Associations between depression and cardiovascular disease (CVD) have been reported in a number of studies. However, in many of these, the results have not been adjusted for potentially important confounding factors. As a result, some researchers argue that the observed association is primarily due to residual confounding. With the large amount of data collected, the UK Biobank offers a unique opportunity to assess to what extent the association between depression and CVD may be due to residual confounding. We investigated whether different measures of depression remain independent risk factors for major cardiovascular events (MCVE) after controlling for a number of potential confounding factors.

We used data from 275,759 UK Biobank participants without a history of cardiovascular disease and major mental disorders other than depression who had complete data available. In primary analyses, depression was defined as one or more of self-reported depression, antidepressant use, or hospital admission for depression. In secondary analyses, the effects of each of the subcategories of depression were analyzed separately. MCVE were defined as first-ever fatal or non-fatal stroke or myocardial infarction ascertained from hospital admission and cause of death records. We performed Cox Proportional Hazards models to estimate the risk of MCVE among participants with depression compared to those without depression. We obtained hazard ratios and 95% confidence intervals for unadjusted models, age, sex, ethnicity, education, income and area-based deprivation adjusted models, and fully adjusted models. Fully adjusted models additionally included measures of body mass index, alcohol intake, physical activity, smoking, homeownership, fruit and vegetable consumption, oily fish intake, and family history of stroke, heart disease, hypertension and/or depression.

We identified a total of 21,842 (7.9%) participants with depression at baseline of whom 64.1% had a self-reported depression diagnosis, 7.4% had a history of hospital admission for depression, and 76.6% reported using antidepressants. The proportion of people who were previous or current smokers, obese, physically inactive, or reported having diabetes or high cholesterol levels was higher among participants with depression than participants without depression. During a median of 6.7 years of follow-up, an incident MCVE occurred among 326 participants with depression and 3,718 participants without depression. In fully adjusted models, hazard ratios (95% confidence intervals) for MCVE were 1.13 (1.01–1.27) for participants with any indication of depression, 1.14 (1.00–1.31) for participants with self-reported depression, 1.60 (1.18–2.17) for participants with history of a hospital admission with depression, and 1.12 (0.99–1.27) for participants reporting antidepressant use. In all analyses, hazard ratios increased after adjustment for age, sex, ethnicity, and socioeconomic factors but were attenuated after full adjustment. Similar patterns were observed when stroke and myocardial infarction were used separately as outcomes.

All measures of depression remained independent risk factors for MCVE after adjustment for a variety of potential confounding factors and effect estimates were similar for all sub-categories of depression. The adjusted hazard ratios should only be interpreted causally if one assumes that the covariates are common sources of depression and MCVE. This assumption remains controversial. Some argue that lifestyle factors such as body mass index, smoking, and physical activity are mediators in the relationship between depression and cardiovascular events. Future studies should apply more advanced statistical methods in order to determine the effect of lifestyle factors as potential mediators and explore potential for interactions.