Systematic Review: Antipsychotic Medication in Pregnancy and Neurodevelopmental Outcomes in Children

Abstract

Objective:

Antipsychotic medications are widely prescribed, including during pregnancy, and pregnant individuals worry about the potential sequelae for the child. Although antipsychotics do not seem to be teratogenic, the long-term neurodevelopmental impact of prenatal exposure remains unclear. A systematic review was conducted to determine if intrauterine antipsychotic exposure increases the risk of adverse neurodevelopmental outcomes.

Method:

A systematic search was performed in MEDLINE, Cochrane, Embase, and PsycINFO for studies published before September 7, 2024. We included original studies assessing cognitive, motor, behavioral, social, and psychiatric outcomes in children prenatally exposed to antipsychotics, excluding case reports, reviews, preclinical studies, and studies without a control group. Quality and risk of bias were assessed using the Newcastle-Ottawa Scale.

Results:

Of 1,349 studies identified, full text of 56 was screened, and 16 were included in the review. The number of exposed participants ranged from 11 to >15,000. In the 8 studies assessing motor development, early motor delays were observed but did not persist into later childhood. Neurodevelopmental disorders were assessed in 7 studies. Crude estimates showed greater risk in exposed children, but after adjusting for confounders, most studies found no significant risk. The mean NOS score was 7.1.

Conclusion:

Transient motor delays may be associated with antipsychotic use during pregnancy, although future studies adjusting for confounding factors should clarify this risk. After adjustment for confounders, the risk of neurodevelopmental disorders in school-age children does not seem to be increased. Studies with longer follow-up time are required to further investigate the risk of neurodevelopmental disorders.

Study registration information:

Neurodevelopmental Consequences of Antenatal Exposure to Antipsychotic Medication: A Systematic Review and Meta-Analysis; https://www.crd.york.ac.uk/PROSPERO/view/CRD42024499352.

Use of medications in the antipsychotic (AP) class has increased over the last several decades,¹ including the use by many women of child-bearing age. Due to their expanding range of indications, as well as their relatively rapid onset of effect and limited potential for acute toxicity compared with alternatives, there is a trend in increased use of AP medication during pregnancy, as well as off-label use.^2–4 However, many unanswered questions about safety and efficacy of prenatal use of drugs of this class remain. Because there are many ethical and practical limitations to conducting pharmaceutical trials in pregnant women, our knowledge base regarding this special population is necessarily limited and incomplete, relying heavily on observational rather than interventional data and on extrapolation from studies in nonpregnant populations and animal data.

Reassuringly, congenital malformations do not appear to be a major concern with APs. A landmark 2016 study⁵ of 1.3 million pregnant participants, including more than 9,000 taking APs, was the first clinical cohort study of sufficient size to show that the risk of congenital malformations was not meaningfully increased. Since then, additional studies^6,7 have further bolstered clinician confidence and resulted in increased prevalence of AP use during pregnancy.⁸

Beyond congenital malformations, however, many important safety concerns for pregnant patients remain poorly explored. Fetal exposure is highly individually variable but can be significant, with numerous reports of fetal serum levels higher than maternal levels.^9,10 Because APs cross the placenta and the blood-brain barrier, and due to their effect on dopamine and serotonin receptors, potential long-term neurodevelopmental, cognitive, and behavioral effects are of particular concern, as crucial stages of brain development occur in the fetal period.^11–17 Research in this area is complicated by the well-established adverse associations between maternal mental illness and child developmental outcomes, an issue known as confounding by indication.¹⁸ Therefore, adequate controls for confounding by indication are both critical for accurate results and reporting and extremely complex to develop and apply.¹⁹ A clear and detailed understanding of the relative risks and benefits conferred by maternal mental illness and pharmacological treatment, respectively, remains elusive. In this article, we offer a conceptual overview of the potential pathways by which antenatal AP exposure may affect brain development, combined with an updated systematic review of available clinical cohort studies investigating behavioral, socioemotional, psychomotor, and neurocognitive outcomes in children with prenatal exposure to AP medication.

Potential Mechanisms for Developmental Effects

The first medications in the AP class, the phenothiazines and butyrophenones, were developed in the 1950s.²⁰ Sometimes termed first-generation or typical APs, their common mechanism is dopamine receptor blockade.^21,22 Especially at higher doses, these medications put patients at risk of motor side effects, including parkinsonism, akathisia, and dyskinesia. A search for effective alternatives without the disabling motor side effects yielded a second generation of so-called atypical APs. For a majority of atypical APs, adverse motor side effects are indeed attenuated, but most carry an increased risk of metabolic side effects, including obesity, hyperlipidemia, and impaired glucose tolerance.²³ Notably, the mechanisms of action and side-effect profiles of APs differ markedly from drug to drug, and the distinction between first- and second-generation APs does not encapsulate all differences between agents. There are several pathways through which APs have the potential to influence long-term neurodevelopment. For example, potential effects of dopaminergic blockade in the fetal brain includes effects on proliferation and differentiation of neural progenitor cells. In addition, dopamine is a regulator of differentiation and migration of cortical GABAergic interneurons,^13,14 which could alter the balance of excitation and inhibition in the cortex. Metabolic disruption caused by APs can have adverse effects on the developing central nervous system, eg, via fetal hyperglycemia and oxidative stress or epigenetic effects due to the imbalanced glucose levels in the uterine environment.²⁴ Some APs have an affinity for serotonergic, cholinergic, or adrenergic receptors, which may also influence long-term neurodevelopment.

Neurodevelopmental Consequences Related to Severe Maternal Mental Illness

In addition to both direct effects of AP exposure and genetic susceptibility to schizophrenia and bipolar disorder that may be passed from mother to child, women with severe mental illness are at increased risk of numerous adverse obstetrical and fetal outcomes. Proposed mechanisms include epigenetics, microbiome differences, lifestyle factors, and metabolic health associated with severe maternal mental illness. Previous studies have found that women with severe mental illness are at increased risk of outcomes such as cesarean or instrumental delivery,²⁵ preterm birth,²⁶ low birth weight,²⁵ and neonatal intensive care unit admission²⁷ and thus neurodevelopmental consequences related to these outcomes.^28–30 Hence there are numerous pathways aside from direct medication effects by which maternal mental illness may be associated with adverse neurodevelopmental outcomes in children, underlining the importance of careful controls in all observational research.

