Authors' reply

The nested case–control approach that we used is the most effective design to study rare outcomes such as major congenital malformations. ^1,2 This is even truer since it was performed in a well-established cohort of women with pre-pregnancy diagnosed psychiatric disorders. We disagree with Rajkumar & Jacob that a cohort approach would have been better, based on the fact that it lacks power for research in perinatal pharmacoepidemiology. This was clearly apparent when several small human cohort studies published in the 1990s did not suggest an overall increased risk for birth defects with first-trimester exposure to any selective serotonin reuptake inhibitors but later studies with more efficient designs such as the case–control approach started showing low-to-moderate increased risks for the more commonly occurring birth defects such as heart defects, neural tube defects and oral clefts. Therefore, using a cohort approach would have resulted again in a null finding, contrary to Rajkumar & Jacob's comments.

We excluded pregnancies ending with abortion or miscarriage per design since malformation outcomes of these foetuses were not available in the Quebec Pregnancy Registry. We agree that this resulted in prevalent cases of malformations in our study but this is highly comparable to studies performed in similar populations. We do not, however, agree that this methodological choice resulted in biasing our study estimates towards the null. Indeed, although Hemels et al ³ reported an association between antidepressant use during pregnancy and risk of spontaneous abortion, this was based on women's self-report and likely resulted in an overestimation of the rate of miscarriage and an underestimation of the rate of abortion, hence a significant association.

Major congenital malformations are structural abnormalities that affect the way a person looks and require medical and/or surgical treatment. Minor defects are abnormalities that do not cause serious health or social problems. Major defects were the focus of interest in our study and, although the risk of minor malformations is interesting, it is a different research question. Several other authors have previously made this distinction. ^4,5

We agree that results from observational studies always need to be interpreted with caution. However, given that from an ethical point of view it is almost impossible to randomise pregnant women to receive medications not known to be safe for the foetus, the collection and follow-up of observational data is the only ethical way to close the knowledge gap between the limited value of animal studies and human pregnancy exposures.

Finally, our study was not designed to look at the effect of the duration of specific antidepressants on the risk of specific major congenital malformations. Therefore, we only looked at duration of antidepressant use during the first trimester of gestation and its risk for major congenital malformations, all types and all malformations combined. Results should be interpreted in this context.