Introduction

During pregnancy, the point prevalence is 11 % for depression and 13 % for anxiety [1]. Given that psychotropic medications are first-line treatments for both depression and anxiety, it is unsurprising that many pregnant women will be prescribed medicine to treat these symptoms; antidepressant prescribing during pregnancy has increased almost fourfold since the 1990s [2]. In a recent study of privately insured Americans, 10 % of pregnant women received a prescription for some psychotropic medication, of which 25 % were diagnosed with depression and 24.4 % anxiety [3].

As a recent New York Times article highlighted, psychiatric illness during pregnancy remains both stigmatized and fraught with complex decisions for the mother-to-be [4]. Furthermore, survey and interview data suggest that depressed pregnant women experience moderate to high levels of decisional conflict about whether to take medication, regardless of socioeconomic status, illness severity, or current treatment with medication [5]. Clinicians also face a large and conflicting body of research on the risk-benefit ratio of medications in pregnancy as well as fear of litigation if the decision to intervene is associated with a negative outcome.

In recognition of the challenges inherent in making treatment decisions in cases of pregnant women with depression, Wisner and colleagues developed a model for risk-benefit decision-making in pregnancy [6]. In addition to guiding the psychiatrist in structuring the presenting problem, the model recommends that risk for each option be discussed with the patient in five major groups: pregnancy loss, physical malformations, growth impairment, behavioral teratogenicity, and neonatal toxicity. The risk of the illness itself is also reviewed. The patient’s active participation, and that of her support team, is imperative, which allows for her values to inform the decision as well as assesses her capacity to give informed consent.

This clinically focused review is intended to support clinicians in shared decision-making and informed consent with patients regarding the treatment of unipolar depression and anxiety during pregnancy. We discuss the most recent evidence on the risks of anxiety and depression in pregnancy. Guideline-supported interventions for depression and anxiety will also be discussed with an emphasis on the risks and benefits of these treatments during pregnancy; treatments reviewed will include psychopharmacology, neuromodulation, and psychotherapy [7–9]. The discussion of risks for medication options will be grouped according to Wisner’s domains, with the addition of mother-specific risks, in order to facilitate collaborative decision-making with patients. Finally, risks and benefits of neuromodulation and psychotherapy will be reviewed more briefly.

Risk of Depression and Anxiety in Pregnancy

In order to fully appreciate the risks and benefits of the treatment options for depression and anxiety during pregnancy, it is essential to consider the risks of depression and anxiety themselves during pregnancy. One such risk is mortality from the illness, such as suicide or infanticide. Relapse is also of concern: although even women who continue medication during pregnancy risk relapse, the risk of depression relapse during pregnancy is increased in women who have had depressive symptoms within the preceding 6 months and discontinue antidepressants, in women with a lifetime history of more than 4 depressive episodes, or a history of moderate to severe pre-pregnancy depressive episodes [10, 11]. Though there is limited evidence that generalized anxiety disorder may actually improve throughout pregnancy [12], pregnancy may be a vulnerable time for development or worsening of other anxiety disorders. For instance, pregnant women have nearly doubled prevalence of obsessive-compulsive disorder (OCD) relative to the general population [13], a third of women with OCD will worsen during pregnancy [14], and 5 % will have onset during pregnancy.

There is also growing evidence that psychiatric illness during pregnancy carries its own somatic risk to the baby and mother. A recent large retrospective cohort study of pregnant veterans found an association between post-traumatic stress disorder (PTSD) and preterm birth [15]. This is consistent with data from another large cohort, this one prospective, that found that PTSD with comorbid depression was as strong a risk factor for preterm birth as a history of previous preterm birth (OR 4.08, 95 % CI 1.27–13.15) [16•]. Preterm birth and low birth weight were also associated with anxiety symptoms at any time in pregnancy in a meta-analysis of studies using interview or questionnaire diagnostic information [17]. Pregnant women with depressive symptoms are more likely to have preterm birth than non-depressed pregnant women, even in the absence of medications, and some, but not all, studies have associated depression with intrauterine growth restriction (IUGR) [18•, 19, 20]. A recent prospective study suggests that high levels of perceived maternal stress at 20 weeks’ gestational age correlates with increased risk for small for gestational age (SGA) at delivery even in babies who had been growing normally when measured at 20 weeks [21]. Additionally, mothers scoring in the highest range of depression and anxiety had an increased risk of SGA, especially with male infants. This study, however, does not address whether any of the patients were taking antidepressants. Several studies have shown that depression is itself a risk factor for preeclampsia, or maternal hypertension with proteinuria, which poses a serious risk to the mother [14, 22], and a recent prospective study revealed a statistically significant increased risk of hypertension during pregnancy in women with comorbid depression and anxiety [23].

