Preschool outcomes following prenatal antidepressant exposure: Differences in language and behavior, but not cognitive function

Abstract

OBJECTIVE

To test the hypothesis that prenatal exposure to serotonin reuptake inhibitors (SRIs) would be associated with language and behavioral outcomes in preschool-aged children, while accounting for confounds such as concomitant exposures and maternal mental illness.

METHOD

An observational, prospective, longitudinal study of mental illness in pregnancy was conducted at a university-based women’s mental health clinic (April 2010–November 2012). A sample of 178 mother-child dyads participated in a laboratory visit at preschool age (2.5–5.5 years). The majority of women (93%) received psychotropic medication during pregnancy. Psychiatric status (based on DSM-IV), other medication use and substance use were serially assessed and tested as confounds. Primary outcome measures included standardized measures of expressive language and cognitive function, and mother and alternate caregiver ratings of child behavior problems, including a Persavive Developmental Disorders (PDD) subscale.

RESULTS

Linear regression analyses revealed that, after controlling for relevant covariates, expressive language scores from the Test of Early Language Development-3^rd edition were negatively associated with prenatal SRI exposure (β=−0.15, t=−2.41), while the PDD behavioral problems subscales completed by alternate caregivers and mothers were positively associated with prenatal SRI exposure (β=0.17, t=2.01; β=0.16, t=2.00). Cognitive function, measured using the Differential Ability Scales-2^nd edition, was not associated with any medication exposures.

CONCLUSION

The current data suggest a small but significant association between prenatal SRI exposure and preschool outcomes, including expressive language and behavior problems. These data corroborate data from recent, population-based studies; though overall, published findings are mixed. Replication and identification of moderating risk factors are needed to understand potential clinical implications.

Introduction

The recent increase in pharmacological treatment for mood disorders during pregnancy^1–3 highlights the need for careful scrutiny of outcomes. Common treatment options include selective serotonin reuptake inhibitors (SSRIs) and other drugs such as venlafaxine and clomipramine, collectively referred to as serotonin reuptake inhibitors (SRIs)^{1, 4}. Human research investigating the effects of prenatal SRI exposure on infant and child development has produced mixed results, with several studies reporting negative findings^5–11. In a recent review¹², only two of 12 included studies reported significant effects of SRI exposure on neurodevelopmental outcomes, and most focused on infants.

A Kaiser Permanente sample¹³ noted a two-fold increase in maternal prenatal SSRI use (and a four-fold increase in maternal first trimester SSRI use) in children diagnosed with an ASD compared to non-diagnosed controls. Dual-action antidepressants were was not associated with an increased risk of ASD, though only ten women were prescribed those, and thus analyses were likely underpowered¹⁴. Small but significant differences in sample characteristics were noted and statistically controlled for. Two additional recent cohort studies also reported links between prenatal antidepressant use and risk of ASDs^{15, 16}, while in a recent Danish cohort study, the association between antidepressant exposure and ASDs was non-significant once the analyses were limited to mothers with an affective disorder¹⁷. Potentially related, prenatal antidepressant exposure was associated with altered developmental trajectories of infant speech perception, which authors warn may be an indicator of future communication difficulties¹⁸.

Limitations across the majority of investigations include reliance on maternal report for perinatal exposures, and mostly retrospective tracking of maternal prenatal symptoms, which our group has demonstrated to produce a recall bias resulting in an under-reporting of prenatal depression¹⁹. Further, most of these studies did not compare potential developmental effects of SRI exposure to those associated with exposure to maternal depressive symptoms. The current study sought to address these gaps in the literature using a prospectively characterized cohort of mother-child dyads to investigate potential associations between prenatal depressive symptoms, level of exposure to SRIs during gestation, and language and social behavioral outcomes measured in preschool-aged children. Specifically, we attempt to replicate recent findings^{13, 15, 18} that maternal SRI use during pregnancy will increase risk for language deficits as well as autism spectrum behaviors.

