Shaded areas around the lines indicate pointwise 95% confidence intervals. Panels A and C show the full cohort. Panels B and D show siblings discordant for first-trimester antidepressant exposure. The median and interquartile range (IQR) follow-up time in the study (age since birth in years) were estimated separately for each exposure group. Median follow-up time for autism spectrum disorder in the full cohort was 8.71 years (IQR, 4.69-13.19) for the unexposed group and 5.82 years (IQR, 3.32-9.08) for the exposed group. Median follow-up time for autism spectrum disorder in the sample of discordant siblings was 9.24 years (IQR, 5.54-12.96) for the unexposed group and 6.27 years (IQR, 3.77-9.02) for the exposed group. Median follow-up time for attention-deficit/hyperactivity disorder in the full cohort was 8.54 years (IQR, 4.54-13.54) for the unexposed group and 5.54 years (IQR, 3.54-9.54) for the exposed group. Median follow-up time for attention-deficit/hyperactivity disorder in the sample of discordant siblings was 9.54 years (IQR, 5.54-13.54) for the unexposed group and 6.54 years (IQR, 3.54-9.54) for the exposed group.
eAppendix 1. Descriptive Statistics Based on Maternal Self-reported Antidepressant Use
eTable 1. Descriptive statistics stratified by maternal self-reported first-trimester use of any antidepressant in a subsample of differentially exposed siblings
eTable 2. Descriptive statistics stratified by maternal self-reported first-trimester SSRI use
eTable 3. Descriptive statistics stratified by maternal self-reported first-trimester use of SSRIs in a subsample of differentially exposed siblings
eTable 4. Information on families with outcome discordant siblings in the cohort born 1996 to 2012
eAppendix 2. Descriptive Statistics and Analyses Based on Antidepressant Dispensation Records
eTable 5. Descriptive statistics stratified by maternal dispensation windows for any antidepressants
eTable 6. Descriptive statistics stratified by maternal dispensation windows for SSRIs
eTable 7. Descriptive statistics stratified by paternal first-trimester dispensations of any antidepressants
eTable 8. Descriptive statistics stratified by paternal first-trimester dispensations of SSRIs
eTable 9. Adjusted associations between maternal antidepressant dispensations before pregnancy, during the first trimester of pregnancy, during the second and/or third trimester of pregnancy, and after pregnancy and offspring birth and neurodevelopmental outcomes
eFigure 1. Dispensation windows
eAppendix 3. Test of Exposure Misclassification
eTable 10. Adjusted associations between four definitions of first-trimester antidepressant use and offspring birth and neurodevelopmental outcomes
eAppendix 4. Test of Generalizability of Sibling Comparisons
eTable 11. Baseline and adjusted associations between maternal self-reported first-trimester antidepressant use and offspring birth and neurodevelopmental outcomes in multiple-offspring families
eAppendix 5. Test of Confounding from Exposure to Other Psychotropic Medications
eTable 12. Drug names and Anatomical Therapeutic Chemical Classification codes for other psychotropic medications
eTable 13. Baseline, adjusted, and sibling comparison associations between maternal self-reported first-trimester antidepressant use and offspring birth and neurodevelopmental outcomes in a subsample of offspring not exposed to other psychotropic medications
eAppendix 6. Test of Bias from Left Censoring and Cohort Effects
eTable 14. Baseline, adjusted, and sibling comparison associations between maternal self-reported first-trimester antidepressant use and offspring neurodevelopmental outcomes in a subsample born in 2001 or after
eAppendix 7. Test of Validity of Early Autism Spectrum Disorder and Attention-Deficit/Hyperactivity Disorder Diagnoses
eTable 15. Baseline, adjusted, and sibling comparison associations between maternal self-reported first-trimester antidepressant use and offspring neurodevelopmental outcomes in a sample excluding offspring diagnosed before age 2 years
eAppendix 8. Associations Between Dispensations During Later Pregnancy and Offspring Outcomes Among Those with a First-trimester Dispensation
eTable 16. Adjusted associations between continuation of antidepressant dispensations later in pregnancy and offspring birth and neurodevelopmental outcomes among those with a first-trimester dispensation
Customize your JAMA Network experience by selecting one or more topics from the list below.