There is also evidence that the offspring of mothers who suffer from depression during pregnancy are at increased risk for behavioral and emotional problems. Although maternal depression during pregnancy is a predictor of behavioral and temperament problems in children irrespective of antenatal medication exposure [24], many studies do not control for antidepressant exposure during pregnancy. One study evaluating the risk of depression and behavioral problems at age 5 found that the greatest risk is when children are exposed to high levels of maternal depression during the preschool years, but the risk of these symptoms was also increased in children exposed to depression from pregnancy through early childhood, highlighting the risk of allowing major depression to persist in and past the antenatal phase inadequately treated [25•]. This risk may extend even into adulthood: a retrospective cohort study recently indicated that children exposed to antenatal maternal anxiety, depression, and stress are at increased risk for behavioral problems and depression at age 21, even when adjusted for health habits and life events [26]. Notably, children of mothers with antenatal depression also appear to be at increased risk of maltreatment during childhood, the etiology of which appears to involve interplay between offspring temperamental changes resulting from in utero exposure to maternal depression and the depressed mother’s attachment and caregiving style [27]. In combination, maternal depression and childhood maltreatment significantly increase the risk of offspring adulthood depression.

Treatment Options

Antidepressants

In any discussion of the risks and benefits of medication, it should be noted that few studies examine the benefit of antidepressant treatment to the infant. One notable exception found that antidepressant treatment during pregnancy alleviated a cognitive change referred to as P50 gating in children of mothers with antenatal anxiety [28]. This underscores the need to study the positive as well as negative effects of psychotropic medication use during pregnancy. Further, the literature is complicated by multiple limitations, including lack of access to psychiatric histories of participants and associated poor health habits of pregnant women who take these medications relative to other women [29]. It may be difficult to ascertain the extent to which adverse events are related to the medication, the disease itself, or the health habits of the mother taking the medication.

Pregnancy Loss

In sum, available data suggest that there is a small increased risk of miscarriage for women taking any antidepressant, though this risk becomes less statistically significant when excluding studies of poor quality [29, 30•]. Moreover, most studies do not differentiate between induced and spontaneous abortion, so it could be that depressed women may be more likely to pursue pregnancy termination.

Regarding selective serotonin reuptake inhibitors (SSRIs), a recent large study indicated that SSRI exposure during pregnancy is associated with a very small increased risk of miscarriage in the first trimester (HR 1.08, 1.04–1.13), but not of second-trimester pregnancy loss [31•]. Notably, the risk of first-trimester miscarriage was lower in women who took SSRIs but were not identified as having a diagnosis of depression or anxiety than for women who carried a diagnosis but were not taking an SSRI, suggesting that the association may be due to the mental health disorder itself. Also of note, pregnancies in which an SSRI was used were also associated with unhealthier lifestyle profiles, an important potential confounder. Another large cohort study found that there was a small increased risk of miscarriage (HR 1.27, 1.22–1.33) in women who took SSRIs during the first 35 days of pregnancy, but that this was roughly equivalent to the increased risk of miscarriage in women who discontinued SSRIs 3–12 months prior to pregnancy (HR 1.24, 1.18–1.30); the risk of miscarriage was not related to dosing of SSRIs [32]. Study authors concluded that the risk of miscarriage is likely not to be modified by discontinuing SSRI prior to pregnancy.

Physical Malformations

Major congenital anomalies, especially cardiac defects, are of concern when a fetus is exposed to medication during organogenesis in the first trimester of pregnancy. The baseline against which the relative risk of major congenital malformations in this population must be considered is that of the general population: 2–3 % [33]. Given this, although there is conflicting evidence that there is an increased risk of such malformations with medication treatment, any increase in risk is likely to result in a very small absolute risk change. There is differing data by class of medication, so we will review SSRIs, serotonin-norepinephrine reuptake inhibitors (SNRIs), tricyclic antidepressants (TCAs), bupropion, and mirtazapine.

