Using National Routine Data to Explore the Utilisation and Outcomes of Multimodal Treatment in the Management of Colorectal Cancer

Background: The multimodal treatment of colorectal cancer (CRC) is becoming progressively more complex with multidisciplinary input required to make appropriate decisions about patient suitability for neo-adjuvant and adjuvant therapies, local excision, watch-and-wait strategies, and choice of surgical procedure, whilst simultaneously taking into account an increasingly old and comorbid population. In addition, there are rapidly evolving advancements, for example, the approval and use of novel systemic anti-cancer therapy (SACT) treatments and surgical innovations such as the increasing uptake of robotic surgery. Large national routinely collected datasets such as the National Bowel Cancer Audit (NBOCA) and SACT dataset are becoming increasingly important for facilitating a better understanding of these complex multimodal care pathways. They allow the identification of unwarranted variation and can help to better understand disparitiesin care and outcomes. The continuous collection of this data facilitates ongoing public reporting and benchmarking processes via performance indicators and can be used to stimulate national quality improvement processes. The work within this thesis aims to utilise national routinely collected data with the broad intention of translating findings into clinical practice in two important areas within the multimodal management of CRC: (i) the use and outcomes of SACT, and (ii) the volume-outcome relationship for rectal cancer surgery. It is presented in the format of six observational studies, including two methodological development studies and four clinical research studies. Methods: In this research, detailed CRC patient care pathways are constructed using NBOCA data linked to Hospital Episode Statistics Admitted Patient Care (HES-APC) (hospital administrative data), Office for National Statistics mortality data, the SACT dataset (chemotherapy data), the National Radiotherapy Dataset (radiotherapy data), and General Medical Council (GMC) surgeon-level data. The unique linkage of these datasets provides a wealth of information, but also requires careful interpretation and validation. This thesis involves two essential components of underpinning methodological work. First, the validation of critical information is undertaken to ensure that the routinely collected data is robust. This includes the validation of routinely collected chemotherapy information using SACT and HES-APC data, and the validation of surgeon-level information for the rectal cancer volume-outcome work using NBOCA, HES-APC, and GMC data. Second, performance indicators are identified and developed to help evaluate the quality of the national delivery of CRC care. For the SACT work, these performance indicators are derived through the development and validation of a series of coding frameworks and clinical algorithms using SACT and HES-APC data. The first enables the identification of adjuvant chemotherapy use (process measure), and the second identifies severe acute toxicity from SACT (outcome measure). For the rectal cancer volume-outcome work, hospital-level rectal cancer surgery volume is established as a performance indicator (process measure). In addition, a panel of relevant performance indicators (outcome measures) are selected and adapted to evaluate the volume outcome relationship. This methodological work is then used to address pertinent clinical research gaps in the use and outcomes of multimodal treatment for CRC patients. For the SACT work, this involves three areas: (i) the exploration of determinants of variation in the use of adjuvant chemotherapy for stage III colon cancer, (ii) the impact of completion of oxaliplatin-based adjuvant chemotherapy and treatment modifications on survival for stage III colon cancer, and (iii) the evaluation of the severe acute toxicity coding framework as a performance indicator for examining between-hospital variation in toxicity rates. The volume-outcome relationship for rectal cancer surgery is explored at hospital- and surgeon-level with volume modelled as a continuous variable in relation to the selected panel of performance indicators. Results: The methodological work demonstrates that chemotherapy information is accurately captured in national routinely collected data, including receipt, regimen, and cycle number. The findings indicate that both SACT and HES-APC should be used in conjunction, where possible, to give the most robust information. A broad and comprehensive coding framework using diagnostic codes in HES-APC is used to identify severe acute toxicity (requiring overnight hospitalisation, mapped across organ systems. The coding framework demonstrates validity by identifying differential rates of toxicity according to clinical group (no chemotherapy, adjuvant cohort, and metastatic cohort), and regimen (toxicity profiles in keeping with those expected from clinical trials). In addition, severe acute toxicity is associated with expected patient and clinical factors. For the SACT work, the first clinical research study demonstrates unwarranted variation in the use of adjuvant chemotherapy in stage III colon cancer, particularly in elderly patients (>70 years). The second clinical research study shows that, for patients having oxaliplatin-based adjuvant chemotherapy, those who complete all of their treatment have significantly better survival outcomes compared to those who complete less than 50% of their treatment. However, only half of patients actually complete their treatment. Amongst patients that complete all of their chemotherapy, there are no survival differences if they have treatment modifications. The third clinical research study shows unwarranted variation in the rates of severe acute toxicity between individual hospitals for both adjuvant and metastatic CRC patients. For the rectal cancer volume-outcome work, the clinical research study demonstrates that 45% of surgeons are not meeting minimum annual rectal cancer surgery volumes as per national recommendations. Adjusting for patient and tumour characteristics, length of stay is significantly lower for high volume surgeons. No other volume-outcome relationships are demonstrated at hospital- or surgeon-level. Conclusions: This research demonstrates that multiple national routinely collected datasets can be effectively combined and subjected to novel analysis with clinically important findings. It shows that the SACT dataset provides a unique, rich, and accurate source of data, with a huge scope for addressing clinical research gaps. In addition, it demonstrates that chemotherapy information can be derived from hospital administrative data to supplement this data, or else provide chemotherapy information when bespoke SACT datasets do not exist. This work provides a rationale and basis to adapt the novel methodology across different tumour types. This work also demonstrates the translation of findings from routinely collected data into clinical practice through the development of performance indicators which facilitate the ongoing reporting and monitoring of important aspects of the multimodal treatment of CRC patients. These will be used to identify and better understand unwarranted variation as already demonstrated in this work with rates of adjuvant chemotherapy use, severe acute toxicity, and rectal cancer surgery volumes. In addition, they will trigger ongoing targeted quality improvement initiatives in order to improve the quality of CRC care on a national scale. With progressively more complex multidisciplinary decisions and management in CRC care, and an increasingly old and comorbid population, routinely collected data is paramount for the exploration of use and outcomes in “real-world” clinical practice, and to complement trial findings. Finally, it is essential for the continued development of performance indicators for timely and ongoing monitoring across the whole CRC pathway and to inform policy-makers and commissioners with regards to areas such as the specialisation of CRC services