Sujan AC, Rickert ME, Öberg AS, Quinn PD, Hernández-Díaz S, Almqvist C, Lichtenstein P, Larsson H, D’Onofrio BM. Associations of Maternal Antidepressant Use During the First Trimester of Pregnancy With Preterm Birth, Small for Gestational Age, Autism Spectrum Disorder, and Attention-Deficit/Hyperactivity Disorder in Offspring. JAMA. 2017;317(15):1553–1562. doi:10.1001/jama.2017.3413
Copyright 2017 American Medical Association. All Rights Reserved.
Is first-trimester maternal antidepressant use related to offspring birth problems, neurodevelopmental problems, or both?
In this retrospect cohort study of 1 580 629 Swedish offspring using multiple statistical and methodical approaches to adjust for confounding, first-trimester antidepressant exposure was significantly associated with preterm birth (odds ratio, 1.3 in a sibling comparison analysis) but not with risk of being born small for gestational age or later autism spectrum disorder or attention-deficit/hyperactivity disorder.
After accounting for confounding factors, first-trimester antidepressant exposure, compared with no exposure, was associated with a small increased risk of preterm birth but no increased risk of small for gestational age, autism spectrum disorder, or attention-deficit/hyperactivity disorder.
Prenatal antidepressant exposure has been associated with adverse outcomes. Previous studies, however, may not have adequately accounted for confounding.
To evaluate alternative hypotheses for associations between first-trimester antidepressant exposure and birth and neurodevelopmental problems.
Design, Setting, and Participants
This retrospective cohort study included Swedish offspring born between 1996 and 2012 and followed up through 2013 or censored by death or emigration. Analyses controlling for pregnancy, maternal and paternal covariates, as well as sibling comparisons, timing of exposure comparisons, and paternal comparisons, were used to examine the associations.
Maternal self-reported first-trimester antidepressant use and first-trimester antidepressant dispensations.
Main Outcomes and Measures
Preterm birth (<37 gestational weeks), small for gestational age (birth weight <2 SDs below the mean for gestational age), and first inpatient or outpatient clinical diagnosis of autism spectrum disorder and attention-deficit/hyperactivity disorder in offspring.
Among 1 580 629 offspring (mean gestational age, 279 days; 48.6% female; 1.4% [n = 22 544] with maternal first-trimester self-reported antidepressant use) born to 943 776 mothers (mean age at childbirth, 30 years), 6.98% of exposed vs 4.78% of unexposed offspring were preterm, 2.54% of exposed vs 2.19% of unexposed were small for gestational age, 5.28% of exposed vs 2.14% of unexposed were diagnosed with autism spectrum disorder by age 15 years, and 12.63% of exposed vs 5.46% of unexposed were diagnosed with attention-deficit/hyperactivity disorder by age 15 years. At the population level, first-trimester exposure was associated with all outcomes compared with unexposed offspring (preterm birth odds ratio [OR], 1.47 [95% CI, 1.40-1.55]; small for gestational age OR, 1.15 [95% CI, 1.06-1.25]; autism spectrum disorder hazard ratio [HR], 2.02 [95% CI, 1.80-2.26]; attention-deficit/hyperactivity disorder HR, 2.21 [95% CI, 2.04-2.39]). However, in models that compared siblings while adjusting for pregnancy, maternal, and paternal traits, first-trimester antidepressant exposure was associated with preterm birth (OR, 1.34 [95% CI, 1.18-1.52]) but not with small for gestational age (OR, 1.01 [95% CI, 0.81-1.25]), autism spectrum disorder (HR, 0.83 [95% CI, 0.62-1.13]), or attention-deficit/hyperactivity disorder (HR, 0.99 [95% CI, 0.79-1.25]). Results from analyses assessing associations with maternal dispensations before pregnancy and with paternal first-trimester dispensations were consistent with findings from the sibling comparisons.
