Association between individual-level socioeconomic position and incident dementia using UK Biobank data: a prospective study

Background Under-education and living in poverty are known risk factors for dementia. However, single-variable makers of socioeconomic position (SEP) are often correlated and cannot reflect the overall SEP. We examined association between composite SEP and incident dementia using UK-wide data. Methods We leveraged data from the UK Biobank, a nation-wide cohort of half-a-million participants recruited across 22 assessment centres during 2006–10. Participants with data on SEP and without dementia at baseline were included. A composite individual-level metric of SEP (low, medium, or high) was developed through latent class analysis using Stata's gsem command and identified through item-response probabilities based on participants' single socioeconomic factors (ie, education, employment, and household income). Cox proportional hazard models were developed to examine the association between composite SEP and incident dementia after adjusting for age, ethnicity, lifestyle factors (eg, living alone), social interaction (eg, frequency of visits from acquaintances), urbanicity, clinical variables (eg, central obesity), and stratified by sex. As sensitivity test, we repeated our analysis with single socioeconomic factors. Electronic informed consent was obtained from participants at the UK Biobank assessment centres before participations and UK Biobank acquired ethical approval from the National Health Service National Research Ethics Service. Findings We included 340 366 adult participants (17 8195 [52·4%] were women and 32 6753 [96·0%] White) aged 38–73 with 3541 incident dementia cases over a mean follow-up period of 12·0 years (SD 1·7). Relative to participants in the highest SEP, those in the medium (HR 1·53 [95% CI 1·33–1·77]; p<0·0001) and low (2·38 [2·05–2·77]; p<0·0001) SEP were associated with higher risks of incident dementia. Sensitivity analyses consistently found higher risks of incident dementia in participants of low educational attainment (HR 1·14 [95% CI 1·03–1·27]; p=0·0107), low household income (HR 2·33 [95% CI 2·03–2·68]; p