[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
Purchase Options:
[Skip to Content Landing]
Figure 1.
Association Between Serotonergic Antidepressant Exposure and Child Autism Spectrum Disorder by Trimester of Exposure
Association Between Serotonergic Antidepressant Exposure and Child Autism Spectrum Disorder by Trimester of Exposure

aBased on a high-dimensional propensity score. The high-dimensional propensity score was reestimated and the inverse probability of treatment weights was regenerated for each subanalysis.

Figure 2.
Association Between Serotonergic Antidepressant Exposure and Child Autism Spectrum Disorder by Type of Serotonergic Antidepressant
Association Between Serotonergic Antidepressant Exposure and Child Autism Spectrum Disorder by Type of Serotonergic Antidepressant

SNRI indicates serotonin-norepinephrine reuptake inhibitor (98% of exposures were with venlafaxine); SSRI, selective serotonin reuptake inhibitor.

aBased on a high-dimensional propensity score. The high-dimensional propensity score was reestimated and the inverse probability of treatment weights was regenerated for each subanalysis.

Table 1.  
Baseline Characteristics of Serotonergic Antidepressant Users and Nonusers Compared Using Standardized Differencesa
Baseline Characteristics of Serotonergic Antidepressant Users and Nonusers Compared Using Standardized Differencesa
Table 2.  
Association Between In Utero Serotonergic Antidepressant Exposure and Child Autism Spectrum Disorder
Association Between In Utero Serotonergic Antidepressant Exposure and Child Autism Spectrum Disorder
1.
Le Strat  Y, Dubertret  C, Le Foll  B.  Prevalence and correlates of major depressive episode in pregnant and postpartum women in the United States.  J Affect Disord. 2011;135(1-3):128-138.PubMedGoogle ScholarCrossref
2.
Stein  A, Pearson  RM, Goodman  SH,  et al.  Effects of perinatal mental disorders on the fetus and child.  Lancet. 2014;384(9956):1800-1819.PubMedGoogle ScholarCrossref
3.
Howard  LM, Molyneaux  E, Dennis  CL,  et al.  Non-psychotic mental disorders in the perinatal period.  Lancet. 2014;384(9956):1775-1788.PubMedGoogle ScholarCrossref
4.
Fournier  JC, DeRubeis  RJ, Hollon  SD,  et al.  Antidepressant drug effects and depression severity.  JAMA. 2010;303(1):47-53.PubMedGoogle ScholarCrossref
5.
Andrade  SE, Reichman  ME, Mott  K,  et al.  Use of selective serotonin reuptake inhibitors (SSRIs) in women delivering liveborn infants and other women of child-bearing age within the US Food and Drug Administration’s Mini-Sentinel program.  Arch Womens Ment Health. 2016;19(6):969-977.PubMedGoogle ScholarCrossref
6.
American Psychiatric Association.  Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Arlington, VA: American Psychiatric Publishing; 2013.
7.
Harrington  RA, Lee  LC, Crum  RM,  et al.  Serotonin hypothesis of autism: implications for selective serotonin reuptake inhibitor use during pregnancy.  Autism Res. 2013;6(3):149-168.PubMedGoogle ScholarCrossref
8.
Borue  X, Chen  J, Condron  BG.  Developmental effects of SSRIs.  Int J Dev Neurosci. 2007;25(6):341-347.PubMedGoogle ScholarCrossref
9.
Brown  HK, Hussain-Shamsy  N, Lunsky  Y,  et al.  The association between antental exposure to selective serotonin reuptake inhibitors and autism.  J Clin Psychiatry. 2017;78(1):e48-e58.PubMedGoogle ScholarCrossref
10.
Hviid  A, Melbye  M, Pasternak  B.  Use of selective serotonin reuptake inhibitors during pregnancy and risk of autism.  N Engl J Med. 2013;369(25):2406-2415.PubMedGoogle ScholarCrossref
11.
Vigod  SN, Gomes  T, Wilton  AS,  et al.  Antipsychotic drug use in pregnancy.  BMJ. 2015;350:h2298.PubMedGoogle ScholarCrossref
12.
Ray  JG, Schull  MJ, Urquia  ML,  et al.  Major radiodiagnostic imaging in pregnancy and the risk of childhood malignancy.  PLoS Med. 2010;7(9):e1000337.PubMedGoogle ScholarCrossref
13.
Burke  JP, Jain  A, Yang  W,  et al.  Does a claims diagnosis of autism mean a true case?  Autism. 2014;18(3):321-330.PubMedGoogle ScholarCrossref
14.
Frenette  P, Dodds  L, MacPherson  K,  et al.  Factors affecting the age at diagnosis of autism spectrum disorders in Nova Scotia, Canada.  Autism. 2013;17(2):184-195.PubMedGoogle ScholarCrossref
15.
Austin  PC.  Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples.  Stat Med. 2009;28(25):3083-3107.PubMedGoogle ScholarCrossref
16.
Hux  JE, Ivis  F, Flintoft  V, Bica  A.  Diabetes in Ontario.  Diabetes Care. 2002;25(3):512-516.PubMedGoogle ScholarCrossref
17.
Quan  H, Khan  N, Hemmelgarn  BR,  et al.  Validation of a case definition to define hypertension using administrative data.  Hypertension. 2009;54(6):1423-1428.PubMedGoogle ScholarCrossref
18.
Parkerson  GR  Jr, Harrell  FE  Jr, Hammond  WE, Wang  XQ.  Characteristics of adult primary care patients as predictors of future health services charges.  Med Care. 2001;39(11):1170-1181.PubMedGoogle ScholarCrossref
19.