Conclusions and Relevance
Among offspring born in Sweden, after accounting for confounding factors, first-trimester exposure to antidepressants, compared with no exposure, was associated with a small increased risk of preterm birth but no increased risk of small for gestational age, autism spectrum disorder, or attention-deficit/hyperactivity disorder.
Given the increasing prevalence of antidepressant use among pregnant women,1 gaining knowledge on the safety of use during pregnancy is a public health priority. Prenatal antidepressant exposure is associated with birth and neurodevelopmental problems, including shorter gestation,2 reduced fetal growth,2 autism spectrum disorder,3-6 and attention-deficit/hyperactivity disorder.7 These associations may be due to causal mechanisms (eg, dysfunctional serotonin signaling).8 However, there are alternative explanations for the associations. Maternal depression and stress are associated with birth and neurodevelopmental problems,9,10 suggesting that antidepressant associations could be attributable to confounding by indication for such treatment. Furthermore, autism spectrum disorder and attention-deficit/hyperactivity disorder have strong genetic influences,11 and these influences partially overlap with genetic contributions to depression.12,13 Thus, genetic transmission of shared risk for neurodevelopmental problems and depression could explain the associations (ie, passive gene-environmental correlation). Other factors, such as poor health practices during pregnancy, could also account for the associations.14
Randomized clinical trials have not been able to test the safety of antidepressant use during pregnancy because pregnant women are typically excluded from these studies. Thus, researchers must use observational designs to rule out alternative explanations for the associations.15 The present study used 4 such designs to explore associations between first-trimester antidepressant exposure (assessed via maternal self-report and registered medication dispensations) and offspring birth and neurodevelopmental problems (ie, preterm birth, small for gestational age, autism spectrum disorder, and attention-deficit/hyperactivity disorder). In addition to statistical controls to adjust for measured pregnancy and maternal and paternal characteristics, this study used sibling comparisons to account for unmeasured genetic and environmental factors that make siblings similar; timing-of-exposure comparisons to account for selection factors related to maternal antidepressant treatment around the time of pregnancy; and paternal comparisons to further account for familial confounding.
The institutional review board at Indiana University and the regional ethical review board in Stockholm, Sweden, approved this study. By Swedish law, informed consent was not necessary because the study used data available from national registries.
We obtained a population-based data set by linking information from the following Swedish registries: the Multi-Generation Register, which includes biological relationships for all individuals residing in Sweden (since 1961); the Prescribed Drug Register, which includes prescription medication dispensation records (since 2006); the Medical Birth Register, which includes information on 96% to 99% of births (since 1973); the National Patient Register, which includes diagnoses from all hospital admissions (since 1987) and specialist outpatient care (since 2001); the National Crime Register, which includes criminal convictions (since 1973); and the Swedish Register of Education, which includes highest level of completed formal education (through 2013).
The main exposures evaluated were first-trimester exposure to any antidepressants (medications with Anatomical Therapeutic Chemical Classification [ATC] codes beginning with N06A) and selective serotonin reuptake inhibitors (SSRIs; medications with ATC codes beginning with N06AB). Exposure was defined according to 2 sources of information: maternal self-reports (available for offspring born between 1996 and 2012) and dispensation records (available for both parents of offspring born between 2006 and 2012).
Information about maternal self-reported medication use during the first trimester of pregnancy came from the Medical Birth Register, which contains information obtained from standardized interviews conducted by midwives at the first antenatal visit. Medication reported in these interviews is presumed to represent first-trimester use because interviews typically occur between weeks 10 and 12 of pregnancy.