Most studies evaluating rates of congenital malformations overall following SSRI exposure find no elevated risk with medication use [34–36], with one well-controlled study as an exception (OR 1.3, 1.1–1.6) [37]. A recent large Bayesian analysis of registry data considering both specific anomalies and specific medications revealed an equivocally increased risk of neural tube defects following citalopram and ventricular septal defects (VSD) after fluoxetine [38]. More significant were associations between paroxetine and anencephaly, atrial septal defects, right ventricular outflow tract obstruction, gastroschisis, and omphalocele. The study authors were not able to control for maternal depression and noted that if causal, the associations result in a small absolute increased risk.

Older meta-analyses evaluating the risk of cardiovascular malformations have found a small but notable association with SSRI use during pregnancy with paroxetine, fluoxetine, and sertraline specifically implicated [34, 39–42]. More recently, a registry-based study including live births, fetal death after 20 weeks, and termination due to known genetic diagnosis found an increased risk of congenital heart defects, particularly severe defects (OR 1.56, 1.02–1.39), tetralogy of Fallot (OR 3.16, 1.52–6.58), and Ebstein’s anomaly (OR 8.23, 2.92–23.16) following SSRI exposure [43]. Importantly, though women with diabetes or taking anti-epileptic drugs were excluded, there was no adjustment for confounding variables, including psychiatric diagnosis or lifestyle behaviors.

Regarding SNRIs and TCAs, many recent studies show no association between these medications and major congenital malformations [34, 35, 44–46]. One Canadian database study did find an association between venlafaxine and a small number of noncardiac malformations [47], and two large older studies did find an increased association between cardiac malformations and tricyclic antidepressant use, particularly clomipramine [34, 48].

There is less data investigating bupropion and mirtazapine. In one well-controlled study, antenatal exposure to bupropion was associated with a higher risk of VSD relative both to controls and to women taking bupropion in combination with other antidepressants (OR 2.5, 1.3–5.0) [49]. A multicenter prospective cohort study of 357 pregnancies exposed to mirtazapine found no difference between overall rates of congenital anomalies relative to SSRI-exposed pregnancies; though there was an increased rate relative to controls, the rate was nonsignificant when given after the first trimester [50].

As previously indicated, confounders may play an important role in biasing study outcomes, and some of these confounders may be not yet fully identified. For example, a large Scandinavian cohort analysis found that although there was a small increased risk of cardiac and other birth defects in babies exposed to antenatal SSRIs or venlafaxine when controlled for maternal age, diabetes, teratogens, and smoking (OR 1.13, 1.06–1.20), this association disappeared in a sibling-controlled analysis, suggesting that familial and lifestyle factors may be important confounders [51]. Similarly, a large Medicaid data study found that despite an increased risk of cardiac defects across all antidepressant categories in an unadjusted analysis, when restricting to those mothers with depression and adjusting for known cardiac defect risk factors and proxies for depression severity, the increased risk vanished, from OR 1.25 down to 1.06 (0.93–1.22), including for fluoxetine, paroxetine, bupropion, and TCAs [52•].

In sum, there are conflicting data regarding the risk of congenital malformations, and in particular cardiac defects, following early trimester treatment with antidepressants. Importantly, paroxetine has been most consistently associated with major congenital anomalies. However, especially for the SSRIs, this appears to result in a small absolute risk increase and must be weighed against the risk of untreated depression or anxiety, which may be an important confounding factor in most studies and may confer independent risk for such malformations.

Growth Impairment

Being SGA or having IUGR puts the infant at risk for adverse health outcomes, including increased mortality. Low birth weight is defined as <2500 g, but if consistent with the infant’s gestational age, then the infant is not SGA. Being SGA can occur because of genetic programming (i.e., a small parent), but when pathological it is referred to as IUGR. IUGR is diagnosed when (a) an infant is less than the 10th percentile on the basis of gestational age, and (b) the abdominal circumference is less than the 2.5 percentile.

An older, large database study indicates that there is no effect of TCA use on birth weight overall, though the analysis for classification of SGA (i.e., 2 standard deviations below the expected weight) was of borderline significance (OR 1.37, 1.00–1.95) [34]. The same study found significant effects of SNRI use on low birth weight (OR 1.87, 1.37–1.64) and SGA (OR 1.84, 1.20–2.81). More recently, a review of studies with small data pools showed no relation between SGA babies and the use of TCAs during pregnancy [29]. Results have been much more mixed for the SSRIs, but the most rigorously conducted studies controlling for variables like mental health diagnoses find no association between SSRI use and SGA [53–56]. One small study found no association between use of bupropion and birth weight [57].

