To our knowledge, this large population-based study is the first Asian investigation to examine the association between antidepressant exposure during pregnancy and subsequent neurodevelopmental disorders. In the present study, a positive association between neurodevelopmental outcomes and antidepressant exposure was observed in all three trimesters as well as before and after pregnancy in population-wide or adjusted analysis controlling for paternal, maternal and child covariates. However, the association between antidepressant exposure and risks of neurodevelopmental outcomes was no longer observed in the sibling comparison models which account for confounding genetic factors. Paternal antidepressant exposure was noted to be associated with ADHD in offspring which suggested unmeasured confounding factors existed. These results indicated that the correlation between neurodevelopmental outcomes and antidepressant exposure during pregnancy may be explained by maternal depression, genetic factors or shared environmental factors rather than the direct intrauterine antidepressant effect.
Our finding of no increased risk for ASD and ADHD associated with antidepressant exposure during pregnancy after controlling for maternal depression, genetic and potential environmental confounding factors is in agreement with previous studies [28,29]. A previous Swedish study revealed no significant association between antidepressant exposure during the first trimester of pregnancy and risks of ADHD and ASD [29]. Our findings supported previous research and added to the evidence that antidepressant exposure regardless of trimesters of pregnancy is not substantially associated with neurodevelopmental outcomes. Nonetheless, another sibling comparison analysis revealed that maternal antidepressant use during pregnancy was associated with an increased risk of autism, though the absolute risk of autism demonstrated in that study was small [27]. Moreover, the authors of that study noted that the results may be influenced by genetic factors or severity of depression. It is important that sibling comparison cannot perfectly eliminate shared familial confounding factors [37,38]. Our findings provide important information into the effect of antidepressants on the risk for ASD and ADHD in offspring. Clinicians should consider the severity of depressive symptoms, the suicidal or self-harm risk, the possible adverse effect of medication, and the risk for neonatal and neurodevelopmental outcomes when treating depression in pregnant women [39].
Another notable observation is that a positive association was observed between ASD and ADHD and paternal use of antidepressants in adjusted analysis and this association was eliminated in the sibling comparison. This finding proposed the existence of unmeasured confounding factors such as genetic influences, environmental factors or familial stressors, because paternal antidepressant exposure should not contribute to an intrauterine medication effect. Prior study has reported that paternal antidepressants use may play a partial role in neurodevelopmental outcomes in offspring [40]. The possible mechanisms include that SSRI may impair semen quality or may lead to abnormal sperm DNA fragmentation, which was associated with adverse outcomes on fertilization and pregnancy [41–43]. Another potential explanation for the positive association between paternal antidepressant use and neurodevelopmental outcomes in the offspring may include mate selection [44]. Future research is warranted to investigate the relationship between paternal antidepressant use and neurodevelopmental outcomes.
Previous research has proposed that exposure to SSRI during the early brain development stage may cause serotonin receptor down-regulation leading to neurodevelopmental disorders [22,23]. Therefore, we divided antidepressants into SSRI and other antidepressants to explore the serotonin effect on neurodevelopmental outcomes. We found that SSRI was associated with an increased risk of both ASD and ADHD in all three trimesters after adjusting for child, maternal, and paternal covariate. Other antidepressants were also linked to a higher risk of ADHD throughout the three trimesters and were associated with ASD only when used during the first trimester. However, when considering unmeasured confounding factors, all these association became non-significant in sibling comparison. Prior research investigated serotonin effects on neurodevelopmental outcomes but produced contradictory results [28,45–47]. A Swedish population study reported that prenatal exposure to escitalopram, but not with other SSRIs, was associated with an increased risk for ASD [45]. Another population study reported that exposure to escitalopram correlated to higher risk for all neurodevelopmental outcomes except for ASD [28]. However, there were also a number of studies which indicated no association between SSRI and neurodevelopmental outcomes including ASD [46,47]. The potential reason for this discrepancy may have resulted from the differences of data sources, measurements of medication prescription and outcomes, confounding by indications, distinct methodologic design and difficulties in adjusting for the confounding factors. The detailed mechanism of serotonin affecting brain neurodevelopment remains elusive and requires further clarification.
There are several strengths of this study. This is a 13-year cohort study utilizing a large population-based dataset integrating birth registration records, clinical information, medical care utilization, diagnoses, and medication prescriptions, with sufficient statistical power. The use of data on medical diagnoses and prescription records from physicians could avoid recall bias. In addition, several analyses, including a population-wide analysis, comparisons of maternal siblings, and paternal control analyses, were conducted to account for measured and unmeasured confounding factors to strengthen confidence in the results. Furthermore, antidepressants were divided into SSRI and other antidepressants in order to clarify the effect of medications on neurodevelopmental outcomes in children.
Several limitations of this study should also be mentioned. First, some essential information such as lifestyle, alcohol drinking, cigarette smoking, and socioeconomic status was not obtained in this dataset. Moreover, the use of prescription-filling data may not accurately reflect patients’ adherence to antidepressant treatment and antidepressant exposure. However, previous research has indicated that prescription-filling patterns remain the most accurate source for estimating medication intake in large observational studies [48,49]. Second, lack of comprehensive measures of the severity of maternal depression was a limitation. We utilized sibling comparisons to adjust for maternal depression, treating it as a constant variable over time. Nonetheless, such comparisons cannot account for non-familial confounding factors and may introduce bias by adjusting for shared mediators [38,50,51]. Third, despite the large sample size, the number of participants with neurodevelopmental disorders in the sibling comparison was small; therefore, inferences should be made with caution. Fourth, the absence of data on ASD traits or ADHD-like symptoms in our database might lead to an underestimation of the risk for ASD and ADHD among siblings. Furthermore, some participants had a short follow-up period, potentially affecting the assessment of ADHD incidence. However, we calculated the relative risk by comparing both antidepressant-exposed and -unexposed groups. Hence, the effect on the estimated risk ratio would be modest.