Background: Systematic reviews examining associations of depressive disorder with coronary heart disease and stroke produce mixed results. Failure to consider reverse causation and dose-response patterns may have caused inconsistencies in evidence. Design: This prospective cohort study on depressive disorder, coronary heart disease, and stroke analysed reverse causation and dose-response effects using four 5-year and three 10-year observation cycles (total follow up 24 years) based on multiple repeat measures of exposure. Methods: Participants in the Whitehall II study (n=10,036, 31,395 person-observations, age at start 44.4 years) provided up to six repeat measures of depressive symptoms via the 30-item General Health Questionnaire (GHQ- 30) and one measure via Center for Epidemiologic Studies Depression Scale (CES-D). The cohort was followed up for major coronary events (coronary death/nonfatal myocardial infarction) and stroke (stroke death/morbidity) through the national mortality register Hospital Episode Statistics, ECG-screening, medical records, and self-report questionnaires. Results: GHQ-30 caseness predicted stroke over 0-5 years (age-, sex- and ethnicity-adjusted HR 1.60, 95% CI 1.1-2.3) but not over 5-10 years (HR 0.94, 95% CI 0.6-1.4). Using the last 5-year observation cycle, cumulative GHQ-30 caseness was associated with incident coronary heart disease in a dose-response manner (1-2 times a case: HR 1.12, 95% CI 0.7-1.7; 3-4 times: HR 2.06, 95% CI 1.2-3.7), and CES-D caseness predicted coronary heart disease (HR 1.81, 95% CI 1.1-3.1). Conclusions: There was evidence of a dose-response effect of depressive symptoms on risk of coronary heart disease. In contrast, prospective associations of depressive symptoms with stroke appeared to arise wholly or partly through reverse causation. © 2013 The European Society of Cardiology.