Of note, in a meta-analysis of antidepressant use during pregnancy, the difference of 74 g between infants exposed and not exposed to antenatal antidepressants was not statistically significant after controlling for mental health diagnoses [30•]. As above, antenatal depression in and of itself has been associated with IUGR, and stress with SGA, again underscoring the importance of exploring the contribution of the depressive or anxiety illness to such findings.

Behavioral Teratogenicity

As discussed previously, offspring of mothers with antenatal depression are at elevated risk for behavioral and emotional problems, even into adulthood. There is also concern that treatment with medication might lead to long-term cognitive and emotional consequences. Some studies have found that SSRI exposure does not affect IQ scores, language development, or behavioral problems [58–60]. A recent large cohort study did find a statistically significant adjusted association between language delay at age 3 and exposure to SSRIs in at least two time periods during pregnancy [61]. However, the same study relied on maternal reports of language competence, was not powered to distinguish between moderate and severe language impairment, and also found an association between language delay and exposure to short- and long-term symptoms of depression and anxiety during pregnancy. As such, and consistent with other similar positive findings concerning language and motor development, the contribution of medication versus illness as well as the clinical impact of findings are uncertain [62, 63].

Regarding behavioral problems, a prospective cohort study demonstrated a significant adjusted association between antenatal use of SSRIs and maternal reports of internalizing and anxious behaviors that was stable at child ages 3 and 6 [64]. Of note, there was no association with externalizing behavior in the same group. A similar association was also found with antenatal exposure to maternal depression and internalizing (especially anxious) and externalizing behavioral problems at child ages 3 and 6. These findings are consistent with another large cohort study, which included matched siblings discordant for antidepressant exposure [65]. In the adjusted, sibling-matched analyses, there was a small increase in the anxiety subscale of internalizing behaviors at 3 years old for children exposed to antenatal antidepressants (beta coefficient 0.64, 0.26–1.02). Of note, a maternal lifetime history of depression (reported at 17 weeks’ gestation) was also associated with internalizing behaviors in offspring in the adjusted, sibling-matched analysis (beta coefficient 0.19, 0.02–0.38). In concert, these data suggest that antidepressants may be associated with a small increase in internalizing, particularly anxiety, behavioral problems in children, but that genetic factors and maternal depression are also important contributors.

There is an emerging and complex database concerning the risk for autism spectrum disorder (ASD) following antenatal antidepressant treatment. Two case–control studies, both well-controlled but relying on retrospective reporting, found that women who took an SSRI during pregnancy were more likely to have a child diagnosed with ASD relative to women who did not take an SSRI during pregnancy [66, 67]. It is important to acknowledge that in their conclusion, the authors of the first study emphasized that the proportion of cases of ASD attributable to SSRIs was less than 3 % in their population, and the second study only found an increase in ASD with normal intellectual ability. Two other studies, both retrospective cohort analyses of registry data [68, 69•] and both controlling for confounders, did not find an association between SSRI use during pregnancy and ASD in children, though the latter did note an association with risk of attention-deficit/hyperactivity disorder (OR 1.81, 1.22–2.70). The mixed literature highlights the importance of considering maternal psychiatric illness as a contributing factor and suggests that if there is an effect of antidepressant treatment, it is likely quite small and relatively mild.

Neonatal Toxicity

In this domain, we consider the perinatal risks of exposure to medications, including preterm birth, persistent pulmonary hypertension of the newborn (PPHN), and poor neonatal adaptation syndrome (PNAS).

Preterm birth, defined as birth before 37 weeks’ gestational age, is associated with an increased risk of neonatal death and a host of adverse outcomes in child development, including intellectual, behavioral disorders, cerebral palsy as well as respiratory, ophthalmic, and auditory dysfunction; the cost associated with preterm birth in the United States in 2007 was over 26 billion dollars [70]. The baseline preterm birth rate in the United States is approximately 11 % [71]. Large database studies and meta-analyses have demonstrated that the use of TCAs, SSRIs, and SNRIs is associated with preterm birth, the highest risk being with the TCAs and the lowest risk with SSRIs, with no apparent risk difference among SSRIs [30•, 34, 38, 54, 72]. Generally, the literature indicates that prolonged exposure to SSRIs during pregnancy is associated with shorter gestational age. It must be noted that gestational length in many studies is shortened by only a few days and is of questionable clinical significance.