Schneeweiss  S, Rassen  JA, Glynn  RJ,  et al.  High-dimensional propensity score adjustment in studies of treatment effects using health care claims data.  Epidemiology. 2009;20(4):512-522.PubMedGoogle ScholarCrossref
20.
Hernán  MA, Robins  JM.  Estimating causal effects from epidemiological data.  J Epidemiol Community Health. 2006;60(7):578-586.PubMedGoogle ScholarCrossref
21.
Guertin  JR, Rahme  E, Dormuth  CR, LeLorier  J.  Head to head comparison of the propensity score and the high-dimensional propensity score matching methods.  BMC Med Res Methodol. 2016;16:22.PubMedGoogle ScholarCrossref
22.
Enders  D, Ohlmeier  C, Garbe  E.  The potential of high-dimensional propensity scores in health services research: an exemplary study on the quality of care for elective percutaneous coronary interventions [published online January 16, 2017].  Health Serv Res. doi:10.1111/1475-6773.12653PubMedGoogle Scholar
23.
Schneeweiss  S, Eddings  W, Glynn  RJ,  et al.  Variable selection for confounding adjustment in high-dimensional covariate spaces when analyzing healthcare databases.  Epidemiology. 2017;28(2):237-248.PubMedGoogle ScholarCrossref
24.
Austin  PC, Stuart  EA.  Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies.  Stat Med. 2015;34(28):3661-3679.PubMedGoogle ScholarCrossref
25.
Austin  PC.  Variance estimation when using inverse probability of treatment weighting (IPTW) with survival analysis.  Stat Med. 2016;35(30):5642-5655.PubMedGoogle ScholarCrossref
26.
Donovan  SJ, Susser  E.  Advent of sibling designs.  Int J Epidemiol. 2011;40(2):345-349.PubMedGoogle ScholarCrossref
27.
Idring  S, Magnusson  C, Lundberg  M,  et al.  Parental age and the risk of autism spectrum disorders.  Int J Epidemiol. 2014;43(1):107-115.PubMedGoogle ScholarCrossref
28.
Croen  LA, Grether  JK, Yoshida  CK,  et al.  Antidepressant use during pregnancy and childhood autism spectrum disorders.  Arch Gen Psychiatry. 2011;68(11):1104-1112.PubMedGoogle ScholarCrossref
29.
Harrington  RA, Lee  LC, Crum  RM,  et al.  Prenatal SSRI use and offspring with autism spectrum disorder or developmental delay.  Pediatrics. 2014;133(5):e1241-e1248.PubMedGoogle ScholarCrossref
30.
Gidaya  NB, Lee  BK, Burstyn  I,  et al.  In utero exposure to selective serotonin reuptake inhibitors and risk for autism spectrum disorder.  J Autism Dev Disord. 2014;44(10):2558-2567.PubMedGoogle ScholarCrossref
31.
Boukhris  T, Sheehy  O, Mottron  L, Bérard  A.  Antidepressant use during pregnancy and the risk of autism spectrum disorder in children.  JAMA Pediatr. 2016;170(2):117-124.PubMedGoogle ScholarCrossref
32.
Sørensen  MJ, Grønborg  TK, Christensen  J,  et al.  Antidepressant exposure in pregnancy and risk of autism spectrum disorders.  Clin Epidemiol. 2013;5:449-459.PubMedGoogle ScholarCrossref
33.
Kyriacou  DN, Lewis  RJ.  Confounding by indication in clinical research.  JAMA. 2016;316(17):1818-1819.PubMedGoogle ScholarCrossref
34.
Cross-Disorder Group of the Psychiatric Genomics Consortium.  Identification of risk loci with shared effects on five major psychiatric disorders.  Lancet. 2013;381(9875):1371-1379.PubMedGoogle ScholarCrossref
35.
Oberlander  TF, Vigod  SN.  Developmental effects of prenatal selective serotonin reuptake inhibitor exposure in perspective.  J Am Acad Child Adolesc Psychiatry. 2016;55(5):351-352.PubMedGoogle ScholarCrossref
36.
Jokiranta-Olkoniemi  E, Cheslack-Postava  K, Sucksdorff  D,  et al.  Risk of psychiatric and neurodevelopmental disorders among siblings of probands with autism spectrum disorders.  JAMA Psychiatry. 2016;73(6):622-629.PubMedGoogle ScholarCrossref
37.
Franklin  JM, Eddings  W, Glynn  RJ, Schneeweiss  S.  Regularized regression versus the high-dimensional propensity score for confounding adjustment in secondary database analyses.  Am J Epidemiol. 2015;182(7):651-659.PubMedGoogle ScholarCrossref
38.
Robins  JM.  Data, design, and background knowledge in etiologic inference.  Epidemiology. 2001;12(3):313-320.PubMedGoogle ScholarCrossref
39.
Chaste  P, Leboyer  M.  Autism risk factors: genes, environment, and gene-environment interactions.  Dialogues Clin Neurosci. 2012;14(3):281-292.PubMedGoogle Scholar
Original Investigation
April 18, 2017

Association Between Serotonergic Antidepressant Use During Pregnancy and Autism Spectrum Disorder in Children

Author Affiliations
  • 1Women’s College Research Institute, Women’s College Hospital, Toronto, Ontario, Canada
  • 2Department of Psychiatry, University of Toronto, Toronto, Ontario, Canada
  • 3Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada
  • 4Department of Obstetrics and Gynaecology, St Michael’s Hospital, Toronto, Ontario, Canada
  • 5Department of Medicine, University of Toronto, Toronto, Ontario, Canada
  • 6Centre for Addiction and Mental Health, Toronto, Ontario, Canada
  • 7Leslie Dan Faculty of Pharmacy, University of Toronto, Toronto, Ontario, Canada
 

Copyright 2017 American Medical Association. All Rights Reserved.

