Intervention-generated inequalities in lung cancer care Lynne Forrest Submitted for the degree of Doctor of Philosophy Institute of Health and Society Faculty of Medical Sciences Newcastle University September 2013 i Abstract Lung cancer survival is poorer in more socio-economically deprived patient groups. It has been suggested that socio-economic inequalities in receipt of, and time to, treatment may contribute to inequalities in cancer outcome. Unintended variations in outcome that result from the way that interventions are organised and delivered have been described as intervention-generated inequalities. The aim of this thesis was to determine if there are socio-economic inequalities in lung cancer care and, if so, to identify where on the pathway of care these inequalities might occur: looking at receipt of treatment; referral, diagnostic and treatment time intervals; and survival. A systematic review and meta-analysis was conducted in order to examine the published evidence for socio-economic inequalities in lung cancer treatment. A secondary analysis of cancer registry data for 65,210 patients diagnosed between 1999-2010 with a primary diagnosis of lung cancer [ICD10 C33 and C34], linked to Hospital Episode Statistics and lung cancer audit data, was conducted. Logistic regression was used to examine the likelihood of receipt of treatment; of receiving timely referral, diagnosis and treatment within guidelines; and of being alive two years after diagnosis, by socio-economic position [SEP]. Cox regression was used to assess the likelihood of early referral, diagnosis and treatment and hazard of death, by SEP. Socio-economic inequalities in receipt of lung cancer surgery and chemotherapy, but not radiotherapy, were found in the systematic review and meta-analysis, and in the linked-data analysis. Socio-economic inequalities in the GP referral to first hospital appointment time interval were identified. Socio-economic inequalities in survival from lung cancer were statistically explained by socio-economic inequalities in receipt of treatment, but not by inequalities in timeliness of referral and treatment, in this cohort. However high levels of missing stage, performance status and co-morbidity data were a limitation. Research into the unexplained variance in treatment rates is required in order to develop interventions that address socio-economic inequalities in receipt of treatment and reduce socio-economic inequalities in survival. ii Acknowledgements I would like to thank my supervisors, Dr Jean Adams, Professor Martin White and Professor Greg Rubin, for allowing me the opportunity to do something productive during my mid-life crisis and offering me this PhD. I am grateful to them all for their help and support but I would particularly like to thank Jean, who has always given timely, detailed feedback on work, pushed me to write papers and present at conferences, and supported me in my career development, as well as being a fantastic academic inspiration. Obtaining the secondary data was problematic so I would also like to thank Becky at NYCRIS for agreeing to take this project on, and for linking and supplying some of the data, and her replacement Sarah for completing the job. Thanks also to IHS statisticians Andy and Colin for their helpful advice re meta-analysis and multiple imputation, Fuse student Helen for being the second reviewer on the systematic review, and to Richard Thomson and Svetlana Glinianaia for their helpful advice as my internal reviewers. I am grateful to all my fellow Fuse PhD students and IHS colleagues for keeping me going, and also to the supportive network of ‘virtual’ public health and academic colleagues I ‘met’ on Twitter. I’d also like to thank my old friends and family for not trying to dissuade me when I said I was going to do a PhD (again), but I’ve been able to reassure them that it was a lot less painful second time around. And hopefully I’ve finally sorted myself out a ‘proper’ career. I’d like to thank my husband John for being so supportive, as well as for taking on much of the burden of weekend childcare and transportation whilst I got on with writing up; and my children, Rachel and Ben, for being understanding when I couldn’t always come along to school events or to watch their sporting activities as often as I used to. Finally I’d like to thank my parents for everything they’ve done for me and, although it’s taken a bit longer than I’d planned, I hope that I’ve finally fulfilled my academic promise and made them proud. iii Publications from the thesis Peer reviewed papers Forrest LF, Adams JM, Wareham H, Rubin, G, White M (2013) Socioeconomic inequalities in lung cancer treatment: systematic review and meta-analysis. PLoS Medicine. Feb; 10(2):e1001376. doi: 10.1371/journal.pmed.1001376. Editorials Forrest LF (2013) Why are socioeconomic inequalities in receipt of treatment found for lung cancer? Lung Cancer Management. 2(3) 177-180 doi:10.2217/lmt.13.12 Abstracts Forrest LF, White M, Rubin G, Adams J (2013) The effect of socio-economic inequalities in receipt of, and time to, treatment on socio-economic inequalities in lung cancer survival: an observational, data-linkage study. The Lancet: 382:S37. Forrest LF, White M, Adams J (2013) Socio-economic inequalities in lung cancer treatment: the role of histological subtype and performance status. Journal of Epidemiology and Community Health: 67(Suppl 1):A25 doi:10.1136/jech-2013-203126.50 Forrest LF, Adams JM, Wareham H, Rubin G, White M (2012) Socio-economic inequalities in lung cancer treatment: systematic review and meta-analysis. Journal of Epidemiology and Community Health 66: A38-A39 doi:10.1136/jech-2012-201753.099 iv Table of Contents Abstract .............................................................................................................................. i Acknowledgements ........................................................................................................... ii Publications from the thesis ............................................................................................ iii Table of Tables ................................................................................................................ vii Table of Figures ................................................................................................................. x List of abbreviations ........................................................................................................ xii Chapter 1. Introduction ..................................................................................................... 1 1.1 Health inequalities .............................................................................................................. 1 1.2 Intervention-generated inequalities ................................................................................... 2 1.3 Cancer ................................................................................................................................. 2 1.4 Cancer inequalities .............................................................................................................. 8 Chapter 2. Literature review – Intervention-generated inequalities ............................. 10 2.1 IGIs - Background and framework .................................................................................... 10 2.2 Differential receipt of healthcare ..................................................................................... 12 2.3 IGIs – examples from the literature .................................................................................. 15 2.4 Summary ........................................................................................................................... 18 Chapter 3. Intervention-generated inequalities in cancer care: receipt of treatment .. 20 3.1 Introduction ...................................................................................................................... 20 3.2 Inequalities in receipt of treatment .................................................................................. 21 3.3 Summary: Socio-economic inequalities in receipt of treatment ...................................... 27 Chapter 4. Intervention-generated inequalities in cancer care: time intervals ............. 29 4.1 Background ....................................................................................................................... 29 4.2 Evidence for inequalities in cancer care time intervals .................................................... 33 4.3 Summary: Socio-economic inequalities in time intervals ................................................. 43 Chapter 5. Inequalities in cancer survival ....................................................................... 45 5.1 Introduction ...................................................................................................................... 45 5.2 SEP and cancer survival ..................................................................................................... 45 5.3 Other factors influencing survival ..................................................................................... 51 5.4 Delay and survival ............................................................................................................. 53 5.5 Summary of chapters 3-5 .................................................................................................. 54 Chapter 6. Aims and Objectives ...................................................................................... 56 6.1 Development of aims and objectives ................................................................................ 56 6.2 Aims and objectives .......................................................................................................... 56 v Chapter 7. Systematic review and meta-analysis of socio-economic inequalities in receipt of lung cancer treatment .................................................................................... 58 7.1 Background ....................................................................................................................... 60 7.2 Review Objectives ............................................................................................................. 60 7.3 Methods ............................................................................................................................ 60 7.4 Results ............................................................................................................................... 67 7.5 Discussion ........................................................................................................................ 102 7.6 Chapter summary............................................................................................................ 107 Chapter 8. Exploring IGIs in lung cancer care using linked secondary data: Methods . 108 8.1 Introduction .................................................................................................................... 108 8.2 Data sources .................................................................................................................... 109 8.3 Data ................................................................................................................................. 110 8.4 Data variables.................................................................................................................. 113 8.5 Ethical Approval .............................................................................................................. 119 8.6 Overview of secondary data analysis analytical methods .............................................. 120 8.7 Chapter summary............................................................................................................ 122 Chapter 9. IGIs in lung cancer care – dataset overview and descriptive statistics ....... 123 9.1 Introduction .................................................................................................................... 123 9.2 Methods .......................................................................................................................... 123 9.3 Results - descriptive statistics ......................................................................................... 123 9.4 Discussion ........................................................................................................................ 131 9.5 Chapter summary............................................................................................................ 133 Chapter 10. Socio-economic inequalities in receipt of lung cancer treatment ............ 134 10.1 Introduction .................................................................................................................. 136 10.2 Methods ........................................................................................................................ 136 10.3 Results ........................................................................................................................... 137 10.4 Discussion...................................................................................................................... 155 10.5 Chapter summary.......................................................................................................... 162 Chapter 11. Inequalities in referral, diagnostic and treatment time intervals ............. 164 11.1 Introduction .................................................................................................................. 166 11.2. Methods ....................................................................................................................... 167 11.3 Results ........................................................................................................................... 169 11.4 Discussion...................................................................................................................... 184 11.5 Chapter summary.......................................................................................................... 190 Chapter 12. Inequalities in survival ............................................................................... 192 vi 12.1 Introduction .................................................................................................................. 194 12.2 Methods ........................................................................................................................ 194 12.3 Results ........................................................................................................................... 195 12.4 Discussion...................................................................................................................... 220 12.5 Chapter summary.......................................................................................................... 223 Chapter 13. Overall summary - discussion and conclusions ......................................... 225 13.1 Background ................................................................................................................... 225 13.2 Methods ........................................................................................................................ 225 13.3 Results ........................................................................................................................... 226 13.4 Discussion...................................................................................................................... 227 13.5 Reflections on the PhD process .................................................................................... 233 13.6 Chapter summary and overall conclusions ................................................................... 233 Chapter 14. Appendices ................................................................................................ 235 14.1 Appendix A: Variables requested from NYCRIS ............................................................ 235 14.2 Appendix B: Data linkage and data completeness ........................................................ 238 14.3 Appendix C: Supporting Information for chapter 7 ...................................................... 240 14.4 Appendix D: Supplementary tables for chapters 11 and 12 ......................................... 265 Chapter 15. References ................................................................................................. 270 vii Table of Tables Table 4.1. Summary of previous evidence from studies examining lung cancer time intervals and SEP ........................................................................................................................................ 34 Table 5.1. Summary of previous evidence from studies examining likelihood of survival by SEP (in most deprived [low SEP] compared to least deprived [high SEP] group), for lung cancer .... 49 Table 7.1. Characteristics of included studies potentially suitable for meta-analysis (universal health care systems). .................................................................................................................. 69 Table 7.2. Characteristics of included studies not suitable for meta-analysis (universal health care systems). ............................................................................................................................. 72 Table 7.3. Characteristics of included studies potentially suitable for meta-analysis (non-universal health care systems). ................................................................................................... 75 Table 7.4. Characteristics of included studies not suitable for meta-analysis (non-universal health care systems). .................................................................................................................. 77 Table 7.5. Likelihood of receipt of surgery by SEP group (universal health care systems). ....... 81 Table 7.6. Likelihood of receipt of surgery by SEP group (non- universal health care systems). 83 Table 7.7. Likelihood of receipt of chemotherapy by SEP group (universal health care systems). .................................................................................................................................................... 88 Table 7.8. Likelihood of receipt of chemotherapy by SEP group (non- universal health care systems). ..................................................................................................................................... 89 Table 7.9. Likelihood of receipt of radiotherapy by SEP group (universal health care systems).93 Table 7.10. Likelihood of receipt of radiotherapy by SEP group (non- universal health care systems). ..................................................................................................................................... 94 Table 7.11. Likelihood of receipt of any type of unspecified treatment by SEP group (universal health care systems). .................................................................................................................. 98 Table 7.12. Likelihood of receipt of any type of unspecified treatment by SEP group (non- universal health care systems). ................................................................................................... 99 Table 7.13. Likelihood of receipt of any type of unspecified curative treatment by SEP group (universal health care systems). ............................................................................................... 100 Table 8.1. lung cancer histology and tumour type ................................................................... 115 Table 8.2. Referral, diagnostic and treatment intervals ........................................................... 117 Table 8.3. Calculation of Charlson co-morbidity (CCM) score .................................................. 118 Table 8.4. Performance status scale ......................................................................................... 119 Table 9.1. Demographic and clinical characteristics of patients diagnosed with lung cancer (DCO excluded) between 1999 and 2010, and cohorts diagnosed between 1999-2005, 2006-2010, and 2006-2010 subset with stage recorded ................................................................... 124 Table 9.1b. Comparison of variable distribution in the datasets .............................................. 125 Table 9.2. Rates of treatment over time ................................................................................... 127 Table 9.3. Percentage of patients in each SEP quintile (1=least deprived, 5=most deprived), by age at diagnosis, stage at diagnosis, tumour type, CCM score, PS, referral route, and percentage recorded as death-certificate only (DCO) .............................................................. 128 Table 9.4. Tumour type and receipt of treatment (1999-2010) ............................................... 129 Table 9.5. Percentage rates of PS and co-morbidity by stage .................................................. 130 Table 9.6. Percentage rates for number of co-morbidities by performance status score ....... 131 Table 10.1. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors for those diagnosed between 1999 and 2005 (DCO cases excluded) .......................... 140 viii Table 10.2. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort ................................................................ 141 Table 10.3. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort with stage recorded ............................... 142 Table 10.4. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort with probable NSCLC ............................. 143 Table 10.5. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors (DCO cases excluded) for those diagnosed between 1999 and 2005 .............. 144 Table 10.6. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort .................................................... 145 Table 10.7. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort with stage recorded................... 147 Table 10.8. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for probable NSCLC ........................................................................................................................................ 148 Table 10.9. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for SCLC 149 Table 10.10. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors (DCO cases excluded) for those diagnosed between 1999 and 2005 .......................... 151 Table 10.11. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort ................................................................ 152 Table 10.12. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for probable NSCLC .................................................................................................................................................. 153 Table 10.13. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for SCLC ............ 154 Table 11.1. Exclusion of data outliers ....................................................................................... 167 Table 11.2. Date completeness for 1999-2005 and 2006-2010 data ....................................... 169 Table 11.3. Median time per interval compared to target time, for 1999-2005 and 2006-2010 data ........................................................................................................................................... 170 Table 11.4. Hazard ratio of early first hospital appointment from GP referral (for those referred within 1 year) in 2006-2010 cohort .......................................................................................... 171 Table 11.5. Odds of FHA within 14 days from GP referral (for those with FHA within 1 year of GP ref) 2006-2010 ..................................................................................................................... 172 Table 11.6. Hazard ratio of early diagnosis (for those referred by GP within 1 year and with diagnosis date within 1 year from FHA), for 2006 to 2010 cohort ........................................... 177 Table 11.7. Odds of diagnosis within 31 days of GP referral (excluding those with FHA > 1 year from GP ref and diagnosis> 1 year from FHA) 2006-2010 ........................................................ 178 Table 11.8. HR of early treatment from diagnosis (for those treated within 1 year of diagnosis and excluding those untreated) 2006 to 2010 ......................................................................... 179 Table 11.9. Hazard ratio of early treatment for 2006-2010 cohort, including those who did not receive treatment, looking at time from diagnosis to first treatment ( n treated =15,373) .... 180 Table 11.10. Odds of receiving any treatment within 31 days of diagnosis (for those treated within 1 year of diagnosis) 2006-10 .......................................................................................... 181 Table 11.11. HR of early treatment from GP referral (for those treated within 1 year of diagnosis and excluding those untreated) 2006 to 2010.......................................................... 182 ix Table 11.12. Odds of receiving any treatment within 62 days of GP referral (for those with FHA within 1 year of referral, diagnosis within 1 year of FHA and treated within 1 year of diagnosis) .................................................................................................................................................. 183 Table 12.1. Median survival (weeks), by selected patient, tumour and system factors for the cohort diagnosed between 2006 and 2010, and for the 2006-2010 subset with stage (DCO cases excluded) ......................................................................................................................... 204 Table 12.2. Likelihood of mortality (hazard ratio of death), by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) .............. 206 Table 12.3. Likelihood of mortality (hazard ratio of death), by selected patient, tumour and system factors for the subset diagnosed between 2006 and 2010 with stage recorded (DCO cases excluded) ......................................................................................................................... 208 Table 12.4 Survival at 2 years: descriptive data and univariable ORs for 2006-2009 cohort ... 212 Table 12.5. Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 2006 and 2009 (DCO cases excluded) ....... 214 Table 12.6. Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 2006 and 2009 (DCO cases excluded) with stage recorded .......................................................................................................................... 217 Table A.1. Variables that are held in the NYCRIS regional dataset, or can be derived from this .................................................................................................................................................. 235 Table A.2. Variables that are available from the linked, anonymised, regional subset of NYCRIS and HES data, or that can be derived from them ..................................................................... 237 Table A.3. Variables available from the linked NYCRIS regional lung cancer and LUCADA dataset .................................................................................................................................................. 237 Table B.1. Completeness of stage data in the linked dataset, by year ..................................... 238 Table B.2. Completeness of performance status data in the linked dataset, by year .............. 238 Table B.3. Completeness of HES linkage in the linked dataset, by year ................................... 239 Table C2.1. PRISMA 2009 Checklist........................................................................................... 247 Table C3.1. MEDLINE search strategy 05/05/11 ....................................................................... 249 Table C3.2. EMBASE search strategy 05/05/11 ........................................................................ 250 Table D.1. Odds of FHA within 2 weeks from GP referral (for those with FHA within 1 year of GP ref) 1999-2005 .......................................................................................................................... 265 Table D.2. Odds of receiving any treatment within 31 days of diagnosis (for those treated within 1 year of diagnosis) 1999-2005 ...................................................................................... 266 Table D.3. Odds of receiving any treatment within 62 days of GP referral (for those with FHA within 1 year of referral, diagnosis within 1 year of FHA and treated within 1 year of diagnosis) 1999-2005 ................................................................................................................................. 267 Table D.4. Hazard ratio of early first hospital appointment from GP referral (for those referred within 1 year) in 2006-2010 cohort with stage recorded ......................................................... 268 Table D.5. Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 1999 and 2005 (DCO cases excluded) .............. 269 x Table of Figures Figure 1.1.Treatment matrix for NSCLC (2005 recommendations) .............................................. 7 Figure 1.2. NICE guidelines (2011) Fitness assessment clinical pathway: lung cancer ................. 8 Figure 4.1. The NAEDI pathway .................................................................................................. 30 Figure 4.2. Diagnostic and staging clinical pathway for lung cancer .......................................... 32 Figure 4.3. Updated Anderson model of delay ........................................................................... 32 Figure 4.4. Hansen model of delay ............................................................................................. 33 Figure 7.1. Flow diagram of study selection and exclusion. ....................................................... 68 Figure 7.2. Funnel plot to assess publication bias. ..................................................................... 79 Figure 7.3. Meta-analysis of odds of receipt of surgery in low (most deprived) versus high (least deprived) SEP. ............................................................................................................................ 85 Figure 7.4. Meta-analysis of odds of receipt of chemotherapy in low (most deprived) versus high (least deprived) SEP. ........................................................................................................... 90 Figure 7.5. Meta-analysis of odds of receipt of radiotherapy in low (most deprived) versus high (least deprived) SEP. ................................................................................................................... 95 Figure 7.6. Meta-analysis of odds of receipt of any type of treatment in low (most deprived) versus high (least deprived) SEP (non- overlapping populations; n=31). ................................... 97 Figure 7.7. Meta-analysis of odds of receipt of unspecified treatment in low (most deprived) versus high (least deprived) SEP. .............................................................................................. 101 Figure 12.1. Kaplan-Meier survival curve: SEP .......................................................................... 196 Figure 12.2. Kaplan-Meier survival curve: age .......................................................................... 196 Figure 12.3. Kaplan-Meier survival curve: sex .......................................................................... 197 Figure 12.4. Kaplan-Meier survival curve: tumour type ........................................................... 197 Figure 12.5. Kaplan-Meier survival curve: PS ............................................................................ 199 Figure 12.6. Kaplan-Meier survival curve: co-morbidity ........................................................... 199 Figure 12.7. Kaplan-Meier survival curve: treatment type ....................................................... 200 Figure 12.8. Kaplan-Meier survival curve: time to, and type of, treatment ............................. 200 Figure 12.9. Kaplan-Meier survival curve: time to first treatment ........................................... 201 Figure 12.10. Kaplan-Meier survival curve: timely surgery ...................................................... 201 Figure 12.11. Kaplan-Meier survival curve: timely chemotherapy ........................................... 202 Figure 12.12. Kaplan-Meier survival curve: timely radiotherapy ............................................. 202 Figure 12.13. Kaplan-Meier survival curve: stage ..................................................................... 203 Figure B.1. LUCADA case ascertainment, and stage & performance status data completeness, for England and Wales, by year ................................................................................................ 239 Figure C7.1. Meta-analysis of odds of receipt of surgery in low (most deprived) versus high (least deprived) SEP. (over-lapping populations; n=16) ........................................................... 261 Figure C7.2. Meta-analysis of odds of receipt of surgery for NSCLC in low (most deprived) versus high (least deprived) SEP. (non-overlapping populations; n=8) .................................... 261 Figure C7.3. Meta-analysis of odds of receipt of chemotherapy in low (most deprived) versus high (least deprived) SEP (over-lapping populations; n=10)..................................................... 262 Figure C7.4. Meta-analysis of odds of receipt of radiotherapy in low (most deprived) versus high (least deprived) SEP (overlapping populations; n=9). ....................................................... 262 Figure C7.5. Meta-analysis of odds of receipt of any type of treatment in low (most deprived) versus high (least deprived) SEP (overlapping populations; n=35). ......................................... 263 xi Figure C7.6. Meta-analysis of odds of receipt of surgery in low versus high SEP. (partially-overlapping populations, n=14) ................................................................................................ 264 xii List of abbreviations CCM score Charlson co-morbidity score CI Confidence interval FHA First hospital appointment HES Hospital Episode Statistics HR Hazard ratio ICD International Classification of Disease IGIs Intervention-generated inequalities LUCADA Lung cancer audit MDT Multi-disciplinary team NAEDI National Awareness and Early Diagnosis Initiative NHS National Health Service NICE National Institute for Health and Clinical Excellence NSCLC Non-small cell lung cancer NYCRIS Northern and Yorkshire Cancer Registry and Information Service OR Odds ratio PRISMA Preferred reporting items for systematic reviews and meta-analyses PS Performance status SCLC Small cell lung cancer SEER National Cancer Institute’s Surveillance , Epidemiology and End Results database SEP Socio-economic position UHCS Universal healthcare system 1 Chapter 1. Introduction The aim of this thesis is to explore the evidence for intervention-generated inequalities (IGIs) in lung cancer treatment: to examine the published evidence for socio-economic inequalities in lung cancer treatment; to use routine secondary data to determine if there are inequalities in lung cancer care, and to identify where on the pathway of care these inequalities might occur – looking at inequalities in receipt of treatment, time to treatment, and survival. Chapter 1 presents a general introduction to the thesis. Chapter 2 reviews the literature on IGIs, and chapters 3, 4 and 5 examine the evidence for IGIs and cancer, whilst chapter 6 details the aims and objectives. Chapter 7 examines the published evidence for socio-economic inequalities in lung cancer treatment using a systematic review and meta-analysis, and chapters 8-12 present the research methods and results from the secondary data analysis. Chapter 8 details the methods and chapter 9 presents descriptive analyses, before going on to explore socio-economic inequalities in lung cancer treatment [chapter 10], referral, diagnostic and treatment intervals [chapter 11] and survival [chapter 12]. An overall discussion of the thesis findings is then presented in chapter 13. The general background to the topic will now be examined. 1.1 Health inequalities Despite improvements in living standards, introduction of the welfare state and free NHS care for all, socio-economic differentials in health are still observed in the UK (1). Health inequalities can be defined as differences in health outcomes between groups within populations [or between populations]. Although health inequalities due to biological differences are inevitable, it is recognised that inequitable differences in health also exist, which are unfair and potentially avoidable (2, 3). Health inequalities have been observed using a number of different measures of socio-economic position [SEP] (2, 4). 2 1.1.1 Socio-economic position Socio-economic position is a construct used to measure the social and economic factors that influence the position of an individual or group within society (5), using area-based, household, and individual measures of deprivation, poverty, wealth, income and education (6). There are advantages and disadvantages to using each of these markers of SEP, which need to be taken into account when examining health inequality. Education is easy to measure, applicable to all and is generally stable over adulthood. However, it is strongly influenced by parental and societal characteristics and there are cohort effects. It is a less sensitive measure of SEP than income, but income is often poorly reported and excludes non-working groups (5). Household rather than individual income can often be a better measure of SEP for women. Poverty is a subjective measure and is determined by societal norms. Using a neighbourhood measure of SEP means there is danger of ecological fallacy but it can apply to all ages and both sexes (6). 1.2 Intervention-generated inequalities Intervention-generated inequalities have been described as health inequalities that result from the way that health interventions are organised and delivered (7) so that, although overall health may improve as the result of an intervention, differences in access to the intervention, differential uptake, delays in uptake, differential compliance with, or effectiveness of, an intervention might result in inequalities in outcome. Inequalities are likely to occur at many different stages of intervention pathways and act in a cumulative way. It is also likely that intervention-generated inequalities contribute to overall socio-economic inequalities in morbidity and mortality, although this has not been conclusively demonstrated (7). 1.3 Cancer Cancer is a term used for a number of diseases in which normal cells change so that they grow in an uncontrolled way and are able to invade other tissues. The uncontrolled growth causes a tumour to form. Cancer cells can spread to other parts 3 of the body through the blood and lymph systems. There are more than 200 different types of cancer and most are named for the organ or type of cell in which they start. Most cancers are carcinomas, which are cancers of the epithelial cells [tissue that covers and lines the body]. Carcinomas make up about 85% of cancers. There are different types of epithelial cells and these can develop into different subtypes of carcinoma, including adenocarcinoma and squamous cell carcinoma. Staging is a way of describing the size of a cancer and how far it has grown and spread to another part of the body, whereas the grade of cancer describes how similar the cancer cell is to a normal cell. Staging is important in determining treatment. For localised cancer, surgery or radiotherapy could be curative. If a cancer has spread, then systemic treatments such as chemotherapy or hormone therapy [that circulate throughout the bloodstream] may also be required. Cancer patients often suffer from a number of concurrent health conditions which are termed co-morbidities. Performance status [PS] is a measure of general well-being and ability to care for oneself that is assessed by the care team. Both these variables may be used to help determine the most suitable cancer treatment when patients are assessed. More details on these variables will be presented in chapter 8. 1.3.1 Cancer incidence, mortality and inequality Breast, lung, colorectal and prostate cancer are the top four cancers for incidence and mortality in the UK and together account for over 54% of UK cancer incidence (8) and 47% of cancer mortality (9). Lung cancer is the leading cause of cancer death accounting for 22% of cancer mortality followed by colorectal cancer at 10% (9). Inequalities in cancer incidence are found for a number of common cancers (8). Higher cancer incidence within more deprived populations is associated with higher rates of smoking and alcohol consumption, unhealthy diet and lack of exercise, all of which contribute to a higher risk of cancer (8). A statistically significant association between higher colorectal cancer incidence and socio-economic deprivation has been reported for men in England and Wales (10). However, this association is not always seen worldwide. A review reported that those 4 with lower SEP had higher colorectal cancer incidence compared to higher SEP groups in the USA and Canada, but that the reverse was true in Europe, where those with lower SEP had lower risk (11). However, no UK studies were included and a number of European studies did show an increased risk with lower SEP. Conversely higher incidence in less deprived populations is found for breast cancer and malignant melanoma in England (12). Around 63% of cancers are diagnosed in people aged 65 and over, and more than a third are diagnosed in the elderly [aged 75 and over]. Survival varies markedly by cancer type. Whereas around 50% of those diagnosed with colorectal cancer and 80% with breast cancer can expect to still be alive five years after diagnosis only around 6% of those with lung cancer can (9). Whilst survival rates have greatly improved for most cancers in recent years this improvement has not been seen for lung cancer. 1.3.2 Lung cancer Lung cancer is the most common incident cancer, worldwide. In the USA and the UK it is the second most incident cancer after breast cancer (8, 13) [and the second most common for men after prostate cancer and for women after breast cancer], as well as the most common cause of cancer mortality (9, 13). Survival differs internationally. In the UK less than 10% of those diagnosed with lung cancer survive for 5 years (9, 14), with higher survival rates found in Nordic countries (14, 15), the USA (13, 15), Australia and Canada (14). A strong socio-economic gradient for lung cancer incidence is seen in the UK, with rates 2-3 times higher in the more deprived (8). Higher incidence and mortality is found in the north of England compared to the rest of the UK also (8), and the incident deprivation gap is wider (12). A systematic review and meta-analysis found that, worldwide, lung cancer incidence was associated with low education and low SEP [both occupational and income-based] and remained so in subgroup analyses adjusted for smoking (16). Smoking has been strongly associated with incidence of small cell lung cancer, and also squamous cell non-small cell lung cancer but less so with adenocarcinoma tumour type. There is some suggestion that past smoking behaviour 5 may account for much of the socio-economic inequalities in risk of lung cancer (17), but not all, as an SEP gradient in incidence remains when smoking status is taken into account (16). Lung cancers are classified into small cell [SCLC] and non-small cell [NSCLC] cancers, with NSCLC accounting for 80% of lung cancers. NSCLC can be further divided into squamous cell carcinomas, adenocarcinomas and large cell carcinomas (18). Studies have found different proportions of NSCLC subtypes. Squamous cell carcinoma has historically been strongly associated with smoking but, with the introduction of lower tar and nicotine cigarettes, the incidence of this subtype is falling (19). Adenocarcinomas are a morphologically heterogeneous group and although they are also associated with smoking [generally of low-tar filtered cigarettes (17)] they are also found in those who have never smoked, particularly in women (19). Large cell carcinomas are the least common subtype and are often misclassified (19). SCLC are thought to account for around 20% of lung cancers although this proportion has been reported to be falling in more recent years (20). SCLC is also strongly associated with smoking and the decrease in incidence of SCLC over time, similarly to the squamous subtype for NSCLC, may be related to the reduced prevalence of smoking and type of cigarettes smoked (20). 1.3.2.1 Lung cancer staging Lung cancer stage is determined using the ‘Tumour, Node, Metastasis’ [TNM] staging system (21). The TNM staging system describes the size of the primary tumour, whether the cancer has spread to the lymph nodes and whether the cancer has spread to a different part of the body [metastasised]. 'T' refers to the size of the cancer, going from 1 [small] to 4 [large]. The letters A, B or C can be used to further divide the number categories. 'N' refers to whether the cancer has spread to the lymph nodes - from 0 [no positive nodes] to 3 [many positive nodes]. 'M' refers to whether the cancer has spread to another part of the body - either 0 [the cancer has not spread] or 1 [the cancer has spread] (21). In stage 1 lung cancer the cancer is small [1A: up to 3cm, 1B: 3-5cm] and localised, and has not spread to the lymph nodes (21). In stage 2A, the cancer is between 5-7cm and 6 has not spread to the lymph nodes. Stage 2B includes cancers that are 5-7cm and have spread to the lymph nodes close to the affected lung; cancers larger than 7cm but with no lymph node spread or that have spread to the chest wall, diaphragm, phrenic nerve, layers covering the heart, or bronchus; cancers larger than 7cm where part of the lung has collapsed; or cancers of any size where there is more than one tumour in the same lobe of the lung (21). In stage 3A lung cancer the cancer is larger than 7cm and has spread to the lymph nodes close to the affected lung; or any size but spread to the heart, trachea, oesophagus, nerve, spinal bone or blood vessel. In stage 3B the cancer is in the mediastinum lymph nodes and has spread to the chest wall, diaphragm, heart, trachea, oesophagus or major blood vessels. In stage 4 the cancer is in both lungs; or has spread to another part of the body; or cancer cells are in a fluid collection around the heart or lungs (21). 1.3.2.2 Lung cancer treatment National Institute for Health and Clinical Excellence [NICE] guidelines [2005] recommend radical surgery [pneumonectomy or lobectomy] for stage I or II NSCLC. Chemotherapy and radical radiotherapy are recommended for stage IIIa, with chemotherapy for stage IIIb and stage IV lung cancer patients with good performance status. Radiotherapy may be given as a palliative option for stage IV patients with poor performance status (18). Figure 1.1 summarises these NICE recommendations. Updated guidelines from 2011 now recommend radical radiotherapy for stage I-III NSCLC patients (22). Chemotherapy and radiotherapy were the treatments of choice for SCLC but surgery is now recommended for early stage SCLC (22). Intervention with surgery, chemotherapy or radiotherapy has been shown to improve survival (18). Fitness for treatment is assessed using the following NICE guidance pathway [Fig 1.2] (22).The type of cancer treatment given is generally determined by cancer stage, but also by the performance status of the patient and the number and type of co-morbidities suffered. Lung cancer sufferers tend to be old [age 70+] and to be smokers so that high rates of chronic obstructive pulmonary disease [COPD] and ischaemic heart disease are common (23), often making radical surgery unsuitable for these patients. 7 Figure 1.1.Treatment matrix for NSCLC (2005 recommendations)(18) 8 Figure 1.2. NICE guidelines (2011) Fitness assessment clinical pathway: lung cancer (22) ©2011 by British Medical Journal Publishing Group 1.4 Cancer inequalities Inequalities in cancer care within the UK have been noted and the NHS Cancer Plan in 2000 pledged to reduce cancer mortality, reduce delay in diagnosis and treatment and increase survival whilst acting to reduce inequalities (24). More recently the National Cancer Equality Initiative has been set up to address some of these issues (10) as, although the incidence and survival of many common cancers varies with SEP (8, 25, 26), little is known about inequalities in receipt of, and time to, treatment and how these might contribute to inequalities in outcome. However, in a 2006 review that summarised a decade of research on the association between SEP and cancer survival, it was suggested that socio-economic differences in ‘access to optimal treatment’ (26) might at least partially explain survival differences. 9 Lung cancer is the leading cause of cancer death within the UK [with mortality rates higher in the North of England and Scotland than in the rest of the UK] and it has been suggested that reducing the socio-economic gradient in survival could prevent many thousands of avoidable cancer deaths (9). It has been estimated that over 1300 deaths could be avoided annually in England and Wales if the survival rate in the more deprived socio-economic groups were similar to that of the most affluent (27). However, in order to do this it is necessary to identify those factors that may contribute to outcome inequalities. Access to treatment may be one such factor. For example, resection rates in the north-east are lower than those for England as a whole (28) and this might contribute to the poorer survival observed locally. Socio-economic inequalities in receipt of cancer care have been found in individual studies for a number of cancers and assessed in a review for colorectal cancer (11) but there has been no systematic review of the evidence to demonstrate if such inequalities in receipt of treatment exist for lung cancer. A previous study conducted in the Institute of Health and Society [IHS] at Newcastle University explored socio-economic inequalities in the delivery of pancreatic cancer care, using Northern and Yorkshire Cancer Registry and Information Service [NYCRIS] data, and found evidence of socio-economic differences in time from GP referral to first hospital appointment and in receipt of treatment (29). This study will build on these findings, using the intervention-generated inequalities framework to explore inequalities in cancer care for lung cancer, whilst also undertaking a systematic review of the evidence into inequalities in receipt of treatment. The next chapter will describe the concept of, and explore the evidence for, Intervention-generated inequalities. 10 Chapter 2. Literature review – Intervention-generated inequalities This chapter will define intervention-generated inequalities [IGIs], review the evidence for IGIs in the literature, as well as exploring the potential reasons for inequalities in receipt of healthcare and public health interventions. 2.1 IGIs - Background and framework Health inequalities can be found across a number of dimensions including: place of residence [urban/rural], race/ethnicity, occupation, gender, religion, education, socio-economic position [SEP] and social capital, which have been given the acronym PROGRESS (30). If other dimensions such as age, disability and sexual orientation are included then this the term PROGRESS–Plus is applied (30). The intervention-generated inequalities [IGI] concept was developed by White et al [2009], in order to produce a framework for a more evidence-based theory and model of intervention inequality (7), as previous equity ‘laws’ and hypotheses had been used in a fairly loose and interpretive way throughout the equity literature. Intervention-generated inequalities have been described as health inequalities that result from the way that health interventions are organised and delivered (7). Although overall health may improve as the result of an intervention, differences in access to the intervention; differential uptake; delays in uptake; differential compliance with; or effectiveness of; an intervention, may result in inequalities in outcome. There may also be inequalities in timeliness of the offer of the intervention and to whom it is offered. The IGI framework expands on previous equity hypotheses that have attempted to describe variations in the provision and uptake of interventions. The ‘Inverse Care Law’ [ICL] first described by Tudor-Hart in 1971 stated that “the availability of good quality health care is inversely related to need in the population served” (31). The ICL has been cited in a number of studies concerning access to primary care, healthcare utilisation (32), GP prescription practices and satisfaction with care (7), but it only looks at provision of care in relation to need. Whilst SEP can be seen as a proxy marker for ‘need’ [with the assumption that those of lower SEP have more need of care] the ICL does not consider intervention effectiveness. The focus is on the organisation and delivery of standardised care whereas, in order to address inequalities, it might be 11 more important to consider different types of care and the context in which care is delivered. The ‘inverse Prevention Law’ suggests that a similar socio-economic gradient also exists in disease prevention but this was an idea that appeared within the Acheson Report in 1999 (33) without any empirical evidence to back up this claim (7). The ‘inverse equity hypothesis’ proposed that public health interventions initially widen socio-economic inequalities due to preferential uptake by more advantaged groups before the less advantaged follow suit, eventually improving health overall (34). Although evidence-based using child health studies from Brazil, it was developed for low-income countries but has been used as a framework elsewhere, for example in looking at uptake of new cancer treatments in the UK (35). The hypothesis builds on the ‘diffusion of innovation’ theory (34) and makes the assumption that all interventions are appropriate for everyone and so could be seen to ascribe blame to those who are slower to ‘innovate’ and uptake. It does not consider that some interventions may not be suitable for all and so will never diffuse downwards. The Tugwell ‘staircase’ model suggests that inequalities at different stages of an intervention may combine multiplicatively in order to reduce intervention effectiveness in the more deprived (36). Lower access to the intervention [including awareness or coverage], inequalities in diagnosis [screening or targeting], provider compliance, and consumer adherence in more deprived populations means that this staircase effect increases the relative equity-effectiveness gap between rich and poor (36). None of the above equity laws and models examines how interventions can be designed to reduce inequality. It needs to be considered whether it is enough to provide a standard level of care to all or whether targeted interventions tailored to specific groups might be more effective. Therefore, in order to expand the previous equity laws into a framework to address intervention-generated inequalities it is necessary to examine all the possible stages where inequalities may be introduced, determine which aspects of an intervention may lead to inequalities, and consider ways in which these can be reduced or eliminated. 12 2.2 Differential receipt of healthcare A number of reasons have been suggested for differential receipt of healthcare, ascribed to both patient and system factors (37). However, when considering inequalities in access to, and receipt of, care, it is important to first take into account the type of healthcare system. 2.2.1 Healthcare system In the UK, the National Health Service [NHS] is free at the point of use. Access to care is therefore, in theory, equally available to all UK citizens who require it. Other countries, including the United States (US), have a non-universal health care system, where access to healthcare is determined by insurance status. Citizens [generally those who are in work] can purchase private insurance, and social insurance programmes - Medicare [for those aged over 65] and Medicaid [for certain categories of people who are on a low income or are disabled] - are also available (38). There is a wealth of evidence to show that lack of a universal healthcare system can result in disparities in access to treatment (38, 39). In 2002, 17% of Americans were uninsured (38), and there is considerable inequity, by income and race, in insurance coverage (38). 2.2.2 Patient and practitioner factors Eligibility for healthcare has been described as ‘candidacy’ where it ‘is jointly negotiated between individuals and health services’ (40). How individuals identify themselves as requiring medical attention may depend on their health expectations and how they manage health and recognise symptoms. They must then be able to navigate the system and access care (40). Socio-economic barriers to uptake may be financial, psychological and educational. For example, specialist care centres may be geographically distant and it may be more difficult to access services for those who cannot afford childcare, are unable take time off from work or do not have access to a car (37). Lower SEP patients report struggling to navigate a complicated appointments system in order to access GP care (41). It may be easier to use emergency services that do not involve making an appointment or require a referral, but this could result in 13 patients delaying presenting until symptoms become more severe and when they are at a more advanced stage of the disease pathway. A realist review [a review method that falls between a systematic and narrative review] found evidence that, as middle-class service users tend to be more vocal, this may influence their interaction with health services and the likelihood of referral and receipt of treatment, as well as determining the type and quality of information provided to them by health professionals (42). Differences in communication patterns between health professionals and patients by SEP have been described that may influence the treatment prescribed (37). The patient must be able to articulate the problem and request help and the ability to do so may affect the likelihood of referral (40). Healthcare professionals may make treatment decisions based on which patients they consider likely to do well, using factors such as age, weight and co-morbidity. There is some evidence that those who are in work are more likely to be referred (40) and these subjective judgements may disadvantage those of lower SEP. Adherence to treatment protocols has also been shown to differ by SEP, as has treatment refusal (40), which may be related to understanding of risk. 2.2.3 Health literacy The ability to interpret symptoms, understand risk and effectively navigate the healthcare system can be included under the general heading of health literacy. Health literacy has been defined as ‘the degree to which individuals have the capacity to obtain, process and understand basic health information and services needed to make appropriate health decisions’ (43). Health literacy has been proposed as a major determinant of population health and health inequalities, although Nutbeam [2008] suggests that ‘the corrosive impact’ that poor literacy may have on health remains well hidden (43). It is likely that poorer health literacy plays a role in the higher prevalence of adverse health behaviours that result in higher disease incidence in more deprived SEP populations. 14 Much of the early research examining health literacy and healthcare interactions has been conducted in the US. In a recent systematic review the authors found evidence of an association between poor health literacy and reduced ability to manage chronic conditions and take medication appropriately, lower use of mammography screening and flu vaccine uptake, higher rates of hospital admission and emergency admissions, and poorer health outcomes in the elderly, but the evidence for a relationship between health literacy and access to care was judged as insufficient (44). It was also suggested that lower health literacy may partially explain racial disparities in some health outcomes (44). SEP was not considered but it seems feasible that health literacy may also play some role in disparities in health outcome by SEP. A recent UK study found that low functional health literacy [the ability to apply basic reading skills in a health context] was associated with higher mortality (45). There is some evidence to suggest that low health literacy is found at higher rates amongst low SEP groups, as well as in older patients and ethnic minorities (46) although much of this research has been carried out by UK consumer groups rather than in academia. A recent qualitative UK study found that patients from GP practices in low SEP areas did appear to have lower health literacy than those from higher SEP areas and, as well as a difference in health knowledge, there was also a difference in role expectation (41). Patients’ willingness to actively participate in the consultation depended on their cultural and normative expectations of the doctor and patient role. Lower SEP patients felt less able to question the doctor and often felt it was not their place to do so (41). Patients of higher SEP are more likely to state a preference for involvement in medical decision making [as are women and younger patients] (47) but in order to make an informed decision patients must be able to understand the information provided. Poor understanding of health information is not a minor problem as studies have estimated that over 50% of the UK adult population struggle to understand narrative health information, and this rises to over 70% when numbers are included in the information and where basic numeracy skills are required (46). Patients often do not understand complex medical terminology and this problem is compounded by the fact that health care professionals in turn do not recognise the extent of the problem and tend to over-estimate their patients’ understanding (48). Health literacy can be context-specific and, within a healthcare environment, the use of unfamiliar 15 vocabulary and terminology can be stressful and alienating. It is possible that health literacy may influence patient choice. If patients have poor capacity to process and understand basic health information then they are less able to make appropriate health decisions. It has been suggested that the current UK government emphasis on patient involvement in health decision making may increase health inequalities (46), as only those with good health literacy are fully able to participate in effective shared decision making. This might be seen as a further example of an intervention-generated inequality. 2.3 IGIs – examples from the literature I will now consider what, if any, evidence exists for the occurrence of intervention-generated inequalities, the types of interventions that are likely to produce IGIs and those which may reduce inequalities. 2.3.1 Review evidence Two early systematic reviews examined interventions to reduce health inequalities (49, 50), looking at interventions to reduce cancer incidence [screening, smoking cessation] and heart disease, accident prevention, reduce risky sexual health behaviour and unwanted pregnancy, and improve accessibility of healthcare. Arblaster et al (1996) found that successful interventions for reducing health inequalities used systematic and intensive health-care delivery approaches, improved access and used prompts to encourage service use, designed tailored interventions specifically for the target population and used peers to help deliver the intervention (49). However, they also found that not all these features were necessary to reduce inequalities but, conversely, that these characteristics alone were not enough to ensure equality. The scope of the review was quite narrow as it considered only interventions that could be carried out by a health service, and the authors were only able to conduct a narrative synthesis of the evidence as the settings, populations and outcomes were so diverse. A similar review conducted in the same year was wider in scope but included some of the same interventions and studies. Gepkens et al [1996] found that structural interventions were most effective. Interventions involving health education only 16 seemed to be effective in lower SEP groups when supplemented by personal support or structural measures (50). However, the majority of the interventions examined aimed to increase knowledge or change behaviour rather than having a specific health outcome. And, again, the diversity of the interventions meant that meta-analysis was not possible. A 2008 systematic review looked at social inequalities [assessed using PROGRESS criteria] in tobacco control interventions (51) and identified multiple gaps in the evidence base. There was some evidence that workplace restrictions on smoking may be more effective in the more affluent (51). However, much of the evidence was from the USA and so may not be relevant in other countries. There was little evidence on differential effects by SEP, sex or ethnicity and, in order to address this, the authors suggested that in population-level studies pre-planned subgroup analyses should be carried out to examine inequalities (51). A short systematic review of reviews published in 2012 used a limited search strategy but employed systematic screening of papers to identify evidence for the types of interventions that might generate inequalities (52). However, this ‘rapid review’ did not look at interventions involving access to healthcare. The authors again found that data was lacking but, the 12 reviews that they were able to include, there was some evidence that interventions that involved structural change, pricing and the provision of resources appeared to reduce socio-economic health inequalities, whereas those that merely provided information, such as media campaigns, increased inequalities (52). That is, more ‘upstream’ social, fiscal or policy-level legislative interventions appeared more equitable than those that relied on education or information-provision to engender voluntary behaviour change. Further evidence has been found to support this theory, as is described below. 2.3.1.1 Interventions that reduce inequalities A health voucher scheme in Bangladesh found that the introduction of vouchers for maternal health care reduced socio-economic inequalities in service utilisation and, contrary to the inverse equity hypothesis, better early uptake was seen in the poorer rather than more affluent women (53). Other financially-related interventions have also been shown to reduce inequalities; for example, tobacco price increases are more 17 likely to reduce smoking in lower-income adults and those in manual occupations whereas smoking cessation services are less effective in these groups (51). Legislative changes may also favour the more deprived, as seen in water fluoridation (54), and motorbike helmet use (55). The introduction of a mandatory motorcycle helmet law in Taiwan reduced regional socio-economic inequality in mortality from road traffic accidents and this was most marked for men aged under 25 (55). However, inequalities began to increase again 5 years after introduction of the law, so it may depend on how strictly a law is enforced as to whether people continue to comply equitably. Rear-seatbelt use in the UK is mandatory but a socio-economic gradient in compliance is found (56). Although not all of these studies may be directly relevant to the UK, they provide examples of the principle that the equitable effectiveness of an intervention may depend on the level of control and voluntary engagement an individual has. 2.3.1.2 Interventions that increase inequalities Inequalities in uptake of population-based non-mandatory public health interventions such as cancer screening and immunisation, were identified in a narrative review of evidence for IGIs (7). A UK intervention designed to improve uptake of childhood vaccinations improved overall uptake but widened socio-economic inequalities in uptake (57). A study that examined smoking inequalities in Sweden and Denmark found that, in Sweden, where the government had implemented an active, population-based anti-smoking campaign, overall levels of smoking were lower but inequalities were higher, with better educated individuals having responded best to the strategy, compared to Denmark (58), where less forceful anti-tobacco policies had been implemented. 2.3.2 Gaps in the evidence base It does appear that there is a lack of systematic evidence on the type of interventions that may result in inequalities in health, as well as in where on the intervention pathway that intervention-generated inequalities may be occurring. 18 Equity assessments are rarely conducted as part of systematic reviews and, in a random sampling, only one in 95 Cochrane Reviews examined outcome effects across PROGRESS factors (30). In an, as yet unpublished, review on the variations in effectiveness of obesity treatment and prevention interventions by SEP it was noted that the majority of papers recovered did not detail outcome by SEP in the abstract, even if it was detailed in the main text [Martin White, personal communication]. A 2013 pilot review examining systematic reviews of physical activity interventions found that only 26% of these were prospectively designed to examine inequalities (59). Comparing the effects of interventions in different populations often requires that studies carry out sub-group analyses. However studies may not be suitably powered to allow detection of sub-group differences, and if post-hoc subgroup analyses are carried out, without suitable consideration of plausibility in relation to inequity theory, then researchers may find themselves accused of ‘data-dredging’ by disapproving statisticians (60). In a methodological study of equity assessment in systematic reviews, of 224 systematic reviews indexed on Medline in one month, only 13% of these examined were found to have conducted subgroup analysis by PROGRESS-plus factors (61). It is often difficult, therefore, to assess the effects of interventions on health equity as few studies report their results in this way. But it is important to do so, as unequal benefits across socio-economic groupings are likely to contribute to decreased health equity (30) which can be considered as unethical and unfair, as well as sub-optimally cost-effective. There are also implications for generalisability, as, if only the more affluent take up an intervention, the data may not be applicable at a population-level (62). However, this scarcity of evidence means that the contribution of intervention-generated inequalities to overall socio-economic inequalities in morbidity and mortality is still unclear. 2.4 Summary Although the evidence base for intervention-generated inequalities remains limited, it does appear that financial and legislative interventions, and those which are specifically targeted at particular groups, are less likely to introduce intervention-generated inequalities than those that rely on voluntary uptake or that provide a ‘one 19 size fits all’ solution for prevention and treatment. However, the patient and system factors that influence access to, and uptake of, healthcare and public health interventions are also not well evidenced and need to be determined in order to eliminate IGIs. Chapters 3-5 will examine the evidence for IGIs in cancer care. 20 Chapter 3. Intervention-generated inequalities in cancer care: receipt of treatment As described in chapter 2, the evidence base for IGIs is somewhat limited. I will now look at the evidence for IGIs specifically in relation to receipt of cancer treatment, and why these might be found. 