Methods

Participants

Mother-child dyads (N=178) were recruited from a larger sample of 229 women enrolled in an observational, prospective, longitudunal study at the Emory Women's Mental Health Program (WMHP), specializing in treatment of perinatal mental illness. Women were assessed at multiple timepoints throughout pregnancy using self-report and clinician-administered measures of psychiatric symptoms and health, urine toxicology, thyroid function, and weekly documentation of all exposures in pregnancy (prescription, over-the-counter, and environmental)^20–25. To be recruited for the current study, mothers had to participate in the prospective study above, and their children needed to fall in the specific age range. Women with active substance abuse during pregnancy were excluded, and children with identified congenital malformations or serious medical conditions were also excluded.

Children falling between 2.5–5.5 years of age and their mothers were assessed during a 2.5 hour lab visit (April 2010–November 2012). Following informed consent, maternal interviews and questionnaires were administered. Child evaluations included a developmental assessment of motor, language, and cognitive functioning. The study was approved by the Emory University Institutional Review Board.

Measures

Prenatal Measures of Psychiatric Illness and Mood Symptoms

The Structured Clinical Interview for DSM-IV Disorders (SCID)²⁶ is a widely-validated semi-structured research interview with strong psychometric properties used to assess current and lifetime history of DSM-IV Axis I disorders ²⁷. The SCID was administered by a trained research interviewer at enrollment. At each subsequent prenatal visit, the mood modules from the SCID were administered. The number of visits in which a woman met criteria for mood episode divided by the number of times she was assessed in pregnancy, was used as a covariate in the current analyses, reflecting the degree to which the mother experienced mood episodes (accounting for her frequency of visits). The Global Assessment of Functioning (GAF), which reflects the clinician's rating of the patients worst level of functioning in the month prior to assessment, was recorded at each visit. The Beck Depression Inventory (BDI)²⁸ is a self-report measure of depressive symptoms administered at each visit. For women with at least three prenatal visits, area under the curve measures were calculated for GAF and BDI scores.

Use of psychotropic medications, in addition to alcohol, tobacco and caffeine was assessed using a detailed weekly tracking sheet completed by a study physician. Medications were grouped by class: antidepressants-serotonin reuptake inhibitors (SRIs; e.g., citalopram, fluoxetine, sertraline, venlafaxine), antidepressants-bupropion, mood stabilizers, including anti-epileptics (e.g., lamotrigine, carbamazepine), atypical antipsychotics (e.g., risperidone, quetiapine, olanzapine), anxiolytics (e.g., alprazolam, lorazepam, diazepam), hypnotics (e.g., zolpidem, trazodone, diphenhydramine). Women's medication regimen with respect to dosage, timing and number of medications was adjusted based on clinical symptoms and tolerability. Medication exposure was estimated by multiplying the number of medications within a given class by the number of weeks taken during pregnancy. The primary predictor in the current study was SRI DrugWeeks (i.e., number of SRI medications X number of weeks taken, standardized to 40 weeks gestation). DrugWeeks of exposure to other medication classes were tested as potential confounds using the same approach (number of drugs in a given class X number of weeks exposed, standardized to 40 weeks gestation)²⁹.

Child Measures

Standardized tests and caregiver report measures were administered. The Expressive Language subtest of the Test of Early Language Development, 3rd edition (TELD-3)³⁰ assesses the child's ability to provide personal information, answer open-ended questions, and use age-appropriate expressive language skills including grammar, syntax, and vocabulary. The TELD-3 can be administed to children ages 2 years, 0 months to 7 years, 11 months. The Differential Ability Scales, 2nd edition (DAS-II)³¹ is a normed measure of current cognitive functioning administered to children aged 2 years, 6 months to age 17 years, 11 months. The TELD-3 and DAS-II were administed by graduate students blind to maternal medication and psychiatric status.