JAMA. 2017;317(15):1544-1552. doi:10.1001/jama.2017.3415
Key Points

Question  What is the risk of autism spectrum disorder in children with in utero exposure to a serotonergic antidepressant compared with unexposed children?

Findings  In this retrospective cohort study of 35 906 singleton births, there was not a statistically significant association between exposure to serotonergic antidepressants compared with no exposure in inverse probability of treatment-weighted analyses or when comparing exposed with unexposed siblings.

Meaning  Maternal use of serotonergic antidepressants during pregnancy compared with no use was not associated with autism spectrum disorder in their children. Although a causal relationship cannot be ruled out, the previously observed association may be explained by other factors.

Abstract

Importance  Previous observations of a higher risk of child autism spectrum disorder with serotonergic antidepressant exposure during pregnancy may have been confounded.

Objective  To evaluate the association between serotonergic antidepressant exposure during pregnancy and child autism spectrum disorder.

Design, Setting, and Participants  Retrospective cohort study. Health administrative data sets were used to study children born to mothers who were receiving public prescription drug coverage during pregnancy in Ontario, Canada, from 2002-2010, reflecting 4.2% of births. Children were followed up until March 31, 2014.

Exposures  Serotonergic antidepressant exposure was defined as 2 or more consecutive maternal prescriptions for a selective serotonin or serotonin-norepinephrine reuptake inhibitor between conception and delivery.

Main Outcomes and Measures  Child autism spectrum disorder identified after the age of 2 years. Exposure group differences were addressed by inverse probability of treatment weighting based on derived high-dimensional propensity scores (computerized algorithm used to select a large number of potential confounders) and by comparing exposed children with unexposed siblings.

Results  There were 35 906 singleton births at a mean gestational age of 38.7 weeks (50.4% were male, mean maternal age was 26.7 years, and mean duration of follow-up was 4.95 years). In the 2837 pregnancies (7.9%) exposed to antidepressants, 2.0% (95% CI, 1.6%-2.6%) of children were diagnosed with autism spectrum disorder. The incidence of autism spectrum disorder was 4.51 per 1000 person-years among children exposed to antidepressants vs 2.03 per 1000 person-years among unexposed children (between-group difference, 2.48 [95% CI, 2.33-2.62] per 1000 person-years; hazard ratio [HR], 2.16 [95% CI, 1.64-2.86]; adjusted HR, 1.59 [95% CI, 1.17-2.17]). After inverse probability of treatment weighting based on the high-dimensional propensity score, the association was not significant (HR, 1.61 [95% CI, 0.997-2.59]). The association was also not significant when exposed children were compared with unexposed siblings (incidence of autism spectrum disorder was 3.40 per 1000 person-years vs 2.05 per 1000 person-years, respectively; adjusted HR, 1.60 [95% CI, 0.69-3.74]).

Conclusions and Relevance  In children born to mothers receiving public drug coverage in Ontario, Canada, in utero serotonergic antidepressant exposure compared with no exposure was not associated with autism spectrum disorder in the child. Although a causal relationship cannot be ruled out, the previously observed association may be explained by other factors.

Introduction

Depression is one of the most common complications during pregnancy, affecting 1 in every 10 women.1 Untreated depression is associated with adverse consequences for exposed children such as poor fetal growth, preterm birth, and developmental problems.2 It is also associated with increased risk for postpartum and chronic maternal depression, conditions linked to adverse child outcomes.3 Women with mild depression may be effectively treated with nonpharmacological strategies; however, use of an antidepressant medication may be required for women with more severe illness.4 In a study of 15 health insurance plans in the United States (data collected from 2001-2013),5 6% of pregnant women had taken selective serotonin reuptake inhibitor (SSRI) antidepressants. Accurate information about the risks and benefits of SSRIs is needed.

Autism spectrum disorders are neurodevelopmental disorders characterized by deficits in social communication and repetitive behavior patterns.6 Both environmental and genetic factors are thought to cause autism spectrum disorder. The serotonin hypothesis of autism posits that fetal exposure to elevated serotonin levels could result in autism spectrum disorder.7 Animal studies suggest that in utero exposure to serotonergic antidepressants such as SSRIs could lead to dysfunctional serotonin signaling and loss of serotonin terminals in exposed fetuses, resulting in the autism phenotype.8

A meta-analysis including studies of children born from 1995 through 2009 found a significant association between first trimester SSRI exposure and autism spectrum disorder in 4 studies (adjusted pooled odds ratio, 1.7; 95% CI, 1.1-2.6).9 Included studies were limited in their ability to account for potential confounders. The association was not present in 1 of the only studies that controlled for a comprehensive set of key confounders.10

The aim of this study was to examine autism spectrum disorder risk in a large cohort of children who were either exposed or unexposed to serotonergic antidepressants in utero using a variety of methods to address confounding.