Systematic Review of the Literature

Here, we systematically review and summarize the current literature investigating AP use during pregnancy and child neurodevelopmental outcomes to provide physicians and patients alike with updated information about the risks associated with use of AP medication during pregnancy.

METHOD

Search Strategy

A systematic search was performed by a trained librarian in the following databases: MEDLINE, Cochrane, Embase, and PsycINFO to identify studies that investigated neurodevelopmental outcomes of human offspring prenatally exposed to any AP medication. No restrictions were made regarding the language or the year of publication. The initial search was conducted on February 13, 2024. Keywords included words such as antipsychotic, specific drug names, development, behavior, and prenatal exposure. The exact search terms used in each database can be found in Supplement 1, available online. An additional final search was conducted using identical search terms on September 6, 2024, to capture any newly published work. The studies included here represent all studies that met inclusion criteria from both searches.

Study Selection

Study selection criteria were determined before initiating searches, and the review was registered in PROSPERO (ID: CRD42024499352). Studies were eligible for inclusion if they were original, peer-reviewed research that assessed human offspring of any age who were prenatally exposed to any AP medication, regardless of maternal indication for use. Animal studies and postmortem tissue studies were excluded. In a deviation from our preregistration, studies without a control group were also excluded. Although we did not expect to identify studies without control groups, we found 2 such studies and excluded them as they do not fully address the question of interest. Case reports, case series, systematic reviews, and book chapters were not eligible for inclusion. Studies with any neurodevelopmental outcome including cognitive, motor, communication, behavioral, and social development were included. Two reviewers (C.A.K. and E.M.P.P.) independently screened the title and abstract of all studies identified in the initial search. Full text articles were then assessed for eligibility based on the above-mentioned inclusion and exclusion criteria. In cases of disagreement, the reviewers met to discuss the conflicts and were able to reach consensus on all studies. Covidence software was used throughout the screening and data extraction process.³¹

Data Extraction

For each included study, 2 authors independently extracted data. Four authors participated in data extraction (C.A.K., A.-S.R., E.M.P.P., T.R.), and studies were divided among them such that each study was evaluated by 2 of the 4 authors. Data on study design, country of study, number of exposed cases and unexposed controls, control type, medication, dose, maternal diagnoses, exposure period, follow-up time, outcome measure, and results were extracted (Tables 1–4; Table S1, available).

TABLE 1.

Characteristics and Results of Studies That Assessed Motor Outcomes

Authors, year, reference	Study design	Country	Exposed (n)	Controls (n)	Control description	Medication	Maternal diagnosis	Exposed period	Follow-up time after birth	Measurements	Results	NOS
Auerbach et al., 199233	Cohort	Israel	11	H: 26 MI: 18	Healthy controls recruited from public well-baby clinics; control group with psychiatrically ill mothers recruited from well-baby clinics and psychiatric clinics, with diagnosis based on interviews and psychiatric records	Chlorpromazine, fluphenazine, perphenazine, thioridazine, trifluoperazine	Schizophrenia (n = 7), MDD (n = 1), bipolar disorder (n = 1), affective disorder (n = 1), antisocial personality disorder (n = 1)	All exposed in third trimester, majority exposed throughout pregnancy	3 and 14 days	Neonatal Behavioral Assessment Scale Abstinence syndrome score	Significantly higher abstinence scores among exposed infants at day 3 (β = .36, t = 2.62, p < .05) and day 14 (β = .362, t = 3.01. p < .005) after accounting for confounders	7
Hurault-Delarue et al., 201639	Register-based	France	70	H: 32,303	General population	APs/neuroleptics (ATC class N05A); most common were chlorpromazine, cyamemazine, and sulpiride	NR	At least 1 dispensation of psychotropic drugs during second and/or third trimester of pregnancy	9 and 24 mo	14-item checklist related to psychomotor development completed by general practitioner or pediatrician	Abnormal motor development among exposed at 9 mo, but no difference at 24 mo	6
Johnson et al., 201234	Prospective cohort	USA	22	32 H 53 no medication with diagnosis; 202 antidepressant medication control	No psychotropic exposure with MI diagnosis, antidepressant exposure	Haloperidol, aripiprazole, olanzapine, quetiapine fumarate, risperidone, ziprasidone	Exposed: 11 (50%) depressive disorder, 10 (45.5%) bipolar disorder, 16 (72.2%) anxiety disorder, 2 (9.1%) psychotic disorder	First trimester (n = 12), second trimester (n = 14), third trimester (n = 16) Median number of weeks exposed: 32	6 mo	INFANIB assessing neuromotor functioning; infant habituation	Exposed infants had lower INFANIB scores than antidepressant and no medication groups: 71.5% transiently abnormal vs 62.1% in antidepressant group and 50.0% in unexposed; 9.5% sustained abnormal vs 6.0% in antidepressant group and 0% in unexposed; (p < .01)	6
Mortensen et al., 200340	Register-based	Denmark	63	H: 755	General population	Any neuroleptic	NR	Any point during pregnancy	7–10 mo	Boel test for psychomotor development	Exposed babies had increased odds of abnormal Boel test at 7–10 mo compared with controls (adjusted OR 4.1 [95% CI 1.3–13.0])	7
Peng et al., 201335	Prospective cohort	China	76	H: 76	Healthy controls, no mental illness	Clozapine, risperidone, sulpiride, olanzapine, quetiapine	Schizophrenia	NR	2, 6, and 12 mo	BSID-III	Exposed infants at 2 mo were more likely to have delayed development in motor domain; there was no difference at 6 and 12 mo	5
Platt et al., 198936	Prospective cohort	USA	192	H: 99 MI: 116	Psychotic/neurotic parent; no drug exposure	Any antipsychotic neuroleptic	Any psychotic/ neurotic	>7 days exposure within any month of pregnancy	7 y	Motor development as measured by Bayley Scales, Infant Behavior Profile, and pediatric neurological examination; assessed at newborn, 8 mo, 4 y, 7 y	Increase in abnormal motor activity during newborn period. At 8 mo, trend toward increased failure on Bayley gross motor assessment (RR = 2.08, p < .07) No difference at 4 or 7 y	7
Schrijver et al., 202337	Cohort	Netherlands	17	MI: 74	Women with mental illness treated at perinatal psychiatry centers, but unexposed to antipsychotic medications	Quetiapine, haloperidol, pipamperone, flupentixol, olanzapine, penfluridol, bromperidol	Among exposed: bipolar I disorder (n = 7), schizoaffective disorder (n = 1), borderline personality disorder (n = 4), schizophrenia (n = 4), OCD (n = 1)	Anytime during pregnancy: first D second trimester (n = 1), second D third trimester (n = 2), second trimester (n = 1), whole pregnancy (n = 13)	6–14 y	Categories and Mosaics subtests from SON-R 6–40 9 subtests from NEPSY-II-NL Learning problems and psychiatric diagnosis based on parental report (CBCL)	No association between exposure and outcome on sensorimotor NEPSY subtest	6
Swetlik et al., 202441	Register-based	USA	178	MI: 174	Maternal exposure to psychiatric medication other than APs, majority of control group exposed to SSRIs	Second-generation APs (most commonly quetiapine, lurasidone, aripiprazole)	Bipolar disorder (64.4%), schizophrenia (1.1%), depression (17%), anxiety (4%)	Any point during pregnancy	9 mo to 6 y	Structured parent report (ASQ and CBCL)	No difference in odds of a fail score on ASQ-3 fine motor and gross motor domains	6