3.1 Introduction Previous work looking at inequality in pancreatic cancer care, using NYCRIS data (29), found a significant socio-economic trend in receipt of any treatment, with the odds of receiving treatment increasing from the most to the least deprived quintile when age, sex and co-morbidity were taken into account. This trend was also seen for receipt of any radical surgery but not for adjuvant chemotherapy. It was also seen for any palliative oncology treatment, any palliative chemotherapy, but not radiotherapy or hormone therapy. Differences in access to treatment might be partially explained by differences in stage at presentation, but unfortunately staging information was not available. Time from GP referral to hospital investigation was significantly shorter in the less socio-economically deprived groups compared to the most deprived [and for older compared to younger patients], but this trend was not seen when looking at time from first hospital appointment to diagnosis, or time from diagnosis to any therapy or treatment [apart from in time to palliative radiotherapy] (29). Therefore, there is some evidence of socio-economically patterned delay in time from referral to first hospital appointment for pancreatic cancer. Building on this first attempt to examine intervention-generated inequalities in cancer care, looking at pancreatic cancer, I will now use the intervention-generated inequalities framework to assess where on the pathway of cancer care inequalities might occur and why they might occur. The evidence for inequalities in receipt of treatment in all cancers is considered in this chapter, including the evidence specifically relating to lung cancer. Inequalities in time intervals on the care pathway are examined in chapter 4 and inequalities in survival in chapter 5. Socio-economic 21 inequalities in receipt of treatment and in time to presentation, referral, diagnosis and treatment may all contribute to inequalities in survival. 3.2 Inequalities in receipt of treatment Existing published evidence for inequalities in receipt of treatment for cancer and why these might occur will be examined here. 3.2.1 Why might socio-economic inequalities in receipt of treatment be found? Patient, tumour and system factors may all be relevant for differential receipt of cancer care. Patients may delay in accessing healthcare, choose not to undergo active cancer treatment or to comply with treatment protocols for a number of reasons, often involving a complex mixture of financial, psychological, cultural, practical and educational factors (37), as previously discussed. Low SEP has been associated with colorectal cancer treatment refusal (63) and this may be related to poor health literacy. However, a Scottish study of 882 lung cancer patients found no association between refusal of treatment and SEP (64). Tumour factors such as stage may also be relevant. Later-stage patients are less likely to receive surgery and there is some evidence for a socio-economic gradient in stage at presentation for some cancers (65-67) but not all (68). The use of co-morbidity and PS to determine suitability for treatment may offer a partial explanation for socio-economic and age-related inequalities in treatment, as the number of co-morbidities and PS vary by SEP (69) and age (70). Higher co-morbidity has been associated with older age, female sex and lower SEP in lung cancer patients (71) and might reduce the likelihood of surgery for these patients. However, another study found that chronological age rather than performance status or the number of co-morbidities determined whether elderly patients with lung cancer were actively treated. Many doctors adopted a ‘nihilistic approach’ to treatment and guidelines were not followed for older patients (72). SEP was not considered in this study, but differences in doctor-patient interactions between higher and lower socio-economic groups (73) have also been suggested as a reason for different socio-22 economic cancer treatment patterns where ‘unrecognised personal biases and beliefs may affect recommendations of cancer treatment’ (37). Doctor-patient communication has been shown to differ by race in the US. Black lung cancer patients and those in ‘racially discordant interactions’ received less information and participated less actively in the consultations (74). It is possible that similar discrepancies by SEP are also found. In the English National Cancer Patient Experience Survey for 2011/12, differences in responses from patients were found by SEP. More deprived patients reported that their views were taken into account when the clinical team discussed which treatment they should have, and possible side effects of treatment were explained in a way they could understand (75). However, they also reported delays in diagnosis and that they were less likely to receive understandable information or explanations (75). Patients in the least deprived quintile were more likely to report a better patient experience than the most deprived, for a number of questions that could relate to their level of health literacy. They reported ‘being given the right amount of information about their condition and treatment, given understandable answers to questions by ward nurses all or most of the time, given easy to understand information about the type of cancer they had and given easy to understand written information about tests beforehand’ (75). These survey results may suggest that there are, indeed, differences in the way that doctors communicate, or that differences in health literacy between higher and lower SEP groups determine how well understood the information given is. 3.2.2 Evidence for socio-economic inequalities in receipt of treatment Socio-economic inequalities in receipt of treatment have been found for some cancers, with the majority of evidence from studies looking at colorectal cancer. In a 1998 UK study utilising Hospital Episode Statistics [HES] data from patients admitted between 1992 and 1995 in south-east England, and using Townsend scores as an area-based measure of SEP, patients from less affluent areas were less likely to receive surgery for lung and breast cancer but not for colorectal cancer (76), in a multivariable analysis adjusted for age and sex. This was a well-conducted study but a 23 number of potentially important confounders such as stage were not included. In a similar study in 2010 looking at UK-wide HES data from 1999-2006, patients from deprived areas were less likely to receive optimal surgical treatment for breast, lung and also rectal cancer, taking into account age, sex and year of admission, but again not stage (77). Inequalities in stage at presentation might account for some of these socio-economic differences. In a UK paper that was primarily concerned with examining geographical access to cancer treatment, the authors noted that surgery was less likely for deprived patients compared to the more affluent using data for breast, colon, rectum, lung, ovary and prostate cancer diagnosed between 1994 and 2002, also taking into account age, sex and travel time to hospital, and stage for all except for lung or prostate, as staging information was not well recorded for these cancers (78). Deprived patients were also less likely to receive chemotherapy [except for breast cancer] and less likely to receive radiotherapy [except for colon and rectal cancer] (78). 3.2.2.1. Colorectal cancer There is some evidence to suggest that deprived patients with colorectal cancer are less likely to receive treatment (11, 79, 80). According to a 2010 review examining socio-economic position and inequalities in risk, treatment and outcome for colorectal cancer (11), low SEP colon cancer patients were less likely to receive treatment [surgery, radiotherapy and chemotherapy] but results for rectal cancer were less clear cut. However, within this review the search terms were narrow and a number of relevant papers were not included. In the above-mentioned review (11) nine studies conducted multivariate analysis and eight included stage. Two only included patients aged 65+ (81, 82). Three (83-85) out of seven studies looking at access to radiotherapy (81, 83-88) found a significant difference by SEP, as did two(82, 85) out of seven studies looking at access to chemotherapy (81-87). Three studies looked at access to surgery (83, 86, 89) but only one (83) showed a significant difference in likelihood of receiving treatment for high compared to low SEP patients. For colon cancer significant differences in receipt of chemotherapy by SEP were found in two (82, 85) out of three papers but, as more mixed results were found for rectal and colorectal cancer and for access to surgery and 24 radiotherapy, it is difficult to conclude, as the authors did, that treatment differences by SEP can be ‘demonstrated consistently’(11). Also, no meta-analysis was conducted. Further studies have also been identified that were not included in the review. In a large, good quality UK study, deprived patients were less likely to receive treatment within six months of NHS contact [surgery in 95% of cases], adjusted for age and stage (79). However, in a Swedish study the odds of receiving any surgical treatment or any resection were not associated with income or education level [taking into account age, sex, stage, region, hospital type and marital status] (90) and in a UK study deprivation was not associated with lower likelihood of surgery (76) but age, sex, stage and co-morbidities were not taken into account. Another UK study using audit data also found no association between deprivation and receipt of surgery but did find that affluent patients were more likely to undergo curative resection in a univariable analysis (91). Therefore the evidence that receipt of colorectal surgery is affected by SEP is inconclusive and the quality of the studies variable. However, there is some evidence to suggest that the type of surgery received may vary by SEP. The traditional treatment for colorectal cancer is excisional surgery. For rectal cancer this may require abdomino-perineal excision [APE] which involves the removal of the anal sphincter and leaves the patient with a permanent stoma. The preferred procedure for colonic or high rectal cancer is anterior resection [AR] which allows sphincter retention and appears to improve survival (92). A review reported that the chances of undergoing APE were higher if the patient was of lower SEP [three UK studies] (11). In a small UK study of 486 patients, those in the most deprived group were less likely to undergo surgery and, if they did, were more likely to have a permanent stoma (93). However, this was a univariable analysis and so other relevant factors were not taken into account. But similar results have been found using better quality multivariable analysis studies. In a 2010 UK study, patients from deprived areas were less likely to receive anterior resection [AR], the preferred procedure for rectal cancer, taking into account age, sex and year of admission, but not stage (77). A good quality study using NYCRIS cancer registry data also found that a patient was more likely to receive APE if of deprived SEP (92), in a multivariable analysis including stage. In a Swedish study where stage 25 was included, lower income groups were less likely to receive AR compared to the highest quartile income group. However, when education level was used as the measure of SEP no significant differences were found (90). As education is a less sensitive measure of SEP than income (5), this could help explain the different pattern of results seen. 3.2.2.2 Lung cancer A systematic review of the published evidence for socio-economic inequalities in receipt of lung cancer treatment [surgery, chemotherapy, radiotherapy and any unspecified treatment] has been carried out as part of this PhD. See chapter 7 for the full protocol, methods and results of this review. 3.2.3 Evidence for inequalities: other factors 3.2.3.1 Age In a 2001 review (94), it was found that older colorectal patients were less likely to receive adjuvant therapy but, as many of the studies included in the review did not control for stage or co-morbidities, it is difficult to know how much these factors might have impacted on the decision. However, in studies that did take these confounders into account older people did appear to receive less intense adjuvant treatment (94). Other, more recent, studies also confirm that age is significantly associated with receipt of treatment (81-89, 95). Older patients were significantly less likely to receive treatment of any sort. 3.2.3.1.2 Age: Lung cancer Under-treatment of lung cancer in elderly patients has now been recognised as a worldwide issue (70). Older patients tend to be under-represented in clinical trials and so there is less of an evidence base to determine whether treatment would be suitable for them. Elderly patients are often less able to tolerate treatment complications and more likely to experience chemotherapy toxicity but there is no evidence to suggest increased toxicity from radiotherapy, compared to younger patients, and elderly patients who undergo surgery do appear to benefit (70). However, decisions to treat tend to be made using fairly crude measures of suitability such as number of co-morbidities or PS which may result in sub-optimal care for older patients, as the 26 interaction between age, co-morbidity, patient well-being and suitability for treatment is likely to be complex. 3.2.3.2 Co-morbidity and performance status Higher levels of co-morbidity were significantly associated with a lower likelihood of surgery and adjuvant therapy in some colorectal cancer studies (82, 83, 85, 89), but not all (81). Although co-morbidity and PS measure different things [number of concurrent health conditions and general health status respectively] both variables may be used as a measure of suitability for lung cancer treatment when patients are assessed. However, it is unclear how well co-morbidity and PS capture this ‘fitness for treatment’ concept. Where PS is not available co-morbidity score may be being used as a surrogate measure of this. In one of the few studies that looked at both co-morbidity and PS, the authors found that the number of co-morbidities did contribute to PS but were also additionally associated with likelihood of treatment (64). A recent study found that PS but not the number of co-morbidities, was associated with poorer NSCLC survival (96) suggesting that the presence of co-morbidities is not a particularly good measure of how well a patient was likely to do and so should not be used as a barrier to receipt of treatment if the patient has a good PS score and appears otherwise well. Another study found that PS did not appear to be a major influence on receipt of treatment according to guidelines, whereas those with a higher number of co-morbidities were less likely to receive guideline treatment for early stage lung cancer (72). Interestingly, they found that co-morbidity was higher in early-stage compared to late-stage patients possibly because lung cancer was picked up earlier in patients who were regularly monitored for other conditions. However, patient age rather than PS or comorbidity appeared to be the main factor on which choice of treatment was based (72). 3.2.3.3 Sex Sex has been less consistently associated with access to treatment. In some colorectal cancer studies men were found to be more likely to undergo surgery resulting in 27 sphincter loss [APE], probably due to differences in pelvic anatomy between the sexes (94). A study using NYCRIS cancer registry data also found that a patient was more likely to receive APE if male (92). Women were more likely to receive AR compared to men (77, 90). Sex was also found to be a significant factor in the likelihood of any surgery in other studies (83, 86, 89, 90). Sex was not a significant factor in access to chemotherapy or radiotherapy in the majority of studies examined (81, 83, 86-88) but was in two (84, 85) and in a further study that looked at chemotherapy alone (82). 3.2.3.4 Ethnicity In the USA differences in access to colorectal cancer treatment and outcome have been observed by race (94, 97) but some of these differences may also be related to differences in income and SEP (85, 97). There are also issues when comparing American and UK studies due to the differences in the healthcare systems, as within non-universal health care systems the type of insurance held can affect access to treatment (85). 3.2.3.5 System factors The introduction of a multidisciplinary team [MDT] to a hospital in Glasgow resulted in an increase in the number of non-operable NSCLC patients receiving chemotherapy and was also associated with an increased rate of staging, compared to an earlier cohort who were assessed only by a medical oncologist (98). A 2001 review of patient and provider characteristics on treatment and outcomes in colorectal cancer found that patients were more likely to receive sphincter-saving surgery [i.e. AR] when they were treated by a surgeon with a high volume caseload or who had undergone subspecialty training (94). Removal of sphincter [i.e. APE] was more likely if operated on by a surgeon undertaking fewer than seven cases per year (92). 3.3 Summary: Socio-economic inequalities in receipt of treatment From this narrative review of socio-economic inequalities in receipt of treatment for a number of cancers, it can be seen that there is some evidence that receipt of treatment is socio-economically patterned. However, this is not clear cut and many studies do not take into account cancer stage at diagnosis, number of co-morbidities or 28 the general health status of the patient, all factors that might potentially vary by SEP and thus help explain socio-economic inequalities in treatment. No systematic review of the evidence taking into account study quality [including control for appropriate confounders] has been conducted and this is required in order to clarify whether inequalities in receipt of treatment are found for lung cancer. The next chapter will examine the evidence for socio-economic inequalities in time to presentation, referral, diagnosis, and treatment for cancer, as well as the potential reasons for this. 29 Chapter 4. Intervention-generated inequalities in cancer care: time intervals 4.1 Background In this chapter, the background relating to the importance of early diagnosis and measures to improve this will be described. The evidence for socio-economic inequalities in time to presentation, referral, diagnosis and treatment, and the reasons for this, will then be examined. 4.1.1 Delay or interval? In the earlier literature the term ‘delay’ appears frequently. However, latterly this mostly appears to have been replaced with the term ‘interval’ and the discussion centres more around early diagnosis rather than a delay in diagnosis. The use of ‘delay’ implies that the time taken is longer than is desirable or acceptable but, within studies that use this term, it is not always the case. It may be appropriate to use ‘delay’ when a time interval is longer than a specified target time but otherwise the term ‘interval’ will be used. However as previous studies have used ‘delay’ the term will still appear in this chapter, when referencing these studies. 4.1.2 Early diagnosis Early diagnosis of cancer is thought to be important for improving outcomes, as survival is better for those diagnosed with early-stage cancer (99). Early diagnosis may also result in longer intermediate survival for patients with SCLC and later stage NSCLC. Early palliative treatment can greatly relieve distressing lung cancer symptoms such as pain, breathlessness and persistent cough [Dr M Peake, personal communication]. Patient anxiety can also be greatly ameliorated by prompt referral and diagnosis. In England the National Awareness and Early Diagnosis Initiative [NAEDI] is a scheme designed to encourage early presentation of patients to primary care and to improve GP cancer recognition and referral. It proposes that delays may lead to diagnosis at a later disease stage and thus result in ‘potentially-avoidable’ deaths (99). It has also been suggested that inequalities in delay might contribute to socio-economic differences in cancer survival (99). 30 A NAEDI delay pathway has been constructed to suggest why delay at different points on the pathway may occur and how these might impact on survival [fig 4.1]. This model dovetails in many respects with the proposed model for intervention-generated inequalities (7), as it considers access to primary care, as well as uptake of, and delay in, care; all of which may result in inequalities in outcome. However, the model does not break down into detail the stages in primary and secondary care where delays might occur. Figure 4.1. The NAEDI pathway (99) 4.1.3 Definition of important time points and target times In the research to date, poorly defined definitions of the important time points that characterise delay periods have resulted in a number of inconsistent and incomparable studies. To address this issue, an international consensus working group [CWG] has now formulated a standard set of definitions relating to delay intervals and time points and a checklist for researchers examining early diagnosis (100). Therefore these will be utilised for this present study. In the UK the National Cancer Plan [2000] proposed target time intervals of 14 days from urgent GP referral to first outpatient assessment, one month [31 days] from diagnosis [decision to treat] to treatment, and two months [62 days] from GP referral to first treatment, to be in place by 2005 (24). A 2007 study of 342 patients seen in a 31 Manchester hospital between 2003 and 2005 found that although all patients were seen within the 14 day referral interval, the 62 day treatment target was not being met (101). This was attributed to complex patient pathways and delays in investigations and initiation of treatment. 4.1.4 Theoretical models of delay The CWG suggested that there is lack of an underlying theoretical model to explain delay (100). However, two delay models have been used in the literature: the Hansen and Anderson models. An early model of cancer delay, the Anderson model, attributed the majority of delay to patient rather than system factors. However, it has been suggested that this may be ‘an artefact of research focus’ and that system delay may be an equally important but under-researched area (102). For example, figure 4.2 shows the detailed diagnostic and staging clinical pathway for lung cancer [NICE guidance 2011] (22). Patients with suspected lung cancer are often initially referred for a chest x-ray and the diagnosis confirmed by bronchoscopy. Other investigations including CT scans may then be deployed to help determine the most appropriate treatment (24). Hence there are a number of system stages where delays can be introduced. A recent review looked at the application of the Anderson Model of delay in cancer diagnosis, as the authors proposed that the use of a robust theoretical framework could improve the investigation of delay, and help develop interventions to reduce time to presentation, diagnosis and treatment (103). They produced a refined Anderson model that took into account system and disease factors, as well as patient factors, and split time intervals into appraisal, help-seeking, diagnostic, and pre-treatment intervals [Fig 4.3]. This results in a model that is fairly similar to the Hansen model which also considers patient, doctor and system delay, with further sub-divisions [Fig 4.4] (104). These models, when used in conjunction with NICE treatment decision matrices [Fig 1.1 shows the matrix for lung cancer] (22), can allow examination of intervention-generated inequalities at all stages of the cancer care pathway. 32 Figure 4.2. Diagnostic and staging clinical pathway for lung cancer (22) Figure 4.3. Updated Anderson model of delay (103) 33 Figure 4.4. Hansen model of delay (104) I will now consider the intervals on the care pathway where delay might occur and the potential reasons for this. I will also examine some of the previous problems in measuring delay and the different ways that this has been done, with the resultant difficulties in comparing studies. Evidence for socio-economic differences in time to presentation, referral, diagnosis and treatment within the cancer care pathway, taking into account patient and system factors that may influence delay, will also be examined, including the evidence specifically relating to lung cancer. 4.2 Evidence for inequalities in cancer care time intervals Table 4.1 summarises the previous evidence from studies examining inequalities in lung cancer care time intervals, by SEP. The table details the time interval examined, the country of the study, years of diagnosis, size of the study, and the median time for the interval in the lowest and highest SEP group [if reported], as well as the type of analysis carried out [whether univariable or adjusted for confounders in a multivariable analysis]. The evidence from the lung cancer studies included in this table, as well as the evidence for other cancers, is discussed in more detail below. 34 Table 4.1. Summary of previous evidence from studies examining lung cancer time intervals and SEP Interval low SEP associated with: Specific Interval/ proxy measure study Country N Years of Diagnosis Median/ mean delay Method Further Info Longer delay Shorter delay No assoc Patient Stage at presentation Lyratzopoulos et al (2012) (68) England 13,286 2006-2009 Not calculated OR of advanced stage at diagnosis X Stage at presentation Lyratzopoulos et al (2012) (66) England 16,714 2006-2010 Not calculated OR of advanced stage at diagnosis X Patient Macleod et al (2009) (105) review USA 34, 23, 119 Pre 1973, 1975 + 1990 Not shown Not detailed X (1/3) X (2/3) Primary care First consultation to referral Macleod et al (2009) (105) review Norway 40 pre1996 Not shown Not detailed X Referral Contact with GP – hospital appointment Neal and Allgar (2005) (106) England 2950 2002 patient survey 30 (professional) 30 (unskilled) Generalised linear modelling X Diagnostic FHA to diagnosis (secondary care delay) Neal and Allgar (2005) (106) England 3199 2002 patient survey 17 (professional) 13 (unskilled) Generalised linear modelling X Referral to diagnosis Berglund et al (2010) (73) Sweden 3370 1996-2004 17 (high education) 32 (low education) Kaplan Meier. Time to diagnosis by education level X Image-diagnosis Yorio et al (2009) (107) USA 482 2000-2005 16 (6-43) Cox regression X Referral - diagnosis Dalton et al (2011) (108) Denmark 18,103 2001-2008 Not calculated Logistic regression: OR of diagnostic delay > 28 days 16,713 with stage and diag date X Treatment Referral-treatment Campbell et al (2002) (86) Scotland 661 1995-1996 33 (15-104) high SEP 25 (13-77) low SEP Kaplan Meier, Cox regression X Image-treatment Yorio et al (2009) (107) USA 397 2000-2005 50 (30-84) high income 65 (39-110) low income Cox regression Sig at uni but not multivariable level X Diagnosis-treatment Yorio et al (2009) (107) USA 299 2000-2005 29 (17-55) high income 36 (21-56) low income Cox regression Sig at uni but not multivariable level X Detection- surgery Saint-Jacques et al (2008) (109) Canada 108 2005 107 (73-141) results by education level not shown Multifactorial regression Those who received surgery inc (108 out of 540 NSCLC) X 35 4.2.1. Patient interval 4.2.1.1 Definition and background Patient delay can be defined as the time between onset of symptoms and initial presentation to the GP or other healthcare professional (110). It has recently been recommended that the term ‘patient delay’ should no longer be used, and that ‘patient interval’, split into an ‘appraisal interval’ and help-seeking interval’, is more meaningful and so should be adopted (100). However as I am reviewing previous literature where only the term ‘patient delay’ has been utilised, it seems appropriate to still include this term here. It is difficult to measure patient delay as this requires the patient to accurately remember when they first noticed the onset of symptoms, which may be subject to recall bias. Unlike system delay, which can be examined using secondary data, the calculation of patient delay generally requires patients to be interviewed or to complete a questionnaire post-diagnosis. Alternatively, stage at diagnosis can be used as a proxy measure of late presentation. 4.2.1.2 Evidence for inequalities in patient interval Time to consultation might be influenced by fear or by poor symptom recognition and disease awareness, and this awareness may be lower in more deprived socio-economic groups (111). In a systematic review summarising risk factors for delayed presentation for a number of common cancers an association was found between lower SEP and increased delay for upper GI and urological cancer, older age and patient delay for breast cancer and lower education level and delay for breast and colorectal cancer (105). Generally, not recognising the seriousness of symptoms was found to be a significant factor in increasing delay in presenting, especially when symptoms were vague and non-specific (105). The availability of social support positively influenced time to presentation for breast, colorectal and endometrial cancer but did not appear to reduce delay for lung, upper GI or urological cancer (105). Older age but not SEP was associated with shorter patient delay in a Dutch study including 10 cancer types (104). 36 Systematic review evidence identified co-morbidity, pain and social support as potentially important factors in reducing patient delay in presenting with suspected colorectal cancer but age and sex were not found to have an influence (112). The evidence for SEP was more mixed. A number of older studies found no relationship between lower SEP and increased delay. Other studies found that lower levels of education and residing in a rural location increased delay (112). Patient delay was found to be longer for rectal than for colon cancer (112, 113) possibly because the symptoms associated with rectal cancer such as rectal bleeding are often found in more benign conditions but embarrassment may also be a factor (112). An Italian study found that patients with lower levels of education had non-significantly longer time intervals between colorectal cancer symptom onset and first consultation (114). 4.2.1.2.1 Evidence for inequalities in patient interval: Lung cancer A 2009 review of two systematic reviews looking at risk factors for delay in common cancers found only three small, old studies that considered SEP and patient delay for lung cancer and only one of three found an association (105). Stage at diagnosis can also be used as a proxy measure of late presentation. Although socio-economic inequalities in stage at diagnosis have been noted for some cancers which might suggest that there are differences in timeliness of patient presentation by SEP (66), three studies examining socio-economic inequalities in stage at diagnosis for lung cancer did not find any significant differences by SEP (66, 68, 115). The availability of social support did not appear to reduce delayed presentation for lung cancer (105). A small qualitative study of patient delay in lung cancer did however find that family support decreased delay in reporting symptoms. Many patients delayed going to the GP as they thought that they would be blamed for their illness due to their smoking or, if they were not a smoker, they thought that lung cancer was highly unlikely (116). Guilt and fear may also be factors that influence delay (102). In a Scottish study, more deprived rural-dwelling lung cancer patients did not take longer to consult than affluent urbanites. However, living alone and having a long smoking history increased delay (117). A small Turkish study that interviewed 48 lung cancer patients who delayed presenting found that the most common reason for patient 37 delay was due to ignoring symptoms but that socio-cultural and economic factors were also important (118). 4.2.2 Primary care interval 4.2.2.1 Definition and background Within the primary care setting, the time from first presentation to the GP until the referral date can be defined as primary care interval or referral interval. Delays could occur in initiating primary-care investigations and ordering tests before ultimately referring the patient for secondary-care investigation. NICE Cancer referral guidelines were introduced in 2005 and are designed to assist GPs to identify patients at risk of cancer (119). However, although lung and colorectal cancer are common an average GP is likely to see only one or two new patients each year (23, 120) and many patients present with vague symptoms. If symptoms are not immediately suggestive of cancer a ‘treat, watch and wait’ period is a suggested management option for colorectal cancer (119), which can increase delay. If colorectal cancer is suspected then abdominal and rectal examinations and a full blood count should be carried out. Patients presenting with co-morbidities can also make diagnosis more difficult (102). 4.2.2.2 Evidence for inequalities in primary care interval Higher SEP patients [measured by income or education] had shorter delay from first contact with the GP to initiation of primary care investigation [doctor delay] than less privileged patients, in a Danish cancer study including but not separating out lung cancer (104). The authors suggest that wealthier and better educated patients are better able to describe symptoms and thus speed up the investigation process or that doctors relate better to these patients and that, possibly unintentionally, these patients are given a speedier investigation (104). However patient numbers were small and inconsistencies were found between men and women. It has been suggested that socio-economic differences in communicating and presenting cancer symptoms to health professionals may result in longer delays for those who are less ‘convincing’ (111). Some evidence of delay was seen for those 38 (mainly women) who consulted frequently and were subsequently diagnosed with colorectal cancer, although SEP was not examined (112). A small Italian study of 137 colorectal cancer patients found that patients with lower levels of education had longer time intervals from first consultation to surgical referral but that this delay was not significant (114). A 2008 review of influences on pre-hospital delay of colorectal cancer found that ‘initial misdiagnosis, inadequate examination and inaccurate investigation increased practitioner delay’ (112). Older patients were referred more quickly and there was some evidence of lower SEP patients experiencing longer delay but no relationship was shown between sex and primary care delay (112). Appropriate use of guidelines may also reduce colorectal cancer delay although the evidence was limited (112). 4.2.2.2.1 Evidence for inequalities in primary care interval: Lung cancer No individual lung cancer studies were identified that examined inequalities by SEP over the primary care interval. One study identified in a systematic review (105), that looked at a number of cancers and included only 40 lung cancer patients, suggested that higher education reduced delay. However, lower SEP was not associated with pre-hospital delay for any of the six cancers [breast, colorectal, lung, ovarian, prostate, non-Hodgkin’s lymphoma] looked at in a UK study of over 65,000 patients (106) although within this it was not possible to separate-out patient and primary care delay. 4.2.3 Primary care to secondary care referral interval 4.2.3.1 Definition and problems The date of referral is considered to be the time point where there is a transfer of responsibility from primary to secondary care (100). Delay from GP referral date to secondary care investigation [generally first outpatient assessment or first hospital appointment] can be categorised as primary care to secondary care referral delay or as investigation delay. However, within the literature this is mostly just referred to as referral delay (121, 122). However, referral delay can also be measured from first contact with the GP until first hospital appointment (123), so can include both primary care referral delay, which is delay prior to the referral date, as well as post-referral delay (11). 39 Patients in England and Wales with suspected cancer should be seen by a specialist within two weeks of GP referral (24) but it appears that delays are still experienced in waiting for tests and in the use of non-urgent referrals, as well as other delays caused by administrative and communication problems. 4.2.3.2 Evidence for inequalities in referral interval Younger age was associated with increased referral delay for lung, breast, colorectal and prostate cancer and women experienced longer delays compared to men for colorectal cancer but lower SEP was not associated with referral delay [here defined as time from first contact with GP to first hospital appointment] for any of the six cancers looked at in a UK study (106). However, mean rather than median time was used here and as delay intervals tend to be non-normally distributed [right skewed] then the use of mean can artificially inflate the delay values seen and so is not necessarily a very useful measure. This can be seen in a lung cancer study that reports mean and median delay values, where mean delay is much larger than median (121). Referral delay was examined in an MPhil thesis looking at pancreatic cancer (29) which found that time from GP referral to hospital investigation was significantly shorter for less deprived patients in a multivariable analysis including age, sex and co-morbidity. In another study, age, sex, ethnicity and marital status did not appear to influence the likelihood of urgent referral for lung, colorectal, prostate or ovarian cancer (124). SEP was not examined. 4.2.3.2.1 Evidence for inequalities in referral interval: Lung cancer In a small UK study of 247 lung cancer patients, 54% of these were referred from their GP and the other 46% were non-GP referrals such as those presenting via casualty. Urgent patient referral intervals were short, with a median of 1 day [IQR 1-5]. However no time period was reported for non-urgent referrals (125). Median referral delay was 8 days in a Finnish study (121). Although SEP was not examined, age, sex, ethnicity and marital status did not appear to influence the likelihood of urgent referral for lung cancer (124). Younger age was associated with increased referral delay [here defined as time from first contact with GP to first hospital appointment] for lung cancer but lower SEP was not associated 40 with referral delay, in a large UK study (106). However, mean rather than median time was used. 4.2.4 Secondary care interval The secondary care interval can be further broken down into diagnostic and treatment intervals although both of these measures can also include elements of patient and primary care delay, depending on how they have been measured. 4.2.4.1 Diagnostic interval – definition and problems Diagnostic delay is defined as the delay in receiving a diagnosis but there is confusion in the literature over the time period measured. It can be measured from onset of symptoms (113, 126), referral date (113) and from secondary-care investigation (127) to diagnosis. Diagnostic delay may contain elements of both patient delay and healthcare delay (113) and it has been suggested that patient and primary care delays account for over two-thirds of reported diagnostic delay (128). However, a Cuban study [96 patients] estimated that patient delay accounted for 18 days of delay and the health-care system for 62 days (126) [but again mean rather than median time was used]. The lack of consistency in defining diagnostic delay makes comparison between studies difficult. The type of referral route might be important in determining diagnostic delay. In a UK study using national cancer survey data, patients who were referred via their GP had longer diagnostic delay than those who were not (123). Date of diagnosis can be defined in a number of different ways and it is not always clear within studies how date of diagnosis has been derived. The European Network of Cancer Registries uses a priority hierarchy to define date of diagnosis, with 6 categories; starting off with the date of histological or cytological confirmation of malignancy [which can be derived from 3 different hierarchical time points]; then date of hospital admission; down to, with lowest priority; date of death if cancer is only identified at autopsy (100). Therefore, it is possible that within each study patients have a date of diagnosis derived at a number of different time points. In order to calculate meaningful diagnostic and treatment intervals it is necessary to exclude those who have a diagnosis only after death but, even after excluding these records, a 41 high level of inconsistency in diagnosis date may remain. If differences in diagnostic date occur in a random manner then this increases error. However, if there are systematic differences in the way that date is recorded according to SEP [although I could find no evidence for this] then this may lead to bias and have a serious impact on our ability to accurately measure interval inequalities. 4.2.4.2 Evidence for inequalities in diagnostic interval In a UK study lower SEP was associated with increased total diagnostic delay for prostate but not colorectal or breast cancer (106), but again this included an element of patient delay. In the same study SEP was associated with secondary care diagnostic delay for colorectal, ovarian, prostate and breast cancer (106). However, no inequalities in time from first hospital appointment to diagnosis in high compared to low SEP groups were found in a pancreatic cancer study in a model including age, sex and co-morbidity but not stage (29). In a Scottish study diagnostic delay for colorectal cancer was shown to be longer for those living in a rural location further from a cancer centre (86). 4.2.4.2.1 Evidence for inequalities in diagnostic interval: Lung cancer Socio-economic position was not associated with secondary care delay [from first hospital appointment to diagnosis date] for lung cancer in a UK study of over 3,000 patients (106). Nor was it associated with increased total diagnostic delay for lung cancer (106), but this calculated the time from first symptom to diagnosis, so included an element of patient delay. Mean rather than median delay was also used. A small US study looking at time from imaging to diagnosis also found no association with SEP when using either education or income as the measure (107). In contrast, a Swedish study did find that lower SEP lung cancer patients experienced longer delays between referral and diagnosis, with a median wait of 32 days for low education and 17 days for high education level patients (73). However, this was a univariable analysis. But a large, good quality Danish study of over 18,000 lung cancer patients also found that those with a higher education level were more likely to have a diagnosis within 28 days from referral [the target waiting time] compared to those with shorter education, adjusted for age, sex cohabitation status and co-morbidity, in 42 both early and late stage patients (108). A significant association was only found for early stage patients when income was used as the measure of SEP however. 4.2.4.3 Treatment interval – definition and problems The treatment interval is generally defined as the time between diagnosis and treatment (113, 129). However, it can also be measured as time from first symptoms (130), time from first consultation (79) or receipt of referral letter (122) to treatment, again making comparison between studies difficult. There may be valid reasons why adjuvant therapy is delayed after curative surgery. These could include post-operative complications, longer recovery due to cancer complications, and poor pre-operative performance status due to presence of co-morbidities, advanced stage and age (131). Post-operative complications may also reduce the likelihood of chemotherapy after surgery. 4.2.4.4 Evidence for inequalities in treatment interval SEP [measured by household income] did not affect time from diagnosis to treatment for cancer in a Canadian study. Older patients and women had lower waiting times (129). Women with high household income experienced shorter system delays [from initial primary care investigation to treatment] than those less economically privileged, in a Danish cancer study (104). In a large, good quality study using data from over 71,000 patients from 3 UK cancer registries and employing an area-based measure of SEP, deprived colorectal cancer patients were more likely to receive late treatment (79) [in a model including SEP, stage, age and time to treatment from first contact]. In a German study of 86 patients treatment delay for colorectal cancer [measured as total time from first symptoms to treatment] was associated with SEP [as was type of cancer and marital status but not age or sex] (130). However, this measurement of treatment delay included patient as well as practitioner delay and the patient delay actually accounted for over 70% of the overall delay. A French study found no influence of occupation [as a measure of SEP] on time to treatment for colorectal cancer (132). 43 4.2.4.4.1 Lung cancer 4.2.4.4.1.1 Length of time to treatment interval A number of lung cancer studies report time to treatment intervals. One small UK study found that the time from GP referral to treatment was a median of 60 days (IQR 44-85) with the time from secondary care specialist referral to treatment varying by treatment type [25 days for surgery, 16.5 days for chemotherapy and 43 days for radiotherapy] (101). A small Indian study of 165 patients found that median time from diagnosis to treatment was 20 days (133) and, when they looked at a number of older studies, these found times varied from 10 to 30 days [but, as they reported, these used mean rather than median time so need to be interpreted with caution]. A small UK study from 1998 found that median time from receipt of referral letter to treatment was 48 days (122). A Finnish study reported lung cancer treatment delay as median 15 days from diagnosis and 30 days from first visit to specialist (121). 4.2.4.4.1.2 Evidence for inequalities in treatment interval In a Swedish study older patients had longer delay in an unadjusted analysis of time to treatment for non-small cell lung cancer. SEP was not measured (134). A small Scottish study of time between referral and first treatment for lung cancer found no association between SEP and time to treatment (86) in both univariable and multivariable analyses [taking into account age, stage and health board]. A small American study found a univariable association between income [as a measure of SEP] and time from image to treatment, but not when education was used. In the multivariable analysis only type of hospital remained significantly associated, with those treated in a public hospital having significantly longer delay [median 76 days] than those treated privately [45 days] (107). A small Canadian study found no association between education level and time from detection to surgery (109). 4.3 Summary: Socio-economic inequalities in time intervals The evidence for the influence of SEP on time to consultation, referral, diagnosis, and cancer treatment is inconclusive and further investigation is required. A review of the current evidence for lung cancer, as discussed in this chapter, is shown in table 4.1. Results are contradictory for time to diagnosis [two studies found 44 inequalities in time from referral to diagnosis and two found no association in time from FHA to diagnosis and image to diagnosis, with SEP]. No association was found between SEP and time to presentation [using stage at diagnosis as a proxy] and time to treatment. However, very few studies used median time, most studies were very small and had low statistical power [particularly in the time to treatment analyses] and many employed only univariable analysis and so were of poor quality. Inconsistent measures of time period were also used making comparison between studies difficult. Therefore the evidence for socio-economic inequalities in delay for lung cancer is inconclusive. The evidence for whether delay contributes to socio-economic differentials in survival is considered in section 5.4 of chapter 5, which examines inequalities in survival. 45 Chapter 5. Inequalities in cancer survival This chapter examines the evidence for inequalities in cancer survival, why inequalities might be found and what factors might influence inequalities in survival. 5.1 Introduction Socio-economic inequalities in survival have been described for most common cancers (135) and although introduction of the NHS Cancer Plan in 2000 may have improved overall cancer survival there does not appear to be a concurrent reduction in the deprivation survival gap in England (136). Three recent studies have looked at relative survival for common cancers in England, Wales and Scotland. They all found that although cancer survival was improving the deprivation gradient was getting steeper (25, 136, 137). Cancer survival may be influenced by patient-related factors such as performance status and co-morbidities (138), tumour factors relating to stage and aggressiveness of the cancer, and to service-related factors such as treatment decisions by the multi-disciplinary team [MDT] (102). Post-operative complications may also be relevant, as might delay. 5.2 SEP and cancer survival 5.2.1 Why might socio-economic inequalities in cancer survival be found? Variation in co-morbidity, stage at diagnosis and in receipt of treatment have all been suggested as potential explanations for socio-economic inequalities in cancer survival, although few studies have adjusted for these factors or looked at how they vary by SEP (137). Poorer survival is generally attributed to later presentation by lower SEP patients. A 2006 review summarised a decade of research on the association between SEP and cancer survival (26) and although stage at diagnosis did seem important in determining likelihood of survival [particularly for colorectal and breast cancer] the authors felt that socio-economic differences in ‘access to optimal treatment’ (26) partially explained survival differences. One recent study has attempted to examine receipt of treatment and survival by SEP for breast and colorectal cancer, using the theoretical framework of the ‘inverse equity 46 law’ to determine whether survival inequalities widen after the introduction of a new treatment, on the assumption that it is more likely to be offered to the more affluent patients (35). However, as they were not able to determine the actual treatment received by patients but were only able to graph the date of introduction of a new treatment and the survival trends after that time, it was difficult to draw any clear conclusions. It has been suggested that social support and being married may lead to better cancer survival (26, 139). The less affluent tend to have lower levels of support although the effects may differ between the sexes (139). Differences in smoking rates may be important. A Norwegian study of women found that smoking status prior to diagnosis was a predictive factor for socio-economic variation in cancer survival (140). 5.2.2 Evidence for socio-economic inequalities in cancer survival Stage at diagnosis appears to be the main prognostic factor for colorectal cancer survival (141) but survival rates have consistently been shown to be worse for those with lower SEP (10, 11, 79, 142-144). A number of studies have also shown that survival rates are lower for breast cancer patients who reside in more deprived areas (137, 143, 145-148). There is evidence for a socio-economic gradient in survival for lung cancer in some studies, with poorer survival in more deprived groups (73, 149-151), but not in others (143, 152, 153), although no studies found an inverse gradient – see table 5.1.The evidence will be explored in more detail below. 5.2.2.1 Colorectal cancer In a UK study looking at five year colorectal cancer survival, area-based socio-economic deprivation was associated with increased mortality (142). Age, stage and number of co-morbidities were also significant. Similar results were found in a Danish study (144) where it was suggested that co-morbidity may partly explain the social gradient in survival. However, in another study it was shown that survival was less likely if co-morbidities were present but a socio-economic gradient was not found (138). Deprived colorectal cancer patients had poorer three-year survival compared to the more affluent using data from three UK cancer registries, in a large, good quality study (79). However, no differences in survival were seen for those patients who had 47 treatment within one month of diagnosis but the socio-economic survival gradient remained in those who had later or no treatment, even when age and stage were taken into account (79). Earlier access to treatment attenuated socio-economic differences in survival in this cohort suggesting that improved access to treatment for all might lead to an overall improvement in outcome (79). Another study appears to support this finding, as within a randomised controlled clinical trial of chemotherapy and radiotherapy for colorectal cancer, when crude and relative survival at one and five years were analysed by socio-economic deprivation then the excess hazard of death was not significantly higher in the more deprived groups, adjusted for age, sex, stage and cancer site (154), suggesting that equal treatment results in equal outcome, regardless of SEP (154). However, it may also be that the type of patient who is considered for a clinical trial [generally younger and fitter] may also have an influence here. A recent study concluded that excess deaths occurred in the first month post-diagnosis in more deprived colon cancer patients but this excess persisted for longer in deprived rectal cancer patients (155) although stage and co-morbidity were not included in the analysis. Lower SEP was associated with a higher risk of 30-day post-operative mortality (65) although much of this effect was explained by differences in grade and stage of the tumour and by emergency compared to elective surgery. In this study more deprived patients presented with later stage cancer (65) in contrast to another study which did not find this (156). This small but good quality study (156) found better survival for more affluent patients who underwent curative surgery. However, there were no socio-economic differences in stage, mode of presentation or treatment offered and so they suggested that outcome differences might be related to a more compromised tumour-host response in the more deprived patients. 5.2.2.2 Lung cancer Table 5.1 summarises the results from studies located in a review of the literature that examined the socio-economic gradient in survival for lung cancer. Details of the country of study, study size, years of diagnosis, and confounders included are shown, as well as the hazard ratio of survival in the lowest compared to the highest SEP group 48 for univariable and where conducted, multivariable analysis. Some contradictory findings can be seen. A recent UK study found that three-year survival in 1828 early-stage NSCLC patients was 50% in the most affluent group and 39% in the most deprived (157) with a univariable HR=1.35 [95% CI 1.07 to 1.70]. However this HR was attenuated and became non-significant [HR=1.24 (95% CI 0.98 to 1.56)] in a multivariable model including age, sex, co-morbidity and receipt of surgery. A similar result was seen in an earlier, smaller UK study that examined one-year survival. A significantly decreased univariable OR of survival was found in the lowest compared to the highest SEP group and this remained when age, sex, histology, basis of diagnosis, number of symptoms, consultant specialty and stage were included in the model but not when receipt of treatment was included (158). The authors suggest that poorer survival in more deprived patients may therefore be due to differences in receipt of treatment rather than case mix variation (158). In a clinical trial education-level was not associated with survival when standard treatment was received by all (159) which would support this theory. However, two studies using lung cancer audit data did not find socio-economic inequalities in survival whether or not treatment was included (152, 153). 49 Table 5.1. Summary of previous evidence from studies examining likelihood of survival by SEP (in most deprived [low SEP] compared to least deprived [high SEP] group), for lung cancer Univariable HR Multivariable HR Low SEP associated with: Study Country N Years of Diagnosis Confounders Further information HR (IMD=5) HR (IMD=5) Higher HR Lower HR No association Berglund et al (2012) (157) England 1,828 2006-2008 Age, sex, co-morbidity, receipt of surgery 3 year survival – early stage 1.35 (1.07 to 1.70) 1.24 (0.98 to 1.56) X X Cheyne et al (2013) (115) England 1,432 2008-2010 Univariable Kaplan-Meier only 1 year survival 39% in high SEP 33% in low SEP X Jack et al (2006) (158) England 695 1998 Age, sex, histology, basis of diagnosis, stage, number of symptoms, consultant specialty, treatment OR of survival NOT HR of death 1-year survival 0.63 (0.35 to 1.13) P for trend = 0.02 0.84 (0.24 to 2.45) P for trend = 0.19 X X Jones et al (2008) (160) England 34,923 1994-2002 Age, sex, stage, histology, treated at cancer centre, travel time Cont measure NYCRIS data -- 1.001 (1.000 to 1.002) X Riaz et al (2012) (161) England 77,349 2004-2006 Age, sex HR inc with age and derivation in full dataset but not by deprivation in receipt of surgery dataset Data not shown Data not shown X X Rich et al (2011) (153) England 48,453 2004-2007 Age, sex, PS, histology, stage 1.03 (0.97 to 1.09) 1.00 (0.95 to 1.06) X Rich et al (2011) (152) England 34,365 2004-2008 Age, sex, stage, PS, ethnicity, CCM, surgery, surgery centre, radio centre, trial centre Clustering by NHS trust 1.04 (1.00 to 1.08) 0.99 (0.94 to 1.05) X Berglund et al (2010) (73) Sweden 3,369 (1,405 women) 1996-2004 Age, sex, histopathology, PS, smoking status, treatment Sig diff for women but not men or overall. Results for high compared to low education in women shown 0.76 (0.60 to 0.96) exc treatment 0.84 (0.65 to 1.08) inc treatment X X Hall et al (2004) (162) Western Australia 9,080 1982-2001 and 1991-2001 Age, sex, yr of diagnosis, marital status, indigenous status, histology, surgery, insurance status, hospital status, remoteness index, rural 5-year survival --- 1.07 (0.94 to 1.20) and 1.05 (0.93 to 1.20) X Hardy et al (2009) (163) USA – I SEER area 83,101 1991-2002 Age, sex, race, standard therapy 5 year survival 19,519 early stage, 80,519 late stage --- 1.38 (1.10 to 1.71) early 1.09 (1.06 to 1.12) late X 50 Table 5.1 (cont). Summary of previous evidence from studies examining likelihood of survival by SEP (in most deprived [low SEP] compared to least deprived [high SEP] group), for lung cancer Univariable HR Multivariable HR low SEP associated with: Study Country N Years of Diagnosis Confounders Further information HR (IMD=5) HR (IMD=5) Higher HR Lower HR No association Schrivers et al (1994) (164) various Review inc 3 studies X Yim et al (2012) (165) Korea 261 1999-2002 Age, sex, stage, PS, family history, no of outpatient visits 1.25 (0.87 to 1.80) 1.46 (0.99 to 2.14) X Table 5.1b. Summary of previous evidence from studies examining the likelihood of survival by SEP (in most deprived compared to least deprived group), for lung cancer: relative survival Relative survival (RS) low SEP associated with: Study Country N Years of Diagnosis Confounders Further information Poorer RS Better RS No association Coleman et al(2004) England and Wales 107,317 1996-1999 Age, sex 5 year relative survival Sig in men only -1.4%(dep gap in men) X (men) X (women) Rachet et al (2010) England 303,422 1996-2006 Age, sex 1 year relative survival (2006 results shown) -1.6%(dep gap in men) -3.1%(dep gap in women) X Shack et al (2007) Scotland 20,851 1996-2000 Age, sex 5 year relative survival Sig in men only -1.6%(dep gap in men) X (men) X (women) 51 5.3 Other factors influencing survival Local variation in cancer management such as access to specialist surgeons might affect survival (26). A review found that cancer mortality was lower when treated by a specialist surgeon or centre (166) but the poor quality of some of the studies included meant that the evidence was not conclusive and no consideration was given to outcome differences by age, sex or SEP. In contrast, a 2001 review found that, in the majority of studies it examined, surgeon volume or experience did not significantly affect colorectal cancer mortality (94). However, there is some more recent evidence to suggest that outcomes are better for colorectal patients who are managed by a specialist surgeon. The odds of death within 30-days post-operatively and within five-years were significantly lower for those managed by a specialist, taking into account age, sex, stage, SEP, tumour site and whether elective or emergency surgery was performed (167). Breast cancer survival in the UK is lower than in other European countries with similar healthcare systems such as Norway and Sweden, possibly due to more advanced stage at presentation (168). Older patients also had poorer survival with a suggestion that they may be less likely to receive optimal treatment (168). 5.3.1 Lung cancer Five-year survival for lung cancer is lower in England than in other European countries with similar healthcare systems, with a suggestion that this may be due to differences in management and access to treatment (169). A recent study found that lung cancer survival differences between six countries [including the UK] were partly due to differences in stage at diagnosis (170) but it was possible that some of the differences observed might also be due to problems with data quality and comparability, and to national differences in clinical decision making when dealing with borderline suitable patients such as older patients with poor PS and co-morbidities (170). A study that simultaneously evaluated the management and survival of lung cancer patients in Teesside in the UK and in Varese in Northern Italy, two areas with similar lung cancer incidence, found that the resection rate was higher [24% compared to 7%], and three-year survival better [14% compared to 7%], in Italy compared to the UK 52 (171). The lower resection rate in the UK was ascribed to patients presenting with later stage cancer, having more aggressive tumours, and higher rates of co-morbidity. SEP and time interval delays were not measured, however. A review examining the prognostic and predictive role of tumour type on survival found contradictory evidence, with some studies showing better survival for those diagnosed with adenocarcinoma and others suggesting more favourable outcomes for squamous cell cancers (19). Poor general health as measured by PS, but not the number of co-morbidities, was associated with poorer lung cancer survival (96) in a recent study. Better survival is found for those with good PS. In a study of 26,957 patients in Japan, the group with good PS [PS=0] contained the highest percentage of non-smokers and over half had stage 1 cancer (172). Smoking status was also found to be an independent prognostic factor for survival in this study, as were stage, sex and age (172). Women were shown to have better survival than men in a meta-analysis although the reasons for this are not clear. For early-stage survival it may be related to women being more likely to have a more slow-growing adenocarcinoma subtype. However, sex appeared to be an independent prognostic factor when stage, histology and smoking status were taken into account (173). Age was shown to be strongly associated with early post-operative death [within 90 days of surgery], as was poor PS in a UK study using lung cancer audit data (174). A study conducted using NYCRIS data did not find that increased travel time to hospital was associated with poorer survival, in fact the reverse (160). There is a correlation between resection rates and survival (175). A recent UK study found that patients who were resident in trusts that had high surgical resection rates had reduced hazard of mortality compared to those with lower surgical rates, suggesting that survival rates could be improved if a higher proportion of patients were resected (161). The authors also suggest that as these rates were insensitive to adjustment for patient age and SEP then these associations were due to physician or hospital-level factors (176). Patients assessed by an MDT had better survival than those in a previous cohort, who were not (98). 53 5.4 Delay and survival Does delay affect survival? The influence of patient and healthcare delay on cancer survival is unclear, according to one review (26). A scoping review identified 47 studies looking at 13 different cancers, of which 11 found a positive association, 9 a negative association and 29 no association between increased delay and poorer outcome (177). If survival is measured from diagnosis or surgery rather than from the onset of symptoms then there is a possibility of lead-time bias (178) and many studies on delay do not take this into account (177). Diagnostic delay has been implicated as a factor that contributes to the poorer survival of UK cancer patients compared to the European average (99, 102). However, there have been some contradictory findings regarding the importance of delay on survival. In some cases it appears that those with shorter system delays may, in fact, have poorer survival. This has been termed the Waiting Time Paradox [WTP]. The WTP suggests that sicker people are referred more quickly and, as they are more ill, have shorter survival (179). A small study of colorectal cancer patients found that shorter diagnostic interval was associated with higher mortality, but this was only seen for those with obvious alarm symptoms and not found for those presenting with vague symptoms, although there were only 67 patients in this latter group (179). A 2013 Korean study found that a delay of greater than 12 weeks from diagnosis to curative surgery was associated with increased mortality for colorectal and breast cancer patients but not for lung or thyroid cancer (180), so that different effects may be seen depending on the specific cancer. However, another study found that an excess wait [>62 days from diagnosis to curative surgery] for breast cancer had no impact on survival (181) so that contradictory results are also found within cancer types. Although some individual studies have shown an association between delay and colorectal cancer survival (79) others have not (113). A 2007 review could not reach a definitive conclusion as to whether diagnostic and therapeutic delay affected survival for colorectal cancer patients (178), although there appeared to be delay differences between colon and rectal cancer (178, 182). A weak degree of association was found between increased delay and improved survival [which again might support this ‘sicker quicker’ WTP theory], but not all the studies considered tumour stage. The review 54 authors also remarked on the scarcity of studies that included other confounding factors such as age, sex, SEP and co-morbidities (178). A different review found that delaying adjuvant chemotherapy for longer than 8 weeks was associated with poorer overall colorectal survival (131). A systematic review of the influence of delay on breast cancer survival found that patients with treatment delay had poorer survival but this was not seen when stage at diagnosis was taken into account (183). 5.4.1 Lung cancer No association between timely care and survival for lung cancer has been shown in two literature reviews, as both found the evidence inconclusive (184, 185). The most recent review found three studies where timely care was associated with better lung cancer survival, four that found the opposite and eight that found no association. No meta-analysis was carried out. The quality of the studies included was mixed and most did not adequately control for age, stage, histology and co-morbidity (185). The authors suggest that this lack of control for confounding factors is likely to account for why those with more timely care had poorer survival, in the studies that found this. However, a recent study looking at delays in diagnosis and treatment on survival in SCLC patients found that patients who were diagnosed earlier had a poorer prognosis that those diagnosed later and they were able to take into account age, sex, stage and PS (186). 5.5 Summary of chapters 3-5 From the narrative review of inequalities in receipt of treatment for a number of cancers, there is some evidence that receipt of treatment and time to treatment is socio-economically patterned and that this may impact on socio-economic inequalities in survival, but it is not clear cut. The evidence is mixed and there are contradictory findings both within and between cancer types. A number of studies look at ‘cancer’ but analysing combined cancers is likely to mask any between-cancer differences. Differences in tumour development and prognosis may account for some of the between-cancer differences in time to, and receipt of, treatment. 55 Even in studies that focus on a particular cancer, the quality of the studies and different confounders included make comparisons difficult. A 2007 study noted that socio-economic inequalities in co-morbidity, stage at diagnosis and receipt of treatment, factors that may be potential explanations for socio-economic inequalities in cancer survival, have not been well explored (137). The role that delay may play in survival is unclear. In time interval analyses there is a lack of consistency in definitions of delay intervals, and a lack of control for potentially important confounders. 5.5.1 Lung cancer Socio-economic gradients in lung cancer survival are found in a number of studies but these gradients appear to be attenuated or eliminated in a clinical trial where participants received equal treatment and in some observational studies where receipt of treatment was included in the model. If access to treatment does influence survival it is important to investigate factors such as SEP that might influence whether treatment is received. Further work is required to determine whether socio-economic inequalities in receipt of lung cancer care do occur, what factors may be influencing this, and what role treatment inequality and delay might play in survival inequality. Having determined the gaps in knowledge regarding the evidence for inequalities in lung cancer treatment, time to diagnosis and treatment, and survival, both in the lack of research in these areas as well as the poor methodological quality of the studies that have been conducted, and the lack of synthesis of the evidence, suitable research aims and objectives were developed. These are detailed in chapter 6. 56 Chapter 6. Aims and Objectives 6.1 Development of aims and objectives The initial narrative literature review identified that further work was required to determine whether socio-economic inequalities in receipt of, and time to, lung cancer are found; what factors may be influencing this; and what role inequalities in treatment and in the referral, diagnostic and treatment intervals might play in survival inequality. No systematic synthesis of the evidence for socio-economic inequalities in lung cancer treatment, time to diagnosis and treatment, or survival has previously been conducted. 6.2 Aims and objectives The following aims and objectives were developed in order to fill the gaps in knowledge that had been identified, and to help determine whether IGIs in lung cancer care are found: Aim: • To determine if there are socio-economic inequalities in lung cancer care and, if so, to identify where on the pathway of care these inequalities might occur: looking at receipt of treatment; referral, diagnostic and treatment intervals; and survival Objectives: • To conduct a systematic review of the literature on socio-economic inequalities in receipt of treatment for lung cancer • To link routine secondary data in order to identify where on the pathway of cancer care inequalities might occur for lung cancer; specifically to describe the relationship between SEP and: o receipt of treatment; o delay in referral, diagnosis and treatment; o survival; whilst also taking into account age, sex, cancer stage, performance status, presence of co-morbidities and other potential confounding factors 57 This thesis includes two areas of work: a systematic review and meta-analysis of the evidence from the published literature on socio-economic inequalities in receipt of lung cancer treatment; an analysis of secondary data examining socio-economic inequalities in receipt of, and time to, treatment, and survival, for lung cancer. Chapter 7 details the systematic review and meta-analysis of socio-economic inequalities in lung cancer treatment. 