Main

The need to manage maternal depression during pregnancy has been well established, yet conflicting evidence regarding the developmental consequences of use of serotonin reuptake inhibitor antidepressants (SRIs) during pregnancy (1,2) continues to cause uncertainty about how to manage antenatal maternal mood disturbances while minimizing the risks associated with exposure to maternal depression. Up to 20% of mothers experience antenatal depression and anxiety disorders and 5–8% of infants are born with in utero exposure to SRIs (3).

SRIs include both selective serotonin reuptake inhibitors (SSRI) and serotonin norepinephrine reuptake inhibitors (fluoxetine, paroxetine, sertraline,citalopram, and venlafaxine, respectively). SRIs inhibit the reuptake of serotonin (5HT), via 5HT transporter blockade at presynaptic neurons, thereby increasing synaptic 5HT concentrations. Long before birth, serotonin plays a key developmental role in regulating neuronal proliferation, differentiation, migration, and synaptogenesis (4). Central to our understanding of how prenatal SRI exposure might influence early brain development is an appreciation of the diverse role played by the neurotransmitter 5HT. Before 5HT becomes a neurotransmitter in the mature brain, it plays a role as a neurodevelopmental signal, regulating cells and influencing serotonin receptor growth (5). Given that SRIs readily cross the placenta (6) and presumably alter central 5HT levels during critical periods of development, in utero exposure to SRIs may influence behavioral outcomes during childhood.

Genetic and early-life pharmacological manipulations of 5HT levels in animal models reveal a link between developmental changes in 5HT and anxious behaviors later in life (7). In mice, fluoxetine increases anxiety-like behavior when early SRI exposure occurs within a specific developmental time frame, akin to the human third trimester. Similar manipulations, analogous to SRI exposure, in animal models where genetically absent 5HT transporter molecules increase intrasynaptic 5HT, also lead to adult onset of depression and anxiety-like behaviors (8). As SRI exposed animals mature they demonstrate decreased serotonergic tone, possibly reflecting prolonged activation of inhibitory receptors (i.e., 5HT1a) (9). Together these findings suggest that altered 5HT auto-inhibitory feedback in the presence of high serotonergic tone (i.e., secondary to SRI exposure or genetically absent 5HT transporter (5HTT) protein) results in a constrained maturation and function of the 5HT system in ways that paradoxically link prenatal SRI exposure with an increased risk for affective disorders later in life (10,11). To date, there is a remarkable paucity of human studies of childhood outcomes following fetal modulation of 5HTT functions or exposure to high 5HT levels during fetal brain development.

To further examine the behavioral effects of in utero SRI exposure, we undertook the following study of internalizing, and externalizing behaviors at an early school age. Longitudinal maternal reports of child behavior at 3 and 6 y were compared between children with and without in utero SRI antidepressant exposure. Based on preclinical models and our previous findings, we expected increased levels of internalizing behaviors in the SRI exposed children even when controlling for the effect of prenatal and concurrent levels of maternal depression and anxiety. Thus, to further examine hypothesized differences in internalizing behaviors, we also examined whether these effects were primarily driven by differences in anxious behaviors (as the animal models suggest) or differences in attention problems.

Results

Maternal and Neonatal Characteristics

Maternal characteristics were similar across exposure groups with the exception that mothers treated with an SRI during pregnancy were recruited earlier in pregnancy (25.0 vs. 26.7 wk, P = 0.003), were slightly less educated with half of exposed mothers having completed some postsecondary or less compared to 22.7% of unexposed mothers (P = 0.022), were more likely to have used other psychotropic medicines, and had significantly higher scores on the scales measuring depression and anxiety at all time points ( Table 1 ). The majority of mothers were already on an SRI at conception and had previously had a diagnosis of depression and/or anxiety. Approximately half of the women in our cohort reported consuming alcohol at some point during their pregnancy (47%); however, among this 47% of women, the mean number of servings consumed across the entire pregnancy was 6.4, suggesting very little total alcohol consumption. Servings were defined as 12 ounces of beer, 5 oz of wine, or 1.5 oz of spirits. Alcohol consumption did not differ significantly by exposure status.