Information about medication use based on dispensation records came from the Prescribed Drug Register, which covers all medication dispensations and accompanying prescriptions written in Sweden since July 2005. The only medication use not covered by the register is medication administered while in hospital, purchased over the counter, or obtained illegally. The Prescribed Drug Register was used to obtain information on maternal antidepressant dispensations that covered the periods before pregnancy and during the first trimester of pregnancy and paternal antidepressant dispensations during the first trimester of pregnancy. First-trimester exposure was defined as having at least 1 dispensation between 90 days before estimated conception and 90 days after estimated conception (eFigure in the Supplement). The time window included 90 days before conception because chronic disease medication is typically prescribed for at least 3 months in Sweden. Use before pregnancy only was defined as having at least 1 dispensation between 270 and 90 days before estimated conception and no dispensations during pregnancy or during the first 180 days after delivery.
The birth outcomes were preterm birth (<37 gestational weeks) and small for gestational age (birth weight <2 SDs below the mean for gestational age). The neurodevelopmental outcomes were first diagnosis of autism spectrum disorder and attention-deficit/hyperactivity disorder, which were identified using inpatient and outpatient diagnoses made by specialists according to International Classification of Diseases, Ninth Revision (ICD-9) and ICD-10 criteria. Previous research has validated these diagnoses in the Swedish registries.16,17 Participants were followed up through 2013 or were censored because of death or emigration. More details about the registries and variables are available in previous studies.18,19
Pregnancy covariates included parity (categorized as first, second, third, or fourth or higher) and year of birth. Maternal and paternal covariates included country of birth (Sweden or outside of Sweden), age at childbearing (categorized into 6 levels), highest level of completed education (categorized into 7 levels), history of any criminal conviction, history of severe psychiatric illnesses (inpatient diagnosis of ICD-8, ICD-9, or ICD-10 schizophrenia, bipolar disorder, or other non-drug-induced psychoses), and history of any suicide attempts (definite or uncertain). History of criminal convictions is commonly used in Swedish register studies to index problems with behavior regulation.20,21
SAS 9.4 and STATA 13.1 were used to manage and analyze data. A complete case analysis was performed, and 95% CIs were calculated based on 2-sided hypothesis testing.
Descriptive statistics were based on the distribution of covariates and outcomes in the whole sample and subsamples of exposed and unexposed offspring, and the occurrence of the outcomes and covariates in differentially exposed and unexposed siblings was also presented. Proportions and unadjusted risk differences were used to show the birth outcomes. Because follow-up time was censored, Kaplan-Meier estimates were used to show the probability of the neurodevelopmental diagnoses.
Logistic regression was used to estimate the model-based associations for the 2 birth outcomes (ie, binary response). Cox proportional hazards regression (using calendar age in years as the timescale) was used to estimate the associations for the 2 neurodevelopmental outcomes to account for censored observations in the data. The associations between antidepressant exposure and outcomes were examined by estimating a sequence of 3 models with increasing degrees of control for potential confounding factors.
First, the baseline models assessed population-wide associations while only adjusting for pregnancy covariates (parity and year of birth).
Second, the population-wide associations were further adjusted for all maternal and paternal covariates. These population models used robust standard errors to account for clustering of individuals within nuclear families (ie, siblings) bound by the same biological mother.
Third, sibling comparison models compared exposure- and outcome-discordant offspring within families and included covariates that could vary among siblings born to the same mother. By design, these models accounted for all factors that made siblings similar (eg, shared genetic and early environmental influences), as well as measured covariates that vary within families (eg, parity and maternal and paternal age at childbearing), thereby producing a stronger test of the associations than the adjusted population models.22 As recommended,23 fixed-effects models were fit using conditional logistic and stratified Cox regression to make purely within-family comparisons.
To explore whether intrauterine exposure was specifically associated with outcomes over and above maternal depression treatment around the time of pregnancy, associations for maternal first-trimester antidepressant dispensations were compared with associations for dispensations before pregnancy, while adjusting for measured pregnancy covariates, maternal covariates, and paternal covariates. Wald χ2 tests were used to evaluate whether these associations differed statistically. Additionally, the fit of models that included separate parameters for before-pregnancy dispensations and first-trimester dispensations were compared with models that included 1 parameter for both dispensation windows.
