A summary of studies on the use of SSRIs in human pregnancy is presented in Table 1. Many studies on the use of SSRIs during pregnancy have not shown an increase in the overall risk of major malformations [¹⁰–²⁸]. Several studies have suggested that the use of SSRIs, particularly paroxetine, may be associated with a small increased risk for cardiovascular malformations [²⁹–³⁸]. Other studies have noted associations between SSRIs in pregnancy and specific types of relatively rare major malformations (neural tube defects, craniosynostosis, omphalocele, or right ventricular outflow tract defects) [³⁹–⁴¹]. Some of these studies are retrospective and burdened with recall and selection bias. The largest dataset with prospective exposure information available to date is from the Swedish Medical Birth Registry. Their initial report has been negative [¹⁶]. Later, increased risk of cardiovascular defects was observed with paroxetine, predominantly of septal defects [³⁰, ³¹]. Recently, new associations were noted, for cystic kidney with SSRIs, for relatively severe malformations with fluoxetine, and for any cardiovascular defect and hypospadias with paroxetine [³⁸]. Three recently published meta-analyses on paroxetine exposure in pregnancy and cardiovascular malformations have not been consistent [⁴²–⁴⁴]. In the meta-analysis published by Bar-Oz et al. [⁴²], first trimester paroxetine exposure was associated with a significant increase in the risk of cardiac anomalies. Women using antidepressants in pregnancy were more likely to utilize ultrasound in pregnancy and postnatal echocardiograms compared with women who did not. In this study, significantly more women receiving paroxetine used the medication for anxiety or panic disorders compared to women using other SSRIs. Detection bias was suggested as a contributing factor to the observed risk of cardiovascular malformations with paroxetine. In the meta-analysis published by O'Brien et al. [⁴³], no increased risk of congenital malformations was associated with paroxetine. Cardiac malformation rates were similar and within population norms. In the meta-analysis published by Wurst et al. [⁴⁴], there was an increased risk for combined cardiac defects and aggregated congenital defects with first trimester paroxetine use. Two opposing commentaries on this topic were recently published [⁴⁵, ⁴⁶]. The definition of cardiovascular malformations varied among studies, some including small septal defects, while others excluded them. The inconsistency across these studies may be explained by differences in study design, by confounding factors, for example, maternal underlying psychiatric disorder, coadministered medications, lifestyle factors (smoking, drinking), maternal BMI, and diabetes, or they may be spurious. Overall, there are over 33,000 reported pregnancy outcomes after prenatal exposure to various SSRIs.

We have calculated the overall rate of major congenital anomalies and of cardiovascular anomalies in the published prospective studies after prenatal exposure to SSRIs, where rates were available [¹⁰–¹⁵, ²⁰, ²⁵, ³², ³³] and found 3.8% (189/4920) and 0.9% (53/6094), respectively, both well within their baseline risk in the general population. It can be summarized that the majority of the prospective studies have not shown an increase in the overall risk of major malformations. The studies which have suggested that SSRIs may be associated with a small increased risk for malformations were particularly with paroxetine.

Seven criteria for proof of human teratogenicity have been amalgamated by Shepard [⁵¹] and are presented in Table 2 and discussed as follows.