Note: AP = antipsychotic; ASQ = Ages and Stages Questionnaire; ATC = Anatomical Therapeutic Chemical; BSID-III = Bayley Scales of Infant and Toddler Development, Third Edition; CBCL = Child Behavior Checklist; H = healthy; HCl = hydrochloride; INFANIB = Infant Neurological International Battery; MI = mentally ill; MDD = major depressive disorder; NOS = Newcastle-Ottawa Scale; NR = not reported; OCD = obsessive-compulsive disorder; RR = relative risk; SON-R 6–40 = Snijders-Oomen Nonverbal Intelligence Test Revision; SSRI = selective serotonin reuptake inhibitor.

TABLE 4.

Characteristics and Results of Studies That Assessed Other Neurodevelopmental Outcomes

Authors, year, reference	Study design	Country	Exposed (n)	Controls (n)	Control description	Medication	Maternal diagnosis	Exposed period	Follow-up time after birth	Measurements	Results 1	NOS
Peng et al., 201335	Prospective cohort	China	76	H: 76	Healthy controls, no < mental illness	Clozapine, risperidone, sulpiride, olanzapine, quetiapine	Schizophrenia	NR	2, 6, and 12 mo	BSID-III	Exposed infants at 2 mo were more likely to have delayed development in cognitive, social-emotional, and adaptive behavior domains No significant differences between the 2 groups in all mean composite scores of BSID-III at 6 and 12 mo of age	5
Schrijver et al., 202337	Cohort	Netherlands	17	MI: 74	Women with mental illness treated at perinatal psychiatry centers, but unexposed to APs	Quetiapine, haloperidol, pipamperone, flupentixol, olanzapine, penfluridol, bromperidol	Among exposed: bipolar I disorder (n = 7), schizoaffective disorder (n = 1), borderline personality disorder (n = 4), schizophrenia (n = 4), OCD (n = 1)	Anytime during pregnancy: first + second trimester (n = 1), second + third trimester (n = 2), second trimester (n = 1), whole pregnancy (n = 13)	6–14 ys	Categories and Mosaics SON-R 6–40 Nine subtests from NEPSY-II-NL Learning problems and psychiatric diagnosis based on parental report (CBCL)	No association between exposure and child IQ scores, NEPSY subtest, or psychiatric diagnosis	6
Swetlik et al., 202441	Register-based	USA	178	MI: 174	Maternal exposure to i psychiatric medication other than APs, majority of control group exposed to SSRIs	Second-generation APs (most commonly quetiapine, lurasidone, aripiprazole)	Bipolar (64.4%), schizophrenia (1.1%), depression (17%), anxiety (4%)	Any point during pregnancy	9 mo to 6 y	Structured parent report (ASQ and CBCL)	No significant differences in odds of a fail score on all ASQ domains, after correction for confounding variables. No significant difference in odds of an abnormal clinical score on CBCL after correction for potential confounding variables	6
Johnson et al., 201234	Prospective cohort	USA	22	32 H 53 no medication with diagnosis 202 antidepressant medication control	No psychotropic I exposure with mental illness diagnosis, antidepressant exposure	Haloperidol, aripiprazole, olanzapine, quetiapine fumarate, risperidone, ziprasidone	Exposed: 11 (50%) depressive disorder, 10 (45.5%) bipolar disorder, 16 (72.2%) anxiety disorder, 2 (9.1%) psychotic disorder	First trimester (n = 12), second trimester (n = 14), third trimester (n = 16) Median number of weeks exposed: 32	6 mo	INFANIB assessing neuromotor functioning; infant habituation	No difference in habituation between groups	6

Note: AP = antipsychotic; ASQ = Ages and Stages Questionnaire; BSID-III = Bayley Scales of Infant and Toddler Development, Third Edition; CBCL = Child Behavior Checklist; H = healthy; INFANIB = Infant Neurological International Battery; MI = mentally ill; NOS = Newcastle-Ottawa Scale; NR = not reported; OCD = obsessive-compulsive disorder; SON-R 6–40 = Snijders-Oomen Nonverbal Intelligence Test Revision; SSRI = selective serotonin reuptake inhibitor.

