assessment of mediation by behavioral risk factors using the additive hazards model
Educational-related gradients in coronary heart disease (CHD) and mediation by behavioral risk factors are plausible given previous research; however this has not been comprehensively addressed in absolute measures. Questionnaire data on health behavior of 69,513 participants, 52 % women, from seven Danish cohort studies were linked to registry data on education and incidence of CHD. Mediation by smoking, low physical activity, and body mass index (BMI) on the association between education and CHD were estimated by applying newly proposed methods for mediation based on the additive hazards model, and compared with results from the Cox proportional hazards model. Short (vs. long) education was associated with 277 (95 % CI: 219, 336) additional cases of CHD per 100,000 person-years at risk among women, and 461 (95 % CI: 368, 555) additional cases among men. Of these additional cases 17 (95 % CI: 12, 22) for women and 37 (95 % CI: 28, 46) for men could be ascribed to the pathway through smoking. Further, 39 (95 % CI: 30, 49) cases for women and 94 (95 % CI: 79, 110) cases for men could be ascribed to the pathway through BMI. The effects of low physical activity were negligible. Using contemporary methods, the additive hazards model, for mediation we indicated the absolute numbers of CHD cases prevented when modifying smoking and BMI. This study confirms previous claims based on the Cox proportional hazards model that behavioral risk factors partially mediates the effect of education on CHD, and the results seems not to be particularly model dependent.
European Journal of Epidemiology, 2013, Vol 28, Issue 2, p. 149-57