The Child Behavior Checklist (CBCL)³² measures the frequency of common behavior problems in children aged 1.5–5 years. The Pervasive Developmental Disorder subscale includes 13 items assessing behaviors such as being afraid of new things, avoiding eye contact, not getting along with others, speech problems, and having difficulty when things are out of order. Although the CBCL is not used as a diagnostic tool, children scoring high on the PDD subscale are more likely to evidence behaviors commonly associated with ASD. The child's mother and an alternate caregiver (e.g., father, babysitter, teacher, grandmother) completed the CBCL; 140 out of 178 (79%) of the alternate caregiver CBCLs were returned.

Maternal Measures

At the preschool follow-up, SCIDs were administered by one clinical psychologist and three doctoral students. Inter-rater reliability based on 15% of the sample were adequate for the most common diagnoses (Κs=0.80–1.00).

Potential Confounds

Sociodemographic variables were assessed at the preschool follow-up included mother and child age, child gender, mother's highest level of education, mother's marital status, the number of adults and children living in the home, whether the mother worked outside of the home and the average number of hours worked per week, and whether the child attended a structured early learning program (e.g., daycare, preschool). Area under the curve measures were derived for the following prenatal exposures: alcohol, illicit drugs (marijuana), tobacco, and caffeine, as well as additional psychotropic medication classes: bupropion, mood-stabilizers, including anti-epileptics and atypical antipsychotics, anxiolytics, and hypnotics. Postpartum psychotropic medication exposure via lactation was coded as a binary (yes/no) variable, collected during postpartum follow-up visits from 0–6 months. Labor and delivery records for all mother-child dyads were obtained and extracted to yield the following variables: birth weight, head circumference, estimated gestational age at delivery, delivery method (vaginal vs. Cesarean-section), Apgar scores at one and five minutes, number of delivery complications, and number of pregnancy complications. The prenatal measures of depressive symptoms, mood episodes, and global functioning, described above, as well as maternal epileptic status, were also tested as potential confounds.

Statistical Analyses

The primary predictor in the current study was SRI DrugWeeks. Dependent measures were the 1) Expressive Language standard score from the TELD-3, 2) overall cognitive functioning composite from the DAS-II, 3) PDD subscale completed by mothers and 4) PDD subscale completed by alternate caregivers. The potential confounds were tested using Pearson correlations, t-test and chi-square tests. Hierarchical linear regression was used to evaluate study hypotheses that prenatal SRI exposure would be associated with increased PDD subscale scores and decreased TELD-3 scores. A separate regression model was used for each dependent variable; covariates that were significantly related to that dependent variable were included in block 1 of the regression model, the predictor SRI DrugWeeks was included in step 2 of the model. Post hoc analyses used logistic regression, partial correlations, and analysis of covariance to further evaluate significant associations. Generalized linear mixed models were incorporated to account for potential effects of 18 sibling pairs in the data.

The PDD subscales completed by both raters were skewed and thus log transformed for all analyses. DAS Congitive and TELD-3 Expressive Language scores were normally distributed.

Results

Sample Characteristics

The sample of women who participated in the preschool follow-up study showed no differences from those who passively or actively refused participation (n=51) with respect to age at pregnancy (t_1,227=0.17, p=0.86), highest level of education attained (t_{1, 227}=−0.65, p=0.52), gestational age of the target child at birth (t_1,227=−0.96, p=0.34), or method of delivery (i.e., vaginal vs. C-section; Χ²_{3, 229} =5.09, p=0.17).

Mothers participating in the preschool follow-up study had a mean age of 37.2 years (SD=4.7). Their median education level was college graduate, and approximately 61% of the mothers worked outside of the home (number of hours ranged from 2–70, M=20.0, SD=19.6). Approximately 84% of the women were married or cohabitating with a partner, and another 8% of the sample lived with another adult (e.g., parent); thus about 8% of the sample resided in a single-adult household. Most of the women received psychotropic medication at some point during pregnancy (n=155; 87%), and 93% of women (n=165) had a lifetime history of at least one Axis I disorder. BDI scores at the preschool visit ranged from 0–53 (Mean=6.9, SD=8.3).Children participating in the current study ranged in age from 30–65 months (M=45.2, SD=11.1 months), and 51.7% were males.