Methods
Data Sources and Study Cohort

This was a retrospective cohort study using health administrative data from Ontario, Canada (population: approximately 13 million). Ontario provides universal health care, including ambulatory and in-hospital mental health care, obstetric, and pediatric care to its residents. Ontarians who meet eligibility criteria such as having either a disability or low income relative to drug costs receive publicly funded drug benefits. The data were analyzed at the Institute for Clinical Evaluative Sciences, an independent, nonprofit organization that evaluates health care services in Ontario.

At the Institute for Clinical Evaluative Sciences, patient-level records across several databases are linked using unique encoded identifiers. Births were identified using the MOMBABY database that successfully links 98% of maternal and newborn health records for in-hospital births. Other data sources are listed in eTable 1 in the Supplement.

For this study, we considered singleton children born in Ontario hospitals between April 1, 2002, and March 31, 2010, whose mothers were between the ages of 16 and 50 years and eligible for public drug benefits during pregnancy.11 Date of conception was estimated using the delivery date and gestational age at birth; in most cases (approximately 70%), gestational age was based on first-trimester ultrasound.12 We excluded children born to non-Ontario residents, those without a valid health card number (<1%), and those who died before the age of 2 years. The Institute for Clinical Evaluative Sciences is a prescribed entity under §45 of Ontario’s Personal Health Information Privacy Act; consent is not required for use of personal data. The Sunnybrook Health Sciences Centre granted ethics approval for this study.

Exposure Status

Serotonergic antidepressant exposure during pregnancy was defined as 2 or more consecutive prescriptions for an SSRI or selective norepinephrine reuptake inhibitor medication filled between conception and delivery (eTable 2 in the Supplement). A prescription was deemed consecutive when the second prescription was filled within 1.5 times the number of days’ supply of the first.11

Pregnancies were considered exposed during a specific trimester if 1 or more prescriptions were filled during that trimester, or if the prescription duration overlapped with that trimester. The unexposed (reference) group comprised women with no serotonergic antidepressants prescribed during pregnancy or within 90 days prior to conception. Within each group, 1 birth per woman was randomly selected to contribute to the analyses. If there was record of only 1 serotonergic antidepressant prescription during pregnancy, the pregnancy was excluded to avoid exposure misclassification.

Outcome

We followed up children to March 31, 2014, to identify autism spectrum disorder, which was defined as 2 or more outpatient diagnoses by either a pediatrician or psychiatrist, 1 or more diagnoses in hospital databases after the age of 2 years, or both (eTable 2 in the Supplement). A similar definition using US insurance data had a positive predictive value of 87.4%.13 Follow-up was to a minimum age of 4 years and a maximum age of 10 years (median age of autism spectrum disorder diagnosis is approximately 4.6 years).14

Covariates

We generated the following set of baseline characteristics on which to assess balance between exposed and unexposed groups: maternal age, parity, neighborhood income quintile, rural residence, medical and psychiatric diagnoses, health service use before and during pregnancy,15-17 use of other prescribed medications, prenatal care, child sex, and gestational age at birth. We used inverse probability of treatment weighting based on a high-dimensional propensity score (HDPS) in an attempt to balance exposure group differences.18 The HDPS is an approach to confounding used with health administrative data that relies on the principle that by measuring a battery of surrogate variables (diagnoses, procedures, drug claims), the likelihood increases that in combination they could behave as a good overall proxy for relevant unobserved confounders.19,20

We ran HDPS algorithms using the following data dimensions captured during the 365 days prior to conception: hospital (medical diagnoses, procedures, psychiatric diagnoses; 3 dimensions), emergency department diagnoses and procedures (2 dimensions), outpatient fee codes and diagnoses (2 dimensions), and prescription drug claims (1 dimension). Within each dimension, the top 200 most prevalent codes were each converted into 3 binary empirical covariates according to how often a code was received (once, sporadic, frequent). Covariates were selected based on associations between exposure and confounder and between confounder and outcome.

The relative risk of being exposed was calculated for each covariate. The top 500 covariates were used in the HDPS estimation (eTable 3 in the Supplement), along with a variable denoting the presence or absence of a woman’s diagnosis of a mood or anxiety disorder within the 2 years prior to conception and the following prespecified covariates not captured in these data dimensions: maternal age, parity, calendar year of delivery, neighborhood income quintile, rural residence, child sex, and gestational age at birth.21-23 We used HDPS scores to create inverse probability of treatment weights for each individual.