Assessment of Risk of Bias and Study Quality

As with data extraction, ratings of methodological quality and risk of bias were assessed by 2o reviewers independently for each article (C.A.K., A.-S.R., E.M.P.P, T.R.) using the Newcastle-Ottawa Scale (NOS),³² which rates studies based on selection, comparability, and outcome criteria on a total scale from 0 to 9. A score of 9 represents studies with the least risk of bias. In the selection domain, authors reviewed Method sections of articles to determine whether cohorts were recruited selectively or unselectively, and control groups were considered “drawn from the same population as the exposed group” if they compared children of mentally ill mothers with offspring exposed to APs. The ascertainment of exposure domain was awarded a point if AP use was confirmed by either medical records or structured interview. All studies received a point for demonstrating that the outcome of interest was not present before the study, as there is no possible presence of neurodevelopmental outcomes in offspring before birth. To achieve full points in the comparability domain, either matching or adjusting for maternal smoking and at least 1 measure of socioeconomic status or education was required. These confounders were chosen because they are associated with maternal mental illness and also known to affect child outcomes. A point was awarded for assessment of the outcome if either an independent assessment was conducted by a researcher or a health professional or the outcomes were derived from medical records rather than by maternal report. Follow-up length was deemed appropriate based on whether it is feasible to identify the outcome measured at the age at which the child was followed up. Finally, follow-up of the cohort was deemed adequate if at least 80% of the cohort was retained in the study or the authors provided a description of those lost to follow-up. Any disagreements between the reviewers were resolved in discussions.

RESULTS

The study selection process is described in Figure 1. Our initial search yielded 1,315 articles, of which 83 were duplicates. During title and abstract screening, 53 articles were included for full text screening. After full text screening, 14 articles were included for data extraction. In our follow-up search for articles published between February 13, 2024, and September 6, 2024, 34 additional articles were identified. During title and abstract screening, 3 articles were included for full text review, and 2 additional articles were ultimately included for data extraction. See Table S2, available online, for a list of excluded studies and reasons for exclusion.

FIGURE 1. PRISMA Flow Diagram.

Note: Please note color figures are available online.

A total of 16 studies were included in the systematic review, including 6 cohort studies^33–38 and 10 register-based studies.^39–48 In 11 studies, a control group with maternal mental illness was included in at least 1 of their analyses. These control groups varied with some including any maternal mental illness, whereas others used an AP discontinuation control, and still others used non-AP psychiatric medication use as a control. Five studies included only a general or healthy maternal population control. In 13 studies, exposure to any type of AP medication was assessed as 1 group, although some of these studies included subgroup analyses separating typical and atypical APs.^43,45–48 Three studies specifically assessed exposure to second-generation AP only.^35,37,41 Additionally, 3 studies conducted secondary analyses limiting their exposure to specific individual APs.^42,46,48 Studies generally included exposure at any point during pregnancy, without limiting to a specific trimester of exposure, although 8 studies did limit exposure timing or conducted sensitivity analyses limited by timing of exposure.^33,42–48 Results for all studies are summarized in Table S1, available online.

Motor Development

Measures of motor function were assessed in 8 studies,^{33–37, 39–41} ranging from newborn assessments up to 14 years of follow-up (Table 1). Six studies assessed motor outcomes in the first year of life, all of which found some degree of motor abnormality or delay compared with controls.

In general, studies performed at earlier ages were more likely to identify differences, potentially related to neonatal abstinence, between AP-exposed and comparison infants than studies performed or reassessed at later ages. Neonatal abstinence syndrome is a clinical manifestation observed in neonates following withdrawal of an intrauterine drug exposure. Symptoms typically include abnormal movements, agitation, change in muscle tone, tremor, and difficulty breathing or feeding.^49,50 Auerbach et al.,³³ for example, reported higher abstinence scores, as measured by general tonus, tremulousness, and motor maturity, among exposed infants at 3 and 14 days old compared with unexposed controls. Platt et al.³⁶ were the first to report increased abnormal motor activity in the newborn period among 192 AP-exposed infants compared with 216 unexposed infants, as assessed by the Bayley Scales of Infant and Toddler Development.⁵¹ They continued to report a trend toward increased failure in exposed babies on gross motor assessment at 8 months, but found no difference at follow-up at 4 and 7 years Similarly, Peng et al.³⁵ followed AP-exposed babies from 2 months to 1 year, and although there were significant differences in mean motor scores at 2 months, the difference was no longer significant at 1 year old. Hurault-Delarue et al.³⁹ mirrored these results, reporting a motor delay at 9 months based on a physician examination and psychomotor checklist that was no longer seen at 24 months.

When comparing 63 exposed infants with 755 infants of mothers who had not redeemed any prescriptions for psychotropic or antiepileptic drugs during their pregnancy, Mortensen et al.⁴⁰ found increased odds of an atypical Boel test for psychomotor development at 7 to 10 months of age among AP-exposed infants.⁵² However, this difference may be driven by the portion of the test related to hearing, on which 14% of exposed infants had atypical results vs 4% of the unexposed infants. When excluding the hearing-related portion of the test, only 3% of exposed infants had abnormal results compared with 1% of unexposed infants. Similarly, at 6 months of age, Johnson et al.³⁴ reported significantly lower scores on the Infant Neurological International Battery (INFANIB), an infant neuromotor examination, among AP-exposed children than children with or without antidepressant exposure. Yet, they also reported significant effects of maternal mental illness on motor scores, making results difficult to interpret. In studies that included longer follow-up periods,^35,37,39,41 ranging from 1 to 14 years, no differences were noted in adjusted analyses between AP-exposed children and control groups.