1. Many of the studies with positive findings on SSRIs in pregnancy are prescription studies, and women may not have actually taken the drugs. Exact timing of exposure during sensitive periods is often problematic, although exposure preceded the outcome.
2. There are several isolated studies with inconsistent findings of statistically significant associations. Confounding factors are often insufficiently controlled for. Many of the prospective studies are underpowered for associations between exposure and specific malformations. Many of the retrospective studies are burdened with potential biases. All studies considered here are observational. The relative risk even in positive studies is below six and the lower bound of the 95% confidence interval often close to one. The findings in regard to the type of malformations are inconsistent in the underlying studies and even in studies from the same database published at different time points.
3. In the positive studies, there was some dominance of cardiovascular malformations, septal defects in some studies, and right ventricular outflow tract obstruction defects in others. In the case of SSRIs, particularly paroxetine; however, they were non-specific. In many of the large studies, diagnosis of malformations uses classification codes and lacks careful delineation of clinical cases.
4. The fourth criterion is not relevant in the context of SSRIs. SSRIs are common exposures in pregnancy, and most of the described defects are also relatively common.
5. Animal reproductive studies in rats and rabbits administered paroxetine [⁵²], fluoxetine [⁵³], or sertraline [⁵⁴] during organogenesis did not show a teratogenic effect.
6. There is evidence based on mouse whole-embryo studies to suggest that serotonin plays a role in cardiovascular and craniofacial development [⁵⁵–⁵⁸]. Paroxetine 1 μM was shown to decrease serotonin-mediated proliferation of dissociated rat embryonic cardiac myocytes [⁵⁹]. Rat whole-embryo culture results showed an increase in branchial bar fusion, but not cardiac malformations, after exposure to paroxetine at concentrations much higher than those achieved clinically [⁶⁰]. It has been speculated that the observed malformations in vitro may be early ontogenetic indicators for infrequent cardiovascular anomalies observed in vivo. Fluoxetine was found to adversely affect cell viability and differentiation to cardiomyocytes at higher concentrations than those achieved clinically in a dose-dependent manner using mouse embryonic stem cell system [⁶¹]. SSRIs inhibit the serotonin transporter. However, developmental defects were not observed in mutant serotonin transporter knockout mice [⁶²]. In another study, gross morphologic abnormalities were not seen in serotonin transporter knockout mice, but association was found with sudden death of the newborn mice in the first week after delivery [⁶³]. Histologic analysis of heart sections of these mice showed that they develop cardiac fibrosis. In terms of biological plausibility, in vitro studies have been helpful in suggesting mechanistic information that adds to the plausibility of the suspected association, though in concentrations much higher than plasma concentrations in humans in clinical settings. However, as stated earlier, in vivo animal studies to date have not supported an association between in utero exposure to SSRIs and major anomalies.
7. There is evidence for placental transfer of SSRIs.

In summary, despite some troubling associations between SSRIs and major malformations, especially cardiovascular, the overall current scientific evidence has not fulfilled the criteria for proof of human teratogenicity of SSRIs. Despite having over 33,000 reported pregnancy outcomes after prenatal exposure to various SSRIs, the differences in the design of these studies and their conflicting results are confusing. One, therefore, wonders whether further well-designed epidemiologic studies, with sufficient power and good control of potential confounders will be helpful in verifying whether SSRIs are indeed associated with a small increased teratogenic risk, especially regarding cardiovascular anomalies. In our opinion, the current data do not support teratogenicity of SSRIs.

Most studies did not specifically focus on the impact of SSRIs on the risk of miscarriage. It was often a secondary outcome without observing a significantly increased risk. There was an increase in the miscarriage risk in two meta-analyses [⁶⁴, ⁶⁵]. However, in the included prospective cohort studies, crude rates were reported, and the effect of earlier gestational age at contact, which is an important factor [⁶⁶], was not corrected for. In two studies that specifically focused on the risk of miscarriage, an increased risk was found with the use of antidepressants during pregnancy [⁶⁷, ⁶⁸], SSRIs alone, serotonin-norepinephrine reuptake inhibitors alone, and combined use of antidepressants [⁶⁸]. When looking at antidepressant use by type, paroxetine alone and venlafaxine alone were associated with increased miscarriage risk. Despite an attempt to adjust for psychiatric history, the possibility of confounding by underlying psychiatric disorder could not be ruled out.

In a Finnish study, there was no increase in the rate of preterm delivery, SGA or LBW [¹⁹]. The risk of both low birth weight and preterm delivery was increased in infants who were born to mothers who had received SSRI therapy [¹⁷, ²¹]. Infants exposed to SSRIs had shorter gestation and lower birth weight compared to nonexposed infants [⁶⁹]. The increased risk of low birth weight remained significant, even when maternal illness severity was accounted for. The adjusted OR for preterm delivery was doubled in SSRI-exposed women compared to two groups of women who had not used SSRIs during pregnancy, one with psychiatric history and another without [⁷⁰]. In another study, the risk of preterm delivery was not significantly increased among SSRI users, but the risk of SGA offspring was increased among women who continued SSRI use beyond the first trimester [⁷¹]. In a study from the Quebec Pregnancy Registry, no association was found between SSRIs and the risk of SGA regardless of trimester of exposure [⁷²]. In other studies, there was an increased risk for preterm delivery among women exposed to SSRIs in the second or third trimesters [³⁸] or to antidepressants [⁷³] with no increased risk for LBW or SGA. The underlying psychiatric disorder is a potential confounder in most of these studies.

In summary, associations were found in some studies between the use of SSRIs during pregnancy and risk of miscarriage, IUGR, or preterm delivery. Most of these studies are potentially confounded by the gestational age at initial contact and the underlying psychiatric disorder.

