In Reply We appreciate this opportunity to reiterate some methodological principles of the nested case-control design (see reviews1,2). It is useful to emphasize that our study3 (and the nested case-control design in general) was a cohort study with a case-control approach to the analysis. This design is widely used in pharmacoepidemiologic studies, particularly when dealing with very large cohorts with time-dependent exposures and comorbidities. Our strategy aimed to control for several potential biases including confounding by age and other variables. We controlled for age by matching cases and controls. In a nested case-control analysis, such matching is equivalent to comparing cohorts of selective serotonin reputake inhibitor and tricyclic antidepressant users of the same age. Although the age of participants on cohort entry may vary between classes of antidepressants, the base cohort is only used to identify cases and their matched controls. Adjustment for age is performed later, at the level of the selection of the matched cases and controls. The mean age of 66 years in Table 1 in our article3 corresponds to the age at the time of occurrence of an intracranial hemorrhage for cases and their matched controls. This is not the mean age on cohort entry (the time of receiving the first prescription of an antidepressant) or the mean age of the cohort. The matching based on age is therefore not done independently for each exposure group. In fact, each case is matched to controls on several variables (in our study these included age, sex, calendar time, and duration of follow-up) without a knowledge of exposure status. Consequently, the mean age of all cases and controls, regardless of their antidepressant exposure status at this index date, is the same, and each case and its matched controls have—by design—the same age regardless of their exposure status. The ratio of exposure to nonexposure (in our study this was the use of selective serotonin reputake inhibitors vs tricyclic antidepressants) in cases and controls is then compared. This ratio provides an estimate of the increase (or decrease) in risk among the participants who were exposed. Several limitations must be considered in observational studies of unwanted adverse effects of drugs. We believe, however, that the comparison group in our study was appropriate and, in particular, that it adequately controlled for age by virtue of the matching process.