Long-term neurodevelopmental consequences of intrauterine exposure to lithium and antipsychotics: a systematic review and meta-analysis

Search strategy for systematic review

A systematic search was performed by a trained librarian in the following databases: (1) Embase, (2) MEDLINE, (3) Web of Science, (4) PsychINFO, (5) Cochrane, and (6) Google Scholar, from their respective inceptions through June 8, 2017 to identify studies that investigated the long-term neurodevelopmental consequences of intrauterine exposure to lithium or antipsychotics. The search included the following elements: lithium, antipsychotics, (neuro) development and intrauterine exposure. All elements were transformed into a thesaurus suitable for each specific database. The exact search terms per database are reported in the Supplementary material (Supplement 1).

Study selection

Studies were considered eligible for inclusion if they were written in the English language and investigated the long-term neurodevelopment, defined as neurodevelopment beyond the newborn period, of offspring exposed to lithium or antipsychotics during gestation. Experimental preclinical investigations and clinical investigations were considered eligible for inclusion. Case reports were also included in this review. Two reviewers (EP and LS) independently screened the title and abstract of all records identified by our database search. Full text articles were obtained from the studies selected during this first screening step. Both reviewers independently selected the full text articles that met the eligibility criteria. The inclusion of both reviewers was compared and consensus was made on the final inclusion. An additional search was performed on the reference section of relevant studies and review articles to screen for other eligible articles that were otherwise not identified by our structured search.

Data extraction

Two authors (EP and LS) independently extracted data on study design, sample size and characteristics, medication dosage and exposure period, follow-up time, and behavioural, cognitive and neurological outcome measures. The data were summarized in a data extraction form. Studies were categorized by medication type and study design, and results were reported descriptively in accordance with the PRISMA statement [24].

Assessment of the risk of bias and the quality of studies

Methodological quality and risk of bias was assessed independently for each study by two reviewers (EP and LS). Risk of bias in preclinical studies was assessed with the SYRCLE’s risk of Bias tool [25]. This tool was recently developed to assess risk of bias and has been adjusted for specific aspects of animal intervention studies. The Newcastle–Ottawa Scale (NOS) [26] was used for clinical studies. The NOS assesses the risk of bias of observational studies based on selection, comparability and outcome criteria. The NOS rating scale varies from zero to nine; with zero representing the highest risk of bias and nine the lowest risk.

Procedure for meta-analysis

For our meta-analysis, we used the same search strategy as mentioned before. Only clinical investigations were included in the meta-analysis with the goal to enhance further insight into the risks associated with the use of these medications during pregnancy in humans. Pooling was performed per type of neuropsychological outcome and per group of medication exposure (lithium or antipsychotics) over a minimum of two studies. Fixed and random-effect estimation was used. In case of substantial heterogeneity, a random-effect estimation provides more reliable pooled results. Results are plotted in a forest plot. Cochran’s Q test, and I² statistics were used to quantify heterogeneity across trials. I² >40% was considered as substantial heterogeneity. The influence of intrauterine exposure to lithium or antipsychotics on neuropsychological development over time was estimated using random effects meta-regression analysis. Statistical analyses were performed using the ‘Metan package’ in Stata 15 [27].

Study selection

The study selection process is presented in Fig. 1. Our initial search produced a total of 1985 articles. After duplicates were removed, 1427 articles remained. Based on the screening of title and abstract, 182 full text articles were examined for eligibility, of which 118 were excluded (Fig. 1). In total, 73 studies were included in the qualitative synthesis, of which nine studies were included through manual (non-structured) identification. Additionally, three studies were included in the quantitative synthesis.

Open image in new window

Study characteristics

Of the 73 studies included in the qualitative analysis, 29 were preclinical investigations of which seven examined lithium exposure and 22 examined antipsychotics exposure. Most preclinical studies were performed in rats, some in mice and one study on lithium exposure used zebrafish. There is a large variety of the measurements used to assess neurodevelopment in animal models (Table 1, 2). The exposed period was generally during gestation, although several studies also investigated the effect of exposure during lactation. Postnatal brain development in rodents up to postnatal day 10 is considered analogous to prenatal brain development in humans [28].