When performing inverse probability of treatment weighting, weights are applied to exposed and unexposed individuals to create a pseudopopulation in which groups are more balanced with respect to confounder distribution.24 We weighted serotonergic antidepressant users by the inverse of the HDPS and nonusers by 1 minus the inverse of the HDPS. Individuals with an unexpected treatment were given an increased weight and those who received expected treatment were given a decreased weight. Weights were stabilized to improve precision.24

Main and Other Analyses

Baseline characteristics were used to describe the cohort and to examine balance between the exposed and unexposed groups with respect to potential confounders before and after applying the inverse probability of treatment weights.15 Standardized differences were used to assess balance. The standardized difference describes between-group differences in units of standard deviation and is not influenced by sample size (and thus may be a better alternative to use of the P value in large cohorts). Standardized differences greater than 0.10 were considered clinically meaningful.15

For outcome analyses, we used Cox proportional hazards regression to estimate hazard ratios (HRs) and 95% CIs using a robust sandwich-type estimator.25 The child’s age was the underlying time scale, with estimation of person-time starting at the age of 2 years. Observations were censored if the child died or moved out of the province before the end of follow-up. We compared exposed children with unexposed children in crude and inverse probability of treatment-weighted analyses based on the HDPS. We also generated a multivariable model not using the HDPS that was adjusted for baseline variables with a standardized difference of greater than 0.10 between groups to provide a comparison with the results of the main analysis. Subanalyses were conducted by class and type of antidepressant, by trimester of antidepressant exposure, and in a cohort restricted to women with a documented diagnosis of a mood or anxiety disorder within 2 years prior to conception (eTable 2 in the Supplement).

To control for shared genetic risk in an additional analysis,26 we compared risk between siblings, in which one was exposed to a serotonergic antidepressant in utero and the other was not exposed. For this analysis, we considered all women with 2 or more live singleton infants born between April 1, 1997, and March 31, 2010. We selected births discordant for in utero serotonergic antidepressant exposure in a 1:1 ratio. We selected the closest consecutive sibling to ensure that the other characteristics were as similar as possible.

We similarly used Cox proportional hazards regression to estimate the HR and 95% CI, with the child’s age as the underlying time scale, and the same censoring criteria as above, and only controlled for maternal age, parity, and calendar year of delivery.27 The HDPS methods were not used for this analysis. Because missing data were uncommon (0.27% for neighborhood income quintile, 0.04% for rurality, and <6 cases for parity), we used a complete case analysis in regression models.

Sensitivity Analyses

We conducted 2 sensitivity analyses. First, because the inverse probability of treatment weighting can be sensitive to large size differences between exposure groups, we conducted a sensitivity analysis matching 1:1 on the HDPS to test the robustness of the findings in an analysis that retained the original size of the exposed group. Individuals for whom a match was not found were excluded from this analysis. Second, outcome risk among children of women prescribed serotonergic antidepressants 90 to 365 days prior to but not during pregnancy (recent users) was compared with risk among children in the unexposed group.

Analyses were conducted using SAS version 9.2 (SAS Institute Inc).

Results

Of 1.1 million deliveries among women between the ages of 16 and 50 years whose children were alive and eligible for follow-up at the age of 2 years, there were 45 985 deliveries to women covered under the public prescription drug plan (4.2% of deliveries). We excluded 1501 deliveries to women who filled only 1 prescription for a serotonergic antidepressant during pregnancy and 2210 deliveries to women who filled a prescription within the 90 days prior to pregnancy (but not during pregnancy), leaving a total of 42 274 deliveries among 35 906 unique women. The final cohort included 35 906 singleton children born at a mean gestational age of 38.7 weeks; 18 230 (50.4%) were male. Mothers were a mean age of 26.7 years and 2837 (7.9%) were serotonergic antidepressant users during pregnancy. Characteristics of women excluded because they were not eligible for Ontario’s publicly funded drug plan appear in eTable 4 in the Supplement.

Serotonergic antidepressant users and nonusers differed in their baseline characteristics, with users having a more severe psychiatric history (Table 1). After weighting, small- to medium-sized differences remained with respect to psychiatric diagnoses and psychiatric emergency department visits during pregnancy, a possible indicator of illness severity. This was also true in the subgroup restricted to women with mood and anxiety disorders (eTable 5 in the Supplement). In the sibling analysis, 620 pairs of siblings discordant for serotonergic antidepressant exposure were successfully identified; several differences in maternal characteristics were observed at the time of the exposed vs unexposed pregnancy (eTable 6 in the Supplement).

Primary Analysis

In the overall cohort, 1.1% of children (95% CI, 1.0%-1.2%) were diagnosed with autism spectrum disorder by the end of follow-up (mean duration of follow-up, 4.95 years). In the 2837 pregnancies (7.9%) exposed to antidepressants, 2.0% (95% CI, 1.6%-2.6%) of children were diagnosed with autism spectrum disorder. Risk of autism spectrum disorder was significantly higher with serotonergic antidepressant exposure (4.51 exposed vs 2.03 unexposed per 1000 person-years; between-group difference, 2.48 [95% CI, 2.33-2.62] per 1000 person-years) in crude (HR, 2.16 [95% CI, 1.64-2.86]) and multivariable-adjusted analyses (HR, 1.59 [95% CI, 1.17-2.17]) (Table 2). After inverse probability of treatment weighting based on the HDPS, the association was not significant (HR, 1.61 [95% CI, 0.997-2.59]) (Table 2).

In subanalyses by trimester of exposure and antidepressant class and type, serotonergic antidepressant exposure was significantly associated with the outcome in crude analyses, but was not significant after inverse probability of treatment weighting (Figure 1 and Figure 2). In contrast, in the subgroup restricted to women with mood and anxiety disorders, the association remained significant after inverse probability of treatment weighting overall (HR, 1.62 [95% CI, 1.04-2.52]; eTable 7 in the Supplement), for SSRIs as a class (HR, 1.78 [95% CI, 1.07-2.95]), and for second or third trimester antidepressant exposure (HR, 1.94 [95% CI, 1.20-3.15]; eFigure 1 and eFigure 2 in the Supplement).