Neurodevelopmental and Psychiatric Disorders

Registered neurodevelopmental or psychiatric diagnoses were assessed as primary outcomes in 6 studies (Table 2).^42–47 These were primarily large, register-based studies using AP medication dispensation or prescription during pregnancy as exposure and registered diagnoses in medical records to assess results. Wang et al.⁴⁷ and Halfdanarson et al.⁴³ examined results specifically for attention-deficit/hyperactivity disorder (ADHD) and autism spectrum disorder diagnoses, whereas Petersen et al.,⁴⁵ Momen et al.,⁴⁴ Straub et al.,⁴⁶ and Bruno et al.⁴² included any neurodevelopmental diagnosis, based on diagnostic codes in medical records. One additional study by Schrijver et al.,³⁷ a smaller cohort study, included any neurodevelopmental diagnosis based on maternal report rather than registered diagnosis.

TABLE 2.

Characteristics and Results of Studies That Assessed Neurodevelopmental Disorders

Authors, year, reference	Study design	Country	Exposed (n)	Controls (n)	Control description	Medication	Maternal diagnosis	Exposed period	Follow-up time after birth	Measurements	Results	NOS
Bruno et al., 202442	Register-based	Denmark, Finland, Iceland, Norway, Sweden	11,626	MI: 197,296	Any psychiatric condition based on ICD codes	Any AP: most commonly quetiapine, olanzapine, prochlorperazine, perphenazine, aripiprazole, levomepromazine	Among exposed: psychotic or bipolar disorders (5,049; 43.4%), other psychiatricdisorders (5,984;51.5%), psychiatric disorders recorded by prescription reimbursement (593; 5.2%)	At least 1 prescription filled for any AP anytime in pregnancy	3–19 y (median follow-up 6.7y)	Neurodevelopmental disorder based on ICD codes and poor academic performance (bottom 25th percentile) on national standardized test	No class effect in most fully adjusted analyses for neurodevelopmental disorders. Possible increased risk of neurodevelopmental disorders in sibling analysis aHR: 1.52 (95% CI 1.00–2.30) Possible relation with neurodevelopmental disorders for chlorpromazine	9
Halfdanarson et al., 202243	Register-based	Denmark, Finland Iceland, Norway, Sweden	15,466	H:4,308,620 MI: 8,356	Full study is all unexposed, sensitivity analysis with maternal psychotic and bipolar diagnosis	Typical and atypical APs	In exposed group: psychotic or bipolar disorders (n = 4,385), other psychiatric disorders (n = 5,729), no recorded psychiatric conditions (n = 5,352)	At least 1 AP prescription during pregnancy (from LMP to birth): first trimester only (n = 6,478), second or third trimester only (n = 3,349), whole pregnancy (n = 5,639)	Median follow-up: 9.8 y (25th to 75th percentile 5.5–14.3 y)	ADHD and ASD diagnoses based on ICD code in chart	No significant difference in either ADHD aHR 1.10 (95% CI 1.00–1.21) or ASD aHR 1.12 (95% CI 0.97–1.29) in adjusted analysis Decreased risk of outcomes when comparing exposure during pregnancy vs pre-pregnancy use only; fully aHR 0.74 (95% CI 0.62–0.87) for ADHD and aHR 0.88 (95% CI 0.70–1.10) for ASD	9
Momen et al., 202244	Register-based	Denmark	2,035	MI: 6,976	Maternal discontinuation of AP medication before pregnancy	Typical and atypical APs	All indications included	Prescription dispensed 1 mo before conception until delivery	5–19 y postpartum	Registered psychiatric diagnosis	No significant difference 9 in risk of psychiatric diagnosis between exposed vs maternal discontinuation group	9
Petersen et al., 201645	Register-based	United Kingdom	290	H: 210,966 MI:492	Unexposed and those who had discontinued APs	All typical and atypical APs	Psychosis (bipolar disorder, schizophrenia, overlap symptoms)	Prescriptions issued between 31 and 105 days after start of pregnancy	Up to 5 y	Neurodevelopmental and behavioral disorder coded in chart	After adjusting for confounders, no association between prenatal AP exposure and neurodevelopmental and behavioral disorder codes (RRR 1.22 [95% CI 0.80–1.84])	7
Schrijver et al., 202337	Cohort	Netherlands	17	MI: 74	Women with mental illness treated at perinatal psychiatry centers, but unexposed to APs	Quetiapine, haloperidol, pipamperone, flupentixol, olanzapine, penfluridol, bromperidol	Among exposed: bipolar I disorder (n[7), schizoaffective disorder (n[1), borderline personality disorder (n = 4), schizophrenia (n = 4), OCD (n = 1)	Anytime during pregnancy: first + second trimester (n = 1), second + third trimester (n = 2), second trimester (n = 1), whole pregnancy (n = 13)	6–14y	Categories and Mosaics subtests from SON-R 6–40 Nine subtests from NEPSY-II-NL Learning problems and psychiatric diagnosis based on parental report (CBCL)	No association between exposure and parent-reported psychiatric diagnosis	6
Straub et al., 202246	Register-based	USA	10,765	H: 3,341,291	Public and private insurance data of unexposed pregnancies	NR	All indications included, largest proportion bipolar disorder (n = 5,752) and depression (n = 5,477)	Prescription dispensed at least once during second half of pregnancy	8–14 y	Registered diagnosis of any neurodevelopmental disorder	After adjusting for confounders, no significant association between AP exposure and neurodevelopmental disorder risk	7
Wang et al., 202147	Register-based	Hong Kong	706 in ASD analysis 547 in ADHD analysis	410,545 not exposed included in ASD analysis 333,202 not exposed included in ADHD analysis	No AP exposure, discontinuation, discordant siblings	Any APs in British National Formulary Chapter 4.2.1, excluding lithium and antidepressants	NR	For ADHD analysis: first trimester only (94/547), both first and second trimesters (61/547), all trimesters (285/547) For ASD analysis: first trimester only (115/706), both first and second semesters (73/706), all trimesters (385/706)	Included children with at least 3-y follow-up for ASD Included children with at least 6-y follow-up for ADHD	ADHD ICD code or prescription of ADHD medication ASD ICD diagnosis code in chart	No associations between prenatal AP exposure and ADHD (wHR 0.99 [95% CI 0.60–1.61]) or ASD (wHR 1.10 [95% CI 0.58–2.08]) when comparing prenatally exposed vs past maternal exposure	8

Note: ADHD = attention-deficit/hyperactivity disorder; aHR = adjusted hazard ratio; AP = antipsychotic; ASD = autism spectrum disorder; ASQ = Ages and Stages Questionnaire; CBCL = Child Behavior Checklist; H = healthy; LMP = last menstrual period; MI = mentally ill; NOS = Newcastle-Ottawa Scale; NR = not reported; OCD = obsessive-compulsive disorder; RRR = relative risk reduction; SON-R6–40 = Snijders-Oomen Nonverbal Intelligence Test Revision; wHR = weighted hazard ratio.