Table 1 Maternal characteristics

Neonates who were exposed to SRIs in utero were more likely to be born preterm (13.6 vs. 3.0%, P = 0.036) and at an earlier gestational age (39.1 wk gestational age vs. 40.1 wk, P = 0.001), they were smaller both in terms of weight (3,214.6 ± 491.6 vs. 3,575.0 ± 516.0, P < 0.001) and length (50.3 ± 2.5 vs. 51.8 ± 2.8, P = 0.006), were significantly more likely to have an Apgar score < 7 at 1 min (27.3 vs. 9.1%, P = 0.012) and were more likely to have been jittery after delivery (31.8 vs. 4.6%, P < 0.001) ( Table 2 ). Unexposed children also completed their follow-up visits or data collection at a younger age (P = 0.003 and 0.016 respectively).

Table 2 Infant and child characteristics

Childhood Behavioral Outcomes: 3 and 6 y

At 3 y, exposed children had higher levels of internalizing behavior and the anxious/depressed symptoms (50.25 ± 12.6 vs. 45.7 ± 8.2, P = 0.023, and 2.45 ± 2.64 vs. 1.23 ± 1.5, P = 0.003, respectively) ( Table 3 ). They were also more likely to meet a clinical threshold for internalizing behaviors (15.9 vs. 1.5%, P = 0.004). Similarly, at 6 y of age, exposed children continued to have significantly higher levels of internalizing behaviors (HBQ-P: 0.39 ± 0.28 vs. 0.29 ± 0.21, P = 0.037) and were still more likely to exceed clinical thresholds for internalizing behaviors (13.6 vs. 3.0%, P = 0.036). Differences in the anxious subscale of the HBQ-P did not reach significance at 6 y of age, and there were no significant differences in the externalizing or attention scales at either 3 or 6 y of age between prenatal SRI exposure groups.

Table 3 Crude CBCL and HBQ-P scores by exposure status

Figure 1 illustrates the mean scores in internalizing (a subscale examining the extent to which children direct their emotional responses, which includes an assessment of anxiety and depressed behaviors) externalizing (a subscale that examines the degree of aggression, delinquency, and hyperactivity) as well as the subscales that more directly assess children’s levels of anxiety and attention by exposure status at both ages 3 and 6. Increased levels of internalizing (mean Z-score for exposed children was 0.28 at 3 y and 0.22 at 6 y vs. −0.16 and −0.18 in unexposed) ( Figure 1a ) and anxious behaviors (mean Z-score for exposed children was 0.37 at 3 y and 0.15 at 6 y vs. −0.26 and −0.14 in unexposed children) ( Figure 1c ) were observed in SRI-exposed children (ANOVA model results F = 7.52, P = 0.0072 and F = 7.91, P = 0.0058, respectively). No significant group differences in externalizing behaviors (mean Z-score for exposed children was 0.0 at 3 y and −0.08 at 6 y vs. 0.09 and 0.04 in unexposed children) and attention problems (mean Z-score for exposed children was 0.09 at 3 y and −0.15 at 6 y vs. 0.01 and 0.05 in unexposed children) were observed between exposed and unexposed children ( Figure 1b (F = 0.43, P = 0.5144) and 1D (F = 0.13, P = 0.7220), respectively). No significant time by group interactions were observed for any of the child behaviors.

Figure 1
figure 1

Means of standardized: internalizing behaviors (a); externalizing behaviors (b) anxious subscale (c); attention problems (d) at 3 and 6 y or age according to serotonin reuptake inhibitor exposure in user. The broken line represents exposed children and the solid line represents unexposed children. * indicates a significant group difference at the 95% confidence level as reported by the repeated measure ANOVA. There were no significant groups by time interactions.