In total, we found 13 clinical cohort studies of which three involved lithium exposure and ten involved antipsychotics exposure (Table 3, 4). Study samples varied from 14 to 2141 exposed subjects. Mean follow-up duration ranged from 1 to 15 years in studies involving lithium exposure and from 14 days to 5 years in studies involving antipsychotic exposure. Assessment of neurodevelopment varied between cohort studies. Out of the three clinical studies involving lithium exposure, one used standardized assessments, while the other two relied on an invalidated questionnaire or telephone interview. Most studies involving antipsychotic exposure used standardized objective assessments, but some studies relied solely on invalidated questionnaires or interview. Additionally, 31 case studies were included, of which 5 involved intrauterine lithium exposure and 26 involved intrauterine antipsychotics exposure (Supplement 2).

Preclinical investigations

Sechzer et al. [29] investigated the long-term developmental consequences of prenatal and early postnatal lithium exposure in rats. Female rats were treated with lithium during pregnancy and lactation. Development of the startle response and depth perception in the offspring were delayed. At the age of 4 months, pups exhibited lower spontaneous activity during open field activity testing. A similar study investigated the neurodevelopmental effect of lithium exposure from day 1 of pregnancy until postnatal day 15 [30]. Decreased locomotor activity and delayed development of sensory motor reflexes were observed in lithium-exposed mice. Whether these developmental delays were caused by prenatal or early postnatal exposure to lithium could not be determined. Nery et al. [31] studied the behavioural effects of lithium exposure on the development of zebrafish embryos and reported decreased locomotion compared to non-exposed embryos. Additionally, several studies have replicated a delay in eye opening [29, 30, 32] and decreased avoidance behaviour [32, 33] in mice and rats exposed to lithium during gestation and/or lactation. One study found impaired performance on the T-maze test [33]. Messiha et al. [34] found lower brain weights in lithium exposed offspring at the age of 37 days. No changes in social, defensive, threatening or aggressive behaviour was observed in lithium-exposed mice [35].

In summary, preclinical studies suggest a deleterious effect of lithium on motor activity, developmental milestones and reflexes, spatial memory and brain weight.

Clinical investigations

Neurodevelopment of 97 children with in utero exposure to lithium has been investigated in clinical cohort studies. Overall, most children were reported to have typical neurodevelopmental trajectories. Schou analysed data from the Scandinavian Register of Lithium Babies to compare neurodevelopmental outcomes in lithium-exposed children (n = 60) with their non-exposed siblings (n = 57) (average age, 7 years) [36]. Outcomes were assessed by questionnaire and based solely on mothers’ subjective retrospective assessment of their children’s developmental milestones. No significant differences were observed between the lithium-exposed children and their siblings.

In a prospective multicenter study, major developmental milestones were examined between a sample of 22 lithium-exposed children with non-exposed children [37]. Subjects were screened for study inclusion from among mothers who contacted the public teratogen information services to discuss the potential risks of prescription medication use during pregnancy. Data were collected by telephone interview. No differences were observed between lithium-exposed versus non-exposed children in the age at which they achieved major developmental milestones.

In an observational cohort study, 15 lithium-exposed children between 3 and 15 years old were investigated [38]. Standardized validated tests were used to assess growth, neurological, cognitive and behavioural outcomes. When compared to norms from the general population, most lithium-exposed children scored lower on the Block patterns subtest of the Wechsler Intelligence Scale for Children (WISC-III-NL). In contrast, no differences in growth or behavioural outcomes were observed. One child in this study exhibited subclinical neurological findings. Importantly, however, the conclusiveness of this study was hampered by the lack of a matched non-exposed control group ascertained in parallel with the lithium-exposed group, but rather relied upon an independently collected general population cohort dataset.

In summary, there is a paucity of clinical data on the neurodevelopment of children with in utero exposure to lithium. The three clinical studies published in the literature report normal neurodevelopment.