Crude analyses generally showed increased risk of neurodevelopmental and psychiatric conditions with AP exposure.^42,43,46,47 However, when including confounders such as indication for use, maternal socioeconomic factors, and maternal smoking, studies did not find a statistically significant difference in likelihood of offspring neurodevelopmental diagnoses in those exposed to AP. As such, discordant sibling analyses, controlling for all genetic and environmental factors shared between siblings, in a population-based cohort spanning 5 Nordic countries found no significant associations between prenatal AP exposure and a range of neurodevelopmental outcomes, including intellectual disorders (ICD-10 codes F70.0-F73.0), developmental speech or language disorders (ICD-10 codes F80.0-F80.2, in the absence of a hearing loss diagnosis), and developmental learning disorders (ICD-10 codes F81.0-F81.9).⁴² In contrast to other studies reporting neurodevelopmental disorders, Schrijver et al.,³⁷ in a smaller cohort study including 17 exposed children and 74 children exposed to severe maternal mental illness but not exposed to APs, assessed psychiatric disorders and learning problems in children ages 6 to 14 years old based on parental report rather than medical record. Similar to most of the large register-based studies, no significant associations were found between AP exposure and psychiatric diagnosis or learning problems. Halfdanarson et al.⁴³ even found decreased risk of ADHD in adjusted analysis when comparing AP continuation during pregnancy vs AP use before pregnancy only (hazard ratio [HR] 0.74 [95% CI 0.62–0.87]), suggesting that treatment for mothers may be beneficial for child neurodevelopmental outcomes.

In sensitivity and subgroup analyses, a number of studies found possible increased risk following exposure to certain types of AP. For instance, Straub et al.⁴⁶ reported increased risk of neurodevelopmental diagnosis with aripiprazole exposure (HR 1.36 [95% CI 1.14–1.63]). Furthermore, Momen et al.⁴⁴ found an increased risk of offspring psychiatric diagnosis with thioxanthene derivatives exposure (HR 1.28 [95% CI 1.01–1.63]) as well as in a moderation analysis of child sex, with an increased risk among boys in the continuation relative to the discontinuation group (HR 1.33 [95% CI 1.08–1.62]).

School Performance

Three studies explored school performance as an outcome (Table 3).^38,42,48 Stika et al.³⁸ used a blinded teacher behavioral evaluation of children ages 9 to 10 years exposed in late pregnancy to chlorpromazine or chlorprothixene compared with unexposed control students with healthy control mothers. They found no difference in teachers’ behavior evaluations between exposed and unexposed students. Liu et al.⁴⁸ examined standardized national Danish language and math tests in grades 2 to 8 (ages 8–15 years) and found that in crude analysis, children exposed to APs had significantly lower mean language and math scores compared with unexposed children with healthy mothers. However, mean score differences were no longer significant in the fully adjusted model, which included maternal factors including age, parity, smoking, marital status, year of delivery, education, psychiatric morbidities such as metrics for disease severity and other prescriptions filled during pregnancy, and paternal factors (ie, country of origin, psychiatric diagnosis, self-harm history, and income). Similarly, in a discordant sibling analysis, performance did not differ between exposed and unexposed siblings. Finally, Bruno et al.⁴² investigated poor academic performance, defined as scoring in the lowest 25th percentile in the first national standardized math and language arts (defined as native language skills, including reading and/or writing and grammar) tests in 5 Nordic countries, administered between ages 8 and 10 years. Only children born to mothers with a psychiatric diagnosis were included in the analysis of AP-exposed and unexposed children. Exposure to APs was not associated with poor academic performance in mathematics (RR 1.04 [95% CI 0.91–1.18]) or language arts (RR 1.00 [95% CI 0.87–1.15]) in fully adjusted analyses. In sensitivity analysis of exposure timing and specific APs, results remained similar.

TABLE 3.

Characteristics and Results of Studies That Assessed School Behavior and Performance

Authors, year, reference	Study design	Country	Exposed (n)	Controls (n)	Control description	Medication	Maternal diagnosis	Exposed period	Follow-up time after birth	Measurements	Results	NOS
Bruno et al., 202442	Register-based	Denmark, Finland, Iceland, Norway, Sweden	11,626	MI: 197,296	Any psychiatric condition based on ICD codes	Any AP: most commonly quetiapine, olanzapine, prochlorperazine, perphenazine, aripiprazole, levomepromazine	Among exposed: psychotic or bipolar disorders (5,049; 43.4%), other psychiatric disorders (5,984; 51.5%), psychiatric disorders recorded by prescription reimbursement: (593; 5.2%)	At least 1 prescription filled for any AP anytime in pregnancy	3–19 y (median follow-up 6.7 y)	Neurodevelopmental disorder based on ICD codes and poor academic performance (bottom 25th percentile) on national standardized test	No class effect in most fully adjusted analyses for academic performance. Possible relation between mathematics performance (aRR 1.19 [95% CI 1.00–1.42]) and levomepromazine, but with low event numbers	9
Liu et al., 202248	Register-based	Denmark	1442	666,075 total controls MI: 33,874	Unexposed: sensitivity analysis with AP discontinuation Stratified by maternal mental illness, discordant siblings	Typical and atypical APs	Exposed group: schizophrenia and related disorders (n = 449), mood disorders (n = 242), neurotic, stress-related, and somatoform disorders (n = 210), other mental illnesses (n = 108), no psychiatric diagnosis (n = 433)	Between 30 days before pregnancy up to delivery: only first trimester (n = 739), only second or third trimester: (n = 170), multiple trimesters (n = 524)	8–15 y	Standardized language and mathematics tests in National School Test Program in grades 2–8 (mean ages 8.9–14.9 y)	After adjusting for confounders, no association between maternal AP prescription and performance in language (adjusted difference 0.5 [95% CI −0.8 to 1.7]) or in mathematics (adjusted difference 0.4 [95% CI −1.0 to 1.8]) tests	9
Stika et al., 199038	Cohort	Czech Republic	66	H: 66	Students from the same school	Chlorpromazine, chlorprothixene	NR	>20 wk	9–10 y	Teacher behavioral evaluation	No difference	6