58 Chapter 7. Systematic review and meta-analysis of socio-economic inequalities in receipt of lung cancer treatment Summary Background Socio-economic inequalities in treatment may occur for some common cancers. Although the incidence and outcome of lung cancer varies with socio-economic position, it is not known whether socio-economic inequalities in treatment occur. A systematic review and meta-analysis of existing research on socio-economic inequalities in receipt of treatment for lung cancer was conducted to investigate this. Methods MEDLINE, EMBASE and Scopus were searched up to September 2012 for cohort studies of participants with a primary diagnosis of lung cancer [ICD10 C33 or C34], where the outcome was receipt of treatment (rates or odds of receiving treatment) and where the outcome was reported by a measure of socio-economic position. Forty six papers met the inclusion criteria and 23 of these papers were eligible for meta-analysis. Results Socio-economic inequalities in receipt of lung cancer treatment were observed. Lower socio-economic position was associated with a reduced likelihood of receiving any treatment [OR=0.79 (95% CI 0.73 to 0.86), p<0.001], surgery [OR=0.68 (95% CI 0.63 to 0.75), p<0.001] and chemotherapy [OR=0.82 (95% CI 0.72 to 0.93), p<0.001], but not radiotherapy [OR = 0.99 (95% CI 0.86 to 1.14), p=0.89], for lung cancer. The association remained when stage was taken into account for receipt of surgery, and was found in both universal and non-universal health care systems. Conclusions Lung cancer patients living in more socio-economically deprived circumstances are less likely to receive any type of treatment, surgery and chemotherapy. These inequalities cannot be accounted for by socio-economic differences in stage at presentation or by 59 differences in health care system. Further investigation is required to determine the patient, tumour, clinician and system factors that may contribute to socio-economic inequalities in receipt of lung cancer treatment. 60 7.1 Background Socio-economic inequalities in incidence of, and survival from, the majority of cancers have been reported (8, 9, 136). The evidence for socio-economic inequalities in cancer treatment was explored in chapter 3. Although a recent non-systematic review found some evidence for socio-economic inequalities in receipt of treatment for colorectal cancer (11) overall the evidence from studies included in my narrative review was inconclusive. Study quality was variable and many studies did not account for important confounders. However, despite the lack of conclusive evidence for socio-economic inequalities in cancer treatment it has been suggested that socio-economic differences in access to treatment might at least partially explain socio-economic differences in cancer survival (26). Incidence of lung cancer is higher (8, 16), and survival poorer (136), in the most deprived patient groups. However, it is not known whether socio-economic inequalities in receipt of treatment exist for lung cancer and, if so, what contribution they make to overall socio-economic inequalities in outcome. In order to explore the first of these questions a systematic review and meta-analysis of cohort studies examining the association between socio-economic position (SEP) and receipt of lung cancer treatment was undertaken. 7.2 Review Objectives The aim of this review was to summarise the existing literature and to assess the published evidence for socio-economic differentials in receipt of treatment for lung cancer. 7.3 Methods A protocol [Appendix C-C1] was developed and systematic methods used to identify relevant studies, assess study eligibility for inclusion and evaluate study quality. The review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] guidelines (187) [see Appendix C-C2 for PRISMA checklist]. 61 7.3.1 Literature Search The online databases of MEDLINE and EMBASE were searched up to September 2012 to locate studies examining receipt of treatment for lung cancer reported by a measure of SEP. This researcher [LF] developed the search strategy, which was refined with the help of a medical librarian. Slightly different strategies were required for each database [for example MEDLINE recognises the MESH term Lung Neoplasms/ whereas EMBASE does not and uses Lung cancer/]. No language restriction was applied [see Appendix C C3 for full search strategies]. A search of Scopus uncovered no further papers. Additional studies were identified by reviewing the reference lists of all included studies and by using a forward citation search to identify more recent studies that had cited included studies. EndNote X5 software was used to manage the references. 7.3.2 Study Eligibility Studies that met the following criteria were included in the review: primary, cohort studies of participants with a primary diagnosis of lung cancer [ICD10 C33 or C34] reported separately from other cancers; published in a peer-reviewed journal; where at least one reported outcome was receipt of treatment [measured by rates or odds of receiving treatment]; and where receipt of this outcome was reported by a measure of SEP. Any curative or palliative treatment for lung cancer including surgery, chemotherapy and radiotherapy was included. Studies where SEP was included as a descriptive variable or potential confounder but where outcomes for receipt of treatment by SEP were not presented were not eligible for inclusion but the authors were contacted to determine whether relevant data were available that might allow for inclusion in the review. Studies where multivariable analysis was conducted [and included control for a minimum of age and sex as confounders]; receipt of treatment was compared to not receiving treatment; odds ratios [ORs] and 95% confidence intervals of receipt of treatment in low [most deprived] compared to high [least deprived] SEP was calculated; and SEP was not further stratified by another variable, were considered suitable for inclusion in meta-analysis. 62 Acceptable measures of SEP were: area-based indices of deprivation [e.g. Index of Multiple Deprivation (IMD), Townsend Score, Carstairs Index]; and area or individual measures of income; poverty; or education level. 7.3.3 Study selection and data extraction Papers obtained from the database searches were independently assessed by two researchers [LF and a fellow PhD student HW] in three phases: title, abstract and full paper screening. Initial screening of titles was carried out to remove obviously irrelevant papers. However, from a preliminary scoping review by LF, the early pilot searches recovered studies that, although they conducted analyses of receipt of treatment stratified by SEP, did not always mention this in the abstract or title. Therefore, in the title search, any titles that referred to uptake or receipt of surgery, chemotherapy and radiotherapy for lung cancer were retained. Papers with titles that examined disparities in cancer survival/mortality were also included as further checking of the abstract was required to determine if inequalities in access to treatment were also examined. Abstracts were then screened and a subset of studies selected for further review and the full article obtained. Abstracts that referred to socio-economic inequalities in receipt of treatment were retained. Others that referred to racial, ethnic, geographical, sex and age-related disparities in treatment as well as disparities by insurance type were also retained as often these papers also looked at SEP, even if this was not mentioned in the abstract. Papers that considered delay were not included. Nor were studies examining access to hospice care as although some papers were recovered that considered this type of care the search was not designed to pick up delay or hospice care. Therefore not all papers looking at this are likely to have been found and so there was a risk of presenting a selective rather than systematic analysis of these areas. Papers that were not available through Medline or Embase were obtained from Google scholar. Six papers were obtained via Inter-library loans and two were accessed in paper format only from Newcastle University library. 63 Two researchers [LF and HW] independently assessed the selected full papers for eligibility according to the study-eligibility criteria. Any disagreements at any of the screening stages were resolved by discussion between the two researchers in the first instance and with a third reviewer [JA] if agreement could not be reached. Data extraction was carried out by LF using an Access database pro-forma developed for this purpose. Data relating to study authors, journal, study design, year of study, data source, number of participants, years of diagnosis, measure of SEP, confounding variables included in the analysis [such as age, sex, stage, co-morbidities, cancer type/site, vital/performance status, marital status, smoking status, cancer network, health board, hospital, emergency or elective treatment, distance from hospital/travel time, ethnicity, insurance status], type of treatment received [any, surgery, chemotherapy, radiotherapy], statistical tests carries out, outcome measures [treatment rates or odds of treatment], comparator used, significance [p values], precision [confidence intervals], other variables that were significant; were recorded. Data extraction was double-checked by HW in a random sample of 10% of the included studies. There is evidence to suggest that health-insurance status is an important factor relating to access to lung cancer care in countries such as the USA that rely on insurance-based health care systems (188). Insurance status is less relevant and rarely measured in most other countries. Therefore studies were grouped into three categories during analysis: those conducted in a universal healthcare system [UHCS], free at the point of access [similar to the UK]; those conducted in countries with primarily private insurance healthcare systems [non-UHCS, similar to the USA] (189) and those conducted in countries with social insurance healthcare systems [similar to many European countries]. No studies were identified that fell into the third category. 7.3.4 Population definitions A number of papers included the same or overlapping study populations or used a regional study population that was also analysed in another paper as part of a national data-set. Some rules had to be derived in order to define patient populations, deal with study data that was being utilised by more than one paper and to choose data from the most suitable paper for inclusion. 64 A study population was defined as being determined by the geographical area, the source of the data and the time period of diagnosis. If two papers used data from the same source and time period then these were defined as fully over-lapping populations. If a study presented separate data for early and late stage patients these were regarded as separate populations. If a paper contained a regional population that was a subset of a paper containing a national population then these were classified as separate but over-lapping populations. If papers used populations from the same data source where there was only one year of overlapping data and a number of different years either side then these were defined as partially over-lapping study populations [substantially non-over-lapping]. Appendix C4 examines papers with over-lapping populations in more detail, with a breakdown of how the inclusion decisions were made. Quality score was used to determine the most appropriate paper to choose for inclusion in the final meta-analysis. 7.3.5 Quality A quality tool was used to divide eligible papers into six study quality categories with 1 being the lowest, and 6 the highest, quality [see Appendix C5]. Quality assessment was based on internal validity [for measures of SEP and outcome]; type of analysis conducted [multivariable or univariable]; population included [population-based or selective]; quality of reporting; and adjustment for relevant confounders. Quality assessment was carried out by LF and checked by HW. 7.3.5.1 Existing study quality tools A number of existing tools potentially suitable for assessing cohort study quality were considered including the Cochrane Tool for Assessment of Risk of Bias (190); three other tools recommended by Cochrane Review Groups: the Jadad Scale, the Moncrief Scale and the Newcastle-Ottawa Scale (191); the Scottish Intercollegiate Guidelines Network [SIGN] methodology checklist for cohort studies (192); the EPHPP Quality Assessment Tool for Quantitative Studies (193); and The STROBE [Strengthening the Reporting of Observational Studies in Epidemiology] guidelines [a checklist of 22 items 65 that should be included in cohort studies] (194). Details of these tools are found in Appendix C6 with a discussion of the suitability of each. In a systematic review it is necessary to examine the methodological quality, that is - the internal validity, of the primary studies. However, previous systematic reviews have used quality tools that defined quality in a number of different ways. Some quality scales look at internal validity [risk of bias] but also measure external validity and the quality of the reporting, which are not measures of bias within the study. Assessment of quality can therefore be sub-divided into a number of categories such as internal validity [including selection, confounding, statistical analysis and outcome], external validity and reporting. If quality scores are employed then they can be used in a number of ways. It is possible to include all papers regardless of quality, include only high quality papers or include all but use quality to divide papers into sub-groups so that high and low quality papers are analysed separately. Quality can also be used as part of a narrative synthesis to help explain any differences found (195). Having examined a number of potential tools and considered the different ways that quality could be utilised, a quality tool was developed for this review. 7.3.5.2 Development of a quality tool for this review As none of the existing tools considered were entirely appropriate for the type of studies included in this review and, as has been done for previous reviews (185, 188), we devised a suitable tool, adapting and utilising aspects of other available tools. This approach has the benefit of producing a quality tool that is highly specific for the type of studies examined and there was good agreement between researchers when using the tool. Quality in this systematic review was appraised using criteria adapted from existing quality tools [SIGN, EPHPP and STROBE guidelines]. In order to be considered for meta-analysis papers had to present adjusted odds ratios and report confidence intervals. A quality checklist was constructed with a screening question used to split papers into those potentially suitable for meta-analysis [adjusted ORs and CIs reported], and those suitable for narrative synthesis [meet inclusion 66 criteria but do not present adjusted ORs]. Then a 19-point scale that examined study quality under the following headings: internal validity [selection], external validity, quality of reporting of the study, and confounding; with a scoring system, was utilised. The scoring system was then used to further sub-divide studies into 6 quality categories [with 1 being the lowest, and 6 the highest, quality], with the final score determined by the type of population included, internal validity, reporting of the study and confounders included. Cohort studies reporting only univariable analysis are of lower quality than those conducting multivariable analysis, in terms of their ability to control for confounding. However, it was decided to include those papers in the narrative analysis, with their overall lower quality taken into account. Initially Harvest Plot analysis was considered as a method for pictorially displaying the narrative findings, as the authors of this method have suggested that the Harvest Plot could be adapted so that, instead of looking at suitability of study design, quality can be used to distinguish studies of the same design (196). However, the initial Harvest Plot graphs produced were difficult to interpret and so it was decided against this approach for the narrative review. 7.3.6 Statistical analysis Trends in receipt of treatment across SEP groups were assessed in the descriptive narrative analysis of all studies that met the inclusion criteria. Meta-analysis of eligible studies was undertaken using Cochrane Collaboration Review Manager 5.1. Only studies reporting multivariable analysis [quality scores 3-6] were included in the meta-analysis. Natural logs of the ORs and their standard errors [SEs] were calculated for use in Forest plots. Random-effects meta-analysis of the odds of treatment in the lowest compared to the highest SEP group was conducted. Where a study reported the most deprived class as the comparator then reverse ORs were calculated. Studies that presented a single OR as either an OR for a one unit increase in deprivation score or incremental quintile increase in income were not included. Subgroup analyses by treatment type and healthcare system were conducted. The I2 statistic was used to examine heterogeneity. In meta-analyses where a ‘substantial’ percentage [previously defined as I2 >50%] (197) of the variability appeared to be due 67 to the heterogeneity of the studies rather than to chance, further subgroup analyses by stage, histology and quality score were conducted, where appropriate, in order to examine potential sources of heterogeneity. A funnel plot was utilised to assess potential publication bias. Multiple papers using the same or overlapping study data were included. Sensitivity analyses were conducted including all eligible papers and using different combinations of included-papers but only data from the better quality or more detailed paper in each overlapping study group were included in the final meta-analyses to ensure independence of results. Sensitivity meta-analyses [showing results including partially- overlapping populations, and fully-overlapping populations] are included in Appendix C7. 7.4 Results 7.4.1 Included studies A total of 46 papers met the inclusion criteria and were included in the review [see Figure 7.1 PRISMA flow diagram]. Twenty eight papers were from UHCS countries [tables 7.1-7.2]. Of these, 19 UK papers examined 13 study populations, although as these included national and regional populations from different sources [cancer registries, Hospital Episode Statistics (HES) and lung cancer audit (LUCADA) data], there was some further population overlap. One UK paper also compared treatment in Scotland and Canada (198). A further nine papers from Canada [2], Sweden [1], Australia [1], Italy [1], France [1] and New Zealand [3] were included. The three New Zealand papers all examined the same population. These UHCS studies were published between 2001 and 2012, and examined populations diagnosed between 1986 and 2008. 68 Figure 7.1. Flow diagram of study selection and exclusion. Abbreviations: SEP = socio-economic position, CI = confidence interval PRISMA 2009 Flow Diagram Records identified through database searching (n =1483 ) Screening Included Eligibility Identification Additional records identified through other sources (n = 2 ) Records after duplicates removed (n = 1346 ) Records screened (n = 1346 ) Records excluded (n = 1250), where 1176 excluded on title screening and 74 on abstract screening Full-text articles assessed for eligibility (n = 96 ) Full-text articles excluded (n = 50), where 25 did not examine SEP, 5 did not present lung cancer results separately from other cancers, 17 where outcome was not receipt of treatment, 3 where numerical results for treatment by SEP not reported Studies included in narrative synthesis (n = 46 ) Studies included in quantitative synthesis (meta-analysis) (n = 23) (17 in final meta-analyses + 6 in sensitivity analyses) Full-text articles excluded (n = 23), where 6 calculated a single OR, 10 reported only univariable analysis, 2 did not report CIs, 3 used a different comparator, 1 had issues with quality, 1 stratified outcome by another variable 69 Table 7.1. Characteristics of included studies potentially suitable for meta-analysis (universal health care systems). Confounders controlled for: Paper Country of study Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Quality score Berglund et al, 2010(73) Sweden Regional Lung Cancer Register (RLCR) - Sweden, Cause of Death Register and LISA (insurance and demographics) Uppsala/ Orebro region in central Sweden 1996-2004 Education level (a) 3 NR 30+ Yes Yes Yes Yes Performance status, year of diagnosis, smoking status 6 Berglund et al, 2012(157) England Thames Cancer Registry, HES, LUCADA South-east England 2006-2008 IMD 2007 income domain 5 NR 0-80+ Yes Yes Yes Yes Co-morbidity 6 Campbell et al, 2002(86) Scotland Scottish Cancer Registry and hospital case notes Random sample from North/NE Scotland (with hospital record) 1995-1996 Carstairs Index 5 12 months NR Yes Yes Yes Yes Health board, distance to cancer centre, mode of admission 5 Crawford et al, 2009(199) England Northern and Yorkshire Cancer Registry and Information Service (NYCRIS) Northern and Yorkshire region 1994-2002 IMD 2004 (access to services domain removed) 4 6 months NR Yes Yes No Yes Travel time (but overall results not stratified by travel time used here). Histology not included in receipt of any treatment analysis. 4 Erridge et al, 2002(200) Scotland Scottish Cancer Registry and medical records Scotland (with hospital record) 1995 Carstairs Index 5 6 months <60- 80+ Yes Yes Yes Yes Health board (not inc in receipt of radiotherapy), diagnosis by specialist, management by oncologist 6 Erridge et al, 2009(198) Scotland/ Canada Scottish Cancer Registry and medical records; British Columbia Cancer Registry Scotland/ British Columbia 1995 Carstairs Index/ average household income 2 6 months <60- 80+ Yes Yes Yes Yes Travel time, CT scan 4 Gregor et al, 2001(201) Scotland Scottish Cancer Registry and medical records Scotland (with hospital record) 1995 Carstairs Index 5 6 months <60- 80+ Yes Yes Yes Yes Referral to specialist within 6 months of diagnosis 6 70 Table 7.1 (cont). Characteristics of included studies potentially suitable for meta-analysis (universal health care systems). Confounders controlled for: Paper Country of study Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Quality score Jack et al, 2003 (202) England Thames Cancer Registry South-east England 1995-1999 Townsend (median score per health authority) continuous (b) NR <35 -85+ Yes Yes Yes Yes First hospital visited is a radiotherapy centre, basis of diagnosis, incidence. Health authority/hospital used as 2nd level in multi-level model. 4 Jack et al, 2006 (158) England Thames Cancer Registry and medical records South-east London (with hospital record) 1998 IMD 2000 5 6 months <55- 85+ Yes Yes Yes Yes Consultant specialty, basis of diagnosis, (hospital, number of symptoms in some analyses) 6 Jones et al,2008(78) England Northern and Yorkshire Cancer Registry and Information Service (NYCRIS) Northern and Yorkshire region 1994-2002 IMD 2004 (access to services domain removed) continuous (c) NR NR Yes Yes No Yes Travel time to hospital 4 Mahmud et al, 2003(203) Ireland National Cancer Registry of Ireland (NCRI) Republic of Ireland 1994-1998 SAHRU area-based material deprivation index 3 6 months 15-80+ Yes Yes No Yes Health board, year of diagnosis 4/2(d) McMahon et al, 2011(204) England Eastern Cancer Registry and Information Centre (ECRIC) East of England 1995-2006 IMD 2004 (access to services domain removed) 5 NR <60 -80+ Yes Yes No Yes Year of diagnosis 4 Pollock &Vickers, 1998(76) England HES FCEs North/South Thames (admitted to hospital) 1992-1995 Townsend 10 NR <100 Yes Yes No No Hospital, mode of admission 3 Raine et al, 2010(77) England HES FCEs England (admitted to hospital) 1999-2006 IMD 5 NR 50-- 90+ Yes Yes No No Trust, year of admission, mode of admission 3 71 Table 7.1 (cont). Characteristics of included studies potentially suitable for meta-analysis (universal health care systems). Confounders controlled for: Paper Country of study Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Quality score Riaz et al, 2012(161) England NCIN/UKACR cancer registries England 2004-2006 IMD 2004 5 NR 0-- 85+ Yes Yes No No Government Office Region 4 Rich et al, 2011(1)(153) England LUCADA supplied by 157 NHS trusts England 2004-2007 Townsend 5 NR NR Yes Yes Yes Yes Performance status. Adjusted for clustering by NHS trust 5 Rich et al, 2011(2)(152) England LUCADA and HES England 2004-2008 Townsend 5 NR 30-100 Yes Yes Yes Yes Co-morbidity, ethnicity, surgery centre, radiotherapy centre, trial entry. Adjusted for clustering by NHS trust 5 Stevens et al, 2007(205) New Zealand Regional hospital and oncology databases checked against NZ cancer registry Auckland-Northland region patients managed in secondary care 2004 NZ Deprivation Index 2 NR <60-80+ Yes Yes Yes Yes Co-morbidity, private sector care, care discussed at MDM 3 Stevens et al, 2008(206) New Zealand Regional hospital and oncology databases checked against NZ cancer registry Auckland-Northland region patients managed in secondary care 2004 NZ Deprivation Index 10 NR <60- 80+ Yes Yes Yes Yes Co-morbidity, private sector care, ethnicity 5 (a) Socio-economic index (SEI) and household income also measured but individual education level used in analyses as it contained least missing data (b) Odds ratio for 1 unit increase in deprivation score, range unknown (c) Odds ratio for 1 unit increase in deprivation score, range 1-80 (d) Quality score 4 where adjusted OR used and 2 where unadjusted rates used Quality score ranges from 1 (lowest quality) to 6 (highest quality) Abbreviations: HES = Hospital Episode Statistics, HES FCE= Hospital Episode Statistics Finished Consultant Episode, IMD = Index of Multiple Deprivation, LUCADA = Lung Cancer Audit, MDM = multi-disciplinary meeting, NCIN/UKACR = National Cancer Information Network/ UK Association of Cancer Registries, NR = not reported, OR = odds ratio, SEP = socio-economic position, UHCS=universal healthcare system72 Table 7.2. Characteristics of included studies not suitable for meta-analysis (universal health care systems). Confounders controlled for: Paper Country of study Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Reason for exclusion Quality score Battersby et al, 2004(207) England HES and East Anglian Cancer Intelligence Unit 17 PCTs in Norfolk, Suffolk and Cambridgeshire with HES record 1997-2000 IMD (weighted average for PCT) NR NR NR Yes Yes No Yes Incidence Rate correlated against deprivation, by sex 1 Bendzsak et al, 2011(208) Canada Ontario Cancer Registry linked to CIHI hospital data, Insurance data and RPD database Ontario 2003-2004 Neighbourhood income 5 12 months 20-75+ Yes Yes No No Univariable analysis Univariable rate 2 Cartman et al, 2002(209) England Northern and Yorkshire Cancer Registry and Information Service (NYCRIS) Yorkshire region 1986-1994 NR NR NR <65- 75+ Yes Yes No Yes Univariable analysis Univariable rate 1 Hui et al, 2005(210) Australia NSW Central Cancer Registry and hospital records Residents of 2 Area Health Services 1996 SEIFA-IRSD 5 NR <50- 70+ Yes Yes Yes Yes Univariable analysis Univariable rate 2 Madelaine et al, 2002(211) France Manche Dept Cancer Registry Manche 1997-1999 INSEE 4 NR <54 -75+ Yes Yes Yes Yes Urban/rural Unemployed used as low SEP group and SEP group 2 used as baseline 2 Pagano et al, 2010(212) Italy Piedmont Cancer Registry of Turin Turin 2000-2003 Education level 3 12 months <65 -75+ Yes Yes Yes Yes Marital status Different comparator – other not no treatment 2 73 Table 7.2 (cont). Characteristics of included studies not suitable for meta-analysis (universal health care systems). Confounders controlled for: Paper Country of study Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Reason for exclusion Quality score Patel et al, 2007(213) England Thames Cancer Registry South-east England 1994-2003 IMD 5 6 months 0--100 Yes Yes Yes Yes Cancer network, year of diagnosis Adjusted rates with no CIs. Possible errors in numbers. 2 Stevens et al, 2009(214) New Zealand Regional hospital and oncology databases checked against NZ cancer Registry listing Auckland-Northland region patients managed in secondary care 2004 NZ Deprivation Index 10 NR <60- 80+ Yes Yes Yes Yes Univariable analysis Univariable OR. Multivariable SEP results not shown 2 Younis et al, 2008(215) Canada Nova Scotia cancer registry and chart review Nova Scotia 2005 Median household income 2 NR 65- 75+ Yes Yes Yes Yes Co-morbidity, PS, hospital, surgery type, post-op complications, surgeon, med onc, education level, distance to cancer centre, marital status, smoking history Univariable rate. Multivariable OR only for referral by SEP 2 Quality scores range from 1 (lowest quality) to 6 (highest quality) Abbreviations: CI = confidence interval, HES = Hospital Episode Statistics, IMD = Index of Multiple Deprivation, NR = not reported, NSW = New South Wales, OR = odds ratio, PCT = Primary Care Trust, PS=performance status, SEIA-IRSD = Socio-economic Indexes for Areas - Index of Relative Social Disadvantage, SEP = socio-economic position, UHCS=universal healthcare system74 Eighteen papers were from non-UHCSs – all of which were from the USA [tables 7.3-7.4]. The papers were published between 1995 and 2010 and examined populations diagnosed between 1978 and 2005. The majority of non-UHCS papers used sub-groups of the National Cancer Institute’s Surveillance, Epidemiology and End Results [SEER] database population and, again, some population overlap was found. SEER has changed in size over the years. In earlier papers SEER covered 9, 10, and mostly 11 registries and included around 14% of the population but this increased to 16-17 registries and 26% of the population in later papers. Over half [11/18 = 61%] of the non-UHCS papers [7/10 in meta-analysis and 4/8 in narrative analysis] looked at receipt of treatment in a Medicare population aged over 65. Twenty nine papers met the criteria for meta-analysis, 19 from UHCS (73, 76-78, 86, 152, 153, 157, 158, 161, 198-206) and 10 from non-UHCSs (216-225). However, six studies that examined receipt of treatment in more compared to less deprived SEP groups presented the results as a single OR [for a one unit increase or incremental quintile increase in deprivation score] and so could not be included in the meta-analyses. Seventeen studies were included in the final meta-analyses and a further six in the sensitivity meta-analyses. Seventeen papers did not meet the criteria for meta-analysis and were excluded for the following reasons: Five studies did not conduct multivariable analysis and calculated only univariable rates (150, 208-210, 214). One paper calculated a rate of treatment by deprivation, stratified by sex (207). A further 11 papers were also excluded as, although they did include multivariable analysis, they stratified SEP by race (226), calculated but did not report adjusted rates by SEP, but did report univariable rates (71, 149, 227, 228), did not report confidence intervals (213, 229), had other quality problems (230), or used a different comparator (211, 212, 215). In the study including Scottish and Canadian data (198) the numbers in some tables for the Scottish data did not appear to add up correctly. It was decided not to exclude this paper on quality grounds but to only include data from the Canadian cohort in the meta-analysis as this did not contain any apparent errors. 75 Table 7.3. Characteristics of included studies potentially suitable for meta-analysis (non-universal health care systems). Confounders controlled for: Paper Country of study Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Quality score Bradley et al, 2008(216) USA Michigan Cancer Registry and Michigan Medicare and Medicaid data Medicare and Medicare/ Medicaid patients in Michigan 1997-2000 Census tract median household income (high v low) 2 6 months 66-80+ Yes Yes Yes Yes Co-morbidity, insurance type, ethnicity, urban/rural 4 Davidoff et al, 2010(217) USA SEER cancer registry linked to Medicare data Medicare patients from 16 SEER registries 1997-2002 Census tract median household income 4 90 days 66- 85+ Yes Yes Yes Yes Co-morbidity, performance status, ethnicity, marital status, rural/urban, prior Medicaid, tumour grade 5 Earle et al, 2000(218) USA SEER cancer registry linked to Medicare data Medicare patients from 11 SEER registries 1991-1993 Census tract median household income(increase in OR per quintile) 5 4 months 65-104 Yes Yes Yes Yes Co-morbidity, year of diagnosis, ethnicity, rural/urban, teaching hospital, SEER area 5 Esnoala et al, 2008(219) USA South Carolina central cancer Registry linked to inpatient and outpatient surgery files South Carolina 1996-2002 Income, zip code level (poverty/not living in poverty) 2 NR <50-80+ Yes Yes Yes Yes Co-morbidity, year of diagnosis, insurance type, ethnicity, rural/urban, education, marital status, tumour location 4 Greenwald et al, 1998(220) USA SEER cancer registry 3 (Detroit, San Francisco, Seattle) out of 9 SEER registries 1978-1982 Census tract median household income (inc in OR per decile) 10 NR <=75 Yes Yes Yes Yes Performance status, ethnicity 6 Hardy et al, 2009(221) USA SEER cancer registry linked to Medicare data Medicare patients from 17 SEER registries 1991-2002 % individuals below poverty line at census tract level 4 NR 65- 85+ Yes Yes Yes Yes Co-morbidity, year of diagnosis, ethnicity, marital status, SEER area, other treatment 5 Hayman et al, 2007(222) USA SEER cancer registry linked to Medicare data Medicare patients from 11 SEER registries 1991-1996 Census tract median household income 5 4 months/ 2 years 65- 85+ Yes Yes Yes Yes Co-morbidity, year of diagnosis, ethnicity, SEER area, hospitalisation, teaching hospital, distance to nearest RT centre, receipt of chemotherapy 5 76 Table 7.3 (cont). Characteristics of included studies potentially suitable for meta-analysis (non-universal health care systems). Confounders controlled for: Paper Country of study Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Quality score Lathan et al, 2008(223) USA SEER cancer registry linked to Medicare data Medicare patients from 11 SEER registries 1991-1999 Census tract median household income (inc in OR per quintile) 5 NR 65+ Yes Yes Yes Yes Co-morbidity, ethnicity, SEER registry, urban, non-profit hospital, patient volume, % of black patients in hospital 5 Polednak, 2001(224) USA Connecticut Tumor Registry (SEER) and inpatient hospital discharge database (HDD) Connecticut 1992 - 1997 Census tract poverty rate 5 NR <55-80+ Yes Yes Yes No Co-morbidity, ethnicity, marital status 4 Smith et al, 1995(225) USA Virginia Cancer Registry and Medicare claims database Medicare patients from Virginia cancer registry 1985-1989 Census tract: median household income by race and age continuous (a) 6 months 65- 85+ Yes Yes Yes Yes Co-morbidity, ethnicity, county of residence, distance to oncologist 5 (a) Odds ratio for increase per $10,000 income Quality scores range from 1 (lowest quality) to 6 (highest quality) Abbreviations: CI = confidence interval, non –UHCS = non-universal healthcare system, NR = not reported, OR = odds ratio, SEER = National Cancer Institute’s Surveillance, Epidemiology and End Results database, SEP = socio-economic position 77 Table 7.4. Characteristics of included studies not suitable for meta-analysis (non-universal health care systems). Confounders controlled for: Paper Country Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Hist ology Other Reasons for exclusion Quality score Bach et al, 1999(226) USA SEER cancer registry linked to Medicare data Medicare patients from 10 SEER registries 1985-1993 Median income in zip code of residence (lowest quartile compared to highest 3) 2 NR 65- 75+ Yes Yes Yes Yes Co-morbidity, ethnicity, SEER area OR of surgery for black v white, univariable rates of surgery used here 2 Earle et al, 2002(227) USA SEER cancer registry linked to Medicare data Medicare patients from 11 SEER registries 1991-1996 Census tract median household income 5 any time NR Yes Yes Yes Yes Co-morbidity, ethnicity, year of diagnosis, teaching hospital, seen by oncologist, SEER area SEP non sig in multivariable analysis but only univariable rate shown. 2 Lathan et al, 2006(230) USA SEER cancer registry linked to Medicare data Medicare patients from 11 SEER registries 1991-1999 Census tract median household income 5 NR 65+ Yes Yes Yes Yes Co-morbidity, ethnicity, SEER region, teaching hospital, rural/urban Quality problems 2 Ou et al, 2008(149) USA California Cancer Registry (part of SEER) California 1989-2003 Composite measure (7 indicators of education, income and occupation) 5 NR 0--89 Yes Yes Yes Yes Ethnicity, tumour grade, tumour location, histologic grade, marital status SEP not reported in multivariable analysis. Univariable rate shown. 2 Suga et al, 2010(229) USA California Cancer Registry Sacramento region in northern California 1994-2004 Census tract composite variable - income, education, employment, poverty, rent, housing value 5 NR NR Yes Yes Yes Yes Ethnicity, residence (urban/rural) No CIs 2 78 Table 7.4 (cont). Characteristics of included studies not suitable for meta-analysis (non-universal health care systems). Confounders controlled for: Paper Country Data Source (s) Population included Years of Diagnosis Measure of SEP No of SEP groups Treatment given within Age range Age Sex Stage Histology Other Reasons for exclusion Quality score Tammemagi et al, 2004(71) USA Josephine Ford Cancer Center Tumor Registry Detroit (receiving care at Henry Ford Health System) 1995-1998 Census tract median household income continuous (a) NR NR Yes Yes Yes Yes Co-morbidity, ethnicity, marital status, smoking history, alcohol use, drug use SEP not reported in multivariable analysis. Univariable OR shown. 2 Wang et al, 2008(228) USA SEER cancer registry linked to Medicare data Medicare patients 11 SEER registries 1992-2002 % below census tract poverty level 4 4 months 66-85 Yes Yes Yes Yes Co-morbidity, ethnicity, year of diagnosis, grade, SEER region, census tract education, marital status, teaching hospital, radiation SEP not reported in multivariable analysis. OR for consultation but not treatment shown. 1 Yang et al, 2010(150) USA Florida Cancer registry linked to inpatient and outpatient medical records Florida 1998-2002 Census tract poverty level 4 NR <45- 70+ Yes Yes Yes Yes Univariable analysis only Univariable rate 2 (a) Odds ratio for increase per $10,000 income Quality scores range from 1 (lowest quality) to 6 (highest quality) Abbreviations: CI = confidence interval, non-UHC S= non-universal healthcare system, NR = not reported, OR = odds ratio, SEER = National Cancer Institute’s Surveillance, Epidemiology and End Results database, SEP = socio-economic position 79 A funnel plot to assess potential publication bias did not appear to show obvious bias [figure 7.2]. Individual measures of SEP were only available in one study (73) and SEP was otherwise measured throughout at an area-based level using a deprivation index, income, poverty or education level. The number of SEP groupings utilised varied from 2 to 10, and again this was taken into account in the quality scores. Figure 7.2. Funnel plot to assess publication bias. Abbreviations: non-UHCS = non universal health care system, NSCLC = non-small cell lung cancer, UHCS = universal health care system 80 7.4.2 Quality Assessment of included studies Rates of treatment are not directly comparable between health care systems. Non-UHCS studies generally included NSCLC-patients only and used as the denominator only those patients eligible for treatment [e.g. stage I and II patients for surgery analyses]. Many also included details of co-morbidity. The SEER database only includes biopsy-confirmed cases of lung cancer and so systematically excludes those patients considered too old or unwell for biopsy [and therefore treatment]. UHCS papers rarely included co-morbidity, tended to look at rates of surgery for all lung cancers and did not stratify by stage. In terms of quality the non-UHCS studies that carried out multivariable analysis had better control for confounding than UHCS studies, as they tended to stratify by stage and histology. However, half of the non-UHCS papers used a Medicare-only population where Medicare is a type of social insurance to which those aged over 65 are eligible. Non-UHCS studies are therefore less generalizable in population terms than the UHCS studies. 7.4.3 Receipt of treatment 7.4.3.1 Surgery – overall analysis Thirty one papers [29 study populations] included receipt of surgery as an outcome, 18 UHCS papers [15 study populations] and 13 non-UHCS papers [14 study populations] [tables 7.5-7.6]. Of the papers that reported measures of significance [CIs or p-values], 20 out of 27 [74%] [64% of UHCS (9 /14) and 85% of non-UHCS (11/13) unique-population studies] reported that lower SEP was significantly associated with lower likelihood of surgery when comparing the lowest with the highest SEP group, although three of these 20 papers did not find a significant trend across groups. Seven papers found no significant association between SEP and receipt of surgery. Four papers did not report CIs or p values so were not considered here. 81 Table 7.5. Likelihood of receipt of surgery by SEP group (universal health care systems). Study No. receiving surgery Cohort no./ no. eligible Rate Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality score Meta-analysis Further Information Bendzsak et al, 2011[49] 1220 6499 18.77 any 21.1 18.3 19.7 18.8 16.8 0.02 2 N Univariable rate Campbell et al, 2002[35] 85 653 13.02 any 1.00 0.76 (0.28 to 2.09) 0.70 (0.27 to 1.84) 0.88 (0.35 to 2.22) 0.59 (0.23 to 1.53) 0.423 5 Y P for trend Hui et al, 2005[51] NR 526 any 29 28 20 27 20 0.19 2 N Univariable rate Jack et al, 2003[39] NR 32818 any 0.98 (0.95 to 1.01) 0.7759 4 N Jack et al, 2006[40] 42 695 6.04 any 1.00 0.82 (0.33 to 2.07) 0.89 (0.35 to 2.25) 0.16 (0.03 to 0.73) 0.75 (0.27 to 2.09) 0.1326 6 Y Subset of Jack et al (2003) pop, p for trend Jones et al,2008[41] 3552 34923 10.17 any 0.99 (0.99 to 1.00) <0.01 4 N Pollock &Vickers, 1998[44] 2869 38668 7.42 any 1.00 0.83 (0.69 to 1.00) 0.73 (0.61 to 0.88) 0.82 (0.68 to 0.98) 0.58 (0.48 to 0.70) <0.05 3 Y Hospital population, p for trend Raine et al, 2010[45] 8790 36902 23.82 any 1.63 (1.49 to 1.77) 1.58 (1.46 to 1.72) 1.45 (1.35 to 1.57) 1.34 (1.25 to 1.45) 1.00 <0.001 3 Y Elective admission population Raine et al, 2010[45] 8923 186741 4.78 any 5.5 5.2 4.8 4.4 4.5 NR 2 N All admissions, univariable rate Battersby et al, 2004[48] 387 4092 9.46 NSCLC -0.10 (-0.55 to 0.40) NR 1 N Rate by sex correlated with deprivation score (men), with overall treatment rate Battersby et al, 2004[48] NSCLC -0.16 (-0.59 to 0.35) NR 1 N Rate by sex correlated with deprivation score (women) Berglund et al, 2010[19] 626 3369 18.58 NSCLC 1.93 (1.25 to 3.00) 1.33 (0.98 to 1.81) 1.00 NR 6 Y Berglund et al, 2010[19] 534 932 57.30 NSCLC 2.84 (1.40 to 5.79) 1.53 (1.01 to 2.31) 1.00 NR 6 Y(S) Early stage only - stage IA-IIB Berglund et al, 2012[22] 899 1826 49.18 NSCLC 1.00 0.74 (0.51 to 1.06) 0.71 (0.49 to 1.02) 0.73 (0.52 to 1.03) 0.67 (0.48 to 0.95) 0.29 6 Y Early stage only – stage IA-IIB, p for trend Cartman et al, 2002[50] 2401 12570 19.10 NSCLC 19.1 18.6 NR 1 N Univariable rate 82 Table 7.5 (cont). Likelihood of receipt of surgery by SEP group (universal health care systems). Study No receiving surgery Cohort no / no eligible Rate Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality score Meta-analysis Further Information Crawford et al, 2009[36] 3335 18324 18.20 NSCLC 1.00 0.90 (0.81 to 1.00) 0.82 (0.74 to 0.91) 0.80 (0.72 to 0.89) <0.05, <0.01, <0.01 4 Y Individual P values reported Mahmud et al, 2003[42] 866 4451 19.46 NSCLC 19.8 18.0 21.0 NR 2 N Univariable rate McMahon et al, 2011[43] 2374 18813 12.62 NSCLC 1.00 0.95 (0.83 to 1.09) 0.95 (0.83 to 1.08) 0.90 (0.79 to 1.03) 0.78 (0.65 to 0.94) 0.018 4 Y P for trend McMahon et al, 2011[43] 0.96 (0.93 to 0.99) 0.018 N Paper presents results in 2 different ways Riaz et al, 2012[34] 6900 77349 8.92 NSCLC 1.00 0.88 (0.80 to 0.96) 0.91 (0.83 to 0.99) 0.82 (0.76 to 0.89) 0.76 (0.70 to 0.83) <0.01 4 Y(S) P for trend Rich et al, 2011(1)[46] 3427 24175 14.18 NSCLC 1.00 1.13 (0.98 to 1.32) 1.18 (1.02 to 1.37) 1.01 (0.87 to 1.16) 1.11 (0.96 to 1.27) 0.77 5 Y(S) Subset of Rich et al 2011 (2) pop, p for trend Rich et al, 2011(2)[21] 4481 34436 13.01 NSCLC 1.00 0.99 (0.88 to 1.11) 1.04 (0.92 to 1.19) 0.98 (0.84 to 1.13) 0.86 (0.71 to 1.04) 0.132 5 Y(S) P for trend Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position , UHCS=universal healthcare system 83 Table 7.6. Likelihood of receipt of surgery by SEP group (non- universal health care systems). Study No. receiving surgery Cohort no/ no eligible Rate Stage(s) included Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality Score Meta-analysis Further Information Bradley et al, 2008[57] 1336 2626 50.88 I,II,IIIa NSCLC 1.00 0.80 (0.67 to 0.98) <0.05 4 Y Esnoala et al, 2008[60] NR 2791 local NSCLC 1.00 0.67 (0.51 to 0.88) 0.005 4 Y Greenwald et al, 1998[61] 3053 5157 59.20 I NSCLC 1.076 <0.0001 6 N SE=0.011 (no CIs shown) Hardy et al, 2009[62] 11834 19658 60.20 I,II NSCLC 1.00 0.92 (0.84 to 1.14) 0.78 (0.75 to 1.03) 0.68 (0.60 to 0.77) >0.05, >0.05, <0.05 5 Y Individual p values reported corrected OR supplieda Lathan et al, 2008[64] 4563 9688 47.10 I,II,III NSCLC 1.06 (1.02 to 1.11) NR 5 N Subset of Lathan et al (2006) pop Ou et al, 2008[70] 16185 19700 82.16 I NSCLC 86.9 84.8 81.1 79.6 74.5 <0.001 2 N Smith et al, 1995[66] 801 2813 28.47 local NSCLC 1.04 (0.90 to 1.19) >0.001 5 N Tammemagi et al, 2004[72] NR 1155 I,II NSCLC 1.19 (1.03 to 1.30) 0.02 2 N Univariable OR Bach et al, 1999[67] 550 860 63.95 I,II NSCLC 67.5 61.9 NR 2 N Surgery (blacks) Bach et al, 1999[67] 7763 10124 76.68 I,II NSCLC 78.0 70.7 NR 2 N Surgery (whites) Polednak, 2001[65] 1385 1564 88.55 I,II NSCLC 1.00 1.27 (0.74 to 2.18) 1.15 (0.65 to 2.03) 1.17 (0.67 to 2.04) 1.78 (1.05 to 3.01) >0.05, >0.05, >0.05, <0.05 4 Y Odds of not receiving surgery, individual p values reported Smith et al, 1995[66] 57 2396 2.38 distant NSCLC 1.27 ( 0.97 to 1.67) >0.001 5 N Suga et al, 2010[71] NR 12395 NSCLC 1.17 <0.001 2 N Surgery after invasive staging, no CIs 84 Table 7.6 (cont). Likelihood of receipt of surgery by SEP group (non- universal health care systems). Study No. receiving surgery Cohort no/ no eligible Rate Stage(s) included Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality Score Meta-analysis Further Information Suga et al, 2010[71] NR 12395 NSCLC 1.18 <0.001 2 N Surgery after non-invasive staging, no CIs Lathan et al, 2006[69] NR 14224 NSCLC 1.05 (1.02 to 1.08) NR 2 Yang et al, 2010[74] NR NR all all 24.6 22.2 20.7 18.3 <0.01 2 Univariable analysis Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis a We are grateful to the authors for supplying a corrected OR to allow inclusion of this study in the meta-analysis Abbreviations: CI = confidence interval, non –UHCS = non-universal healthcare system, NR = not reported, OR = odds ratio, pop = population, SE=standard error, SEP = socio-economic position85 Meta-analysis of all 16 populations that were suitable for inclusion showed a significant negative effect of lower SEP on the likelihood of receiving surgery [OR=0.72 (95% CI 0.65 to 0.80) I2=80%, p<0.001] [Appendix C, Fig C7.1]. Including only non-overlapping study populations [n=12] gave a similar result: [OR=0.68 (95% CI 0.63 to 0.75) I2=53%, p<0.001] [Fig 7.3]. Similar results were also seen for the subgroup of 8 papers including NSCLC patients only: [OR=0.73 (95% CI 0.68 to 0.80) I2=24%, p<0.001] [Appendix C, Fig C7.2] and with further stratification by health care system; NSCLC [UHCS]: [OR=0.75 (95% CI 0.66 to 0.85), I2=29%, p<0.001]; NSCLC [non-UHCS, early stage only, co-morbidity included]: [OR=0.71 (95% CI 0.64 to 0.78) I2=2%, p<0.001]; [Fig 7.3]. Lower SEP was associated with a lower likelihood of receiving lung cancer surgery, in both types of health care system, and in studies where histology and stage at diagnosis were taken into account. Figure 7.3. Meta-analysis of odds of receipt of surgery in low (most deprived) versus high (least deprived) SEP. Abbreviations: CI = confidence interval, non-UHCS = non universal health care system, NSCLC = non-small cell lung cancer, OR = odds ratio, SE=standard error, SEP = socio-economic position, UHCS = universal healthcare system 86 7.4.3.1.1 Surgery - detailed analysis by healthcare system: UHCS Eight UHCS papers looked at receipt of surgery in any-histology populations [six registry studies (including two with overlapping populations) and two HES studies, so seven unique populations]. Rates of surgery varied from 4.8 to 18.8% [although a higher rate of 23.8% was seen for electively-admitted patients]. A significant association was seen between receipt of surgery and higher SEP in the HES studies [although neither of the two studies included stage or histology] but was seen in only one out of three non-overlapping multivariable-analysis registry studies [in the only study that did not include stage] and in one out of two univariable registry studies, although all studies showed trends in a reduction in OR as SEP decreased. Three of the studies where no association was found were very small [all less than 700 participants]. All of these studies included patients with SCLC who are rarely eligible for surgery. Ten UHCS papers looked at receipt of surgery in NSCLC-only populations [seven registry studies, one HES and two audit studies]. Rates of surgery varied from 9.5 to 19.5% [although rates of 49.2% and 57.3% were seen for early stage patients]. All five multivariable-analysis registry studies found an association between lower SEP and reduced likelihood of receipt of surgery when comparing the lowest SEP group with the highest. Two studies looked at surgery for early stage patients only and both found higher odds of receiving surgery for patients in the highest SEP compared to the lowest but one did not find a significant trend across groups. The other two univariable studies calculated rates not odds ratios and did not present p values. One study produced a correlated rate by sex. The two LUCADA papers using an overlapping study population did not find an association between SEP and receipt of surgery. Similar meta-analysis results to those seen overall were found with stratification by health care system. Eight non-overlapping UHCS studies were suitable for inclusion, with an OR=0.67, 95% CI 0.54 to 0.77, I2=65%, p<0.001, [with an OR=0.72, 95% CI 0.63 to 0.83, I2=85%, p<0.001 when 12 over-lapping studies included], but again heterogeneity was high as these included studies that contained eligible NSCLC patients and some that also contained ineligible SCLC patients. Not all studies 87 accounted for stage. Including non-overlapping NSCLC-only UHCS studies [n=4] reduced the heterogeneity: [OR=0.75, 95% CI 0.66 to 0.85, I2=29%, p<0.001] [fig 7.3]. In a subgroup analysis of the five partially-overlapping UHCS studies including stage [three studies that looked at NSCLC and two studies that looked at any-histology in registry and audit populations, all high quality studies, score 5 or 6] the association between lower SEP and reduced likelihood of surgery remained [OR=0.72, 95% CI 0.58 to 0.89, I2=23%, p=0.002]. This was also found using any combination of non-overlapping studies. In the four studies that did not conduct histology-specific analyses [two registry and two HES studies, where one HES study looked at receipt of surgery in electively admitted patients only] the summary statistic was OR=0.61, 95% CI 0.56 to 0.66, I2=0%, p<0.001 [fig 7.3]. The two HES studies included only age and sex as confounders and so were of lower quality but accounted for over 98% of the weight in this meta-analysis. 7.4.3.1.2 Surgery - detailed analysis by healthcare system: non-UHCS Ten US papers [11 populations] looked at receipt of surgery for early stage NSCLC. A significant association was seen between receipt of surgery and higher SEP in seven out of eight papers. However, in two of these studies the pattern across socio-economic groupings was not significant, only the OR in the lowest compared to the highest group was significant. A further 2 papers did not report CIs or p values. Rates of surgery varied from 28.5% to 88.6%. Four studies compared the odds of surgery in the lowest compared to the highest SEP group. All four studies used different study populations and included co-morbidity as a confounder. The meta-analysis summary statistic was OR=0.71, 95% CI 0.64 to 0.78, I2=2%, p<0.001; [Fig 7.3]. 88 Table 7.7. Likelihood of receipt of chemotherapy by SEP group (universal health care systems). Study No receiving chemo Cohort no/ no eligible Rate Hist ology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality Score Meta-analysis Further Information Berglund et al, 2012[22] 3661 10039 36.47 any 1.00 0.90 (0.77 to 1.06) 0.78 (0.67 to 0.91) 0.77 (0.66 to 0.89) 0.75 (0.65 to 0.87) <0.01 6 Y NSCLC stage IIIA-IV & all stage SCLC, p for trend Campbell et al, 2002[35] 124 653 18.99 any 1.00 0.58 (0.21 to 1.57) 0.72 (0.29 to 1.78) 0.41 (0.16 to 1.05) 0.39 (0.16 to 0.96) 0.028 5 Y Jack et al, 2003[39] NR 32818 any 0.96 (0.94 to 0.98) 0.0001 4 N Subset of Patel et al (2007) pop Jack et al, 2006[40] 108 695 15.54 any 1.00 1.04 (0.50 to 2.16) 0.81 (0.38 to 1.70) 0.89 (0.43 to 1.85) 1.04 (0.48 to 2.25) 0.9130 6 Y Subset of Patel et al (2007) pop, p for trend Jones et al,2008[41] 5783 34923 16.56 any 0.99 (0.99 to 0.99) <0.01 4 N Patel et al, 2007[54] 11217 67312 16.66 any 18.3 15.7 14.5 12.8 12.8 <0.001 2 N Adjusted rates, no CIs Rich et al, 2011(1)[46] 14168 59592 23.78 any 1.00 0.97 (0.90 to 1.04) 0.89 (0.83 to 0.96) 0.83 (0.77 to 0.89) 0.85 (0.79 to 0.91) <0.01 5 Y(S) Hui et al, 2005[51] NR 526 any 31 34 36 27 26 0.15 2 N Univariable rate Berglund et al, 2010[19] 1285 3369 38.14 NSCLC 1.35 (1.00 to 1.81) 1.25 (1.03 to 1.52) 1.00 NR 6 Y Pagano et al, 2010[53] 430 1231 34.93 NSCLC 1.00 0.98 (0.64 to 1.50) 1.63 (1.08 to 2.44) NR 2 N Odds of receiving chemo +/or radio rather than surgery Younis et al, 2008[56] 29 108 26.85 NSCLC 4.7 (1.3 to 17.8) 1.0 0.015 2 N Odds of referral for adjuvant chemo after surgery, stage I,II,III Cartman et al, 2002[50] 1349 2448 55.11 SCLC 52.1 56.8 NR 1 N Univariable rate Crawford et al, 2009[36] 3619 5510 65.68 SCLC 1.00 1.10 (0.94 to 1.30) 0.91(0.78 to 1.08) 0.94 (0.80 to 1.11) >0.05 4 Y Individual p values, all reported as >0.05 Mahmud et al, 2003[42] 425 1002 42.42 SCLC 37.8 40.5 50.2 NR 2 N Univariable rate Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position , UHCS=universal healthcare system 89 Table 7.8. Likelihood of receipt of chemotherapy by SEP group (non- universal health care systems). Study No receiving chemo Cohort no/ no eligible Rate Stage Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality Score Meta-analysis Further Information Bradley et al, 2008[57] 643 2348 27.39 I,II,IIIa NSCLC 1.00 1.09 (0.87 to 1.37) >0.05 4 Y Hardy et al, 2009[62] 2951 19658 15.01 I,II NSCLC 1.00 0.91 (0.81 to 1.02) 0.96 (0.85 to 1.09) 0.85 (0.74 to 0.98) >0.05, >0.05, <0.05 5 Y Individual p values reported Ou et al, 2008[70] 1175 19700 5.96 I NSCLC 5.3 5.7 5.3 6.9 7.4 0.001 2 N Univariable analysis Davidoff et al, 2010[58] 5499 21285 25.84 IIIB, IV NSCLC 1.43 (1.28 to 1.60) 1.17 (1.05 to 1.30) 1.11 (1.00 to 1.22) 1.00 <0.01, <0.01, <0.05 5 Y Individual p values reported Earle et al, 2000[59] 1356 6308 21.50 IV NSCLC 1.07 (1.02 to 1.12) 0.0077 5 N Subset of Earle (2002) Earle et al, 2002[68] 8813 12015 73.35 IV NSCLC 41 41 36 31 27 >0.05 2 N Univariable analysis only. SEP was included in multivariable analysis but non-sig (figs not reported) Hardy et al, 2009[62] 26417 51243 51.55 III,IV NSCLC 1.00 0.87 (0.78 to 0.96) 0.76 (0.63 to 0.90) 0.60 (0.45 to 0.79) <0.05, <0.05, <0.05 5 Y(S) Individual p values reported Tammemagi et al, 2004[72] NR 1155 III,IV NSCLC 1.09 (1.01 to 1.18) 0.03 2 N Univariable OR Davidoff et al, 2010[58] 749 1946 38.49 IIIB, IV NSCLC 0.86(0.69 to 1.08) 0.96 (0.77 to 1.19) 0.99 (0.81 to 1.22) 1.00 NR 5 N Odds of single agent compared to doublet chemo. Wang et al, 2008[73] 1521 3196 47.59 II,IIIa NSCLC 1.00 1.08 (0.97 to 1.21) 1.08 (0.97 to 1.21) 0.97 (0.85 to 1.10) NR 1 N Odds of receiving oncology consultation. Yang et al, 2010[74] NR NR All any 32.2 30.7 29.9 30.1 <0.01 2 N Univariable analysis Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, non –UHCS = non-universal healthcare system , NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position 90 7.4.3.2 Chemotherapy – overall analysis Twenty three papers included chemotherapy as an outcome – 14 UHCS papers [12 populations] and nine non-UHCS papers [10 populations] [tables 7.7-7.8]. Of the 21 papers that reported measures of significance, 15 [71%] reported that lower SEP was significantly associated with lower likelihood of receipt of chemotherapy. Meta-analysis of the ten populations that were suitable for inclusion found a significant negative effect of lower SEP on the likelihood of receiving chemotherapy: [OR=0.81, 95% CI 0.73 to 0.89, I2=68%, p<0.001] [Appendix C, Fig C7.3]. Similarly, in a meta-analysis of the eight papers containing non-overlapping populations that were selected for inclusion, the odds of receiving chemotherapy were significantly lower for those in the most deprived SEP group compared to those in the least deprived [OR=0.82, 95% CI 0.72 to 0.93, I2=67%, p=0.003], overall. A similar pattern was found in UHCS [OR = 0.80, 95% CI 0.68 to 0.95, I2=46%, p=0.01]; and in non-UHCS settings [OR=0.85, 95% CI 0.68 to 1.07, I2=85%, p=0.16], although this did not reach significance [Fig 7.4]. Figure 7.4. Meta-analysis of odds of receipt of chemotherapy in low (most deprived) versus high (least deprived) SEP. Abbreviations: CI = confidence interval, non-UHCS = non universal health care system, OR = odds ratio, SE=standard error, SEP = socio-economic position, UHCS = universal healthcare system 91 7.4.3.2.1 Chemotherapy - detailed analysis by healthcare system: UHCS Eight papers looked at receipt of chemotherapy in any-histology populations but three used overlapping populations. One paper included a NSCLC-only population. Rates of chemotherapy varied from 15.5% to 38.1% with the highest rate found in the paper including only a NSCLC population. Seven out of nine studies found an association between SEP and receipt of chemotherapy. Two small studies did not, one of which (158) was a subset of a population used in two larger studies where an association was seen (202, 231), and one which conducted only univariable analysis (210). Three studies looked at receipt of chemotherapy in SCLC patients. Rates here varied from 42.4-65.7%. Two studies reported only univariable rates and no CIs but both showed a trend for increased rates of chemotherapy in more deprived groups. However, in the one study that conducted multivariable analysis no association between SEP and receipt of chemotherapy was found, although the trend was for reduced likelihood of chemotherapy in deprived populations, in contrast to the univariable studies. In the meta-analysis five non-overlapping studies [three any histology, one NSCLC, one SCLC] reported OR for the lowest compared to the highest SEP group and a significant association between SEP and receipt of chemotherapy was seen [OR = 0.80, 95% CI 0.68 to 0.95, I2=46%, p=0.005]. 7.4.3.2.2 Chemotherapy - detailed analysis by healthcare system: non-UHCS Eight non-UHCS papers looked at receipt of chemotherapy in 10 populations. One further study looked at the odds of receiving an oncology consultation but didn’t report the odds of actually receiving treatment. Six out of 10 found an association between low SEP and lower likelihood of chemotherapy as a trend across SEP groups and one further study found an association only in the lowest SEP group compared to the highest. Breaking this down further, eight studies looked at receipt of chemotherapy for NSCLC, three for early stage and five for late stage NSCLC. Two out of three early stage studies found an association between SEP and receipt of chemotherapy [where one included only univariable analysis and one where an association was seen only in the most 92 deprived group]. Only one out of five late stage studies did not find an association and this was in a univariable analysis (227). The same authors did find SEP to be significant in an earlier paper (218), with a significantly increased OR of receiving chemotherapy of 1.07, 95% CI 1.02 to 1.12, for each incremental quintile increase in SEP. Three papers [four populations] were potentially eligible for meta-analysis, two had data on early stage and two on later stage. Including only the three non-overlapping populations gave a non-significant result [OR=0.85, 95% CI 0.68 to 1.07, I2=85%, p=0.16]. In the two early-stage studies the results were contradictory. One study using Michigan registry data found non-significantly increased odds of receiving chemotherapy for lower-SEP patients (216) but a much larger study using SEER-linked Medicare data (221) found significantly reduced odds. When meta-analysis was conducted no association was found for SEP and receipt of chemotherapy [OR=0.95, 95% CI 0.74 to 1.21, I2=70%, p=0.66]. However, heterogeneity was high and as both studies showed contradictory effects meta-analysis may not be suitable. In the two studies that looked at late stage NSCLC and compared the odds of receiving chemotherapy in high and low SEP, the odds of receiving chemotherapy were significantly lower for those in the most deprived SEP group compared to those in the least deprived [OR=0.68, 95% CI 0.62 to 0.76, , I2=2%, p<0.001]. Both these studies used Medicare-linked SEER data with some overlap in the study populations. 93 Table 7.9. Likelihood of receipt of radiotherapy by SEP group (universal health care systems). Study No receiving radio Cohort no/ no eligible Rate Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality Score Meta-analysis Further Information Berglund et al, 2012[22] 1054 2771 38.04 any 1.00 1.16 (0.88 to 1.54) 1.17 (0.90 to 1.53) 1.18 (0.91 to 1.53) 0.99 (0.77 to 1.29) 0.67 6 Y Stage III only, p for trend Campbell et al, 2002[35] 412 653 63.09 any 1,00 2.08 (1.11 to 3.91) 2.27 (1.24 to 4.16) 1.47 (0.83 to 2.60) 1.86 (1.05 to 3.28) 0.378 5 Y P for trend Jack et al, 2003[39] NR 32818 any 1.00 (0.99 to 1.02) 0.2048 4 N Jack et al, 2006[40] 338 695 48.63 any 1.00 1.24 (0.76 to 2.02) 0.76 (0.46 to 1.26) 0.98 (0.60 to 1.59) 0.68 (0.41 to 1.14) 0.0978 6 Y Subset of Jack et al (2003) pop, p for trend Jones et al,2008[41] 13857 34923 39.68 any 0.99 (0.99 to 1.00) <0.01 4 N Rich et al, 2011(1)[46] 12079 59592 20.27 any 1.00 1.08 (1.01 to 1.16) 1.12 (1.04 to 1.20) 1.12 (1.04 to 1.20) 1.02 (0.95 to 1.09) 0.80 5 Y(S) P for trend Hui et al, 2005[51] NR 526 any 52 62 51 55 55 0.84 2 N Univariable rate Stevens et al, 2009[55] 222 555 40.00 any 1.0 0.8 (0.4 to 1.5) 0.6 (0.3 to 1.2) 0.9 (0.5 to 1.6) 0.7 (0.4 to 1.3) >0.05 2 N Hosp pop, univariable OR Berglund et al, 2010[19] 863 3369 25.62 NSCLC 0.91 (0.67 to 1.22) 1.12 (0.93 to 1.36) 1.00 NR 6 Y Erridge et al, 2002[37] 824 3177 25.94 NSCLC/ unknown 1.00 0.94 (0.70 to 1.26) 1.04 (0.79 to 1.38) 1.33 (1.01 to 1.75) 1.13 (0.84 to 1.51) 0.10 6 Y Mahmud et al, 2003[42] 1265 4451 28.42 NSCLC 26.1 29.0 29.9 NR 2 N Univariable rate Cartman et al, 2002[50] 693 2448 28.31 SCLC 37.1 39.5 NR 1 N Univariable rate Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position , UHCS=universal healthcare system 94 Table 7.10. Likelihood of receipt of radiotherapy by SEP group (non- universal health care systems). Study No receiving radio Cohort no/ no eligible Rate Stage Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality Score Meta-analysis Further information Bradley et al, 2008[57] 950 2348 40.46 I,II,IIIa NSCLC 1.00 0.97 (0.79 to 1.19) >0.05 4 Y Ou et al, 2008[70] 2779 19700 14.11 I NSCLC 11.7 12.6 14.7 16.5 16.6 <0.001 2 N Univariable analysis Smith et al, 1995[66] 1323 2813 47.03 local NSCLC 0.95 (0.83 to 1.09) >0.001 5 N Hardy et al, 2009[62] 43519 51243 84.93 III,IV NSCLC 1.00 1.01 (0.96 to 1.07) 0.93 (0.88 to 0.99) 0.88 (0.82 to 0.93) 0.05, <0.05, <0.05 5 Y Individual p values reported Hayman et al, 2007[63] 6436 11084 58.07 IV NSCLC 1.48 (1.17 to 1.87) 1.50 (1.17 to 1.91) 1.32 (1.01 to 1.72) 1.25 (0.93 to 1.69) 1.00 <0.001 5 Y(S) Smith et al, 1995[66] 1438 2396 60.02 distant NSCLC 1.00 (0.90 to 1.12) >0.001 5 N Yang et al, 2010[74] NR NR ?? any 32.0 32.1 31.4 33.1 0.02 2 N Univariable analysis Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, non –UHCS = non-universal healthcare system , NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position 95 7.4.3.3 Radiotherapy – overall analysis Eighteen papers [18 populations] examined receipt of radiotherapy for lung cancer - 12 in UHCS settings [11 populations] and six in non-UHCS settings [seven populations - six NSCLC-only populations and one study including any-histology] [tables 7.9-7.10]. Only one UHCS study found an association between SEP and receipt of radiotherapy. The non-UHCS studies had very heterogeneous outcomes. Overall, no association between SEP and receipt of radiotherapy was seen in the meta-analysis of the seven studies with non-overlapping populations selected for inclusion [OR = 0.99, 95% CI 0.86 to 1.14, I2=54%, p=0.89] [Fig 7.5], or when all nine studies were included [OR = 0.95, 95% CI 0.85 to 1.06, I2=71%, p=0.40] [Appendix C, Fig C7.4]. A significant association was seen for non-UHCS studies but only two studies were included here, each looking at different stage patients. Figure 7.5. Meta-analysis of odds of receipt of radiotherapy in low (most deprived) versus high (least deprived) SEP. Abbreviations: CI = confidence interval, non-UHCS = non universal health care system, OR = odds ratio, SE=standard error, SEP = socio-economic position, UHCS = universal healthcare system 96 7.4.3.3.1 Radiotherapy - detailed analysis by healthcare system: UHCS Eight papers using six registry populations [two with overlapping populations], one audit and one hospital-based population looked at receipt of radiotherapy for lung cancer in any-histology patients. A further three registry papers looked at NSCLC only and one at SCLC. Only one of these study populations reported an association between low SEP and reduced likelihood of receipt of radiotherapy and this study used SEP as a continuous variable, although the OR was very close to 1.00 [0.99] and the CI actually reached 1.00. Five non-overlapping studies looked at the odds of radiotherapy in the lowest compared to the highest SEP quintile and the meta- analysis found no association [OR = 1.07, 95% CI 0.87 to 1.32]. 7.4.3.3.2 Radiotherapy - detailed analysis by healthcare system: non-UHCS Five non-UHCS NSCLC papers [including six populations] looked at receipt of radiotherapy, three for early stage and three for late stage NSCLC. One study examined an all-histology population. Two out of three late-stage studies found an association between low SEP and reduced likelihood of radiotherapy but one out of three early-stage [univariable analysis] study and one any-histology [also univariable analysis] study found an association between low SEP and increased likelihood of radiotherapy. Hence these are contradictory results. However, the univariable studies are of a low quality and look at rates of treatment rather than ORs. 7.4.3.4 Surgery, chemotherapy and radiotherapy summary meta-analysis When the surgery, chemotherapy and radiotherapy papers included in the separate treatment meta-analyses in this systematic review were analysed together to produce an overall summary effect meta-analysis OR, low SEP was associated with a lower likelihood of receiving any type of treatment. This was found when including only studies with non-overlapping populations [OR = 0.79, 95% CI 0.73 to 0.86, I2=77%, p<0.001] [Fig 7.6] and when including all eligible studies [OR = 0.80, 95% CI 0.75 to 0.86, I2=82%, p<0.001] [Appendix C, Fig C7.5]. 97 Figure 7.6. Meta-analysis of odds of receipt of any type of treatment in low (most deprived) versus high (least deprived) SEP (non- overlapping populations; n=31). 98 Table 7.11. Likelihood of receipt of any type of unspecified treatment by SEP group (universal health care systems). Study No. receiving treatment Cohort no/ no eligible Rate Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality score Meta-analysis Further Information Crawford et al, 2009[36] 19667 34923 56.32 any 1.00 0.91 (0.86 to 0.97) 0.82(0.77 to 0.88) 0.79 (0.74 to 0.84) <0.01 4 Y Individual P values, all reported as <0.01 Erridge et al, 2009[18] 2186 3833 57.03 any 1.3 (1.1 to 1.5) 1.00 <0.05 4 Y(S) Scottish pop Erridge et al, 2009[18] 1372 2073 66.18 any 1.3 (1.1 to 1.7) 1.00 <0.05 4 Y(S) Canadian pop Jack et al, 2003[39] NR 32818 any 0.98 (0.96 to 0.99) 0.0091 4 N Jack et al, 2006[40] 414 695 59.57 any 1.00 0.91 (0.53 to 1.55) 0.69 (0.40 to 1.19) 0.57 (0.34 to 0.97) 0.65 (0.37 to 1.13) 0.03 6 Y Subset of Jack et al (2003) pop, p for trend Stevens et al, 2007[23] 285 565 50.44 any 1.0 0.9 (0.6 to 1.5) 0.773 3 Y(S) Hospital pop Mahmud et al, 2003[42] 2678 4451 60.17 NSCLC 1.0 0.9 (0.8 to 1.1) 1.0 (0.8 to 1.2) 0.39, 0.958 4 Y(S) Odds of NOT receiving treatment – individual p values reported Mahmud et al, 2003[42] 694 1002 69.26 SCLC 1.0 1.0 (0.6 to 1.5) 0.8 (0.5 to 1.3) 0.888, 0.358 4 Y(S) Odds of NOT receiving treatment – individual p values reported Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position, UHCS=universal healthcare system 99 Table 7.12. Likelihood of receipt of any type of unspecified treatment by SEP group (non- universal health care systems). Study No. receiving treatment Cohort no/ no eligible Rate Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality score Meta-analysis Further Information Ou et al, 2008[70] 18216 19700 92.47 NSCLC 94.7 94.1 92.2 91.9 87.2 <0.001 2 N Stage I. Univariable analysis Smith et al, 1995[66] 1697 2396 70.83 NSCLC 1.00 (0.91 to 1.11) >0.001 5 N Distant stage Smith et al, 1995[66] 2343 2813 83.29 NSCLC 1.00 (0.88 to 1.13) >0.001 5 N Local stage Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, non –UHCS = non-universal healthcare system , NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position 100 Table 7.13. Likelihood of receipt of any type of unspecified curative treatment by SEP group (universal health care systems). Study No. receiving treatment Cohort no/ no eligible Rate/ eligible rate Histology OR/rate in Q1 (95% CI) OR/rate in Q2 (95% CI) OR/rate in Q3 (95% CI) OR/rate in Q4 (95% CI) OR/rate in Q5 (95% CI) P value Quality score Meta-analysis Further Information Erridge et al, 2009[18] 548 3833 14.30 any 1.1(0.9 to 1.4) 1.00 >0.05 4 Y (S) Scottish pop – subset of Gregor et al (2001) pop Erridge et al, 2009[18] 546 2073 26.34 any 1.4(1.1 to 1.8) 1.00 <0.05 4 Y Canadian pop Gregor et al, 2001[38] 627 3855/ 1423 16.26/ 44.06 any 1.00 1.14 (0.72 to 1.80) 1.07 (0.69 to 1.66) 0.95 (0.62 to 1.47) 0.77 (0.51 to 1.16) 0.25 6 Y Eligible = early stage Stevens et al, 2008[47] 109 565 19.29 any 1.0 3.1 (1.0 to 9.7) 1.4 (0.4 to 4.4) 1.1 (0.4 to 0.3) 0.6 (0.2 to 1.8) 0.05, 0.60, 0.86, 0.40 5 Y Hospital pop - subset of Stevens et al (2007) pop, individual P values reported Q1= high socio-economic position, Q5=low socio-economic position Some studies reported SEP quintiles but others reported SEP in 2, 3 or 4 categories or as a continuous variable. Details of the number of SEP groups per study are given in Tables 1-4 in the column entitled 'No. of SEP groups’ Quality scores range from 1 (lowest quality) to 6 (highest quality) Meta-analysis: Y=included in final meta-analysis, Y(S) = included in sensitivity meta-analysis, N=not included in meta-analysis Abbreviations: CI = confidence interval, NR = not reported, OR = odds ratio, pop = population, SEP = socio-economic position, UHCS=universal healthcare system 101 7.4.3.5 Treatment type not specified – overall analysis Seven papers [eight study populations] examined receipt of unspecified treatment and three papers considered receipt of unspecified curative treatment in three populations [tables 7.11-7.13]. In the meta-analysis of five non-overlapping studies [all UHCS studies], low SEP was associated with a lower likelihood of receiving unspecified treatment [OR = 0.78, 95% CI 0.74 to 0.83, I2=0, p<0.001] [Fig 7.7]. This was also seen when studies with overlapping populations were included [OR = 0.80, 95% CI 0.77 to 0.84, I2=17%, p<0.001]. Figure 7.7. Meta-analysis of odds of receipt of unspecified treatment in low (most deprived) versus high (least deprived) SEP. Abbreviations: CI = confidence interval, OR = odds ratio, SE=standard error, SEP = socio-economic position 7.4.3.5.1 Treatment type not specified - detailed analysis by healthcare system: UHCS Five UHCS papers looked at receipt of any unspecified treatment [table 7.11] in six populations [although two contained overlapping populations]. All studies included any-histological-type and rates of treatment ranged from 50.4% to 66.2%. A significant association was seen between receipt of any treatment and SEP in four out of five non-overlapping populations, where those with lower SEP were less likely to receive any treatment. This association was not seen in one study that included a small hospital population. Three UHCS papers also looked at receipt of any type of curative treatment [table 7.13] in a subset of the any-type-of-treatment populations. One study included a Scottish and Canadian population (198) and the Scottish population overlapped with that included in a second, better quality, study (201) hence three unique study populations 102 were examined. In only one out of three non-overlapping study populations was lower SEP associated with lower likelihood of any curative treatment. Combining the any type of treatment and any type of curative treatment studies gave five non-overlapping populations suitable for meta-analysis. Low SEP was associated with a lower likelihood of receiving unspecified treatment [OR = 0.78, 95% CI 0.74 to 0.83, I2=0, p<0.001] [Fig 7.7]. 