Interestingly, there is evidence from a very small study that SSRI treatment may in fact extend gestational age and raise birth weight relative to women who are not treated for depression; though these findings were significant by ANOVA testing, they were not by χ 2 testing, likely due to the small size of the study [73]. Nevertheless, as above, pregnant women with depressive symptoms are more likely to have preterm birth than are non-depressed women, even in the absence of medications, again highlighting the theme that the risk of medications is confounded by the associated risk of the disease state, which may or may not be mediated with treatment [18•, 20].

PPHN is a life-threatening condition in which pulmonary vascular resistance fails to decrease following delivery, resulting in ineffective oxygenation and potential respiratory failure [29]. Some studies have shown that the rate of PPHN in infants of mothers who took SSRIs after 20 weeks of pregnancy is above the average rate of 1–2 cases per 1000 births [34, 74, 75], most recently as 2.9–3.5 infants per 1000 births following third-trimester exposure to SSRIs [76]. However, not all studies find this association, and many other factors, like obesity, cesarean section, and smoking, all of which have elevated rates in depressed women, might account for an association between PPHN and SSRI use [29]. Most recently, a large Medicaid cohort study found that the odds ratio adjusted for depression for primary PPHN in women taking SSRIs was only slightly elevated and of borderline significance (OR 1.28, 1.01–1.64) and not significantly elevated for non-SSRI antidepressants [77•]. In accordance with this data, the FDA has withdrawn a 2006 black box warning regarding the risk of PPHN and antidepressants [78].

PNAS is marked by a broad constellation of physical and behavioral symptoms observed in 10–30 % of neonates exposed to antidepressants near delivery [79, 80]. Symptoms are most commonly associated with withdrawal from short half-life antidepressants, such as venlafaxine or paroxetine, or toxicity of long half-life medicines, like fluoxetine, but have also been associated with all antidepressants [81]. The risk of PNAS also appears to be dose-related. Specific symptoms include agitation, feeding difficulties, hyperreflexia, hypertonia, jaundice, respiratory distress, tremor, and vomiting [29]. Most of these symptoms are mild in nature and resolve within 2 weeks of delivery, although more severe symptoms may require special care. One recent study found that only 3 % of infants with PNAS demonstrated severe symptoms on a standardized PNAS scale [80]. Tapering the medication near to delivery will decrease the risk of PNAS, but may put the mother at risk for postpartum depression, which may outweigh concern for mild symptoms. Interestingly, there is only one case series evaluating the incidence of PNAS after exposure to mirtazapine; this found a significantly diminishment of PNAS in children who were partially or fully breastfed, raising a question as to whether breastfeeding might reduce risk following exposure to other antidepressants [82].

Maternal Risks

Beyond those risks associated with antidepressant treatment in any given individual, risks identified for the pregnant person include post-partum hemorrhage and gestational hypertension. Acute postpartum hemorrhage is of concern in treatment with SSRIs as these medications may be mild platelet inhibitors [83]. Four studies have examined women exposed to antidepressants proximate to delivery, two of which found an increased risk [34, 84] and two of which did not [85, 86]. As with PNAS, tapering antidepressants before delivery may reduce the risk, but the increased risk of bleeding is controversial, and doing so may lead to a postpartum relapse of depression.

Some studies have indicated an association between gestational hypertension and antidepressants [87, 88]. Currently, however, antidepressants do not appear to “cause” gestational hypertension [89], and a woman’s controlled hypertension is not a contraindication to antidepressant use during pregnancy. As above, depression in and of itself is a risk factor for gestational hypertension and preeclampsia.

Benzodiazepines

Benzodiazepines, which reduce anxiety symptoms through the stimulation of the gamma-Aminobutyric acid (GABA) receptor, are commonly used as treatment strategies for anxiety, including in pregnancy: one study found that 3.9 % of pregnant women were prescribed benzodiazepines, and 1.2 % a combination of antidepressant and anxiolytic [3]. Benzodiazepines also cross the placenta, and one recent study found a higher fetus to mother ratio of diazepam in women with diabetes and hypertension relative to healthy control patients [90]. The authors noted that this ratio did not correlate with change in Apgar scores and postulated that the increased ratio may be due to decreased protein binding and increased placental transfer in diabetes and hypertension.