Paternal first-trimester antidepressant dispensations were used as a negative control to further explore the role of familial confounding. First, the association between maternal and paternal first-trimester dispensations was assessed. Next, estimates were obtained for associations between paternal first-trimester antidepressant dispensations and the 4 outcomes while adjusting for the pregnancy covariates.
First, to evaluate the influence of exposure misclassification, adjusted associations with 4 additional exposure definitions were assessed in the cohort of offspring born between 2006 and 2012. This cohort was used because information on exposure based on both maternal self-reports and maternal dispensations were available. The 4 additional definitions included the following: first-trimester exposure defined as use according to either self-reports or dispensation records; first-trimester exposure defined as use according to self-reports and also to dispensation records; a narrower first-trimester dispensation window of 30 days before conception to 90 days after conception; and at least 2 dispensations during the original first-trimester exposure window.
Second, given that single-offspring families cannot contribute to sibling-comparison analyses, the population models in the subsample of offspring with siblings were reassessed to evaluate the generalizability of sibling-comparison results.
Third, to assess whether exposure to other psychotropic medications confounded the associations, the analyses were restricted to offspring not exposed to other psychotropic medications.
Fourth, given that prior to 2001, outpatient psychiatric diagnoses were not included in the National Patient Register, analyses were conducted on a subsample of offspring born after 2000 to assess whether left censoring of the neurodevelopmental outcomes biased the findings. These analyses also enabled exploration of whether cohort effects influenced the results.
Fifth, the associations with the neurodevelopmental outcomes were estimated in subsamples excluding offspring with diagnoses before age 2 years to address concerns about the validity of early neurodevelopmental diagnoses.
Sixth, because the main analyses focused on first-trimester exposure, the association was examined between dispensations during the second or third trimester and each outcome in the subsample of offspring whose mothers had a dispensation during the first trimester.
The target sample included 1 670 237 offspring born between 1996 and 2012. Cases of multiple births (48 979 offspring), those with a missing father identifier (16 295), missing or invalid responses on covariates (20 118), and missing the small for gestational age variable (4216) were sequentially excluded.
The final analytic cohort of 1 580 629 offspring (48.6% female) represented 95% of target singleton births and included 943 776 distinct mothers and 946 579 distinct fathers. According to maternal self-reports, 22 544 (1.4%) of the offspring in the final cohort were exposed to any antidepressant during the first trimester, and of these, 82% (18 470) were exposed to SSRIs.
The timing of exposure and paternal comparisons were conducted on the subsample of 708 450 offspring (born between 2006-2012) with dispensation-based exposure data. There were 26 477 (3.7%) offspring with first-trimester maternal antidepressant dispensations. Of these, 84% (22 125) had first-trimester maternal SSRI dispensations specifically. There were 8203 (1.2%) offspring who had mothers with antidepressant dispensations before pregnancy only. Of these, 81% (6674) had mothers who were specifically dispensed SSRIs before pregnancy. There were 18 727 (2.6%) offspring who had fathers with first-trimester antidepressant dispensations. Of these, 72% (13 521) had fathers with first-trimester SSRI dispensations specifically.
The same pattern of results was observed for associations with first-trimester exposure to any antidepressant as for first-trimester exposure to SSRIs specifically. Therefore, results for exposure to any antidepressant are presented in the text and tables, and the results for SSRIs can be found in the tables and in the Supplement.
In the whole sample, 6.98% of exposed and 4.78% of unexposed offspring were preterm (Table 1), which equates to 220 (95% CI, 187-254) additional preterm birth cases per 10 000 offspring. Approximately 2.54% of exposed and 2.19% of unexposed offspring were born small for gestational age (risk difference, 35 additional cases per 10 000 offspring [95% CI, 14-56]). When compared with unexposed offspring, exposed offspring also had a higher probability of the neurodevelopmental diagnoses (Figure, panels A and C). For example, by age 15 years, Kaplan-Meier estimates indicated a cumulative risk of autism spectrum disorder of 5.28% for exposed and 2.14% for unexposed offspring. By age 15 years, the cumulative risk of attention-deficit/hyperactivity disorder was 12.63% for exposed and 5.46% for unexposed offspring (see eTables 1-3 in the Supplement for more descriptive information).