7.4.3.5.2 Treatment type not specified - detailed analysis by healthcare system: non-UHCS Two studies looked at receipt of unspecified treatment for NSCLC in three populations using registry data from California and Virginia [table 7.12]. Rates of treatment ranged from 83.3% to 92.5% for early stage patients and 70.8% for late stage. A significant association was seen between receipt of unspecified treatment and SEP in one of the two studies, where those with lower SEP were less likely to receive any treatment. However, this was in an unadjusted analysis. In the study conducting multi-variable analysis no significant association was seen in either the early or late-stage population. 7.5 Discussion 7.5.1 Principal findings To my knowledge, this is the first systematic review and meta-analysis examining socio-economic inequalities in receipt of lung cancer treatment. It showed an association between low SEP and reduced likelihood of receipt of any type of treatment, surgery and chemotherapy but not radiotherapy. The results were generally consistent across different health care systems. 7.5.2 Interpretation of results Surgery is only suitable for early stage NSCLC patients and it has been suggested that cancer patients with lower SEP are more likely to present later and with later stage disease (108). This may help explain why socio-economic inequalities in receipt of surgery are observed in some studies. Therefore if stage is not included as a confounder then the absence of controlling for stage may account for some of the reduced likelihood of receipt of surgery in lower SEP groups. However, presentation 103 with later stage cancer in lower SEP patients has not been consistently observed (73). In this review, when receipt of treatment was examined in studies of early-stage patients only [from non-UHCS studies] and in UHCS studies where stage was taken into account, low SEP remained associated with reduced likelihood of surgery. Thus, the association between SEP and receipt of surgery appears to be independent of stage. Similar results were seen for NSCLC studies in both health care systems. If SCLC patients are included within the surgery denominator and histology is not controlled for, as greater levels of SCLC are seen in lower SEP groups due to smoking (17), this may be a valid reason for lower rates of surgery in this group. Receipt of treatment may also be influenced by clinical suitability for treatment, and socio-economic differences in the number of co-morbidities present may explain socio-economic inequalities in treatment. In the three UHCS studies that took co-morbidity into account, SEP was not associated with receipt of surgery (152, 157) or of any treatment (205) when the trend across SEP groups was examined, suggesting that co-morbidity may be a potential mediator of socio-economic inequalities in treatment in UHCSs. However, most of the non-UHCS studies did include co-morbidity as a confounder and socio-economic inequalities in treatment were still observed, suggesting that there may be potential differences between healthcare systems in relation to the mediating effect of co-morbidity. 7.5.3 Strengths and weaknesses of the review and of the available evidence This is one of the first equity reviews conducted (51, 61), the first systematic review of the literature on intervention-generated inequalities in lung cancer treatment, as well as the first cancer equity review to conduct meta-analysis. Extensive searches were carried out to identify studies. However, it is possible that not all relevant studies were obtained. The included studies reported observational data only. The suitability of meta-analysis for observational studies has been questioned, as it may produce precise but spurious results (232). Examining the possible sources of heterogeneity by conducting sensitivity analyses across different sub-groups may be less prone to bias than calculating an overall summary effect (232). Here, although an overall summary effect OR was calculated, heterogeneity was taken into account. Separate analyses by type of 104 treatment were carried out, with further stratification by stage and histology. Universal and non-UHCSs were examined separately and random effects rather than fixed effects meta-analyses were conducted. These precautions did not change the overall pattern of results seen. Significant heterogeneity remained in some cases, which could be considered as a limitation, although this is not surprising due to the characteristics of the studies included. For studies examining receipt of chemotherapy and radiotherapy it was generally not possible to differentiate between curative and palliative treatment and, if patterns of care differ for these by SEP, this might explain the high degree of heterogeneity seen. However, although there is some suggestion that heterogeneity can be considered high at >50% (197), when confidence intervals were calculated around the I2 statistic these were wide, and so it was difficult to be confident about the degree of heterogeneity present (233). Results for receipt of radiotherapy differed in the non-UHCS sub-group compared to overall but, as only two studies were included in this sub-group, it is difficult to be sure whether different patterns of receipt of radiotherapy by SEP are due to differences in healthcare system. Many of the non-UHCS studies used over-lapping population sub-groups from the SEER database. There was also population overlap between some UHCS datasets. We attempted to include only substantially non-overlapping populations within the final meta-analyses to ensure independence of results. A judgement had to be made as to which was the best quality and most appropriate paper to include, but sensitivity analyses using different inclusion combinations [Appendix C, Fig C7.6] did not change the overall findings and nor did including all suitable studies regardless of population overlap [Appendix C, Figs C7.1, C7.3-C7.5]. Included papers contained data for patients diagnosed between 1978 and 2008. As treatment guidance has changed over time, older studies may be less applicable to current clinical practice. However, the majority of included studies were published within the last five years and sensitivity analyses excluding studies published prior to 2000 did not change the overall findings. 105 Various measures of SEP were utilised and these were categorised differently, which is an acknowledged problem in equity reviews (196). All but one study measured SEP at area-level. This is a further limitation, as area-based measures of SEP are unlikely to be accurate markers of individual-level circumstances and access to resources (234). Area-based measures of SEP can be calculated using address, making them easy to add to disease registers, such as those used in many of the studies synthesised here. However, the reliance on area-based markers of SEP may under-estimate the strength of the true association between SEP and receipt of treatment. Not all studies reported details of stage and histology, both of which influence treatment type, and very few UHCS studies took co-morbidity into account. Thus, the ORs used in the meta-analyses were not consistently adjusted for the same covariates. However, I attempted to take these factors into account in the quality scores and by conducting subgroup sensitivity analyses. Examining only high quality studies did not alter findings nor did sensitivity analyses [for example, including only NSCLC studies or studies adjusting for stage], although consequent reduction in numbers did result in loss of significance in some analyses, potentially due to lack of power to detect differences. In order to conduct meta-analysis it is necessary to compare the odds of treatment in the lowest SEP group with the odds in the highest, which simplifies what may be a complex relationship across SEP groups. However, studies that reported a change in odds ratios across the SEP categories and thus explored trends in receipt of treatment generally supported the overall findings of the review. A number of existing tools suitable for assessing cohort study quality were considered (192, 194). However, none of these tools was entirely appropriate for the type of studies included and, as has been done in previous reviews (185, 188), we devised a unique tool, adapting and utilising aspects of other available tools. This approach has the benefit of producing a quality tool that is highly specific for the type of studies examined. As with any systematic review, I was unable to exclude the possibility of publication bias. It was considered whether to search the grey literature including published abstracts, theses and reports from cancer registries and public health observatories, 106 but time and scope restraints prevented this. Studies reporting null findings are less likely to be published or, if they are published, not to report numerical outcomes (197). A funnel plot to assess potential publication bias did not appear to show obvious bias [Figure 7.2]. However, a number of papers recovered in the search included SEP in the description of the study population but did not report receipt of treatment by SEP (64, 161, 162, 235). Study authors were contacted and asked to provide further information, but only one supplied the requested data (161). It is likely that SEP was not significantly associated with receipt of treatment in the other studies but this was not always clearly reported. However, publication bias is thought to be less important than other sources of bias such as confounding, in meta-analyses of observational studies (232). 7.5.4 Implications for policy and practice/ future research Socio-economic inequalities in receipt of treatment may exacerbate socio-economic inequalities in incidence of lung cancer, which is strongly associated with higher smoking rates in more deprived populations, and so may further contribute to the poorer outcomes in lower SEP groups. Socio-economic inequalities in treatment may be due to differences in access to care. Within a non-UHCS it might be expected that socio-economic differences in receipt of treatment would be observed due to income-related differences in insurance status. Patients with lung cancer in the USA who do not have insurance have been reported to have more limited access to care (188). However, as socio-economic inequalities in receipt of lung cancer treatment were also observed in UHCS that do not depend on ability to pay, this would suggest that other system factors may be contributing to this inequality. In non-UHCS, studies in younger populations, examining a range of insurance providers, are required. The extent to which receipt of treatment is influenced by factors such as patient choice is not known. Variability at patient, tumour, system and individual clinician level needs to be investigated before clear recommendations for changes to policy and practice can be made. Further investigation into the factors that might contribute to socio-107 economic inequalities in receipt of lung cancer care is necessary, to help develop interventions that ensure equitable receipt of appropriate treatment. This could include quantitative exploration of inequalities at each stage of the care pathway as well as qualitative work exploring reasons for inequality. It is likely that inequalities in receipt of treatment may contribute to inequalities in cancer survival and so cohort survival analyses are warranted in order to investigate intervention-generated inequalities in lung cancer outcomes. 7.6 Chapter summary This review has demonstrated an association between lower SEP and reduced likelihood of receiving surgery, chemotherapy and any type of unspecified treatment, but not radiotherapy, for lung cancer. If lower SEP patients are more likely to present with later stage cancer [although the evidence for this is unclear] they will be ineligible for surgery. Therefore if stage is not included as a confounder then the absence of controlling for stage may account for some of the reduced likelihood of receipt of surgery in lower SEP groups. Similarly, If SCLC patients are included within the surgery denominator and histology is not controlled for, as greater levels of SCLC are seen in lower SEP groups due to smoking (17), this may be a valid reason for lower rates of surgery in this group. Receipt of treatment may be influenced by clinical suitability for treatment, and socio-economic differences in the number of co-morbidities present may explain socio-economic inequalities in treatment. Better quality UHCS studies, including statistical control for co-morbidity, stage and histology, are required. Chapter 10 describes just such a study using a secondary linked data-set. Chapter 8 now goes on to describe the data-set employed for the secondary data-analysis and an overview of the analytical methods. 108 Chapter 8. Exploring IGIs in lung cancer care using linked secondary data: Methods 8.1 Introduction Socio-economic inequalities in receipt of surgery and chemotherapy for lung cancer were found in my systematic review and meta-analysis (236), as described in the previous chapter [chapter 7]. However, the quality of the included studies was mixed and many did not include important potential confounders such as stage, histology, PS and co-morbidity in the analyses. To further investigate socio-economic inequalities in receipt of lung cancer treatment and the factors that might be influencing this, as well as exploring the role treatment inequality and delay might play in socio-economic inequalities in survival, Intervention-generated inequalities in lung cancer care in the North of England were explored using cancer registry [NYCRIS] data linked to Hospital Episode Statistics [HES] data and National Lung Cancer Audit [LUCADA] data. These linked secondary data sources were used to determine if there were socio-economic inequalities in lung cancer care and to identify where on the pathway of care these inequalities might occur – looking at time from GP referral to first hospital appointment, diagnosis and treatment; receipt of treatment; and survival. This chapter describes the methods employed to do this: the data sources used; variables included and the reasons for their inclusion; problems with data linkage and missing data; ethical approval; and an overview of the analytical methods. Potential variables of interest, particularly those that had not previously been well explored, were determined from the literature and systematic reviews. An initial scoping exercise was then conducted to determine the availability and accessibility of data held by NYCRIS and HES that could be supplied in anonymised form. When it became clear that all variables of interest were not contained within these two datasets further linkage to the lung cancer audit dataset was considered. Tables A1-A3 in Appendix A show the variables that were included in the initial request and the reasons for their inclusion. 109 8.2 Data sources The Northern and Yorkshire Cancer Registry and Information Centre [NYCRIS] is one of eight English regional cancer registries [11 in the UK] that collect a common minimum cancer dataset of information to obtain population-based figures for incidence of, and survival from, cancer. It covers a population of 6.6 million and registers approximately 50,000 incident cancer cases annually. Of these, around 5800 are lung cancer cases [2010 data] (237). From 1st April 2013 it became part of Public Health England. The National Cancer Dataset uses data collected from all the registries. As well as their traditional registry role the registries also provide a cancer information and intelligence function. Validity and completeness of the data is also monitored. Active cancer treatments given within 6 months of diagnosis [surgery, chemotherapy, and radiotherapy] are well recorded in NYCRIS but not always so well recorded in other registries (237). Therefore, in order to look at adjuvant and palliative therapy as well as surgery, the investigative study uses NYCRIS regional, rather than national, lung cancer data. Co-morbidity may be a factor determining whether treatment is offered (238) but details of co-morbidity are not collected by cancer registries. However, this data can be obtained from HES and used to calculate a co-morbidity score. Hospital Episode Statistics are stored in a HES ‘data warehouse’ containing details of all admissions to English NHS hospitals, and recording episodes of care from 1989 onwards. Other details of care including referring GP are also held in HES. The National Cancer Data Repository [NCDR] is a merged dataset of cancer registry data linked to an extract of HES (239). Lack of stage data has been identified as a major limitation of cancer registry data (240). Definitive cancer staging requires extensive clinical investigation and levels of registry data on stage vary by cancer type. Lung cancer staging data is not routinely collected by cancer registries. However, the Lung Cancer Audit [LUCADA], a non-mandatory register of clinical information on patients diagnosed with lung cancer, does collect staging data. The audit initially included only a subset of registry patients whose data has been entered into LUCADA, which began in 2004. The audit examines the care delivered for patients diagnosed with lung cancer, and is run by The Health 110 and Social Care Information Centre in partnership with the Royal College of Physicians on behalf of the Healthcare Quality Improvement Partnership [HQIP]. 8.3 Data Data for patients with a primary diagnosis of lung cancer [ICD10 C33 and C34], diagnosed between 1997 and 2010 were requested from NYCRIS, and where linked data were available, with a linkage to HES [to supply co-morbidity score] and LUCADA [for stage and PS data]. The three data sources [NYCRIS cancer registry, HES and LUCADA data] were anonymised by NYCRIS and the three data-sets supplied by NYCRIS. The Thames Cancer Registry, the lead registry for lung cancer were contacted and they informed NYCRIS of the appropriate methods for linking, processing and extracting data from LUCADA, to ensure consistent rules were applied by NYCRIS. Records were allocated a unique randomly-generated key number, derived from the NHS number by NYCRIS, and the HES and LUCADA data were then linked by LF using this key. 8.3.1 Data problems 8.3.1.1 Delays in obtaining data, and data linkage Registry data for patients diagnosed from 1997 to 2010 were supplied by NYCRIS in March 2012 but at that point it was still unclear whether HES and LUCADA data would be available. A national problem with the way that the co-morbidity score had been calculated meant that the data was withdrawn from use and NYCRIS were unclear about when this problem would be resolved. After much uncertainty the data eventually became available in November 2012. LUCADA data were obtained in December 2012. Data from HES and LUCADA were then imported and linked to the cleaned NYCRIS dataset. 8.3.1.2 Data format When the initial registry data was received, in order to ensure that the data were non-identifiable I was not able to have access to the variable ‘date of death’. Instead survival time from date of diagnosis was calculated by NYCRIS. However, this was given in years, which was not appropriate for meaningful lung cancer survival analysis as 111 lung cancer has such a short survival period. After some negotiation it was agreed that survival time in weeks could be supplied. An estimated survival time in days could then be calculated with an accuracy of within 7 days of actual death. However, when the ‘survival in weeks’ variable did become available it was unfortunately not possible to append the new variable onto the existing dataset due to the way that the unique key for the dataset had been randomly generated by NYCRIS. Therefore a whole new un-cleaned dataset had to be sent by NYCRIS, meaning that all data cleaning and the generation of new variables had to be carried out again. Luckily as Stata ‘do’ files had been created for the initial data cleaning, this was not as big a set-back as it could have been. 8.3.1.3 Missing data In the linked dataset the levels of stage, PS and co-morbidity data missing were high. Inpatient HES data up until 31/3/2010 were theoretically available. However, although linked HES co-morbidity data were available for between 88.4% up to 92.0% of patients diagnosed between 1999 and 2008 [figures that are similar to those found in a similar study linking Thames cancer registry and HES data (157)], only 1.9% of those diagnosed in 2009 and 1.8% diagnosed in 2010 had a linked HES record [Appendix B]. National problems with the methodology used to calculate the co-morbidity score contained in the National Cancer Dataset meant that these all had to be re-calculated and so there was a resultant time-lag in their availability (NYCRIS, personal communication). Stage and PS data were available from 2006 onwards. However, stage was only recorded for 27% of the 2006-2010 cohort, resulting in a large ‘missing’ category. Unfortunately the years in which the highest levels of stage and PS data were available [2009-2010] were also the years in which HES-linked data [and hence co-morbidity score] were not available for over 98% of the cohort. Patients diagnosed between 2006 and 2008 [n= 17,096] potentially had stage, PS and co-morbidity recorded although levels of stage data in these years were low and only 2080 had both details of stage and co-morbidity recorded. Those diagnosed between 1999 and 2005 had co-morbidity score but not stage or PS available. 112 The method for calculating date of diagnosis was changed in 2010 which might influence the comparability of the time intervals used in the time to diagnosis and treatment analyses for this year. Some decisions therefore had to be made regarding which years of data to include in the analyses. 8.3.2 Data-sets Data for patients with a primary diagnosis of lung cancer [ICD10 C33 and C34], diagnosed between 1997 and 2010 were initially requested from NYCRIS. However, patients diagnosed prior to 1999 did not have a valid IMD code and so could not be utilised. Data for 66,891 patients with a primary diagnosis of lung cancer [ICD10 C33 and C34], diagnosed between 1st Jan 1999 and 31st December 2010, were available for use. Of these, 1681 had tumour registration based on death-certification only and so were excluded from analyses, leaving an eligible cohort of 65,210. 7776 [11.9%] had stage recorded in LUCADA [7769 in 2006-2010] and 51,614 [79.2%] had a HES linkage. It was decided to split the cohort into two to best deal with missing data. Group 1 consisted of patients diagnosed between 1999 and 2005 [n=36,477], of whom 32,974 [90.3%] had a linked HES record. Stage and PS were not available. Patients diagnosed between 2006 and 2010 [n= 28,733] made up the 2nd group and, of these, only 18,650 [64.9%] had a linked HES record, 7769 [27.0%] had stage recorded in LUCADA and 8885 [30.9%] had a PS score. Analyses with and without the inclusion of co-morbidity were also conducted for group 2. Multiple imputation for variables with missing data was considered but it is not recommended where over 50% of the data in a variable are missing (241). An alternative way to address the problem of missing data is to analyse only complete cases, although results from complete-case analyses can be biased (242). It was decided to analyse the full 2006-2010 dataset and include a ‘stage missing’ category and also to conduct a sub-group analysis of the 7769 staged-patients diagnosed in 2006-2010. 113 The distribution of each variable in the 1999-2005 and 2006-2010 subgroups and in the staged subgroup were compared to the overall cohort and to each other using the Chi squared (χ2) test, to determine the representativeness of each of these groups. Statistical analyses were performed using STATA v 12.0. 8.4 Data variables 8.4.1 Registry data 8.4.1.1 Socio-demographic factors - SEP Socio-economic position [SEP] was calculated according to the agreed methodology for all English cancer registries, as the rank of the income domain of the Index of Multiple Deprivation [IMD], grouped into quintiles, based on the England-wide distribution of this variable. IMD is an area-based composite measure of SEP and the UK government’s preferred measure of deprivation. The English IMD provides a relative measure of deprivation at small area level across England. Areas are ranked from least deprived to most deprived, on seven different dimensions of deprivation as follows: income deprivation; employment deprivation; health deprivation and disability; education deprivation; crime deprivation; barriers to housing and services deprivation; and living environment deprivation (243). Postcodes of residence were used to assign individuals to small administrative areas known as lower-level super output areas [LSOA], containing an average of 1500 individuals. To avoid any risk of including an indicator incorporating a health score, some studies have previously used the composite measure of IMD with the health domain removed (78, 199), but the agreed methodology for all English cancer registries is now to use only the income domain of the IMD and not to provide actual deprivation scores per LSOA, as scores are not a proportional measure of deprivation. An LSOA with a score of 40 is not twice as deprived as one with a score of 20 and so raw scores are not particularly meaningful (244). IMD is not directly comparable between England, Scotland and Wales as there are differences in sub-measures due to differences in rules for benefit qualification. 114 IMD quintiles were supplied by NYCRIS, where 5 is the most deprived and 1 the least deprived. The income domain of IMD 2010 was used for the most recent years of data, which is based on 2008 data and thus populations were derived from 2008 data for patients diagnosed between 2007 and 2010. For those diagnosed between 2003 and 2006 the income domain of IMD2007 was used [derived from 2005 data] and for 1999-2002 the income domain of IMD2004 [derived from 2001 data]. This changing-deprivation-score-over-time methodology was thought to be preferable to applying deprivation based on 2008 data back in time to much earlier years of diagnosis, as this may not properly reflect the deprivation of that area if, for example, there had been regeneration over the last decade (244). Although data for patients diagnosed in 1997 and 1998 were available, valid IMD data were only available from 1999 onwards and so the records for the two earlier years were not used. 8.4.1.2 Other socio-demographic factors Age at diagnosis was categorised into 4 groupings: age <60, 60-69, 70-79 and 80+ years. Sex was categorised as M [male] and F [female]. Marital status is recorded by NYCRIS. This variable was requested but due to poor data completeness was not able to be supplied. Data on ethnicity was supplied. However as ethnicity information was missing for 38% of the dataset and, of the 62% that had this variable coded, less than 1% of the dataset were characterised as non-white, this variable was not included in the final analyses. As inequalities in receipt of treatment may have changed over time and to take into account the effect of the introduction of new guidelines for referral in 2005 (18), year of diagnosis was split into 4 categories: 1999-2001, 2002-2004, 2005-2007, 2008-2010. Individual years were utilised for the 1999-2005 and 2006-2010 datasets. 115 8.4.1.3 Tumour factors There were 67 different cancer type codes listed in the registry dataset. Nine codes encompassed 96.1% of the records, relating to seven recognisable tumour types [table 8.1]. The remaining 3.9% had one of 58 different codes and these were classified as other-specified carcinomas. Lung cancer was therefore categorised into eight morphological subtypes: adenocarcinoma, large cell carcinoma, non-small cell carcinoma, squamous cell carcinoma, small-cell carcinoma, other specified carcinoma [of which 26.6% were carcinoid tumours, codes M82413 and M82463], unspecified carcinoma (20) and neoplasm [table 8.1]. The neoplasm subtype included patients with a clinical diagnosis only. Table 8.1. lung cancer histology and tumour type Cancer code Description N % Tumour type Histology M81403 Adenocarcinoma unspecified 9,463 14.5 Adenocarcinoma NSCLC M80103 Carcinoma unspecified 4,615 7.1 Unspecified Other M80463 Non-small cell carcinoma 8,661 13.3 Non-small NSCLC M80123 Large cell carcinoma unspecified 2,140 3.3 Large cell NSCLC M80003 Neoplasm malignant 15,390 23.6 Neoplasm Other M80413 Small cell carcinoma unspecified 8,599 13.2 Small cell SCLC M80723 Squamous cell carcinoma non-keratinising 13,831 21.2 Squamous NSCLC M80723 Squamous cell carcinoma NSCLC M80713 Squamous cell carcinoma keratinising NSCLC Other cancer type codes (n=54) 2,511 3.9 Other specified carcinomas Other Total 65,210 100 Cancer type [histology] was classified as confirmed non-small cell lung cancer [NSCLC], including adenocarcinoma, large cell carcinoma, non-small cell carcinoma and squamous cell carcinoma subtypes; small cell lung cancer [SCLC]; and Other histology [including unspecified carcinoma, neoplasm, and other specified carcinomas (including carcinoid tumours)] [table 8.1]. When including only morphologically-specified lung cancers in analyses then the ‘unspecified carcinoma/neoplasm’ subtypes were excluded and the ‘other specified carcinoma’ subtype was included as probable NSCLC. 116 8.4.1.4 Treatment The types of treatment [surgery, chemotherapy and radiotherapy] were used to derive a number of different treatment variables: three binary receipt of treatment variables - Surgery [y/n], Chemotherapy [y/n], Radiotherapy [y/n]; Type of First Treatment – categorised as surgery first, chemotherapy first, radiotherapy first, and no treatment; Types of Treatment – categorised as surgery, surgery + chemotherapy/radiotherapy, chemotherapy, chemotherapy + radiotherapy, radiotherapy, no treatment. 8.4.1.5 Dates, interval periods and target times 8.4.1.5.1 Dates Dates of GP referral, first hospital appointment, diagnosis, and treatment received [surgery, chemotherapy and radiotherapy], were extracted from the cancer registry data. For the 1999-2009 NYCRIS data, diagnosis date was determined by searching through case notes to collect information. Diagnosis date was determined as the first time the tumour was identified either by imaging or histology. From 2010 onwards UKACR guidance was followed. The guidance uses a priority hierarchy with 6 categories to determine date of diagnosis; the date of histological or cytological confirmation of malignancy [which can be derived from 3 different hierarchical time points]; then date of hospital admission; down to, with lowest priority, date of death if cancer is only identified at autopsy (100). In order to calculate meaningful diagnostic and treatment intervals it is necessary to exclude those who have a diagnosis only after death. All death certificates that mention cancer are returned to the registry so that date of death can be recorded. If this cancer has not previously been entered on the register then further details are searched for retrospectively, but if medical records cannot be located then the cancer is registered as Death Certificate Only [DCO].These cases [n=1681] were excluded from analyses. 117 8.4.1.5.2 Intervals and target times The following referral, diagnostic and treatment time intervals were potentially available for investigation: Table 8.2. Referral, diagnostic and treatment intervals Interval Definition of interval NCP* target time Referral GP referral date to first hospital appointment date 14 Diagnostic GP referral date to diagnosis date 31 (62-31) first hospital appointment (FHA) date to diagnosis date 17 (31-14) Treatment diagnosis date to first treatment date 31 GP referral date to first treatment 62 *National Cancer Plan In England, three of these intervals have been the subject of performance management. Since 2000, urgent referrals for suspected cancer have been required to have a first hospital appointment [FHA] within 14 days from the date of referral [referral interval]. Since 2005, intervals of 62 days from date of urgent GP referral to first treatment and 31 days from diagnosis/decision to treat to first treatment [treatment intervals] have been in place (24). Interim time target periods [italics] can be inferred from these stated target times and calculated as follows: GP referral to diagnosis [62-31=31 days] and FHA to diagnosis [31-14=17 days]. Time from GP referral date to FHA was categorised as <=14 days [within target], >14 days, no referral interval recorded [either no GP referral date or no FHA date]. Time from GP referral date to treatment was categorised as <=62 days [within target], >62 days, no referral interval recorded [either no GP referral date or no FHA date]. Time from diagnosis to first treatment was categorised as: <=31 days [within target], 32-62 days, >62 days, no treatment. 8.4.2 HES data – co-morbidity score Co-morbidities were derived from diagnostic codes in HES. A calculated co-morbidity score was then derived using the Charlson co-morbidity index, a validated instrument (245), and recorded in the National Cancer Registry Dataset. Only patients who had an inpatient episode with a condition that counted towards co-morbidity were included. Inpatient HES data up until 31/3/2010 were theoretically available. 118 A Charlson co-morbidity [CCM] score was calculated by NYCRIS using the number of inpatient HES admissions for 17 specified conditions [other than lung cancer] in the 3- 18 months prior to diagnosis. Co-morbidities were assigned a weighted score [table 8.3] and the total score was the sum of weighted scores for the co-morbidities experienced. No HES data linkage was available for 98.3% of patients diagnosed in 2009-10 as national problems in calculating the co-morbidity score meant that there was a time-lag in their availability. Patients without a CCM score were split into those who did and did not have a HES linkage. Those who had a linked HES record but no CCM score recorded on HES [CCM missing] were analysed separately from those who had no linked HES record and therefore no CCM score [no HES linkage]. CCM score was categorised as 0, 1-2, 3+, CCM missing, and no HES linkage. Table 8.3. Calculation of Charlson co-morbidity (CCM) score Charlson Group Description Charlson Score Notes 1 Acute Myocardial Infarction 1 2 Congestive Heart Failure 1 3 Peripheral Vascular Disease 1 4 Cerebral Vascular Accident 1 5 Dementia 1 6 Pulmonary Disease 1 7 Connective Tissue Disorder 1 8 Peptic Ulcer 1 9 Diabetes 1 Only highest score is counted 10 Diabetes Complications 2 11 Paraplegia 2 12 Renal Disease 2 13 Cancer 2 Derived from cancer registry data rather than HES data. 14 Metastatic Cancer N/A 17 Liver Disease 1 Only highest score is counted 15 Severe Liver Disease 3 16 HIV 6 8.4.3 LUCADA data – stage and PS Lung cancer staging data is very poorly recorded in NYCRIS. Stage and performance status were therefore obtained from LUCADA records. These variables were only substantially available for patients who were diagnosed between 2006 and 2010. The audit includes only a subset of registry patients whose data has been entered into LUCADA, which began in late 2004. Nationally, the percentage completeness varies 119 from only 40% for those diagnosed in 2005, with gradual increases over the years [66% in 2006, 75% in 2007, 85% in 2008, 95% in 2009, 93% in 2010 and 93% in 2011] (246). The number with stage recorded was also initially low [55% in 2006, although this is 55% of the 66% included so is only around 36.3% of the full number diagnosed that year] but again is increasing over time [85% in 2010, equivalent to 79% of all those diagnosed]. For my linked NYCRIS dataset the numbers were lower than this. Stage was recorded for only 11.8% in 2006 and increased to 63.1% in 2010 [see Appendix B]. Stage was assigned using the TNM staging system and categorised as I [IA,IB], II [IIA,IIB], III [IIIA,IIIB], IV and missing/uncertain. Performance status [PS] is a measure of general well-being for cancer patients, as assessed by the Multi-Disciplinary Team [MDT], on a scale of 0-4 using the Eastern Co-operative Group performance status scale [ECOG PS] (18) [table 8.4]. A code of 5 signifies that PS is missing. Table 8.4. Performance status scale ECOG/ WHO (Zubrod) scale(119) Short Description LUCADA Description 0 Asymptomatic Able to carry out all normal activity without restriction 1 Symptomatic but ambulatory Restricted in physically strenuous activity but able to walk and do light work 2 In bed < 50% of day Able to walk and capable of all self care but unable to carry out any work. Up and about more than 50% of waking hours 3 In bed >50& of day Capable of only limited self care, confined to bed or chair more than 50% of waking hours 4 Bedridden Completely disabled. Cannot carry on any self care. Totally confined to bed or chair 5 -- Not recorded 8.5 Ethical Approval Ethical Approval was applied for through the Integrated Research Application System [IRAS] for NHS Research Ethics Committee [REC] approval. The shorter Proportionate Review process was used as only anonymised data was requested. The National Information Governance Board for Health and Social Care [NIGB], which deals with applications to access patient data without consent, were contacted and agreed that, providing mechanisms to prevent small numbers being disclosed and thus potentially-identifying patients were discussed with NYCRIS, NIGB approval was not necessary. 120 Nor was NHS R&D approval required, although I was initially informed that this would be required from Leeds Teaching Hospitals NHS Trust, as the trust that holds the NYCRIS registry. This did delay the Ethics process whilst this was resolved. A favourable ethical opinion was obtained from the Proportionate Review sub-committee of the NRES Committee East of England REC on the 13th December 2011 [REC reference 11/EE/0537]. Access to NYCRIS data was applied for using the UK Association of Cancer Registries [UKACR] Information Request form. The UKACR considers the following to be identifiable data items: date of birth, postcode, date or cause of death; therefore I did not request these items but rather anonymised data containing variables derived from these data items. Data items that could be classified as potentially-identifiable are: individual records even if they do not contain identifiable information and tables of data with low cell counts. However, as I was requesting large datasets for common cancers with all identifiable data anonymised, individual patients should not be identifiable. 8.6 Overview of secondary data analysis analytical methods The triple linked dataset was used to examine the factors that may influence socio-economic inequalities in lung cancer treatment; referral, diagnosis and treatment time; and survival. The analytical methods are summarised here and detailed statistical methods are given in chapters 9, 10, 11 and 12. 8.6.1 Socio-economic inequalities in receipt of treatment Receipt of treatment by socio-economic position was examined using logistic regression. Other variables that were of interest [age, sex, histology, year of diagnosis, co-morbidity, stage, PS] were examined in unadjusted analyses and included in the adjusted model if significant in the univariable analysis at p<0.05 or if thought to be of a priori importance. Outcome was receipt of treatment by SEP for:  Surgery  Chemotherapy  Radiotherapy 121 These results are in chapter 10. 8.6.2 Socio-economic inequalities in referral, diagnostic and treatment intervals Socio-economic inequalities in referral, diagnosis and treatment time for lung cancer patients were examined using Cox proportional hazard models. The following intervals were considered:  Referral [time from referral to secondary care investigation/first hospital appointment]  Diagnosis [time from first hospital appointment to diagnosis, and time from GP referral to diagnosis]  Treatment [time from diagnosis to first treatment, and time from GP referral to first treatment] Multivariable Cox Proportional Hazard models for each time period were constructed including variables that were thought to be of a-priori importance, to examine the hazard ratio [HR] of early referral, diagnosis and treatment, by SEP. Since 2000, urgent referrals for suspected cancer have been required to have a first hospital appointment [FHA] within 14 days from the date of referral [referral interval]. Since 2005, intervals of 62 days from date of urgent GP referral to first treatment and 31 days from diagnosis to first treatment [treatment intervals] have been in place. The likelihood of referral, diagnosis and treatment within recommended target times by SEP was examined using logistic regression, in multivariable models including other factors [age, sex, histology, year of diagnosis, co-morbidity, stage, PS] that may impact on these time intervals. These results are in chapter 11. 8.6.3 Socio-economic inequalities in survival Survival time was calculated as the interval from the date of diagnosis to the date of death or to the end of follow up at 31/12/2011 for those still alive, when data was censored. All-cause mortality was analysed using Kaplan-Meier plots and Cox proportional hazard models. Kaplan Meier graphs were used to examine univariable influences on all-cause mortality. Cox regression models were used to calculate hazard ratios [HRs] and 95% CIs for all-cause mortality in relation to SEP in multivariable 122 models. Logistic regression was used to examine the odds of still being alive two years [for 2006 to 2009 and 1999-2005 cohorts] after diagnosis, by SEP. 8.7 Chapter summary This chapter has given an overview of the secondary data analysis analytical methods and has described the data sources used and variables included, as well as problems with data linkage and missing data. In chapter 9 descriptive results for the dataset are presented. The demographic and clinical characteristics of the patients are examined, as are associations between SEP and a number of variables of interest. In chapter 10 multivariable analyses to examine socio-economic inequalities in lung cancer treatment are presented. Chapter 11 looks at inequalities in the referral, diagnostic and treatment intervals and chapter 12 examines inequalities in lung cancer survival. Each chapter presents detailed methods, self-contained results and discussion, which places results in context with the literature, discusses the strengths and weaknesses of the study and the potential implications for policy and practice and further research required. Chapter 13 then summarises the overall findings of the thesis, as well as detailing overall methodological strengths and weaknesses of the secondary data analysis, detailing any further implications for policy and practice, and further research required. 123 Chapter 9. IGIs in lung cancer care – dataset overview and descriptive statistics 9.1 Introduction In this chapter a descriptive overview of the dataset is given. The demographic and clinical characteristics of the patients in the entire dataset and in the 1999-2005 cohort, 2006-2010 cohort, and 2006-2010 cohort with stage recorded, are examined and compared. Associations between SEP and a number of variables of interest are also examined. Time trends and differences between the datasets are considered and the implications of these differences for the representativeness of the datasets discussed in relation to the proposed multivariable analyses to be carried out. 9.2 Methods Descriptive statistics for the datasets were calculated using the Chi squared [χ2 ] test to describe the characteristics and representativeness of each dataset; to describe characteristics over time; to examine univariable associations between SEP and other patient and tumour characteristics [stage, age, tumour type, PS and CCM score]; and to examine univariable associations between stage, PS and CCM score. 9.3 Results - descriptive statistics 9.3.1 Characteristics of the datasets Table 9.1 shows the demographic and clinical characteristics of patients diagnosed with lung cancer between 1999 and 2010 [n=65,210, excluding those patients who were recorded as DCO] and comparing them with the subset diagnosed from 1999-2005 [n=36,477], 2006-2010 [n=28,733] and, within this, the subset in 2006-2010 with stage recorded [n=7769]. The highest percentage of patients were found in the most deprived quintile [34.8%] with the smallest percentage in the most affluent quintile [11.1%] in the full dataset. A higher percentage of affluent patients were found in the 2006 to 2010 cohort [11.8%] and in the stage subset [12.0%] [table 9.1]. 124 Table 9.1. Demographic and clinical characteristics of patients diagnosed with lung cancer (DCO excluded) between 1999 and 2010, and cohorts diagnosed between 1999-2005, 2006-2010, and 2006-2010 subset with stage recorded Variable All patients 1999-2010 cohort 1999-2005 cohort 2006-2010 2006-2010 subset with stage N % N % N % N % Deprivation quintile 65210 100 36,477 100 28733 100 7769 100 1 (least deprived) 7246 11.1 3,857 10.6 3,389 11.8 931 12.0 2 9265 14.2 5,087 14.0 4,178 14.5 1118 14.4 3 10853 16.6 6,005 16.5 4,848 16.9 1300 16.7 4 15163 23.3 8,453 23.2 6,710 23.4 1831 23.6 5 (most deprived) 22683 34.8 13,075 35.8 9,608 33.4 2589 33.3 Sex 65210 100 36,477 100 28733 100 7769 100 Female 28668 44.0 15,414 42.3 13254 46.1 3559 45.8 Male 36542 56.0 21,063 57.7 15479 53.9 4210 54.2 Age group 65210 100 36,477 100 28733 100 7769 100 <60 8905 13.7 5,223 14.3 3,682 12.8 1041 13.4 60-69 17130 26.3 9,535 26.1 7,595 26.4 2189 28.2 70-79 24830 38.1 14,582 40.0 10,248 35.7 2843 36.6 80+ 14345 22.0 7,137 19.6 7,208 25.1 1696 21.8 Histology 65210 100 36,477 100 28733 100 7769 100 NSCLC 34095 52.3 18,972 52.0 15,123 52.6 5116 65.9 SCLC 8599 13.2 5,104 14.0 3,495 12.2 582 7.5 Other 22516 34.5 12,401 34.0 10,115 35.2 2071 26.7 Co-morbidity score 65210 100 36,477 100 28733 100 7769 100 0 10063 15.4 6,053 16.6 4,010 14.0 995 12.8 1 4968 7.6 2,588 7.1 2,380 8.3 579 7.5 2 2316 3.6 1,165 3.2 1,151 4.0 278 3.6 3+ 1639 2.5 705 1.9 934 3.3 226 2.9 CCM score missing 32,628 50.0 22,453 61.6 10,175 35.4 1,977 25.5 No HES link 13,596 20.9 3,513 9.6 10,083 35.1 3,714 47.8 Stage 65210 100 36,477 100 28733 100 7769 100 I -- -- -- -- 1,186 4.1 1186 15.3 II -- -- -- -- 552 1.9 552 7.1 III -- -- -- -- 2,273 7.9 2273 29.3 IV -- -- -- -- 3,758 13.1 3758 48.4 Missing/ unknown -- -- -- -- 20,964 73.0 -- -- Performance Status 65210 100 36,477 100 28733 100 7769 100 0 -- -- -- -- 1,842 6.4 1,493 19.2 1-2 -- -- -- -- 4,865 16.9 3,870 49.8 3+ -- -- -- -- 2,178 7.6 1,763 22.7 Missing/unknown -- -- -- -- 19,848 69.1 643 8.3 Surgery 65210 100 36,477 100 28733 100 7769 100 Yes 6407 9.8 3,513 9.6 2,894 10.1 1001 12.9 No 58803 90.2 32,964 90.4 25,839 89.9 6768 87.1 Chemotherapy 65210 100 36,477 100 28733 100 7769 100 Yes 15996 24.5 7,648 21.0 8,348 29.1 2732 35.2 No 49214 75.5 28,829 79.0 20,385 71.0 5037 64.8 Radiotherapy 65210 100 36,477 100 28733 100 7769 100 Yes 22216 34.1 12,602 34.6 9,611 33.5 3069 39.5 No 42994 65.9 23,875 65.5 19,122 66.6 4700 60.5 GP referral date 65210 100 36,477 100 28733 100 7769 100 Yes 32,140 49.3 16,688 45.7 15,452 53.8 5,351 68.9 No 33,070 50.7 19,789 54.3 13,281 46.2 2,418 31.1 125 Table 9.1(cont). Demographic and clinical characteristics of patients diagnosed with lung cancer (DCO excluded) between 1999 and 2010, and cohorts diagnosed between 1999-2005, 2006-2010, and 2006-2010 subset with stage recorded Variable All patients 1999-2010 cohort 1999-2005 cohort 2006-2010 2006-2010 subset with stage N % N % N % N % Year of Diagnosis 65210 100 36,477 100 28733 100 7769 100 1999 5,154 7.9 5,154 14.1 -- -- -- -- 2000 5,227 8.0 5,227 14.3 -- -- -- -- 2001 5,200 8.0 5,200 14.3 -- -- -- -- 2002 5,219 8.0 5,219 14.3 -- -- -- -- 2003 5,221 8.0 5,221 14.3 -- -- -- -- 2004 5,216 8.0 5,216 14.3 -- -- -- -- 2005 5,240 8.0 5,240 14.4 -- -- -- -- 2006 5,533 8.5 -- -- 5,533 19.3 671 8.6 2007 5,712 8.8 -- -- 5,712 19.9 866 11.2 2008 5,851 9.0 -- -- 5,851 20.4 1,556 20.0 2009 5,871 9.0 -- -- 5,871 20.4 2,140 27.6 2010 5,766 8.8 -- -- 5,766 20.1 2,536 32.6 Tumour type 65210 100 36,477 100 28733 100 7769 100 Adenocarcinoma 9,463 14.5 5,001 13.7 4,462 15.5 1473 19.0 squamous 13,831 21.2 8,602 23.6 5,229 18.2 1850 23.8 Large-cell 2,140 3.3 1,372 3.8 768 2.7 169 2.2 non-small cell 8,661 13.3 3,997 11.0 4,664 16.2 1624 20.9 Small cell 8,599 13.2 5,104 14.0 3,495 12.2 582 7.5 Other specified 2,511 3.9 1,356 3.7 1,155 4.0 298 3.8 Unspecified carcinoma 4,615 7.1 4,095 11.2 520 1.8 93 1.2 neoplasm 15,390 23.6 6,950 19.1 8,440 29.4 1680 21.6 Alternative histology 65210 100 36,477 100 28733 100 7769 100 Probable NSCLC 36,606 56.1 20,328 55.7 16,278 56.7 5,414 69.7 SCLC 8,599 13.2 5,104 14.0 3,495 12.2 582 7.5 Unspecified/neoplasm 20,005 30.7 11,045 30.3 8,960 31.2 1,773 22.8 Note: stage and performance status data only available for patients diagnosed between 2006 and 2010 Table 9.1b. Comparison of variable distribution in the datasets Variables 2006-10 cohort v 1999-2005 cohort Staged subset v non-staged 2006-10 cohort χ2 P χ2 P SEP 54.16 <0.001 0.86 0.930 Age 328.21 <0.001 62.65 <0.001 Sex 97.78 <0.001 0.44 0.511 Histology 48.73 <0.001 765.53 <0.001 CCM 244.22 <0.001 39.02 <0.001 Surgery 3.53 0.06 92.99 <0.001 Chemotherapy 567.80 <0.001 192.96 <0.001 Radiotherapy 8.76 0.003 175.30 <0.001 126 Rates of treatment were higher in the 2006-2010 sub-group that had stage recorded, compared to the 2006-2010 cohort. However, the population distribution by SEP and sex was similar. The age distribution amongst the cohorts was significantly different [table 9.1b]. A higher percentage of patients were diagnosed at an older age in 2006-2010 [25.1% were aged 80+] compared to the earlier years cohort [19.6% were diagnosed at age 80+ in 1999-2005]. However in the cohort with stage recorded higher percentages of younger patients were found [table 9.1]. The sub-population with stage had a much higher percentage of patients diagnosed with NSCLC [65.9% compared to 52.6% in the full 2006-10 cohort]. This is not surprising as staging is not generally carried out for SCLC. In my systematic review of published research [chapter 7], rates of surgery in UK studies ranged from 4.8% to 13.0% [table 7.5]. Higher rates were found in NSCLC only studies [8.9 to 19.1%]. Chemotherapy rates [any histology] ranged from 15.5 to 23.8% and radiotherapy rates from 20.3 to 63.1%. Any type of treatment rates ranged from 56.3 to 59.6 [table 7.11]. The results from this dataset fall within these ranges for surgery [9.6% in 1999-2005, 10.1% in 2006-2010], chemotherapy [21.0% in 1999-2005, 29.1% in 2006-2010] and radiotherapy [34.6% on 1999-2005 and 33.5% in 2006-2010] [table 9.1]. 9.3.2 Time trends Looking at distribution by SEP over time, in 1999 8.5% of patients were in the highest SEP group and 38.7% in the lowest, in 2006 this was 11.2% and 33.5% respectively, and in 2010 this had changed to 12.0% and 32.2% respectively. When looking at the full 1999-2010 dataset these differences were significant [χ2=225.53, p<0.001] with a higher percentage of people in a higher social class in later years. However, when looking at only the 2006-10 subset there were no significant differences in SEP distribution over this timescale [χ2 =20.15, p=0.21]. Rates of surgery have increased slightly in recent years but there has been a near doubling of rates of chemotherapy from 1999 [15.2%] to 2010 [29.9%] [table 9.2]. Radiotherapy rates have fallen slightly over time but overall treatment rates have changed little, with rates consistently between 52- 55%. 127 Table 9.2. Rates of treatment over time Year Surgery Chemotherapy Radiotherapy Total N % N % N % N 1999 494 9.6 784 15.2 2,059 40.0 5,154 2000 508 9.7 937 17.9 2,013 38.5 5,227 2001 489 9.4 923 17.8 1,881 36.2 5,200 2002 513 9.8 1,048 20.1 1,730 33.2 5,219 2003 505 9.7 1,170 22.4 1,639 31.4 5,221 2004 480 9.2 1,291 24.8 1,615 31.0 5,216 2005 524 10.0 1,495 28.5 1,665 31.8 5,240 2006 505 9.1 1,506 27.2 1,857 33.6 5,533 2007 547 9.6 1,667 29.2 1,865 32.7 5,712 2008 502 8.6 1,681 28.7 1,865 31.9 5,851 2009 682 11.6 1,773 30.2 2,060 35.1 5,871 2010 658 11.4 1,721 29.9 1,964 34.1 5,766 Total 6,407 9.8 15,996 24.5 22,213 34.1 65,210 9.3.3 SEP and patient and tumour characteristics Table 9.3 shows the percentage of patients in each SEP group by stage at diagnosis, age at diagnosis, tumour type, PS, and CCM score, as well as the percentage who were recorded as DCO, for the 1999-2005 and 2006-2010 cohorts. There was a significant association between IMD quintile and being DCO [χ2 =26.89, p<0.001] in the 1999-2005 cohort, with 1.6% of those in the highest SEP group reported as DCO compared to 3.1% in the lowest. This was not found in the 2006-10 dataset however, where 2.1% of high and low SEP patients were DCO [χ2 = 6.88, p=0.14] [table 9.3]. IMD quintile was significantly associated with age at diagnosis in both cohorts where those in the most deprived quintile had a younger age at diagnosis [χ2 = 116.19, p<0.001] in 2006-2010, with 11.4% of those in the most affluent SEP quintile diagnosed at <60 years old compared to 14.2% in the lowest SEP quintile. Similar results were found in 1999-2005 [13.9% compared to 15.3%, χ2=149.27, p<0.001] [table 9.3]. For the 7769 patients diagnosed in 2006-2010 who had stage recorded, although the percentage who were diagnosed with stage 1 cancer was slightly higher in the highest SEP group compared to the lowest [16.1% compared to 15.0%] and the percentage diagnosed with stage 4 cancer lower [46.6% compared to 49.3%], overall no significant difference in stage at diagnosis was found by SEP [χ2=4.87, p=0.96] [table 9.3]. Stage was not available for the 1999-2005 cohort. 128 Table 9.3. Percentage of patients in each SEP quintile (1=least deprived, 5=most deprived), by age at diagnosis, stage at diagnosis, tumour type, CCM score, PS, referral route, and percentage recorded as death-certificate only (DCO) SEP quintile: 1999-2005 SEP quintile: 2006-2010 Variables: 1 2 3 4 5 1 2 3 4 5 DCO DCO=yes 1.6 2.5 2.8 2.8 3.1 2.1 2.2 2.0 2.6 2.1 Age at diagnosis <60 13.9 13.9 14.2 13.3 15.3 11.4 11.4 11.6 13.4 14.2 60-69 23.8 24.7 24.8 25.9 28.2 25.8 25.4 25.5 25.8 28.0 70-79 39.8 39.5 39.8 40.9 39.7 34.8 36.1 35.5 35.8 35.7 80+ 22.6 21.9 21.1 20.0 16.8 28.0 27.3 27.4 25.0 22.0 Stage I -- -- -- -- -- 16.1 15.9 15.1 14.9 15.0 II -- -- -- -- -- 7.2 7.2 7.8 7.2 6.7 III -- -- -- -- -- 30.1 28.2 29.9 29.4 29.0 IV -- -- -- -- -- 46.6 48.8 47.2 48.6 49.3 Tumour type Adenocarcinoma 16.3 15.5 14.0 13.5 12.2 19.3 17.2 15.6 14.5 14.2 squamous 19.9 22.6 23.3 23.5 25.2 15.8 17.2 17.6 17.9 20.0 Large-cell 4.7 4.0 3.8 3.7 3.4 2.9 3.1 2.3 2.9 2.4 non-small cell 12.2 11.4 10.4 10.6 10.9 16.8 15.9 16.2 16.9 15.8 Small cell 12.9 13.3 13.2 14.1 14.9 11.8 12.1 12.8 12.3 12.0 Other specified 5.1 4.0 3.8 3.6 3.2 5.0 4.7 3.9 4.0 3.5 Unspecified carcinoma 10.2 10.7 12.0 11.6 11.1 2.1 1.6 1.6 2.0 1.7 neoplasm 18.6 18.5 19.4 19.5 19.0 26.4 28.2 30.0 29.5 30.5 Co-morbidity score 0 61.9 60.2 59.5 56.2 55.6 54.5 52.3 50.1 44.8 43.3 1-2 32.6 33.5 33.9 36.6 37.6 36.7 37.0 40.4 42.9 45.0 3+ 5.6 6.3 6.7 7.2 6.9 8.9 10.7 9.6 12.3 11.6 Perf Status 0 -- -- -- -- -- 26.3 25.9 20.9 19.0 17.7 1-2 -- -- -- -- -- 53.0 52.8 54.2 55.9 55.7 3+ -- -- -- -- -- 20.6 21.4 24.9 25.1 26.6 GP referral GP ref date = yes 47.2 47.3 46.7 45.7 44.3 53.6 51.5 53.8 54.7 54.2 129 SEP was associated with histological subtype [tumour]. 16.3% of those in the highest SEP group were diagnosed with adenocarcinoma and 19.9% with squamous cell, whereas in the lowest SEP group 12.2% had adenocarcinoma and 25.2% squamous cell in the 1999-2005 cohort [χ2=173.74, p<0.001]. Similarly, in the 2006-2010 dataset 19.3% in the highest SEP group were diagnosed with adenocarcinoma and 15.8% squamous cell whereas in the lowest SEP group 14.2% have adenocarcinoma and 20.0% squamous cell [χ2=145.71, p<0.001] [table 9.3]. There was also an association between tumour type and receipt of surgery. In those diagnosed with an adenocarcinoma 22.7% received surgery, compared to 12.6% with large cell, and 17.6% with squamous cell tumours [table 9.4]. However, the highest rates of surgery were found in the ‘other-specified carcinomas’ group which encompassed the 3.9% of the dataset that had 54 different tumour types specified. Table 9.4. Tumour type and receipt of treatment (1999-2010) Surgery Chemotherapy Radiotherapy Total N % N % N % Adenocarcinoma 2,148 22.7 2,708 28.6 3,154 33.3 9,463 Squamous 2,437 17.6 3,047 22.0 6,706 48.5 13,831 Large 270 12.6 539 25.2 948 44.3 2,140 Non-small 390 4.5 2,654 30.6 4,175 48.2 8,661 Small 117 1.4 5,783 67.3 3,387 39.4 8,599 Other specified 956 38.1 726 28.9 574 22.9 2,511 Unspecified carcinoma 67 1.5 271 5.9 1,261 27.3 4,615 Neoplasm 22 0.1 268 1.7 2,008 13.1 15,390 Total 6,407 9.8 15,996 24.5 22,213 34.1 65,210 Of the 8885 patients diagnosed in 2006-2010 who had PS recorded, 26.6% in the lowest SEP group had poor performance status [PS 3-4] compared to 20.6% in the highest SEP group, whereas 26.3% in the highest SEP group had good PS [PS=0] compared to 17.7% in the lowest group [table 9.3]. These differences were statistically significant [χ2= 69.02, p<0.001]. Of the 8475 patients who had a CCM score recorded in 2006-2010, 11.6% in the lowest SEP group had three or more co-morbidities recorded compared to 8.9% in the highest SEP group, whereas 54.4% in the highest SEP group had 0 co-morbidities recorded compared to 43.3% in the lowest SEP group [table 9.3]. These differences were statistically significant (χ2=63.44, p<0.001). Similar results were seen for the 10,511 patients who had a CCM score recorded in 1999-2005 [χ2=23.73, p=0.003]. 130 SEP was significantly associated with having a GP referral date [χ2=20.62, p<0.001 for 1999-2005 cohort and χ2=11.998, p=0.02 for 2006-2010 cohort] although contrasting patterns of results were seen in the two cohorts. In 1999-2005 those in the lowest SEP group had lower rates of GP referral date recorded compared to those in the highest but for 2006-2010, in contrast to what might have been expected, those in the lowest SEP group had a higher rate of GP referral date recorded than those in the highest SEP group. 9.3.4 Stage, PS and co-morbidity In the 2006-2010 dataset 7126 patients had stage and PS recorded. A significant association between stage and PS was found [χ2 = 436.95, p<0.001] with high levels of good PS in those with early stage cancer compared to late stage lung cancer [table 9.5]. Only 2078 patients had both stage and co-morbidity score recorded. A slightly higher rate of recorded co-morbidity was found in those who had early stage cancer [of those with stage 1 cancer 56.8% had co-morbidity compared with 49.4% for those with stage 4 cancer] but no significant association between stage and co-morbidity was found [χ2 = 10.17, p=0.12] [table 9.5]. Table 9.5. Percentage rates of PS and co-morbidity by stage % PS (n=7126) Co-morbidity Score (n=2078) Stage 0 1-2 3-4 0 1-2 3+ 1 35.9 51.0 13.1 43.2 43.7 13.1 2 29.0 54.5 16.5 41.8 48.4 9.8 3 22.7 59.0 18.3 48.3 40.7 11.0 4 14.1 52.4 33.5 50.5 39.4 10.0 Both co-morbidity and PS can be used as surrogate measures of suitability for lung cancer treatment. The number of co-morbidities was significantly associated with PS score, where those with good PS had lower rates of high co-morbidity [χ2 =78.53, p<0.001]. Only 2434 patients had both a PS and CCM score recorded and 63.0% of those who had a PS score of 0 [good health] also had no co-morbidities, 31.5% had 1-2 co-morbidities and 5.5% had 3+ co-morbidities. However, 36.3% of those with a PS score of 3-4 [poor health] also had no co-morbidities recorded [table 9.6]. 131 Table 9.6. Percentage rates for number of co-morbidities by performance status score % Co-morbidity Score Performance Status 0 1-2 3+ 0 (good PS) 63.0 31.5 5.5 1-2 48.6 41.1 10.4 3-4 (poor PS) 36.3 48.6 15.2 9.4 Discussion As these are descriptive results only, a general discussion is presented here. Implications for policy and practice and further research will be discussed in the next four chapters when the multivariable results are presented. When the demographic and clinical characteristics of the patients in the different cohorts were compared, differences were found. The 2006-2010 cohort was significantly different from the 1999-2005 cohort on all variables examined except for sex. Rates of treatment were significantly higher in the 2006-2010 sub-group that had stage recorded and it had a higher percentage of younger patients and NSCLC patients, compared to the full 2006-2010 cohort. The population distribution by SEP and sex was not significantly different however. The staged cohort does therefore appear to be younger and more likely to undergo treatment. Over a third of the northern lung cancer population were in the most deprived SEP quintile. National deprivation quintiles are used by cancer registries so that, across England, the definition of most and least deprived is consistent (244). Quintiles are based on population so that each quintile has an equal number of residents. This does mean that, in areas with deprivation higher than the national average [such as NYCRIS] there may be low numbers of people within the population in the least deprived quintiles. This could be considered a limitation when using regional data. Unfortunately the population data for each quintile was not available, to examine any skew in population distribution in regional compared to national data. However as there were data for over 65,000 patients available, the numbers in all quintiles were large. In order to determine patient delay [time between onset of symptoms and initial presentation to GP], data should ideally be obtained from patients or from GP records. 132 Patient records were not available for this study and as a date is not recorded in the Registry dataset for the onset of symptoms it was not possible to look at patient delay by SEP unless a proxy measure such as stage of disease at presentation or mode of presentation [elective/emergency] was employed. Death within one year of diagnosis can be used as a proxy measure of patient delay for some cancers but would not be suitable for lung cancer which has short expected survival. Mode of presentation was requested but was not available from NYCRIS [due to problems with its calculation] but stage at diagnosis was available, as was the number of patients recorded as DCO. It has previously been suggested that DCO registrations are more common in deprived communities (143) but a study using NYCRIS data did not find this, using breast and colorectal cancer data (247). In this study a socio-economic gradient in DCO cases by SEP was found in the 1999-2005 cohort, with DCO rates nearly double in the most deprived quintile compared to the least deprived. However this was not found in the 2006-2010 cohort. Socio-economic inequalities in stage at diagnosis have been noted for some cancers which might suggest that there are differences in timeliness of patient presentation by SEP (66), where stage is used as a marker of late presentation and thus as a proxy measure of patient delay. However, previous studies examining socio-economic inequalities in stage at diagnosis for lung cancer did not find any significant differences by SEP (66, 68) suggesting that timeliness of patient presentation to the GP with lung cancer does not differ by SEP. We also found no evidence of socio-economic inequalities in stage at diagnosis. Results using DCO as a proxy measure of patient delay suggest that there were inequalities in time to patient presentation in the early years but unfortunately stage at diagnosis was not available for the 1999-2005 cohort to confirm this. However, using both stage at diagnosis and DCO, no evidence for inequalities in delay in patient presentation by SEP was found for the 2006-2010 cohort, although these were univariable analyses. A non-significantly higher rate of co-morbidity was found in those who had early stage compared to late stage cancer. Other studies have found that patients with severe co-morbidity were likely to have early-stage lung cancer (157), suggesting that those who suffer from a number of different medical conditions may have greater contact with 133 health professionals and so are more likely to have their cancer detected at an earlier stage. A significant association between stage and PS was found but none was found between stage and CCM score. Although co-morbidity and PS measure different things [number of concurrent health conditions and general health status respectively], both can be used as surrogate measures of suitability for lung cancer treatment. Those with good PS were less likely to have severe co-morbidity [65.7% of those who had a PS score of 0 had 0 co-morbidities]. However 41.9% of those with a PS score of 3-4 [poor health] also had 0 co-morbidities recorded. There is therefore an association between the two measures but, as can be seen here, a large percentage of those who have poor health do not have any co-morbidities recorded. Descriptive statistics for the univariable associations between SEP and receipt of treatment, time to referral, diagnosis and treatment, and survival are described in chapters 10-12. 9.5 Chapter summary Associations between SEP and a number of variables of interest were examined. SEP was univariably associated with age at diagnosis, tumour type, PS, CCM score and GP referral, but not with stage at diagnosis. An association between tumour type and receipt of treatment was also found. The demographic and clinical characteristics of the patients in the entire dataset and in the 1999-2005 cohort, 2006-2010 cohort, and 2006-2010 cohort with stage recorded, were examined and compared. Significant differences were found between cohorts. The implications of these differences on the representativeness of the cohorts and any subsequent differences in outcome observed between cohorts will be considered in more detail in the following chapters. Chapter 10 now goes on to examine socio-economic inequalities in lung cancer treatment. 134 Chapter 10. Socio-economic inequalities in receipt of lung cancer treatment Summary Background Socio-economic inequalities in receipt of lung cancer treatment have been demonstrated in my systematic review and meta-analysis, in both universal [UHCS] and non-universal healthcare systems. These findings could not be explained by type of healthcare system or stage at diagnosis. However, not all of the included studies reported details of stage and histology, both of which influence treatment type, and very few UHCS studies took co-morbidity or performance status into account. The review recommended that the reasons for socio-economic inequalities in treatment should be more thoroughly investigated. Cancer registry [NYCRIS], Hospital Episode Statistics [HES] and lung cancer audit [LUCADA] data-sets from the North of England were linked in order to examine the influence of stage, histology, performance status and co-morbidity on socio-economic inequalities in lung cancer treatment. Methods Multivariable logistic regression was used to examine the likelihood of receipt of surgery, chemotherapy and radiotherapy by socio-economic position [SEP] in patients diagnosed between 1999-2005 [n=36,477] of whom 32,974 had a linked HES record, and in patients diagnosed between 2006-2010 [n=28,733], and in a subset of whom had stage [n=7769] and PS recorded in LUCADA. Results (for 2006-2010 cohort) Socio-economic inequalities in receipt of surgery and chemotherapy, but not radiotherapy, were found after control for age, sex, histology, PS and co-morbidity. The odds of receiving surgery were significantly lower in the lowest compared to the highest SEP group in the full cohort [OR=0.75, CI 0.65 to 0.86, p<0.001]. Inequalities in receipt of surgery were substantially attenuated by adjustment for tumour type [OR=0.83, CI 0.71 to 0.96, p=0.01]. 135 Patients in the lowest SEP group were significantly less likely to receive chemotherapy in the full cohort [OR= 0.60, CI 0.54 to 0.67, p<0.001], but not in the stage-subset when PS was included in the model [OR= 0.84, CI 0.68 to 1.02, p=0.08]. Conclusions Socio-economic inequalities in performance status statistically explain socio-economic inequalities in receipt of chemotherapy in the selective subset of patients whose cancer was staged but not in the full cohort. Socio-economic inequalities in receipt of surgery cannot be statistically explained by inequalities in stage, PS or co-morbidity. However, socio-economic inequalities in tumour type account for some of the inequalities in surgery by SEP. Patients in lower SEP groups are more likely to be diagnosed with squamous cell cancer [a tumour type strongly associated with smoking] and are less likely to receive surgery than patients in higher SEP groups, who are more likely to be diagnosed with adenocarcinoma. 136 10.1 Introduction Socio-economic inequalities in receipt of lung cancer treatment were demonstrated in my systematic review and meta-analysis (236), in both universal [UHCS] and non-universal healthcare systems [chapter 6]. These findings could not be explained by the type of healthcare system or by socio-economic inequalities in stage at diagnosis. However, not all of the included UK studies reported details of stage and histology, both of which influence treatment type (236), and very few UHCS studies took co-morbidity into account. Performance status [PS], a measure of patient well-being, is also a factor that has not been previously well explored. The systematic review recommended that the reasons for socio-economic inequalities in treatment should be more thoroughly investigated in better quality UHCS studies, including statistical control for co-morbidity, stage and histology (236). Although co-morbidity and PS measure different things [the number of concurrent health conditions and general health status respectively] both variables may be used as surrogate measures of suitability for lung cancer treatment when patients are assessed. It is unclear how well co-morbidity and PS capture this ‘fitness for treatment’ concept but, as the number of co-morbidities and PS vary by SEP, this may help explain why inequalities are found. I also wanted to explore whether differences in rates of histological subtype by SEP might be important as, in the descriptive analyses [chapter 9], SEP was associated with tumour type, and different rates of treatment were also found by tumour type. Cancer registry, Hospital Episode Statistics [HES] and lung cancer audit [LUCADA] data-sets were linked in order to examine the role that socio-economic inequalities in stage, histological subtype, PS and co-morbidity might play in contributing to socio-economic inequalities in lung cancer treatment in the North-East of England. 10.2 Methods 10.2.1 Data For details of data sources and variables see chapter 8, sections 8.2 and 8.4. 137 10.2.2 Regression analyses Multivariable logistic regression was used to examine the likelihood of receipt of surgery, chemotherapy and radiotherapy by SEP, in the 1999-2005 and 2006-2010 cohorts. The analyses were conducted as follows: in the 1999-2005 dataset controlling for age, sex, histology [or histological subtype], co-morbidity, year of diagnosis and GP referral; in the 2006-2010 dataset and, within this, on the subgroup that had stage recorded [n=7769], controlling for age, sex, histology [or histological subtype], year of diagnosis, GP referral, PS and stage, and including and excluding co-morbidity. Interaction between SEP and histology was also explored. The R2 statistic was examined to determine the amount of variance in receipt of treatment explained by each model. Odds ratios [ORs] with 95% confidence intervals [CIs] for the likelihood of receipt of treatment in the lower SEP groups compared to the highest SEP group were reported. A likelihood ratio test was performed to determine the overall significance of each categorical variable. Cancer type [histology] was classified as confirmed non-small cell lung cancer [NSCLC], SCLC and other histology as previously described. Sub group analyses including only morphologically-specified lung cancers were also carried out. The ‘unspecified carcinoma/neoplasm’ subtypes were excluded and the ‘other specified carcinoma’ subtype was included as probable NSCLC. Receipt of surgery was examined for probable NSCLC-only patients. Receipt of chemotherapy and radiotherapy were examined separately in probable NSCLC and SCLC populations. 10.3 Results Table 9.1 shows the demographic and clinical characteristics of the lung cancer patients included in the study. 10.3.1 Surgery Socio-economic inequalities in receipt of surgery were found. When age, sex, histology or tumour type, year of diagnosis, co-morbidity, performance status, stage and GP referral were taken into account, socio-economic inequalities in receipt of treatment remained. 