Pregnancy Loss

Overall, there are very limited data on the risk of pregnancy loss following benzodiazepine exposure. A systematic review of the risks of benzodiazepines in pregnancy found elevated rates of spontaneous abortion in small studies exploring the effect on the fetus of intentional suicidal overdose with high-dose alprazolam, diazepam, medazepam, nitrazepam, and chlordiazepoxide [91]. However, an older small case series of women who were prescribed diazepam for hyperemesis gravidarum did not reveal a higher risk of spontaneous abortion [92].

Physical Malformations

Historically, first-trimester benzodiazepine use was thought to be associated with skeletal and oral cleft abnormalities and central nervous system dysfunction, and clinicians were encouraged to avoid first-trimester benzodiazepine use [93]. More recently, a systematic review identified that alprazolam, clonazepam, and diazepam were not associated with major malformations when recall biases were accounted for, though lorazepam was associated with higher rates of anal atresia in a small number of cases relative to other benzodiazepines [91]. Two large cohort studies, one British and the other Swedish, did not find any increased risk for major congenital malformations following antenatal exposure to benzodiazepines, even in combination with SSRIs [94, 95]. Contrary to physician concern, it is therefore unlikely that benzodiazepines significantly increase the risk of major congenital malformations, and these medications should not be avoided for this reason.

Growth Impairment

There are limited, and no recent, data regarding birth weight following benzodiazepine exposure. An older Swedish registry study found a small increased risk of low birth weight in both early- and late-pregnancy benzodiazepine exposure, though the rate was higher in late-pregnancy exposure (early, OR 1.30, 1.06–1.59; late, 1.89, 1.29–2.76) [96]. While the rates of SGA were also somewhat increased, they did not meet statistical significance, raising questions as to the clinical impact of this growth impairment.

Behavioral Teratogenicity

There have been no recent studies specifically examining the long-term neurodevelopmental consequences of antenatal benzodiazepine exposure. A systematic review published in 2014 did summarize several older studies that revealed mixed data regarding these consequences [97]. However, only one study evaluated children beyond 18 months, and this study did not find any effect on behavior at ages 9 to 10 [98].

Neonatal Toxicity

Swedish registry data suggest that benzodiazepine use during later trimesters does increase the odds of preterm birth (OR 1.87, 1.20–2.89) [99]. When benzodiazepines and SSRIs were both used, odds of preterm birth increased markedly (OR 3.43, 2.33–5.06) after adjustment for year of birth, maternal age, parity, smoking status, BMI, and number of previous miscarriages, but not for severity of mental illness, which may be a serious confounder.

Third trimester exposure to benzodiazepines may be associated with withdrawal symptoms in infants. Symptoms of benzodiazepine withdrawal in the neonate are usually not serious and easily manageable; they include mild sedation, hypotonia, reluctance to suck, apnea, cyanosis, and temperature dysregulation [100]. As with PNAS, discontinuing benzodiazepines during the third trimester may reduce this risk, but could worsen anxiety symptoms in the postpartum period.

Maternal Risks

There do not appear to be risks for the mother associated with benzodiazepine use that is unique to the pregnancy state.

Antipsychotics

Though primarily indicated in psychotic and bipolar disorders, antipsychotics are often prescribed as augmentation agents for difficult to treat depression and anxiety. In one British primary care cohort, not limited to pregnant patients, 57 % of quetiapine prescriptions were coded for diagnoses other than serious mental illness, of which 20 % were for anxiety and 35 % for depression [101]. It is important to note that there has been little systematic investigation into the teratogenic effects of antipsychotic medications on the fetus, and as in depression, many studies do not account for the risk posed by having a serious mental health disorder, which increases risk of a host of adverse outcomes [102]. Further, there are very limited data available on the use of long-acting injectable antipsychotics in pregnancy [103], and data are still emerging on newer second-generation antipsychotics, such as aripiprazole and paliperidone [104, 105].

Pregnancy Loss

There is no current evidence that antipsychotics increase risk of spontaneous abortion or intrauterine fetal death.

Physical Malformations

A comprehensive literature review found no association between use of first- or second-generation antipsychotics during pregnancy and major structural malformations [106]. Some recent studies have shown a small absolute increased risk for major malformations following second-generation antipsychotic exposure when controlling for risk factors for major malformations, but not for psychiatric illness (adjusted OR 2.17, 1.20–3.91, 9.3 % exposed vs. 6.4 % unexposed [107]; adjusted OR 1.44, 0.94–2.22, 6.1 % exposed vs. 4.5 % exposed [108]).