Among differentially exposed siblings, 6.24% of exposed and 5.07% of unexposed siblings were born preterm (eTable 1 in the Supplement). However, 1.87% of exposed and 1.98% of unexposed siblings were small for gestational age. By age 15 years, the cumulative risk for autism spectrum disorder was 5.52% for exposed and 4.55% for unexposed siblings; the cumulative risk for attention-deficit/hyperactivity disorder was 12.38% for exposed and 12.73% for unexposed siblings (Figure, panels B and D show the probabilities of the neurodevelopmental diagnoses among differentially exposed siblings through age 15 years).
In the baseline models (Table 2), maternal self-reported first-trimester antidepressant use was associated with preterm birth (odds ratio [OR], 1.47 [95% CI, 1.40-1.55]), small for gestational age (OR, 1.15 [95% CI, 1.06-1.25]), autism spectrum disorder (hazard ratio [HR], 2.02 [95% CI, 1.80-2.26]), and attention-deficit/hyperactivity disorder (HR, 2.21 [95% CI, 2.04-2.39]). In the adjusted models, first-trimester exposure to antidepressants was also statistically significantly associated with all outcomes (preterm birth OR, 1.35 [95% CI, 1.28-1.42]; small for gestational age OR, 1.12 [95% CI, 1.03-1.22]; autism spectrum disorder HR, 1.64 [95% CI, 1.46-1.83]; attention-deficit/hyperactivity disorder HR, 1.58 [95% CI, 1.46-1.71]).
In the sibling comparison models, first-trimester exposure was associated with preterm birth (OR, 1.34 [95% CI, 1.18-1.52]; P < .001). However, it was not associated with small for gestational age (OR, 1.01 [95% CI, 0.81-1.25]), autism spectrum disorder (HR, 0.83 [95% CI, 0.62-1.13]), or attention-deficit/hyperactivity disorder (HR, 0.99 [95% CI, 0.79-1.25]). See eTable 4 in the Supplement for information on offspring who could contribute to sibling comparison analyses.
Dispensation data was used in timing of exposure and paternal comparisons (eTables 5-9 in the Supplement). For preterm birth, the association with maternal dispensations before pregnancy, but not during or after pregnancy (OR, 1.17 [95% CI, 1.07-1.28]; Table 3), was statistically significantly weaker than the association with first-trimester maternal dispensations (OR, 1.40 [95% CI, 1.33-1.47]). For all other outcomes, the associations with maternal dispensations, before but not during or after pregnancy, did not statistically significantly differ from the associations with first-trimester maternal dispensations.
Paternal first-trimester antidepressant dispensations were associated with maternal first-trimester antidepressant dispensations (OR, 3.42 [95% CI, 3.26-3.58]). Paternal first-trimester antidepressant dispensations (Table 4) had very modest associations with preterm birth (OR, 1.13 [95% CI, 1.05-1.20]) and small for gestational age (OR, 1.06 [95% CI, 0.96-1.17]), with the latter not being statistically significant. Paternal dispensations during pregnancy were associated with autism spectrum disorder (HR, 1.31 [95% CI, 1.05-1.62]) and attention-deficit/hyperactivity disorder (HR, 1.73 [95% CI, 1.38-2.17]).
Sensitivity analyses showed a consistent pattern of results across analyses using more strict criteria for exposure and narrower exposure windows, suggesting that exposure misclassification was not responsible for the pattern of findings (eTable 10 in the Supplement). Results from population models conducted on a subsample that excluded offspring who did not have siblings also were essentially identical to the main results (eTable 11 in the Supplement). These results provide support for the generalizability of sibling comparison results. Sensitivity analyses also suggested that confounding by exposure to other psychotropic medications (eTable 12 and eTable 13 in the Supplement), left censoring of the neurodevelopmental outcomes and cohort effects (eTable 14 in the Supplement), and measurement error of the neurodevelopmental outcomes (eTable 15 in the Supplement) had very little influence on the results. In addition, among offspring whose mothers received a dispensation during the first trimester, a dispensation during the second or third trimester was associated with increased risk of the pregnancy outcomes, although the associations with the neurodevelopmental diagnoses were not statistically significant (eTable 16 in the Supplement).