138 10.3.1.1 1999-2005 The rate of surgery in the highest SEP group was 11.7% compared to 8.5% in the lowest SEP group, with a univariable OR for receipt of surgery of 0.70, 95% CI 0.62 to 0.78, p<0.001 in the lowest compared to the highest. The odds of surgery were significantly lower in the lowest compared to the highest SEP group [OR=0.64, 95% CI 0.57 to 0.73, p<0.001] in a multivariable analysis including age, sex, histology, co-morbidity and GP referral [table 10.1]. When histology was further broken down into tumour type then the SEP OR was attenuated [OR=0.70, 95% CI 0.61 to 0.79, p<0.001] and the treatment variance explained by the model greatly improved [from R2=13.33 to 22.91] [table 10.1]. 10.3.1.2 2006-2010 The rate of surgery in the highest SEP group was 11.8% compared to 9.5% in the lowest SEP group, with a univariable OR for receipt of surgery of 0.78 [95% CI 0.69 to 0.89, p<0.001] in the lowest compared to the highest. The odds of surgery were significantly lower in the lowest compared to the highest SEP group in the multivariable analysis when stage and PS [but not CCM] were included [OR=0.75, 95% CI 0.65 to 0.86, p<0.001] [table 10.2]. The addition of co-morbidity to the model made no substantial change to this OR. Significant interaction between SEP and histology was found. Similarly to the 1999-2005 cohort, when histology was further broken down into tumour type then the SEP OR was attenuated [OR=0.83, 95% CI 0.71 to 0.96, p=0.0013] and the treatment variance explained by the model greatly improved [from R2=24.04 to 35.17]. Older patients were significantly less likely to receive surgery [OR=0.14, 95% CI 0.12 to 0.17, p<0.001], for those aged 80+, and [OR =0.65, 95% CI 0.57 to 0.73, p<0.001 for those aged 70-79], compared to the youngest group [aged <60] in a multivariable model, but when histology was further broken down into tumour type then the OR was attenuated [to OR=0.22, 95% CI 0.18 to 0.27, p<0.001, and OR=0.79, 95% CI 0.70 to 0.91, p=0.001, respectively] [table 10.2]. 56% of those aged 80+ had their tumour type recorded as a neoplasm compared to 8% in the youngest age group. Those with late stage cancer and those with poor performance status were also significantly less 139 likely to receive surgery. However, even when taking all these factors into account, age and socio-economic inequalities in receipt of surgery were still found. 10.3.1.2.1 2006-2010: subgroup with stage In the subgroup that had stage recorded [n=7769], although overall 12.9% of these patients received surgery, 51.2% of stage I patients received surgery, with 35.3% in stage II, so that receipt of surgery was highly dependent on stage. The odds of receiving treatment in the lowest SEP group compared to the highest were significantly lower in the univariable [OR=0.67, 95% CI 0.54 to 0.82, p=0.001] and adjusted analysis [OR=0.60, 95% CI 0.45 to 0.81, p=0.001] [table 10.3]. 10.3.1.2.2 2006-2010: probable NSCLC only When looking at receipt of surgery for those with probable NSCLC only [n=16,278] socio-economic inequalities in receipt of treatment were found in the multivariable analysis [OR=0.84, 95% CI 0.72 to 0.98, p=0.025], in the lowest compared to the highest SEP group [table 10.4]. 140 Table 10.1. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors for those diagnosed between 1999 and 2005 (DCO cases excluded) Variable Receipt of Surgery Adjusted – selected1 (n=36,477, R2=13.33) Adjusted – selected2 (n=36,477, R2=22.91) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 3,513 9.6 <0.001 <0.001 1 (least deprived) 452 11.7 1.00 1.00 2 565 11.1 0.92 0.80 1.06 0.25 0.95 0.83 1.10 0.53 3 592 9.9 0.81 0.71 0.93 0.002 0.85 0.73 0.97 0.02 4 794 9.4 0.77 0.67 0.87 <0.001 0.81 0.71 0.93 <0.001 5 (most deprived) 1,110 8.5 0.64 0.57 0.73 <0.001 0.70 0.61 0.79 <0.001 Sex 3,513 9.6 0.009 0.05 Female 1,459 9.5 1.00 1.00 Male 2,054 9.8 0.91 0.84 0.98 0.009 0.92 0.85 1.00 0.05 Age group 3,513 9.6 <0.001 <0.001 <60 852 16.3 1.00 1.00 60-69 1,296 13.6 0.82 0.75 0.91 <0.001 0.91 0.82 1.00 0.06 70-79 1,248 8.6 0.50 0.45 0.55 <0.001 0.60 0.54 0.66 <0.001 80+ 117 1.6 0.10 0.08 0.13 <0.001 0.15 0.12 0.18 <0.001 Histology 3,513 9.6 <0.001 NSCLC 2,897 15.3 1.00 SCLC 63 1.2 0.06 0.05 0.08 <0.001 Other 553 4.5 0.38 0.35 0.42 <0.001 Co-morbidity score 3,513 9.6 <0.001 <0.001 0 740 12.2 1.00 1.00 1-2 318 8.5 0.75 0.65 0.86 <0.001 0.82 0.70 0.95 0.01 3+ 58 8.2 0.79 0.59 1.05 0.11 0.92 0.68 1.24 0.57 CCM missing 2,351 10.5 0.73 0.67 0.80 <0.001 0.78 0.71 0.86 <0.001 No HES link 46 1.3 0.13 0.10 0.18 <0.001 0.18 0.13 0.24 <0.001 Year of Diagnosis 3,513 9.6 0.46 0.07 1999 494 9.6 1.00 1.00 2000 508 9.7 1.04 0.90 1.19 0.61 1.07 0.93 1.23 0.33 2001 489 9.4 0.95 0.83 1.09 0.45 1.10 0.95 1.26 0.21 2002 513 9.8 0.96 0.84 1.10 0.57 1.14 0.99 1.32 0.06 2003 505 9.7 0.95 0.83 1.09 0.46 1.16 1.01 1.34 0.04 2004 480 9.2 0.89 0.77 1.02 0.09 1.06 0.92 1.23 0.39 2005 524 10.0 0.97 0.84 1.11 0.63 1.25 1.08 1.43 0.002 GP referral 3,513 9.6 <0.001 <0.001 No 1,560 7.9 1.00 1.00 Yes 1,953 11.7 1.29 1.20 1.39 <0.001 1.20 1.11 1.30 <0.001 Histological sub-type 3,513 9.6 <0.001 Adenocarcinoma 1,150 23.0 1.00 squamous 1,426 16.6 0.73 0.66 0.80 <0.001 Large-cell 146 10.6 0.41 0.34 0.49 <0.001 non-small cell 175 4.4 0.15 0.13 0.18 <0.001 Other specified 486 35.8 1.81 1.59 2.07 <0.001 Small cell 63 1.2 0.04 0.03 0.05 <0.001 Unspecified carcinoma 53 1.3 0.06 0.05 0.08 <0.001 Neoplasm 14 0.2 0.01 0.01 0.02 <0.001 1 Mutually adjusted for SEP, sex, age, histology, co-morbidity score, year of diagnosis and GP referral 2 Mutually adjusted for SEP, sex, age, co-morbidity score, year of diagnosis, GP referral and histological subtype 141 Table 10.2. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort Variable Receipt of surgery (2894/28733) Adjusted – selected1 (n=28,733, R2=24.04) Adjusted – selected2 (n=28,733, R2=35.17) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 2,894 10.1 <0.001 0.01 1 (least deprived) 400 11.8 1.00 1.00 2 458 11.0 0.92 0.78 1.08 0.32 0.92 0.77 1.09 0.33 3 512 10.6 0.91 0.78 1.07 0.26 0.97 0.82 1.15 0.75 4 613 9.1 0.75 0.65 0.87 <0.001 0.81 0.69 0.95 0.009 5 (most deprived) 911 9.5 0.75 0.65 0.86 <0.001 0.83 0.71 0.96 0.013 Sex 2,894 10.1 <0.001 <0.001 Female 1,405 10.6 1.00 1.00 Male 1,489 9.6 0.80 0.73 0.87 <0.001 0.78 0.71 0.86 <0.001 Age group 2,894 10.1 <0.001 <0.001 <60 560 15.2 1.00 1.00 60-69 1,118 14.7 0.96 0.85 1.08 0.50 1.05 0.92 1.19 0.47 70-79 1,048 10.2 0.65 0.57 0.73 <0.001 0.79 0.70 0.91 0.001 80+ 168 2.3 0.14 0.12 0.17 <0.001 0.22 0.18 0.27 <0.001 Histology 2,894 10.1 <0.001 NSCLC 2,348 15.5 1.00 SCLC 54 1.6 0.08 0.06 0.11 <0.001 Other 492 4.9 0.43 0.38 0.48 <0.001 Year of Diagnosis 2,894 10.1 <0.001 <0.001 2006 505 9.1 1.00 1.00 2007 547 9.6 1.04 0.91 1.19 0.58 1.10 0.95 1.27 0.19 2008 502 8.6 0.95 0.82 1.09 0.47 0.96 0.82 1.11 0.58 2009 682 11.6 1.33 1.16 1.52 <0.001 1.33 1.15 1.53 <0.001 2010 658 11.4 1.38 1.20 1.59 <0.001 1.25 1.08 1.45 0.003 Stage 2,894 10.1 <0.001 <0.001 I 607 51.2 1.00 1.00 II 195 35.3 0.41 0.32 0.53 <0.001 0.43 0.33 0.56 <0.001 III 144 6.3 0.04 0.03 0.05 <0.001 0.04 0.03 0.06 <0.001 IV 55 1.5 0.01 0.01 0.02 <0.001 0.01 0.01 0.02 <0.001 Missing 1,893 9.0 0.12 0.10 0.14 <0.001 0.11 0.09 0.14 <0.001 Performance Status 2,894 10.1 <0.001 <0.001 0 583 31.7 1.00 1.00 1-2 480 9.9 0.30 0.25 0.36 <0.001 0.33 0.28 0.40 <0.001 3-4 15 0.7 0.03 0.02 0.05 <0.001 0.05 0.03 0.09 <0.001 Missing/ unknown 1,816 9.2 0.34 0.28 0.41 <0.001 0.41 0.34 0.51 <0.001 GP referral 2,894 10.1 <0.001 <0.001 No 979 7.4 1.00 1.00 Yes 1,915 12.4 1.37 1.25 1.50 <0.001 1.26 1.14 1.39 <0.001 Histological subtype 2,894 10.1 <0.001 Adenocarcinoma 998 22.4 1.00 squamous 1,011 19.3 0.86 0.77 0.97 0.01 Large cell 124 16.2 0.64 0.51 0.80 <0.001 Non-small 215 4.6 0.19 0.16 0.22 <0.001 Other specified 470 40.7 2.40 2.06 2.80 <0.001 Small 54 1.6 0.06 0.04 0.07 <0.001 Unspecified carcinoma 14 2.7 0.14 0.08 0.24 <0.001 Neoplasm 8 0.1 0.01 0.00 0.01 <0.001 1 Mutually adjusted for SEP, sex, age, histology, year of diagnosis, stage, PS and GP referral 2 Mutually adjusted for SEP, sex, age, year of diagnosis, stage, PS, GP referral and histological subtype 142 Table 10.3. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort with stage recorded Variable Receiving surgery (1001/7769) Adjusted – selected1 (n=7769, R2=46.99) Adjusted – selected2 (n=7676, R2=53.61) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 1,001 12.9 0.0007 0.003 1 (least deprived) 156 16.8 1.00 1.00 2 162 14.5 0.79 0.56 1.10 0.16 0.72 0.51 1.04 0.08 3 177 13.6 0.76 0.55 1.05 0.09 0.77 0.54 1.08 0.13 4 199 10.9 0.54 0.40 0.74 <0.001 0.53 0.38 0.75 <0.001 5 (most deprived) 307 11.9 0.60 0.45 0.81 0.001 0.61 0.44 0.83 0.002 Sex 1,001 12.9 0.02 0.12 Female 485 13.6 1.00 1.00 Male 516 12.3 0.81 0.67 0.97 0.02 0.85 0.70 1.04 0.12 Age group 1,001 12.9 <0.001 <0.001 <60 179 17.2 1.00 1.00 60-69 398 18.2 1.00 0.76 1.32 0.99 1.13 0.85 1.51 0.41 70-79 353 12.4 0.56 0.43 0.75 <0.001 0.66 0.49 0.90 0.007 80+ 71 4.2 0.15 0.11 0.22 <0.001 0.19 0.13 0.28 <0.001 Histology 1,001 12.9 <0.001 NSCLC 852 16.7 1.00 SCLC 18 3.1 0.25 0.14 0.43 <0.001 Other 131 6.3 0.48 0.37 0.62 <0.001 Year of Diagnosis 1,001 12.9 <0.001 0.0001 2006 84 12.5 1.00 1.00 2007 116 13.4 1.14 0.77 1.69 0.52 1.09 0.71 1.66 0.70 2008 162 10.4 1.22 0.84 1.78 0.29 1.11 0.75 1.66 0.60 2009 311 14.5 1.97 1.39 2.78 <0.001 1.81 1.25 2.62 0.002 2010 328 12.9 2.13 1.51 3.01 <0.001 1.74 1.20 2.52 0.003 Stage 1,001 12.9 <0.001 <0.001 I 607 51.2 1.00 1.00 II 195 35.3 0.42 0.32 0.53 <0.001 0.43 0.33 0.56 <0.001 III 144 6.3 0.04 0.03 0.05 <0.001 0.04 0.03 0.05 <0.001 IV 55 1.5 0.01 0.01 0.01 <0.001 0.01 0.01 0.01 <0.001 Performance Status 1,001 12.9 <0.001 <0.001 0 498 33.4 1.00 1.00 1-2 413 10.7 0.31 0.26 0.38 <0.001 0.36 0.29 0.45 <0.001 3-4 12 0.7 0.03 0.02 0.06 <0.001 0.06 0.03 0.11 <0.001 Missing/ unknown 78 12.1 0.33 0.23 0.46 <0.001 0.36 0.25 0.52 <0.001 GP referral 1,001 12.9 0.29 0.85 No 243 10.1 1.00 1.00 Yes 758 14.2 1.12 0.91 1.39 0.29 1.02 0.81 1.28 0.85 Histological subtype 1,001 12.9 <0.001 Adenocarcinoma 372 25.3 1.00 squamous 372 20.1 0.67 0.52 0.85 0.001 Large cell 32 18.9 0.55 0.31 0.98 0.04 Non-small 76 4.7 0.19 0.13 0.26 <0.001 Other specified 130 43.6 2.12 1.44 3.12 <0.001 Small 18 3.1 0.15 0.09 0.27 <0.001 Unspecified carcinoma 0 0.0 -- Neoplasm 1 0.1 0.003 0.00 0.02 <0.001 1 Mutually adjusted for SEP, sex, age, histology, year of diagnosis, stage, PS and GP referral 2 Mutually adjusted for SEP, sex, age, year of diagnosis, stage, PS, GP referral and histological subtype 143 Table 10.4. OR of receipt of lung cancer surgery, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort with probable NSCLC Variable Probable NSCLC Receiving surgery Unadjusted Adjusted – selected1 (n=16,278, R2=23.74) N N % OR (95% CI) P OR (95% CI) P Deprivation quintile 16,278 2,818 17.3 0.001 0.02 1 (least deprived) 2,024 389 19.2 1.00 1.00 2 2,430 451 18.6 0.96 0.82 1.11 0.58 0.93 0.78 1.11 0.43 3 2,696 494 18.3 0.94 0.81 1.09 0.44 0.96 0.81 1.14 0.68 4 3,768 592 15.7 0.78 0.68 0.90 0.001 0.80 0.68 0.95 0.009 5 (most deprived) 5,360 892 16.6 0.84 0.74 0.96 0.009 0.84 0.72 0.98 0.03 Sex 16,278 2,818 17.3 <0.001 <0.001 Female 7060 1359 19.3 1.00 1.00 Male 9218 1459 15.8 0.79 0.73 0.86 <0.001 0.79 0.72 0.87 <0.001 Age group 16,278 2,818 17.3 <0.001 <0.001 <60 2679 544 20.3 1.00 1.00 60-69 5126 1085 21.2 1.05 0.94 1.18 0.37 1.05 0.92 1.20 0.50 70-79 5293 1025 17.3 0.82 0.73 0.92 0.001 0.80 0.70 0.92 0.001 80+ 2550 164 6.4 0.27 0.22 0.32 <0.001 0.22 0.18 0.27 <0.001 Year of Diagnosis 16,278 2,818 17.3 <0.001 <0.001 2006 3141 489 15.6 1.00 1.00 2007 3269 530 16.2 1.05 0.92 1.20 0.48 1.10 0.95 1.27 0.21 2008 3285 493 15.0 0.96 0.84 1.10 0.53 0.97 0.83 1.13 0.71 2009 3358 668 19.9 1.35 1.18 1.53 <0.001 1.34 1.16 1.55 <0.001 2010 3225 638 19.8 1.34 1.17 1.52 <0.001 1.24 1.07 1.44 0.005 Stage 16,278 2,818 17.3 <0.001 <0.001 I 917 595 64.9 1.00 1.00 II 437 192 43.9 0.42 0.34 0.54 <0.001 0.44 0.33 0.58 <0.001 III 1701 141 8.3 0.05 0.04 0.06 <0.001 0.05 0.04 0.06 <0.001 IV 2359 54 2.3 0.01 0.01 0.02 <0.001 0.01 0.01 0.02 <0.001 Missing 10,864 1836 16.9 0.11 0.10 0.13 <0.001 0.12 0.10 0.16 <0.001 Performance Status 16,278 2,818 17.3 <0.001 <0.001 0 1528 574 37.6 1.00 1.00 1-2 3467 470 13.6 0.26 0.23 0.30 <0.001 0.33 0.28 0.40 <0.001 3-4 835 13 1.6 0.03 0.02 0.05 <0.001 0.04 0.02 0.08 <0.001 Missing/ unknown 10,448 1761 16.9 0.34 0.30 0.38 <0.001 0.39 0.32 0.48 <0.001 GP referral 16,278 2,818 17.3 <0.001 <0.001 No 6304 946 15.0 1.00 1.00 Yes 9974 1872 18.8 1.31 1.20 1.43 <0.001 1.26 1.15 1.40 <0.001 Histological sub-type 16,278 2,818 17.3 <0.001 <0.001 Adenocarcinoma 4462 998 22.4 1.00 1.00 squamous 5229 1011 19.3 0.83 0.75 0.92 <0.001 0.86 0.77 0.96 0.009 Large cell 768 124 16.2 0.67 0.54 0.82 <0.001 0.64 0.51 0.80 <0.001 Non-small 4664 215 4.6 0.17 0.14 0.20 <0.001 0.19 0.16 0.22 <0.001 Other specified 1155 470 40.7 2.38 2.08 2.73 <0.001 2.39 2.05 2.79 <0.001 1Mutually adjusted for SEP, sex, age, year of diagnosis, stage, PS, GP referral and histological subtype 144 10.3.2 Chemotherapy 10.3.2.1 1999-2005 In the adjusted model SEP was strongly associated with receipt of chemotherapy. Those in the lowest SEP group were significantly less likely to receive chemotherapy [OR=0.60, 95% CI 0.54 to 0.67, p<0.001] [table 10.5]. Older patients were significantly less likely to receive chemotherapy, as were those with co-morbidities. Rates of chemotherapy increased over time with 15.2% receiving chemotherapy in 1999, increasing to 28.5% in 2005. Table 10.5. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors (DCO cases excluded) for those diagnosed between 1999 and 2005 Variable Receiving chemotherapy Unadjusted Adjusted – selected1 (n=36,477, R2=32.84) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 7,648 21.0 <0.001 <0.001 1 (least deprived) 961 24.9 1.00 1.00 2 1,108 21.8 0.84 0.76 0.93 0.001 0.75 0.67 0.85 <0.001 3 1,196 19.9 0.75 0.68 0.83 <0.001 0.67 0.59 0.76 <0.001 4 1,720 20.4 0.77 0.70 0.84 <0.001 0.69 0.61 0.77 <0.001 5 (most deprived) 2,663 20.4 0.77 0.71 0.84 <0.001 0.60 0.54 0.67 <0.001 Sex 7,648 21.0 0.001 0.39 Female 3,356 21.8 1.00 1.00 Male 4,292 20.4 0.92 0.87 0.97 0.001 0.97 0.91 1.04 0.39 Age group 7,648 21.0 <0.001 <0.001 <60 2393 45.8 1.00 1.00 60-69 2,945 30.9 0.53 0.49 0.57 <0.001 0.48 0.44 0.52 <0.001 70-79 2,074 14.2 0.20 0.18 0.21 <0.001 0.18 0.17 0.20 <0.001 80+ 236 3.3 0.04 0.04 0.05 <0.001 0.04 0.04 0.05 <0.001 Histology 7,648 21.0 <0.001 <0.001 NSCLC 3,582 18.9 1.00 1.00 SCLC 3,402 66.7 8.59 8.02 9.20 <0.001 12.87 11.86 13.96 <0.001 Other 664 5.4 0.24 0.22 0.26 <0.001 0.41 0.37 0.45 <0.001 Co-morbidity score 7,648 21.0 <0.001 <0.001 0 1,302 21.5 1.00 1.00 1-2 556 14.8 0.63 0.57 0.71 <0.001 0.75 0.66 0.85 <0.001 3-4 68 9.7 0.39 0.30 0.50 <0.001 0.47 0.35 0.63 <0.001 CCM missing 5,643 25.1 1.22 1.14 1.31 <0.001 1.06 0.97 1.15 0.21 No HES link 79 2.3 0.08 0.07 0.11 <0.001 0.14 0.11 0.18 <0.001 Year of Diagnosis 7,648 21.0 <0.001 <0.001 1999 784 15.2 1.00 1.00 2000 937 17.9 1.22 1.10 1.35 0.00 1.35 1.19 1.53 <0.001 2001 923 17.8 1.20 1.08 1.33 0.00 1.46 1.28 1.66 <0.001 2002 1,048 20.1 1.40 1.26 1.55 0.00 1.70 1.50 1.93 <0.001 2003 1,170 22.4 1.61 1.46 1.78 0.00 2.20 1.95 2.49 <0.001 2004 1,291 24.8 1.83 1.66 2.02 0.00 2.64 2.34 2.99 <0.001 2005 1,495 28.5 2.23 2.02 2.45 0.00 3.56 3.16 4.02 <0.001 GP referral 7,648 21.0 <0.001 <0.001 No 3,265 16.5 1.00 1.00 Yes 4,383 26.3 1.80 1.71 1.90 <0.001 1.86 1.74 1.98 <0.001 1Mutually adjusted for SEP, sex, age, histology, co-morbidity score, year of diagnosis and GP referral 145 10.3.2.2 2006-2010 SEP was associated with receipt of treatment in the 2006-2010 subset when age, sex, histology, year of diagnosis, GP referral, stage and PS were included in the model [OR=0.60, 95% CI 0.54 to 0.67, p<0.001] [table 10.6]. Table 10.6. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort Variable Receiving chemotherapy (8348/28733) Unadjusted Adjusted – selected1 (n=28,733, R2=33.01) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 8,348 29.1 <0.001 <0.001 1 (least deprived) 1,133 33.4 1.00 1.00 2 1,275 30.5 0.87 0.79 0.96 0.007 0.83 0.74 0.94 <0.001 3 1,425 29.4 0.83 0.75 0.91 <0.001 0.78 0.69 0.88 <0.001 4 1,898 28.3 0.79 0.72 0.86 <0.001 0.67 0.60 0.75 <0.001 5 (most deprived) 2,617 27.2 0.75 0.69 0.81 <0.001 0.60 0.54 0.67 <0.001 Sex 8,348 29.1 0.29 0.96 Female 3,810 28.8 1.00 1.00 Male 4,538 29.3 1.03 0.98 1.08 0.29 1.00 0.94 1.07 0.96 Age group 8,348 29.1 <0.001 <0.001 <60 2,221 60.3 1.00 1.00 60-69 3,332 43.9 0.51 0.47 0.56 <0.001 0.51 0.46 0.55 <0.001 70-79 2,452 23.9 0.21 0.19 0.22 <0.001 0.22 0.20 0.24 <0.001 80+ 343 4.8 0.03 0.03 0.04 <0.001 0.04 0.04 0.05 <0.001 Histology 8,348 29.1 <0.001 <0.001 NSCLC 5,366 35.5 1.00 1.00 SCLC 2,381 68.1 3.89 3.59 4.20 <0.001 5.63 5.12 6.18 <0.001 Other 601 5.9 0.11 0.11 0.13 <0.001 0.22 0.20 0.25 <0.001 Year of Diagnosis 8,348 29.1 0.005 0.03 2006 1,506 27.2 1.00 1.00 2007 1,667 29.2 1.10 1.02 1.20 0.02 1.17 1.06 1.30 <0.001 2008 1,681 28.7 1.08 0.99 1.17 0.07 1.06 0.96 1.17 0.27 2009 1,773 30.2 1.16 1.07 1.25 <0.001 1.11 1.00 1.23 0.05 2010 1,721 29.9 1.14 1.05 1.23 <0.001 1.11 1.00 1.23 0.05 Stage 8,348 29.1 <0.001 <0.001 I 223 18.8 1.00 1.00 II 163 29.5 1.81 1.43 2.29 <0.001 2.11 1.60 2.78 <0.001 III 1,032 45.4 3.59 3.04 4.25 <0.001 5.45 4.46 6.67 <0.001 IV 1,314 35.0 2.32 1.98 2.73 <0.001 4.30 3.54 5.23 <0.001 Missing 5,616 26.8 1.58 1.36 1.83 <0.001 3.46 2.83 4.24 <0.001 Performance Status 8,348 29.1 <0.001 <0.001 0 1,151 62.5 1.00 1.00 1-2 2,067 42.5 0.44 0.40 0.50 <0.001 0.46 0.40 0.53 <0.001 3-4 113 5.2 0.03 0.03 0.04 <0.001 0.04 0.03 0.05 <0.001 Missing/ unknown 5,017 25.3 0.20 0.18 0.22 <0.001 0.29 0.25 0.34 <0.001 GP referral date 8,348 29.1 <0.001 <0.001 No 2,566 19.3 1.00 1.00 Yes 5,782 37.4 2.50 2.37 2.64 <0.001 2.17 2.03 2.32 <0.001 1Mutually adjusted for SEP, sex, age, histology, year of diagnosis, stage, PS and GP referral 146 10.3.2.2.1 2006-2010: subset with stage recorded In the subset of patients who had stage recorded, in a multivariable analysis including age and sex, a significant association was seen [OR=0.72, 95% CI 0.61 to 0.86, P<0.001], and this was also observed with the stepwise addition of other variables; histology, year of diagnosis, and stage, with a similar OR in the lowest compared to the highest SEP group, in all cases [table 10.7]. However, on the addition of performance status to the model the OR was attenuated and this association was no longer observed [OR=0.84, 95% CI 0.68 to 1.02, p=0.08] [table 10.7]. Including performance status also greatly increased the treatment variance explained by the model [R2=25.84 without PS, 34.58 with]. An identical pattern was also seen in the probable NSCLC-only subset with stage recorded [results not shown]. Socio-economic differences in performance status appear to account for the observed socio-economic differences in receipt of chemotherapy in this staged subgroup. 10.3.2.2.2 2006-2010: probable NSCLC and SCLC separately When chemotherapy was examined separately in probable NSCLC and SCLC populations, socio-economic inequalities in receipt of chemotherapy were found for NSCLC [OR=0.66, 95% CI 0.58 to 0.74, p<0.001] [table 10.8] and SCLC [OR=0.57, 95% CI 0.42 to 0.75, p<0.001] [table 10.9]. For probable NSCLC likelihood of chemotherapy increased over time but this was not seen for SCLC. 147 Table 10.7. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort with stage recorded Variable Receiving chemotherapy (2732/7769) Adjusted – selected1 (n=7769, R2=25.84) Adjusted – selected2 (n=7769, R2=34.58) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 2,732 35.2 <0.001 0.16 1 (least deprived) 351 37.7 1.00 1.00 2 414 37.0 0.97 0.78 1.20 0.76 0.99 0.79 1.24 0.91 3 445 34.2 0.85 0.69 1.05 0.14 1.01 0.81 1.26 0.94 4 638 34.8 0.78 0.64 0.94 0.01 0.90 0.73 1.11 0.32 5 (most deprived) 884 34.1 0.68 0.56 0.82 <0.001 0.84 0.68 1.02 0.08 Sex 2,732 35.2 0.79 0.80 Female 1,255 35.3 1.00 1.00 Male 1,477 35.1 0.98 0.88 1.10 0.79 0.98 0.87 1.11 0.80 Age group 2,732 35.2 <0.001 <0.001 <60 690 66.3 1.00 1.00 60-69 1,103 50.4 0.48 0.40 0.56 <0.001 0.54 0.45 0.65 <0.001 70-79 821 28.9 0.19 0.16 0.23 <0.001 0.25 0.21 0.30 <0.001 80+ 118 7.0 0.04 0.03 0.05 <0.001 0.06 0.04 0.07 <0.001 Histology 2,732 35.2 <0.001 <0.001 NSCLC 2,140 41.8 1.00 1.00 SCLC 411 70.6 4.03 3.25 4.99 <0.001 6.21 4.85 7.95 <0.001 Other 181 8.7 0.20 0.17 0.24 <0.001 0.29 0.24 0.35 <0.001 Year of Diagnosis 2,732 35.2 0.21 0.07 2006 211 31.5 1.00 1.00 2007 323 37.3 1.36 1.06 1.74 0.02 1.44 1.11 1.88 0.01 2008 556 35.7 1.22 0.97 1.53 0.08 1.37 1.08 1.74 0.01 2009 758 35.4 1.21 0.97 1.51 0.08 1.29 1.02 1.62 0.03 2010 884 34.9 1.20 0.97 1.49 0.09 1.30 1.04 1.64 0.02 GP referral 2,732 35.2 <0.001 <0.001 No 603 24.9 1.00 1.00 Yes 2,129 39.8 2.15 1.90 2.45 <0.001 2.01 1.75 2.30 <0.001 Stage 2,732 35.2 <0.001 <0.001 I 223 18.8 1.00 1.00 II 163 29.5 1.83 1.41 2.39 <0.001 2.16 1.64 2.84 <0.001 III 1,032 45.4 3.94 3.26 4.77 <0.001 5.58 4.55 6.84 <0.001 IV 1,314 35.0 2.41 2.01 2.89 <0.001 4.43 3.64 5.40 <0.001 Performance Status 2,732 35.2 <0.001 0 933 62.5 1.00 1-2 1,561 40.3 0.39 0.34 0.46 <0.001 3-4 78 4.4 0.03 0.02 0.04 <0.001 Missing 160 24.9 0.25 0.19 0.31 <0.001 1Mutually adjusted for SEP, sex, age, histology, year of diagnosis, GP referral and stage 2Mutually adjusted for SEP, sex, age, histology, year of diagnosis, GP referral, stage and PS 148 Table 10.8. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for probable NSCLC Variable Unadjusted (n=16,278) Adjusted – selected1 (n=16,278, R2=18.26) OR (95% CI) P OR (95% CI) P Deprivation quintile <0.001 <0.001 1 (least deprived) 1.00 1.00 2 0.90 0.80 1.02 0.10 0.88 0.77 1.01 0.08 3 0.84 0.74 0.94 0.004 0.84 0.74 0.97 0.014 4 0.80 0.71 0.89 <0.001 0.73 0.64 0.83 <0.001 5 (most deprived) 0.76 0.68 0.84 <0.001 0.66 0.58 0.74 <0.001 Sex <0.001 0.97 Female 1.00 1.00 Male 0.96 0.90 1.02 0.18 1.00 0.93 1.07 0.97 Age group <0.001 <0.001 <60 1.00 1.00 60-69 0.55 0.50 0.61 <0.001 0.54 0.49 0.60 <0.001 70-79 0.24 0.22 0.27 <0.001 0.23 0.21 0.25 <0.001 80+ 0.05 0.04 0.06 <0.001 0.05 0.04 0.05 <0.001 Year of Diagnosis <0.001 0.007 2006 1.00 1.00 2007 1.23 1.10 1.36 <0.001 1.24 1.11 1.39 <0.001 2008 1.23 1.11 1.36 <0.001 1.14 1.01 1.28 0.03 2009 1.28 1.15 1.42 <0.001 1.13 1.00 1.27 0.04 2010 1.35 1.22 1.50 <0.001 1.17 1.04 1.32 0.01 Stage <0.001 <0.001 I 1.00 1.00 II 1.87 1.46 2.41 <0.001 2.22 1.67 2.94 <0.001 III 3.71 3.09 4.46 <0.001 6.02 4.88 7.43 <0.001 IV 2.80 2.35 3.34 <0.001 5.03 4.10 6.16 <0.001 missing 1.70 1.45 2.00 <0.001 3.30 2.66 4.10 <0.001 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 0.47 0.42 0.53 <0.001 0.48 0.42 0.56 <0.001 3-4 0.04 0.03 0.05 <0.001 0.03 0.02 0.05 <0.001 Missing/ unknown 0.31 0.28 0.34 <0.001 0.38 0.32 0.45 <0.001 GP referral <0.001 <0.001 No 1.00 1.00 Yes 2.08 1.94 2.23 <0.001 2.08 1.93 2.25 <0.001 1Mutually adjusted for SEP, sex, age, year of diagnosis, stage, PS and GP referral 149 Table 10.9. OR of receipt of lung cancer chemotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for SCLC Variable Unadjusted (n=3495) Adjusted – selected1 (n=3495, R2=15.83) OR (95% CI) P OR (95% CI) P Deprivation quintile 0.06 0.0006 1 (least deprived) 1.00 1.00 2 0.73 0.54 0.97 0.03 0.74 0.53 1.02 0.07 3 0.77 0.58 1.03 0.07 0.70 0.51 0.95 0.02 4 0.70 0.54 0.91 0.009 0.58 0.43 0.78 <0.001 5 (most deprived) 0.70 0.54 0.90 0.006 0.57 0.42 0.75 <0.001 Sex 0.87 0.56 Female 1.00 1.00 Male 1.01 0.88 1.17 0.87 0.95 0.81 1.12 0.56 Age group <0.001 <0.001 <60 1.00 1.00 60-69 0.58 0.45 0.75 <0.001 0.55 0.42 0.71 <0.001 70-79 0.32 0.25 0.41 <0.001 0.31 0.24 0.40 <0.001 80+ 0.10 0.07 0.13 <0.001 0.09 0.06 0.12 <0.001 Year of Diagnosis 0.39 0.13 2006 1.00 2007 1.02 0.81 1.28 0.86 1.01 0.79 1.30 0.91 2008 0.91 0.73 1.14 0.42 0.81 0.63 1.03 0.09 2009 1.12 0.89 1.40 0.34 1.00 0.77 1.29 0.98 2010 0.92 0.74 1.15 0.47 0.79 0.61 1.03 0.08 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 0.40 0.24 0.66 <0.001 0.50 0.30 0.84 0.01 3-4 0.04 0.02 0.07 <0.001 0.05 0.03 0.10 <0.001 Missing/ unknown 0.18 0.11 0.29 <0.001 0.22 0.13 0.36 <0.001 GP referral <0.001 <0.001 No 1.00 1.00 Yes 2.16 1.87 2.49 <0.001 2.24 1.91 2.64 <0.001 1Mutually adjusted for SEP, sex, age, year of diagnosis, PS and GP referral150 10.3.3 Radiotherapy 10.3.3.1 1999-2005 In the 1999 to 2005 data no association between SEP and receipt of radiotherapy was found in the unadjusted analysis or in the adjusted model [table 10.10] when co-morbidity was included [OR= 0.98, 95% CI 0.91 to 1.06, p=0.67]. 10.3.3.2 2006-2010 No association between SEP and receipt of radiotherapy was found in the 2006-2010 cohort [adjusted OR= 1.03, CI 0.94 to 1.13, p=0.48] [table 10.11]. Similarly, in the analysis of receipt of radiotherapy in the subset of patients who had stage recorded [n=7769], SEP was not associated with radiotherapy [OR=1.01, 95% CI 0.86 to 1.19, p=0.88] in the adjusted model including age, sex, year of diagnosis, histology, GP referral status, stage and PS [results not shown]. The addition of co-morbidity to the model did not change this. The rate of receipt of radiotherapy was higher [39.5%] in the staged subset compared to in the full cohort [33.5%]. 10.3.3.2.1 2006-2010: probable NSCLC and SCLC separately When radiotherapy was examined separately in probable NSCLC [n=16,278] and SCLC [n=3495] populations, socio-economic inequalities in receipt of radiotherapy were found for probable NSCLC [OR=1.18, 95% CI 1.06 to 1.31, p=0.003] [table 10.12] but not SCLC [OR=0.84, 95% CI 0.66 to 1.07, p=0.16] when comparing OR in the lowest SEP group to the highest [table 10.13]. 151 Table 10.10. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors (DCO cases excluded) for those diagnosed between 1999 and 2005 Variable Receiving radiotherapy (12,602/36,477) Unadjusted (n=36,477) Adjusted – selected1 (n=36,477, R2=7.23) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 12,602 34.6 0.65 0.65 1 (least deprived) 1,312 34.0 1.00 1.00 2 1,781 35.0 1.04 0.96 1.14 0.33 1.02 0.93 1.12 0.63 3 2,092 34.8 1.04 0.95 1.13 0.40 1.03 0.94 1.12 0.55 4 2,875 34.0 1.00 0.92 1.08 1.00 0.99 0.91 1.07 0.76 5 (most deprived) 4,542 34.7 1.03 0.96 1.11 0.41 0.98 0.91 1.06 0.67 Sex 12,602 34.6 <0.001 0.16 Female 5,092 33.0 1.00 1.00 Male 7,510 35.7 1.12 1.08 1.17 <0.001 1.03 0.99 1.08 0.16 Age group 12,602 34.6 <0.001 <0.001 <60 2,354 45.1 1.00 1.00 60-69 3,815 40.0 0.81 0.76 0.87 <0.001 0.83 0.78 0.90 <0.001 70-79 4,938 33.9 0.62 0.59 0.67 <0.001 0.71 0.66 0.76 <0.001 80+ 1,495 21.0 0.32 0.30 0.35 <0.001 0.46 0.43 0.50 <0.001 Histology 12,602 34.6 <0.001 <0.001 NSCLC 8,371 44.1 1.00 1.00 SCLC 1,825 35.8 0.70 0.66 0.75 <0.001 0.68 0.63 0.72 <0.001 Other 2,406 19.4 0.30 0.29 0.32 <0.001 0.38 0.36 0.40 <0.001 Co-morbidity score 12,602 34.6 <0.001 <0.001 0 2,123 35.1 1.00 1.00 1-2 1,200 32.0 0.87 0.80 0.95 0.002 1.01 0.92 1.10 0.87 3+ 175 24.8 0.61 0.51 0.73 <0.001 0.77 0.64 0.92 0.005 CCM missing 8,410 37.5 1.11 1.04 1.18 0.001 1.02 0.96 1.09 0.52 No HES link 694 19.8 0.46 0.41 0.50 <0.001 0.62 0.56 0.69 <0.001 Year of Diagnosis 12,602 34.6 <0.001 <0.001 1999 2,059 40.0 1.00 1.00 2000 2,013 38.5 0.94 0.87 1.02 0.13 0.96 0.89 1.05 0.40 2001 1,881 36.2 0.85 0.79 0.92 <0.001 0.83 0.77 0.91 <0.001 2002 1,730 33.2 0.75 0.69 0.81 <0.001 0.70 0.65 0.76 <0.001 2003 1,639 31.4 0.69 0.63 0.75 <0.001 0.66 0.60 0.72 <0.001 2004 1,615 31.0 0.67 0.62 0.73 <0.001 0.65 0.59 0.70 <0.001 2005 1,665 31.8 0.70 0.65 0.76 <0.001 0.67 0.61 0.73 <0.001 GP referral 12,602 34.6 <0.001 <0.001 No 5,581 28.2 1.00 1.00 Yes 7,021 42.1 1.85 1.77 1.93 <0.001 1.63 1.55 1.70 <0.001 1Mutually adjusted for SEP, sex, age, histology, co-morbidity score, year of diagnosis and GP referral152 Table 10.11. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors (DCO cases excluded) for 2006-2010 cohort Variable Receiving radiotherapy (9611/28,733) Unadjusted (n=28,733) Adjusted – selected1 (n=28,733, R2=11.20) N % OR (95% CI) P OR (95% CI) P Deprivation quintile 9,611 33.5 0.80 0.80 1 (least deprived) 1,126 33.2 1.00 1.00 2 1,401 33.5 1.01 0.92 1.12 0.78 1.04 0.94 1.16 0.41 3 1,612 33.3 1.00 0.91 1.10 0.98 1.03 0.93 1.14 0.60 4 2,215 33.0 0.99 0.91 1.08 0.83 1.00 0.91 1.10 0.96 5 (most deprived) 3,257 33.9 1.03 0.95 1.12 0.48 1.03 0.94 1.13 0.48 Sex 9,611 33.5 <0.001 0.06 Female 4,258 32.1 1.00 1.00 Male 5,353 34.6 1.12 1.06 1.17 <0.001 1.05 1.00 1.11 0.06 Age group 9,611 33.5 <0.001 <0.001 <60 1,707 46.4 1.00 1.00 60-69 3,140 41.3 0.82 0.75 0.88 <0.001 0.86 0.79 0.93 <0.001 70-79 3,330 32.5 0.56 0.52 0.60 <0.001 0.67 0.62 0.73 <0.001 80+ 1,434 19.9 0.29 0.26 0.31 <0.001 0.49 0.45 0.54 <0.001 Histology 9,611 33.5 <0.001 <0.001 NSCLC 6,612 43.7 1.00 1.00 SCLC 1,562 44.7 1.04 0.97 1.12 0.30 1.06 0.98 1.14 0.15 Other 1,437 14.2 0.21 0.20 0.23 <0.001 0.30 0.28 0.32 <0.001 Year of Diagnosis 9,611 33.5 0.003 0.009 2006 1,857 33.6 1.00 1.00 2007 1,865 32.7 0.96 0.89 1.04 0.30 0.95 0.88 1.03 0.25 2008 1,865 31.9 0.93 0.86 1.00 0.06 0.87 0.80 0.94 0.001 2009 2,060 35.1 1.07 0.99 1.16 0.09 0.99 0.91 1.08 0.78 2010 1,964 34.1 1.02 0.95 1.11 0.58 0.95 0.87 1.04 0.28 Stage 9,611 33.5 <0.001 <0.001 I 370 31.2 1.00 1.00 II 234 42.4 1.62 1.32 2.00 0.00 1.56 1.25 1.94 <0.001 III 1,206 53.1 2.49 2.15 2.89 0.00 2.42 2.07 2.83 <0.001 IV 1,259 33.5 1.11 0.97 1.28 0.14 1.22 1.05 1.42 <0.011 Missing 6,542 31.2 1.00 0.88 1.14 1.00 1.51 1.29 1.76 <0.001 Performance Status 9,611 33.5 <0.001 <0.001 0 832 45.2 1.00 1.00 1-2 2,375 48.8 1.16 1.04 1.29 0.01 1.43 1.27 1.60 <0.001 3-4 387 17.8 0.26 0.23 0.30 <0.001 0.52 0.44 0.61 <0.001 Missing/ unknown 6,017 30.3 0.53 0.48 0.58 <0.001 0.89 0.78 1.01 0.07 GP referral 9,611 33.5 <0.001 <0.001 No 3,262 24.6 1.00 1.00 Yes 6,349 41.1 2.14 2.04 2.25 <0.001 1.70 1.61 1.79 <0.001 1Mutually adjusted for SEP, sex, age, histology, year of diagnosis, stage, PS and GP referral 153 Table 10.12. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for probable NSCLC Variable Unadjusted (n=16,278) Adjusted – selected1 (n=16,278, R2=2.98) OR (95% CI) P OR (95% CI) P Deprivation quintile 0.005 0.04 1 (least deprived) 1.00 1.00 2 1.07 0.95 1.21 0.24 1.08 0.96 1.22 0.21 3 1.08 0.96 1.21 0.20 1.08 0.96 1.22 0.20 4 1.12 1.01 1.25 0.04 1.11 0.99 1.24 0.08 5 (most deprived) 1.20 1.08 1.33 0.001 1.18 1.06 1.31 0.003 Sex 0.08 0.09 Female 1.00 1.00 Male 1.06 0.99 1.13 0.08 1.06 0.99 1.13 0.09 Age group <0.001 <0.001 <60 1.00 1.00 60-69 0.94 0.86 1.03 0.21 0.92 0.84 1.02 0.11 70-79 0.78 0.71 0.86 <0.001 0.77 0.70 0.85 <0.001 80+ 0.69 0.62 0.77 <0.001 0.70 0.63 0.79 <0.001 Year of Diagnosis 0.56 0.03 2006 1.00 1.00 2007 1.00 0.91 1.10 1.00 0.98 0.89 1.09 0.74 2008 0.95 0.86 1.05 0.36 0.87 0.79 0.97 0.01 2009 1.04 0.94 1.15 0.43 0.94 0.84 1.04 0.20 2010 1.00 0.91 1.11 0.97 0.88 0.79 0.98 0.02 Stage <0.001 <0.001 I 1.00 1.00 II 1.90 1.50 2.40 <0.001 1.86 1.47 2.36 <0.001 III 3.09 2.61 3.66 <0.001 3.04 2.55 3.61 <0.001 IV 1.54 1.31 1.81 <0.001 1.58 1.34 1.86 <0.001 Missing 1.46 1.27 1.69 <0.001 1.77 1.48 2.12 <0.001 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 1.47 1.30 1.66 <0.001 1.53 1.34 1.73 <0.001 3-4 0.60 0.50 0.71 <0.001 0.66 0.54 0.79 <0.001 Missing/ unknown 0.92 0.82 1.02 0.13 1.03 0.89 1.20 0.71 GP referral <0.001 <0.001 No 1.00 1.00 Yes 1.63 1.53 1.74 <0.001 1.58 1.48 1.69 <0.001 1Mutually adjusted for SEP, sex, age, histology, year of diagnosis, stage, PS and GP referral 154 Table 10.13. OR of receipt of lung cancer radiotherapy, by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) for SCLC Variable Unadjusted (n=3495) Adjusted – selected1 (n=3495, R2=6.63) OR (95% CI) P OR (95% CI) P Deprivation quintile 0.48 0.35 1 (least deprived) 1.00 1.00 2 0.84 0.64 1.09 0.19 0.86 0.65 1.13 0.28 3 1.01 0.78 1.29 0.96 0.98 0.75 1.27 0.85 4 0.88 0.69 1.12 0.31 0.82 0.64 1.06 0.13 5 (most deprived) 0.90 0.72 1.13 0.38 0.84 0.66 1.07 0.16 Sex 0.06 0.04 Female 1.00 1.00 Male 0.88 0.77 1.00 0.06 0.86 0.75 0.99 0.04 Age group <0.001 <0.001 <60 1.00 1.00 60-69 0.69 0.57 0.84 <0.001 0.67 0.55 0.82 <0.001 70-79 0.46 0.38 0.56 <0.001 0.47 0.39 0.58 <0.001 80+ 0.31 0.24 0.41 <0.001 0.32 0.25 0.42 <0.001 Year of Diagnosis <0.001 0.008 2006 1.00 1.00 2007 1.03 0.83 1.28 0.77 1.02 0.82 1.28 0.84 2008 1.14 0.93 1.41 0.22 1.07 0.86 1.34 0.53 2009 1.53 1.24 1.89 <0.001 1.45 1.16 1.82 0.001 2010 1.48 1.19 1.83 <0.001 1.41 1.12 1.77 0.003 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 0.47 0.34 0.65 <0.001 0.53 0.38 0.75 <0.001 3-4 0.11 0.07 0.17 <0.001 0.14 0.09 0.22 <0.001 Missing/ unknown 0.31 0.23 0.42 <0.001 0.41 0.30 0.56 <0.001 GP referral <0.001 <0.001 No 1.00 1.00 Yes 1.89 1.64 2.16 <0.001 1.78 1.54 2.05 <0.001 1Mutually adjusted for SEP, sex, age, year of diagnosis, PS and GP referral 155 10.4 Discussion 10.4.1 Principal findings Socio-economic inequalities in receipt of surgery and chemotherapy but not radiotherapy for lung cancer were found in the full cohort analyses [1999-2005 and 2006-2010]. However, socio-economic inequalities in receipt of chemotherapy were not found in the subset with stage recorded when PS was included in the model. Socio-economic inequalities in receipt of lung cancer treatment were demonstrated in my systematic review and meta-analysis, in both universal [UHCS] and non-universal healthcare systems (236). However, the quality of the included studies varied and not all of the UK studies reported details of stage and histology, both of which influence the type of treatment that is likely to be offered. Factors such as co-morbidity and performance status [PS], that might help to explain socio-economic inequalities in receipt of treatment, had not been previously well explored. We were able to take all these factors into account and still socio-economic inequalities in receipt of surgery remained. Socio-economic inequalities in performance status appear to account for much of the socio-economic inequalities in receipt of chemotherapy in the staged subset. Socio-economic differences in tumour type may partially account for some of the socio-economic differences in receipt of surgery observed. 10.4.2 Strengths and limitations To our knowledge this is one of the first studies to use multiple dataset linkage [NYCRIS, HES and LUCADA] in order to examine the factors that may influence socio-economic inequalities in lung cancer treatment in a large dataset. Only one other recent 2012 study has used these three datasets but did not include PS or tumour type in the analysis (157). In the 1999-2005 cohort I was able to take into account co-morbidity and in the 2006-2010 analyses I was also able to take into account stage and performance status. Therefore I have been able to include a range of potential confounders that previous studies have not, and inequalities in receipt of surgery remained. The high levels of missing stage, PS and co-morbidity data are, however, a limitation. 156 It appeared that those patients referred via the GP were more likely to receive treatment. Having a GP referral date recorded was used to determine mode of presentation [via the GP or directly to secondary care], as a proxy measure of urgency. However, it is possible that some patients may have been referred via their GP, but this has not been recorded within the registry dataset. When histology was further broken down into histological [tumour] subtypes the OR of likelihood of receipt of treatment in the most compared to the least deprived SEP group was attenuated. The highest rates of surgery were found in the ‘other-specified carcinomas’ group, of which over a quarter were highly-operable carcinoid tumours, and the remaining 73% consisted of 52 different tumour-type codes. It may be that poor histological classification of non-standard tumour types [by putting these into the ‘other’ rather than NSCLC histology category] may account for this attenuation, as patients in the higher SEP groups had higher rates of these non-standardly coded tumour types than did the lowest SEP groups. 10.4.3 Interpretation of results and comparison with other studies 10.4.3.1 Receipt of surgery Socio-economic inequalities in receipt of surgery may be partially explained by socio-economic differences in tumour type as, when histology was further broken down into histological subtypes, the OR of likelihood of receipt of treatment in the most compared to the least deprived SEP group was attenuated. It appears that higher levels of more operable tumour types are found in the least deprived SEP group. A previous study using Thames Cancer Registry data found that adenocarcinoma is less clearly associated with deprivation compared to other histological subtypes, as its development is less strongly linked to smoking (248). This current study also found a higher proportion of adenocarcinoma and lower proportion of the squamous subtype in higher SEP patients compared to the lowest SEP group, with a significant association between SEP group and histological subtype. In the 2006-2010 cohort, in the highest SEP group 19.3% of patients had adenocarcinoma compared to 14.2% in the lowest; and 15.8% had squamous subtypes compared to 20.0% in the lowest. As 22.7% of patients with adenocarcinoma received 157 surgery but only 17.6% with squamous cell carcinoma did it does appear that some types of tumours are more operable than others and that socio-inequalities in receipt of surgery may be partly explained by socio-economic differences in tumour type. However, although no other UK studies could be identified that examined surgery by tumour type, a study using Danish cancer registry data for patients diagnosed between 2005-2010 (249) found similar rates of surgery in patients with adenocarcinoma [24%] but found higher rates of surgery in squamous cell patients [23%] than I did, with squamous rate similar to their adenocarcinoma rate. They therefore concluded that tumour type did not greatly influence the likelihood of receiving surgery in that population (249). Some further smoking-related confounding could however be occurring as, although I was unable to determine smoking status from this study, smokers have generally poorer health (71) and it may be this, rather than specifically tumour type, that determines receipt of surgery. However, I was able to include PS in the analysis, which is a measure of general well-being, as well as co-morbidity score. Smokers have also been shown to be less likely to receive treatment even after adjustment for co-morbidity (250) although the reasons for this are not clear. Around 30% of tumour type was recorded as morphologically unspecified [unspecified carcinoma and neoplasm] although this is a similar level to that recorded in other UK registries (20). Only 0.1% of those diagnosed with a neoplasm received surgery. Interestingly, 56% of those aged 80+ had their tumour type recorded as a neoplasm compared to 8% in the youngest age group and so this is likely to explain much of why those in the older age group did not receive surgery. Neoplasm is a general term for an unclassified tumour and is used for those patients who have undergone clinical investigation only. This could suggest that older patients do not receive surgery because their cancer has not been as thoroughly investigated and classified as those in younger patients. It is unclear whether older patients do not receive such thorough investigation because they are less well, have poorer PS and so are less able to tolerate this, or whether there is genuine bias occurring here, where older patients receive less thorough investigation than younger patients due to more nihilistic attitudes from clinicians. 158 Previous studies have shown age-related inequalities in histological diagnosis and treatment even after co-morbidity and PS were taken into account, which would suggest age-discrimination (251). A Scottish study found that the selection of investigations initially carried out varied by geographical location and according to individual clinician practice rather than being guided by the therapeutic intention. Patients who were referred for surgery tended to be younger and had good PS (252). Although the study was carried out prior to treatment guidelines being introduced it is possible that similar practices still occur. A study in the Netherlands that examined lung cancer treatment found that less than half the patients were treated according to guidelines and the proportion decreased with age (72). Age inequalities in treatment remained in the multivariable analyses here when PS and co-morbidity were taken into account which would suggest that older-patients who are fit enough for treatment are not receiving it. 10.4.3.3 Receipt of radiotherapy Socio-economic inequalities in receipt of radiotherapy were not found apart from in the analysis of NSCLC-only patients, where more deprived NSCLC patients were significantly more likely to receive radiotherapy. However, it was not possible to distinguish between palliative and radical radiotherapy. Low dose palliative radiotherapy is most commonly given, whereas fewer than 10% of patients receive high dose radiotherapy with potentially-curative intent. It is possible that differential effects by SEP might be seen if these two groups were separated, with more deprived SEP patients more likely to get palliative care, and less deprived patients, curative care. Potentially, these differential effects could cancel each other out and help explain why no overall association was found. However, the low R2 value [2.98] suggests that the variables in the model do not well account for the factors that are important determinants of receipt of radiotherapy. 10.4.3.4 Receipt of chemotherapy In the multivariable analysis in the subset of patients who had stage recorded, lower SEP was associated with a reduced likelihood of receipt of chemotherapy when including age, sex, histology and stage in the model. However, on the addition of performance status to the model this association was no longer observed, suggesting 159 that socio-economic differences in performance status may therefore account for the observed socio-economic differences in receipt of chemotherapy in this subgroup. Only two other UK studies [using early LUCADA audit data] have included performance status in a multivariable analysis of receipt of treatment (152, 153). They found that SEP was not associated with receipt of surgery but was associated with receipt of chemotherapy when performance status was included. In contrast we found that, when including performance status, SEP remained associated with a lower likelihood of receipt of surgery but SEP was no longer associated with receipt of chemotherapy in the staged subset. The two previous studies using LUCADA data also found that number of co-morbidities was significantly associated with receipt of surgery but other studies have not found this (157) and nor did this study. Previous studies used national LUCADA audit data [entered from 2004-2007 and 2004-2008] which, for those early years, included only a small subset of registry patients. Although entry of lung cancer stage data in LUCADA was noted to be 85% complete nationally for 2008 in the Audit report (246), the validity of the 2005 and 2006 LUCADA data has been queried due to the poor entry of staging data (28). In these LUCADA studies no association between SEP and receipt of surgery was found in the unadjusted and adjusted analyses. This result was in contrast to the majority of other UK studies included in a systematic review of inequalities in treatment that found an association between SEP and receipt of surgery (236). It should therefore be queried whether patients included in LUCADA in the early years of the audit are representative of the full spectrum of patients diagnosed with lung cancer in England. One of the previous LUCADA studies concluded that the data within LUCADA were unbiased and representative of all lung cancer patients in England, despite the variable levels of case ascertainment in the Trusts that supplied data in the early years (153). They also suggested that there was little variation in patient demographics across participating Trusts with different levels of missing data (153), but this does not seem to be quite true. Closer inspection shows that there were significant differences in stage at diagnosis, histology and PS when comparing patients from Trusts with high levels of missing data against those who had low levels of missing data. Patients from Trusts with high levels of missing data were significantly more likely to have early-stage disease but these differences were explained away as being proportionately different 160 but small at an absolute level (153). Unfortunately SEP was not examined. However, it does appear that results obtained from studies using LUCADA data do differ from those using Registry and HES data, and some caution should be employed when interpreting data from studies using only early-years LUCADA data, as those patients from Trusts with high levels of missing data do appear to be selectively different. An NCIN report also noted that although concordance of recording of data on receipt of surgery between the combined national registry-HES dataset and the LUCADA dataset was high this was not so for chemotherapy and radiotherapy, with 48% of patients with chemotherapy and 58.6% with radiotherapy recorded in the former having no record in the latter (253), so again this might account for some of the differences seen in my results to that found in previous LUCADA studies. We were able to include later years of LUCADA data [2009-2010] which are more complete. However, we used local registry data whereas previous studies used England-wide data, which might explain some of the differences observed. Higher levels of missing data for stage and PS were found in our linked dataset than have been reported nationally (246). Higher levels of deprivation are found in the north compared to England as a whole. Although other regional analyses have found inequalities in lung cancer treatment, none have investigated the role of PS and tumour type. It would be useful to perform these analyses using other regional or national registry data to confirm results. 10.4.4 Implications for policy and practice, and future research The results from this study suggest that socio-economic inequalities in performance status substantially explain socio-economic inequalities in receipt of chemotherapy in the subset of patients whose cancer was staged. However, this staged subset may not be representative of the full regional cohort as patients within this were more likely to be younger and of higher SEP. A previous study has shown a socio-economic gradient in completeness of data on stage and grade of cancer, which could be interpreted as inequality in investigative intensiveness (247). It may be that younger patients receive more intensive investigation and so are more likely to be staged (247) and so, although PS may explain inequalities in chemotherapy in this group, they are a selective cohort. This is a relationship that needs to be clarified in other datasets, ideally with lower 161 levels of missing data for stage and PS. Later years of data in LUCADA have high levels of data completeness and so could be utilised in future. Guidelines indicate that chemotherapy should be offered to stage III patients and to stage IV patients with good PS [0-1] (18) so that poor PS is a valid reason not to offer chemotherapy to patients who would not be able to tolerate this. Chemotherapy can be offered to patients who are ambulatory and not bed-bound (18). Socio-economic differences in health status may determine whether a patient receives chemotherapy. Patients who do not smoke, eat a healthy and balanced diet, are not overweight or obese, and undertake exercise are more likely to be in better health and so might have a greater chance of being able to undergo chemotherapy. Policy advice regarding healthy lifestyle would therefore apply here, although there is clearly a long chain of causality from health behaviours earlier in life and their specific implications for health status in later life. It is also debateable whether making lifestyle changes once diagnosed with cancer is likely to do much to improve PS, although a recent systematic review and meta-analysis produced preliminary evidence for improved survival [using life table modelling] for early-stage lung cancer patients who quit smoking (254). A higher percentage of patients in the lowest SEP group had a squamous cell subtype which is strongly associated with smoking. Surgery rates were also lower for this histological subtype. Non-smokers are less likely to get lung cancer and if they do then it appears that they are more likely to get a histological subtype that is more amenable to surgery. This is a further reason, if any other were required, to emphasise why it is important not to smoke. However, we cannot rule out uncontrolled confounding related to smoking-status, where smokers may be less likely to undergo treatment for smoking-related reasons [such as suffering from a serious, smoking-related co-morbidity such as COPD or heart disease] that we cannot measure within this study. Unfortunately smoking status is not recorded in the cancer registry data-set. This relationship between tumour type and receipt of treatment was not found in a Danish cohort (249) and further UK studies are needed to confirm this association. I have been able to investigate a number of factors that may be important in the relationship between SEP and receipt of treatment, such as stage and performance status, but the high levels of missing data limit the conclusions that can be drawn. 162 Although tumour type may account for some of the socio-economic gradient in surgery it does not explain it all. Socio-economic inequalities in PS did not explain inequalities in chemotherapy in the full cohort analyses [although it did within the staged subset], and nor did stage or number of co-morbidities, suggesting that other factors are at play. I was unable to look at patient choice. Poorer health literacy may influence patient choice and understanding of risk and this may vary by SEP (41). If patients have poor capacity to process and understand basic health information then they are less able to make appropriate health and treatment decisions (43). It is important that clinicians take this into account when discussing treatment options. Further research is required to investigate the unexplained variance in treatment rates, looking at factors such as patient choice, doctor-patient communication of risk and benefit, and possible system variation by trusts or within this, by hospital and individual consultant. Previous studies have suggested that there may be variation in the level of surgery offered by trusts or, within this, by hospital, so that area-level rather than individual factors may account for treatment variation observed (161). Multi-level modelling is required to fully investigate the effect of area-level factors such as trust, in this data-set. 10.5 Chapter summary Socio-economic inequalities in receipt of surgery, chemotherapy but not radiotherapy were found in the full cohort analyses. Socio-economic inequalities in performance status statistically explain socio-economic inequalities in receipt of chemotherapy in the selective subset of patients whose cancer was staged, but not in the full cohort in this study. Socio-economic inequalities in receipt of surgery cannot be statistically explained by inequalities in stage, PS or co-morbidity. However, socio-economic inequalities in tumour type may account for some of the inequalities in surgery by SEP. Patients in lower SEP groups are more likely to be diagnosed with squamous cell cancer [a tumour type strongly associated with smoking] and are less likely to receive surgery than 163 patients in higher SEP groups, who are more likely to be diagnosed with adenocarcinoma. In the next chapter socio-economic inequalities in referral, diagnostic and treatment intervals are examined. 164 Chapter 11. Inequalities in referral, diagnostic and treatment time intervals Summary Background Diagnostic delay has been implicated as a factor that contributes to the poor survival of UK cancer patients compared to the European average. In England, urgent referrals for suspected cancer are required to have a first hospital appointment [FHA] within 14 days from the date of GP referral [referral interval], and first treatment within 62 days from date of urgent GP referral and within 31 days from diagnosis [treatment intervals]. There has been little work conducted on socio-economic inequalities and delay in lung cancer. Cancer Registry [NYCRIS], Hospital Episode Statistics [HES] and lung cancer audit [LUCADA] data-sets were linked in order to investigate socio-economic inequalities in system delay for lung cancer. Methods System delay was examined in patients diagnosed between 1999 and 2005 [n=36,477] of whom 32,974 [90.4%] had a linked HES record, and in patients diagnosed between 2006 and 2010 [n=28,733] a subset of whom had PS and stage [n=7769, 27.0%] recorded in LUCADA. Socio-economic inequalities in the likelihood of receipt of referral, diagnosis and treatment within the target intervals and for the calculated interim date targets were explored using logistic regression models. Cox regression analysis was used to examine the influence of SEP on the likelihood of shorter time to first hospital appointment, diagnosis and treatment. Results (for 2006-2010 cohort) In the 2006 to 2010 cohort 70% of patients received a hospital appointment within the 14 day referral target. Time to treatment targets were only being met for 42.5% when measured from diagnosis and for 62% from GP referral. Socio-economic inequalities in 165 time from GP referral to FHA, but not in time to diagnosis or treatment, were found. However no linear trend association between SEP and referral time was found. Late-stage, poor performance status, and SCLC histology were associated with a higher likelihood of a first hospital appointment within 14 days of GP referral. Older patients were less likely, whereas late stage and poor PS patients were more likely, to receive treatment within guideline time limits. Conclusions Patients who appeared sick were more likely to receive early referral, diagnosis and treatment. However, older patients, who were more likely to have poorer PS, were less likely to receive early treatment, indicating possible age-discrimination. There is some evidence for socio-economic inequalities in the referral but not in the diagnostic or treatment interval. It is possible that the WTP ‘sicker quicker’ effect may effectively ‘cancel out’ system inequalities that might result in longer time intervals for lower SEP patients. 166 11.1 Introduction Delays between the onset of cancer symptoms and the time to treatment may contribute to the poor survival of UK cancer patients compared to that found in other European countries (99). Early diagnosis of cancer is thought to be important for improving outcomes, as survival is better for patients who are diagnosed at an early stage, because they are more likely to be suitable for receipt of potentially curative treatment (99). Early diagnosis may also result in longer intermediate survival for patients with SCLC and later stage NSCLC [Dr M Peake, personal communication]. An early model of cancer delay, the Anderson model [Fig 4.3], attributed the majority of delay to patient rather than system [primary and secondary healthcare] factors. However, it has been suggested that this may be ‘an artefact of research focus’ and that system delay may be an equally important but under-researched area (102). Current theoretical models of the pathway from first symptom to cancer treatment identify key intervals and related health care settings (104). Best practice in the definition of key time points demarcating these intervals has also been described (100). In England, three intervals have been the subject of performance management. Since 2000, urgent referrals for suspected cancer have been required to have a first hospital appointment [FHA] within 14 days from the date of referral [referral interval]. Since 2005, intervals of 62 days from date of urgent GP referral to first treatment and 31 days from diagnosis/decision to treat to first treatment [treatment intervals] have been in place (24). Although some research has been conducted into the factors that might influence delay for some common cancers (105, 112) there has been little work conducted on socio-economic inequalities and delay in lung cancer. Table 4.1 in chapter 4 summarises the current evidence. In this study cancer registry, Hospital Episode Statistics [HES] and lung cancer audit [LUCADA] data-sets were linked in order to investigate system delay, examining socio-economic inequalities in referral, diagnosis and treatment time for lung cancer patients; the likelihood of referral, diagnosis and treatment within recommended National Cancer Plan [NCP] guideline target times, by SEP; and to examine the other 167 factors [age, sex, histology, stage, and health status] that may impact on these time intervals. 11.2. Methods 11.2.1 Data For details of data sources and variables see chapter 8, sections 8.2 and 8.4. 11.2.2 Dates, interval periods and target times For details of dates, intervals and target times see chapter 8, table 8.2 and section 8.4.1. Records were excluded from a particular analysis if they had a negative interval for that time period e.g. if they had a GP referral date later then diagnosis date, as this was likely to be a data entry error. Records were included if they had 0 time between dates [e.g. treatment date was recorded on the same day as date as diagnosis. In order to include these records in the analysis, 0 time intervals were recoded as 0.1]. Analysis was also restricted to those cases with interval dates within one year of the previous interval endpoint. Records were excluded [table 11.1] if the dates fell out-with these timeframes, as extremely long gaps between dates were considered likely to be data entry errors, such as transposition of numbers on date entry. Some patients had very long [one to five year] gaps from GP referral to FHA and these dates might refer to other illness episodes. Table 11.1. Exclusion of data outliers Time period N N interval > 1 year % excluded GP referral to FHA 31415 36 0.11 FHA to diagnosis 31524 20 0.06 Diagnosis to treatment 15051 57 0.38 11.2.3 Analysis NYCRIS data for 28,733 patients diagnosed between 2006 and 2010 and 36,473 diagnosed between 1999 and 2005 were analysed. As date of diagnosis was determined in a different way in 2010 from previous years it was thought that this 168 might affect calculation of the diagnosis to treatment interval. Therefore inequalities in time from diagnosis to treatment were also examined in the 2006-2009 cohort. The number and percentage of patients referred within guidelines were calculated and inequalities in the likelihood of receipt of referral, diagnosis and treatment within the targets set by the National Cancer Plan (24) and for the calculated interim date targets [see chapter 8 section 8.4.1.4.2], by SEP, adjusted for age, sex, histology, year of diagnosis, co-morbidity score, stage, and PS, were explored using logistic regression models mutually adjusted for all co-variables. Referral route was also included in the models for the FHA to diagnosis and diagnosis to treatment intervals. Type of first treatment was included in the models for the FHA to treatment and diagnosis to treatment intervals. Receipt of treatment was included in the models for the GP referral to diagnosis and FHA to diagnosis intervals. Interactions between SEP and histology were also explored. The R2 statistic was examined to determine the amount of variance in receipt of treatment explained by each model. Odds ratios [ORs] with 95% confidence intervals [CIs] for the likelihood of referral, diagnosis and treatment within target in the lowest compared to the highest SEP group were reported. A likelihood ratio test was performed to determine the overall significance of each categorical variable. The problem with using logistic regression as described above is that one can only look at the odds of receiving treatment within one year for those who have received any treatment, thus excluding around half of the dataset who are untreated. If Cox regression is used, with ‘failure’ defined as receipt of treatment [equivalent to death in survival analyses, except it is a positive rather than negative outcome], then it is possible to include the whole dataset, even those who did not receive treatment. Hazard ratios [HRs] are a measure of time to treatment in the whole dataset, with a lower hazard ratio equating to a longer time to treatment. This is only relevant for time to treatment analyses as the dates required to calculate time to FHA and diagnosis were well recorded and fairly complete for the whole dataset. Median time and inter-quartile range [IQR], from GP referral to first hospital appointment [FHA], diagnosis and first treatment, and from diagnosis to treatment, was calculated. Univariable and multivariable Cox regression analysis [with a hazard 169 ratio of <1.00 indicating longer time to treatment] was used to examine the influence of SEP, adjusted for age, sex, stage, performance status, co-morbidity, histology and year of diagnosis, on the likelihood of shorter time to first hospital appointment, diagnosis and treatment, for those who had dates recorded for those time periods. For some analyses receipt of treatment, type of first treatment and whether the patient had a GP referral date were also included. Hazard ratios [HRs] with 95% confidence intervals [CIs] for the likelihood of early referral, diagnosis and treatment in the lowest compared to the highest SEP group were reported. A likelihood ratio test was performed to determine the overall significance of each categorical variable. Cox regression was also used to assess the likelihood of shorter time to treatment from diagnosis and GP referral date, for the whole dataset, including those who did not receive treatment. 11.3 Results 11.3.1 Descriptive statistics Table 9.1 shows the demographic and clinical characteristics of the lung cancer patients included in the study. Table 11.2 shows the level of data completeness for diagnosis, FHA and GP referral date and the number receiving treatment. Of the 28,733 patients diagnosed between 2006 and 2010, all had a date of diagnosis, 28, 704 [99.9%] had FHA date recorded, 15,452 [53.8%] had a GP referral date and 15,373 [53.5%] received any treatment within one year of diagnosis. Table 11.3 shows the overall median time per interval compared to target time, and the numbers included in the analysis for each time period. Table 11.2. Date completeness for 1999-2005 and 2006-2010 data Date 1999-2005 (n=36,477) 2006-2010 (n=28,733) N % N % Diagnosis 36,477 100 28,733 100 FHA 36,409 99.8 28,704 99.9 GP referral 16,688 45.7 15,452 53.8 Treatment (any) 19,516 53.5 15,442 53.7 Any treatment within 1 year and with first treatment date >=diagnosis date 19,510 53.5 15,373 53.5 170 Table 11.3. Median time per interval compared to target time, for 1999-2005 and 2006-2010 data 1999-2005 2006-2010 Interval Target time* N Median time IQR N Median time IQR GP referral date to first hospital appointment date 14 16649 10 6-15 14730 10 6-17 GP referral date to diagnosis date 31 16644 16 9-33 14865 13 7-24 First hospital appointment date to diagnosis date 17 36330 3 0-13 28284 0 0-0 Diagnosis date to first treatment date 31 19510 36 20-62 15373 36 22-56 GP referral date to first treatment date 62 10844 63 41-98 10090 56 39-79 To avoid repetition of results only the results for 2006-2010 are presented below. The pattern of results found was broadly similar in the 1999-2005 cohort [see appendix D tables D1-D3] and for the 2006-2009 dataset [results not shown]. No significant interactions between SEP and histology were found in any of the models. 11.3.2 GP referral to first hospital appointment 2006-2010 11.3.2.1 Hazard ratio of early FHA from GP referral Median time from GP referral to FHA was 10 days [IQR 6-17], n=14,730 [table 11.4]. A linear association between SEP and likelihood of earlier FHA within one year of GP referral was not found, although those in the middle SEP groups had significantly decreased likelihood [as measured by HR] of early FHA from referral compared to the least deprived SEP group. Patients with poorer PS, those subsequently diagnosed with later stage cancer and those diagnosed with SCLC all had an increased likelihood [HR] of early FHA from referral in the multivariable analysis [table 11.4]. However, median referral time was between 8-11 days for these variables and so any ‘delay’ appeared short. Those diagnosed in later years were significantly more likely to have an early FHA with a median time to FHA of 9 days in 2010 compared to 14 days in 2006. A similar pattern was seen for the cohort that had stage recorded [Appendix D, table D4] although median referral time was shorter in this sub-group [n=5100, median = 9 days (IQR 4-14)]. 171 Table 11.4. Hazard ratio of early first hospital appointment from GP referral (for those referred within 1 year) in 2006-2010 cohort FHA Median time to FHA (days) Univariable analysis (n=14,730) Multivariable analysis1 (n=14,730) N N IQR HR 95% CI P HR 95% CI P IMD 14730 10 6-17 0.03 0.08 1 (least deprived) 1735 9 5-15 1.00 1.00 2 2054 10 6-17 0.94 0.88 1.00 0.05 0.96 0.90 1.02 0.18 3 2496 10 6-17 0.93 0.88 0.99 0.03 0.94 0.88 1.00 0.04 4 3490 11 6-18 0.91 0.86 0.97 0.002 0.93 0.87 0.98 0.008 5 (most deprived) 4955 10 6-17 0.95 0.90 1.01 0.08 0.96 0.91 1.02 0.15 Age Range 14730 10 6-17 0.004 0.004 <60 2043 10 6-17 1.00 1.00 60-69 4218 10 6-16 1.01 0.96 1.06 0.75 1.00 0.94 1.05 0.89 70-79 5336 11 6-18 0.94 0.89 0.99 0.02 0.93 0.88 0.98 0.007 80+ 3133 10 5-17 0.98 0.92 1.03 0.41 0.96 0.90 1.01 0.13 Sex 14730 10 6-17 0.36 0.96 Female 6696 10 6-17 1.00 1.00 Male 8034 10 6-17 0.99 0.95 1.02 0.36 1.00 0.97 1.03 0.96 Histology 14730 10 6-17 <0.001 <0.001 NSCLC 8940 11 6-18 1.00 1.00 SCLC 1929 9 5-14 1.20 1.15 1.26 <0.001 1.20 1.14 1.26 <0.001 Other 3861 10 4-17 1.03 0.99 1.07 <0.101 1.02 0.98 1.06 0.28 Year of Diagnosis 14730 10 6-17 <0.001 <0.001 2006 2498 14 9-25 1.00 1.00 2007 2603 13 8-21 1.12 1.06 1.19 <0.001 1.12 1.06 1.19 <0.001 2008 3089 8 5-13 1.72 1.63 1.81 <0.001 1.70 1.61 1.79 <0.001 2009 3232 8 3-14 1.69 1.61 1.78 <0.001 1.79 1.67 1.91 <0.001 2010 3308 9 3-14 1.63 1.54 1.71 <0.001 1.69 1.58 1.81 <0.001 Co-morbidity score 14730 10 6-17 <0.001 <0.001 0 1,929 11 6-19 1.00 1.00 1-2 1,418 10 4-19 1.02 0.95 1.09 0.56 0.96 0.90 1.03 0.25 3+ 298 9 2-20 1.07 0.95 1.21 0.26 0.93 0.82 1.05 0.25 CCM missing 5,432 11 7-19 1.02 0.96 1.07 0.55 1.14 1.08 1.20 <0.001 No HES link 5,653 9 4-14 1.23 1.17 1.30 <0.001 1.00 0.94 1.06 1.00 Stage 14730 10 6-17 <0.001 <0.001 I 773 11 7-20 1.00 1.00 II 399 11 5-16 1.21 1.08 1.37 0.002 1.20 1.06 1.35 0.004 III 1578 9 6-15 1.28 1.17 1.39 <0.001 1.23 1.13 1.34 <0.001 IV 2350 8 3-14 1.47 1.36 1.60 <0.001 1.36 1.25 1.48 <0.001 Missing/ unknown 9630 11 6-19 1.08 1.00 1.16 0.04 1.21 1.11 1.32 <0.001 Performance Status 14730 10 6-17 <0.001 <0.001 0 1,367 11 6-15 1.00 1.00 1-2 3,257 9 5-15 1.07 1.01 1.14 0.03 1.05 0.99 1.12 0.13 3-4 1,230 7 0-13 1.39 1.28 1.50 <0.001 1.33 1.23 1.44 <0.001 Missing/ unknown 8,876 11 6-19 0.91 0.86 0.96 0.001 1.01 0.94 1.08 0.88 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS172 Table 11.5. Odds of FHA within 14 days from GP referral (for those with FHA within 1 year of GP ref) 2006-2010 FHA within 1 year FHA within 14 days Univariable analysis (n=14,730) Multivariable analysis1 (n=14,730, R2=6.80) N N % OR 95% CI P OR 95% CI P IMD 14730 10319 70.1 0.005 0.001 1 (least deprived) 1735 1274 73.4 1.00 1.00 2 2054 1439 70.1 0.85 0.73 0.98 0.02 0.85 0.74 0.99 0.04 3 2496 1738 69.6 0.83 0.72 0.95 0.007 0.81 0.70 0.93 0.003 4 3490 2385 68.3 0.78 0.69 0.89 <0.001 0.75 0.66 0.86 <0.001 5 (most deprived) 4955 3483 70.3 0.86 0.76 0.97 0.01 0.83 0.73 0.94 0.004 Age Range 14730 10319 70.1 0.01 0.03 <60 2043 1467 71.8 1.00 1.00 60-69 4218 3005 71.2 0.97 0.87 1.09 0.64 0.96 0.85 1.08 0.49 70-79 5336 3661 68.6 0.86 0.77 0.96 0.008 0.86 0.76 0.97 0.01 80+ 3133 2186 69.8 0.91 0.80 1.02 0.12 0.89 0.78 1.02 0.09 Sex 14730 10319 70.1 0.17 0.80 Female 6696 4729 70.6 1.00 1.00 Male 8034 5590 69.6 0.95 0.89 1.02 0.17 1.01 0.94 1.09 0.80 Histology 14730 10319 70.1 <0.001 <0.001 NSCLC 8940 6135 68.6 1.00 1.00 SCLC 1929 1492 77.4 1.56 1.39 1.75 <0.001 1.62 1.43 1.83 <0.001 Other 3861 2692 69.7 1.05 0.97 1.14 0.22 1.04 0.95 1.14 0.42 Year of Diagnosis 14730 10319 70.1 <0.001 <0.001 2006 2498 1250 50.0 1.00 1.00 2007 2603 1484 57.0 1.32 1.19 1.48 <0.001 1.33 1.19 1.49 <0.001 2008 3089 2491 80.6 4.16 3.69 4.68 <0.001 4.17 3.69 4.70 <0.001 2009 3232 2543 78.7 3.68 3.28 4.13 <0.001 4.24 3.66 4.91 <0.001 2010 3308 2551 77.1 3.36 3.01 3.77 <0.001 3.80 3.26 4.43 <0.001 Co-morbidity score 14730 10319 70.1 <0.001 <0.001 0 1,929 1,291 66.9 1.00 1.00 1-2 1,418 956 67.4 1.02 0.88 1.18 0.76 0.91 0.78 1.06 0.23 3+ 298 204 68.5 1.07 0.83 1.39 0.60 0.82 0.62 1.08 0.17 CCM missing 5,432 3,577 65.9 0.95 0.85 1.06 0.39 1.25 1.10 1.41 <0.001 No HES link 5,653 4,291 75.9 1.56 1.39 1.74 <0.001 0.95 0.83 1.09 0.46 Stage 14730 10319 70.1 <0.001 <0.001 I 773 508 65.7 1.00 1.00 II 399 286 71.7 1.32 1.01 1.72 0.04 1.33 1.01 1.75 0.04 III 1578 1182 74.9 1.56 1.29 1.88 <0.001 1.50 1.24 1.83 <0.001 IV 2350 1870 79.6 2.03 1.70 2.43 <0.001 1.74 1.44 2.10 <0.001 Missing/ unknown 9630 6473 67.2 1.07 0.92 1.25 0.39 1.42 1.17 1.72 <0.001 Performance Status 14730 10319 70.1 <0.001 0.0005 0 1,367 997 72.9 1.00 1.00 1-2 3,257 2,422 74.4 1.08 0.93 1.24 0.31 1.01 0.87 1.18 0.88 3-4 1,230 998 81.1 1.60 1.33 1.92 <0.001 1.38 1.13 1.69 <0.001 Missing/ unknown 8,876 5,902 66.5 0.74 0.65 0.84 <0.001 0.94 0.80 1.12 0.49 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS173 11.3.2.2 Likelihood of FHA within 14 days of GP referral Of those referred by their GP [which could include both urgent and non-urgent referrals], 70.1% had a first hospital appointment within 2 weeks of referral. Low SEP was associated with a lower likelihood of FHA within 2 weeks of referral in the multivariable analysis [OR=0.83, CI 0.73 to 0.94, p=0.004] but the lowest likelihood was for those in the middle SEP groups. A similar pattern was seen for the 1999-2005 cohort [Appendix D, table D1]. Late stage cancer, SCLC, poor performance status and referral post-2006 were associated with increased likelihood of FHA within 14 days of referral, but older age [80+], sex and number of co-morbidities were not [table 11.5]. 11.3.3 GP referral to diagnosis 2006-2010 11.3.3.1 Hazard ratio of early diagnosis from GP referral In those patients with GP referral and diagnosis dates [n=14,860] median time from GP referral to diagnosis was 13 days [IQR 7-24]. Similarly to the pattern seen in the GP referral to FHA interval, those in the middle SEP groups had the lowest likelihood of earlier diagnosis. Patients with poorer PS, co-morbidities, those subsequently diagnosed with later stage cancer and those with SCLC had an increased likelihood [HR] of early diagnosis [table 11.6]. Those who subsequently went on to receive treatment had longer median time to diagnosis. Similar results were seen for the cohort subset [n=5148] who had stage recorded [data not shown]. 11.3.3.2 Likelihood of diagnosis within 31 days of GP referral Neither SEP nor age was associated with likelihood of diagnosis within 31 days of GP referral overall, although those in the middle SEP groups had significantly lower likelihood of diagnosis within 31 days [table 11.7]. Patients with poorer PS, later stage cancer and those with SCLC had an increased likelihood of diagnosis within 31 days of GP referral. Those diagnosed in in 2007-2009 appeared to have a higher likelihood of diagnosis within 31 days compared to those diagnosed in 2010. However, similar overall results were seen for the 2006-2009 data-set [results not shown]. 174 11.3.4 Diagnosis to first treatment 2006-2010 11.3.4.1 Hazard ratio of early treatment for those treated within 1 year of diagnosis Median time from diagnosis to first treatment was 36 days [IQR 22-56], for the 15,373 patients who had first treatment within 1 year of diagnosis [table 11.8]. No association between SEP and likelihood [HR] of early treatment was found. Patients with late stage cancer, those with SCLC and those with poor performance status were more likely to receive early treatment. Older patients had longer time to treatment [median time to treatment was 41 days for those aged 80+ compared to 33 days for those aged <60]. Patients with stage 1 lung cancer waited a median of 51 days for treatment, whereas those with stage 4 cancer waited 30 days. This may reflect the type of treatment they were likely to receive as patients who received surgery as a first treatment had a median waiting time 17 days longer than patients who received chemotherapy and 10 days longer than those who had radiotherapy first. Patients with poor performance status waited a median of 28 days to treatment compared with 37 days for those with good performance status. No clear pattern of waiting time was seen for number of co-morbidities, however. For those treated in 2010 it appeared that time to treatment was shorter than in all other years [median =29 days compared to 35-41 days in other years]. Again this may be related to how date of diagnosis was calculated in this year. 11.3.4.2 Hazard ratio of early treatment from diagnosis for entire cohort, including untreated I also examined the likelihood of receiving early treatment for the whole dataset, including those who did not receive treatment [table 11.9]. Using this whole-dataset methodology resulted in some contrasting findings compared to including only those who received treatment. Those with low SEP were significantly less likely to receive early treatment in the lowest compared to the highest SEP group [HR=0.83, CI 0.79 to 0.88), p<0.001], as were those with co-morbidities. Those with later stage cancer were less likely to receive earlier treatment, as were those with poorer performance status whereas the 175 reverse was found in the treatment-only cohort. Those referred via the GP were more likely to receive early treatment and, again, the reverse was found in the treatment-only cohort. Older patients were significantly less likely to receive early treatment as were those with co-morbidities. Those diagnosed with SCLC were more likely to receive early treatment. 