Note: AP = antipsychotic; aRR = adjusted relative risk; H = healthy; MI = mentally ill; NOS = Newcastle-Ottawa Scale; NR = not reported.

Other Neurodevelopmental Measures

Four studies used standardized neurodevelopmental assessments to investigate neuropsychological domains, including communication, social, problem-solving, emotional, attention, memory, and emotional domains (Table 4).^34,36,38,40 As with the motor outcomes, studies of younger infants were more likely to detect differences in cognition and behavior between study groups.

On the Bayley Scales of Infant and Toddler Development, Third Edition (BSID-III), Peng et al.³⁵ reported lower mean scores on the cognitive, social-emotional, and adaptive behavior scales among 76 children exposed to APs at 2 months of age compared with unexposed controls. However, by 12 months of age, no significant differences were found on any Bayley III scale. Johnson et al.,³⁴ evaluating infants at 6 months of age in a habituation paradigm, found no significant effect of prenatal medication exposure on the number of trials to habituate or on average looking time during the habituation task.

Studies of older children generally reported nonsignificant results on other neurodevelopmental measures. Schrijver et al.³⁷ found no differences in IQ scores in school-age children exposed to APs in utero compared with children exposed to mothers with severe maternal mental illness. Likewise, they reported no differences between any NEPSY-II-NL assessment subtest, a test of neuropsychological development that included measures of attention and executive functioning, social perception, memory and learning, sensorimotor, visuospatial processing, and language.⁵³ Lastly, Swetlik et al.⁴¹ assessed 178 children exposed to second-generation APs compared with 174 children exposed prenatally to other psychiatric medications (mostly selective serotonin reuptake inhibitors) on parent-rated behavioral problems and developmental delay using the Child Behavior Checklist (CBCL)⁵⁴ and Ages and Stages Questionnaire (ASQ),⁵⁵ respectively. The authors found no significant differences between the exposed and unexposed groups with regard to the numbers of children who scored above the clinical cutoff for internalizing (odds ratio [OR] 2.27 [95% CI 0.70–7.36), externalizing (OR 1.78 [95% CI 0.77–4.14]), or total (OR 2.72 [95% CI 0.95–7.69]) behavior problems as measured by the CBCL. Additionally, no significant difference in risk of developmental delay was reported when combining all of the ASQ-3 subscales (OR 1.24 [95% CI 0.74–2.09]). However, when considering the subscales individually, exposed children were at higher risk of developmental delay on the communication subscale (OR 2.96 [95% CI 1.21–7.24]).

Risk of Bias and Quality of Included Studies

NOS scores varied between 5 (moderate quality) and 9 (high quality). The mean NOS score was 6.25 for studies that included motor outcomes (Table 1), 7.86 for studies that included neurodevelopmental disorders (Table 2), 8 for studies that included school behavior and performance (Table 3), and 5.75 for studies reporting other neurodevelopmental outcomes (Table 4). Studies most frequently lost points for selection of the nonexposed cohort due to use of healthy controls rather than mentally ill maternal controls. Of note, none of the studies included a power analysis to justify their sample size. See Table S3, available online, for detailed NOS scoring of each study.

DISCUSSION

Here, we systematically reviewed the literature on the associations between in utero exposure to AP medications and neurodevelopmental outcomes. We found that the majority of studies did not identify differences in neurodevelopmental outcomes between AP-exposed and unexposed infants and children, after appropriately accounting for important confounding factors. However, there was large heterogeneity in the populations studied and outcome measures assessed.

Specifically, large population-based register studies did not find associations between prenatal AP exposure and higher risk of neurodevelopmental disorders, including ADHD and autism spectrum disorder, in childhood. Other cohort studies, which used more detailed assessments of neuropsychological functioning and IQ in school-age children, equally reported no associations with AP exposure in pregnancy. Notably, the follow-up periods of these studies ranged significantly within and between studies from early childhood into late adolescence. However, the only cohort study with follow-up into adolescence included only 17 participants.³⁷ The length of the follow-up periods is important to consider because studies with a shorter follow-up period were more likely to report significant effects, particularly in motor development outcomes. Yet, 3 longer-term studies found that significant motor and behavioral effects in early infancy were no longer apparent by 1 to 7 years.^35,36,39

Whereas the longer-term results of these studies are reassuring, the frequent finding of transient early behavioral and motor deficits requires attention. One possible explanation for these findings is that AP use during pregnancy does influence neurodevelopment, with associated delays in motor development in infancy, but once the AP exposure is discontinued after birth, the infant’s brain is able to adapt and overcome these deficits within the first year of life, leaving no observed long-term motor deficits. Another possible explanation is that these findings represent an extended abstinence syndrome. Neonatal abstinence syndrome, described as clinical manifestations observed in neonates following withdrawal of an intrauterine drug exposure, has been documented with AP medications with symptoms of abnormal movements, agitation, change in muscle tone, tremor, or difficulty breathing or feeding.^49,50 However, abstinence syndrome has previously been assumed to be an acute withdrawal reaction that subsides hours to days after birth. Given the motor development results observed here, a more prolonged reaction to the withdrawal of in utero exposure to APs is conceivable. In interpreting these findings, one should also consider the possible influence of confounding by indication. It is possible that severe mental illness, the indication for which APs are prescribed, is in itself associated with delays in motor development.^56,57 Several studies^35,39,40 used a control group drawn from the general population, rather than a control group of children born to mothers with severe mental illness. Thus, these studies likely did not sufficiently account for the impact of confounding by indication.

