Objective Because it is not known whether particular clusters of depressive symptoms are associated with a greater risk of adverse cardiac outcomes, we compared the utility of four clusters in predicting incident coronary artery disease (CAD) events during a 15-year period in a large cohort of primary care patients 60 years and older.

Methods Participants were 2537 primary care patients 60 years or older who were screened for depression between 1991 and 1993 and had no existing CAD diagnosis. Depressive symptoms cluster scores (depressed affect, somatic symptoms, interpersonal distress, and positive affect) were computed from responses on the Center for Epidemiologic Studies Depression Scale administered at baseline. CAD events, defined as the occurrence of a nonfatal acute myocardial infarction or CAD death during the follow-up period, were identified using electronic medical record and National Death Index data.

Results There were 678 CAD events. In separate fully adjusted Cox proportional hazard models (controlling for demographics and cardiovascular risk factors), the depressed affect (hazard ratio [HR] = 1.11, 95% confidence interval [CI] = 1.04–1.20), somatic (HR = 1.17, 95% CI = 1.08–1.26), and positive affect (HR = 0.88, 95% CI = 0.82–0.95) clusters each predicted CAD events. When the depressive symptom clusters were entered simultaneously into the fully adjusted model, however, only the somatic cluster remained predictive of CAD events (HR = 1.13, 95% CI = 1.03–1.23).

Conclusions Our findings suggest that the longitudinal relationship between overall depressive symptom severity and incident CAD events may be driven primarily by the somatic cluster.