Several limitations of our study merit emphasis. First, we used administrative data rather than clinical data to establish our case definition. The use of clinical data would likely have resulted in a more sensitive case definition than the one we used. However, our definition is likely to have a very high positive predictive value because independent clinical review was required for drug reimbursement. Because pulmonary arterial hypertension is rare, very few of the control subjects in our study are likely to have had pulmonary arterial hypertension. Therefore, a specific but insensitive case definition is unlikely to result in significant bias. Second, we only considered medication exposure in the 180 days preceding the index date. It is possible that more remote SSRI use, perhaps over a longer period, might have reduced the likelihood of pulmonary arterial hypertension. However, because of the nature of the administrative data available to us, a longer exposure window would have reduced our ability to demonstrate a potential effect. Furthermore, the hemodynamic effects of serotonin in animal models and in vitro studies are not delayed. Similarly, the deleterious effects of fenfluramine were observed within a time frame similar to the one we used in our study.4 Nevertheless, we are unable to exclude the possibility that SSRIs exert a protective effect over a longer time period. Similarly, because the cases had New York Heart Association class III or IV symptoms, we are also unable to exclude the possibility that SSRIs would prevent the development of asymptomatic or mildly symptomatic pulmonary arterial hypertension. Third, because public drug coverage in Ontario covers only some individuals under the age of 65 years, the age distribution of our study population is skewed in comparison with the age distribution of the pulmonary arterial hypertension population and also in comparison with the age distribution of the population studied by Shah et al.9 However, as noted previously, our outcome definition is highly specific, and it is unlikely that any of the older case patients in our study did not have pulmonary arterial hypertension. Nevertheless, pulmonary arterial hypertension remains a heterogeneous group of conditions, and it is possible that SSRIs might have a preventive or therapeutic effect in younger patients or in another subgroup. Finally, we could not adjust for unmeasured confounders, including depression and anxiety, as well as risk factors for pulmonary arterial hypertension that are not yet elucidated or were unmeasured in our study (eg, genetic polymorphisms). We believe that confounding is likely to be the reason for the positive association between SSRI exposure and pulmonary arterial hypertension in our study.