Table 1 compares sample characteristics between SRI-exposed and unexposed children and describes the type of SRI-exposures for the sample. Table 2 describes statistically significant associations between potential covariates and primary outcomes of interest. Of note, other psychotropic exposures (bupropion, mood stabilizers, anxiolytics, and hypnotics) were not significantly associated with outcome variables.

Table 1.

Description of Sample by Exposure Group

	Assessed in Current Study (N = 178)	Not Exposed to SRIs (n = 76)	Exposed to SRIs (n = 102)
Child age, mean (SD), mo	45.2 (11.2)	45.1 (12.1)	45.2 (10.4)
Child gender, % male	52	55	49
Prenatal BDI score,a mean (SD)	10.0 (7.9)	8.3 (6.3)	11.3 (8.8)
Preschool BDI score,a mean (SD)	6.9 (8.3)	5.0 (5.2)	8.3 (9.8)
Prenatal BDI < 15, n (%)	88 (54.0)	45 (67.2)	43 (44.8)
SRI exposure type,b n
Citalopram			15
Escitalopram			14
Fluoxetine			21
Paroxetine			7
Sertraline			26
Desvenlafaxine			1
Duloxetine			2
Venlafaxine			20

^aDenotes statistical difference between groups, P <.05.

^bOther antidepressant exposures: bupropion = 27, mirtazapine = 3; some women were exposed to multiple antidepressants.

Abbreviations: BDI = Beck Depression Inventory, SRI = serotonin reuptake inhibitor.

Table 2.

Multiple Linear Regression Results for TELD-3^a,^b

Independent Variable	β	Unstandardized B Coefficient	95% CI for B	P Value
Maternal educationc	0.118	1.187	−0.025 to 2.398	.055
Prenatal BDI AUC	0.023	0.001	−0.007 to 0.009	.818
Prenatal mood episodes	−0.083	−5.332	−17.590 to 6.926	.391
Prenatal caffeine	−0.043	−0.003	−0.011 to 0.005	.480
Prenatal alcohol	−0.029	−0.018	−0.103 to 0.068	.678
Prenatal tobacco	0.038	0.001	−0.003 to 0.006	.604
Gestational age at delivery	0.089	0.887	−0.308 to 2.081	.144
Apgar at 5 minutes	0.059	1.783	−1.792 to 5.357	.326
Child age (months)	−0.509	−0.693	−0.862 to −0.525	<.001
DAS-II (cognitive ability)	0.556	0.550	0.422 to 0.679	<.001
Prenatal SRI (drug weeks)	−0.146	−0.105	−0.190 to −0.019	.017

^aTELD-3 score: mean = 100, SD = 15. Bolded values show significance at P < .05, t = −2.41.

^bThe following variables were tested as covariates and were not significantly associated with the TELD: mother's age; child's gender; mother's marital status; the number of adults and children living in the home; mother's work status; average number of hours mother worked per week; whether the child attended a structured early learning program; maternal depressive symptoms (BDI) at preschool visit; area under the curve measures for prenatal exposure to marijuana, bupropion, mood stabilizers, anxiolytics, and hypnotics; postpartum exposure to SRIs, bupropion, mood stabilizers, anxiolytics, and hypnotics; birth weight; head circumference; delivery method; number of delivery complications; number of pregnancy complications; maternal prenatal global assessment of functioning scores; and maternal epileptic status.

^cMaternal education was measured on an 6-point scale ranging from < 8th grade to completing graduate degree.