Risk of autism spectrum disorder did not differ significantly between exposed and unexposed siblings in the crude analyses (incidence: 3.40 vs 2.05 per 1000 person-years, respectively; between-group difference, 1.35 [95% CI, −0.84 to 3.54] per 1000 person-years; HR, 1.51 [95% CI, 0.63 to 3.60]) or after adjustment for maternal age, parity, and calendar year of delivery (adjusted HR, 1.60 [95% CI, 0.69 to 3.74]; Table 2).

Sensitivity Analyses

In the HDPS-matched sensitivity analysis, there was a significantly higher risk for autism spectrum disorder in the exposed group (4.48 per 1000 person-years vs 2.73 per 1000 person-years in the unexposed group; HR, 1.64 [95% CI, 1.07-2.53]). Children of women prescribed serotonergic antidepressants 90 to 365 days prior to but not during pregnancy (recent users; eTable 8 in the Supplement) were at higher risk of autism spectrum disorder compared with children in the unexposed group whose mothers were not recent users (4.06 vs 1.83 per 1000 person-years, respectively; adjusted HR, 1.85 [95% CI, 1.37-2.51]; Table 2).

Discussion

In this retrospective cohort study, children exposed to serotonergic antidepressants were at higher risk for autism spectrum disorder compared with unexposed children prior to confounder adjustment. After inverse probability of treatment weighting using HDPS, the difference in outcome risk between exposed and unexposed children was no longer statistically significant. Even though this finding was not robust to all subanalyses and sensitivity analyses, when antidepressant-exposed children were compared with their unexposed siblings, there was no observed association between in utero antidepressant exposure and risk of autism spectrum disorder.

Furthermore, among children whose mothers were prescribed antidepressants shortly before but not during pregnancy, outcome risk was as high compared with unexposed children as it was for exposed children. Taken together, attenuation of the association with weighting and the results of the sibling and recent user analysis suggest that confounding by indication for the medication may explain previously observed associations between in utero serotonergic antidepressant exposure and autism spectrum disorder.

Crude and multivariable-adjusted findings in the current study are similar in both magnitude and direction to those of several previous observational studies that largely showed statistically significant associations; however, those studies had variable control for confounding.28-31 The nonsignificant associations observed in the inverse probability of treatment-weighted analysis and sibling analysis of the current study are aligned with the results reported in 2 large Danish cohort studies. One of these studies found that serotonergic antidepressant exposure was associated with child autism spectrum disorder risk in the main analysis (adjusted HR, 1.5; 95% CI, 1.2-1.9), but not in an analysis restricted to maternal mood disorder (adjusted HR, 1.2; 95% CI, 0.7-2.1), nor in a sibling analysis (adjusted HR, 1.1; 95% CI, 0.5-2.3).32

In the second study, analyses controlling for a large number of maternal characteristics were not significant (adjusted rate ratio, 1.20; 95% CI, 0.90-1.61).10 Furthermore, use of antidepressants during the 6 to 24 months prior to but not during pregnancy was significantly associated with autism spectrum disorder risk (adjusted rate ratio, 1.46; 95% CI, 1.17-1.81). These findings, along with those of the current study, lend support to the hypothesis that the indication for the medication (ie, depression or anxiety) may explain observed increased risks for autism spectrum disorder and not the use of antidepressant medication.33,34

To our knowledge, the current study is the first to use HDPS to try to balance exposure group differences in an evaluation of the association between in utero serotonergic antidepressant exposure and autism spectrum disorder. Although baseline characteristics were fairly balanced after inverse probability of treatment weighting using HDPS, imbalance remained with respect to specific psychiatric diagnoses and psychiatric emergency department visits, suggesting that the HDPS did not completely balance the groups on the potentially important confounders of psychiatric illness diagnosis and severity.35,36 This may explain the significant association between antidepressant exposure and autism spectrum disorder that was observed in the HDPS-matched sensitivity analysis and some of the subanalyses.

Even though HDPS approaches are a sophisticated method for attempting to account for confounders, no statistical technique can improve comparability of groups that do not sufficiently overlap in confounder distribution, nor can it recover information about variables not contained in the data set by themselves or through proxy representation.37 Furthermore, it is possible that the HDPS missed 1 or more key low-prevalence covariates because the semiautomated selection method first selects prevalent codes before assessing for associations between covariate and exposure or covariate and outcome.38

The sibling and recent user analyses reported herein, by the nature of their designs, may have been better able to account for confounding by the underlying indication compared with the HDPS approach. Future studies can draw on this and similar studies to refine observational methods further to disentangle the roles of serotonergic antidepressants and maternal mental illness in their associations with child autism spectrum disorder and other important outcomes.

Limitations

This study has several limitations common to research using administrative data. It was limited to women in Ontario’s publicly funded drug plan. These women had lower socioeconomic status and were more likely disabled than women in the broader population, so caution should be used in making generalizations about overall autism spectrum disorder prevalence estimates. Exposure status may have been misclassified if prescriptions were filled but not used, creating a bias toward the null. The outcome definition was restricted in outpatient data to diagnoses made by a pediatrician or a psychiatrist to improve specificity, so cases diagnosed by family physicians or psychologists may have been missed. However, the autism spectrum disorder prevalence of more than 1% is consistent with current estimates, suggesting that few cases were missed.