Growth Impairment

Much of the existing data regarding the association between antipsychotics and low birth weight are conflicting and ambiguous [108–112]. Atypical antipsychotics are well known to be linked to metabolic syndrome, which is of special concern in pregnancy in that diabetes during pregnancy is associated with infants who are large for gestational age (LGA). A recent Swedish registry-based study did find an association between antipsychotic use and large head circumference, but not LGA, in a group of patients prescribed either clozapine or olanzapine, after adjusting for risk factors for LGA, including early pregnancy BMI, smoking, and diabetes (OR 3.02, 1.60–5.71) [102].

Behavioral Teratogenicity

Data evaluating the long-term outcomes following antipsychotic exposure in pregnancy is currently lacking. One study indicated that babies exposed to antipsychotics antenatally score lower on a measure of child development at 2 months [109]. Another study suggested that infants exposed to antipsychotics in utero demonstrated impairment on a test of neuromotor development at 6 months [113].

Neonatal Toxicity

As of 2011, the antipsychotics carry an FDA label for risk of neonatal extrapyramidal symptoms at birth and withdrawal symptoms. This caution was based on a small number of safety reports to the FDA. A recent prospective cohort study demonstrated a 15 % rate of withdrawal symptoms, which some babies had for up to 6–8 weeks, and which was associated (though not statistically significant) with higher doses of antipsychotics [114].

Antipsychotics have also been associated with a higher rate of preterm birth and shortened gestational age in multiple studies [104, 108, 115, 116], though a recent large cohort study did not find that the association was statistically significant after adjusting for confounding risk factors [117]. Of note, a prospective cohort study comparing women treated with typical versus atypical antipsychotics with controls found that the rate of preterm birth was increased only for the group treated with typical antipsychotics [107]. The authors did observe, however, that there was a higher incidence of treatment with antidepressants in the typical relative to the atypical antipsychotic cohort, which may confound the results.

That said, a recent large meta-analysis, including 6289 antipsychotic-exposed and 1,618,039 unexposed pregnancies, found an increased rate of preterm birth (absolute risk increase 0.03 %), low birth weight (weighted mean difference of 57.89 g), SGA (absolute risk increase 0.05 %), and no increased risk of LGA [118]. It must be emphasized that the authors noted that limited adjustment was possible given the studies included.

Maternal Risks

As previously indicated, antipsychotic exposure during pregnancy has also been associated gestational diabetes, in addition to cesarean section [102, 108]. The risk of gestational diabetes attributable to antipsychotics is controversial. A recent large matched cohort study did not find an increased risk of gestational diabetes, hypertensive disorders, or venous thromboembolism [117], while a recent descriptive, prospective Australian cohort study found a moderately increased incidence of gestational diabetes and caesarean section in women treated with antipsychotics relative to the national averages, even when adjusted for DSM-IV diagnoses [104].

Neuromodulation

Much of the specific benefit of neuromodulation in pregnancy is the ability to avoid or minimize psychotropic medication and the teratogenic effects thereof. Although ECT does require short-acting sedatives and paralytics [119, 120], hospitalization for initiation of ECT can and should be considered as a treatment option for severe or emergent cases of depression in pregnancy [121]. A recent review suggests that the overall chance of a transient, non-life-threatening adverse event is 10 % for ECT during pregnancy, with the most common events being vaginal bleeding, uterine contractions, abdominal pain, and fetal arrhythmias [122]. Fetal death, major congenital malformations, and preterm birth appear to occur at rates similar to the psychiatric population not receiving ECT.

Vagal nerve stimulation (VNS), the stimulation of the left vagus nerve via surgical placement under anesthesia of a small pulse generator, has been FDA-approved for treatment-resistant depression. A review of case reports suggests that this may be viable alternative to medication, though little is known about the specific risk of VNS during pregnancy [122]. Transcranial magnetic stimulation (TMS) refers to the application of focal magnetic pulses to the dorsolateral prefrontal cortex for 15–45 min a day, 5 days a week, for 4–6 weeks. Case series have shown feasibility and safety in the treatment of depressive symptoms in pregnancy, but randomized controlled trials are needed for efficacy [122]. Supine hypotensive syndrome, occurring when the gravid uterus compresses the inferior vena cava (IVC) as the pregnant woman is supine, may cause symptoms of pallor, dizziness, hypotension, and tachycardia that resolve with position change [123]. For TMS devices that require the patient be in the supine position, the supine hypotensive syndrome may be prevented by positioning the patient with a right-sided wedge cushion to reduce the pressure on the IVC. Other risks of TMS in pregnancy have yet to be elucidated; though in one follow-up study, the offspring of women treated with TMS during pregnancy were perceived by their mothers to have a language delay, no objective delay was documented similar to that observed in the offspring of untreated mothers [124]. Other major drawbacks to TMS include the time and financial costs of treatment, though treatments are now covered by some health insurances.

