Interpretation
In line with prior studies showing HR of magnitude 0.75-0.9836,37 with 1.20 as upper bound of the pooled 95% CI,38 we found no substantial difference in ADHD risk between prenatally SSRI/SNRI exposed children and children born to non-medicated women. Albeit with some uncertainty, the averaged hazard for ADHD with SSRI/SNRI exposure at any time during pregnancy was elevated (53% increased risk) when comparing to discontinuers, and likewise following 9-20 weeks exposure duration (2.1-fold increase). These contrasting results across comparisons may be in part due to lower impulsive traits and thereby lower ADHD risk in discontinuers than in the other groups; however they do also suggest that antenatal depression/anxiety is possibly a key confounder. While confounding by indication was limited in the comparison with non-medicated, by restriction, discontinuers had no active psychiatric illness in pregnancy.16 This risk of confounding did not emerge in the timing analysis, possibly because we fit methods able to account for time-varying depression symptom severity in pregnancy.30
Causal interpretation of HR is however risky, and effect estimates averaged over the duration of study’s follow-up may not be informative.39 In this study, the HRs changed over time and this was not due to a cohort effect or to sex-specific differences. We observed lower or at least equal ADHD risk in SSRI/SNRI exposed children compared with unexposed in early childhood. Yet, an increased risk emerged in mid childhood (7-9 years). This temporal trend was apparent across all the window of exposures, except for SSRI/SNRI in late pregnancy, which partly aligns with the result of Boukhris et al .40 Further comparison with prior research is difficult since adjusted survival curves are often not presented, the follow-up time is too short, or it is unclear whether the HRs were constant over time.36,39,41-43
The observed ADHD risk reduction at early childhood aligns with our analysis by child age 5 years; nevertheless, the effect sizes were small and unlikely to reach clinical relevance. This absence of risk aligns with results of prior studies that controlled for maternal mood disorders, genetic liability or familial environment.44 Alternative explanations are however possible: chance, few ADHD cases in early childhood, or distorted maternal report on child ADHD symptoms.45
The apparent elevated risk for ADHD observed in mid childhood, at age 7-9 years, needs careful interpretation. We found some evidence for an association between SSRI/SNRI ever exposure (93% increased hazard) or for 9-20 weeks duration (2.8-fold increased hazard), relative to non-medicated in this age band. Evidence was weak for longer duration of exposure, and there were no substantial timing associations. Inattention symptoms are more easily detected as children grow older.46 Our age-specific results may then be explained by measurement issues, but bias due to frailty, small sample size, or competing risks cannot be ruled out. If the former explanation holds true, the question remains as to why measurement bias would be differential across the exposure groups. It could be argued that children prenatally exposed to an active, non-medicated depression are more susceptible to the combined type of ADHD,46 often detected in earlier childhood. This would reduce the number of susceptible children in this group over time, and in turn produce a fictitious increased hazard for the SSRI/SNRI-exposed.39 At the same time, an age-specific association between prenatal SSRI/SNRI and predominantly inattentive ADHD subtype,46 cannot be completely ruled out. The pathophysiology of ADHD involves multiple neuronal circuits, and serotonin has been shown to modulate the default mode network.47,48 Yet, the role of age-specific genetic influences, or the importance of the environmental exposure to depression or SSRI/SNRI on ADHD across ages, remains untestable in the current study.49
Our results document the confounding role of maternal pre-existing clinical depression/anxiety, yielding a risk attenuation of 18-20% in the weighted effect estimates. This factor was independently associated with child ADHD (34% increased risk), which replicates prior research.40 It is reasonable that maternal psychiatric status prior to pregnancy confound the antidepressant-ADHD association via unfavorable health behaviors, poor nutrition, or genetic make-up.