PowerPoint slide

Using multilevel mixed model regression analysis, we observed that in utero SRI exposure was associated with significantly higher scores on the internalizing behavior (β = 0.38 (0.01, 0.79)) and anxious behaviors subscales (β = 0.49 (0.04, 0.94)). These positive and significant associations persisted when controlling for concurrent maternal depression and during the second and third trimester and 6 mo postpartum, delivery by cesarean section, child sex, use of other psychotropic medicine during pregnancy, drinking alcohol during pregnancy, and maternal education ( Table 4 ). There were no statistically significant differences in any of our outcomes by child sex. There were also no differences in externalizing behaviors or attention problems according to SRI exposure status. Concurrent maternal depression (maternal depression when the child was 3 and/or 6) was statistically significantly associated with increased levels of internalizing (β = 0.64 (0.24, 1.05)) and externalizing symptoms (β = 0.87 (0.47, 1.27)) in children and increased anxious behaviors (β = 0.53 (0.08, 0.98)).

Table 4 Results from multilevel mixed effects regression models, β-coefficients (95% CI)

Discussion

Our results suggest that in utero exposure to maternal mood disturbances and to the antidepressants used to manage these symptoms are associated with increased levels of internalizing behaviors and anxiety at school entry. Children who were exposed to SRIs in utero were reported to have more anxious and internalizing behaviors at 3 y of age and this pattern persists at 6 y of age. There was no interaction with time, suggesting that children do not outgrow these behaviors by age 6. Moreover, in utero SRI exposure did not appear to influence risk for externalizing and attention problems, while maternal depressive symptoms during childhood were associated with increased reports of these behaviors. Significantly more exposed children in our cohort reached clinically significant thresholds for anxiety and internalizing behaviors at 3 and 6 y of age; however, we still cannot rule out that this is the result of the underlying maternal depression.

Our findings are consistent with SRI-related behavioral changes observed in animal models that reflect the effects of early changes in central serotonin signaling associated with SRI antidepressant exposure at developmental times comparable with the human third trimester (8,12,13,14). Our results are also consistent with the previous findings reporting no differences in externalizing and attention behaviors in SRI-exposed animals (13,14) and children (15,16). The lack of association between SRI exposure in utero and externalizing and attention behaviors was expected based on previous reports, and follows from the hypothesized biological mechanism of the downregulation of the developing fetus’ serotonergic system (17). In a mature brain, acute SRI exposure may have an inverse modulatory effect whereby it upregulates 5HT and downregulates norepinephrine-locus coeruleus (NE-LC) (17). In a rodent model, Maciag et al. (14) reported downregulated 5HT-raphe circuit function following citalopram exposure during a period analogous to the human third trimester, whereas enhanced NE-LC system function has been reported following neonatal citalopram administration (18). Such early (or fetal) exposure leading to increased locus ceruleus (NE-LC) function raises an intriguing question of whether anxious childhood behavior in exposed children may also reflect abnormal noradrenergic (NE) transmission (19). The developmental effects of altered NE-LC system and underlying mechanisms associated with such signal modulation remains to be demonstrated in humans with in utero exposure to SRIs.

In human studies, distinguishing the impact of maternal mood disturbances from in utero SRI exposure remains challenging. Previous research has illustrated that the duration of maternal depression and number of depressive episodes predicts lower IQ and language scores in fluoxetine-exposed children (15–70 mo) (20). At 3 y of age, both in utero SRI exposure and concurrent maternal mood equally predicted increased internalizing behaviors (21). At 4 y, behavioral disturbances in prenatally SRI exposed children in another cohort were also influenced by maternal mood symptoms (15,16). Our current findings add longitudinal data that indicates that in utero SRI exposure is associated with sustained increases in internalizing behaviors beyond early childhood.

At this point, it remains unclear whether increased anxious behaviors represent an emerging clinically relevant mood disorder in children with in utero SRI exposure. Anxiety is one of the most common debilitating lifetime psychiatric disorders occurring in 5–20% of individuals, with symptoms frequently emerging in middle to late adolescence (22,23). In early childhood, these symptoms are expressed as behavioral inhibition, reflecting a tendency to withdraw from novel social situations, heightened vigilance, and fear (24). Such inhibition may represent a potential endophenotype that may evolve into an anxiety disorder in adolescence. Whether our findings point to the effects of early life shifts in central 5HT signaling, secondary to in utero SRI exposure, with a clinically significant condition in later childhood remains to be determined.