Psychotherapy

A major appeal of psychotherapy is the ability to avoid pharmacotherapy, and it is perhaps for this reason that perinatal women tend to prefer psychotherapy to medication [125]. Psychotherapy has been recommended as the first-line treatment for women with mild illness during pregnancy and psychotherapy plus medication for women with moderate to severe illness [126]. Interpersonal therapy, a time-limited therapy focusing on relational contributors to mental health, has nearly two decades worth of research supporting its efficacy in the treatment of antepartum depression across a range of sociodemographic groups as well as individual and group settings [127]. More recently, cognitive behavioral therapy has also been identified as effective for both the treatment and prevention of perinatal depression, with individual treatment more effective than group treatment [128]. Commonly cited disadvantages to psychotherapy during pregnancy include time, transportation, and lack of access to childcare, as well as instances in which a woman’s illness is so severe that she is unable to fully participate in and benefit from psychotherapy [129].

Conclusions

In summary, the treatment of depression and/or anxiety during pregnancy requires a careful and collaborative weighing of the risks and benefits of treatment options versus the illness itself by both patient and provider. Medications for depression and anxiety can be prescribed during pregnancy, but may carry with them small risks of pregnancy loss, physical malformations, growth impairment and pre-term birth, neonatal toxicity, behavioral teratogenicity, and risks to the mother unique to pregnancy. Other factors, including the severity of illness, the patient’s functional impairment, the risk of illness relapse, and the patient’s preferences, must also be taken into consideration for what may be the patient’s first parenting decision. To further complicate matters, interpreting the broad and conflicting available database is limited by parsing out the individual contributions of treatment and illness. Although there is an extensive literature on the risks of antidepressants in pregnancy, there are as yet no randomized control trials evaluating the efficacy of these medications in pregnancy, and the data on the risks of antipsychotics and benzodiazepines are globally lacking.

We encourage clinicians to adapt Wisner’s model to facilitate this complex and fraught decision-making and informed consent process. We also offer the following specific recommendations regarding treatment informed by this review and our clinical experience:

  1. 1.

    Clinicians should tailor recommendations to the patient’s fears, concerns, preferences, and goals, which should be elicited as part of the history-taking process. Equal weight should be given to the benefit–risk profile of the treatment and that of not treating the illness, including relapse, suicide, and infanticide.

  2. 2.

    If a patient prefers to avoid medication, psychotherapy is most appropriate as primary treatment for mild to moderate illness and neuromodulation for more severe or life-threatening illness.

  3. 3.

    Clinicians may first decide what medication they would recommend if the patient were not pregnant, including medications the patient is currently taking. This helps to clarify the thought and decision making process.

  4. 4.

    It is important to document all exposures to medications and other substances with the potential to impact fetal development as well as to choose medication with the greatest amount of research regarding efficacy and adverse effects in this population [130].

  5. 5.

    Clinicians should aim to prescribe the lowest dose possible while ensuring that the dose is significant enough to achieve a tangible therapeutic effect [131]. Otherwise, the fetus risks exposure to both disease state and medication.

  6. 6.

    Similarly, when possible, monotherapy is preferable to polytherapy [132].

  7. 7.

    The tapering of medication in the weeks prior to delivery is not recommended given the little evidence that it will alleviate the risk of neonatal toxicity or withdrawal and may induce a worsening or relapse of the patient’s symptoms in the postpartum period.

  8. 8.

    Collaboration with an accepting and knowledgeable obstetrician is paramount, particularly as some women may require alternative monitoring during pregnancy.

  9. 9.

    Because pregnant women with psychiatric illnesses have an elevated rate of unhealthy lifestyle behaviors, and these behaviors may influence pregnancy outcomes, a routine part of psychiatric care should also include support, guidance, and counseling on smoking cessation, healthy eating, and exercise.

  10. 10.

    It may be helpful to remind oneself and the patient that no pregnancy is “risk-free,” but an emotionally healthy mother is more resilient and more able and available to deal with any unexpected outcomes.