Case reports

Neurodevelopmental delay after intrauterine exposure to lithium was reported in four case studies, encompassing a total of eight children [39, 40, 41, 42]. Kozma et al. [39] reported on a neonate with neurodevelopmental deficits, including decreased muscle tone, depressed reflexes and diminished social response, during the 2.5 months after birth. However, by 13 months of age, no deficits were observed using the Bayley Scale of Infant Development. Morrel et al. [43] described a case of lithium toxicity at 35 weeks of gestation (lithium blood level: 2.6 mmol/L). The baby was born with primary cardiac muscle dysfunction and treated with isoprenaline at birth. At 12 months of age, cardiac function had normalized but there was evidence of delayed motor development and a concomitant strabismus. Delayed gross motor function was also reported in two cases with prenatal lithium exposure that did not involve lithium intoxication [41, 42]. One case report reported normal psychomotor development [44].

Preclinical investigations

Eleven studies have investigated the long-term neurodevelopmental consequences of prenatal haloperidol exposure in rats. Emergence of the surface righting reflex was found to be delayed [45]. Two studies found deficits in a circling training test [45, 46], a measure of motor performance and associative learning. Impairments in spatial learning were also found [45, 47, 48] using the Morris water maze, T-maze and radial arm maze. Moreover, the open field test revealed increased rearing, ambulation and general activity [47, 49, 50, 51]. One study reported finding no difference in ambulation on the open field test [52]. Notably, since behaviour in the open field test is influenced not only by locomotor activity, but also by anxiety and exploratory behaviour [53], additional studies have been performed to further differentiate these phenotypes. Indeed, consistent with an increase in anxiety, rats made fewer entries and spent less time in open arms during elevated zero and plus-maze tests [49, 51]. Moreover, aggressive behaviour was also increased [54]. Duration of shock-precipitated wall climbing was reduced on postnatal days 9 and 11, and there were no differences in stimulant induced behavioural stereotypes [55]. Lastly, from a neuroanatomical perspective, rats with intrauterine exposure to haloperidol exhibited significantly lower brain weight in adulthood [56].

Eight studies have investigated the long-term neurodevelopmental effects of intrauterine exposure to chlorpromazine in rats. One study systematically investigated the onset of neurodevelopmental milestones [57]. They found no difference in onset of eye opening, but emergence of the righting reflex was delayed. Studies investigating motor development found both increases [58] and decreases [59] in wheel running activity and impairments in a hanging task [57]. In the open field test, latency time was decreased [59, 60] and locomotor activity was increased [60, 61]. Similarly, spontaneous activity was normal in one study [62], but decreased in another [63]. Spatial memory was found to be impaired in a mother-goal maze task, whereas no differences were found in a T-maze task [61]. Studies focusing on other types of learning have reported impaired avoidance conditioning [57, 59], reversal learning [62] and operant conditioning [61]. Avoidance conditioning was observed to be normal [64]. A study investigating exploratory behaviour found that rats made fewer hole dippings [64]. Hoffeld et al. [58] did not find changes in emotionality testing. Susceptibility to audiogenic seizures was increased [63]. Brain weights did not differ between chlorpromazine and placebo exposed groups [60].

Several other antipsychotics were examined for neurodevelopmental effects. Rosengarten et al. [65] investigated the possible sequelae of intrauterine exposure to quetiapine, risperidone or olanzapine and found impaired spatial learning in a radial arm maze task for both risperidone and quetiapine, and disrupted short-term retention for quetiapine. Intrauterine exposure to olanzapine did not affect learning or retention. A recently published study also found impaired spatial learning and retention capability in rats with prenatal exposure to quetiapine [66]. Two studies investigated the effects of risperidone exposure during gestation. Singh et al. [67] reported increased ambulation and rearing in the open field test, and increased anxiety-like exploratory behavior in the elevated plus maze test. Intrauterine exposure to risperidone led to a dose-dependent reduction of adult brain weight. Zuo et al. [68] also found increased ambulation, while righting reflexes and spatial memory were normal. In the same study, rats with prenatal exposure to sulpiride exhibited an impaired cue response in a visual task performance and reduced spontaneous activity, while righting reflexes and spatial memory were normal.

In summary, most preclinical studies found a deleterious effect of antipsychotics on motor activity, developmental milestones and reflexes and spatial memory. Additionally, exposure to either haloperidol or risperidone led to decreased brain weight.