Abbreviations: AUC = area under the curve; BDI = Beck Depression Inventory; DAS-II = Differential Ability Scales, 2nd edition; SRI = serotonin reuptake inhibitor; TELD-3 = Test of Early Language Development, 3rd edition.

Expressive Language and Cognitive Functioning

Linear regression analyses revealed a negative association between SRI DrugWeeks and TELD-3 Expressive Language scores (Table 2). To assess magnitude of the effect, mean Expressive Language scores were compared between those children prenatally exposed to SRI medications versus unexposed. After controlling for significant confounds, ANCOVA revealed a modest mean difference of approximately 5 points in Expressive Language scores (F(1,144)=8.594, p=.004). Prenatal SRI exposure was not associated with the DAS-2 composite measure of cognitive functioning (r=.016, p=.833, n=175).

Pervasive Development Disorder Subscale Ratings

Linear regression analyses revealed a positive association between SRI DrugWeeks and PDD subscale completed by the alternate caregiver (Table 3). Consistently, prenatal SRI exposure was also positively associated with the PDD subscale completed by the mother (Table 4). PDD subscales scores rated by the mother and alternate caregiver were modestly correlated (r=0.29, p<.001, n=142).

Table 3.

Multiple Linear Regression Results for Alternate Caregiver-Rated PDD^a,^b

Independent Variable	β	Unstandardized B Coefficient	95% CI for B	P Value
Maternal epilepsy	−0.044	−0.028	−0.134 to 0.078	.607
No. of children in the home	−0.124	−0.015	−0.036 to 0.005	.146
Prenatal SRI (drug weeks)	0.171	0.001	< 0.001 to 0.002	.047

^aPDD subscale score reflects a T score with a mean of 50 and an SD of 10. Bolded values show significance at P < .05, t = 2.01.

^bThe following variables were tested as covariates and were not significantly associated with the PDD subscale: child's age; child's gender; mother's age; mother's marital status; number of adults in the home; mother's work status; average hours mother worked per week; whether the child attended a structured early learning program; maternal depressive symptoms (BDI) at preschool visit; area under the curve measures for prenatal exposure to marijuana, alcohol, caffeine, tobacco, bupropion, mood stabilizers, anxiolytics, and hypnotics; postpartum exposure to SRIs, bupropion, mood stabilizers, anxiolytics, and hypnotics; birth weight, Apgar scores; gestational age; head circumference; delivery method; number of delivery complications; number of pregnancy complications; and maternal prenatal measures of depressive symptoms, mood episodes, and global assessment of functioning.

Abbreviations: PDD = Pervasive Developmental Disorder subscale of the Child Behavior Checklist, SRI = serotonin reuptake inhibitor.

Table 4.

Multiple Linear Regression Results for Mother-Rated PDD Scores^a,^b

Independent Variable	β	Unstandardized B Coefficient	95% CI for B	P Value
Preschool BDI	0.036	< 0.001	−0.002 to 0.002	.646
Postpartum antidepressant (yes/no)	0.124	0.027	−0.008 to 0.061	.131
Prenatal tobacco	0.243	< 0.001	< 0.001 to < 0.001	.002
Prenatal SRI (drug weeks)	0.163	0.001	< 0.001 to 0.002	.048

^aPDD subscale score reflects a T score with a mean of 50 and an SD of 10. Bolded values show significance at P < .05, t = 2.00.

^bThe following variables were tested as covariates and were not significantly associated with the PDD: child's age; child's gender; mother's age; mother's marital status; number of adults and children living in the home; mother's work status; average hours mother worked per week; whether the child attended a structured early learning program; maternal depressive symptoms (BDI) at preschool visit; area under the curve measures for prenatal exposure to marijuana, alcohol, caffeine, bupropion, mood stabilizers, anxiolytics, and hypnotics; postpartum exposure to bupropion, mood stabilizers, anxiolytics, and hypnotics; birth weight; head circumference; Apgar scores; gestational age; delivery method; number of delivery complications; number of pregnancy complications; maternal prenatal measures of depressive symptoms, mood episodes, global assessment of functioning; and maternal epileptic status.