11.3.4.3 Likelihood of treatment within 31 day of diagnosis Of the 15,373 patients who were treated within one year, 42.5% [6,537] were treated within the 31 day target time from diagnosis [table 11.10]. There was no association between low SEP and likelihood of treatment within 31 days [OR=0.91, CI 0.81 to 1.02, p=0.11], for those treated within 1 year of diagnosis. Although 73.4% of SCLC patients were treated within target, this was true for only 35.4% of NSCLC patients. Histology appeared the most important factor in the likelihood of receiving timely treatment, with a median time of 22 days for SCLC compared to 40 days for NSCLC. Patients with SCLC often deteriorate quickly and need to receive chemotherapy as soon as possible within a short ‘window of opportunity’ to improve survival time. Older patients were significantly less likely to receive treatment within the 31 day target even when factors such as stage and PS were taken into account [OR=0.78, CI 0.68 to 0.89, p<0.001]. The type of treatment first received also influenced the likelihood of treatment within target. Those receiving surgery were less likely to receive treatment within target times. Those with poor performance status and those receiving chemotherapy were more likely to do so. 11.3.5 GP referral to first treatment 2006-2010 11.3.5.1 Hazard ratio of early treatment from GP referral Median time from GP referral to first treatment date was 56 days [IQR 39-79] in the 10,090 patients who had both dates recorded [table 11.11]. Patients in the middle and low SEP groups were less likely to receive early treatment compared to the highest SEP group. 176 Early stage patients, older patients and those receiving surgery were significantly less likely to receive early treatment, whereas later stage patients, those with poor performance status and those receiving chemotherapy as first treatment were more likely to do so. 11.3.5.2 Likelihood of treatment within 62 days of GP referral 61.8% of those who received treatment were treated within the 62 day target from GP referral [although this could include patients who were not urgently referred] [table 11.12]. No linear trend association between SEP and likelihood of starting treatment within the target time was found [p=0.08] although those in the middle SEP groups were significantly less likely to start treatment within 62 days. Early stage patients, older patients and those receiving surgery were significantly less likely to start treatment within 62 days, whereas late stage patients, those with poor PS and those receiving chemotherapy were more likely to do so. The likelihood of receiving treatment within guidelines significantly improved over time. 177 Table 11.6. Hazard ratio of early diagnosis (for those referred by GP within 1 year and with diagnosis date within 1 year from FHA), for 2006 to 2010 cohort Variable Diag nosis Median time to diagnosis (days) Univariable analysis (n=14,860) Multivariable analysis1 (n=14,860) N N IQR HR 95% CI P HR 95% CI P IMD 14,860 13 7-24 0.04 0.04 1 (least deprived) 1,746 13 7-21 1.00 1.00 2 2,074 13 7-24 0.93 0.87 0.99 0.02 0.94 0.88 1.00 0.07 3 2,519 13 7-25 0.91 0.86 0.97 0.004 0.92 0.86 0.97 0.005 4 3,521 14 7-25 0.93 0.88 0.98 0.01 0.92 0.87 0.98 0.005 5 (most deprived) 5,000 13 7-23 0.95 0.90 1.00 0.04 0.95 0.90 1.00 0.06 Age Range 14,860 13 7-24 0.03 0.04 <60 2,064 13 7-23 1.00 1.00 60-69 4,252 13 7-23 1.00 0.95 1.05 0.97 1.01 0.96 1.07 0.70 70-79 5,394 14 7-24 0.96 0.92 1.01 0.16 0.95 0.91 1.01 0.08 80+ 3,150 13 7-23 1.03 0.97 1.09 0.34 0.97 0.91 1.03 0.30 Sex 14,860 13 7-24 0.88 0.63 Female 6,760 13 7-24 1.00 1.00 Male 8,100 13 7-23 1.00 0.97 1.03 0.88 1.01 0.98 1.04 0.63 Histology 14,860 13 7-24 <0.001 <0.001 NSCLC 9,027 14 8-25 1.00 1.00 SCLC 1,942 12 7-19 1.28 1.22 1.34 <0.001 1.27 1.21 1.34 <0.001 Other 3,891 12 6-22 1.11 1.07 1.16 <0.001 1.06 1.01 1.10 0.01 Year of diagnosis 14,860 13 7-24 <0.001 <0.001 2006 2,501 16 9-28 1.00 1.00 2007 2,601 14 8-24 1.15 1.09 1.22 <0.001 1.15 1.08 1.21 <0.001 2008 3,140 11 6-18 1.49 1.41 1.57 <0.001 1.45 1.38 1.53 <0.001 2009 3,260 10 5-19 1.51 1.43 1.59 <0.001 1.62 1.52 1.73 <0.001 2010 3,358 16 8-30 1.01 0.96 1.06 0.81 1.07 1.00 1.14 0.07 Co-morbidity score 14,860 13 7-24 <0.001 <0.001 0 1,958 14 7-26 1.00 1.00 1-2 1,430 13 7-27 1.00 0.93 1.07 0.89 0.94 0.88 1.01 0.10 3+ 303 13 5-29 0.93 0.83 1.05 0.27 0.84 0.74 0.95 0.006 CCM missing 5,464 13 7-21 1.13 1.08 1.19 <0.001 1.20 1.14 1.27 <0.001 No HES link 5,705 14 7-25 1.05 1.00 1.11 0.05 1.02 0.96 1.08 0.46 Stage 14,860 13 7-24 <0.001 <0.001 I 783 18 10-37 1.00 1.00 II 403 16 8-28 1.18 1.04 1.33 0.008 1.21 1.08 1.37 0.002 III 1,592 13 7-22 1.45 1.33 1.58 <0.001 1.38 1.27 1.51 <0.001 IV 2,369 11 5-19 1.77 1.64 1.92 <0.001 1.64 1.51 1.78 <0.001 Missing/ unknown 9,713 13 7-24 1.34 1.24 1.44 <0.001 1.29 1.19 1.41 <0.001 Performance Status 14,860 13 7-24 <0.001 <0.001 0 1,376 14 8-27 1.00 1.00 1-2 3,287 13 7-23 1.14 1.07 1.21 <0.001 1.08 1.01 1.15 0.02 3-4 1,240 8 3-17.5 1.54 1.43 1.66 <0.001 1.33 1.23 1.44 <0.001 Missing/ unknown 8,957 14 7-24 1.07 1.01 1.13 0.03 1.05 0.98 1.13 0.18 Any treatment 14,860 13 7-24 <0.001 <0.001 No 5063 12 6-21 1.00 Yes 9767 14 8-25 0.84 0,82 0.88 <0.001 0.89 0.85 0.93 <0.001 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS, any treatment 178 Table 11.7. Odds of diagnosis within 31 days of GP referral (excluding those with FHA > 1 year from GP ref and diagnosis> 1 year from FHA) 2006-2010 Diagnosis within 2 years Diagnosis within 31 days Univariable analysis (n=14,860) Multivariable analysis1 (n=14,860, R2=4.68) N N % OR 95% CI P OR 95% CI P IMD 14,860 12,287 82.7 0.22 0.11 1 (least deprived) 1,746 1,469 84.1 1.00 1.00 2 2,074 1,710 82.5 0.89 0.75 1.05 0.17 0.90 0.76 1.07 0.25 3 2,519 2,061 81.8 0.85 0.72 1.00 0.05 0.84 0.71 0.99 0.04 4 3,521 2,889 82.1 0.86 0.74 1.01 0.06 0.83 0.71 0.97 0.02 5 (most deprived) 5,000 4,158 83.2 0.93 0.80 1.08 0.35 0.93 0.80 1.08 0.32 Age Range 14,860 12,287 82.7 0.57 0.37 <60 2,064 1,705 82.6 1.00 1.00 60-69 4,252 3,535 83.1 1.04 0.90 1.19 0.60 1.07 0.93 1.24 0.33 70-79 5,394 4,431 82.2 0.97 0.85 1.11 0.64 0.98 0.85 1.12 0.74 80+ 3,150 2,616 83.1 1.03 0.89 1.19 0.68 0.98 0.83 1.15 0.81 Sex 14,860 12,287 82.7 0.09 0.009 Female 6,760 5,550 82.1 1.00 1.00 Male 8,100 6,737 83.2 1.08 0.99 1.17 0.09 1.12 1.03 1.23 0.009 Histology 14,860 12,287 82.7 <0.001 <0.001 NSCLC 9,027 7,330 81.2 1.00 1.00 SCLC 1,942 1,714 88.3 1.74 1.50 2.02 <0.001 1.76 1.51 2.05 <0.001 Other 3,891 3,243 83.4 1.16 1.05 1.28 0.004 1.04 0.93 1.17 0.47 Year of Diagnosis 14,860 12,287 82.7 <0.001 <0.001 2006 2,501 1,925 77.0 1.00 1.00 2007 2,601 2,159 83.0 1.46 1.27 1.68 <0.001 1.47 1.28 1.70 <0.001 2008 3,140 2,782 88.6 2.33 2.01 2.69 <0.001 2.29 1.98 2.65 <0.001 2009 3,260 2,853 87.5 2.10 1.82 2.41 <0.001 2.44 2.06 2.90 <0.001 2010 3,358 2,568 76.5 0.97 0.86 1.10 0.66 1.04 0.88 1.24 0.64 Co-morbidity score 14,860 12,287 82.7 <0.001 <0.001 0 1,958 1,575 80.4 1.00 1.00 1-2 1,430 1,137 79.5 0.94 0.80 1.12 0.50 0.86 0.73 1.03 0.11 3+ 303 229 75.6 0.75 0.57 1.00 0.05 0.67 0.50 0.90 0.008 CCM missing 5,464 4,664 85.4 1.42 1.24 1.62 <0.001 1.53 1.33 1.77 <0.001 No HES link 5,705 4,682 82.1 1.11 0.98 1.27 0.11 1.12 0.96 1.31 0.14 Stage 14,860 12,287 82.7 <0.001 <0.001 I 783 553 70.6 1.00 1.00 II 403 312 77.4 1.43 1.08 1.89 0.01 1.53 1.15 2.04 0.003 III 1,592 1,343 84.4 2.24 1.83 2.75 <0.001 2.11 1.71 2.61 <0.001 IV 2,369 2,111 89.1 3.40 2.78 4.16 <0.001 2.93 2.38 3.62 <0.001 Missing/ unknown 9,713 7,968 82.0 1.90 1.61 2.23 <0.001 1.64 1.33 2.03 <0.001 Performance Status 14,860 12,287 82.7 <0.001 0.005 0 1,376 1,089 79.1 1.00 1.00 1-2 3,287 2,754 83.8 1.36 1.16 1.60 <0.001 1.23 1.04 1.45 0.02 3-4 1,240 1,107 89.3 2.19 1.76 2.74 <0.001 1.66 1.30 2.12 <0.001 Missing/ unknown 8,957 7,337 81.9 1.19 1.04 1.37 0.01 1.19 0.98 1.44 0.08 Any treatment 14,860 12,287 82.7 <0.001 No 5,063 4,328 85.5 1.00 1.00 Yes 9,797 7,959 81.2 0.74 0.67 0.81 <0.001 0.75 0.66 0.84 <0.001 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS, any treatment 179 Table 11.8. HR of early treatment from diagnosis (for those treated within 1 year of diagnosis and excluding those untreated) 2006 to 2010 Treated Median time to treatment (days) Univariable analysis (n=15,373) Multivariable analysis1 (n=15,373) N N IQR HR 95% CI P HR 95% CI P IMD 15,373 36 22-56 0.70 0.25 1 (least deprived) 1,928 35 21-55 1.00 1.00 2 2,298 35 21-56 1.00 0.94 1.06 0.89 1.00 0.94 1.06 0.95 3 2,605 36 22-57 0.98 0.92 1.03 0.41 0.97 0.92 1.03 0.35 4 3,510 36 22-56 0.97 0.91 1.02 0.22 0.95 0.90 1.00 0.07 5 (most deprived) 5,032 36 22-56 0.98 0.93 1.03 0.38 0.97 0.92 1.02 0.25 Age Range 15,373 36 22-56 <0.001 <0.001 <60 2,941 33 20-50 1.00 1.00 60-69 5,322 35 21-54 0.93 0.89 0.97 0.002 0.96 0.92 1.01 0.10 70-79 5,360 38 23-61 0.79 0.75 0.82 <0.001 0.84 0.80 0.88 <0.001 80+ 1,750 41 23-61 0.74 0.70 0.79 <0.001 0.86 0.81 0.92 <0.001 Sex 15,373 36 22-56 <0.001 0.07 Female 6,918 35 21-55 1.00 1.00 Male 8,455 36 22-57 0.94 0.91 0.97 <0.001 0.97 0.94 1.00 0.07 Histology 15,373 36 22-56 <0.001 <0.001 NSCLC 10,638 40 26-59 1.00 1.00 SCLC 2,615 22 14-33 2.17 2.08 2.27 <0.001 1.96 1.87 2.05 <0.001 Other 2,120 38.5 20-64 0.94 0.89 0.98 0.005 0.98 0.94 1.03 0.45 Diagnosis year 15,373 36 22-56 <0.001 <0.001 2006 2956 35 21-54 1.00 1.00 2007 3006 36 22-57 0.96 0.92 1.01 0.15 0.95 0.90 1.00 0.06 2008 3010 41 25-62 0.87 0.82 0.91 <0.001 0.83 0.79 0.88 <0.001 2009 3262 40 24-60 0.89 0.85 0.94 <0.001 0.88 0.83 0.94 <0.001 2010 3139 29 16-46 1.25 1.19 1.32 <0.001 1.26 1.17 1.35 <0.001 Co-morbidity score 15,373 36 22-56 <0.001 0.34 0 2,323 37 22-57 1.00 1.00 1-2 1,617 40 24-63 0.93 0.87 0.99 0.02 0.94 0.89 1.01 0.08 3+ 352 37.5 20-69 0.92 0.82 1.02 0.12 0.98 0.87 1.09 0.68 CCM missing 5,996 36 22-56 1.02 0.97 1.07 0.40 1.00 0.95 1.05 0.96 No HES link 5,085 34 20-52 1.11 1.06 1.17 <0.001 0.97 0.91 1.03 0.32 GP referral date 15,373 36 22-56 0.28 0.21 No 5,253 35 19-58 1.00 1.00 Yes 10,120 36 23-55 1.01 0.99 1.05 0.28 1.02 0.99 1.06 0.21 Stage 15,373 36 22-56 <0.001 <0.001 I 922 51 32-74 1.00 1.00 II 403 48 32-69 1.07 0.95 1.20 0.27 1.00 0.89 1.12 0.95 III 1,602 39 27-57 1.33 1.22 1.44 <0.001 1.16 1.07 1.26 0.001 IV 2,009 30 20-44 1.84 1.70 1.99 <0.001 1.52 1.40 1.65 <0.001 Missing/ unknown 10,437 35 21-56 1.37 1.28 1.47 <0.001 1.30 1.20 1.41 <0.001 Performance Status 15,373 36 22-56 <0.001 <0.001 0 1,673 37 24-57 1.00 1.00 1-2 3,594 35 22-54 1.05 0.99 1.11 0.11 1.06 1.00 1.13 0.05 3-4 466 28 15-48 1.27 1.15 1.41 <0.001 1.31 1.18 1.46 <0.001 Missing/ unknown 9,640 36 21-57 0.97 0.92 1.02 0.23 0.95 0.88 1.01 0.12 Ist treatment 15,373 36 22-56 <0.001 <0.001 Chemotherapy 6863 30 21-45 1.00 1.00 Surgery 2830 47 20-68 0.65 0.62 0.68 0.65 0.83 0.79 0.87 <0.001 Radiotherapy 5680 40 24-63 0.64 0.62 0.67 0.64 0.75 0.73 0.79 <0.001 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS, first treatment 180 Table 11.9. Hazard ratio of early treatment for 2006-2010 cohort, including those who did not receive treatment, looking at time from diagnosis to first treatment ( n treated =15,373) Univariable analysis (n=28,664) Multivariable analysis1 (n=28,664) Multivariable analysis2 (n=28,664) HR 95% CI P HR 95% CI P HR 95% CI P IMD <0.001 <0.001 <0.001 1 (least deprived) 1.00 1.00 1.00 2 0.96 0.90 1.01 0.14 0.95 0.90 1.01 0.13 0.95 0.89 1.01 0.09 3 0.91 0.86 0.97 0.003 0.90 0.85 0.95 <0.001 0.91 0.86 0.96 0.002 4 0.88 0.83 0.93 <0.001 0.84 0.80 0.89 <0.001 0.85 0.80 0.90 <0.001 5 (most deprived) 0.88 0.83 0.93 <0.001 0.82 0.78 0.86 <0.001 0.83 0.79 0.88 <0.001 Age Range <0.001 <0.001 <0.001 <60 1.00 1.00 1.00 60-69 0.76 0.73 0.79 <0.001 0.81 0.78 0.85 <0.001 0.82 0.78 0.86 <0.001 70-79 0.47 0.45 0.49 <0.001 0.57 0.54 0.59 <0.001 0.58 0.56 0.61 <0.001 80+ 0.18 0.17 0.19 <0.001 0.29 0.27 0.31 <0.001 0.31 0.29 0.32 <0.001 Sex 0.002 0.21 0.29 Female 1.00 1.00 1.00 Male 1.05 1.02 1.09 0.002 0.98 0.95 1.01 0.21 0.98 0.95 1.01 0.29 Histology <0.001 <0.001 <0.001 NSCLC 1.00 1.00 1.00 SCLC 1.50 1.44 1.57 <0.001 1.55 1.49 1.62 <0.001 1.64 1.57 1.72 <0.001 Other 0.21 0.20 0.22 <0.001 0.30 0.28 0.31 <0.001 0.33 0.31 0.34 <0.001 Diagnosis year <0.001 <0.001 <0.001 2006 1.00 1.00 1.00 2007 0.97 0.92 1.02 0.20 0.98 0.93 1.03 0.35 0.97 0.92 1.02 0.24 2008 0.91 0.87 0.96 0.00 0.88 0.83 0.92 <0.001 0.86 0.82 0.91 <0.001 2009 1.02 0.97 1.07 0.39 1.14 1.07 1.22 <0.001 1.10 1.03 1.17 0.004 2010 1.09 1.04 1.15 0.00 1.37 1.28 1.47 <0.001 1.35 1.26 1.44 <0.001 Co-morbidity score <0.001 <0.001 <0.001 0 1.00 1.00 1.00 1-2 0.71 0.66 0.75 <0.001 0.85 0.80 0.91 <0.001 0.86 0.81 0.92 <0.001 3+ 0.56 0.50 0.62 <0.001 0.75 0.67 0.84 <0.001 0.77 0.69 0.86 <0.001 CCM missing 1.03 0.98 1.08 0.27 1.00 0.95 1.05 1.00 1.01 0.96 1.06 0.76 No HES link 0.84 0.80 0.89 <0.001 0.77 0.73 0.82 <0.001 0.78 0.74 0.83 <0.001 GP referral <0.001 <0.001 <0.001 No 1.00 1.00 1.00 Yes 2.02 1.95 2.09 <0.001 1.64 1.58 1.70 <0.001 1.58 1.52 1.63 <0.001 Stage <0.001 <0.001 I 1.00 1.00 II 0.92 0.82 1.04 0.17 0.85 0.75 0.95 0.005 III 0.94 0.87 1.02 0.16 0.87 0.80 0.95 0.001 IV 0.68 0.63 0.73 <0.001 0.78 0.72 0.85 <0.001 Missing/unknown 0.58 0.54 0.62 <0.001 0.84 0.77 0.91 <0.001 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 0.70 0.66 0.74 <0.001 0.90 0.85 0.96 0.001 3-4 0.13 0.12 0.15 <0.001 0.27 0.24 0.30 <0.001 Missing/ unknown 0.36 0.34 0.38 <0.001 0.65 0.61 0.70 <0.001 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, GP referral 2Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, GP referral, stage, PS 181 Table 11.10. Odds of receiving any treatment within 31 days of diagnosis (for those treated within 1 year of diagnosis) 2006-10 Treated Treated within 31 days Univariable analysis (n=15,373) Multivariable analysis1 (n=15,373, R2=9.43) N N % OR 95% CI P OR 95% CI P IMD 15,373 6,537 42.5 0.39 0.29 1 (least deprived) 1,928 849 44.0 1.00 1.00 2 2,298 993 43.2 0.97 0.86 1.09 0.59 0.97 0.85 1.11 0.67 3 2,605 1,089 41.8 0.91 0.81 1.03 0.13 0.89 0.78 1.01 0.06 4 3,510 1,504 42.9 0.95 0.85 1.07 0.40 0.94 0.84 1.06 0.34 5 (most deprived) 5,032 2,102 41.8 0.91 0.82 1.01 0.09 0.91 0.81 1.02 0.11 Age Range 15,373 6,537 42.52 <0.001 <0.001 <60 2,941 1,409 47.9 1.00 1.00 60-69 5,322 2,355 44.3 0.86 0.79 0.94 0.001 0.89 0.81 0.98 0.02 70-79 5,360 2,111 39.4 0.71 0.65 0.77 <0.001 0.75 0.68 0.83 <0.001 80+ 1,750 662 37.8 0.66 0.59 0.75 <0.001 0.78 0.68 0.89 <0.001 Sex 15,373 6,537 42.5 0.001 0.27 Female 6,918 3,045 44.0 1.00 1.00 Male 8,455 3,492 41.3 0.89 0.84 0.95 0.001 0.96 0.90 1.03 0.27 Histology 15,373 6,537 42.5 <0.001 <0.001 NSCLC 10,638 3,765 35.4 1.00 1.00 SCLC 2,615 1,920 73.4 5.04 4.58 5.55 <0.001 4.38 3.94 4.86 <0.001 Other 2,120 852 40.2 1.23 1.11 1.35 <0.001 1.26 1.14 1.39 <0.001 Co-morbidity score 15,373 6,537 42.5 <0.001 0.24 0 2,323 940 40.5 1.00 1.00 1-2 1,617 598 37.0 0.86 0.76 0.98 0.03 0.91 0.79 1.04 0.17 3+ 352 145 41.2 1.03 0.82 1.29 0.80 1.08 0.85 1.38 0.52 CCM missing 5,996 2,482 41.4 1.04 0.94 1.15 0.44 1.02 0.92 1.14 0.68 No HES link 5,085 2,372 46.7 1.29 1.16 1.42 <0.001 1.05 0.92 1.19 0.46 Diagnosis year 15,373 6,537 42.5 <0.001 <0.001 2006 2,956 1,314 44.5 1.00 1.00 2007 3,006 1,250 41.6 0.89 0.80 0.99 0.03 0.88 0.79 0.98 0.02 2008 3,010 1,071 35.6 0.69 0.62 0.77 <0.001 0.64 0.57 0.71 <0.001 2009 3,262 1,196 36.7 0.72 0.65 0.80 <0.001 0.68 0.59 0.79 <0.001 2010 3,139 1,706 54.4 1.49 1.34 1.65 <0.001 1.54 1.32 1.79 <0.001 GP referral 15,373 6,537 42.5 <0.001 <0.001 No GP referral 5,253 2,384 45.4 1.00 1.00 FHA<=14 days 7,178 3,081 42.9 0.91 0.84 0.97 0.006 0.89 0.83 0.97 0.006 FHA >14 days 2,942 1,072 36.4 0.69 0.63 0.76 <0.001 0.73 0.66 0.81 <0.001 1st treatment 15,373 6,537 42.5 <0.001 Chemotherapy 6,863 3,563 51.9 1.00 1.00 Surgery 2,830 903 31.9 0.43 0.40 0.48 <0.001 0.77 0.69 0.86 <0.001 Radiotherapy 5,680 2,071 36.5 0.53 0.49 0.57 <0.001 0.77 0.71 0.84 <0.001 Stage 15,373 6,537 42.5 <0.001 <0.001 I 922 229 24.8 1.00 1.00 II 403 98 24.3 0.97 0.74 1.28 0.84 0.86 0.65 1.14 0.28 III 1,602 580 36.2 1.72 1.43 2.06 <0.001 1.33 1.09 1.61 0.005 IV 2,009 1,065 53.0 3.41 2.87 4.06 <0.001 2.48 2.05 3.00 <0.001 Missing/unknown 10,437 4,565 43.7 2.35 2.02 2.75 <0.001 1.77 1.45 2.14 <0.001 Performance Status 15,373 6,537 42.5 <0.001 0.0001 0 1,673 647 38.7 1.00 1.00 1-2 3,594 1,551 43.2 1.20 1.07 1.36 0.002 1.10 0.96 1.25 0.15 3-4 466 256 54.9 1.93 1.57 2.38 <0.001 1.68 1.33 2.12 <0.001 Missing/ unknown 9,640 4,083 42.4 1.17 1.05 1.30 0.005 1.01 0.87 1.17 0.92 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, GP referral, 1st treatment, stage, PS 182 Table 11.11. HR of early treatment from GP referral (for those treated within 1 year of diagnosis and excluding those untreated) 2006 to 2010 Treat ment Median time to treatment (days) Univariable analysis (n=10,090) Multivariable analysis1 (n=10,090) N N IQR HR 95% CI P HR 95% CI P MD 10,090 56 39-79 0.09 0.049 1 (least deprived) 1,250 54 36-75 1.00 1.00 2 1,460 55 38-79 0.94 0.87 1.01 0.10 0.95 0.88 1.03 0.22 3 1,696 56 39-80.5 0.92 0.85 0.99 0.02 0.92 0.86 0.99 0.03 4 2,306 56 38-79 0.91 0.85 0.98 0.009 0.90 0.84 0.97 0.005 5 (most deprived) 3,378 56 39-79 0.92 0.86 0.98 0.01 0.92 0.86 0.98 0.01 Age Range 10,090 56 39-79 <0.001 <0.001 <60 1,861 51 35-71 1.00 1.00 60-69 3,512 55 37-76 0.94 0.89 0.99 0.02 0.97 0.92 1.03 0.27 70-79 3,531 59 41-84 0.79 0.75 0.84 <0.001 0.85 0.81 0.90 <0.001 80+ 1,186 59 41-85 0.76 0.71 0.82 <0.001 0.86 0.80 0.93 <0.001 Sex 10,090 56 39-79 0.24 0.83 Female 4,508 56 38-78 1.00 1.00 Male 5,582 56 39-79 0.98 0.94 1.02 0.24 1.00 0.97 1.04 0.83 Histology 10,090 56 39-79 <0.001 <0.001 NSCLC 7,241 59 43-82 1.00 1.00 SCLC 1,646 36 26-51 2.19 2.07 2.31 <0.001 1.94 1.83 2.05 <0.001 Other 1,203 62 41-92 0.86 0.80 0.91 <0.001 0.91 0.86 0.97 0.005 Diagnosis year 10,090 56 39-79 0.0001 0.02 2006 1,682 57 40-83 1.00 1.00 2007 1,756 56 39-79 1.09 1.02 1.16 0.01 0.90 0.82 0.98 0.01 2008 2,090 55 37-77 1.10 1.03 1.18 0.003 0.87 0.74 1.03 0.11 2009 2,262 56 39-78 1.12 1.05 1.19 0.001 1.07 1.01 1.15 0.04 2010 2,300 55 38-76 1.16 1.09 1.24 <0.001 1.02 0.94 1.10 0.69 Co-morbidity score 10,090 56 39-79 <0.001 <0.001 0 1,408 57 40-80 1.00 1.00 1-2 892 62 41-89.5 0.90 0.82 0.98 0.01 1.06 0.99 1.13 0.11 3+ 163 56 40-89 0.88 0.75 1.04 0.14 1.07 1.00 1.14 0.05 CCM missing 3,943 55 38-77 1.07 1.01 1.14 0.03 1.15 1.06 1.26 0.001 No HES link 3,684 55 38-76 1.11 1.05 1.18 0.001 1.16 1.06 1.28 0.001 Stage 10,090 56 39-79 <0.001 <0.001 I 665 71 57-100 1.00 1.00 II 332 67 54-92 1.14 1.00 1.30 0.05 1.07 0.94 1.22 0.30 III 1,239 56 42-74 1.60 1.46 1.76 <0.001 1.27 1.16 1.41 <0.001 IV 1,454 44 31-61 2.23 2.04 2.45 <0.001 1.69 1.53 1.86 <0.001 Missing/ unknown 6,400 56 38-80 1.46 1.35 1.59 <0.001 1.26 1.14 1.39 <0.001 Performance Status 10,090 56 39-79 <0.001 <0.001 0 1,305 58 42-76 1.00 1.00 1-2 2,629 54 38-73 1.07 1.00 1.14 0.05 1.03 0.96 1.10 0.44 3-4 311 42 25-64 1.39 1.23 1.57 <0.001 1.37 1.20 1.55 <0.001 Missing/ unknown 5,845 57 39-82 0.95 0.89 1.00 0.07 0.97 0.89 1.05 0.40 1st treatment 10,090 56 39-79 <0.001 <0.001 Chemotherapy 4715 48 34-65 1.00 1.00 Surgery 1856 70 56-94 0.54 0.51 0.57 <0.001 0.69 0.65 0.74 <0.001 Radiotherapy 3519 58 40-85 0.65 0.63 0.68 <0.001 0.80 0.76 0.84 <0.001 1Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS, 1st treatment 183 Table 11.12. Odds of receiving any treatment within 62 days of GP referral (for those with FHA within 1 year of referral, diagnosis within 1 year of FHA and treated within 1 year of diagnosis) Treat ment Treated within 62 days Univariable analysis (n=10,090) Multivariable analysis1 (n=10,090, R2=8.75) N N % OR 95% CI P OR 95% CI P IMD 10,090 6,232 61.8 0.22 0.08 1 (least deprived) 1,250 806 64.5 1.00 1.00 2 1,460 902 61.8 0.89 0.76 1.04 0.15 0.89 0.75 1.05 0.16 3 1,696 1,033 60.9 0.86 0.74 1.00 0.05 0.84 0.71 0.98 0.03 4 2,306 1,399 60.7 0.85 0.74 0.98 0.03 0.80 0.69 0.94 0.005 5 (most deprived) 3,378 2,092 61.9 0.90 0.78 1.03 0.11 0.88 0.76 1.01 0.07 Age Range 10,090 6,232 61.8 <0.001 <0.001 <60 1,861 1,263 67.9 1.00 1.00 60-69 3,512 2,241 63.8 0.83 0.74 0.94 0.003 0.88 0.78 1.00 0.06 70-79 3,531 2,033 57.6 0.64 0.57 0.72 <0.00 0.70 0.62 0.80 <0.001 80+ 1,186 695 58.6 0.67 0.58 0.78 <0.00 0.77 0.65 0.91 0.002 Sex 10,090 6,232 61.8 0.53 0.11 Female 4,508 2,769 61.4 1.00 1.00 Male 5,582 3,463 62.0 1.03 0.95 1.11 0.53 1.07 0.98 1.17 0.11 Histology 10,090 6,232 61.8 <0.001 NSCLC 7,241 4208 58.1 1.00 1.00 SCLC 1,646 1393 84.6 3.97 3.44 4.57 <0.001 3.05 2.62 3.55 <0.001 Other 1,203 631 52.5 0.80 0.70 0.90 <0.001 0.87 0.76 0.99 0.03 Co-morbidity score 10,090 6,232 61.8 <0.001 0.0002 0 1,408 838 59.5 1.00 1.00 1-2 892 469 52.6 0.75 0.64 0.89 0.001 0.78 0.65 0.94 0.008 3+ 163 89 54.6 0.82 0.59 1.13 0.23 0.83 0.59 1.17 0.29 CCM missing 3,943 2,443 62.0 1.11 0.98 1.25 0.11 1.13 0.98 1.30 0.10 No HES link 3,684 2,393 65.0 1.26 1.11 1.43 <0.001 1.07 0.91 1.25 0.44 Diagnosis year 10,090 6,232 61.8 0.0003 2006 1,682 983 58.4 1.00 1.00 2007 1,756 1,032 58.8 1.01 0.88 1.16 0.85 1.01 0.87 1.16 0.90 2008 2,090 1,305 62.4 1.18 1.04 1.35 0.01 1.13 0.99 1.30 0.08 2009 2,262 1,410 62.3 1.18 1.03 1.34 0.01 1.24 1.04 1.50 0.02 2010 2,300 1,502 65.3 1.34 1.18 1.52 <0.001 1.39 1.14 1.69 <0.001 1st treatment 10,090 6,232 61.8 <0.001 Chemotherapy 4,715 3,441 73.0 1.00 1.00 Surgery 1,856 718 38.7 0.23 0.21 0.26 <0.001 0.38 0.34 0.44 <0.001 Radiotherapy 3,519 2,073 58.9 0.53 0.48 0.58 <0.001 0.76 0.69 0.85 <0.001 Stage 10,090 6,232 61.8 <0.001 I 665 257 38.7 1.00 1.00 II 332 147 44.3 1.26 0.97 1.65 0.09 1.07 0.81 1.41 0.64 III 1,239 799 64.5 2.88 2.37 3.50 <0.001 1.56 1.27 1.93 <0.001 IV 1,454 1,136 78.1 5.67 4.65 6.92 <0.001 2.74 2.20 3.41 <0.001 Missing/unknown 6,400 3,893 60.8 2.47 2.09 2.90 <0.001 1.63 1.31 2.03 <0.001 Performance Status 10,090 6,232 61.8 0.01 0 1,305 790 60.5 1.000 1.00 1-2 2,629 1,747 66.5 1.29 1.13 1.48 <0.001 1.07 0.92 1.24 0.40 3-4 311 226 72.7 1.73 1.32 2.28 <0.001 1.31 0.97 1.77 0.08 Missing/ unknown 5,845 3,469 59.4 0.95 0.84 1.08 0.43 0.88 0.73 1.05 0.16 1Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS, 1st treatment 184 11.4 Discussion 11.4.1 Principal Findings Socio-economic inequalities in the time intervals from GP referral to FHA, diagnosis, and treatment were found, but socio-economic inequalities were not found in the time intervals from FHA to diagnosis [results not shown], or from diagnosis to treatment. This would suggest that interval inequalities originate from the GP referral to FHA interval. However, a linear trend was not seen. Patients in the middle SEP groups were less likely to receive early FHA, diagnosis or treatment from GP referral, compared to the highest SEP group, but this was not consistently observed for the lowest SEP group. Generally, those with SCLC, those with poorer PS and those with more advanced stage cancer were more likely to receive FHA, diagnosis and treatment within guidelines. Older patients were less likely to receive first treatment within guidelines. Patients undergoing surgery and radiotherapy as their first treatment had longer time to treatment than those receiving chemotherapy as a first treatment. 11.4.2 Strengths and weaknesses To my knowledge this is the first study to examine factors associated with timely care within UK guidelines for referral, diagnosis and treatment for lung cancer. In order to accurately determine whether there are inequalities in referral, diagnostic and treatment intervals consistent recording of dates is required. It is likely that FHA and treatment dates are accurately recorded but as a number of different dates can be used for the diagnosis date then, unless there is consistent application of the rules, this could affect calculation of the diagnostic and treatment intervals and thus introduce error. There is no evidence to suggest that the accuracy of recording date of diagnosis is likely to vary by SEP and introduce bias, however. As the results for time from GP referral to diagnosis and from diagnosis to treatment for those diagnosed in 2010 seemed markedly different to those of earlier years it may be that the different methodology used for determining date of diagnosis in 2010 contributed to these observed differences. It is difficult to say, therefore, whether the likelihood of receiving quicker first treatment in 2010 is a genuine system 185 improvement or not. However, using the 2006-2009 dataset gave similar results overall. I examined the interim time interval from FHA to diagnosis. However, the FHA and diagnosis date often appeared to be recorded as the same date [there was a median interval of 0 with 0-0 IQR for 2006-2010 data], which may relate to problems with variability in recording of date of diagnosis in the registry dataset. There did not appear to be socio-economic inequalities within this diagnostic interval but it was difficult to draw any firm conclusions from this analysis when the median time interval in all SEP groups was 0. It is of course possible that, for those without a GP referral date, the FHA may be an emergency hospital visit where diagnosis is made on that day. Patients who are diagnosed clinically may also have their first clinical appointment date as the date of diagnosis. A UK study also found that the median time from FHA to diagnosis was 0 days [with mean =11.5 days] for lung cancer and was also 0 days in the five other cancer types they looked at, which would suggest that the majority of patients do find out their diagnosis at the FHA (123). The type of referral route may influence time to diagnosis and treatment. More than half of patients did not have a GP referral date recorded but it was not possible to determine from this data how many of these patients had been referred via a different route and how many may have had a GP referral for which the date was missing. However a large study examining routes to diagnosis found that only 41% of lung cancer patients were referred via their GP and a high proportion of cases (39%) presented as an emergency (255). It was also not possible to determine which patients were urgent GP referrals. However, those who did have a GP date had longer time to treatment than those who did not [a finding similar to that seen in one other study that looked at this for lung cancer (123)], which does suggest that those without a GP date may be presenting as emergency cases and thus receive more urgent investigation and treatment. In this study I was not able to examine inequalities in the primary care interval [from first patient presentation to the GP until the referral was made], as date of first presentation to GP is not recorded within the registry dataset. I was able to look at 186 patient delay by using stage at diagnosis and number of DCO cases as proxy measures of patient delay. No evidence for inequalities in delay in patient presentation by SEP was found for the 2006-2010 cohort (chapter 9), in a univariable analysis. 11.4.3 Interpretation of results and comparison with other studies 11.4.3.1 Diagnostic delay Two studies from Sweden (73) and Denmark (108), (which have similar healthcare systems to the UK), found socio-economic inequalities in time from referral to diagnosis, with higher SEP patients having more timely diagnosis. I found socio-economic inequalities from GP referral to diagnosis and, within this interval, from GP referral to FHA but not from FHA to diagnosis. A similar result was found in a study which examined socio-economic inequalities in these intervals for pancreatic cancer (29) and a UK study of over 3,000 lung cancer patients also found that socio-economic position was not associated with time from FHA to diagnosis (106). Therefore the evidence does suggest that socio-economic diagnostic inequalities are found in the referral to FHA interval rather than the secondary care FHA to diagnostic interval. 11.4.3.2 Referral delay Patients in England and Wales with suspected cancer should be seen by a specialist within two weeks (24), as an urgent referral. A small 2007 study of 247 UK lung cancer patients found that all patients were referred within this 2 week interval (125). In my study, the median time from GP referral to first hospital appointment was 10 days and 70% of patients were seen in secondary care within the 14 day guideline target, although I did not have details on what type of GP referral these were. A non-linear trend in socio-economic inequalities in the likelihood of being referred to secondary care within time interval guidelines was found. Those in the middle SEP groups were less likely to receive timely referral than those in the top SEP group, but this was not seen for those in the lowest SEP group. The Waiting Time Paradox [WTP] suggests that sicker people are referred more quickly and, as they are more ill, actually have shorter survival (179). In our study we found that poorer performance status did result in increased likelihood of earlier referral and diagnosis. This may help to explain why those in the lowest SEP group were not 187 significantly less likely to receive a timely referral or diagnosis compared to the highest SEP group. Patients with lower SEP had poorer PS [see chapter 9] and so were likely to be in generally poorer health. The results of this ‘sicker quicker’ effect may, therefore, act to effectively ‘cancel out’ any system inequalities that might result in longer referral and diagnosis time intervals for lower SEP patients. Those in the bottom SEP group are more likely to receive earlier referral due to being more ill and presenting as an emergency, whereas those at the top are better able to seek urgent referral. It has been suggested that socio-economic differences in communicating and presenting cancer symptoms to health professionals may result in longer delays for those who are less ‘convincing’ (111). Lower SEP patients may be less articulate whereas more educated cancer patients are better able to describe symptoms and thus speed up the referral process (104). Differences in health literacy of patients by SEP may contribute to this, where better-educated, more health-literate patients have looked up their symptoms on-line, found out the correct vocabulary and what might be important to mention to the doctor. Doctors may also relate better to wealthier, better educated patients [as they are more similar to themselves in terms of social class and culture] and this may result, possibly unintentionally, in a more rapid investigation process (104). In a previous study age, sex, ethnicity and marital status did not appear to influence the likelihood of urgent referral for lung cancer (124). We also found that sex was not associated with likelihood of early referral and no clear pattern of referral by age was found. Patients subsequently diagnosed with earlier-stage cancer and those with good PS had a lower likelihood of being referred within 14 days. It could be argued that it is not appropriate to include stage as a variable in the analysis of time from referral to FHA, as stage is not yet known at this time. A similar argument can also be made for the inclusion of performance status here as, again, this is only ascertained post-FHA. However, stage, PS and co-morbidity score can all be considered as proxy measures for how ill a patient is. Those with earlier stage cancer may not have so many obvious clinical symptoms as patients with later stage cancer. A GP will be prompted to investigate or refer by signs and symptoms of possible cancer. Urgent referral requires 188 the presence of ‘alarm’ symptoms that are more likely to be present as the disease advances. A previous study found that co-morbidity delayed diagnosis in around 20% of lung cancer patients seen by GPs (256). In patients with a number of co-morbidities, lung cancer may not be suspected possibly because symptoms (for example cough, weight loss breathlessness) may be ascribed to a known comorbidity [for example COPD] rather than lung cancer. There is some evidence to suggest that those with lower SEP are likely to have more co-morbidity. One study found that more deprived cancer patients had a 50% higher risk of serious co-morbidity compared to less deprived [high SEP] patients (69). Patients in the lowest SEP group in this study had more co-morbidity and poorer PS than those in the higher SEP groups [table 9.3]. However, no clear-cut pattern was observed in referral or diagnosis time by co-morbidity score in this study, but those with poorer performance status had increased odds of timely FHA and diagnosis. Both CCM score and PS can be used as proxy measures of general wellbeing/sickness and it is difficult to say how well they capture this. It may be that the number of co-morbidities is not a particularly sensitive measure and PS may be a better marker. However, it may also depend on a patient’s prior history as to how a GP interprets their symptoms. If someone regularly visits their GP presenting with multi-morbidities [as measured by CCM score] then the GP may be reluctant to refer that person for yet more testing and the patient may also be reluctant to do so. However if someone with few or no previous comorbidities appears very ill [as measured by performance status] then they may be more likely to be referred for investigation. 11.4.3.3 Treatment delay A previous study that looked at time to treatment targets for lung cancer found that these were not being achieved (101) (125). We found that time to treatment targets were only being met for 43% of patients when measured from diagnosis and 62% when measured from GP referral date [although the type of referral route was not known and could include those not referred under the two-week wait route]. However, some of the reasons for longer delay may be valid. Those who receive curative surgery may undergo a longer period of preliminary investigation with a 189 greater number of diagnostic and staging procedures, as well as an assessment for fitness for surgery (125). In this study patients who underwent surgery as a first treatment waited a median of 47 days from diagnosis compared to 30 days for chemotherapy and 40 days for radiotherapy. Lower SEP was associated with a longer time to treatment when including both those who did, and did not, receive treatment. This may be because higher SEP patients are more likely to receive treatment. Socio-economic inequalities in receipt of lung cancer treatment were found in my systematic review and meta-analysis (236) and have also been shown for this dataset [see chapter 10]. Late stage and poor performance status were also associated with longer time to treatment. Again, this may be because those with late stage and poorer performance are less likely to receive treatment and thus the time interval appears long as this is censored at the end of the time period rather than treatment being obtained in the interim. However, when only those patients who received treatment were included in the analysis then SEP was not associated with early treatment from diagnosis. However, in the GP referral to treatment interval, those in the middle SEP groups appeared less likely to receive treatment within the 62 day guidelines, compared to the highest SEP group. Those with later stage cancer and poorer PS were more likely to receive timely treatment within guidelines. Again my previous explanation may be relevant where those in the bottom SEP group are more likely to receive earlier referral and treatment as, due to poorer general health, they are more likely to proceed more speedily through the system and undergo quicker, non-curative treatment such as chemotherapy, whereas those in the top SEP group are more likely to receive curative surgery for which there is a longer interval but they are able to obtain this earlier. Those in the middle groups who are either not so obviously ill or are less able to communicate thus wait the longest for referral and treatment. 11.4.4 Implications for policy and practice, and further research The consistent and accurate recording of GP referral, FHA, diagnosis and first treatment dates by cancer registry staff is important if inequalities in referral, diagnosis and treatment intervals are to be identified. The interval from FHA to diagnosis could not be examined. Due to changes in the way that date of diagnosis was 190 calculated it was difficult to be sure that any differences seen over time were system changes as a result of the introduction of targets or just due to the change in rules for determining date of diagnosis. Similar studies using other data would be useful to verify this. Our findings suggest that lower SEP patients are likely to appear more ill [as measured by PS] and the WTP ‘sicker quicker’ effect may neutralise any patient and system inequalities that might result in longer referral and diagnosis time intervals for lower compared to higher SEP patients, potentially relating to poorer communication and health literacy. Better evidence of inequalities in health literacy by SEP is required. It would also be useful to know more about patient and primary care delay prior to GP referral, and to determine whether socio-economic inequalities are found in these intervals. As around 40% of lung cancer patients present as emergency admissions and there is some evidence that lower SEP patients are more likely to present via this route (77), further investigation into SEP and route to diagnosis is also required. 11.5 Chapter summary Socio-economic inequalities in the time intervals from GP referral to FHA, diagnosis, and treatment were found, but socio-economic inequalities were not found in the interim intervals from FHA to diagnosis or from diagnosis to treatment. This would suggest that interval inequalities originate from the GP referral to FHA interval. However, a linear trend was not seen. Patients in the middle SEP groups were less likely to receive early FHA, diagnosis or treatment from GP referral, compared to the highest SEP group, but this was not consistently observed for the lowest SEP group. It is possible that the WTP ‘sicker quicker’ effect may effectively ‘cancel out’ system inequalities that might result in longer time intervals for lower SEP patients. The main factors that determined early referral, diagnosis and treatment appeared to be how sick patients appeared, as patients with poor PS and late-stage cancer were more likely to receive early referral and treatment. However, older patients, who were more likely to have poor performance status, were less likely to receive early treatment, indicating possible age-discrimination. 191 Socio-economic inequalities in lung cancer survival will now be examined in chapter 12. 192 Chapter 12. Inequalities in survival Summary Background Lung cancer survival is socio-economically patterned and socio-economic inequalities in receipt of treatment have been demonstrated. In England, target waiting times for the referral interval [14 days] and treatment intervals [31 days from diagnosis and 62 days from GP referral] have been set and socio-economic inequalities in the time intervals from GP referral have been found. The contribution of these inequalities to socio-economic inequalities in lung cancer survival is unclear. Cancer registry [NYCRIS], Hospital Episode Statistics [HES] and lung cancer audit [LUCADA] data-sets were linked in order to investigate the factors that may influence socio-economic inequalities in survival. Methods NYCRIS data for 36,477 patients diagnosed between 1999 and 2005 and 28,733 diagnosed between 2006 and 2010 were analysed. Survival time [in weeks] was calculated as the interval from the date of diagnosis to the date of death or to the end of follow up at 31/12/2011 for those still alive. Cox regression models were used to calculate hazard of death, in relation to SEP for those diagnosed in 1999-2005 and 2006-2010. Logistic regression was used to examine the likelihood of still being alive two years after diagnosis for those diagnosed in in 1995-2005 and in 2006-2009. Results Those in the lowest SEP group had a significantly higher risk of death. Socio-economic inequalities in survival were no longer found when treatment was included in the model. Only 15.3% of patients were still alive 2 years after diagnosis but this increased to 70% for those who had surgery. Patients in the lowest SEP group were significantly less likely to still be alive after 2 years, compared to the highest SEP group, in a multivariable analysis adjusted for age, sex, histology, year of diagnosis, GP referral 193 date, co-morbidity, stage and PS [OR=0.78, 95% CI 0.69 to 0.88, p<0.001]. Adding timely-referral did not substantially change this. However if treatment was included the association no longer remained significant [OR=0.87, 95% CI 0.75 to 1.00, p=0.06] but further addition of timeliness-of-treatment made no difference to the outcome. Patients treated later than the treatment target times had better 2 year survival and lower risk of death than those who received timely treatment. Conclusions Socio-economic inequalities in survival are statistically explained by inequalities in receipt of treatment but not by time to treatment in this cohort. However, patients who were treated within the time-to-treatment guideline targets had poorer survival compared to those who had later treatment. Sicker patients had quicker referral through the care pathway and this ‘sicker quicker’ effect may mask any system inequalities that might otherwise result in longer time intervals in lower SEP patients. Interventions that address socio-economic inequalities in receipt of treatment are likely to reduce socio-economic inequalities in survival and thus improve survival rates overall. 194 12.1 Introduction Survival from lung cancer is socio-economically patterned (135). Socio-economic inequalities in receipt of lung cancer treatment have been demonstrated in a systematic review and meta-analysis (236) conducted as part of this PhD thesis [chapter 7] as well as in the secondary data analysis [chapter 10]. It has been suggested that inequalities in receipt of treatment may at least partially contribute to inequalities in outcome (26) although there is little definitive evidence to support this. Socio-economic inequalities in the time intervals from GP referral to FHA, diagnosis and treatment have also been found [chapter 11] but no linear pattern by SEP emerged. Again it is not known what role inequalities in referral, diagnostic and treatment time may play in survival inequalities. Cancer registry [NYCRIS], Hospital Episode Statistics [HES] and lung cancer audit [LUCADA] data-sets were linked in order to investigate the factors that may influence socio-economic inequalities in survival for lung cancer patients, specifically examining the influence of socio-economic inequalities in receipt of treatment, and in timely GP referral and treatment, taking into account age, sex, histology, year of diagnosis, co-morbidity, stage and performance status. 12.2 Methods 12.2.1 Data For details of data sources and variables see chapter 8, sections 8.2 and 8.4. The following variables were included in multivariable analyses: SEP, age, sex, histology, year of diagnosis, co-morbidity, stage, PS, type of treatment [no treatment, surgery, surgery + chemotherapy or radiotherapy, chemotherapy, chemotherapy and radiotherapy, radiotherapy], timely GP referral [FHA within 14 days of GP referral, FHA>14 days from GP referral, or no GP referral date], timely first treatment [first treatment within 31 days of diagnosis, or first treatment >31 days from diagnosis]. 12.2.2 Analysis Kaplan Meier graphs were used to examine univariable influences on all-cause mortality. Cox regression models were used to calculate hazard ratios [HRs] and 95% CIs for all-cause mortality in relation to SEP, in multivariable models for those 195 diagnosed in 1999-2005 and 2006-2010. As date of diagnosis was determined in a different way in 2010 from previous years then analysis of the 2006-2009 dataset [n=22,967] and of the 2006-2009 dataset with stage recorded [n=5233] was also carried out. Nelson-Aalen plots were used to check the proportional hazard assumptions for the potential explanatory variables in each model. Logistic regression was used to examine the likelihood of still being alive two years after diagnosis, by SEP, in the 1995-2005 cohort and in the 2006-2009 dataset. Interactions between SEP and histology, and SEP and type of treatment received, were explored. The R2 statistic was examined to determine the amount of variance in survival explained by each model. 12.3 Results 12.3.1 Descriptive statistics Similar results were found in the 1999-2005 cohort compared to those for 2006-2010 and 2006-2009 and so to avoid repetition the 1999-2005 tables are found in appendix D. Figs 12.1 – 12.13 show univariable Kaplan-Meier survival curves for patients diagnosed between 2006 and 2010, plotting proportion of patients surviving [y axis] over time [x axis]. Results for 1999-2005 and for 2006-2009 all showed similar patterns [results not shown]. Median survival was 24 weeks in the least deprived SEP group and 21 in the most deprived [table 12.1]. Although survival inequalities by SEP were found [Fig 12.1] these were less pronounced than for survival inequalities by age [Fig 12.2] and sex [Fig 12.3]. Survival inequalities by tumour type were observed [Fig 12.4] with the best survival found for the small number of patients who have a non-standard histology code [other-specified histology], with 41% of these still alive at 2 years compared to 25.3% for adenocarcinoma and 24% for those diagnosed with squamous cell carcinoma in the 2006-2009 cohort. Only 8.3% of those diagnosed with SCLC were still alive at 2 years [table 12.4]. 196 Figure 12.1. Kaplan-Meier survival curve: SEP Figure 12.2. Kaplan-Meier survival curve: age 0.000.250.500.751.000 100 200 300analysis time (weeks)imd = 1 imd = 2imd = 3 imd = 4imd = 5Kaplan-Meier survival estimates by SEP group (IMD), 2006-2010 cohort0.000.250.500.751.000 100 200 300analysis time (weeks)age = <60 age = 60-69age = 70-79 age = 80+Kaplan-Meier survival estimates by age, 2006-2010 cohort197 Figure 12.3. Kaplan-Meier survival curve: sex Figure 12.4. Kaplan-Meier survival curve: tumour type 0.000.250.500.751.000 100 200 300analysis time (weeks)Female MaleKaplan-Meier survival estimates by sex, 2006-2010 cohort0.000.250.500.751.000 100 200 300analysis time (weeks)adenocarcinoma largeneoplasm non-smallother specified SCLCsquamous unspecifed carcinomaKaplan-Meier survival estimates by tumour type, 2006-2010 cohort198 Survival also varied markedly by performance status [Fig 12.5] with a median survival of 64 weeks for those with good PS [PS=0] compared to 34 weeks with PS=1-2 and 8 weeks for those with poor PS [PS=3-4] [table 12.1]. However, survival differences by number of co-morbidities was less pronounced [Fig 12.6] with only small median survival differences found between those who had no co-morbidities recorded [23 weeks], 1-2 co-morbidities [20 weeks] and 3 or more co-morbidities [18 weeks] [table 12.1]. Patients who underwent surgery had the best survival [Fig 12.7]. Interestingly survival one year after diagnosis was better for those who surgery plus chemotherapy/radiotherapy compared to surgery alone, but this reversed shortly after, with those having surgery alone having the best overall survival [73.0% at 2 years compared to 64.4% in the multiple treatment group for the 2006-2009 data-set] [table 12.4]. Fig 12.8 shows survival by type of, and time to, first treatment. Those who had surgery later than 31 days from diagnosis had the best survival although after four years those who had surgery within the 31 day target had similar survival. Patients who had radiotherapy as a 1st treatment within the 31 day target from diagnosis had similar survival to those who had no treatment, after one year. Those who had first treatment within 31 days had poorer survival [median survival 32 weeks] compared to those who had first treatment later than 31 days from diagnosis [median survival 55 weeks] [Fig 12.9]. Figs 12.10-12.12 show survival by receipt of and time to, surgery, chemotherapy and radiotherapy respectively and again it can be seen that those who had treatment later than the 31 day target from diagnosis had better survival than those who had timely treatment. For chemotherapy and radiotherapy the best survival was found for those who had treatment more than 62 days after diagnosis but these graphs do not take into account whether any other treatment was also received. Fig 12.13 shows survival by stage. Patients with early stage cancer had better survival than those with later stage cancer. 199 Figure 12.5. Kaplan-Meier survival curve: PS Figure 12.6. Kaplan-Meier survival curve: co-morbidity 0.000.250.500.751.000 100 200 300analysis time (weeks)PS = 0 PS = 1-2PS = 3-4 PS= missingKaplan-Meier survival estimates by Performance Status (PS), 2006-2010 cohort0.000.250.500.751.000 100 200 300analysis time (weeks)No co-morbidities 1-2 comorbidities3+ comorbdities CCM score missingNo HES linkKaplan-Meier survival estimates by co-morbidity (CCM) score, 2006-2010 cohort200 Figure 12.7. Kaplan-Meier survival curve: treatment type Figure 12.8. Kaplan-Meier survival curve: time to, and type of, treatment 0.000.250.500.751.000 100 200 300analysis time (weeks)surgery surgery + chemo/radiotherapychemotherapy chemo + radiotherapyradiotherapy no treatmentKaplan-Meier survival estimates by treatment type, 2006-2010 cohort0.000.250.500.751.000 100 200 300analysis time (weeks)no treatment chemo 1st, > 31 days from diagnosischemo 1st, <=31 days from diagnosis surgery 1st, > 31 days from diagnosissurgery 1st, <=31 days from diagnosis radio 1st, > 31 days from diagnosisradio 1st, <=31 days from diagnosisKaplan-Meier survival estimates by type of, and time to, first treatment, 2006-2010 cohort201 Figure 12.9. Kaplan-Meier survival curve: time to first treatment Figure 12.10. Kaplan-Meier survival curve: timely surgery 0.000.250.500.751.000 100 200 300analysis time (weeks)first treatment > 31 days from diagnosis first treatment within 31 days of diagnosisno treatmentKaplan-Meier survival estimates by time to first treatment, 2006-2010 cohort0.000.250.500.751.000 100 200 300analysis time (weeks)no surgery surgery within 31 days of diagnosissurgery 32-62 days from diagnosis surgery >62 days from diagnosisKaplan-Meier survival estimates by receipt of, and time to, surgery, 2006-2010 cohort202 Figure 12.11. Kaplan-Meier survival curve: timely chemotherapy Figure 12.12. Kaplan-Meier survival curve: timely radiotherapy 0.000.250.500.751.000 100 200 300analysis time (weeks)no chemotherapy chemo within 31 days of diagnosischemo 32-62 days from diagnosis chemo >62 days from diagnosisKaplan-Meier survival estimates by receipt of, and time to, chemotherapy, 2006-2010 cohort0.000.250.500.751.000 100 200 300analysis time (weeks)no radiotherapy radiotherapy within 31 days of diagnosisradio 32-62 days from diagnosis radio >62 days from diagnosisKaplan-Meier survival estimates by receipt of, and time to, radiotherapy, 2006-2010 cohort203 Figure 12.13. Kaplan-Meier survival curve: stage 12.3.2 Risk (HR) of all-cause mortality 12.3.2.1 Descriptive analysis: 2006-2010 cohort Median survival in this cohort was 21 weeks from diagnosis [IQR 6-59.3] [table 12.1]. Socio-economic inequalities in survival were found in the unadjusted analyses. In the univariable Cox regression analyses those in the most deprived group had a significantly higher risk [HR] of death than those in the most affluent [HR=1.10, 95% CI 1.05 to 1.14, p<0.001] [table 12.2]. Those in the oldest age group, men, those with no treatment, later stage cancer and poorer performance status all had significantly higher risk of death. Patients who had their first treatment greater than 31 days after diagnosis had better survival compared to those who had treatment within the 31 day guideline [median survival of 55 weeks compared to 32], as did those who had a FHA greater than 14 days from GP referral compared to those who had referral within guidelines [median survival of 37 compared to 30 weeks] [table 12.1]. Those in the no treatment group had the poorest survival. 0.000.250.500.751.000 100 200 300analysis time (weeks)stage = 1 stage = 2stage = 3 stage = 4stage = missingKaplan-Meier survival estimates by stage, 2006-2010 cohort204 Table 12.1. Median survival (weeks), by selected patient, tumour and system factors for the cohort diagnosed between 2006 and 2010, and for the 2006-2010 subset with stage (DCO cases excluded) Variable 2006-2010 Cohort Median survival (weeks) 2006-2010 Cohort with stage Median survival (weeks) N N IQR N N IQR Deprivation quintile 28,733 21 6-59.3 7769 30 10-68.1 1 (least deprived) 3389 24 7-64 931 34 11-76 2 4178 22 6-60.3 1118 33 10-71 3 4948 21 6-59 1300 30 10-68.6 4 6710 20 6-58 1831 28 9-64.1 5 (most deprived) 9608 21 6-58.6 2589 29 9-67.7 Age group 28,733 21 6-59.3 7769 30 10-68.1 <60 3682 36 12-78 1041 41 16-77 60-69 7595 30 8-72 2189 39 12-80 70-79 10248 21 6-59 2843 28 9-67 80+ 7208 12 3-37 1696 18 6-51 Sex 28,733 21 6-59.3 7769 30 10-68.1 Female 13254 22 6-63 3559 31 11-72 Male 15479 21 6-56 4210 29 9-66 Histology 28,733 21 6-59.3 7769 30 10-68.1 NSCLC 15123 32 11-74.7 5116 38 14-78 SCLC 3495 24 7-50 582 29 8-54 Other 10115 9 2-36 2071 14 4-49 Year of Diagnosis 28,733 21 6-59.3 7769 30 10-68.1 2006 5533 20 6-57 671 36 12-85 2007 5712 21 6-60 866 34.5 11-88 2008 5851 21 6-58 1556 27 10-71 2009 5871 23 6-67 2140 30 9-85 2010 5766 22 6-58 2536 29 8-60.7 Co-morbidity Score 28,733 21 6-59.3 7769 30 10-68.1 0 4010 23 7-65 995 30 11-75 1-2 3531 20 6-57 857 28 9-75 3+ 934 18 4-54 226 21 7-61 CCM missing 10175 22 7-61 1977 32 11-79 No HES link 10083 21 5-58.6 3714 30 9-64.6 Timely GP referral 28,733 21 6-59.3 7769 30 10-68.1 No GP referral date 13281 12 3-42 2418 19 6-57 FHA<=14 days 11019 30 9-66 4087 33 11-71 FHA>14 days 4433 37 13-78 1264 44 17-79.3 Stage 28,733 21 6-59.3 7769 30 10-68.1 I 1186 88 56-139.6 1186 88 56-139.6 II 552 61.5 32-99.6 552 61.5 32-99.6 III 2273 40 17-72.1 2273 40 17-72.1 IV 3758 14 5-35 3758 14 5-35 missing 20964 19 5-55 -- -- -- Performance Status 28,733 21 6-59.3 7769 30 10-68.1 0 1842 64 35-108 1493 65 36-107.6 1-2 4865 34 14-69.9 3870 35 14-70 3-4 2178 8 3-21 1763 8 3-20 Missing 19848 18 5-54 643 22 7-64.6 205 Table 12.1 (cont). Median survival (weeks), by selected patient, tumour and system factors for the cohort diagnosed between 2006 and 2010, and for the 2006-2010 subset with stage (DCO cases excluded) Variable 2006-2010 Cohort Median survival (weeks) 2006-2010 Cohort with stage Median survival (weeks) N N IQR N N IQR Type of treatment 28,733 21 6-59.3 7769 30 10-68.1 No treatment 13291 6 2-19 2815 9 4-23 Surgery 1833 120.4 66-194.3 579 103.1 64.7-153.5 Surgery + chemotherapy/ radiotherapy 1061 103 64.3-166.4 422 96.5 64.1-150.7 Chemotherapy 3404 30 15-55 1051 32 16-56 Chemotherapy+radiotherapy 4117 53 34-84 1329 55 35-82 Radiotherapy 5027 25 11-57 1573 28 12-63 Timely 1st treatment 28,733 21 6-59.3 7769 30 10-68.1 >31 days from diagnosis 8887 55 27-108.6 2979 58 29-104.6 <=31 days from diagnosis 6555 32 13-65 1975 36 15-67.3 12.3.2.2 Multivariable regression analysis: 2006-2010 cohort In the multivariable analysis including age and sex a survival gradient by SEP was seen [HR=1.13, 95% CI 1.09 to 1.18, p<0.001]. This remained when histology, year of diagnosis, co-morbidity and timely GP referral were taken into account [HR=1.13, 95% CI 1.09 to 1.18, p<0.001] [table 12.2] and on addition of stage and PS [HR=1.11, 95% CI 1.07 to 1.16, p<0.001] [table 12.2] but not once receipt of treatment was included in the model [HR=1.02, 95% CI 0.98 to 1.06, p=0.36] [table 12.2]. Addition of timeliness of treatment to the model did not change the findings [HR=1.03, 95% CI 0.98 to 1.07, p=0.22]. 12.3.2.3 2006-2009 cohort An identical pattern of results was seen for the 2006-2009 dataset [results not shown]. 206 Table 12.2. Likelihood of mortality (hazard ratio of death), by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) Variable Unadjusted (n=28,733) Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral (n=28,733) HR 95% CI P HR 95% CI P Deprivation quintile 0.0002 <0.001 1 (least deprived) 1.00 1.00 2 1.06 1.01 1.11 0.02 1.07 1.02 1.12 0.009 3 1.06 1.01 1.11 0.01 1.06 1.01 1.11 0.02 4 1.09 1.05 1.14 <0.001 1.12 1.07 1.17 <0.001 5 (most deprived) 1.10 1.05 1.14 <0.001 1.13 1.09 1.18 <0.001 Age group <0.001 <0.001 <60 1.00 1.00 60-69 1.09 1.04 1.14 <0.001 1.10 1.05 1.14 <0.001 70-79 1.33 1.28 1.39 <0.001 1.29 1.24 1.35 <0.001 80+ 1.83 1.75 1.91 <0.001 1.62 1.55 1.69 <0.001 Sex <0.001 <0.001 Female 1.00 1.00 Male 1.10 1.07 1.12 <0.001 1.14 1.12 1.17 <0.001 Histology <0.001 <0.001 NSCLC 1.00 1.00 SCLC 1.39 1.34 1.44 <0.001 1.43 1.37 1.48 <0.001 Other 1.76 1.71 1.81 <0.001 1.52 1.48 1.57 <0.001 Year of Diagnosis <0.001 0.28 2006 1.00 1.00 2007 0.97 0.94 1.01 0.15 0.98 0.94 1.02 0.25 2008 0.97 0.93 1.01 0.14 0.98 0.94 1.02 0.37 2009 0.90 0.86 0.93 <0.001 0.95 0.91 1.00 0.03 2010 0.91 0.88 0.95 <0.001 0.95 0.91 1.00 0.07 Co-morbidity score <0.001 0 1.00 1.00 1-2 1.12 1.06 1.17 <0.001 1.01 0.96 1.06 0.81 3+ 1.20 1.11 1.29 <0.001 1.02 0.94 1.09 0.70 CCM missing 1.06 1.02 1.10 0.002 1.13 1.08 1.17 <0.001 No HES link 1.06 1.02 1.10 0.004 1.07 1.03 1.12 0.002 Timely GP referral <0.001 <0.001 No GP ref date 1.00 1.00 FHA<=14 days 0.68 0.66 0.70 <0.001 0.73 0.71 0.76 <0.001 FHA>14 days 0.60 0.57 0.62 <0.001 0.63 0.61 0.66 <0.001 Stage <0.001 I 1.00 II 1.54 1.35 1.76 <0.001 III 2.61 2.38 2.87 <0.001 IV 4.86 4.45 5.31 <0.001 Missing 3.63 3.34 3.95 <0.001 Performance Status <0.001 0 1.00 1-2 1.73 1.63 1.85 <0.001 3-4 4.05 3.77 4.35 <0.001 Missing 2.34 2.21 2.49 <0.001 207 Table 12.2 (cont). Likelihood of mortality (HR of death), by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 (DCO cases excluded) Variable Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral, stage, PS (n=28,733) Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral, stage, PS, treatment type (n=28,733) HR 95% CI P HR 95% CI P Deprivation quintile <0.001 0.51 1 (least deprived) 1.00 1.00 2 1.07 1.02 1.12 0.007 1.04 0.99 1.10 0.08 3 1.05 1.00 1.10 0.05 1.03 0.98 1.08 0.23 4 1.10 1.05 1.15 <0.001 1.02 0.98 1.07 0.29 5 (most deprived) 1.11 1.07 1.16 <0.001 1.02 0.98 1.06 0.36 Age group <0.001 <0.001 <60 1.00 1.00 60-69 1.09 1.05 1.14 <0.001 0.96 0.92 1.00 0.07 70-79 1.28 1.22 1.33 <0.001 0.91 0.87 0.95 <0.001 80+ 1.57 1.50 1.65 <0.001 0.85 0.81 0.89 <0.001 Sex <0.001 <0.001 Female 1.00 1.00 Male 1.15 1.12 1.18 <0.001 1.14 1.11 1.17 <0.001 Histology <0.001 <0.001 NSCLC 1.00 1.00 SCLC 1.36 1.31 1.42 <0.001 1.52 1.46 1.59 0.00 Other 1.44 1.40 1.49 <0.001 0.99 0.96 1.03 0.70 Year of Diagnosis 0.61 <0.001 2006 1.00 1.00 2007 0.98 0.94 1.02 0.31 0.99 0.95 1.03 0.63 2008 0.98 0.94 1.02 0.33 0.98 0.95 1.02 0.43 2009 0.97 0.93 1.02 0.22 1.14 1.09 1.19 <0.001 2010 0.96 0.91 1.01 0.12 1.15 1.09 1.21 <0.001 Co-morbidity score <0.001 <0.001 0 1.00 1.00 1-2 1.01 0.96 1.06 0.77 0.92 0.87 0.96 <0.001 3+ 1.01 0.94 1.09 0.75 0.89 0.82 0.96 0.002 CCM missing 1.12 1.07 1.16 <0.001 1.11 1.07 1.16 <0.001 No HES link 1.07 1.02 1.12 0.003 0.92 0.88 0.96 <0.001 Timely GP referral <0.001 <0.001 No GP ref date 1.00 1.00 FHA<=14 days 0.75 0.73 0.77 <0.001 0.84 0.82 0.87 <0.001 FHA>14 days 0.66 0.63 0.68 <0.001 0.75 0.72 0.78 <0.001 Stage <0.001 <0.001 I 1.00 1.00 II 1.61 1.41 1.84 <0.001 1.42 1.24 1.62 <0.001 III 2.68 2.44 2.95 <0.001 2.05 1.87 2.25 <0.001 IV 4.65 4.25 5.08 <0.001 3.29 3.01 3.60 <0.001 Missing 2.82 2.58 3.09 <0.001 2.26 2.06 2.48 <0.001 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 1.39 1.30 1.49 <0.001 1.12 1.05 1.20 <0.001 3-4 2.49 2.31 2.68 <0.001 1.55 1.44 1.67 <0.001 Missing 1.71 1.60 1.84 <0.001 1.34 1.25 1.44 <0.001 Type of treatment <0.001 No treatment 1.00 Surgery 0.09 0.08 0.10 <0.001 Surgery +chemo/radiotherapy 0.12 0.11 0.13 <0.001 Chemotherapy 0.40 0.38 0.42 <0.001 Chemo+radiotherapy 0.25 0.24 0.26 <0.001 Radiotherapy 0.50 0.48 0.51 <0.001 208 Table 12.3. Likelihood of mortality (hazard ratio of death), by selected patient, tumour and system factors for the subset diagnosed between 2006 and 2010 with stage recorded (DCO cases excluded) Variable Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral (n=28,733) Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral, stage (n=28,733) HR 95% CI P HR 95% CI P Deprivation quintile 0.0004 0.002 1 (least deprived) 1.00 1.00 2 1.06 0.96 1.16 0.27 1.06 0.96 1.17 0.22 3 1.07 0.97 1.17 0.18 1.08 0.98 1.18 0.13 4 1.17 1.07 1.27 0.001 1.16 1.06 1.27 0.001 5 (most deprived) 1.17 1.07 1.27 <0.001 1.16 1.06 1.26 0.001 Age group <0.001 <0.001 <60 1.00 1.00 60-69 1.02 0.94 1.11 0.65 1.06 0.98 1.16 0.14 70-79 1.21 1.12 1.32 0.00 1.34 1.24 1.45 <0.001 80+ 1.49 1.37 1.63 0.00 1.70 1.55 1.85 <0.001 Sex <0.001 <0.001 Female 1.00 1.00 Male 1.13 1.07 1.19 <0.001 1.13 1.07 1.19 <0.001 Histology <0.001 <0.001 NSCLC 1.00 1.00 SCLC 1.46 1.33 1.60 <0.001 1.18 1.08 1.30 <0.001 Other 1.48 1.39 1.57 <0.001 1.55 1.46 1.65 <0.001 Year of Diagnosis 0.16 0.96 2006 1.00 1.00 2007 1.01 0.91 1.12 0.90 0.98 0.88 1.10 0.77 2008 1.08 0.98 1.19 0.12 0.98 0.89 1.08 0.71 2009 1.10 0.98 1.23 0.09 1.01 0.90 1.13 0.86 2010 1.13 1.01 1.27 0.03 1.02 0.91 1.15 0.76 Co-morbidity score 0.02 0.03 0 1.00 1.00 1-2 0.95 0.86 1.06 0.36 1.02 0.92 1.13 0.72 3+ 1.07 0.92 1.25 0.40 1.18 1.01 1.38 0.04 CCM missing 1.11 1.01 1.21 0.03 1.04 0.95 1.14 0.43 No HES link 0.97 0.89 1.05 0.45 0.94 0.86 1.02 0.15 Timely GP referral <0.001 <0.001 No GP referral date 1.00 1.00 FHA<=14 days 0.78 0.74 0.83 <0.001 0.77 0.73 0.81 <0.001 FHA>14 days 0.66 0.61 0.71 <0.001 0.70 0.65 0.75 <0.001 Stage <0.001 I 1.00 II 1.73 1.52 1.98 <0.001 III 3.16 2.88 3.48 <0.001 IV 6.54 5.96 7.16 <0.001 209 Table 12.3 (cont). Likelihood of mortality (HR of death), by selected patient, tumour and system factors for those diagnosed between 2006 and 2010 with stage recorded (DCO cases excluded) variable Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral ,stage, PS (n=28,733) Adjusted – IMD, age, sex, histology, year, CCM, stage, PS, timely GP referral, treatment type (n=28,733) HR 95% CI P HR 95% CI P Deprivation quintile 0.18 0.45 1 (least deprived) 1.00 1.00 2 1.05 0.95 1.15 0.37 1.07 0.97 1.18 0.16 3 1.01 0.92 1.11 0.76 1.03 0.93 1.13 0.61 4 1.10 1.01 1.20 0.04 1.07 0.98 1.16 0.16 5 (most deprived) 1.06 0.97 1.15 0.21 1.02 0.94 1.11 0.60 Age group <0.001 <0.001 <60 1.00 1.00 60-69 1.00 0.92 1.09 0.99 0.95 0.87 1.03 0.20 70-79 1.13 1.04 1.22 0.005 0.86 0.79 0.94 0.001 80+ 1.28 1.17 1.40 <0.001 0.81 0.73 0.89 <0.001 Sex <0.001 <0.001 Female 1.00 1.00 Male 1.15 1.10 1.21 <0.001 1.15 1.09 1.21 <0.001 Histology <0.001 <0.001 NSCLC 1.00 1.00 SCLC 1.12 1.02 1.23 0.02 1.42 1.29 1.57 <0.001 Other 1.25 1.18 1.33 <0.001 0.95 0.89 1.02 0.15 Year of Diagnosis 0.65 0.0009 2006 1.00 1.00 2007 0.96 0.87 1.07 0.49 0.98 0.88 1.09 0.65 2008 0.95 0.86 1.04 0.26 0.98 0.89 1.08 0.66 2009 1.00 0.89 1.12 0.99 1.17 1.04 1.30 0.008 2010 1.01 0.90 1.14 0.86 1.18 1.05 1.33 0.005 Co-morbidity score 0.05 <0.001 0 1.00 1.00 1-2 0.95 0.86 1.05 0.34 0.90 0.81 0.99 0.03 3+ 1.05 0.90 1.23 0.55 0.91 0.78 1.06 0.23 CCM missing 1.06 0.97 1.17 0.18 1.04 0.95 1.14 0.43 No HES link 0.93 0.85 1.01 0.09 0.83 0.76 0.91 <0.001 Timely GP referral <0.001 <0.001 No GP referral date 1.00 1.00 FHA<=14 days 0.80 0.76 0.85 <0.001 0.87 0.82 0.92 <0.001 FHA>14 days 0.74 0.69 0.80 <0.001 0.81 0.75 0.87 <0.001 Stage <0.001 <0.001 I 1.00 1.00 II 1.66 1.45 1.90 <0.001 1.54 1.34 1.76 <0.001 III 2.96 2.69 3.25 <0.001 2.44 2.21 2.70 <0.001 IV 5.88 5.36 6.45 <0.001 4.52 4.10 4.98 <0.001 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 1.51 1.40 1.63 <0.001 1.22 1.13 1.32 <0.001 3-4 3.45 3.15 3.77 <0.001 2.08 1.89 2.28 <0.001 Missing 1.94 1.75 2.16 <0.001 1.53 1.37 1.70 <0.001 Type of treatment <0.001 No treatment 1.00 Surgery 0.18 0.15 0.21 <0.001 Surgery +chemo/radiotherapy 0.17 0.14 0.20 <0.001 Chemotherapy 0.42 0.38 0.46 <0.001 Chemo+radiotherapy 0.28 0.26 0.31 <0.001 Radiotherapy 0.59 0.54 0.63 <0.001 210 12.3.2.4 2006-2010 subset with stage recorded Median survival in the staged cohort was better than for the full 2006-2010 cohort [median of 30 compared to 21 weeks]. In the subset with stage recorded [n=7769] socio-economic inequalities in survival were found in a univariable analysis [OR=1.13, 95% CI 1.04 to 1.22, p=0.005] and in multivariable analyses when age and sex, histology, year of diagnosis, co-morbidity, timely GP referral and stage were included in the analysis [OR=1.16, 95% CI 1.06 to 1.26, p=0.001], but not when PS was added to the model [OR=1.06, 95% CI 0.97 to 1.15, p=0.21] [table 12.3] and not when receipt of treatment was included [OR=1.02, 95% CI 0.94 to 1.11, p=0.60]. Further addition of timely treatment did not alter this. 12.3.2.5 2006-2009 subset with stage recorded In the 2006-2009 subset with stage recorded [n=5233] an identical pattern of results to the 2006-2010 cohort with stage recorded was again found [results not shown]. 12.3.3 Two year survival: 2006-2009 data-set 12.3.3.1 Descriptive analysis In the 2006 to 2009 data-set [n=22,967], 15.3% of patients [3513] were still alive two years after diagnosis [table 12.4]. In the unadjusted analysis those in the lowest SEP group were significantly less likely to still be alive after 2 years, compared to the highest SEP group [OR=0.79, 95% CI 0.70 to 0.89, p<0.001] [table 12.5]. Likelihood of two-year survival was better for patients with high SEP, younger patients, women, those diagnosed with NSCLC, those with no co-morbidity, early stage patients, those with good PS, those referred by their GP and those receiving treatment. Patients who received surgery between 32-62 days after diagnosis had the greatest likelihood of two-year survival. For chemotherapy and radiotherapy optimal survival was found for those who had treatment more than 62 days after diagnosis [table 12.4]. 12.3.3.2 Multivariable regression analysis In a multivariable analysis adjusted for age, sex, histology, year of diagnosis, timely GP referral and co-morbidity, inequalities in survival by SEP were observed, with a 211 reduced likelihood of 2 year survival in the lowest compared to the highest SEP group [OR=0.74, 95% CI 0.66 to 0.84, p<0.001] [table 12.5]. The amount of survival variation explained by the model was poor [R2=5.77]. Adding stage and PS improved this [R2=12.31] but did not substantially change the SEP OR [OR=0.77, 95% CI 0.66 to 0.88, p<0.001]. However if treatment type [surgery, surgery plus chemotherapy/radiotherapy, chemotherapy, chemotherapy plus radiotherapy, radiotherapy] was included in the analysis the association no longer remained significant [OR=0.87, 95% CI 0.75 to 1.00, p=0.06]. Receipt of treatment also made the greatest contribution to explaining survival variance in the model [R2=27.97]. Further addition of timeliness of treatment made no difference to the outcome [OR=0.87, 95% CI 0.75 to 1.00, p=0.05]. No significant interactions between SEP and type of treatment [or SEP and histology] were found. Patients treated later than the referral and treatment target times had better 2-year survival than those who received timely treatment. Similar results were found for the 1999-2005 cohort. Socio-economic inequalities in survival were found in a univariable analysis [OR=0.81, 95% CI 0.73 to 0.90, p=0.006] and remained when adjusted for age, sex, year of diagnosis, co-morbidity and timely referral [OR=0.84, 95% CI 0.76 to 0.94, p<0.001, R2=5.80] but were no longer found when treatment was added to the model [OR=1.06, 95% CI 0.94 to 1.20, p=0.38 R2=24.60] [appendix D, table D5]. Stage and PS were not available for these years. 12.3.3.3 2006-2009 subset with stage recorded In the 2006 to 2009 subset with stage recorded [n=5233] socio-economic inequalities in survival were found in a multivariable analysis including age, sex, histology, year of diagnosis, co-morbidity and timely GP referral, with those in the most deprived group having a significantly lower likelihood of 2-year survival than those in the most affluent [OR=0.76, 95% CI 0.61 to 0.76, p=0.02], but with poor explanation of survival variance using this model [R2=3.72]. However, the association was no longer significant when stage was added [OR=0.79, 95% CI 0.61 to 0.82, p=0.07], with a large increase in R2 to 24.39. The addition of PS further attenuated the OR (OR=0.90, 95% CI 0.70 to 1.17, p=0.45) [table 12.6] as did 212 the addition of treatment [OR=1.03, 95% CI 0.78 to 1.36, p=0.85]. Stage and receipt of treatment contributed most to the explanation for survival variance in the model but PS and treatment had the greatest influence on likelihood of survival by SEP. Table 12.4 Survival at 2 years: descriptive data and univariable ORs for 2006-2009 cohort Variable Cohort Survival at 2 years Unadjusted (n=22,967) N N % OR 95% CI P Deprivation quintile 22,967 3,513 15.3 0.004 1 (least deprived) 2,698 474 17.6 1.00 2 3,303 520 15.7 0.88 0.76 1.00 0.06 3 3,827 586 15.3 0.85 0.74 0.97 0.02 4 5,387 815 15.1 0.84 0.74 0.95 0.005 5 (most deprived) 7,752 1,118 14.4 0.79 0.70 0.89 <0.001 Age group 22,967 3,513 15.3 <0.001 <60 3,041 651 21.4 1.00 60-69 6,016 1,199 19.9 0.91 0.82 1.02 0.10 70-79 8,219 1,210 14.7 0.63 0.57 0.70 <0.001 80+ 5,691 453 8.0 0.32 0.28 0.36 <0.001 Sex 22,967 3,513 15.3 <0.001 Female 10,510 1,770 16.8 1.00 Male 12,457 1,743 14.0 0.80 0.75 0.86 <0.001 Histology 22,967 3,513 15.3 <0.001 NSCLC 12,152 2,463 20.3 1.00 SCLC 2,829 236 8.3 0.36 0.31 0.41 <0.001 Other 7,986 814 10.2 0.45 0.41 0.49 <0.001 Histology (alternative) 22,967 3,513 15.3 <0.001 Probable NSCLC 13053 2835 21.7 1.00 SCLC 2,829 236 8.3 0.33 0.29 0.38 <0.001 Unspecified 7085 442 6.2 0.24 0.22 0.27 <0.001 Year of Diagnosis 22,967 3,513 15.3 <0.001 2006 5,533 783 14.2 1.00 2007 5,712 844 14.8 1.05 0.95 1.17 0.35 2008 5,851 861 14.7 1.05 0.94 1.16 0.39 2009 5,871 1,025 17.5 1.28 1.16 1.42 <0.001 Co-morbidity 22,967 3,513 15.3 0.0006 0 3,597 601 16.7 1.00 1-2 3,125 453 14.5 0.85 0.74 0.97 0.01 3+ 766 89 11.6 0.66 0.52 0.83 <0.001 CCM missing 10,133 1,509 14.9 0.87 0.79 0.97 0.009 No HES link 5,346 861 16.1 0.96 0.85 1.07 0.45 Stage 22,967 3,513 15.3 <0.001 I 864 504 58.3 1.00 II 332 128 38.6 0.45 0.35 0.58 <0.001 III 1,587 276 17.4 0.15 0.12 0.18 <0.001 IV 2,450 139 5.7 0.04 0.03 0.05 <0.001 missing 17,734 2,466 13.9 0.12 0.10 0.13 <0.001 Performance Status 22,967 3,513 15.3 <0.001 0 1,298 481 37.1 1.00 1-2 3,414 635 18.6 0.39 0.34 0.45 <0.001 3-4 1,415 65 4.6 0.08 0.06 0.11 <0.001 Missing 16,840 2,332 13.9 0.27 0.24 0.31 <0.001 Timely treatment 22,967 3,513 15.3 <0.001 >31 days from diagnosis 7,443 2,346 31.5 1.00 <=31 days from diagnosis 4,849 708 14.6 0.37 0.34 0.41 <0.001 No treatment 10,675 459 4.3 0.10 0.09 0.11 <0.001 213 Table 12.4 (cont). Survival at 2 years: descriptive data and univariable ORs for 2006-2009 cohort Variable Cohort Survival at 2 years Unadjusted (n=22,967) N N % OR 95% CI P Timely GP referral 22,967 3,513 15.3 <0.001 No GP referral date 11,014 1,254 11.4 1.00 FHA<=14 days 8,284 1,456 17.6 1.66 1.53 1.80 <0.001 FHA>14 days 3,669 803 21.9 2.18 1.98 2.40 <0.001 Tumour type 22,967 3,513 15.3 <0.001 Adenocarcinoma 3,409 861 25.3 1.00 Squamous 4,141 992 24.0 0.93 0.84 1.04 0.19 Large 640 116 18.1 0.66 0.53 0.81 <0.001 Non-small 3,962 494 12.5 0.42 0.37 0.48 <0.001 Other specified 901 372 41.3 2.08 1.78 2.43 <0.001 Small 2,829 236 8.3 0.27 0.23 0.31 <0.001 Unspecified carcinoma 442 26 5.9 0.18 0.12 0.28 <0.001 Neoplasm 6,643 416 6.3 0.20 0.17 0.22 <0.001 Surgery Y/N 22,967 3,513 15.3 <0.001 No surgery 20731 1951 9.4 1.00 Surgery 2236 1562 69.9 22.31 20.15 24.69 <0.001 Chemotherapy Y/N 22,967 3,513 15.3 <0.001 No chemotherapy 16340 2104 12.9 1.00 Chemotherapy 6627 1409 21.3 1.83 1.70 1.97 <0.001 Radiotherapy Y/N 22,967 3,513 15.3 <0.001 No radiotherapy 15320 2108 13.8 1.00 Radiotherapy 7647 1405 18.4 1.41 1.31 1.52 <0.001 Type of 1st treatment 22,967 3,513 15.3 <0.001 No treatment 10,675 459 4.3 1.00 Surgery 1st 2193 1534 70.0 51.81 45.46 59.04 <0.001 Chemotherapy 1st 5490 892 16.3 4.32 3.84 4.86 <0.001 Radiotherapy 1st 4609 628 13.6 3.51 3.10 3.98 <0.001 Type of treatment 22,967 3,513 15.3 <0.001 No treatment 10,675 459 4.3 1.00 Surgery 1,427 1,041 73.0 60.02 51.68 69.71 <0.001 Surgery +chemo/ radiotherapy 809 521 64.4 40.26 33.91 47.80 <0.001 Chemotherapy 2,759 267 9.7 2.38 2.04 2.79 <0.001 Chemotherapy+radiotherapy 3,236 701 21.7 6.15 5.43 6.98 <0.001 Radiotherapy 4,061 524 12.9 3.30 2.89 3.76 <0.001 Surgery 22,967 3,513 15.3 <0.001 No surgery 20,731 1,951 9.41 1.00 Surgery <31 days from diagnosis 515 333 64.7 17.61 14.61 21.22 <0.001 Surgery 32-62 days from diagnosis 943 686 72.8 25.69 22.10 29.87 <0.001 Surgery > 62 days from diagnosis 778 543 69.8 22.24 18.95 26.10 <0.001 Chemotherapy 22,967 3,513 15.3 <0.001 No chemotherapy 16,340 2,104 12.9 1.00 Chemotherapy <31 days from diagnosis 2,782 324 11.7 0.89 0.79 1.01 0.07 Chemotherapy 32-62 days from diagnosis 2,267 420 18.5 1.54 1.37 1.73 <0.001 Chemotherapy >62 days from diagnosis 1,578 665 42.1 4.93 4.42 5.50 <0.001 214 Table 12.4 (cont). Survival at 2 years: descriptive data and univariable ORs for 2006-2009 cohort Variable Cohort Survival at 2 years Unadjusted (n=22,967) N N % OR 95% CI P Radiotherapy 22,967 3,513 15.3 <0.001 No radiotherapy 15,320 2,108 13.8 1.00 Radiotherapy <31 days from diagnosis 1,649 62 3.8 0.24 0.19 0.32 <0.001 Radiotherapy 32-62 days from diagnosis 1,871 268 14.3 1.05 0.91 1.20 0.51 Radiotherapy > 62 days from diagnosis 4,127 1,075 26.1 2.21 2.03 2.40 <0.001 Timely 1st Treatment 22,967 3,513 15.3 <0.001 No treatment 10,675 459 4.3 1.00 Chemotherapy 1st >31 days from diagnosis 2,775 577 20.8 5.84 5.13 6.66 <0.001 Chemotherapy 1st <31 days from diagnosis 2,715 315 11.6 2.92 2.51 3.39 <0.001 Surgery 1st >31 days from diagnosis 1,678 1,201 71.6 56.04 48.65 64.55 <0.001 Surgery 1st <31 days from diagnosis 515 333 64.7 40.72 33.23 49.91 <0.001 Radiotherapy 1st >31 days from diagnosis 2,990 568 19.0 5.22 4.58 5.95 <0.001 Radiotherapy 1st <31 days from diagnosis 1,619 60 3.7 0.86 0.65 1.13 0.27 215 Table 12.5. Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 2006 and 2009 (DCO cases excluded) Variable Adjusted – IMD, age, sex, histology, year, CCM, timely referral (n=22,967, R2=5.77) Adjusted – IMD, age, sex, histology, year, CCM, timely referral, stage, PS (n=22,967, R2=12.31) OR 95% CI P OR 95% CI P Deprivation quintile <0.001 0.002 1 (least deprived) 1.00 1.00 2 0.87 0.76 1.00 0.05 0.87 0.76 1.01 0.07 3 0.86 0.75 0.98 0.03 0.87 0.75 1.00 0.05 4 0.80 0.70 0.91 0.001 0.83 0.72 0.95 0.005 5 (most deprived) 0.74 0.66 0.84 <0.001 0.77 0.68 0.88 <0.001 Age group <0.001 <0.001 <60 1.00 1.00 60-69 0.93 0.83 1.03 0.17 0.92 0.82 1.03 0.14 70-79 0.65 0.58 0.73 <0.001 0.64 0.57 0.71 <0.001 80+ 0.35 0.31 0.41 <0.001 0.34 0.30 0.39 <0.001 Sex <0.001 <0.001 Female 1.00 1.00 Male 0.74 0.69 0.80 <0.001 0.73 0.67 0.79 <0.001 Histology <0.001 <0.001 NSCLC 1.00 1.00 SCLC 0.34 0.30 0.40 <0.001 0.36 0.31 0.42 <0.001 Other 0.60 0.55 0.66 <0.001 0.64 0.58 0.71 <0.001 Year of Diagnosis 0.08 0.20 2006 1.00 1.00 2007 1.06 0.95 1.18 0.30 1.05 0.94 1.18 0.35 2008 1.06 0.95 1.19 0.27 1.08 0.96 1.21 0.18 2009 1.19 1.04 1.36 0.01 1.16 1.01 1.33 0.04 Co-morbidity score 0.007 0.01 0 1.00 1.00 1-2 0.95 0.83 1.09 0.50 0.94 0.81 1.08 0.38 3+ 0.84 0.66 1.07 0.16 0.79 0.61 1.03 0.08 CCM missing 0.82 0.74 0.92 <0.001 0.83 0.74 0.93 0.001 No HES link 0.94 0.82 1.07 0.32 0.97 0.85 1.11 0.66 Timely GP referral <0.001 <0.001 No GP referral 1.00 1.00 FHA<=14 days 1.44 1.32 1.58 <0.001 1.40 1.28 1.53 <0.001 FHA>14 days 1.98 1.79 2.19 <0.001 1.80 1.62 2.01 <0.001 Stage <0.001 I 1.00 II 0.38 0.29 0.50 <0.001 III 0.13 0.11 0.16 <0.001 IV 0.04 0.03 0.05 <0.001 missing 0.17 0.14 0.20 <0.001 Performance Status <0.001 0 1.00 1-2 0.55 0.47 0.65 <0.001 3-4 0.18 0.14 0.24 <0.001 Missing 0.45 0.38 0.54 <0.001 216 Table 12.5 (b). Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 2006 and 2009 (DCO cases excluded) variable Adjusted – IMD, age, sex, histology, year, CCM, timely referral, stage, PS, treatment type (n=22,967, R2=27.97) Adjusted – IMD, age, sex, histology, year, CCM, timely referral, stage, PS, treatment type, timely 1st treatment (n=22,967, R2=28.74) OR 95% CI P OR 95% CI P Deprivation quintile 0.34 0.34 1 (least deprived) 1.00 1.00 2 0.88 0.75 1.04 0.13 0.88 0.75 1.04 0.14 3 0.90 0.77 1.06 0.22 0.90 0.76 1.05 0.18 4 0.94 0.81 1.09 0.41 0.94 0.81 1.09 0.39 5 (most deprived) 0.87 0.75 1.00 0.06 0.87 0.75 1.00 0.05 Age group 0.11 0.04 <60 1.00 1.00 60-69 1.04 0.91 1.18 0.60 1.02 0.89 1.16 0.80 70-79 0.94 0.82 1.07 0.33 0.90 0.79 1.03 0.12 80+ 0.88 0.74 1.03 0.12 0.84 0.71 1.00 0.05 Sex <0.001 <0.001 Female 1.00 1.00 Male 0.72 0.66 0.79 <0.001 0.71 0.65 0.78 <0.001 Histology <0.001 <0.001 NSCLC 1.00 1.00 SCLC 0.46 0.40 0.54 <0.001 0.57 0.49 0.67 <0.001 Other 1.28 1.14 1.44 <0.001 1.33 1.18 1.50 <0.001 Year of Diagnosis 0.009 0.009 2006 1.00 1.00 2007 1.02 0.90 1.15 0.77 1.00 0.88 1.13 0.95 2008 1.08 0.95 1.23 0.22 1.02 0.90 1.16 0.75 2009 0.83 0.71 0.98 0.03 0.80 0.68 0.93 0.005 Co-morbidity score <0.001 <0.001 0 1.00 1.00 1-2 1.10 0.93 1.29 0.25 1.10 0.93 1.29 0.25 3+ 0.96 0.72 1.28 0.79 0.96 0.72 1.29 0.81 CCM missing 0.84 0.73 0.95 0.006 0.83 0.73 0.95 0.006 No HES link 1.37 1.17 1.60 <0.001 1.38 1.18 1.61 <0.001 Timely GP referral <0.001 <0.001 No GP referral date 1.00 1.00 FHA<=14 days 1.16 1.05 1.29 0.004 1.14 1.03 1.26 0.02 FHA>14 days 1.40 1.25 1.58 <0.001 1.36 1.20 1.53 <0.001 Stage <0.001 <0.001 I 1.00 1.00 II 0.48 0.35 0.68 <0.001 0.49 0.35 0.69 <0.001 III 0.29 0.23 0.36 <0.001 0.30 0.24 0.38 <0.001 IV 0.12 0.09 0.15 <0.001 0.13 0.10 0.17 <0.001 Missing 0.31 0.25 0.38 <0.001 0.33 0.27 0.41 <0.001 Performance Status 0.001 0 1.00 1.00 1-2 0.88 0.73 1.05 0.16 0.89 0.74 1.07 0.20 3-4 0.59 0.43 0.81 0.001 0.61 0.44 0.83 0.002 Missing 0.73 0.60 0.90 0.002 0.72 0.59 0.88 0.002 217 Table 12.5 (b, cont). Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 2006 and 2009 (DCO cases excluded) variable Adjusted – IMD, age, sex, histology, year, CCM, stage, PS, timely referral, treatment type (n=22,967, R2=27.97) Adjusted – IMD, age, sex, histology, year, CCM, stage, PS, timely referral, treatment type, timely 1st treatment (n=22,967, R2=28.74) OR 95% CI P OR 95% CI P Type of treatment <0.001 <0.001 No treatment 1.00 1.00 Surgery 49.77 41.93 59.07 <0.001 60.95 51.12 72.66 <0.001 Surgery +chemo/ radiotherapy 33.00 26.99 40.34 <0.001 40.47 32.96 49.69 <0.001 Chemotherapy 3.25 2.71 3.91 <0.001 4.19 3.47 5.05 <0.001 Chemotherapy+radiotherapy 7.83 6.65 9.22 <0.001 10.03 8.48 11.86 <0.001 Radiotherapy 3.34 2.89 3.87 <0.001 4.12 3.55 4.78 <0.001 Timely 1st treatment <0.001 >31 days from diagnosis 1.00 <31 days from diagnosis 0.50 0.45 0.56 <0.001 218 Table 12.6. Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 2006 and 2009 (DCO cases excluded) with stage recorded Variable Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral (n=5233, R2=3.72) Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral, stage (n=5233, R2=24.39) Adjusted – IMD, age, sex, histology, year, CCM, timely GP referral, stage, PS (n=5233, R2=26.59) OR 95% CI P OR 95% CI P OR 95% CI P Deprivation quintile 0.02 0.07 0.22 1 (least deprived) 1.00 1.00 1.00 2 0.92 0.71 1.20 0.54 0.93 0.69 1.25 0.62 0.97 0.72 1.31 0.83 3 0.92 0.72 1.19 0.53 0.92 0.69 1.23 0.56 1.00 0.75 1.34 0.99 4 0.72 0.56 0.91 0.007 0.71 0.54 0.93 0.01 0.77 0.58 1.02 0.06 5 (most deprived) 0.76 0.61 0.96 0.02 0.79 0.61 1.02 0.07 0.90 0.70 1.17 0.45 Age group <0.001 <0.001 <0.001 <60 1.00 1.00 1.00 60-69 1.06 0.86 1.31 0.60 0.94 0.74 1.20 0.63 1.03 0.81 1.32 0.79 70-79 0.79 0.64 0.97 0.03 0.60 0.47 0.77 <0.001 0.75 0.59 0.96 0.02 80+ 0.53 0.41 0.68 <0.001 0.35 0.26 0.47 <0.001 0.49 0.36 0.67 <0.001 Sex <0.001 <0.001 <0.001 Female 1.00 1.00 1.00 Male 0.75 0.65 0.86 <0.001 0.72 0.61 0.84 <0.001 0.70 0.59 0.82 <0.001 Histology <0.001 <0.001 <0.001 NSCLC 1.00 1.00 1.00 SCLC 0.32 0.22 0.47 <0.001 0.49 0.32 0.75 0.001 0.49 0.32 0.76 0.001 Other 0.65 0.54 0.78 <0.001 0.61 0.50 0.76 <0.001 0.80 0.64 1.00 0.05 Year of Diagnosis 0.70 0.29 0.93 2006 1.00 1.00 1.00 2007 1.03 0.80 1.33 0.83 1.04 0.78 1.39 0.78 1.07 0.80 1.43 0.66 2008 0.91 0.72 1.15 0.42 1.04 0.80 1.37 0.75 1.09 0.83 1.44 0.52 2009 0.95 0.72 1.25 0.70 1.09 0.80 1.49 0.58 1.09 0.79 1.50 0.59 Co-morbidity score 0.07 0.22 0 1.00 1.00 1.00 1-2 1.22 0.95 1.58 0.12 1.08 0.81 1.45 0.59 1.21 0.90 1.64 0.20 3+ 1.00 0.63 1.58 1.00 0.67 0.40 1.13 0.14 0.83 0.49 1.40 0.48 CCM missing 0.85 0.68 1.07 0.17 0.94 0.73 1.22 0.63 0.93 0.71 1.21 0.58 No HES link 1.05 0.83 1.34 0.66 1.13 0.86 1.48 0.39 1.19 0.91 1.58 0.21 Timely GP referral <0.001 <0.001 <0.001 No GP referral 1.00 1.00 1.00 FHA<=14 days 1.44 1.22 1.70 <0.001 1.47 1.21 1.78 <0.001 1.36 1.12 1.66 0.002 FHA>14 days 1.87 1.52 2.30 <0.001 1.45 1.15 1.84 0.002 1.33 1.05 1.70 0.02 Stage <0.001 <0.001 I 1.00 1.00 II 0.39 0.29 0.51 <0.001 0.40 0.30 0.53 <0.001 III 0.13 0.11 0.16 <0.001 0.14 0.12 0.17 <0.001 IV 0.04 0.03 0.05 <0.001 0.04 0.04 0.06 <0.001 Performance Status <0.001 0 1.00 1-2 0.53 0.44 0.64 <0.001 3-4 0.18 0.13 0.25 <0.001 Missing 0.47 0.35 0.64 <0.001 219 Table 12.6 (cont). Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 2006 and 2009 (DCO cases excluded) with stage recorded Variable Adjusted – IMD, age, sex, histology, year, CCM, stage, PS, timely referral, treatment type (n=5233, R2=31.73) Adjusted – IMD, age, sex, histology, year, CCM, stage, PS, timely referral, treatment type, timely 1st treatment (n=5233, R2=32.04) OR 95% CI P OR 95% CI P Deprivation quintile 0.76 0.80 1 (least deprived) 1.00 1.00 2 1.02 0.74 1.40 0.92 1.01 0.73 1.39 0.97 3 1.08 0.79 1.47 0.64 1.07 0.78 1.46 0.67 4 0.90 0.67 1.21 0.51 0.91 0.67 1.22 0.51 5 (most deprived) 1.03 0.78 1.36 0.85 1.02 0.77 1.34 0.90 Age group 0.69 0.67 <60 1.00 1.00 60-69 1.12 0.86 1.45 0.41 1.11 0.85 1.44 0.45 70-79 1.03 0.78 1.34 0.85 1.00 0.77 1.31 0.99 80+ 0.95 0.68 1.33 0.79 0.94 0.67 1.31 0.71 Sex <0.001 <0.001 Female 1.00 1.00 Male 0.69 0.58 0.82 <0.001 0.69 0.58 0.82 <0.001 Histology 0.0005 0.007 NSCLC 1.00 1.00 SCLC 0.48 0.31 0.74 0.001 0.55 0.35 0.85 0.008 Other 1.19 0.93 1.52 0.18 1.18 0.92 1.51 0.19 Year of Diagnosis 0.43 0.41 2006 1.00 1.00 2007 1.02 0.75 1.38 0.91 1.00 0.73 1.36 0.99 2008 1.02 0.77 1.37 0.87 0.98 0.74 1.31 0.92 2009 0.82 0.59 1.14 0.24 0.79 0.57 1.11 0.18 Co-morbidity score 0.02 0.02 0 1.00 1.00 1-2 1.35 0.99 1.84 0.06 1.35 0.99 1.84 0.06 3+ 1.06 0.61 1.83 0.84 1.02 0.59 1.78 0.93 CCM missing 0.90 0.68 1.18 0.46 0.90 0.68 1.18 0.43 No HES link 1.40 1.04 1.88 0.03 1.40 1.05 1.89 0.02 Stage <0.001 <0.001 I 1.00 1.00 II 0.47 0.34 0.63 <0.001 0.47 0.35 0.64 <0.001 III 0.23 0.18 0.29 <0.001 0.24 0.19 0.30 <0.001 IV 0.09 0.07 0.12 <0.001 0.10 0.07 0.13 <0.001 Performance Status <0.001 <0.001 0 1.00 1.00 1-2 0.75 0.61 0.92 0.007 0.76 0.62 0.94 0.01 3-4 0.44 0.30 0.65 <0.001 0.46 0.31 0.67 <0.001 Missing 0.70 0.51 0.97 0.03 0.72 0.52 0.99 0.05 Timely GP referral 0.14 0.16 No GP referral date 1.00 1.00 FHA<=14 days 1.23 1.00 1.52 0.05 1.22 0.99 1.50 0.06 FHA>14 days 1.16 0.90 1.50 0.25 1.15 0.89 1.48 0.29 Type of treatment <0.001 <0.001 No treatment 1.00 1.00 Surgery 11.24 7.78 16.24 <0.001 12.86 8.84 18.70 <0.001 Surgery +chemo/radiotherapy 13.13 8.78 19.63 <0.001 14.94 9.93 22.49 <0.001 Chemotherapy 2.48 1.70 3.62 <0.001 2.81 1.92 4.11 <0.001 Chemo+radiotherapy 4.75 3.41 6.61 <0.001 5.43 3.87 7.60 <0.001 Radiotherapy 2.24 1.67 3.00 <0.001 2.47 1.83 3.32 <0.001 Timely 1st treatment <0.001 >31 days from diagnosis 1.00 <31 days from diagnosis 0.63 0.51 0.79 <0.001 220 12.4 Discussion 12.4.1 Principal findings Socio-economic inequalities in receipt of treatment appear to contribute to socio-economic inequalities in lung cancer survival. Survival inequalities were found when patient, tumour and system factors were included in the multivariable model but the addition of treatment type significantly attenuated socio-economic inequalities in survival. The further addition of time to treatment had no significant effect on socio-economic inequalities in survival. Time to treatment was, however, significantly associated with risk of death and those who received treatment within guidelines [within 31 days of diagnosis] had poorer survival than those who had later treatment [within 1 year of diagnosis]. 12.4.2 Strengths and weaknesses An exact date of death was not available as this was considered a potentially-identifiable data item. Survival time in weeks was utilised here. Inconsistent recording of the date of diagnosis and the different methodology used for determining date of diagnosis in 2010, as previously described, might also be a problem when calculating survival time from diagnosis to death. However, the 2006-2009 dataset was used to calculate the likelihood of two-year survival, and both the 2006-2010 and 2006-2009 datasets were used to determine risk of all-cause mortality and similar patterns of results were found in both data-sets. 12.4.3 Interpretation of results and comparison with other studies Inequalities in receipt of treatment [surgery and chemotherapy but not radiotherapy] were found for this dataset [chapter 10] and these treatment inequalities appear to substantially explain inequalities in survival. This has previously been shown in a small study of 695 patients that did not include co-morbidity or PS in the multivariable analysis (158). Number of co-morbidities and PS vary by SEP (69) and age (70) and might help explain socio-economic inequalities in receipt of treatment and survival. But in a multivariable analysis adjusted for stage and PS survival inequalities were still observed and it was only on addition of receipt of treatment that the association no longer remained significant. However, in the subset with stage recorded PS also 221 substantially accounted for socio-economic inequalities in survival. As previously discussed, this staged subset may not be representative of the full cohort, as patients within this were more likely to be younger and had higher rates of treatment. It has been suggested that the poorer survival of UK cancer patients compared to the European average might be related to diagnostic delay (102). However, two literature reviews both found the evidence for any association between timely care and survival for lung cancer inconclusive (184, 185). Contradictory results were found in the studies examined but the quality of the studies included was mixed and most did not adequately control for important confounders such as age, stage, histology and co-morbidity (185). The lack of control for confounding factors may account for why those with more timely care actually had poorer survival, due to the Waiting Time Paradox [WTP], as previously discussed in chapter 11. Adequately controlling for stage, co-morbidity and PS should therefore eliminate this ‘sicker quicker’ effect. In a small study of colorectal cancer patients, shorter diagnostic interval was associated with higher mortality for those who appeared more ill, but not for those presenting with ‘vague’ symptoms (179). However, in our study we found that a shorter diagnostic to treatment interval resulted in a lower likelihood of survival two years after diagnosis and increased hazard of death, compared to those with later treatment. Furthermore, those who had a FHA within 14 days of GP referral had poorer survival than those who waited longer to be seen in secondary care. These associations remained after age, stage, histology, co-morbidity and PS were taken into account in the multivariable analysis. It may be that uncontrolled confounding remains or that the measures of ‘sickness’ used – PS and co-morbidity - are not particularly good. The high levels of missing stage and PS data are also a major limitation. However, we also carried out this analysis in the staged subset [which is effectively a complete-case cohort for stage and PS] and a similar result was found. The NHS Cancer Plan set guideline waiting times for cancer treatment of 31 days from diagnosis and 62 days from GP referral. I have previously shown that, in patients who received treatment within 1 year of diagnosis, those with later stage cancer, poor PS, and those undergoing chemotherapy as a first treatment were more likely to receive timely treatment within guidelines [chapter 11]. Those with higher SEP were not 222 significantly more likely to receive treatment within the 31 day guideline from diagnosis. It appears that sicker patients with later-stage cancer obtain non-curative treatment within guidelines whereas earlier stage, good-PS patients wait longer for curative care, but their survival is better. Patients who were treated within the guideline targets had poorer survival than those who had later treatment, which might bring into question the clinical validity of these guidelines. Two year survival was better for those patients who had surgery later than the 31 day guideline [72.8% of those who had surgery between 31-62 days after diagnosis were still alive compared to 64.7% who had treatment within guidelines] but for four-year survival there was higher survival for those treated within guidelines [see Fig 12.10]. When I looked at time to surgery and survival for patients with early stage cancer [as these are the patients for whom application of the guidelines might improve survival] similar patterns were seen, so the evidence for the effectiveness of the guidelines in improving survival is unclear. I previously found no socio-economic inequalities in the diagnosis to treatment interval but some suggestion of inequalities in the GP referral to FHA interval, with those in the highest and the lowest SEP groups having quicker referral. As no linear association is observed it is perhaps unsurprising that time from diagnosis to treatment had little effect on socio-economic inequalities in survival. A study that looked at breast cancer survival also found that that time to treatment was not associated with deprivation and that adjusting for time to treatment did not attenuate the association of deprivation with survival (181). They suggest that lower survival amongst more deprived patients may therefore relate to factors prior to their entry into the secondary care system although they were not able to consider inequalities in receipt of treatment as they looked at only those patients who received treatment. The results from chapter 10 suggested that socio-economic inequalities in tumour type might account for some of the socio-economic inequalities in receipt of treatment as rates of treatment were higher for adenocarcinoma compared to squamous-cell tumours. The highest treatment rates were found in the ‘other-specified-histology’ and patients in this category had the highest two-year survival. However there was no 223 significant difference in two- year survival rates between patients diagnosed with adenocarcinoma and squamous-cell tumours. 12.4.4 Implications for policy and practice, and further research Inequalities in receipt of treatment appear to substantially account for inequalities in lung cancer survival. However, clinical guidelines focus on target times for referral and treatment. Perhaps a clinical focus on ensuring that those who are eligible for treatment actually receive it, rather than on time-interval targets might have a greater impact on improving survival, as well as reducing inequalities in survival. As patients who were treated within the guideline targets had poorer survival than those who had later treatment, the effectiveness of the guidelines on improving survival appears unclear. Patients with early stage cancer are likely to be the patients for whom application of the guidelines, resulting in earlier referral and treatment, might improve survival. Further research on this group of patients is warranted to help determine whether delays in referral and treatment lead to poorer lung cancer survival, without the confounding effect of the WTP. Further research is also required to determine whether this interval effect on survival is also seen in other cancers. Further examination of the patient, tumour and system factors that determine timeliness of treatment is warranted to determine why those who have later treatment have better survival, and to develop interventions to improve cancer survival. 12.5 Chapter summary Socio-economic inequalities in lung cancer survival appear to be statistically explained by inequalities in receipt of treatment but not by time from GP referral to FHA or from diagnosis to treatment. However, patients who were treated within the time-to-treatment guideline targets had poorer survival compared to those who had later treatment, as explained by the WTP. Interventions that address socio-economic inequalities in receipt of treatment are likely to reduce socio-economic inequalities in survival and thus improve survival rates overall. 224 The next chapter will summarise and discuss all the findings from previous results chapters on socio-economic inequalities in: receipt of treatment [chapters 7 and 10]; referral, diagnostic and treatment intervals [chapter 11]; and survival [chapter 12]; discuss the methodological strengths and weaknesses of the study, the implications for policy and practice, and the further research required. 225 Chapter 13. Overall summary - discussion and conclusions This final chapter presents a summary of the thesis findings, details overall strengths and weaknesses of the data and the analyses, and suggests some detailed further research that might arise from this work, as well as the specific post-doctoral research that I plan to conduct. The findings from each component of the thesis in relation to published studies have been discussed in depth in previous chapters and so a summary of what the thesis has added to knowledge will be detailed here. 13.1 Background This research examined the evidence for socio-economic inequalities in time to, and receipt of, treatment for lung cancer, and the contribution of these inequalities to socio-economic inequalities in survival, using the intervention-generated inequalities framework. 13.2 Methods To examine the existing evidence for socio-economic inequalities in lung cancer treatment a systematic review and meta-analysis of published studies was conducted. Cohort studies of participants with a primary diagnosis of lung cancer [ICD10 C33 or C34], where the outcome was receipt of treatment [rates or odds of receiving treatment] and where the outcome was reported by a measure of socio-economic position were examined. A quality tool was used to assess study quality. Studies conducting multivariable analysis were considered for meta-analysis. Potentially-important confounders that had previously been poorly investigated were identified by the systematic review and the results from this were used to determine variables to include in the secondary data analyses. Cancer registry [NYCRIS], Hospital Episode Statistics [HES] and lung cancer audit [LUCADA] data-sets were linked in order to investigate intervention-generated inequalities in lung cancer care. NYCRIS data for 65,210 patients diagnosed between 1999 and 2010 with a primary diagnosis of lung cancer were analysed. Logistic regression was used to examine the likelihood of receipt of treatment by SEP and of receiving timely referral, diagnosis and treatment within guideline targets, as well as the likelihood of still being alive two 226 years after diagnosis, by SEP. Cox regression was used to assess the likelihood of early referral, diagnosis and treatment and risk [hazard] of death by SEP. 13.3 Results A total of 46 papers were included in the systematic review and 23 in the meta-analysis. Socio-economic inequalities in receipt of lung cancer treatment [surgery, chemotherapy but not radiotherapy] were identified in the meta-analysis. These inequalities could not be accounted for by socio-economic differences in stage at presentation or by differences in health care system. However, not all of the included studies reported details of stage and histology, both of which influence treatment type, and very few studies took co-morbidity or performance status into account. In the linked database analysis socio-economic inequalities in receipt of surgery and chemotherapy, but not radiotherapy, were found after control for age, sex, histology, stage, performance status and co-morbidity. The odds of receiving surgery were significantly lower in the lowest compared to the highest SEP group. Patients in the lowest SEP group were significantly less likely to receive chemotherapy in the full 2006-2010 cohort but not in the staged-subset when performance status was included in the model. In the 2006 to 2010 cohort 70% of patients received a hospital appointment within the 14 day urgent referral target [although patients with non-urgent referrals might also be included here]. Time to treatment targets were only being met for 42.5% when measured from diagnosis and for 62% from GP referral. Socio-economic inequalities in time from GP referral to FHA were identified but a linear trend association by SEP was not seen. Late-stage, poor performance status, high SEP and SCLC histology were associated with a higher likelihood of a first hospital appointment within 14 days of GP referral. Older patients were less likely, whereas late stage and poor PS patients were more likely, to receive treatment within guideline time limits. Only 15.3% of patients were still alive 2 years after diagnosis. This increased to 70% for those who had surgery. Patients referred and treated later than guideline targets had higher likelihood of surviving to 2 years and decreased risk of death. 227 Socio-economic inequalities in survival were found in a multivariable analysis controlling for age, sex, histology, year of diagnosis, timely GP referral, co-morbidity, stage and performance status, with those in the lowest socio-economic group significantly less likely to be alive after 2 years, compared to the highest group. However, when receipt of treatment was included in the analysis the association no longer remained significant. Addition of timeliness of treatment made no difference to the conclusion. 13.4 Discussion 13.4.1 Principal findings This study found socio-economic inequalities in the lung cancer care pathway, with inequalities in receipt of surgery and chemotherapy, and in the time interval from GP referral to FHA. Socio-economic inequalities in survival from lung cancer were statistically explained by socio-economic inequalities in receipt of treatment, but not by inequalities in timeliness of referral and treatment, in this cohort. Socio-economic inequalities in lung cancer treatment could not be explained by inequalities in stage or performance status in the full dataset. However, socio-economic differences in performance status may account for the observed socio-economic differences in receipt of chemotherapy in the staged subset. Socio-economic inequalities in tumour type may account for some of the socio-economic gradient in surgery. Socio-economic inequalities in the GP referral to FHA, diagnosis and treatment intervals were found but were not seen in the interim intervals from FHA to diagnosis or from diagnosis to treatment, suggesting that interval inequalities originate in the GP referral to FHA interval. However, no survival benefit of timely referral and treatment was demonstrated and patients who were treated within the guideline targets had poorer survival than those who had later treatment. 13.4.2 Methodological strengths and weaknesses This is one of the first studies to link three datasets [NYCRIS, HES and LUCADA] in order to examine the factors that may influence socio-economic inequalities in lung cancer treatment, time to treatment and survival. Using a triple-linked dataset allowed 228 inclusion of co-morbidity, stage and performance status data in the analyses whilst also employing the large, comprehensive cancer registry dataset. However, the years in which stage and PS were most completely recorded in LUCADA [2009-2010] were also the years of registry data which did not yet have a co-morbidity score available. Examining receipt of treatment and survival in the 2006-2008 dataset that included all three variables gave the same pattern of results [results not shown] as that for 2006-2009 and 2006-2010, therefore the larger year groups were employed where possible. Co-morbidity was included in the 1999-2005 cohort analysis, and stage and PS in the 2006-2010 cohort, with the addition of co-morbidity to see if this made any difference to the outcome. Differences over time between the earlier- and later-year cohorts could also be observed. A major strength of this study is the completeness of treatment data recorded in the large NYCRIS cancer registry dataset particularly for chemotherapy and radiotherapy, which are often poorly recorded elsewhere. Treatment was also more comprehensively recorded here than in LUCADA (253). LUCADA data has been cited as an ideal data source for lung cancer research (153) but although LUCADA audit data completeness has improved, nationally this was low in the early years [66% entry of patients diagnosed in 2006, increasing to 93% entry in 2010] (246). The number with stage recorded was also initially low [36% of patients diagnosed in 2006] but again is increasing over time [79% in 2010] (246). For my linked north of England dataset the numbers were lower than this. Stage and PS were available within the linked LUCADA data but only 27% [7769] of the 2006-2010 registry cohort had stage recorded in LUCADA. Therefore one major weakness of the analysis was the high level of missing stage and PS data. Stage was recorded for 11.8% in 2006 and data completeness improved over time. However, even in 2010, the year with the most complete stage data, stage was only recorded for 43% of participants. Multiple imputation was considered as a way of dealing with missing data but it is not recommended where over 50% of the data in a variable are missing (241). An alternative way to address the problem of missing data is to analyse only complete cases, although results from complete-case analyses can be biased (242). Therefore I looked at the subset of patients who had stage recorded 229 [as an analysis of complete-case stage patients] and I also analysed the full dataset and included a ‘stage missing’ category. Differences between the datasets were examined and used to help explain any differences in outcome observed. Patients with poorer performance status had increased odds of timely FHA, diagnosis and treatment but a lower likelihood of receiving treatment. No clear-cut pattern was observed in time intervals or receipt of treatment by co-morbidity score in this study. Both co-morbidity score and PS can be used as proxy measures of general wellbeing and my results would suggest that PS may be a better marker of this. PS is a measure of patients’ functional status and need for care measured on an ECOG scale of 0-4, determined by the care team. Although it has been shown to have good prognostic predictive validity (257) it could be considered a somewhat subjective measure. Only moderate inter-observer agreement in allocating PS score was found in an inter-observer variability study, although there was good agreement when allocating good [PS 0-2] compared to poor PS [PS 3-4] (258). Co-morbidity score, obtained from HES data, does not appear to be a particularly good measure of patient well-being, and high levels of missing data were an issue. The Charlson Co-morbidity [CCM] Index is a validated instrument for measuring co-morbidity (245) but only contains details of conditions requiring in-patient care. Patients who suffer from a relevant condition but are treated by their GP will have a score of O, thus resulting in misclassification and underestimation of co-morbidity (259). It is also a crude measure of co-morbidity as patients with mild and severe forms of a disease receive the same score. This is a problem for conditions such as COPD where the severity of the disease is likely to influence the likelihood of surgery (259). Details of the particular condition and the severity of this would be a more useful measure than a crude score. Details of co-morbidity are included within the LUCADA audit dataset although recording of this remains low. One of the key messages of the 2012 audit report (246) related to improving the recording of co-morbidity so that this is a more reliable variable for case-mix adjustment. Future studies may therefore be able to utilise this. As for most cancer registry-based analyses, an area-based, ecological measure of SEP, calculated using postcode, was employed, which is unlikely to be an accurate marker 230 of individual circumstances and access to resources (234), potentially under-estimating the strength of the true association between SEP and receipt of treatment. A changing-deprivation-score-over-time methodology was used, as this is more likely to accurately reflect current area deprivation, rather than previous methods which retrospectively applied current deprivation back in time to earlier years of diagnosis (244). Lack of smoking status data is a major weakness for any study examining lung cancer but is one found for all studies examined as part of this research, as smoking status is not routinely recorded as part of the cancer registry dataset. The focus of this PhD study was lung cancer. One advantage of focusing on a single cancer is that a detailed and in-depth analysis can be carried out within the three-year timescale, which would not be possible if the scope of the PhD were broader. Conducting analyses on a combined number of different cancers may mask any between-cancer differences relating to tumour development and prognosis that might influence likelihood and timing of treatment. However, the generalisability of the results to other cancers is unclear. The systematic review and meta-analysis included studies from UHCS and non-UHCS countries and so the results are relevant internationally. However, for the secondary data analysis, local north of England data were analysed. Local rather than England-wide data were requested from NYCRIS following discussion with the NYCRIS Director and a number of analysts. This was a pragmatic decision relating to the complexity of the linkage, and the manageability and size of the dataset, but affects the potential generalisability of these lung cancer results for the whole of England, the UK and beyond. 13.4.3 Interpretation of findings and comparisons with other studies I have previously interpreted the findings of each of my study components and done a comparison with published studies in the preceding chapters. Here I will make some remarks in summary. Overall, this study has been able to advance knowledge on the topic of socio-economic inequalities in lung cancer care. I have demonstrated socio-economic inequalities in lung cancer treatment for surgery and chemotherapy, but not radiotherapy, in both 231 UHCSs and non-UHCSs, in my systematic review and meta-analysis. I have been able to include a number of important confounders that have not been previously well examined [stage, PS, co-morbidity] in the secondary data analysis, although the levels of missing data were a limitation. Inequalities in the referral, diagnostic and treatment intervals have not been well explored before. My results suggest that socio-economic inequalities are found in the referral interval. I have also been able to show an association between socio-economic inequalities in receipt of treatment and inequalities in lung cancer survival. 13.4.4 Implications for policy and practice, and further research I have been able to investigate a number of factors that may be important in the relationship between SEP and survival such as receipt of, and time to, treatment, stage and performance status. However, high levels of missing data and inconsistent data recording remain important problems. In order to clearly document the relationships between these variables and receipt of treatment and survival, more complete and accurate recording of data is required. The high levels of missing stage and PS data within the early LUCADA data were a major problem. The comparability of the audit data with that held within registries is uncertain as differences were found between trusts that supplied high and low levels of data to the audit in early years (152) that does appear to have been somewhat glossed over in the literature. Data quality and completeness within LUCADA has greatly improved over recent years [see appendix B Fig B1] and this is now an excellent resource that will allow better investigation into lung cancer inequalities in the future. A new single national cancer registration system [Encore] was launched in 2013 in order to improve timeliness and quality of the data, reduce data replication and the costs involved in this, and to support the improvement of staging information (260). The data will be used in conjunction with, or as an alternative to, the existing cancer audits including LUCADA. Therefore, Encore appears to be a potential resource that will improve the quality of research into inequalities in lung cancer care. Smoking status is not recorded in the cancer registry dataset but is available from primary care data. Date of first contact with GP, which could be used to calculate the 232 primary care interval [from first patient presentation to GP referral] could also be obtained from this source. Further linkage of my secondary care linked-dataset to primary care data could allow further exploration of where on the lung cancer care pathway inequalities might occur and also allow the inclusion of smoking status in analyses, the omission of which is a major weakness in current studies. Socio-economic inequalities in were found in the lung cancer care pathway but it is still unclear whether these inequalities are also found for other cancers and whether they occur at the same points in the care pathway. Similar analyses should, therefore, be conducted for other cancers. Interventions that address socio-economic inequalities in receipt of treatment are likely to reduce socio-economic inequalities in survival and thus improve the overall likelihood of survival. Further research is required to investigate the unexplained variance in treatment rates, looking at factors such as patient choice, doctor-patient communication of risk and benefit, and possible variation by trusts or within this, by hospital and individual clinician. I hope to examine these factors in my post-doctoral work. 13.4.4.1 Further research - post-doctoral work My systematic review on socio-economic inequalities in lung cancer treatment has been published in an international journal (236) and press coverage on this paper resulted in an invited editorial (261). Papers are in preparation on the empirical work presented in chapters 10-12 and the preliminary findings have been presented at conferences and published in abstract form (262-264). The rapid reviews of the evidence on socio-economic inequalities in referral, diagnostic and treatment intervals [chapter 4, table 4.1], and in survival [chapter 5, table 5.1], conducted as part of the literature review, will also be developed into full systematic reviews for publication. As a member of Fuse [UKCRC Centre for Translational Research in Public Health], I hope to feedback the findings from this thesis to practitioners and policy makers with support from the Fuse Knowledge Exchange Broker, who has an explicit remit for knowledge translation. Further qualitative work with healthcare professionals to explore potential reasons for inequalities in receipt of treatment, as well as the 233 implications for clinical practice of earlier referral and treatment resulting in poorer survival, are also planned. A future NIHR Fellowship application, to further investigate intervention-generated inequalities in cancer care using a mixed-methods approach, is planned. It is hoped to examine doctor-patient communication using a series of vignettes, and to conduct multi-level modelling on the linked data-set to take better account of surgeon and hospital characteristics in receipt of care. It is planned to repeat the secondary-data analyses detailed in this thesis on NYCRIS colorectal cancer data linked to HES data. The linked colorectal dataset was obtained at the same time as the lung cancer data and ethical approval is in place to begin this work immediately. 13.5 Reflections on the PhD process During the PhD I wrote a number of blog posts for the Fuse Open Science blog, detailing some of the many highs (265), and lows [problems with obtaining the secondary data] (266), of the PhD experience, and also why I decided to undertake a PhD (267). 13.6 Chapter summary and overall conclusions Intervention-generated inequalities in lung cancer care were identified from this research. Socio-economic inequalities in receipt of surgery and chemotherapy, but not radiotherapy, were found in the systematic review and meta-analysis and in the secondary data-analysis. Socio-economic inequalities in lung cancer treatment could not be explained by inequalities in stage or performance status in the full dataset. However, socio-economic differences in performance status may account for the observed socio-economic differences in receipt of chemotherapy in the staged subset. Tumour type may account for some of the socio-economic gradient in surgery. Socio-economic inequalities in the GP referral to FHA interval were found. Socio-economic inequalities in survival from lung cancer were statistically explained by socio-economic inequalities in receipt of treatment, but not by inequalities in timeliness of referral and treatment, in this cohort. 234 Suggestions have been made for future research that could help confirm these thesis findings, with respect to where on the pathway of lung cancer care inequalities occur, and also why they might occur. It will then be possible to develop specific interventions that address socio-economic inequalities in receipt of treatment and reduce socio-economic inequalities in survival. 235 Chapter 14. Appendices 14.1 Appendix A: Variables requested from NYCRIS Table A.1. Variables that are held in the NYCRIS regional dataset, or can be derived from this Variable Derived from NYCRIS variable (by NYCRIS) Reason for inclusion Analyses to be used in IMD ranking Postcode The project is looking at inequalities by SEP and so a measure of SEP is required All analyses, as the measure of SEP Age at diagnosis Date of birth/ Date of Diagnosis Age has been shown to influence receipt of treatment and so will be included as a confounder. Age at diagnosis is required to calculate survival time. Logistic regression analyses of rates of treatment, Cox proportional hazard models of survival and time to treatment Year of diagnosis Date of diagnosis Treatment trends may change over time Logistic regression analyses of rates of treatment, Cox proportional hazard models of survival and time to treatment Sex Sex Sex is included as a potential confounder Logistic regression analyses of rates of treatment, Cox proportional hazard models of survival and time to treatment Marital status Marital status Marital status may affect survival and so will be included as a confounder Logistic regression analyses of rates of treatment, Cox proportional hazard models of survival and time to treatment Age at death Date of Death/ Date of Birth Required to calculate survival time (age at death – age at diagnosis) All survival models Cause of Death = lung cancer (Y/N) Cause of Death Need to differentiate between those who died of cancer of interest and those who died of other causes Survival models DCO (Y/N) DCO Need to differentiate between those who were DCO and those who weren’t Logistic regression analysis comparing profile of those who are DCO with those who are not Cancer site (ICD10 C33, C34 (lung cancer) Cancer site These are the cancer sites that the project will investigate Incidence, rates of treatment, time to treatment and survival analyses Cancer type Cancer morphology/ type Cancer morphology may influence survival and type of treatment received Logistic regression analyses of rates of treatment, Cox proportional hazard models of survival and time to treatment 236 Table A.1 (cont). Variables that are held in the NYCRIS regional dataset, or can be derived from this Variable Derived from NYCRIS variable (by NYCRIS) Reason for inclusion Analyses to be used in Basis of diagnosis Basis of diagnosis Basis of diagnosis may influence time to treatment. Want to look at rates of treatment for histologically-confirmed cases. Logistic regression analyses of rates of treatment, Cox proportional hazard models of time to treatment Date of GP referral Date of GP referral Required to calculate primary care delay and investigation delay Cox proportional hazard models of time to referral and 20 care investigation Date of first hospital appt Date of first hospital appt Required to calculate investigation delay and diagnostic delay Cox proportional hazard model of time to 20 care investigation and diagnosis Date of diagnosis Date of diagnosis Required to calculate diagnostic delay and treatment delay Cox proportional hazard models of time to diagnosis and treatment Date of first surgery Date of surgery Required to calculate treatment delay and therapy delay Cox proportional hazard models of time to treatment and therapy Date of first chemotherapy Date of chemotherapy Required to calculate therapy delay Cox proportional hazard models of time to therapy Date of first radiotherapy Date of radiotherapy Required to calculate therapy delay Cox proportional hazard models of time to therapy Trust: (in pseudo-anonymised form as trust A,B,C etc) PCT code Receipt of, and time to, treatment may be determined by Trust factors and so this will be included as a confounder Multi-level modelling Surgeon: (in pseudo-anonymised form as surgeon 1,2,3 etc) Consultant code Survival may be associated with surgeon and so this will be included as a confounder Multi-level modelling All survival analyses 237 Table A.2. Variables that are available from the linked, anonymised, regional subset of NYCRIS and HES data, or that can be derived from them Variable Derived from HES variable (by NYCRIS) Reason for inclusion Analyses to be used in Charlson co-morbidity score (CCS) Charlson co-morbidity score (in NCDR) Co-morbidity may influence receipt of treatment and survival and so will be included as a confounder Logistic regression analyses of rates of treatment, Cox proportional hazard model of survival Nature of admission (elective/emergency) Method of Admission Nature of admission may influence survival. Emergency rather than elective admission may be a marker of late presentation Logistic regression analysis comparing profile of those who are emergency admissions with those who are not, Cox proportional hazard model of survival Post-operative complications Diagnostic codes ? Post-operative complications may contribute to delay from surgery to adjuvant therapy and influence survival Cox proportional hazard models of survival and time to adjuvant treatment Ethnicity Ethnic category Ethnicity may influence receipt of treatment and so will be included as a confounder Logistic regression analyses of rates of treatment, Cox proportional hazard models of survival and time to treatment GP(in pseudo-anonymised form as GP a,b,c etc) Referring general medical practitioner GP may influence referral time (primary care delay) Cox proportional hazard models of time to treatment Table A.3. Variables available from the linked NYCRIS regional lung cancer and LUCADA dataset Variable Derived from LUCADA variable (by NYCRIS) Reason for inclusion Analyses used in Performance status Performance status Performance status may influence receipt of treatment and survival and so will be included as a confounder Logistic regression analyses of rates of treatment, Cox proportional hazard model of survival Stage Stage Stage has been associated with receipt of treatment and survival and so will be included as a confounder Logistic regression analyses of rates of treatment, Cox proportional hazard model of survival 238 14.2 Appendix B: Data linkage and data completeness Table B.1. Completeness of stage data in the linked dataset, by year Stage Year 1 2 3 4 missing Total N % N % N % N % N % 1999 0 0 0 0 0 0 0 0 5,154 100 5,154 2000 0 0 0 0 0 0 0 0 5,227 100 5,227 2001 0 0 0 0 0 0 0 0 5,200 100 5,200 2002 0 0 0 0 0 0 0 0 5,219 100 5,219 2003 0 0 0 0 0 0 0 0 5,221 100 5,221 2004 0 0 0 0 1 0.02 0 0 5,215 99.98 5,216 2005 2 0.04 0 0 0 0 1 0.02 5,237 99.9 5,240 2006 129 2.3 45 0.8 229 4.1 268 4.7 4,862 88.2 5,533 2007 151 2.6 73 1.2 274 4.7 368 6.3 4,846 85.2 5,712 2008 229 3.8 86 1.4 464 7.8 777 13.0 4,295 74.0 5,851 2009 355 6.0 128 2.2 620 10.4 1,037 17.4 3,731 64.1 5,871 2010 322 5.5 220 3.7 686 11.7 1,308 22.2 3,230 56.9 5,766 Total 1,188 1.8 552 0.8 2,274 3.4 3,759 5.6 57,437 88.4 65,210 Table B.2. Completeness of performance status data in the linked dataset, by year Performance Status Year 0 1 2 3 4 missing Total N % N % N % N N % N % 1999 0 0 0 0 0 0 0 0 0 0 5,154 100 5,154 2000 0 0 0 0 0 0 0 0 0 0 5,227 100 5,227 2001 0 0 0 0 0 0 0 0 0 0 5,200 100 5,200 2002 0 0 0 0 0 0 0 0 0 0 5,219 100 5,219 2003 0 0 0 0 0 0 0 0 0 0 5,221 100 5,221 2004 0 0 1 0.02 0 0 0 0 0 0 5,215 99.98 5,216 2005 1 0.02 1 0.02 1 0.02 1 0.02 0 0 5,236 99.9 5,240 2006 219 4.0 301 5.4 192 3.5 139 2.5 26 0.5 4,656 84.2 5,533 2007 253 4.4 374 6.6 214 3.8 152 2.7 48 0.8 4,671 81.8 5,712 2008 332 5.7 598 10.2 402 6.9 341 5.8 107 1.8 4,071 69.6 5,851 2009 494 8.4 807 13.8 526 9.0 432 7.4 170 2.9 3,442 58.6 5,871 2010 544 9.4 883 15.3 568 9.9 539 9.4 224 3.9 3,008 52.2 5,766 Total 1,843 2.8 2,965 4.6 1,903 2.9 1,604 2.5 575 0.9 56,320 86.4 65,210 239 Table B.3. Completeness of HES linkage in the linked dataset, by year HES link No Yes Year N % N % Total 1999 596 11.6 4,558 88.4 5,154 2000 600 11.5 4,627 88.5 5,227 2001 570 11.0 4,630 89.0 5,200 2002 442 8.5 4,777 91.5 5,219 2003 474 9.1 4,747 90.9 5,221 2004 416 8.0 4,800 92.0 5,216 2005 425 8.1 4,815 91.9 5,240 2006 490 8.9 5,043 91.1 5,533 2007 442 7.7 5,270 92.3 5,712 2008 675 11.5 5,176 88.5 5,851 2009 5,758 98.1 113 1.9 5,871 2010 5,663 98.2 103 1.8 5,766 Total 16,551 25.4 48,659 74.6 65,210 Figure B.1. LUCADA case ascertainment, and stage & performance status data completeness, for England and Wales, by year (246) Figures taken from National Lung Cancer Audit Report 2012 240 14.3 Appendix C: Supporting Information for chapter 7 Appendix C1 Protocol Appendix C2 PRISMA checklist Appendix C3 Full search strategies (Medline and Embase) Appendix C4 Population over-lap and paper selection for inclusion Appendix C5 Quality score checklist Appendix C6 Assessment of existing quality assessment tools Appendix C7 Sensitivity meta-analyses Figure C7.1 Sensitivity Meta-analysis of odds of receipt of surgery in low (most deprived) versus high (least deprived) SEP (overlapping populations) Figure C7.2 Meta-analysis of odds of receipt of surgery for NSCLC in low (most deprived) versus high (least deprived) SEP (non-overlapping populations) Figure C7.3 Sensitivity meta-analysis of odds of receipt of chemotherapy in low (most deprived) versus high (least deprived) SEP (overlapping populations) Figure C7.4 Sensitivity Meta-analysis of odds of receipt of radiotherapy in low (most deprived) versus high (least deprived) SEP (overlapping populations) Figures C7.5 Sensitivity meta-analysis of odds of receipt of any type of treatment in low (most deprived) versus high (least deprived) SEP (overlapping populations) Figure C7.6 Sensitivity meta-analysis of odds of receipt of surgery in low (most deprived) versus high (least deprived) SEP (partially-overlapping populations) 241 Appendix C1. Protocol Summary Background Intervention-generated inequalities are health inequalities that result from the way that health interventions are organised and delivered and there is some evidence to suggest that intervention-generated inequalities in care may occur for some common cancers. Although the incidence and outcome of lung cancer varies with socio-economic position, it is not known whether inequalities in treatment occur and, if they do, how these might contribute to inequalities in outcome. Review objectives To summarise the existing literature and assess whether there are socio-economic differentials in receipt of treatment for lung cancer Population Adults who have a primary diagnosis of lung cancer [ICD10 C33 and C34], participating in studies published in a peer-reviewed journal up to 2011, and where the relevant outcome is analysed according to a measure of socio-economic position [including an individual or area-based measure of socio-economic position, deprivation, income, or education]. Interventions and comparators: Receipt of any curative or palliative treatment for lung cancer including surgery, chemotherapy and radiotherapy compared to not receiving surgery, chemotherapy or radiotherapy Outcomes: i) Rates of treatment; or ii) Odds of receiving treatment; looking at low [most deprived] compared to high [least deprived] SEP or trends by socio-economic strata Study design: Cohort, observational studies conducting appropriate univariable or multivariable analyses 242 C1.1 Background C1.1.1 Lung cancer Worldwide, lung cancer is the most common cancer. In the UK it is the second most common cancer for incidence overall [the second most common for men and third most common for women] (8), as well as the most common cause of cancer mortality (9). Less than 10% of those diagnosed survive for 5 years. Lung cancers are classified into small cell [SCLC] and non-small cell [NSCLC] cancers, with NSCLC accounting for 80% of lung cancers. NSCLC can be further divided into squamous cell carcinomas, adenocarcinomas and large cell carcinomas (18). NICE guidelines recommend radical surgery [pneumonectomy or lobectomy] for stage I or II NSCLC. Chemotherapy and radical radiotherapy are recommended for stage IIIa, with chemotherapy for stage IIIb and good performance-status stage IV lung cancer patients. Radiotherapy may be given as a palliative option for stage IV patients with poor performance status (18). Intervention with surgery, chemotherapy or radiotherapy has been shown to improve survival (18). The incidence and outcome rates of lung cancer vary with socio-economic position [SEP], with incidence and mortality rates 2-3 times higher in the more deprived, within the UK (8). A strong deprivation gradient for incidence (16) and mortality is also seen worldwide. However, it is not known whether inequalities in investigation and treatment occur and, if so, how these might contribute to inequalities in outcome. C1.1.2 Intervention-generated inequalities Intervention-generated inequalities are health inequalities that result from the way that health interventions are organised and delivered (7) so that although overall health may improve as the result of an intervention, differences in access to the intervention, differential uptake and delays in uptake might result in inequalities in outcome. Inequalities are likely to occur at many different stages of intervention pathways and act in a cumulative way. It is also likely that intervention-generated inequalities contribute to overall socio-economic inequalities in morbidity and mortality, although this has not been conclusively demonstrated (7). 243 Inequalities in cancer care within the UK have been noted and the NHS Cancer Plan in 2000 pledged to reduce cancer mortality, reduce delay in diagnosis and treatment and increase survival whilst acting to reduce inequalities (24). More recently the National Cancer Equality Initiative has been set up to address some of these issues (10). This is an important task as, in a 2006 review that summarised a decade of research on the association between socio-economic status and cancer survival, the authors suggested that socio-economic differences in ‘access to optimal treatment’ (26) might at least partially explain survival differences. Inequalities in access to cancer care have been shown in individual studies for a number of cancers and in a non-systematic review for colorectal cancer (11) but the evidence is inconclusive and there has been no systematic review of the evidence to demonstrate if such inequalities in access to care exist for lung cancer. C1.2 Review Objectives To summarise the existing literature and assess whether there are socio-economic differentials in receipt of treatment for lung cancer C1.3 Methods C1.3.1 Search strategy Systematic methods will be used to identify relevant studies, assess study eligibility for inclusion and evaluate study quality. A search will be undertaken to locate all studies published up to May 2011 examining care and treatment for lung cancer associated with socio-economic status. One researcher [LF] will develop the search strategy with support from her supervisors, which will then be refined with the help of a medical librarian and used to search the online databases of MEDLINE, EMBASE and Scopus. Slightly different strategies will be required for each database [for example MEDLINE recognises the MESH term Lung Neoplasms/ whereas EMBASE does not and uses Lung cancer/. See pp6-8 of this protocol for draft MEDLINE and EMBASE search strategies]. Additional studies will be identified by reviewing the reference lists of relevant studies identified from the search and by using a forward citation search to identify more recent studies that have cited an older, relevant study. EndNote software will be used to manage the references. 244 C1.3.2 Study Eligibility Cohort studies of adult participants who have a primary diagnosis of lung cancer [small-cell lung cancer or non-small-cell lung cancer - ICD10 C33 or C34], published in a peer-reviewed journal up to 2011, and where the outcome is receipt of care or treatment [measured by rates or odds of receiving care/treatment] and where the outcome is analysed by a measure of socio-economic position [such as an individual or area-based measure of SEP, deprivation, income or education] will be eligible for inclusion. Receipt of any curative or palliative treatment for lung cancer including surgery, chemotherapy and radiotherapy will be considered. Preliminary independent screening of the titles and abstracts obtained from the database searches will be carried out by two researchers [LF and HW]. Initial screening of titles will be carried out to remove obviously irrelevant papers. However, from a preliminary scoping review by LF, the early pilot searches recovered studies that, although they conducted analyses by SEP, did not always mention this in the abstract or title. Therefore, in the title search, any titles that refer to surgery, chemotherapy and radiotherapy uptake for lung cancer will be retained. Papers that look at disparities in cancer survival/mortality will also be included as further checking of the abstract is required to see if inequalities in access to treatment are also examined. Selected abstracts will then be screened and a subset of studies will be selected for further review and the full article obtained. Abstracts that refer to socio-economic inequalities in receipt of care/treatment will be retained. Abstracts that refer to racial, ethnic, geographical, sex and age-related disparities in treatment as well as disparities by insurance type will also be retained as often these papers also look at SEP, even if this is not mentioned in the abstract. Papers that look at delay will not be included. Two researchers [LF and HW] will then independently assess the selected full papers for eligibility according to the study-eligibility criteria detailed above. Any disagreements at any of the screening stages will be resolved by discussion between the two reviewers in the first instance. If agreement cannot be reached, then a third reviewer [JA or MW] will independently review the title, abstract or full paper, as appropriate, and a majority decision will be taken on inclusion/exclusion. 245 C1.3.3 Data Extraction Data extraction will be carried out by LF and HW using a pro-forma to be developed by LF for this purpose. Data relating to study authors, journal, study design, year of study, data source, number of participants, years of diagnosis, measure of SEP, confounding variables included in the analysis [such as age, sex, stage, co-morbidities, cancer type/site, vital/performance status, marital status, smoking status, cancer network, health board, hospital, emergency or elective treatment, distance from hospital/travel time, ethnicity, insurance status], type of treatment received [any, surgery, chemotherapy, radiotherapy], statistical tests carries out, outcome measures [treatment rates or odds of treatment], comparator used, significance [p values], precision [confidence intervals], other variables that were significant; will be recorded. There is evidence to suggest that insurance status is an important factor relating to access to lung cancer care in the US system (188) but is less relevant or rarely measured in the UK and Europe. Therefore studies will be split into three categories: those carried out in a healthcare system free at the point of access [similar to the UK]; those based on an insurance system [similar to the USA]; those that include a mixture of free care and social insurance-based payment [some European systems] (189). C1.3.4 Study Quality Study quality will be appraised using criteria based on the SIGN guidelines (192) and the STROBE [Strengthening the Reporting of Observational Studies in Epidemiology] guidelines (194) that contain a checklist of 22 items that should be included in cohort studies. Although the STROBE guidelines are a checklist measure of good reporting rather than ‘an instrument to evaluate the quality of observational research’ (194) , a number of other lung cancer systematic reviews use a similar scale for quality ascertainment (185, 188). The quality of reporting on the following criteria will be assessed: study design, size, setting, dates, data sources, eligibility criteria, number of participants potentially eligible, number actually included, number analysed, missing/incomplete data reported, variables included [in terms of outcome, exposure, predictors, confounders], type of statistical analysis carried out, unadjusted and adjusted estimates reported, 246 precision [confidence intervals], significance [p values] given, limitations of the study, potential bias addressed, external validity of results and funding source. C1.3.5 Statistical analysis Meta-analysis will be considered if there are sufficient studies available with suitable data. If it is not possible to conduct a meta-analysis, due to the heterogeneity of the studies, then Harvest Plot methodology will be considered. This is a method that has been devised for synthesising evidence from studies looking at the differential effects of interventions, where meta-analysis is not suitable (196). Meta-regression may also be considered if there are sufficient studies with similar variables available that might enable combined analysis of factors associated with combined outcomes. 247 Table C2.1. PRISMA 2009 Checklist Section/topic # Checklist item Reported in section paragraph #* TITLE Title 1 Identify the report as a systematic review, meta-analysis, or both. Para 1 ABSTRACT Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key findings; systematic review registration number. Para 1-4 INTRODUCTION Rationale 3 Describe the rationale for the review in the context of what is already known. Para 1-4 Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS). Para 4 METHODS Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number. Para 1 Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale. Para 2-3 Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched. Para 2 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated. Table S1 248 Table C2.1 (cont). PRISMA 2009 Checklist Section/topic # Checklist item Reported in section paragraph # Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis). Para 3-7 Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators. Para 8 Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. Para 3-7, 9 Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis. Para 10-11 Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means). Para 13 Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I2) for each meta-analysis. Para 14 From: Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 6(6): e1000097. doi:10.1371/journal.pmed1000097 For more information, visit: www.prisma-statement.org. #*: paragraphs refer to paragraphs within PLoS Medicine paper generated from this research (236) 249 Appendix C3. Search strategies Table C3.1. MEDLINE search strategy 05/05/11 Search Term Number Retrieved 1. Lung Neoplasms/ di, ep, mo, pc, rt, su, 59693 2. Exp carcinoma, Non-Small-cell Lung/or exp Carcinoma, Small Cell/ 37049 3. Or/1-2 80591 4. Social Class/ or Socio-economic Factors/ 110317 5. Socio-economic status.mp 4098 6. Education/ or exp Education, Continuing 64913 7. Income/cl, sn 3109 8. Exp Health Status/sn, td 1046 9. Exp Poverty/pc, sn, td 2309 10. Exp Social Class/ 27097 11. Socio-economic position.mp 213 12. Inequalities.mp 5837 13. Exp Social Environment/td 137 14. Social factors.mp 4869 15. Income.mp 53934 16. Exp Residence Characteristics/cl, sn 3307 17. Social Class.mp 29869 18. Education.mp 463082 19. Exp Health Status Disparities/ 3380 20. Inequities.mp 1569 21. Deprivation.mp 51228 22. Equity.mp 5183 23. Inequity.mp 920 24. Insurance status.mp 1700 25. Or/ 4-24 650350 26. surgery.mp 660227 27. treatment.mp 255751 28. exp Health Services Accessibility/cl, og, st, sn, td, ut 16904 29. exp Healthcare Disparities/ 3400 30. treatment disparities.mp 69 31. exp “Delivery of Health Care”/ 662295 32. exp Primary Health Care/sn, td, ut 6839 33. exp Drug Therapy/ 921829 34. Chemotherapy.mp or 236003 35. Radiotherapy, Adjuvant/ or Radiotherapy/ 45648 36. Radiotherapy.mp 145755 37. Accessibility.mp 56793 38. Access.mp 123272 39. Pattern$.mp 767175 40. Palliative care/ or Patient care/ or Primary Health care/ 83405 41. Care.mp 1154474 42. Investigation.mp 282065 43. Exp “Quality of Health Care”/st, sn, td, ut 99809 250 44. Exp Patient Selection/ or exp Eligibility Determination/or exp Medicaid/ 60372 45. Exp “Referral and Consultation”/ st, sn, td, ut 9243 46. Receipt.mp or exp “Patient Acceptance of Health Care”/ 137892 47. Provision.mp 33164 48. Attendance.mp 11676 49. Or/26-48 5529748 50. 3 and 25 and 49 484 51. News.pt 130842 52. Comment.pt 438297 53. Letter.pt 712489 54. Review pt 1600963 55. Editorial.pt 274165 56. 50 not (or/51-55) 398 Table C3.2. EMBASE search strategy 05/05/11 Search Term Number Retrieved 1. Exp lung cancer/ di, dm, dt, ep, rt, rh, su, th 71888 2. Exp LUNG CARCINOMA/ di, dm, dt, ep, rt, rh, su, th 43143 3. Exp lung non-Small-cell cancer/ di, dm, dt, ep, rt, rh, su, th 25650 4. exp small cell carcinoma/ di, dm, dt, ep, rt, rh, su, th 4284 5. Or/1-4 75298 6. Social Class/ or Socio-economic Factors/ 105019 7. Socio-economic status.mp 5032 8. Education/ or exp Education, Continuing 264216 9. Socio-economic position.mp 254 10. Social factors.mp 5735 11. Income.mp 60981 12. Social Class.mp 24420 13. Education.mp 463082 14. Exp LOWEST INCOME GROUP/ or exp INCOME/ 54541 15. Exp Poverty/ 23046 16. Inequality.mp 7552 17. Inequalities.mp 6594 18. Exp Social Environment/ 224660 19. Exp demography/ 114434 20. Exp health disparity/ 2317 21. Exp Health insurance/ or exp socioeconomics/ or exp Social status/ 308865 22. Inequity.mp 1106 23. Equity.mp 5975 24. Exp CULTURAL DEPRIVATION/ or deprivation.mp 53312 25. Or/ 6-24 1301033 26. surgery.mp 1211103 27. treatment.mp 3334976 28. treatment disparities.mp 92 251 29. exp “Delivery of Health Care”/ 1383093 30. exp Drug Therapy/ 1219237 31. Chemotherapy.mp or 357346 32. Radiotherapy, Adjuvant/ or Radiotherapy/ 68286 33. Radiotherapy.mp 202245 34. HEALTH CARE ACCESS/ 27526 35. Access.mp 176386 36. Care.mp 1749819 37. Pattern$.mp 824237 38. Health service/ or health care policy/ or equity.mp or health care/ or health care delivery 332536 39. Quality.mp or HEALTH CARE QUALITY/ 731294 40. Health care utilization/ 28472 41. Provision.mp 39964 42. Attendance.mp 14211 43. Receipt.mp 7755 44. Terminal care/ 18463 45. Or/26-44 7347896 46. 5 and 25 and 45 1708 47. Letter.pt 726344 48. Editorial.pt 370622 49. Note.pt 440574 50. Review.pt 1692350 51. 46 not (or/47-50) 1208 52. 51 and article.pt 1080 252 Appendix C4. Population over-lap and paper selection for inclusion C4.1 UHCS papers Three papers (198, 200, 201) all looked at the same single-year Scottish population and presented overlapping results. Two papers examined the same NYCRIS population (78, 199) but presented different results. Four papers used Thames Cancer Registry data where two of the papers used overlapping data (202, 213) and one of the papers (158) used a subset population included in both other papers. The fourth paper (ref) used more recent data and did not overlap the other 3 papers. Two papers by the same authors used LUCADA audit data, the first from 2004-2007 (153) and the second paper had a further year of audit data to 2008 (152). These were considered as one population and the paper with the larger population used where both were eligible for inclusion. Three New Zealand papers all used the same study population (205, 206, 214). The size of the UK populations examined ranged from 695 to over 60,000. The non-UK papers examined a smaller population size range – from 108 to 6449. The majority of the papers used cancer registry data, some supplemented with data from medical records, and two used data from a lung cancer audit. Audit data includes only a subset of registry patients whose data has been entered into the audit. Three UK papers obtained data from Hospital Episode Statistics [HES] and the three New Zealand papers included only patients who were admitted to secondary care. Papers that used only HES/hospital data were able to look at inequalities in care only for those admitted to hospital, not the entire registry population. C4.2 Non-UHCS studies There is also overlap within the SEER dataset so again only substantially non-overlapping studies were included in the final meta-analyses. Alternative analyses including all eligible studies were also run, as well as analyses including partially overlapping studies but not those that were fully a subset of another. 253 C4.3 Choosing the most suitable studies for inclusion in meta-analysis Problems did arise where two studies used the same population, or a subset of a population, but found slightly different results. Quality score was used as the initial selection factor for the most appropriate study to choose for inclusion in the meta-analysis. In some cases the smaller regional analyses were of better quality as they included stage and histology, whereas the analysis of the national population did not. For example the Berglund (2012) paper looked at patients diagnosed between 2006 and 2008 in the Thames Cancer Registry region (157). However, Riaz (2012) looked at England-wide data diagnosed between 2004 and 2006 (161) and so there is likely to be some overlap between these populations for 2006. So although these are clearly different populations they are not entirely unique populations. The Berglund study is smaller (1826 in the surgery population) compared to Riaz (77,349) but Berglund includes stage and co-morbidity and so is the better quality study. The two Rich [2011] studies (152, 153) include England-wide data for 2004-2007 (24,175 patients) and 2004-2008 (34,436 patients) respectively, obtained from the LUCADA audit and so these two had clearly overlapping populations where it would only be suitable to include one in the meta-analysis. But again some of the same patients are likely to be included here that are also found in the Riaz (161) and Berglund (157) studies, although different sources have been used [LUCADA rather than Registry data]. Therefore, although the overlap may be small, strictly speaking only one of these four studies should be included in the final meta-analysis. The quality score was therefore used to decide this. Berglund (157) was the smallest study but the best quality. Sensitivity analyses including all the over-lapping populations meant that three large, reasonable quality studies were still able to supply data to the meta-analyses. Further analyses were then run where only non-overlapping populations were included and, where it was not clear which was the best of the over-lapping populations to include, multiple analyses were run using different study combinations, to see if this made any difference to the overall outcome. It was therefore possible to examine whether it might be better to choose a slightly less-good quality study that contained more data, 254 for example, the larger Rich paper (152) as it contained stage and co-morbidity but did not capture everyone diagnosed within the timescale, or the Riaz paper (161) which had better population capture but did not include important confounders such as stage and co-morbidity. 