Clinical investigations

In total, neurodevelopmental data of 2934 children with in utero exposure to antipsychotics have been published involving nine clinical cohort studies. Six studies reported neurodevelopmental delays or deficits after prenatal exposure to antipsychotics [69, 70, 71, 72, 73, 74], while three studies reported normal developmental outcomes [75, 76, 77]. Most studies reported antipsychotic exposure on the basis of a single broad category, which combined a wide variety of antipsychotics. The initial report of abnormalities of motor development in children with intrauterine exposure to antipsychotic was authored by Platt in a cohort of 192 children exposed to antipsychotic neuroleptics and 116 children of women with a history of psychiatric disorders described as psychotic/neurotic but without antipsychotic treatment. Notably, deficits at the neonatal assessment of motor activity were more severe than at 8 months of age, when there was a non-significant trend towards more failures based on the Bayley gross motor assessment [73]. Another study examined 21 children with prenatal antipsychotic exposure, 183 children with prenatal antidepressant exposure and 78 non-exposed children at 6 months of age [78]. Children with prenatal exposure to antipsychotics had lower scores on the infant neurological international battery (INFANIB) compared to children with prenatal antidepressant exposure or non-exposed children. Comparable results were found in a recent register-based study in France [74]. Psychomotor development, assessed by pediatric examination, was compared between 70 children with prenatal neuroleptic exposure and 32.303 non-exposed controls. A higher prevalence of motor deficits was reported in exposed children at 9 months of age, a difference that was no longer present at 24 months of age. No differences in cognitive development were observed. A Danish general population register-based study reported an association between drug prescriptions during pregnancy and results on the Boel test, a psychomotor development test assessed at 7–10 months of age [71]. Specifically, the odds ratio for an abnormal Boel test was 4.1 (95% CI 1.3–13.0) among children with intrauterine exposure to neuroleptic medication after adjustment for several confounders including gestational age, birth weight and breastfeeding. In contrast, Stika et al. [76] reported finding no discernible adverse behavioural outcomes based on evaluations made by their classroom teachers in 10-year-old children with prenatal neuroleptic exposure.

Using data from two large electronic primary care databases in the UK, Petersen et al. investigated the risks and benefits of psychotropic medication during pregnancy. They compared the prevalence of neurodevelopmental and behavioural disorders in children with prenatal exposure to antipsychotics, children with no antipsychotic exposure whose mothers discontinued antipsychotic treatment before pregnancy and non-exposed children whose mother was not prescribed antipsychotic treatment in the 24 months before pregnancy. They found an increased risk of neurodevelopmental and behavioural disorders in children exposed to antipsychotics with a relative risk ratio (RRR) of 1.58 (95% CI 1.04–2.40). However, after adjustment for possible confounders, these differences were no longer statistically significant (RRR 1.22, 95% CI 0.80–1.84) [75]. An earlier study, focused specifically on phenothiazine antipsychotics, similarly reported no difference in intelligence quotient scores among 4-year-old children, of which 2141 had prenatal exposure and 26,217 were non-exposed [77]. Notably, however, a higher burden of neonatal withdrawal symptoms and autonomic instability was reported 14 days after birth in neonates with intrauterine exposure to phenothiazine antipsychotics [69].

More recent studies have also focused on the long-term developmental consequences of intrauterine exposure to atypical antipsychotics. Peng et al. [79] prospectively investigated 76 children with intrauterine exposure to atypical antipsychotics and 76 non-exposed controls, from birth until 12 months of age. Neurobehavioural development was assessed by the Bayley Scale of Infant and Toddler Development (BSID) at 2, 6 and 12 months of age. At 2 months of age, antipsychotic-exposed children exhibited significantly lower scores regarding cognitive, motor, social-emotional, and adaptive behavioural functioning. At 6 months of age, scores regarding social-emotional and adaptive behavioural functioning were still lower, but not significantly different between groups in cognitive or motor scores. In contrast, by 12 months of age none of these effects persisted. A post hoc analysis revealed that children prenatally exposed to clozapine had lower scores on the BSID adaptive behavior scale at the ages of 2 and 6 months compared to children exposed to other atypical antipsychotics [80]. However, this difference was also no longer present at 12 months of age.