Abbreviations: BDI = Beck Depression Inventory, PDD = Pervasive Developmental Disorder subscale of the Child Behavior Checklist, SRI = serotonin reuptake inhibitor.

To facilitate interpretation of these effects, cutoff scores (T>65) were used in logistic regression models to test whether SRI DrugWeeks was associated with increased likelihood of clinically elevated PDD subscales. In clinical practice, scores above 65 are considered "At Risk." For mother-rated CBCLs, PDD scores for 15 children fell above the cut-off, and children with prenatal SRI exposure were statistically more likely to score in the PDD At Risk range (OR=1.05, 95% CI=1.01 to 1.08, p<0.02). However, prenatal SRI exposure was not associated with increased risk of elevated PDD scores rated by alternate caregivers (OR=1.01, (95% CI=0.98 to 1.05, p=0.52); 10 children fell in the at-risk range as rated by alternate caregivers.

Post Hoc Analyses

To examine whether non-independence of siblings (n=18 pairs) biased our results, generalized linear mixed models were run analogous to the linear regression models described above. This technique yielded a nested model with sibling pairs nested within families. Significant results held for associations between prenatal SRI exposure and Expressive Language scores, mother-rated PDD behaviors, and alternate caregiver-rated PDD behaviors (data not shown). In addition, ten women were diagnosed with a psychotic disorder (e.g., schizophrenia, schizoaffective disorder). Given that a family history of a psychotic disorder may act as an independent risk factor for ASDs³³ and therefore elevated PDD behaviors, we removed these mother-child dyads from the sample and re-executed the regression models. Results were mostly unchanged: Expressive Language (β=−.143, t=−2.320, p=.022, R²∆=.019), Mother-rated PDD behavior (β=.160, t=1.927, p=.05, R²∆=.021), Alternate caregiver-rated PDD behavior (β=.179, t=2.077, p=.040, R²∆=.031).

Discussion

The prospective nature of our data collected supported our efforts to delineate associations between SRI exposure in pregnancy and longer-term outcomes in offspring. Population-based studies recently linked prenatal SSRI exposure to increased risk for ASDs^{13, 15, 16}, though at least one large study reported negative findings once maternal mental illness was accounted for¹⁷. While the current study did not assess for a diagnosis of ASD per se, measures of expressive language and behavior problems linked to ASD were assessed using clinician, mother, and alternate caregiver ratings. Several observational studies have failed to find associations between prenatal SRI exposure and neurodevelopmental outcomes, using clinical assessment^{7, 9}, standardized measures^{6, 8, 34}, and physician exams¹⁰. Methodological issues may have contributed to the discordant results. Of note, the number of medication exposed infants in most previous studies is less than 50, the inclusion of prenatal and postnatal maternal symptoms is rare, and many studies did not use standardized developmental measures. The one research group to prospectively examine expressive language outcomes in children prenatally exposed to antidepressants compared to unexposed controls showed no group differences^{5, 35}, but language measures were associated with postpartum depression. Differences in statistical approach, outcome measures, and covariates included may have accounted for the discrepant findings. Given the mixed findings, replication is clearly needed.

Effect sizes in the current study are consistent with the heterogeneous nature of complex disorders such as ASDs, where multiple etiological factors are likely at play. Likewise, only a proportion of the children who were exposed prenatally to SRI’s showed altered language or behavioral outcomes. Future studies should focus on moderating factors that help explain the variability in exposure effects, including genetic variation^36–38. Prenatal exposure to SRIs likely impact a subset of offspring, thus identifying moderators such as genetic markers, is a critical next step. Identifying maternal physiological factors influencing the extent of fetal medication exposure³⁹ may also be relevant.