Caveats and Confounding Factors

In evaluating the literature on this topic, it is important to keep in mind that children of women with severe mental illness are at elevated risk of numerous adverse cognitive, neuromotor, and behavioral outcomes related to the mother’s psychiatric diagnosis itself.^27,58–60 These outcomes may overlap to an unknown degree with neurodevelopmental effects of AP exposure. Therefore, to investigate any additional effect of antenatal AP use during pregnancy, it is essential to compare children of women with severe mental illness with and without AP exposure, rather than using children of women from the general population as a comparison group. Studies that included a control group of children of women with severe mental illness reported early motor deficits in AP-exposed children up to 6 months of age^33,34,36 that were not present at later ages.^36,37,41 Additionally, studies reported no association with school performance,^42,48 neurodevelopmental disorders,^37,42–44 or other neurodevelopmental outcomes.^36,37,41 Other potential confounding factors, such as smoking, licit and illicit drug use, psychotropic medication use, socioeconomic status, and maternal education, are also important to consider. In the studies systematically reviewed here, confounding was considered to varying degrees. Several studies that did adjust for potential confounding factors found significant associations in unadjusted analyses, which were no longer statistically significant in the fully adjusted model. Only approximately 44% of the included studies adjusted for maternal smoking and at least one of either socioeconomic status or maternal education. As above, these studies^{33,36,41,43,44,46} reported early motor deficits in AP-exposed children up to 6 months of age and no associations with school performance, neurodevelopmental disorders, or other neurodevelopmental outcomes.

Conversely, it is important to keep in mind that by modifying the mother’s psychiatric functioning, the child’s prenatal and postnatal environments are also modified and that these modifications may have positive effects as well as negative effects. There is at least a theoretical possibility that pharmacological treatment for more severe mental illness could have protective effects on the child. In line with this idea, Halfdanarson et al.⁴³ even found decreased risk of ADHD in adjusted analysis when comparing AP continuation during pregnancy vs AP use before pregnancy only (HR 0.74 [95% CI 0.62–0.87]). Although no study to date has examined potential protective effects of maternal treatment with APs on child developmental outcomes, several studies have examined this in women treated with antidepressants. For example, Hunter et al.⁶¹ demonstrated improved auditory sensory gating in neonates of women with anxiety who were treated pharmacologically compared with women with anxiety who did not receive medication. Uguz et al.⁶² demonstrated that antenatal antidepressant treatment could reduce the risk of preterm birth, low birth weight, and neonatal intensive care unit admission outcomes associated with maternal panic disorder. More recently, Liu et al.⁶³ demonstrated clinically significant improvements in externalizing behaviors, internalizing behaviors, and attention problems among children whose mothers had received medication for postpartum depression compared with children whose mothers had been diagnosed with postpartum depression but had not received medication. Thus, the potential for beneficial or protective effects of maternal AP treatment on child development is an equally important area to investigate to provide patients with fully informed risk/benefit discussions.

Clinical Recommendations

The current literature, despite its limitations, does not show a consistent, strong indication that AP use during pregnancy is harmful to child neurodevelopment in the long term. On the other hand, studies are heterogeneous, and we do not have enough data on specific medications and long-term outcomes to detail all possible risks and benefits of AP use in pregnancy. Importantly, there are other potential negative effects of AP use during pregnancy, including a higher risk of maternal metabolic syndrome and gestational diabetes, conditions separately known to affect birth outcomes.^64,65 Likewise, however, untreated maternal mental illness during and after pregnancy can have deleterious effects on child development. Therefore, all potential risks should be weighed against clear advantages of protecting the mother from relapse and the child from exposure to severe untreated maternal mental illness. We encourage clinicians to share the existing research with patients and to use shared clinical decision making regarding risks and benefits of AP medications to choose the best approach for the individual. This will ultimately maximize maternal and fetal health in the short and long term.

Future Directions

Several important questions remain unanswered after this systematic review of the literature.

1. Are there drug- or dose-specific neurodevelopmental effects? Most studies investigated the effects of combined prenatal AP exposure, not the individual drug effects, and dose-response analyses were lacking.

2. What are the extent and effect, if any, of continued exposure to APs through lactation? Although some studies used breastfeeding as a covariate,^34–36 none of the studies conducted subgroup analyses comparing outcomes of continued breast milk exposure vs bottle feeding among AP users.

3. Are there sex-specific differences in the neurodevelopmental effect of prenatal AP exposure? Several studies point to the possibility of sex-specific effects, but investigating this properly requires large sample sizes.

4. What are the effects of prenatal AP exposure on structural and functional brain development? Our search of the literature detected no brain imaging studies in humans.

5. Are there any protective effects for children of maternal AP use in pregnancy?

Future studies should focus on trying to answer these questions using adequate sample sizes for stratified analyses, collection of detailed information on AP medication used (including prescribed dose), longitudinal data collection with multiple time points and adequate follow-up periods, validated detailed measures of neurodevelopment, collection of structural and functional infant brain data, sex-specific analyses, and investigations of AP use during lactation. Importantly, control groups should always consist of children born to mothers with severe mental illness. Discordant sibling analyses are ideal when sample sizes permit. Finally, to be able to properly weigh the risks and benefits of AP use during pregnancy, the efficacy of AP use in preventing peripartum episodes should be another focus of future research projects in the field.

Data Sharing:

Data collected for the study will not be made available to others.