We thank Vinkers and colleagues for their comments about our article describing the association between depression and subsequent cardiovascular events.1 Vinkers et al raise the interesting question of whether the increased risk of cardiovascular mortality associated with depression may act through noncompliance with treatment on the part of those depressed rather than through a biological mechanism. We have no direct evidence on this issue because we have no measures of compliance. We know that the women participating in the study were not depressed to an extent that would impair their adherence to protocol, since that was an exclusion criterion, and we know that they complied with completing the annual forms and providing outcomes information about any hospitalizations. We also show that women with risk factors for cardiovascular disease, such as smoking, lack of physical activity, and being overweight, as well as women with hypertension, high cholesterol level, or diabetes, are more likely to be depressed at baseline. Therefore, we controlled for these factors. Because this was an observational study, the appropriate analysis of relative risks was through proportional hazards models, which control for multiple covariates that may be related both to being depressed and to the outcomes of interest, as well as for different follow-up times, rather than the crude relative risks presented in table by Vinkers and colleagues. Adjustment for confounding variables did not change the relationship between depression and cardiovascular mortality or all-cause mortality. We found no increase in cancer diagnosis or death associated with depression. Presumably, if noncompliance were the sole explanation, one might expect that there would be some relationship to cancer as well. It is most likely that biological mechanisms as well as compliance to treatment both are implicated in the mechanism of depression-associated cardiovascular mortality. It is a subject that deserves further study.