A number of key limitations should be acknowledged. We cannot rule out that the relationship reported between SRI exposure in utero and childhood behaviors is a result of confounding by the underlying maternal depression. Given that our behavioral outcomes are based on maternal report, there is some risk of differential outcome reporting by maternal mood or SRI exposure status. The extent of this potential influence is difficult to quantify as there is conflicting evidence in the literature, with some research suggesting that anxious parents are overly pessimistic and biased in their reporting of their anxious children’s behaviors (25) while a recent study has illustrated that neither maternal depression nor anxiety contributes significantly to maternal ratings of their children as compared to independent and blinded observer ratings (26). We are also unable to control for confounding by maternal genetic status that might have predisposed a woman to develop a mood disorder (which was then treated with an SRI), and might be heritable also predisposing her child to increased internalizing behaviors. Thus, our findings could be related to a genetic predisposition towards internalizing behaviors and anxiety, rather than an effect of the drug or maternal depression. Finally, we were unable to examine the relationship between the timing, or the duration, of the SRI exposure and childhood behavioral outcomes, as the majority of our mothers using SRI (80%) did so for more than 200 d of their pregnancy.

Our results suggest that in utero SRI exposure is associated with increased internalizing and anxious behaviors at an early school age. However, equally important is our finding that concurrent exposure to maternal depression during early childhood was also associated with increased anxious, internalizing as well as externalizing behaviors, even in the presence of continued maternal treatment with an SRI. We report a sustained increase in child behavioral disturbances despite the fact that two thirds (72.7%) of our mothers who used SRI treatment in pregnancy were still on SRI when their child was 6 y old. These findings raise critical questions about the long-term benefit, or lack thereof of SRI treatment for perinatal mood disorders. SRIs are prescribed to pregnant women with the expectation they will improve maternal mental health, and thus also to confer benefits on the developing fetus and child. Given current rates of use of antidepressants during pregnancy (3) and the concerns about childhood and adolescent anxiety (27), a clearer understanding of the long-term impact of in utero SRI exposure on children’s mental health is needed to shape care for mothers with mood disorders and their children.

Methods

In this longitudinal cohort study, 191 women were recruited (258 were approached) during their second trimester of pregnancy (mean gestational age 25.7 wk) from community midwife clinics, family physician clinics, and a reproductive mental health clinic in metropolitan Vancouver (British Columbia, Canada). Seventy-six mothers were using SRIs. Of these mother-child pairs, 110 had complete data on child outcomes at 3 and 6 y of age (44 children exposed to SRIs vs. 66 not exposed). We examined the 81 mothers and children who had incomplete data at 3 and 6 y and they were similar in every respect except that children with incomplete data had a significantly different gestational age at birth (39.14 vs. 39.7 wk; P = 0.035) and were more likely to be male (56 vs. 39.1%; P = 0.023).

Among mothers using SRIs, the most common drug used was paroxetine ( Table 5 ) and the mean duration of exposure for all SRIs spanned most or almost all of the pregnancy. Informed consent was obtained from all mothers and the study was approved by the University of British Columbia Research Ethics Board and the BC Women’s Hospital Research Review Committee. All SRI exposed mothers were already on antidepressants at the time of conception and had been diagnosed with an Axis I mood disorder by their psychiatrist or family physician prior to entering the study.

Table 5 Specific SRIs (n), prenatal dose (range), and length of prenatal exposure

Inclusion criteria for this study were: singleton pregnancy, confirmed gestational age at birth, ability to give informed consent, and no fetal anomalies that were detected by antenatal ultrasound. Exclusion criteria included the presence of a maternal bipolar disorder, illicit drug use or abuse and significant maternal medical, obstetrical, or fetal conditions.

Maternal Characteristics and Neonatal Data

Demographics, reproductive and medical history, and information regarding prescribed medications were obtained at the time of enrollment (mean 25.7 wk gestation), at 36 wk gestation, and 6 mo postpartum. Data on neonatal outcomes at birth were obtained from the neonatal health records including gestational age at birth, birth weight, length, head circumference, Apgar scores at 1 and 5 min, and whether the infant exhibited symptoms of postnatal adaptation syndrome—a cluster of symptoms reported in ~30% of neonates exposed to SRIs in utero (28) including rapid breathing, respiratory distress, jitteriness, and abnormal tone (29,30,31,32).