255 Appendix C5: Quality Score Quality Score Characteristics of studies 6 Multi-variable analysis. Population-based sample. Good internal validity/reporting/confounding 5 Multi-variable analysis. Selective population. Good internal validity/reporting/confounding 4 Multi-variable analysis. Population-based sample. Some problems with internal validity/reporting/confounding (eg good internal validity/reporting but stage not included OR less good internal validity/reporting and stage included) 3 Multi-variable analysis. Selective population. Some issues with internal validity/reporting/confounding (eg good internal validity/reporting but stage not included OR less good internal validity/reporting and stage included) 2 Univariable analysis. Good internal validity/reporting OR multivariable analysis but only univariable results reported or no CIs/ different comparator/ stratified by other variable 1 Univariable analysis. Poor internal validity/reporting OR multivariable analysis but results for SEP not shown or errors in data Quality checklist used to derive quality scores Screening questions: Does the study conduct multivariable analysis and report adjusted odds ratios/rates? Yes – consider for meta-analysis No – consider for narrative review What population is included? a. Population-based sample: eg total local or national lung cancer population from a cancer Registry b. Selective population: eg. hospital population from Hospital Episode Statistics (HES) data or similar, or from Registry linked to Medicare records (USA) or incomplete audit data Section 1: Internal validity 1. Appropriate, valid and reliable measure of SEP is used (e.g. IMD, Carstairs, Townsend, similar local index, income, poverty level, education) a. Yes - individual standard measure used 5 b. Yes - area-based standard measure used 3 c. Standard measure used but presented as average for PCT/health authority 1 d. Non-standard measure used/ measure not reported 0 2. SEP categorised as a. Continuous/Deciles 5 b. Quintiles/ Quartiles 3 c. Tertiles/Dichotimised 1 d. Unknown/ not reported 0 3. SEP expressed as an OR to a. Two decimal places 5 256 b. One decimal place 3 c. Whole number/ expressed as rate 1 d. unknown/not reported 0 4. Outcome measures are valid and reliable a. Yes - care/ treatment details obtained from Registry, HES or similar system 3 b. Yes - care/treatment details obtained from audit or hospital records 3 c. Care treatment details obtained from survey/questionnaire/other 1 d. No/unknown/not reported 0 Good: 14-18 OK: 13-11 Poor: 0-10 Section 2: External validity 1. Population a. Study uses multiple Registry or national population(s) or similar 3 b. Study uses regional population from Registry or HES or similar 2 c. Study uses only small subset of population (some health boards/PCTs/areas) 1 d. Study uses small, random sample from Registry or hospital data /other 1 e. Population not stated 0 2. Population date and time period a. Multiple years of diagnosis, some post-2000 3 b. Multiple years of diagnosis, all pre-2000 2 c. Single year of diagnosis, post -2000 2 d. Single year of diagnosis, pre-2000 1 Good: 4-6 Poor: 0-3 Section 3: Reporting of Study 1. Outcome measures are clearly defined/reported a. Yes 1 b. No 0 2. Outcome measure reported a. Rates/odds of treatment compared with no treatment 2 b. Rates/odds of NOT receiving treatment compared to receiving treatment 2 c. Rates/odds of treatment compared with other care 1 d. Rates/odds of treatment stratified by other variable (sex, race etc) 1 e. Results presented in some other way 0 3. Number initially eligible/ number excluded reported a. Eligible/excluded/included all reported or able to be calculated 2 b. Number included reported only 1 c. Numbers not reported 0 4. Inclusion/ exclusion criteria detailed a. Yes 1 b. No 0 257 5. Numbers receiving treatment a. Numerator and denominator populations clearly documented 2 b. Numbers calculable from details given but not clearly specified 1 c. Numbers not specified 0 d. Numbers do not add up correctly and need to be checked with authors 0 6. Other variables that are significant in analysis reported a. Yes 2 b. Yes, but results for only some presented 1 c. No 0 7. Death Certificate Only (DCO) excluded a. Yes 1 b. Not applicable (if using HES type data) 1 c. No/ not reported 0 8. Confidence interval reported a. Yes 1 b. No 0 9. P value reported a. Overall p value/ p for trend 2 b. Individual p values 1 c. Not reported 0 Good: 11-14 OK: 7-10 Poor: 0-6 Section 4: Confounding 1. Multivariable analysis - other important confounders included a. Age and sex 3 b. Age or Sex 2 c. Univariable analysis only reported 1 d. Descriptive only/ no analysis 0 2. Results stratified by stage a. Yes, and only eligible- stage patients used for denominator 3 b. Yes, but all-stage patients used for denominator 2 c. No, but stage included as a confounder 2 d. No 0 3. Results stratified by histology a. Yes, and only histologically-verified cases included 3 b. Yes, but clinically diagnosed and histologically-unknown cases included 2 c. No, but histology included as a confounder 2 d. No 0 4. Other relevant confounders included a. Co-morbidity/ performance status 2 b. Trust/ health board/ hospital/area 1 c. No 0 Good: 7-11 (must include age, sex, stage and histology) OK: 3-5 (must include age, sex)Poor: 0-2 (univariable analysis) 258 Appendix C6. Assessment of existing quality assessment tools The Cochrane tool for assessment of risk of bias is only suitable for randomised controlled trials as the tool examines aspects of studies such as randomisation and blinding which are not relevant in cohort studies. However, the measures of the four different types of biases identified [selection, performance, attrition and detection] can be adapted for other types of studies. For example RCTs use randomisation to prevent selection bias but in cohort studies controlling for confounders can be considered instead (190). Measurement of exposure in cohort studies rather than blinding of participants/investigators can be used to check for performance bias. Completeness of follow-up to assess attrition bias and blinded outcome assessment for detection bias can be examined for cohort studies as they can for RCTs (190). Other tools used by Cochrane Review Groups include the Jadad scale (191) which again is only suitable for RCTs. The Moncrief scale includes items on external validity and reporting of statistical analysis which are not necessarily measures of bias (191). The Newcastle-Ottawa scale is recommended by the Musculoskeletal Cochrane Group (191). This scale has been developed for assessing the quality of non-randomised studies and so is potentially suitable for cohort studies. Studies are assessed on selection, comparability and outcome and so this does appear to be more of a measure of the internal validity of a study than some of the other scales examined. The scale is divided into: Selection of the cohorts [4 items]; Comparability of the cohorts [1 item]; and Assessment of outcome [3 items]. ‘High quality’ outcomes are identified with a star, with a maximum of 1 star per item in the selection category and 2 stars per item in the comparability category [but the number of stars in the outcome category is not stated]. These star charts can then be presented alongside the meta-analysis. Although published Cochrane reviews do utilise scales and many of their Review Groups recommend them (191) their validity has not been demonstrated in empirical research. Therefore the validity of the scale is unclear and the definition of ‘good’ and ‘poor’ scores appears somewhat subjective. Also, the scale does not appear to have been published as, after online searching, only a power point presentation was available. 259 SIGN, the Scottish Intercollegiate Guidelines Network has a methodology checklist for cohort studies (192). This examines internal validity under the following headings: selection of subjects, confounding, statistical analysis. Each criterion [for example: ‘The outcomes are clearly defined’] is assessed as: well covered, adequately addressed, poorly addressed, not addressed, not reported, not applicable. In the overall assessment of the study a score of ++, + or – is given for how well has the study minimised the risk of bias or confounding and a judgment made on applicability and certainty that the overall effect is due to the study intervention. A more extensive checklist than SIGN is utilised by the Effective Public Health Practice Project [EPHPP] Quality Assessment Tool for Quantitative Studies in Canada (193), which looks at selection bias, study design, confounders, blinding, data collection methods, withdrawals and drop-outs, intervention integrity and analyses. Each section is rated as strong, moderate or weak. Not all of these categories are relevant for the type of studies included in this review as all are cohort studies, no blinding takes place and as participants are not actively recruited they cannot drop out, but parts of it could be adapted and utilised. The STROBE [Strengthening the Reporting of Observational Studies in Epidemiology] guidelines (194) contain a checklist of 22 items that should be included in cohort studies. The quality of reporting on the following criteria can be assessed: study design, size, setting, dates, data sources, eligibility criteria, number of participants potentially eligible, number actually included, number analysed, missing/incomplete data reported, variables included [in terms of outcome, exposure, predictors, confounders], type of statistical analysis carried out, unadjusted and adjusted estimates reported, precision [confidence intervals], significance [p values] given, limitations of the study, potential bias addressed, external validity of results, funding source. Although the STROBE guidelines are a checklist measure of good reporting rather than ‘an instrument to evaluate the quality of observational research’ (194), a number of other lung cancer systematic reviews use some of the above items in their scales for quality ascertainment (185, 188). However, just because something is not reported 260 does not necessarily mean it was not done. However, in order to determine this it may be necessary to contact the authors of the study. Conclusion Following review of the potential tools identified above, criteria adapted from SIGN, EPHPP and STROBE guidelines were used to develop a quality assessment tool for this systematic review. 261 Appendix C7: Sensitivity meta-analyses Figure C7.1. Meta-analysis of odds of receipt of surgery in low (most deprived) versus high (least deprived) SEP. (over-lapping populations; n=16) Figure C7.2. Meta-analysis of odds of receipt of surgery for NSCLC in low (most deprived) versus high (least deprived) SEP. (non-overlapping populations; n=8) 262 Figure C7.3. Meta-analysis of odds of receipt of chemotherapy in low (most deprived) versus high (least deprived) SEP (over-lapping populations; n=10). Figure C7.4. Meta-analysis of odds of receipt of radiotherapy in low (most deprived) versus high (least deprived) SEP (overlapping populations; n=9). 263 Figure C7.5. Meta-analysis of odds of receipt of any type of treatment in low (most deprived) versus high (least deprived) SEP (overlapping populations; n=35). 264 Figure C7.6. Meta-analysis of odds of receipt of surgery in low versus high SEP. (partially-overlapping populations, n=14) 265 14.4 Appendix D: Supplementary tables for chapters 11 and 12 Table D.1. Odds of FHA within 2 weeks from GP referral (for those with FHA within 1 year of GP ref) 1999-2005 Variables FHA within 1 year FHA within 2 weeks Univariable analysis (n=16,649) Multivariable analysis1 (n=16,649, R2=3.42) N N % OR 95% CI P OR 95% CI P IMD 16,649 12,241 73.5 0.28 0.16 1 1,815 1,357 74.8 1.00 1.00 2 2,400 1,792 74.7 0.99 0.86 1.14 0.94 0.96 0.83 1.10 0.55 3 2,796 2,043 73.1 0.92 0.80 1.05 0.20 0.89 0.77 1.02 0.09 4 3,847 2,793 72.6 0.89 0.79 1.02 0.09 0.86 0.76 0.98 0.03 5 5,791 4,256 73.5 0.94 0.83 1.06 0.28 0.90 0.80 1.02 0.10 Age Range 16,649 12,241 73.5 <0.001 <0.001 <60 2,561 1,970 76.9 1.00 1.00 60-69 4,707 3,517 74.7 0.89 0.79 0.99 0.04 0.92 0.82 1.03 0.14 70-79 6,635 4,834 72.9 0.81 0.72 0.90 <0.001 0.86 0.77 0.96 0.01 80+ 2,746 1,920 69.9 0.70 0.62 0.79 <0.001 0.79 0.69 0.89 <0.001 Sex 16,649 12,241 73.5 0.76 0.59 Female 6,909 5,071 73.4 1.00 1.00 Male 9,740 7,170 73.6 1.01 0.94 1.08 0.76 1.02 0.95 1.10 0.59 Histology 16,649 12,241 73.5 <0.001 <0.001 NSCLC 9,865 7,236 73.4 1.00 1.00 SCLC 2,341 1,905 81.4 1.59 1.42 1.78 <0.001 1.59 1.42 1.79 <0.001 Other 4,443 3,100 69.8 0.84 0.78 0.91 <0.001 0.88 0.81 0.96 <0.001 Year of Diagnosis 16,649 12,241 73.5 <0.001 <0.001 1999 2,345 1,608 68.6 1.00 1.00 2000 2,355 1,714 72.8 1.59 1.41 1.79 <0.001 1.24 1.10 1.41 <0.001 2001 2,404 1,860 77.4 1.95 1.72 2.20 <0.001 1.60 1.40 1.82 <0.001 2002 2,379 1,865 78.4 2.49 2.20 2.82 <0.001 1.69 1.48 1.93 <0.001 2003 2,422 1,954 80.7 2.64 2.33 3.00 <0.001 1.96 1.71 2.24 <0.001 2004 2,350 1,855 78.9 3.04 2.67 3.46 <0.001 1.76 1.54 2.01 <0.001 2005 2,394 1,385 57.9 2.73 2.40 3.10 <0.001 0.64 0.57 0.72 <0.001 Co-morbidity score 16,649 12,241 73.5 <0.001 <0.001 0 2,681 1,888 70.4 1.00 1.00 1-2 1,151 763 66.3 0.83 0.71 0.96 0.01 0.87 0.75 1.01 0.07 3+ 175 110 62.9 0.71 0.52 0.98 0.04 0.76 0.55 1.06 0.10 CCM missing 11,178 8,438 75.5 1.29 1.18 1.42 <0.001 1.29 1.18 1.42 <0.001 No HES link 1,464 1,042 71.2 1.04 0.90 1.19 0.61 1.19 1.03 1.38 0.02 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score 266 Table D.2. Odds of receiving any treatment within 31 days of diagnosis (for those treated within 1 year of diagnosis) 1999-2005 Treated Treated within 31 days Univariable analysis (n=19,510) Multivariable analysis1 (n=19,510, R2=10.50) N N % OR 95% CI P OR 95% CI P IMD 19,510 8,457 43.4 0.004 0.001 1 2,208 1,036 46.9 1.00 1.00 2 2,828 1,247 44.1 0.89 0.80 1.00 0.05 0.91 0.81 1.03 0.12 3 3,182 1,371 43.1 0.86 0.77 0.96 0.005 0.88 0.78 0.98 0.03 4 4,405 1,866 42.4 0.83 0.75 0.92 <0.001 0.83 0.74 0.92 0.001 5 6,887 2,937 42.7 0.84 0.76 0.93 <0.001 0.82 0.74 0.91 <0.001 Age Range 19,510 8,457 43.4 <0.001 <0.001 <60 4,073 2,085 51.2 1.00 1.00 60-69 6,389 2,886 45.2 0.79 0.73 0.85 <0.001 0.83 0.76 0.90 <0.001 70-79 7,274 2,887 39.7 0.63 0.58 0.68 <0.001 0.77 0.71 0.84 <0.001 80+ 1,774 599 33.8 0.49 0.43 0.55 <0.001 0.68 0.60 0.78 <0.001 Sex 19,510 8,457 43.4 <0.001 0.58 Female 8,060 3,641 45.2 1.00 1.00 Male 11,450 4,816 42.1 0.88 0.83 0.93 <0.001 0.98 0.92 1.05 0.58 Histology 19,510 8,457 43.4 <0.001 <0.001 NSCLC 12,551 4,269 34.0 1.00 1.00 SCLC 3,705 2,867 77.4 6.64 6.09 7.23 <0.001 4.76 4.32 5.25 <0.001 Other 3,254 1,321 40.6 1.33 1.22 1.43 <0.001 1.41 1.30 1.53 <0.001 Co-morbidity score 19,510 8,457 43.4 <0.001 0.28 0 3,464 1,469 42.4 1.00 1.00 1-2 1,790 726 40.6 0.93 0.83 1.04 0.20 0.98 0.87 1.12 0.81 3+ 273 110 40.3 0.92 0.71 1.18 0.50 1.01 0.77 1.32 0.96 CCM missing 13,210 5,866 44.4 1.08 1.01 1.17 0.04 1.06 0.97 1.15 0.18 No HES link 773 286 37.0 0.80 0.68 0.94 0.006 1.16 0.98 1.38 0.08 Diagnosis year 19,510 8,457 43.4 0.10 0.01 1999 2,844 1,246 43.8 1.00 1.00 2000 2,901 1,225 42.2 0.94 0.84 1.04 0.23 0.87 0.77 0.97 0.01 2001 2,776 1,200 43.2 0.98 0.88 1.09 0.66 0.96 0.86 1.08 0.49 2002 2,766 1,155 41.8 0.92 0.83 1.02 0.12 0.86 0.76 0.96 0.009 2003 2,705 1,230 45.5 1.07 0.96 1.19 0.21 1.00 0.89 1.12 1.00 2004 2,701 1,156 42.8 0.96 0.86 1.07 0.45 0.86 0.77 0.97 0.01 2005 2,817 1,245 44.2 1.02 0.91 1.13 0.77 0.95 0.85 1.07 0.39 GP referral 19,510 8,457 43.4 <0.001 <0.001 No GP referral date 8,645 4,117 47.6 1.00 1.00 <=14 days to FHA 8,107 3,424 42.2 0.80 0.76 0.85 <0.001 0.79 0.73 0.84 <0.001 >14 days to FHA 2,758 916 33.2 0.55 0.50 0.60 <0.001 0.59 0.54 0.65 <0.001 Ist treatment 19,510 8,457 43.4 <0.001 <0.001 Chemotherapy 6,949 4,248 61.1 1.00 1.00 Surgery 3,430 1,330 38.8 0.40 0.37 0.44 <0.001 0.76 0.69 0.83 <0.0011 Radiotherapy 9,131 2,879 31.5 0.29 0.27 0.31 <0.001 0.53 0.49 0.58 <0.001 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, GP referral, first treatment 267 Table D.3. Odds of receiving any treatment within 62 days of GP referral (for those with FHA within 1 year of referral, diagnosis within 1 year of FHA and treated within 1 year of diagnosis) 1999-2005 Treated Treated within 62 days Univariable analysis (n=10,844) Multivariable analysis1 (n=10,844, R2=9.71) N N % OR 95% CI P OR 95% CI P IMD 10,844 5,326 49.1 0.005 0.0002 1 1,220 646 53.0 1.00 1.00 2 1,628 825 50.7 0.91 0.79 1.06 0.23 0.92 0.79 1.08 0.32 3 1,796 901 50.2 0.89 0.77 1.03 0.13 0.89 0.76 1.04 0.16 4 2,476 1,173 47.4 0.80 0.70 0.92 0.001 0.78 0.67 0.90 0.001 5 3,724 1,781 47.8 0.81 0.72 0.93 0.002 0.77 0.67 0.88 <0.001 Age Range 10,844 5,326 49.1 <0.001 <0.001 <60 2,146 1,233 57.5 1.00 1.00 60-69 3,616 1,843 51.0 0.77 0.69 0.86 <0.001 0.82 0.73 0.92 0.001 70-79 4,054 1,826 45.0 0.61 0.55 0.67 <0.001 0.74 0.66 0.83 <0.001 80+ 1,028 424 41.3 0.52 0.45 0.60 <0.001 0.70 0.59 0.82 <0.001 Sex 10,844 5,326 49.1 0.38 0.21 Female 4,436 2,201 49.6 1.00 1.00 Male 6,408 3,125 48.8 0.97 0.90 1.04 0.38 1.05 0.97 1.14 0.21 Histology 10,844 5,326 49.1 <0.001 <0.001 NSCLC 7,290 3,091 42.4 1.00 1.00 SCLC 1,911 1,550 81.1 5.83 5.15 6.60 <0.001 3.66 3.19 4.20 <0.001 Other 1,643 685 41.7 0.97 0.87 1.08 0.60 1.04 0.93 1.16 0.54 Co-morbidity score 10,844 5,326 49.1 <0.001 <0.001 0 1,824 799 43.8 1.00 1.00 1-2 729 316 43.4 0.98 0.83 1.17 0.83 1.05 0.87 1.26 0.61 3+ 97 33 34.0 0.66 0.43 1.02 0.06 0.73 0.46 1.15 0.17 CCM missing 7,763 3,990 51.4 1.36 1.22 1.50 <0.001 1.32 1.18 1.47 <0.001 No HES link 431 188 43.6 0.99 0.80 1.23 0.94 1.38 1.11 1.73 0.004 Diagnosis year 10,844 5,326 49.1 0.01 0.04 1999 1,541 745 48.4 1.00 1.00 2000 1,530 697 45.6 0.89 0.78 1.03 0.12 0.86 0.74 1.00 0.05 2001 1,530 767 50.1 1.07 0.93 1.24 0.32 1.07 0.92 1.25 0.37 2002 1,552 747 48.1 0.99 0.86 1.14 0.91 0.92 0.79 1.07 0.28 2003 1,581 778 49.2 1.04 0.90 1.19 0.63 0.96 0.82 1.12 0.60 2004 1,534 766 49.9 1.07 0.93 1.23 0.38 0.96 0.82 1.12 0.58 2005 1,576 826 52.4 1.18 1.02 1.35 0.02 1.08 0.93 1.26 0.33 1st treatment 10,844 5,326 49.1 <0.001 <0.001 Chemotherapy 3,985 2,732 68.6 1.00 1.00 Surgery 1,882 619 32.9 0.22 0.20 0.25 <0.001 0.36 0.32 0.41 <0.001 Radiotherapy 4,977 1,975 39.7 0.30 0.28 0.33 <0.001 0.50 0.45 0.56 <0.001 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, first treatment 268 Table D.4. Hazard ratio of early first hospital appointment from GP referral (for those referred within 1 year) in 2006-2010 cohort with stage recorded Median time to FHA (days) Univariable analysis (n=5100) Multivariable analysis (n=5100) N N IQR HR 95% CI P HR 95% CI P IMD 5100 9 4-14 0.28 0.16 1 594 8 4-14 1.00 1.00 2 705 9 5-14 0.93 0.84 1.04 0.21 0.95 0.85 1.06 0.32 3 839 9 5-15 0.90 0.81 1.00 0.05 0.89 0.80 0.99 0.03 4 1198 9 5-14 0.91 0.83 1.01 0.07 0.90 0.81 0.99 0.03 5 1764 9 4-15 0.95 0.87 1.04 0.27 0.93 0.85 1.02 0.13 Age Range 5100 9 4-14 0.34 <60 682 9 4-14 1.00 1.00 60-69 1490 9 5-14 1.01 0.93 1.11 0.75 1.00 0.91 1.10 1.00 70-79 1876 9 5-15 0.95 0.87 1.04 0.30 0.93 0.85 1.02 0.11 80+ 1052 8 3-15 0.99 0.90 1.09 0.79 0.91 0.82 1.01 0.07 Sex 5100 9 4-14 0.18 0.46 female 2297 9 4-14 1.00 1.00 male 2803 9 5-15 0.96 0.91 1.02 0.18 0.98 0.93 1.04 0.46 Histology 5100 9 4-14 <0.001 NSCLC 3518 9 6-15 1.00 1.00 SCLC 386 8 4-13 1.23 1.11 1.37 0.00 1.10 0.99 1.22 0.09 Other 1196 8 1-14 1.14 1.07 1.22 0.00 1.06 0.99 1.14 0.11 Year of Diagnosis 5100 9 4-14 <0.001 <0.001 2006 395 14 9-23 1.00 1.00 2007 473 14 8-22 1.03 0.90 1.17 0.70 1.00 0.88 1.15 0.98 2008 968 7.5 4-12 1.70 1.51 1.92 <0.001 1.64 1.46 1.85 <0.001 2009 1492 8 3-13 1.68 1.50 1.87 <0.001 1.78 1.56 2.04 <0.001 2010 1772 8 3-14 1.53 1.37 1.71 <0.001 1.59 1.38 1.82 <0.001 Co-morbidity score 5100 9 4-14 0.005 0 618 9 5-15 1.00 1.00 1-2 508 8 3-18 0.93 0.82 1.07 0.31 0.93 0.83 1.05 0.26 3+ 110 8 2-15 1.05 0.88 1.25 0.59 0.88 0.72 1.08 0.24 CCM missing 1259 10 6-16 1.03 0.84 1.26 0.77 1.19 1.06 1.32 0.002 No HES link 2605 8 3-14 1.11 1.02 1.21 0.01 1.02 0.92 1.12 0.75 Stage 5100 9 4-14 <0.001 <0.001 I 773 11 7-20 1.00 1.00 II 399 11 5-16 1.20 1.06 1.35 0.003 1.20 1.06 1.36 0.004 III 1578 9 6-15 1.26 1.16 1.38 <0.001 1.23 1.12 1.34 <0.001 IV 2350 8 3-14 1.45 1.33 1.57 <0.001 1.35 1.24 1.47 <0.001 Performance Status 5100 9 4-14 <0.001 0 1111 11 6-15 1.00 1.00 1-2 2634 9 5-15 1.05 0.97 1.12 0.22 1.02 0.95 1.10 0.56 3-4 1016 6 0-13 1.39 1.28 1.51 <0.001 1.32 1.20 1.45 <0.001 Missing/ unknown 339 11 7-18 0.93 0.82 1.05 0.24 1.01 0.89 1.15 0.84 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, stage, PS 269 Table D.5. Likelihood of still being alive 2 years after diagnosis, by selected patient, tumour and system factors for those diagnosed between 1999 and 2005 (DCO cases excluded) 1 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, timely GP referral 2 Mutually adjusted for SEP, age, sex, histology, year of diagnosis, co-morbidity score, timely GP referral, type of treatment variable Adjusted – selected1 (n=36,477, R2=5.80) Adjusted – selected2 (n=36,477, R2=24.60) OR 95% CI P OR 95% CI P Deprivation quintile 0.003 0.64 1 (least deprived) 1.00 1.00 2 0.97 0.86 1.10 0.68 1.02 0.89 1.18 0.74 3 0.90 0.80 1.01 0.08 1.01 0.88 1.15 0.94 4 0.86 0.77 0.96 0.01 0.98 0.86 1.12 0.80 5 (most deprived) 0.84 0.76 0.94 0.002 1.06 0.94 1.20 0.38 Age group <0.001 0.15 <60 1.00 1.00 60-69 0.85 0.78 0.93 <0.001 0.99 0.89 1.09 0.78 70-79 0.61 0.56 0.66 <0.001 0.95 0.86 1.06 0.35 80+ 0.32 0.29 0.36 <0.001 0.86 0.74 0.99 0.03 Sex <0.001 <0.001 Female 1.00 1.00 Male 0.74 0.69 0.78 <0.001 0.71 0.66 0.76 <0.001 Histology <0.001 <0.001 NSCLC 1.00 1.00 SCLC 0.37 0.33 0.41 <0.001 0.59 0.52 0.67 <0.001 Other 0.59 0.55 0.64 <0.001 1.05 0.96 1.15 0.28 Year of Diagnosis <0.001 <0.001 1999 1.00 1.00 2000 1.17 1.04 1.32 0.009 1.17 1.02 1.34 0.03 2001 1.12 0.99 1.26 0.07 1.17 1.02 1.35 0.02 2002 1.15 1.02 1.29 0.03 1.21 1.05 1.39 0.01 2003 1.22 1.08 1.37 0.001 1.30 1.13 1.49 <0.001 2004 1.20 1.07 1.36 0.002 1.32 1.15 1.52 <0.001 2005 1.18 1.05 1.33 0.01 1.26 1.10 1.44 <0.001 Co-morbidity score <0.001 0.18 0 1.00 1.00 1-2 0.85 0.75 0.96 0.01 0.98 0.85 1.13 0.77 3+ 0.67 0.51 0.88 0.004 0.76 0.55 1.04 0.08 CCM missing 0.82 0.75 0.88 <0.001 0.91 0.83 1.00 0.06 No HES link 0.46 0.39 0.54 <0.001 0.90 0.76 1.07 0.25 Timely GP referral <0.001 <0.001 No GP referral 1.00 1.00 FHA<=14 days 1.42 1.32 1.52 <0.001 1.24 1.14 1.34 <0.001 FHA>14 days 2.08 1.90 2.28 <0.001 1.80 1.62 1.99 <0.001 Type of treatment <0.001 No treatment 1.00 Surgery 44.34 39.20 50.15 <0.001 Surgery + chemo/radiotherapy 19.90 16.68 23.74 <0.001 Chemotherapy 2.13 1.82 2.50 <0.001 Chemotherapy +radiotherapy 5.81 5.06 6.68 <0.001 Radiotherapy 2.73 2.45 3.05 <0.001 270 Chapter 15. References 1. Acheson Report. Independent Inquiry into Inequalities and Health. 1999. 2. Townsend P, Davidson N, editors. Inequalities in Health: The Black Report. Suffolk: Penguin books; 1982. 3. Nuffield Council on Bioethics. Public Health: Ethical Issues Report. London: 2007. 4. Charlton BG, White M. Living on the margin: a salutogenic model for socio-economic differentials in health. Public Health. 1995;109(4):235-43. 5. Galobardes B, Shaw M, Lawlor DA, Lynch JW, Davey Smith G. Indicators of socioeconomic position (part 1). Journal of Epidemiology and Community Health. 2006;60(1):7-12. 6. Krieger N, Williams DR, Moss NE. Measuring Social Class in US Public Health Research: Concepts, Methodologies, and Guidelines. Annual Review of Public Health. 1997;18(1):341-78. 7. White M, Adams J, Heywood P. How and why do interventions that increase health overall widen inequalities within populations? In: Babones S, editor. Health, inequality and society. Bristol: Policy Press; 2009. p. 65-83. 8. Cancer Research UK. Cancer incidence: UK statistics [18/12/13]. Available from: http://info.cancerresearchuk.org/cancerstats/incidence/. 9. Cancer Research UK. Common cancers - UK mortality statistics [18/12/13]. Available from: http://info.cancerresearchuk.org/cancerstats/mortality/cancerdeaths/. 10. Department of Health. Reducing cancer inequality: evidence, progress and making it happen - a report by the National Cancer Equality Initaitive. 2010. 11. Aarts MJ, Lemmens VEPP, Louwman MWJ, Kunst AE, Coebergh JWW. Socioeconomic status and changing inequalities in colorectal cancer? A review of the associations with risk, treatment and outcome. European Journal of Cancer. 2010;46(15):2681-95. 12. Shack L, Jordan C, Thomson C, Mak V, Moller H. Variation in incidence of breast, lung and cervical cancer and malignant melanoma of skin by socioeconomic group in England. BMC Cancer. 2008;8(1):271. 13. Howlader N NA, Krapcho M, Neyman N, Aminou R, Altekruse SF, Kosary CL, Ruhl J, Tatalovich Z, Cho H, Mariotto A, Eisner MP, Lewis DR, Chen HS, Feuer EJ, Cronin KA SEER Cancer Statistics Review, 1975-2009 (Vintage 2009 Populations) National Cancer Institute. Bethesda, MD2011. based on November 2011 SEER data submission, posted to SEER web site 2]. Available from: http://seer.cancer.gov/csr/1975_2009_pops09/. 14. Coleman MP, Forman D, Bryant H, Butler J, Rachet B, Maringe C, et al. Cancer survival in Australia, Canada, Denmark, Norway, Sweden, and the UK, 1995-2007 (the International Cancer Benchmarking Partnership): an analysis of population-based cancer registry data. The Lancet. 2011;377(9760):127-38. 15. Verdecchia A, Francisci S, Brenner H, Gatta G, Micheli A, Mangone L, et al. Recent cancer survival in Europe: a 2000–02 period analysis of EUROCARE-4 data. The Lancet Oncology. 2007;8(9):784-96. 16. Sidorchuk A, Agardh E, Aremu O, Hallqvist J, Allebeck P, Moradi T. Socioeconomic differences in lung cancer incidence: a systematic review and meta-analysis. Cancer Causes and Control. 2009;20(4):459-71. 17. Sharpe KH, McMahon AD, McClements P, Watling C, Brewster DH, Conway DI. Socioeconomic inequalities in incidence of lung and upper aero-digestive tract cancer by age, tumour subtype and sex: A population-based study in Scotland (2000–2007). Cancer Epidemiology. 2012;36(3):e164-e70. 18. NICE. Clinical Guideline 24. Lung Cancer: the diagnosis and treatment of lung cancer. London: NICE; 2005. 19. Hirsch FR, Spreafico A, Novello S, Wood MD, Simms L, Papotti M. The Prognostic and Predictive Role of Histology in Advanced Non-small Cell Lung Cancer: A Literature Review. Journal of Thoracic Oncology. 2008;3(12):1468-81 10.097/JTO.0b013e318189f551. 271 20. Riaz SP, Lüchtenborg M, Coupland VH, Spicer J, Peake MD, Møller H. Trends in incidence of small cell lung cancer and all lung cancer. Lung Cancer. 2012;75(3):280-4. 21. Cancer Research UK. More about staging for lung cancer [27/01/14]. Available from: http://www.cancerresearchuk.org/cancer-help/type/lung-cancer/treatment/more-about-lung-cancer-staging. 22. Baldwin DR, White B, Schmidt-Hansen M, Champion AR, Melder AM. Diagnosis and treatment of lung cancer: summary of updated NICE guidance. BMJ. 2011;342. 23. The Information Centre for health and social care. National Lung Cancer Audit. Leeds. : 2006. 24. Department of Health. The NHS cancer plan: a plan for investment, a plan for reform. 2000. 25. Coleman MP, Rachet B, Woods LM, Mitry E, Riga M, Cooper N, et al. Trends and socioeconomic inequalities in cancer survival in England and Wales up to 2001. Br J Cancer. 2004;90(7):1367-73. 26. Woods LM, Rachet B, Coleman MP. Origins of socio-economic inequalities in cancer survival: a review. Annals of Oncology. 2006;17(1):5-19. 27. Coleman MP, Babb P, Sloggett A, Quinn MJ, De Stavola B. Socio-economic inequalities in cancer survival in England and Wales. Cancer. 2001;91:208-16. 28. Murdoch C, Wilkinson J, Unsworth L. Lung resection rates across the North of England Cancer Network. NEPHO., 2010. 29. Brown JJS. Inequality and pancreatic cancer [MPhil thesis]. Newcastle upon Tyne: Newcastle University; 2006. 30. Welch V, Tugwell P, Wells GA, Kristjansson B, Petticrew M, McGowan JL, et al. How effects on health equity are assessed in systematic reviews of interventions (protocol). The Cochrane Collaboration. 2009(3). 31. Tudor Hart J. The inverse care law. The Lancet. 1971;297(7696):405. 32. McCartney G, Hart C, Watt G. How can socioeconomic inequalities in hospital admissions be explained? A cohort study. BMJ Open. 2013;3(8). 33. Acheson D. Health inequalities impact assessment. Bulletin of the World Health Organization. 2000;78:75-6. 34. Victora CG, Vaughan JP, Barros FC, Silva AC, Tomasi E. Explaining trends in inequities: evidence from Brazilian child health studies. The Lancet. 2000;356(9235):1093-8. 35. Lyratzopoulos G, Barbiere JM, Rachet B, Baum M, Thompson MR, Coleman MP. Changes over time in socioeconomic inequalities in breast and rectal cancer survival in England and Wales during a 32-year period (1973–2004): the potential role of health care. Annals of Oncology. 2011;22(7):1661-6. 36. Tugwell P, de Savigny D, Hawker G, Robinson V. Applying clinical epidemiological methods to health equity: the equity effectiveness loop. BMJ. 2006;332(7537):358-61. 37. Murphy MM, Tseng JF, Shah SA. Disparities in cancer care: An operative perspective. Surgery. 2010;147(5):733-7. 38. Andersen RM, Davidson PL. Improving access to care in America. Changing the US health care system: key issues in health services policy and management 3a edición San Francisco: Jossey-Bass. 2007:3-31. 39. Lasser KE, Himmelstein DU, Woolhandler S. Access to Care, Health Status, and Health Disparities in the United States and Canada: Results of a Cross-National Population-Based Survey. American Journal of Public Health. 2006;96(7):1300-7. 40. Dixon-Woods M, Cavers D, Agarwal S, Annandale E, Arthur A, Harvey J, et al. Conducting a critical interpretive synthesis of the literature on access to healthcare by vulnerable groups. BMC Medical Research Methodology. 2006;6(1):35-47. 41. Protheroe J, Brooks H, Chew-Graham C, Gardner C, Rogers A. ‘Permission to participate?’ A qualitative study of participation in patients from differing socio-economic backgrounds. Journal of Health Psychology. 2013;18(8):1046-55. 272 42. Matthews P, Hastings A. Middle-Class Political Activism and Middle-Class Advantage in Relation to Public Services: A Realist Synthesis of the Evidence Base. Social Policy & Administration. 2013;47(1):72-92. 43. Nutbeam D. The evolving concept of health literacy. Social Science & Medicine. 2008;67(12):2072-8. 44. Berkman ND, Sheridan SL, Donahue KE, Halpern DJ, Crotty K. Low health literacy and health outcomes: an updated systematic review. Ann Intern Med. 2011;155:97-107. 45. Bostock S, Steptoe A. Association between low functional health literacy and mortality in older adults: longitudinal cohort study. British Medical Journal. 2012;doi: 10.1136/bmj.e1602. 46. Protheroe J, Nutbeam D, Rowlands G. Health literacy: a necessity for increasing participation in health care. British Journal of General Practice. 2009;DOI: 10.3399/bjgp09X472584. 47. Say R, Murtagh M, Thomson R. Patients' preference for involvement in medical decision making: a narrative review. Patient Education and Counseling. 2006;60:102-14. 48. Pirisi A. Low health literacy prevents equal access to care. The Lancet. 2000;356(9244):1828. 49. Arblaster L, Lambert M, Entwistle V, Forster M, Fullerton D, Sheldon T, et al. A systemetic review of the effectiveness of health service interventions aimed at reducing inequalities in health. Journal of Health Services Research and Policy. 1996;1(2):93-103. 50. Gepkens A, Gunning-Schepers LJ. Interventions to reduce socioeconomic health differences: A review of the international literature. Eur J Public Health. 1996;6(3):218-26. 51. Thomas S, Fayter D, Misso K, Ogilvie D, Petticrew M, Sowden A, et al. Population tobacco control interventions and their effects on social inequalities in smoking: systematic review. Tobacco Control. 2008;17(4):230-7. 52. Lorenc T, Petticrew M, Welch V, Tugwell P. What types of interventions generate inequalities? Evidence from systematic reviews. Journal of Epidemiology and Community Health. 2012. 53. Ahmed S, Khan MM. Is demand-side financing equity enhancing? Lessons from a maternal health voucher scheme in Bangladesh. Social Science & Medicine. 2011;72(10):1704-10. 54. Jones CM, Worthington H. Fluoridation: The relationship between water fluoridation and socioeconomic deprivation on tooth decay in 5-year-old children. Br Dent J. 1999;186(8):397-400. 55. Lu T-H, Lai C-H, Chiang T-L. Reducing regional inequality in mortality from road traffic injuries through enforcement of the mandatory motorcycle helmet law in Taiwan. Injury Prevention. 2012;18(3):150-7. 56. Colgan F, Gospel A, Petrie J, Adams J, Heywood P, White M. Does rear seat belt use vary according to socioeconomic status? Journal of Epidemiology and Community Health. 2004;58(11):929-30. 57. Reading R, Colver A, Openshaw S, Jarvis S. Do interventions that improve immunisation uptake also reduce social inequalities in uptake? BMJ. 1994;308(6937):1142-4. 58. Eek F, Ostergren P-O, Diderichsen F, Rasmussen N, Andersen I, Moussa K, et al. Differences in socioeconomic and gender inequalities in tobacco smoking in Denmark and Sweden; a cross sectional comparison of the equity effect of different public health policies. BMC Public Health. 2010;10(1):9. 59. Humphreys D, Ogilvie D. Synthesising evidence for equity impacts of population-based physical activity interventions: a pilot study. International Journal of Behavioral Nutrition and Physical Activity. 2013;10(1):76. 60. Petticrew M, Tugwell P, Kristjansson E, Oliver S, Ueffing E, Welch V. Damned if you do, damned if you don't: subgroup analysis and equity. Journal of Epidemiology and Community Health. 2012;66(1):95-8. 273 61. Welch V, Petticrew M, Ueffing E, Benkhalti Jandu M, Brand K, Dhaliwal B, et al. Does Consideration and Assessment of Effects on Health Equity Affect the Conclusions of Systematic Reviews? A Methodology Study. PLoS ONE. 2012;7(3):e31360. 62. Claxton KP, Sculpher MJ. Using value of information analysis to prioritise health research. Pharmacoeconomics. 2006;24(11):1055-68. 63. Hoa L, Argyrios Z, Lipkin SM, Zell JA. Effects of socio-economic status and treatment disparities in colorectal cancer survival. Cancer Epidemiology Biomarkers & Prevention. 2008;17(8):1950-62. 64. Grose D, Devereux G, Brown L, Jones R, Sharma D, Selby C, et al. Variation in comorbidity and clinical management in patients newly diagnosed with lung cancer in four Scottish centers. Journal of Thoracic Oncology. 2011;6 (3):500-9. 65. Smith JJ, Tilney HS, Heriot AG, Darzi AW, Forbes H, Thompson MR, et al. Social deprivation and outcomes in colorectal cancer. British Journal of Surgery. 2006;93(9):1123-31. 66. Lyratzopoulos G, Abel GA, Brown CH, Rous BA, Vernon SA, Roland M, et al. Socio-demographic inequalities in stage of cancer diagnosis: evidence from patients with female breast, lung, colon, rectal, prostate, renal, bladder, melanoma, ovarian and endometrial cancer. Annals of Oncology. 2012. 67. Clegg L, Reichman M, Miller B, Hankey B, Singh G, Lin Y, et al. Impact of socioeconomic status on cancer incidence and stage at diagnosis: selected findings from the surveillance, epidemiology, and end results: National Longitudinal Mortality Study. Cancer Causes and Control. 2009;20(4):417-35. 68. Lyratzopoulos G, Abel GA, Barbiere JM, Brown CH, Rous BA, Greenberg DC. Variation in advanced stage at diagnosis of lung and female breast cancer in an English region 2006-2009. Br J Cancer. 2012;106(6):1068-75. 69. Louwman WJ, Aarts MJ, Houterman S, van Lenthe FJ, Coebergh JWW, Janssen-Heijnen MLG. A 50% higher prevalence of life-shortening chronic conditions among cancer patients with low socioeconomic status. Br J Cancer. 2010;103(11):1742-8. 70. Ng R, de Boer R, Green MD. Undertreatment of elderly patients with non-small cell lung cancer. Clinical Lung Cancer. 2005;7(3):168-74. 71. Tammemagi CM, Neslund-Dudas C, Simoff M, Kvale P. In lung cancer patients, age, race-ethnicity, gender and smoking predict adverse comorbidity, which in turn predicts treatment and survival. Journal of Clinical Epidemiology. 2004;57(6):597-609. 72. de Rijke JM, Schouten LJ, Velde GPMt, Wanders SL, Bollen ECM, Lalisang RI, et al. Influence of age, comorbidity and performance status on the choice of treatment for patients with non-small cell lung cancer; results of a population-based study. Lung Cancer. 2004;46(2):233-45. 73. Berglund A, Holmberg L, Tishelman C, Wagenius G, Eaker S, Lambe M. Social inequalities in non-small cell lung cancer management and survival: a population-based study in central Sweden. Thorax. 2010;65(4):327-33. 74. Gordon HS, Street Jr RL, Sharf BF, Souchek J. Racial differences in doctors' information-giving and patients' participation. Cancer. 2006;107 (6):1313-20. 75. National Cancer Intelligence Network. Cancer and equality groups: key metrics. London: 2013. 76. Pollock AM, Vickers N. Deprivation and emergency admissions for cancers of colorectum, lung, and breast in south east England: ecological study. BMJ. 1998;317(7153):245-52. 77. Raine R, Wong W, Scholes S, Ashton C, Obichere A, Ambler G. Social variations in access to hospital care for patients with colorectal, breast, and lung cancer between 1999 and 2006: retrospective analysis of hospital episode statistics. BMJ. 2010;340. 78. Jones AP, Haynes R, Sauerzapf V, Crawford SM, Zhao H, Forman D. Travel time to hospital and treatment for breast, colon, rectum, lung, ovary and prostate cancer. European Journal of Cancer. 2008;44(7):992-9. 274 79. Lejeune C, Sassi F, Ellis L, Godward S, Mak V, Day M, et al. Socio-economic disparities in access to treatment and their impact on colorectal cancer survival. International Journal of Epidemiology. 2010;39(3):710-7. 80. Palmer R, Schneider E. Social disparities across the continuum of colorectal cancer: a systematic review. Cancer Causes and Control. 2005;16(1):55-61. 81. Schrag D, Gelfand SE, Bach PB, Guillem J, Minsky BD, Begg CB. Who Gets Adjuvant Treatment for Stage II and III Rectal Cancer? Insight From Surveillance, Epidemiology, and End Results–Medicare. Journal of Clinical Oncology. 2001;19(17):3712-8. 82. Lemmens VEPP, van Halteren AH, Janssen-Heijnen MLG, Vreugdenhil G, Repelaer van Driel OJ, Coebergh JWW. Adjuvant treatment for elderly patients with stage III colon cancer in the southern Netherlands is affected by socioeconomic status, gender, and comorbidity. Annals of Oncology. 2005;16(5):767-72. 83. Roetzheim RG, Pal N, Gonzalez EC, Ferrante JM, Van Durme DJ, Krischer JP. Effects of Health Insurance and Race on Colorectal Cancer Treatments and Outcomes. American Journal of Public Health. 2000;90(11):1746-54. 84. VanEenwyk J, Campo JS, Ossiander EM. Socioeconomic and demographic disparities in treatment for carcinomas of the colon and rectum. Cancer. 2002;95(1):39-46. 85. McGory M, Zingmond D, Sekeris E, Bastani R, Ko C. A Patient's Race/Ethnicity Does Not Explain the Underuse of Appropriate Adjuvant Therapy in Colorectal Cancer. Diseases of the Colon & Rectum. 2006;49(3):319-29. 86. Campbell NC, Elliott AM, Sharp L, Ritchie LD, Cassidy J, Little J. Impact of deprivation and rural residence on treatment of colorectal and lung cancer. Br J Cancer. 2002;87(6):585-90. 87. Ayanian JZ, Zaslavsky AM, Fuchs CS, Guadagnoli E, Creech CM, Cress RD, et al. Use of Adjuvant Chemotherapy and Radiation Therapy for Colorectal Cancer in a Population-Based Cohort. Journal of Clinical Oncology. 2003;21(7):1293-300. 88. Vulto JCM, Louwman WJ, Lybeert MLM, Poortmans PMP, Rutten HJT, Brenninkmeijer SJ, et al. A population-based study of radiotherapy in a cohort of patients with rectal cancer diagnosed between 1996 and 2000. European Journal of Surgical Oncology (EJSO). 2007;33(8):993-7. 89. Hall SE, Holman CDAJ, Platell C, Sheiner H, Threlfall T, Semmens J. Colorectal cancer surgical care and survival: do private health insurance, socioeconomic and locational status make a difference? ANZ Journal of Surgery. 2005;75(11):929-35. 90. Olsson LI, Granström F, Påhlman L. Sphincter preservation in rectal cancer is associated with patients' socioeconomic status. British Journal of Surgery. 2010;97(10):1572-81. 91. Bharathan B, Welfare M, Borowski DW, Mills SJ, Steen IN, Kelly SB, et al. Impact of deprivation on short- and long-term outcomes after colorectal cancer surgery. British Journal of Surgery. 2011;98(6):854-65. 92. Morris E, Quirke P, Thomas JD, Fairley L, Cottier B, Forman D. Unacceptable variation in abdominoperineal excision rates for rectal cancer: time to intervene? Gut. 2008;57(12):1690-7. 93. Harris AR, Bowley DM, Stannard A, Kurrimboccus S, Geh JI, Karandikar S. Socioeconomic deprivation adversely affects survival of patients with rectal cancer. British Journal of Surgery. 2009;96(7):763-8. 94. Hodgson DC, Fuchs CS, Ayanian JZ. Impact of Patient and Provider Characteristics on the Treatment and Outcomes of Colorectal Cancer. Journal of the National Cancer Institute. 2001;93(7):501-15. 95. Sankaranarayanan J, Watanabe-Galloway S, Sun J, Qui F, Boilsen EC, Thorson AG. Age and rural residence effects on accessing colorectal cancer treatments: a registry study. American Journal of Managed Care. 2010;16(4):265-73. 96. Ngeow J, Leong SS, Gao F, Toh CK, Lim WT, Tan EH, et al. Impact of comorbidities on clinical outcomes in non-small cell lung cancer patients who are elderly and/or have poor performance status. Critical Reviews in Oncology/Hematology. 2010;76(1):53-60. 275 97. Haas JS, Brawarsky P, Iyer A, Fitzmaurice GM, Neville BA, Earle C. Association of area sociodemographic characteristics and capacity for treatment with disparities in colorectal cancer care and mortality. Cancer. 2011;117(18):4267-76. 98. Forrest LM, McMillan DC, McArdle CS, Dunlop DJ. An evaluation of the impact of a multidisciplinary team, in a single centre, on treatment and survival in patients with inoperable non-small-cell lung cancer. Br J Cancer. 2005;93(9):977-8. 99. Richards MA. The National Awareness and Early Diagnosis Initiative in England: assembling the evidence. Br J Cancer. 2009;101(S2):S1 - S4. 100. Weller D, Vedsted P, Rubin G, Walter FM, Emery J, Scott S, et al. The Aarhus statement: improving design and reporting of studies on early cancer diagnosis. British Journal of Cancer. 2012;106:1262-7. 101. Devbhandari MP, Bittar MN, Quennell P, Barber P, Krysiak P, Shah R, et al. Are We Achieving the Current Waiting Time Targets in Lung Cancer Treatment? Result of a Prospective Study from a Large United Kingdom Teaching Hospital. Journal of Thoracic Oncology. 2007;2(7):590-2 10.1097/JTO.0b013e318070ccf0. 102. National Patient Safety Agency. Delayed diagnosis of cancer: Thematic review. London: National Reporting and Learning Service; 2010. 103. Walter F, Webster A, Scott S, Emery J. The Andersen Model of Total Patient Delay: a systematic review of its application in cancer diagnosis. Journal of Health Services Research & Policy. 2012;17(2):110-8. 104. Hansen R, Olesen F, Sorensen H, Sokolowski I, Sondergaard J. Socioeconomic patient characteristics predict delay in cancer diagnosis: a Danish cohort study. BMC Health Services Research. 2008;8(1):49. 105. Macleod U, Mitchell ED, Burgess C, Macdonald S, Ramirez AJ. Risk factors for delayed presentation and referral of symptomatic cancer: evidence for common cancers. Br J Cancer. 2009;101(S2):S92-S101. 106. Neal RD, Allgar VL. Sociodemographic factors and delays in the diagnosis of six cancers: analysis of data from the 'National Survey of NHS patients: Cancer'. British Journal of Cancer. 2005;92:1971-5. 107. Yorio JT, Xie Y, Yan J, Gerber DE. Lung cancer diagnostic and treatment intervals in the United States. Journal of Thoracic Oncology. 2009;4(11):1322-30. 108. Dalton SO, Frederiksen BL, Jacobsen E, Steding-Jessen M, Osterlind K, Schuz J, et al. Socioeconomic position, stage of lung cancer and time between referral and diagnosis in Denmark, 2001-2008. Br J Cancer. 2011;105(7):1042-8. 109. Saint-Jacques N, Rayson D, Al-Fayea T, Virik K, Morzycki W, Younis T. Waiting Times in Early-Stage Non-small Cell Lung Cancer (NSCLC). Journal of Thoracic Oncology. 2008;3(8):865-70 10.1097/JTO.0b013e318180210c. 110. Korsgaard M, Pedersen L, Laurberg S. Delay of diagnosis and treatment of colorectal cancer--A population-based Danish Study. Cancer Detection and Prevention. 2008;32(1):45-51. 111. Corner J, Brindle L. The influence of social processes on the timing of cancer diagnosis: a research agenda. J Epidemiol Community Health. 2011;65:477-82. 112. Mitchell E, Macdonald S, Campbell NC, Weller D, Macleod U. Influences on pre-hospital delay in the diagnosis of colorectal cancer: a systematic review. Br J Cancer. 2008;98(1):60-70. 113. Terhaar sive Droste J, Oort F, van der Hulst R, Coupe V, Craanen M, Meijer G, et al. Does delay in diagnosing colorectal cancer in symptomatic patients affect tumor stage and survival? A population-based observational study. BMC Cancer. 2010;10(1):332. 114. Ciccone G, Prastaro C, Ivaldi C, Giacometti R, Vineis P. Access to hospital care, clinical stage and survival from colorectal cancer according to socio-economic status. Annals of Oncology. 2000;11(9):1201-4. 115. Cheyne L, Taylor A, Milton R, Fear J, Callister MEJ. Social deprivation does not affect lung cancer stage at presentation or disease outcome. Lung Cancer. 2013(0). 276 116. Tod AM, Craven J, Allmark P. Diagnostic delay in lung cancer: a qualitative study. Journal of Advanced Nursing. 2008;61(3):336-43. 117. Smith SM, Campbell NC, MacLeod U, Lee AJ, Raja A, Wyke S, et al. Factors contributing to the time taken to consult with symptoms of lung cancer: a cross-sectional study. Thorax. 2009;64(6):523-31. 118. Sulu E, Tasolar O, Takir HB, Tuncer LY, Karakurt Z, Yilmaz A. Delays in the diagnosis and treatment of non-small-cell lung cancer. Tumori. 2012;97(6):693-7. 119. National Institute for Health and Clinical Excellence. Referral guidelines for suspected cancer: Clinical Guideline 27. London: NICE; 2005. 120. Hamilton W, Sharp D. Diagnosis of colorectal cancer in primary care: the evidence base for guidelines. Family Practice. 2004;21(1):99-106. 121. Salomaa E-R, Sallinen S, Hiekkanen H, Liippo K. Delays in the Diagnosis and Treatment of Lung Cancer*. Chest. 2005;128(4):2282-8. 122. Bozcuk H, Martin C. Does treatment delay affect survival in non-small cell lung cancer? A retrospective analysis from a single UK centre. Lung Cancer. 2001;34(2):243-52. 123. Allgar VL, Neal RD. Delays in the diagnosis of six cancers: analysis of data from the National Survey of NHS Patients: Cancer. Br J Cancer. 2005;92(11):1959-70. 124. Allgar VL, Neal RD, Ali N, Leese B, Heywood P, Proctor G, et al. Urgent GP referrals for suspected lung, colorectal, prostate and ovarian cancer. British Journal of General Practice. 2006;56:355-62. 125. Devbhandari M, Soon S, Quennell P, Barber P, Krysiak P, Shah R, et al. UK waiting time targets in lung cancer treatment: are they achievable? Results of a prospective tracking study. Journal of Cardiothoracic Surgery. 2007;2(1):5. 126. Valdes S, Garcia E, Perez H, Hernandez M. Length of diagnostic delay in patients with non-small-cell lung cancer. MEDICC Review. 2010;12(1):29-32. 127. Yilmaz A, Damadoglu E, Salturk C, Okur E, Yagci Tuncer L, Halezeroglu S. Delays in the Diagnosis and Treatment of Primary Lung Cancer: Are Longer Delays Associated with Advanced Pathological Stage? Upsala Journal of Medical Sciences. 2008;113(3):287-96. 128. Neal RD, Pasterfield D, Wilkinson C, Hood K, Makin M, Lawrence H. Determining patient and primary care delay in the diagnosis of cancer - lessons from a pilot study of patients referred for suspected cancer. BMC Family Practice. 2008;9 (9). 129. Bardell T, Belliveau P, Kong W, Mackillop WJ. Waiting Times for Cancer Surgery in Ontario: 1984-2000. Clinical Oncology. 2006;18(5):401-9. 130. Langenbach M, Sauerland S, Kröbel K-W, Zirngibl H. Why so late?!—delay in treatment of colorectal cancer is socially determined. Langenbeck's Archives of Surgery. 2010;395(8):1017-24. 131. Des Guetz G, Nicolas P, Perret G-Y, Morere J-F, Uzzan B. Does delaying adjuvant chemotherapy after curative surgery for colorectal cancer impair survival? A meta-analysis. European Journal of Cancer. 2010;46(6):1049-55. 132. Dejardin O, Herbert C, Velten M, Buemi A, Menegoz F, Maarouf N, et al. Social and geographical factors influencing the delay in treatment for colorectal cancer. Br J Cancer. 2004;91(9):1751-2. 133. Chandra S, Mohan A, Guleria R, SIngh V, Yadav P. Delays during the diagnostic evaluation and treatment of lung cancer. Asian Pacific Journal of Cancer Prevention. 2009;10:453-6. 134. Myrdal G, Lambe M, Hillerdal G, Lamberg K, Agustsson T, Stahle E. Effect of delays on prognosis in patients with non-small cell lung cancer. Thorax. 2004;59(1):45-9. 135. Kogevinas M, Porta M. Socioeconomic differences in cancer survival: a review of evidence. Social inequalities and cancer IARC Scientific Publication No 138. Lyon1997. p. 177-206. 136. Rachet B, Ellis L, Maringe C, Chu T, Nur U, Quaresma M, et al. Socioeconomic inequalities in cancer survival in England after the NHS cancer plan. Br J Cancer. 2010;103(4):446-53. 277 137. Shack LG, Rachet B, Brewster DH, Coleman MP. Socioeconomic inequalities in cancer survival in Scotland 1986-2000. Br J Cancer. 2007;97(7):999-1004. 138. Munro A, Bentley A. Deprivation, comorbidity and survival in a cohort of patients with colorectal cancer. European Journal of Cancer Care. 2004;13(3):254-62. 139. Cwikel JG, Behar LC, Zabora JR. Psychosocial Factors That Affect the Survival of Adult Cancer Patients -- A Review of Research. Journal of Psychosocial Oncology. 1998;15(3):1 - 34. 140. Braaten T, Weiderpass E, Lund E. Socioeconomic differences in cancer survival: The Norwegian Women and Cancer Study. BMC Public Health. 2009;9(1):178. 141. Aravani A, Thomas J, Day M, Forman D, Morris E, Tatarek-Gintowt R. Survival by stage of colorectal cancer in England. NYCRIS poster2009. 142. Wrigley H, Roderick P, George S, Smith J, Mullee M, Goddard J. Inequalities in survival from colorectal cancer: a comparison of the impact of deprivation, treatment, and host factors on observed and cause specific survival. Journal of Epidemiology and Community Health. 2003;57(4):301-9. 143. Pollock AM, Vickers N. Breast, lung and colorectal cancer incidence and survival in South Thames Region, 1987-1992: the effect of social deprivation. Journal of Public Health. 1997;19(3):288-94. 144. Frederiksen BL, Osler M, Harling H, Ladelund S, Jørgensen T. Do patient characteristics, disease, or treatment explain social inequality in survival from colorectal cancer? Social Science & Medicine. 2009;69(7):1107-15. 145. Yu X. Socioeconomic disparities in breast cancer survival: relation to stage at diagnosis, treatment and race. BMC Cancer. 2009;9(1):364. 146. Halmin Mr, Bellocco R, Lagerlund M, Karlsson P, Tejler Gr, Lambe M. Long-term inequalities in breast cancer survival : a ten year follow-up study of patients managed within a National Health Care System (Sweden). Acta Oncologica. 2008;47(2):216-24. 147. Franzini L, Williams A, Franklin J, Singletary S, Theriault R. Effects of race and socioeconomic status on survival of 1,332 black, hispanic, and white women with breast cancer. Annals of Surgical Oncology. 1997;4(2):111-8. 148. Schrijvers CTM, Coebergh JWW, van der Heijden LH, Mackenbach JP. Socioeconomic status and breast cancer survival in the southeastern Netherlands, 1980-1989. European Journal of Cancer. 1995;31(10):1660-4. 149. Ou SHI, Zell JA, Ziogas A, Anton-Culver H. Low socioeconomic status is a poor prognostic factor for survival in stage I nonsmall cell lung cancer and is independent of surgical treatment, race, and marital status. Cancer. 2008;112(9):2011-20. 150. Yang R, Cheung MC, Byrne MM, Huang Y, Nguyen D, Lally BE, et al. Do racial or socioeconomic disparities exist in lung cancer treatment? Cancer. 2010;116(10):2437-47. 151. Grunfeld E, Watters JM, Urquhart R, O'Rourke K, Jaffey J, Maziak DE, et al. A prospective study of peri-diagnostic and surgical wait times for patients with presumptive colorectal, lung, or prostate cancer. Br J Cancer. 2008;100(1):56-62. 152. Rich AL, Tata LJ, Free CM, Stanley RA, Peake MD, Baldwin DR, et al. Inequalities in outcomes for non-small cell lung cancer: the influence of clinical characteristics and features of the local lung cancer service. Thorax. 2011;66(12):1078-84. 153. Rich AL, Tata LJ, Stanley RA, Free CM, Peake MD, Baldwin DR, et al. Lung cancer in England: Information from the National Lung Cancer Audit (LUCADA). Lung Cancer. 2011;72(1):16-22. 154. Nur U, Rachet B, Parmar MKB, Sydes MR, Cooper N, Lepage C, et al. No socioeconomic inequalities in colorectal cancer survival within a randomised clinical trial. Br J Cancer. 2008;99(11):1923-8. 155. Møller H, Sandin F, Robinson D, Bray F, Klint Å, Linklater KM, et al. Colorectal cancer survival in socioeconomic groups in England: Variation is mainly in the short term after diagnosis. European Journal of Cancer. 2012;48(1):46-53. 156. Hole DJ, McArdle CS. Impact of socioeconomic deprivation on outcome after surgery for colorectal cancer. British Journal of Surgery. 2002;89(5):586-90. 278 157. Berglund A, Lambe M, Lüchtenborg M, Linklater K, Peake MD, Holmberg L, et al. Social differences in lung cancer management and survival in South East England: a cohort study. BMJ Open. 2012;2(3):10.1136/bmjopen-2012-001048. 158. Jack RH, Gulliford MC, Ferguson J, Møller H. Explaining inequalities in access to treatment in lung cancer. Journal of Evaluation in Clinical Practice. 2006;12(5):573-82. 159. Herndon JE, Kornblith AB, Holland JC, Paskett ED. Patient Education Level As a Predictor of Survival In Lung Cancer Clinical Trials. Journal of Clinical Oncology. 2008;26(25):4116-23. 160. Jones AP, Haynes R, Sauerzapf V, Crawford SM, Zhao H, Forman D. Travel times to health care and survival from cancers in Northern England. European Journal of Cancer. 2008;44(2):269-74. 161. Riaz SP, Lüchtenborg M, Jack RH, Coupland VH, Linklater KM, Peake MD, et al. Variation in surgical resection for lung cancer in relation to survival: Population-based study in England 2004–2006. European Journal of Cancer. 2012;48(1):54-60. 162. Hall SE, Bulsara CE, Bulsara MK, Leahy TG, Culbong MR, Hendrie D, et al. Treatment patterns for cancer in Western Australia: Does being indigenous make a difference? Medical Journal of Australia. 2004;181(4):191-4. 163. Hardy D, Xia R, Liu C-C, Cormier JN, Nurgalieva Z, Du XL. Racial disparities and survival for nonsmall-cell lung cancer in a large cohort of black and white elderly patients. Cancer. 2009;115(20):4807-18. 164. Schrijvers CT, Mackenbach JP. Cancer patient survival by socioeconomic status in seven countries: a review for six common cancer sites [corrected]. Journal of Epidemiology and Community Health. 1994;48(5):441-6. 165. Yim J, Hwang S-s, Yoo K-y, Kim C-y. Contribution of income-related inequality and healthcare utilisation to survival in cancers of the lung, liver, stomach and colon. Journal of Epidemiology and Community Health. 2012;66(1):37-40. 166. Grilli R, Minozzi S, Tinazzi A, Labianca R, Sheldon TA, Liberati A. Do specialists do it better? The impact of specialization on the processes and outcomes of care for cancer patients. Annals of Oncology. 1998;9(4):365-74. 167. Taylor EF, Thomas JD, Finan PJ, Quirke P, Forman D, Coleman MP, et al. English sugical colorectal practice between 1998 and 2006. NCIN/ UKACR conference; Birmingham2010. 168. Møller H, Sandin F, Bray F, Klint Å, Linklater KM, Purushotham A, et al. Breast cancer survival in England, Norway and Sweden: a population-based comparison. International Journal of Cancer. 2010;127(11):2630-8. 169. Holmberg L, Sandin F, Bray F, Richards M, Spicer J, Lambe M, et al. National comparisons of lung cancer survival in England, Norway and Sweden 2001-2004: differences occur early in follow-up. Thorax. 2010;65(5):436-41. 170. Walters S, Maringe C, Coleman MP, Peake MD, Butler J, Young N, et al. Lung cancer survival and stage at diagnosis in Australia, Canada, Denmark, Norway, Sweden and the UK: a population-based study, 2004–2007. Thorax. 2013. 171. Imperatori A, Harrison RN, Leitch DN, Rovera F, Lepore G, Dionigi G, et al. Lung cancer in Teesside (UK) and Varese (Italy): a comparison of management and survival. Thorax. 2006;61(3):232-9. 172. Kawaguchi T, Takada M, Kubo A, Matsumura A, Fukai S, Tamura A, et al. Performance Status and Smoking Status Are Independent Favorable Prognostic Factors for Survival in Non-small Cell Lung Cancer: A Comprehensive Analysis of 26,957 Patients with NSCLC. Journal of Thoracic Oncology. 2010;5(5):620-30 10.1097/JTO.0b013e3181d2dcd9. 173. Nakamura H, Ando K, Shinmyo T, Morita K, Mochizuki A, Kurimoto N, et al. Female Gender Is an Independent Prognostic Factor in Non-small-cell Lung Cancer: A Meta-analysis. Annals of Thoracic and Cardiovascular Surgery. 2011;17(5):469-80. 174. Powell HA, Tata LJ, Baldwin DR, Stanley RA, Khakwani A, Hubbard RB. Early mortality after surgical resection for lung cancer: an analysis of the English National Lung cancer audit. Thorax. 2013. 279 175. Peake MD, Black EA. Increasing the surgical resection rate for lung cancer in the UK: the debate. Lung Cancer. 2013;2(3):205-11. 176. Riaz SP, Linklater KM, Page R, Peake MD, Møller H, Lüchtenborg M. Recent trends in resection rates among non-small cell lung cancer patients in England. Thorax. 2012. 177. Neal RD. Do diagnostic delays in cancer matter? British Journal of Cancer. 2009;101(S2):S9-S12. 178. Ramos M, Esteva M, Cabeza E, Campillo C, Llobera J, Aguiló A. Relationship of diagnostic and therapeutic delay with survival in colorectal cancer: A review. European Journal of Cancer. 2007;43(17):2467-78. 179. Torring ML, Frydenberg M, Hansen RP, Olesen F, Hamilton W, Vedsted P. Time to diagnosis and mortality in colorectal cancer: a cohort study in primary care. Br J Cancer. 2011;104(6):934-40. 180. Shin D, Cho J, Kim S, Guallar E, Hwang S, Cho B, et al. Delay to Curative Surgery Greater than 12 Weeks Is Associated with Increased Mortality in Patients with Colorectal and Breast Cancer but Not Lung or Thyroid Cancer. Annals of Surgical Oncology. 2013;20(8):2468-76. 181. Redaniel MT, Martin RM, Cawthorn S, Wade J, Jeffreys M. The association of waiting times from diagnosis to surgery with survival in women with localised breast cancer in England. Br J Cancer. 2013;109(1):42-9. 182. Ramos M, Esteva M, Cabeza E, Llobera J, Ruiz A. Lack of association between diagnostic and therapeutic delay and stage of colorectal cancer. European Journal of Cancer. 2008;44(4):510-21. 183. Richards MA, Westcombe AM, Love SB, Littlejohns P, Ramirez AJ. Influence of delay on survival in patients with breast cancer: a systematic review. The Lancet. 1999;353(9159):1119-26. 184. Jensen AR, Mainz J, Overgaard J. Impact of Delay on Diagnosis and Treatment of Primary Lung Cancer. Acta Oncologica. 2002;41(2):147-52. 185. Olsson JK, Schultz EM, Gould MK. Timeliness of care in patients with lung cancer: a systematic review. Thorax. 2009;64(9):749-56. 186. Radzikowska E, Roszkowski-Sliz K, Chabowski M, Glaz P. Influence of Delays in Diagnosis and Treatment on Survival in Small Cell Lung Cancer Patients. In: Pokorski M, editor. Neurobiology of Respiration. Advances in Experimental Medicine and Biology. 788: Springer Netherlands; 2013. p. 355-62. 187. Moher D, Liberati A, Tezlaff J, Altman D, The PRISMA Group. Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med. 2009;6(7). 188. Slatore CG, Au DH, Gould MK, on behalf of the American Thoracic Society Disparities in Healthcare G. An Official American Thoracic Society Systematic Review: Insurance Status and Disparities in Lung Cancer Practices and Outcomes. Am J Respir Crit Care Med. 2010;182(9):1195-205. 189. Donaldson C, Gerard K. Health care financing reforms: moving into the new millenium. Economics of Health Care Financing: The Visible Hand. Basingstoke: Palgrave Macmillan; 2005. 190. Deeks J, Dinnes J, D’Amico R, Sowden A, Sakarovitch C, Song F, et al. Evaluating non-randomised intervention studies. Health Technology Assessment. 2003;7(27). 191. Lundh A, Gotzsche P. Recommendations by Cochrane Review Groups for assessment of the risk of bias in studies. BMC Medical Research Methodology. 2008;8(1):22. 192. SIGN. SIGN 50: A Guideline Developer's Handbook. Methodology Checklist 3: Cohort studies. Edinburgh: SIGN; 2011. 193. Effective Public Health Practice Project (EPHPP). [01/06/2011]. Available from: http://www.ephpp.ca/aboutus.html. 194. Vandenbroucke JP, von Elm E, Altman DG, Gøtzsche PC, Mulrow CD, Pocock SJ, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): Explanation and Elaboration. PLoS Med. 2007;4(10):e297. 195. Centre for Reviews and Dissemination. Systematic Reviews: CRD's guidance for undertaking reviews in health care. 2009. 280 196. Ogilvie D, Fayter D, Petticrew M, Sowden A, Thomas S, Whitehead M, et al. The harvest plot: A method for synthesising evidence about the differential effects of interventions. BMC Medical Research Methodology. 2008;8(1):8. 197. The Cochrane Collaboration. Cochrane Handbook for Systematic Reviews of Interventions (version 5.0.2). Higgins J, Green S, editors 2008. 198. Erridge SC, Murray B, Williams L, Brewster D, Black R, Price A, et al. Improved survival from lung cancer in British Columbia compared to Scotland-Are different treatment rates the whole story? Lung Cancer. 2009;64 (3):358-66. 199. Crawford SM, Sauerzapf V, Haynes R, Zhao H, Forman D, Jones AP. Social and geographical factors affecting access to treatment of lung cancer. Br J Cancer. 2009;101(6):897-901. 200. Erridge SC, Thomson CS, Davidson J, Jones RD, Price A. Factors Influencing the Use of Thoracic Radiotherapy in Lung Cancer;an Analysis of the 1995 Scottish Lung Cancer Audit. Clinical Oncology. 2002;14(3):219-27. 201. Gregor A, Thomson CS, Brewster DH, Stroner PL, Davidson J, Fergusson RJ, et al. Management and survival of patients with lung cancer in Scotland diagnosed in 1995: results of a national population based study. Thorax. 2001;56(3):212-7. 202. Jack RH, Gulliford MC, Ferguson J, Moller H. Geographical inequalities in lung cancer management and survival in South East England: evidence of variation in access to oncology services? Br J Cancer. 2003;88(7):1025-31. 203. Mahmud SM, Reilly M, Comber H. Patterns of initial management of lung cancer in the Republic of Ireland: a population-based observational study. Lung Cancer. 2003;41(1):57-64. 204. McMahon M, Barbiere JM, Greenberg DC, Wright KA, Lyratzopoulos G. Population-based trends in use of surgery for non-small cell lung cancer in a UK region, 1995–2006. Thorax. 2011;66(5):453-5. 205. Stevens W, Stevens G, Kolbe J, Cox B. Lung cancer in New Zealand: Patterns of secondary care and implications for survival. Journal of Thoracic Oncology. 2007;2 (6):481-93. 206. Stevens W, Stevens G, Kolbe J, Cox B. Ethnic differences in the management of lung cancer in New Zealand. Journal of Thoracic Oncology. 2008;3 (3):237-44. 207. Battersby J, Flowers J, Harvey I. An alternative approach to quantifying and addressing inequity in healthcare provision: access to surgery for lung cancer in the east of England. Journal of Epidemiology and Community Health. 2004;58(7):623-5. 208. Bendzsak A, Nenshi R, Darling G, Schultz SE, Gunraj N, Wilton AS, et al. Overview of Lung Cancer Surgery in Ontario. The Annals of Thoracic Surgery. 2011;91(2):361-6. 209. Cartman ML, Hatfield AC, Muers MF, Peake MD, Haward RA, Forman D. Lung cancer: district active treatment rates affect survival. Journal of Epidemiology and Community Health. 2002;56(6):424-9. 210. Hui AC, Vinod SK, Jalaludin BB, Yuile P, Delaney GP, Barton MB. Socio-economic status and patterns of care in lung cancer. Australian and New Zealand Journal of Public Health. 2005;29(4):372-7. 211. Madelaine J, Guizard AV, Lefevre H, Lecarpentier MM, Launoy G. [Diagnosis, treatment, and prognosis of lung cancer in the Manche (France) (1997-1999) according to patients socioeconomic characteristics]. Revue d Epidemiologie et de Sante Publique. 2002;50(4):383-92. 212. Pagano E, Filippini C, Di Cuonzo D, Ruffini E, Zanetti R, Rosso S, et al. Factors affecting pattern of care and survival in a population-based cohort of non-small-cell lung cancer incident cases. Cancer Epidemiology. 2010;34(4):483-9. 213. Patel N, Adatia R, Mellemgaard A, Jack R, Moller H. Variation in the use of chemotherapy in lung cancer. Br J Cancer. 2007;96(6):886-90. 214. Stevens G, Stevens W, Purchuri S, Kolbe J, Cox B. Radiotherapy utilisation in lung cancer in New Zealand: Disparities with optimal rates explained. New Zealand Medical Journal. 2009;122 (1306):43-54. 281 215. Younis T, Al-Fayea T, Virik K, Morzycki W, Saint-Jacques N. Adjuvant chemotherapy uptake in non-small cell lung cancer. Journal of Thoracic Oncology: Official Publication of the International Association for the Study of Lung Cancer. 2008;3(11):1272-8. 216. Bradley CJ, Dahman B, Given CW. Treatment and Survival Differences in Older Medicare Patients With Lung Cancer as Compared With Those Who Are Dually Eligible for Medicare and Medicaid. Journal of Clinical Oncology. 2008;26(31):5067-73. 217. Davidoff AJ, Tang M, Seal B, Edelman MJ. Chemotherapy and Survival Benefit in Elderly Patients With Advanced Non–Small-Cell Lung Cancer. Journal of Clinical Oncology. 2010;28(13):2191-7. 218. Earle CC, Venditti LN, Neumann PJ, Gelber RD, Weinstein MC, Potosky AL, et al. Who Gets Chemotherapy for Metastatic Lung Cancer?*. Chest. 2000;117(5):1239-46. 219. Esnaola NF, Gebregziabher M, Knott K, Finney C, Silvestri GA, Reed CE, et al. Underuse of Surgical Resection for Localized, Non-Small Cell Lung Cancer Among Whites and African Americans in South Carolina. The Annals of Thoracic Surgery. 2008;86(1):220-7. 220. Greenwald HP, Polissar NL, Borgatta EF, McCorkle R, Goodman G. Social Factors, Treatment, and Survival in Early-Stage Non-Small Cell Lung Cancer. American Journal of Public Health. 1998;88(11):1681-4. 221. Hardy D, Liu C-C, Xia R, Cormier JN, Chan W, White A, et al. Racial disparities and treatment trends in a large cohort of elderly black and white patients with nonsmall cell lung cancer. Cancer. 2009;115(10):2199-211. 222. Hayman JA, Abrahamse PH, Lakhani I, Earle CC, Katz SJ. Use of Palliative Radiotherapy Among Patients With Metastatic Non-Small-Cell Lung Cancer. International Journal of Radiation Oncology*Biology*Physics. 2007;69(4):1001-7. 223. Lathan CS, Neville BA, Earle CC. Racial composition of hospitals: Effects on surgery for early-stage non-small-cell lung cancer. Journal of Clinical Oncology. 2008;26 (26):4347-52. 224. Polednak AP. Disaparities in surgical treatment of early-stage non-small-cell lung cancer. Yale Journal of Biology. 2001;74:309-14. 225. Smith TJ, Penberthy L, Desch CE, Whittemore M, Newschaffer C, Hillner BE, et al. Differences in initial treatment patterns and outcomes of lung cancer in the elderly. Lung Cancer. 1995;13(3):235-52. 226. Bach PB, Cramer LD, Warren JL, Begg CB. Racial differences in the treatment of early-stage lung cancer. New England Journal of Medicine. 1999;341(16):1198-205. 227. Earle CC, Neumann PJ, Gelber RD, Weinstein MC, Weeks JC. Impact of Referral Patterns on the Use of Chemotherapy for Lung Cancer. Journal of Clinical Oncology. 2002;20(7):1786-92. 228. Wang J, Kuo Y, Freeman J, Goodwin J. Increasing access to medical oncology consultation in older patients with stage II-IIIA non-small-cell lung cancer. Medical Oncology. 2008;25(2):125-32. 229. Suga JM, Nguyen DV, Mohammed SM, Brown M, Calhoun R, Yoneda K, et al. Racial disparities on the use of invasive and noninvasive staging in patients with non-small cell lung cancer. Journal of Thoracic Oncology. 2010;5 (11):1772-8. 230. Lathan CS, Neville BA, Earle CC. The effect of race on invasive staging and surgery in non-small-cell lung cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2006;24 (3):413-8. 231. Patel N, Ing L, Jack R, Moller H. Factors influencing the use of antitumoral chemotherapy in the South East of England. Journal of Chemotherapy. 2006;18 (3):318-24. 232. Egger M, Schneider M, Smith GD. Meta-analysis Spurious precision? Meta-analysis of observational studies. BMJ. 1998;316(7125):140-4. 233. Ioannidis J, Patsopoulos N, Evangelou E. Uncertainty in heterogeneity estimates in meta-analyses. BMJ. 2007;335:914-6. 234. Adams J, Ryan V, White M. How accurate are Townsend Deprivation Scores as predictors of self-reported health? A comparison with individual level data. Journal of Public Health. 2005;27(1):101-6. 282 235. McCann J, Artinian V, Duhaime L, Lewis JW, Kvale PA, DiGiovine B. Evaluation of the Causes for Racial Disparity in Surgical Treatment of Early Stage Lung Cancer*. Chest. 2005;128(5):3440-6. 236. Forrest LF, Adams J, Wareham H, Rubin G, White M. Socioeconomic Inequalities in Lung Cancer Treatment: Systematic Review and Meta-Analysis. PLoS Med. 2013;10(2):e1001376. 237. NYCRIS. Northern and Yorkshire Cancer Registry and Information Centre, 2012 [cited 2013 02/08/2013]. Available from: http://www.nycris.nhs.uk/about/. 238. Ludbrook JJS, Truong PT, MacNeil MV, Lesperance M, Webber A, Joe H, et al. Do age and comorbidity impact treatment allocation and outcomes in limited stage small-cell lung cancer? a community-based population analysis. International Journal of Radiation Oncology*Biology*Physics. 2003;55(5):1321-30. 239. National Cancer Intelligence Network. Access to the National Cancer Data Repository. 2010 21/02/2011. Report No. 240. Threlfall T, Wittorff J, Boutdara P, Heyworth J, Katris P, Sheiner H, et al. Collection of population-based cancer staging information in Western Australia - a feasibilty study. Popul Health Metr. 2005;3(9):10.1 186/I478-7954-3-9. 241. White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and guidance for practice. Statistics in Medicine. 2011;30(4):377-99. 242. Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. British Medical Journal. 2009;338:b2393. 243. HM Government. English Indices of Deprivation 2010 2013. 244. NYCRIS. National deprivation scores - notes from the teleconference. 2011. 245. Charlson M, Pompei P, Ales K, MacKenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of Chronic Disease. 1987;40(373-383). 246. NHS Information Centre. National Lung Cancer Audit Report 2012. Leeds: NHS Information Centre for Health and Social Care,; 2012. 247. Adams J, White M, Forman D. Are there socioeconomic gradients in the quality of data held by UK cancer registries? J Epidemiol Community Health. 2004;58:1052-3. 248. Bennett V, Davies E, Jack R, Mak V, Moller H. Histological subtype of lung cancer in relation to socio-economic deprivation in South East England. BMC Cancer. 2008;8(1):139. 249. Lüchtenborg M, Jakobsen E, Krasnik M, Linklater KM, Mellemgaard A, Møller H. The effect of comorbidity on stage-specific survival in resected non-small cell lung cancer patients. European Journal of Cancer. 2012;48(18):3386-95. 250. Tammemagi CM, Neslund-Dudas C, Simoff M, Kvale P. Smoking and lung cancer survival*: The role of comorbidity and treatment. CHEST Journal. 2004;125(1):27-37. 251. Peake MD, Thompson S, Lowe D, Pearson MG, on behalf of the Participating C. Ageism in the management of lung cancer. Age and Ageing. 2003;32(2):171-7. 252. Fergusson RJ, Gregor A, Dodds R, Kerr G. Management of lung cancer in South East Scotland. Thorax. 1996;51(6):569-74. 253. Riaz SP, Linklater K, Horton M, Peake M, Moller H, Luchtenborg M. Lung cancer data in the National Cancer Data Repository, Hospital Episode Statistics and National Lung Cancer Audit datasets. National Cancer Intelligence Network, London KsC; 2010. 254. Parsons A, Daley A, Begh R, Aveyard P. Influence of smoking cessation after diagnosis of early stage lung cancer on prognosis: systematic review of observational studies with meta-analysis. British Medical Journal. 2010;340:b:5569. 255. Elliss-Brookes L, McPhail S, Ives A, Greenslade M, Shelton J, Hiom S, et al. Routes to diagnosis for cancer - determining the patient journey using multiple routine data sets. Br J Cancer. 2012;107(8):1220-6. 283 256. Bjerager M. Delay in diagnosis and treatment of lung cancer [thesis]. 1 ed: Aarhus: Research Unit and Department of General Practice, Faculty of Health Sciences, University of Aarhus; 2006. 257. Buccheri G, Ferrigno D, Tamburini M. Karnofsky and ECOG performance status scoring in lung cancer: A prospective, longitudinal study of 536 patients from a single institution. European Journal of Cancer. 1996;32(7):1135-41. 258. Sorensen J, Klee M, Palshof T, Hansen H. Performance status in cancer patients. An inter-observer variability study. British Journal of Cancer. 1993;67(773-775). 259. Starr LK, Osler M, Steding-Jessen M, Frederiksen BL, Jakobsen E, Østerlind K, et al. Socioeconomic position and surgery for early-stage non-small-cell lung cancer: A population-based study in Denmark. Lung Cancer. 2013;79(3):262-9. 260. Her Majesty's Government. English National Cancer Online Registration Environment 2013 [cited 2013]. Available from: http://data.gov.uk/dataset/english-national-cancer-online-registration-environment-encore. 261. Forrest LF. Why are socioeconomic inequalities in receipt of treatment found for lung cancer? Lung Cancer Management. 2013;2(3):177-80. 262. Forrest LF, Adams JM, Wareham H, Rubin G, White M. PL03 Socio-Economic Inequalities in Lung Cancer Treatment: A Systematic Review and Meta-Analysis. Journal of Epidemiology and Community Health. 2012;66(Suppl 1):A38-A9. 263. Forrest LF, Adams JM, White M. OP50 Socio-Economic Inequalities in Lung Cancer Treatment: The Role of Histological Subtype and Performance Status. Journal of Epidemiology and Community Health. 2013;67(Suppl 1):A25. 264. Forrest LF, White M, Rubin G, Adams J. The impact of socio-economic inequalities in receipt of, and time to, treatment on socio-economic inequalities in lung cancer survival: an observational, data-linkage study. The Lancet. 2013(In press). 265. Forrest LF. Fuse Open Science blog [Internet]. Adams J, editor. Newcastle upon Tyne2013. Available from: http://fuseopenscienceblog.blogspot.co.uk/2013/02/the-phd-journey.html. 266. Forrest LF. Fuse Open Science blog [Internet]. Adams J, editor. Newcastle upon Tyne2012. Available from: http://fuseopenscienceblog.blogspot.co.uk/2012/05/routine-secondary-data.html. 267. Forrest LF. Fuse Open Science blog [Internet]. Adams J, editor. Newcastle upon Tyne2012. Available from: http://fuseopenscienceblog.blogspot.co.uk/2012/03/mid-life-crisis-or-late-developer.html.

PROTOCOL Open Access

Abstract

Extracted data

PROTOCOL Open Access

Abstract

Extracted data

Related items

Related items