Neonatal symptoms have been described initially following prenatal exposure to fluoxetine [⁷⁴] and later on after exposure to paroxetine and other SSRIs [⁷⁵–⁸¹]. Neonatal toxicity or discontinuation (withdrawal, abstinence) syndromes associated with SSRIs are characterized by irritability, abnormal crying, tremor, and poor neonatal adaptation including respiratory distress, tachypnoea, jitteriness, lethargy, poor tone or colour, and, rarely, convulsions. The neonatal effects have been described in up to 30% of neonates exposed to SSRIs late in pregnancy [⁸²]. Most symptoms are mild and transient.

It can be concluded that SSRI use late in pregnancy, similar to many other psychotropic drugs, is associated with neonatal transient effects.

Some epidemiologic studies have suggested an association between maternal use of SSRIs late in pregnancy and an increased risk of persistent pulmonary hypertension of the newborn (PPHN) [³⁸, ⁸³, ⁸⁴]. In these studies the absolute risk of PPHN was <1%. In the study which used data from the Swedish Medical Birth Register [⁸⁴], the eleven infants whose mothers reported the use of SSRI in pregnancy and had PPHN survived the neonatal period. Contrary to the above, other studies, possibly underpowered, did not find such an association [²⁶, ⁸⁵]. A recent study found PPHN to be associated with mode of delivery, specifically caesarean delivery prior to the onset of labour, but not with SSRI use in the second half of pregnancy [⁸⁶].

In summary, an absolute risk of <1% for PPHN in infants exposed to SSRIs cannot be excluded, although studies are not consistent.

Most studies have focused on possible postnatal neurodevelopmental effects. Children of mothers exposed in pregnancy to fluoxetine or tricyclic antidepressants were neurodevelopmentally assessed and compared to an unexposed control group. Similar global IQ and language scores were found in the three groups [⁸⁷, ⁸⁸]. No significant differences in neurobehavioral scores were found between children whose mothers were taking fluoxetine during pregnancy and nonexposed children [⁸⁹]. Normal development was observed in a small group of children exposed in pregnancy to citalopram and followed up to 1 year [⁹⁰]. Infant developmental assessment done at 2 and 8 months of age revealed no significant differences between SSRI-exposed and unexposed infants [⁹¹]. Levels of internalizing or externalizing behaviours did not significantly differ between children prenatally exposed to SSRIs and unexposed [⁹², ⁹³]. On the other hand, maternal depression and anxiety were associated with increased reports of internalizing and externalizing behaviours in their children.

Mental developmental indexes were similar in children whose mothers were diagnosed with major depressive disorder treated or untreated in pregnancy. However, children exposed to SSRIs scored lower on the psychomotor developmental indexes and the motor quality factor of the behavioural rating scale compared to unexposed children [⁹⁴]. In a follow-up study using a psychomotor developmental test (Boel), abnormal test was more frequent in children prenatally exposed to antidepressants compared to unexposed [⁹⁵]. In another neurobehavioral assessment study, newborns prenatally exposed to SSRIs had abnormal outcomes including increased motor activity, fewer changes in behavioural state, and abnormal sleep patterns [⁹⁶].

Children's developmental milestones were assessed using a questionnaire at 6 and 19 months of age. Second or third trimester exposure to antidepressants was associated with later gross motor developmental milestones, though still within normal range, compared to unexposed children [⁹⁷].

Children who had neonatal abstinence syndrome had similar mean overall developmental results compared to those who did not; however, they were more likely to have abnormal results on the social component of the Denver developmental test [⁹⁸].

A recent prospective study demonstrated that SSRIs during pregnancy affect the neurobehavioral development of the human fetus [⁹⁹]. Fetuses exposed to SSRIs exhibited dose-related increased motor activity and disrupted sleep. The significance of the observed changes on postnatal development is unclear.

In a recent population-based case-control study, a two-fold increased risk of autism spectrum disorders was found with prenatal exposure to SSRIs [¹⁰⁰]. Further studies are needed to verify the suggested association.

In summary, in most of the studies that focused on the possible neurodevelopmental effects of prenatal SSRI exposure, there is no conclusive evidence for an increased risk of adverse long-term effects.

When evaluating the risk/benefit ratio of SSRI treatment in pregnancy, the risks associated with treatment discontinuation should also be considered. Abrupt discontinuation of psychotropic drugs in pregnancy is associated with physical and psychological adverse effects [¹⁰¹]. SSRI treatment discontinuation during pregnancy is associated with a higher frequency of relapse [¹⁰²]. Depression is associated with an increased risk for preterm delivery [¹⁰³–¹⁰⁵]. The risk of preterm delivery increases with increasing severity of depression [¹⁰⁶]. Treated women have lower depressive symptom scores and better functioning [¹⁰⁵]. These risks should be a factor in the decision making in regard to treatment continuation during pregnancy.