There were no data on paternity or postnatal environment characteristics, both of which may be associated with autism spectrum disorder risk.26,39 Furthermore, when the HDPS methods are applied by 2 or more investigators working with different data sets in which patterns of associations are otherwise similar but prevalence of covariates or granularity of coding differs (eg, International Classification of Diseases codes vs another coding system), a different set of covariates could be selected for the propensity score. This could affect reproducibility across studies.

Conclusions

In children born to mothers receiving public drug coverage in Ontario, Canada, in utero serotonergic antidepressant exposure compared with no exposure was not associated with autism spectrum disorder in the child. Although a causal relationship cannot be ruled out, the previously observed association may be explained by other factors.

Back to top
Article Information

Corresponding Author: Simone N. Vigod, MD, MSc, FRCPC, Women’s College Hospital, 76 Grenville St, Toronto, ON M5S 1B2, Canada (simone.vigod@wchospital.ca).

Author Contributions: Drs Brown and Vigod 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: All authors.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Brown, Vigod.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Brown, Ray, Wilton, Gomes.

Supervision: Vigod.

Conflict of Interest Disclosures: The authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: This study was supported by the Institute for Clinical Evaluative Sciences (ICES), which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC).

Role of the Funder/Sponsor: Neither the ICES nor the Ontario MOHLTC had a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Disclaimer: The opinions, results, and conclusions reported in this article are those of the authors and are independent from the funding sources. No endorsement by the ICES or the Ontario MOHLTC is intended or should be inferred. Parts of this material are based on data and information compiled and provided by the Canadian Institute for Health Information (CIHI). However, the analyses, conclusions, opinions, and statements expressed herein are those of the authors, and not necessarily those of the CIHI.