Maternal Depression and Anxiety

Maternal depressed mood was assessed using the clinician-rated Hamilton Rating Scale for Depression (HAMD) (33) and maternal anxiety was assessed using the Hamilton Rating Scale for Anxiety (HAMA) (34) at the time of recruitment (mean 25.7 wk gestation) and during the third trimester of pregnancy (mean 34.6 wk gestation) and again at 6 mo postpartum (mean infant age 26.7 wk) and when the child was 6 y old (mean child age 6.06 y). At 3 y of age (mean age 3.64 y), measures of maternal depression and anxiety were obtained by self-report using the Beck depression and anxiety inventories. The HAMD is a 21-item clinician rated scale that measures the severity of depression in adults, adolescents, and children with a range from 0 to 63. The HAMA is a 14-item scale with a range from 0 to 56 (33). The Beck depression and anxiety inventories are 21-item multiple-choice self-report questionnaires designed to assess the severity of depression and anxiety. Both have a range from 0 to 63 (35,36).

Child Behavioral Outcomes

At 3 y of age, maternal report of child behavior was obtained using the Child Behavior Checklist (CBCL 1.5–5) (37). The CBCL provides a standardized rating of behavior in children aged 1.5–5 y and yields a total problem score, externalizing and internalizing scores, and subscale scores. At 6 y of age, mothers completed the MacArthur Health and Behavior Questionnaire (HBQ-P), which yielded measures of internalizing, externalizing, over anxious, and inattention behaviors. The HBQ was derived from the Ontario Child Health Study Measure designed to map on DSM-II_R symptom criteria consistent with the dimensions of internalizing and externalizing behaviors (38). The HBQ-P has strong psychometric properties and has been used to assess child mental health across multiple ages from 4.5 into adolescence (39).

Statistical Analysis

To compare 3- and 6-y behavioral outcomes, we standardized the scores from the CBCL and the HBQ-P and used z-scores of both scales. Crude differences in maternal, newborn characteristics, and CBCL and HBQ-P scores were compared between mothers and children exposed to SRIs and those unexposed using t-tests (continuous variables) or χ2 tests for continuous or categorical variables respectively. Proportions of children who exceeded a validated clinically significant threshold for the internalizing and externalizing subscales (a score of 70 or higher for the CBCL scales and a threshold of 0.71 for internalizing and 0.68 for externalizing on the HBQ) were compared again according to exposure status (39).

Repeated measures ANOVA was used to examine group differences (SRI exposed vs. unexposed) and time (3 and 6 y) by group differences in child behaviors. We then used multilevel mixed effects models with a random intercept to examine our longitudinal data for each child for each outcome, as we had repeated measures of internalizing, externalizing, anxiety and attention for each child. Multilevel mixed effects regression modeling allowed us to take advantage of our repeated measures data while adjusting for maternal depression or anxiety (during the second and third trimester, 6 mo postpartum, and when the child was 3 and 6 y old). Cut-off scores of 17 or higher on the HAMD, HAMA, and Beck scales were used to create dichotomous variables indicating that the mother was experiencing clinically significant depression or anxiety symptoms (40). Both maternal anxiety and depression were examined; however our final analytic models only controlled for depression as including maternal anxiety in the model did not meaningfully change the effect size of the SRI exposure variable (defined as a greater than 10% change). We also adjusted for sex of the child, maternal drinking during pregnancy, mode of delivery (cesarean section), and prenatal use of other psychotropic medications, and maternal education. Maternal smoking status during pregnancy was obtained, but only two mothers reported smoking, so it was not controlled for.

Statement of Financial Support

G.E.H. was supported by the Canadian Institutes for Health Research, the Michael Smith Foundation for Health Research (British Columbia, Canada), the Women’s Health Research Institute (Vancouver, Canada) and Neurodevnet (Canada). T.F.O. is the R. Howard Webster Professor in Early Child Development at the University of British Columbia and his work is supported by the Child and Family Research Institute and the Canadian Institutes for Health Research. This study was supported by the Child and Family Research Institute (UBC) and the Canadian Institutes of Health Research (MOP-57837). The funders had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review, or approval of the manuscript.

Disclosure

The authors report no conflict of interest.