The fact that no other psychotropic medication classes (e.g., anxiolytics, antipsychotics) were associated with ASD-related behaviors or any other child outcomes should also be highlighted. In addition, prenatal SRI exposure was related to language and PDD subscale scores but not general cognitive functioning. The Rai et al study found prenatal antidepressant exposure associated with increased risk of ASD without intellectual impairment but not increased risk of ASD with intellectual impairment. The children without intellectual impairment reflect a specific subset of the ASD continuum and include those with Asperger’s syndrome and specific deficits in social communication. Future studies should focus on identifying neural mechanisms that may link prenatal SRI exposure to specific deficits observed in ASDs; including measures of pragmatic language and social communication in follow up studies would be particularly helpful.

Despite the detailed information available and thorough statistical interrogation of the data, it is not possible to assign a causal relationship between SRI exposure and risk for ASD. Due to the ethical limitations of randomly assigning women to treatment conditions, this study is by nature, observational. Thus, the type of medication, dosages, and timing vary for each individual, as do the patients’ diagnostic histories. Using a multi-method approach including clinician-administered and self-report measures helped capture this inherent variability. While the observational study design limits our ability to draw causal inferences, our sample closely matches the day-to-day clinical scenarios faced by treating physicians.

The issue of determining the ideal control group remains a potentially contentious debate. As noted, our sample had a limited group of non-exposed infants and non-ill mothers. The significant associations with covariates underscore the potential mediating factors that are often associated with greater maternal psychopathology (e.g., tobacco use). Another point of debate includes the accurate grouping of medications into classes. Grouping medications is necessary given that most studies, including population-based studies, do not have sufficient statistical power to measures the effects of individual medication exposures. We used an a priori approach, as medications were grouped prior to all analyses, based on a series of discussions by study psychiatrists and other investigators. Reasonable alternatives to our groupings could certainly be proposed, and which medications were included and not included may contribute to discrepant results in the literature.

Left untreated, maternal prenatal mood disorders are associated with many adverse neurodevelopmental and behavioral outcomes in children, including lower activity levels, decreased motor tone, altered stress regulation, brain morphology differences, increased irritability, negative affectivity and childhood behavior problems^40–49. Given that both psychiatric illness and its treatment may negatively impact offspring, and that both have independently been associated with ASD risk, patients and providers have to be cautious in terms of discontinuing or avoiding medication use in pregnancy. Discontinuation of medication during pregnancy is associated with a significantly higher rate of relapse for major depressive disorder; 68% compared to 24% of women who continue medication⁵⁰.

Though many studies suggest the ASDs evidence high heritability^51–54, approximately 55% of variance in ASD risk is attributable to environmental factors⁵⁵. Thus, identifying environmentally mediating risk factors of ASDs is paramount to prevention. The current data, while speculative, suggest a small but significant association between prenatal SRI exposure and outcomes typically associated with ASDs, including mild expressive language deficits. While the direct clinical import of these findings is limited pending replication, our data suggest that a subset of children may be vulnerable to the effects of prenatal SRI exposure. Replication with samples that include important comparator groups and identification of moderating risk factors is critical for understanding the clinical implications of these findings. Future studies would benefit from efforts to combine data sets across multiple clinical sites, given the relatively low incidence of prenatal exposure and the modest prevalence rate of ASDs. Whether the current findings reflect a delay or a deficit is currently unknown, and prospective longitudinal data across childhood and adolescence would be needed to test those hypotheses. These novel prospective data indicating an association between SRI exposure and ASD must be balanced with the adverse effects of maternal depression in the clinical decision process. Similarly, these data provide a foundation for future clinical and laboratory investigations in advancing our understanding of the etiopathogenesis of ASDs.

Clinical Points.

• A significant number of pregnant women present with depression. Long-term, prospective studies following the children of these women are lacking.

• These findings do not suggest physicians avoid SRI treatment for all pregnant women, but instead suggest that more research is needed to determine whether certain offspring are at risk for negative outcomes related to prenatal SRI exposure versus untreated and undertreated depression.