References
1.
Le Strat  Y, Dubertret  C, Le Foll  B.  Prevalence and correlates of major depressive episode in pregnant and postpartum women in the United States.  J Affect Disord. 2011;135(1-3):128-138.PubMedGoogle ScholarCrossref
2.
Stein  A, Pearson  RM, Goodman  SH,  et al.  Effects of perinatal mental disorders on the fetus and child.  Lancet. 2014;384(9956):1800-1819.PubMedGoogle ScholarCrossref
3.
Howard  LM, Molyneaux  E, Dennis  CL,  et al.  Non-psychotic mental disorders in the perinatal period.  Lancet. 2014;384(9956):1775-1788.PubMedGoogle ScholarCrossref
4.
Fournier  JC, DeRubeis  RJ, Hollon  SD,  et al.  Antidepressant drug effects and depression severity.  JAMA. 2010;303(1):47-53.PubMedGoogle ScholarCrossref
5.
Andrade  SE, Reichman  ME, Mott  K,  et al.  Use of selective serotonin reuptake inhibitors (SSRIs) in women delivering liveborn infants and other women of child-bearing age within the US Food and Drug Administration’s Mini-Sentinel program.  Arch Womens Ment Health. 2016;19(6):969-977.PubMedGoogle ScholarCrossref
6.
American Psychiatric Association.  Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Arlington, VA: American Psychiatric Publishing; 2013.
7.
Harrington  RA, Lee  LC, Crum  RM,  et al.  Serotonin hypothesis of autism: implications for selective serotonin reuptake inhibitor use during pregnancy.  Autism Res. 2013;6(3):149-168.PubMedGoogle ScholarCrossref
8.
Borue  X, Chen  J, Condron  BG.  Developmental effects of SSRIs.  Int J Dev Neurosci. 2007;25(6):341-347.PubMedGoogle ScholarCrossref
9.
Brown  HK, Hussain-Shamsy  N, Lunsky  Y,  et al.  The association between antental exposure to selective serotonin reuptake inhibitors and autism.  J Clin Psychiatry. 2017;78(1):e48-e58.PubMedGoogle ScholarCrossref
10.
Hviid  A, Melbye  M, Pasternak  B.  Use of selective serotonin reuptake inhibitors during pregnancy and risk of autism.  N Engl J Med. 2013;369(25):2406-2415.PubMedGoogle ScholarCrossref
11.
Vigod  SN, Gomes  T, Wilton  AS,  et al.  Antipsychotic drug use in pregnancy.  BMJ. 2015;350:h2298.PubMedGoogle ScholarCrossref
12.
Ray  JG, Schull  MJ, Urquia  ML,  et al.  Major radiodiagnostic imaging in pregnancy and the risk of childhood malignancy.  PLoS Med. 2010;7(9):e1000337.PubMedGoogle ScholarCrossref
13.
Burke  JP, Jain  A, Yang  W,  et al.  Does a claims diagnosis of autism mean a true case?  Autism. 2014;18(3):321-330.PubMedGoogle ScholarCrossref
14.
Frenette  P, Dodds  L, MacPherson  K,  et al.  Factors affecting the age at diagnosis of autism spectrum disorders in Nova Scotia, Canada.  Autism. 2013;17(2):184-195.PubMedGoogle ScholarCrossref
15.
Austin  PC.  Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples.  Stat Med. 2009;28(25):3083-3107.PubMedGoogle ScholarCrossref
16.
Hux  JE, Ivis  F, Flintoft  V, Bica  A.  Diabetes in Ontario.  Diabetes Care. 2002;25(3):512-516.PubMedGoogle ScholarCrossref
17.
Quan  H, Khan  N, Hemmelgarn  BR,  et al.  Validation of a case definition to define hypertension using administrative data.  Hypertension. 2009;54(6):1423-1428.PubMedGoogle ScholarCrossref
18.
Parkerson  GR  Jr, Harrell  FE  Jr, Hammond  WE, Wang  XQ.  Characteristics of adult primary care patients as predictors of future health services charges.  Med Care. 2001;39(11):1170-1181.PubMedGoogle ScholarCrossref
19.
Schneeweiss  S, Rassen  JA, Glynn  RJ,  et al.  High-dimensional propensity score adjustment in studies of treatment effects using health care claims data.  Epidemiology. 2009;20(4):512-522.PubMedGoogle ScholarCrossref
20.
Hernán  MA, Robins  JM.  Estimating causal effects from epidemiological data.  J Epidemiol Community Health. 2006;60(7):578-586.PubMedGoogle ScholarCrossref
21.
Guertin  JR, Rahme  E, Dormuth  CR, LeLorier  J.  Head to head comparison of the propensity score and the high-dimensional propensity score matching methods.  BMC Med Res Methodol. 2016;16:22.PubMedGoogle ScholarCrossref
22.
Enders  D, Ohlmeier  C, Garbe  E.  The potential of high-dimensional propensity scores in health services research: an exemplary study on the quality of care for elective percutaneous coronary interventions [published online January 16, 2017].  Health Serv Res. doi:10.1111/1475-6773.12653PubMedGoogle Scholar
23.
Schneeweiss  S, Eddings  W, Glynn  RJ,  et al.  Variable selection for confounding adjustment in high-dimensional covariate spaces when analyzing healthcare databases.  Epidemiology. 2017;28(2):237-248.PubMedGoogle ScholarCrossref
24.
Austin  PC, Stuart  EA.  Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies.  Stat Med. 2015;34(28):3661-3679.PubMedGoogle ScholarCrossref
25.
Austin  PC.  Variance estimation when using inverse probability of treatment weighting (IPTW) with survival analysis.  Stat Med. 2016;35(30):5642-5655.PubMedGoogle ScholarCrossref
26.
Donovan  SJ, Susser  E.  Advent of sibling designs.  Int J Epidemiol. 2011;40(2):345-349.PubMedGoogle ScholarCrossref
27.
Idring  S, Magnusson  C, Lundberg  M,  et al.  Parental age and the risk of autism spectrum disorders.  Int J Epidemiol. 2014;43(1):107-115.PubMedGoogle ScholarCrossref
28.
Croen  LA, Grether  JK, Yoshida  CK,  et al.  Antidepressant use during pregnancy and childhood autism spectrum disorders.  Arch Gen Psychiatry. 2011;68(11):1104-1112.PubMedGoogle ScholarCrossref
29.
Harrington  RA, Lee  LC, Crum  RM,  et al.  Prenatal SSRI use and offspring with autism spectrum disorder or developmental delay.  Pediatrics. 2014;133(5):e1241-e1248.PubMedGoogle ScholarCrossref
30.
Gidaya  NB, Lee  BK, Burstyn  I,  et al.  In utero exposure to selective serotonin reuptake inhibitors and risk for autism spectrum disorder.  J Autism Dev Disord. 2014;44(10):2558-2567.PubMedGoogle ScholarCrossref
31.
Boukhris  T, Sheehy  O, Mottron  L, Bérard  A.  Antidepressant use during pregnancy and the risk of autism spectrum disorder in children.  JAMA Pediatr. 2016;170(2):117-124.PubMedGoogle ScholarCrossref
32.
Sørensen  MJ, Grønborg  TK, Christensen  J,  et al.  Antidepressant exposure in pregnancy and risk of autism spectrum disorders.  Clin Epidemiol. 2013;5:449-459.PubMedGoogle ScholarCrossref
33.
Kyriacou  DN, Lewis  RJ.  Confounding by indication in clinical research.  JAMA. 2016;316(17):1818-1819.PubMedGoogle ScholarCrossref
34.
Cross-Disorder Group of the Psychiatric Genomics Consortium.  Identification of risk loci with shared effects on five major psychiatric disorders.  Lancet. 2013;381(9875):1371-1379.PubMedGoogle ScholarCrossref
35.
Oberlander  TF, Vigod  SN.  Developmental effects of prenatal selective serotonin reuptake inhibitor exposure in perspective.  J Am Acad Child Adolesc Psychiatry. 2016;55(5):351-352.PubMedGoogle ScholarCrossref
36.
Jokiranta-Olkoniemi  E, Cheslack-Postava  K, Sucksdorff  D,  et al.  Risk of psychiatric and neurodevelopmental disorders among siblings of probands with autism spectrum disorders.  JAMA Psychiatry. 2016;73(6):622-629.PubMedGoogle ScholarCrossref
37.
Franklin  JM, Eddings  W, Glynn  RJ, Schneeweiss  S.  Regularized regression versus the high-dimensional propensity score for confounding adjustment in secondary database analyses.  Am J Epidemiol. 2015;182(7):651-659.PubMedGoogle ScholarCrossref
38.
Robins  JM.  Data, design, and background knowledge in etiologic inference.  Epidemiology. 2001;12(3):313-320.PubMedGoogle ScholarCrossref
39.
Chaste  P, Leboyer  M.  Autism risk factors: genes, environment, and gene-environment interactions.  Dialogues Clin Neurosci. 2012;14(3):281-292.PubMedGoogle Scholar
×