The present study found that after accounting for measured pregnancy, maternal and paternal traits, and all (unmeasured) stable familial characteristics shared by siblings, maternal antidepressant use during the first trimester of pregnancy, compared with no exposure, was associated with a small increased risk of preterm birth but no increased risk of small for gestational age, autism spectrum disorder, or attention-deficit/hyperactivity disorder. That is, unexposed siblings were at equal risk for small for gestational age, autism spectrum disorder, and attention-deficit/hyperactivity disorder as their exposed siblings. These results are consistent with the hypothesis that genetic factors, familial environmental factors, or both account for the population-wide associations between first-trimester antidepressant exposure and these outcomes. Moreover, results from analyses examining timing of exposure were consistent with the interpretation of the sibling-comparison findings. Specifically, the strength of the associations between antidepressant dispensations before pregnancy and small for gestational age, autism spectrum disorder, and attention-deficit/hyperactivity disorder did not statistically significantly differ from the strength of associations for first-trimester antidepressant dispensations, suggesting that the underlying condition, rather than exposure to antidepressants during the first trimester, explained the associations.
Paternal first-trimester antidepressant dispensations were also associated with both neurodevelopmental disorders. Because paternal antidepressant use during the first trimester is unlikely to contribute to intrauterine exposure, these findings provide further support that associations between first-trimester antidepressant exposure and offspring neurodevelopmental disorders may, at least partially, be explained by familial confounding.
The results also showed that across multiple designs that account for familial confounding factors, first-trimester antidepressant exposure was associated with a slightly elevated risk of preterm birth. Although these results may be consistent with the hypothesis that prenatal antidepressant exposure could lead to a small increased risk of preterm birth, other possible explanations for the findings need to be considered. Most important, the potential role of confounding by maternal depression should be noted because both the existence and severity of depression symptoms in the mother could potentially influence the risk of preterm birth.24
The results of the population-wide models were consistent with numerous observational studies that have demonstrated associations between prenatal antidepressant exposure and birth and neurodevelopmental problems.2-7 The results of the sibling comparisons were also consistent with the limited previous sibling comparison studies that have examined associations between prenatal antidepressant exposure and birth and neurodevelopmental problems. A sibling comparison study using dispensation data from the Swedish registries found statistically significant associations between prenatal antidepressant dispensations and shorter gestation.25 Another sibling comparison reported that prenatal antidepressant exposure was not associated with autism spectrum disorder,26 although CIs were too wide to draw strong conclusions.
The current study had several strengths. First, the study analyzed a large, population-based sample, which provided statistical power to examine rare yet serious outcomes. Second, the conclusions were based on converging evidence from multiple research designs that accounted for both measured and unmeasured confounding factors. Third, first-trimester antidepressant use was indexed by maternal self-report and dispensations. Fourth, the study included 4 outcomes (2 pregnancy-related and 2 neurodevelopmental problems), all of which are associated with significant morbidity and mortality. Fifth, sensitivity analyses suggested that misclassification of antidepressant use, several assumptions of sibling comparison analyses, confounding by other psychotropic medications, and misclassification of the neurodevelopmental problems were unlikely to influence the overall conclusions.