Although studies varied in measurements and follow-up time, five cohort studies [70, 71, 72, 73, 74, 80] investigated motor development of children with in utero exposure to antipsychotics. These studies consistently showed a deficit in motor functioning in the first 9 months of life, but which appeared to spontaneously resolve based on subsequent follow-up assessments.

Case reports

Overall, case studies on intrauterine exposure to first generation antipsychotics have largely reported normal neurodevelopment [81, 82, 83, 84, 85, 86, 87, 88]. Additionally, although most case studies involving prenatal exposure to second generation antipsychotics have also reported normal infant and child development [81, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101], several case reports have found neurodevelopmental delays or deficits. Two cases reported speech delay, one involving risperidone and the other clozapine [102, 103]. One case reported abnormal behavioral development following prenatal exposure to risperidone and ziprasidone [103]. Impaired motor development has been reported following exposure to olanzapine, clozapine or risperidone [104, 105, 106].

Risk of bias and quality of the included studies

The risk of bias for the included preclinical studies is presented in the Supplementary Material (Supplement 3). Notably, many lack descriptions of the assessed domains, thereby making the risk of bias unclear (e.g., selection bias, performance bias, detection bias or attrition bias). In 34% of the preclinical studies, cross-fostering after birth was applied in order to control for medication-induced changes in maternal care.

NOS scores of clinical cohort studies varied between three and eight points (Table 3 and 4). Only three cohort studies properly controlled for maternal mental illness, widely considered the most important confounder in studies of intrauterine exposure to prescription psychotropic medication. In the other studies, neurodevelopment was compared between children with prenatal exposure to lithium or antipsychotics versus unaffected children, thereby leaving unaddressed the risk of confounding by indication. Moreover, few clinical studies controlled for additional confounders such as maternal age, congenital malformations, preterm birth, or smoking and alcohol use during pregnancy, often because information on these factors was not available. In most studies of antipsychotic exposure, developmental assessments were standardized and validated, although some studies based their results on non-validated questionnaires or information obtained exclusively from medical records. The quality of the included studies on lithium exposure is poor, as only one cohort study used validated measurements of neurodevelopment. Unfortunately, this study did not compare their findings with a formal control group. Regarding case studies, their quality is generally considered low with a high risk of publication bias. Indeed, most case studies did not assess neurodevelopment using validated objective measures.

Meta-analysis

Three out of five studies that investigated neuromotor deficits in children with in utero exposure to antipsychotics provided sufficient data and were included in a meta-analysis [70, 72, 74]. Figure 2 shows the relative risk of neuromotor deficits for antipsychotic exposure for all reported follow-up assessments (six effect sizes). Pooled relative risk calculated using fixed effect estimation was 1.97 (95%CI 1.47–2.62; Z value: 4.59, p < 0.001) with absence of heterogeneity (I² 0%, p = 0.622). Since studies reported multiple follow-up outcomes, this pooled estimate should be interpreted with care.

Open image in new window

Two studies [70, 72] reported follow-up outcomes at 6 months. The pooled relative risk was 1.63 (95% CI 1.22–2.19; Z value = 3.29; p = 0.001, fixed effect) with absence of heterogeneity (I² = 0%, p = 0.849), indicating a 63% increased risk for neuromotor deficits at 6 months (Fig. 3).

Open image in new window

Next we performed a random effects meta-regression analysis to study the longitudinal influence of intrauterine exposure to antipsychotics on motor development. For each study, we included the initial follow-up assessment. The direction of the regression coefficient suggested a decrease of the impact of intrauterine exposure to antipsychotics on neuromotor deficits over time. However, there was no statistically significant effect (− 0.03; 95% CI − 1.26 to 1.20; p = 0.80). Residual heterogeneity was substantial (I² = 49%). In conclusion, in this meta-analysis we were able to partially confirm the negative effect of antipsychotic exposure on motor development. However, we were not able to confirm the transient nature of the neuromotor deficits.