The findings from the present study should be considered in light of several limitations. First, and most important, observational designs such as these cannot fully rule out all sources of confounding. In particular, like other register-based approaches,26 this study could not comprehensively assess maternal depression or its severity,27 nor could it compare different antidepressant treatment regimens. Thus, associations could have been influenced by confounding by antidepressant indication. This study used multiple designs to address this limitation, each of which could help rule out some but not all sources of confounding, to provide complementary evidence. For example, sibling comparisons ruled out all stable confounders (eg, chronic maternal depression), but that design may not have been able to account for confounding from maternal depression that varied across pregnancies.28 Thus, the within-family associations with preterm birth may plausibly be driven by unmeasured time-varying maternal depression rather than by antidepressant use.29
Second, this study focused on first-trimester exposure. Whereas one recent study found an association between antidepressant dispensations late, but not early in pregnancy and autism spectrum disorder,3 there has been considerable debate regarding the role of timing.30-32 In fact, several studies have found stronger associations with first-trimester antidepressant use than with use later in pregnancy.4,6 Supplemental analyses indicated that among offspring of mothers who received a dispensation during the first trimester, a dispensation during the second or third trimester was associated with greater risk of offspring being born preterm and small for gestational age. These associations could be due to intrauterine exposure to antidepressants later in pregnancy, increased severity of depression (ie, confounding by indication), or other unmeasured confounding. Therefore, future studies are needed to explicitly examine whether timing of exposure moderates the preterm birth association or whether exposure later in pregnancy is more strongly associated with other outcomes.
Third, the vast majority of antidepressant exposure (82% according to maternal reports) was to SSRIs. Future research should explore class- and drug-specific associations.
Fourth, analyses were conducted on a Swedish sample, and it is not known if results would generalize to other countries. Although the population-wide associations in the present study were commensurate with those from other countries, future research should use designs that help account for unmeasured confounders to explore associations with prenatal antidepressant exposure in the United States and elsewhere.
Fifth, sibling comparisons require large samples to have adequate statistical power.33 Although the large Swedish sample ensured fairly precise parameter estimates in sibling comparisons, small effects of antidepressant exposure cannot be ruled out. However, their magnitudes, particularly for the neurodevelopmental outcomes, would be much smaller than those suggested by population-wide associations.
Among offspring born in Sweden, after accounting for confounding factors, first-trimester exposure to antidepressants, compared with no exposure, was associated with a small increased risk of preterm birth but no increased risk of small for gestational age, autism spectrum disorder, or attention-deficit/hyperactivity disorder.
Corresponding Author: Brian M. D’Onofrio, PhD, Department of Psychological and Brain Sciences, Indiana University, 1101 E 10th St, Bloomington, IN 47405 (firstname.lastname@example.org).
Author Contributions: Dr Rickert and Ms Sujan had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Sujan, Rickert, Oberg, Hernandez-Diaz, Lichtenstein, Larsson, D'Onofrio.
Acquisition, analysis, or interpretation of data: Sujan, Rickert, Quinn, Hernandez-Diaz, Almqvist, Lichtenstein, D'Onofrio.
Drafting of the manuscript: Sujan, D'Onofrio.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Sujan, Rickert.
Obtained funding: Sujan, Quinn, Almqvist, Lichtenstein, Larsson, D'Onofrio.
Administrative, technical, or material support: Lichtenstein.
Supervision: Oberg, Hernandez-Diaz, Larsson, D'Onofrio.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Larsson reports having served as a speaker for Eli Lilly and Shire and has receiving a research grant from Shire; all outside the submitted work. Dr Hernández-Díaz reports receipt of salary support from the North American AED Pregnancy Registry; research funding from GSK, Eli Lilly, and Pfizer; and consulting for UCB, Teva, and Boehringer-Ingelheim. Dr Lichtenstein reports having served as a speaker for Medice. No other disclosures were reported.
Funding/Support: Research reported in this publication was supported by the National Institute of Mental Health of the National Institutes of Health (NIH) (T32MH103213) and the National Institute on Drug Abuse of NIH (K99DA040727). This work was also supported by a National Science Foundation Graduate Research Fellowship (1342962); Indiana Clinical and Translational Sciences Institute: Pediatric Project Development Team; the Swedish Initiative for Research on Microdata in the Social and Medical Sciences (SIMSAM) framework (340-2013-5867); the Swedish Research Council for Health, Working Life, and Welfare (FORTE; 50623213); and the Swedish Research Council (2014-38313831).
Role of the Funder/Sponsor: The funders of the study had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript or the decision to submit for publication.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.