The literature search yielded 7,984 references. After the removal of 2,452 duplicate references, 5,532 were left for the title and abstract screening. Of these, 232 references underwent complete text evaluation, resulting in 68 eligible studies which were included in the review. Figure 1 shows the PRISMA flowchart.
Overview of the studies
An overview of the 68 eligible studies is presented in Table 1. The majority of perinatal pharmacoepidemiologic studies on cognition, language, and education outcomes were conducted in one or several of the Scandinavian countries (n = 24 studies), followed by the USA and the UK with 14 and 13 studies, respectively. The most common study design was a cohort design (n = 46 studies), followed by case-control (n = 7 studies) and cross-sectional study design (n = 3 studies). Eligible studies ranged considerably in study size (10 to 24,825 exposed children). Of the eligible studies, 59 focused on psychotropic medications. Among these, antiepileptics were by far the most studied psychotropic medications with 38 out of 59 studies examining this medication group; 22 studies focused on other psychotropic medications including antidepressants (n = 20 studies), anxiolytics (n = 3 studies), antipsychotics (n = 1 study), hypnotics (n = 3 studies), and sedatives (n = 1 study); and nine focused on analgesics. One study included both antiepileptics and other psychotropics. Due to the substantial number of studies focused on antiepileptics, this medication group is described separately from other psychotropics. In total, 50 studies assessed cognitive outcomes such as IQ and cognitive development, 17 evaluated language and language development, and 11 examined educational outcomes (Fig. 2). Most papers assessed several outcomes within the same study. In 39 studies, neurodevelopmental outcomes were assessed by healthcare professionals such as psychologists, and nurses or by trained researchers (Table 1; Supplemental Table 2, and Fig. 1). Psychologists were the most common administrators of the instruments (n = 23 studies). Confounders were accounted for in 61 studies, either by adjustment or by matching. Typical covariates were maternal age, ethnicity, education, socioeconomic status, maternal cognitive ability, maternal prenatal depressive symptoms, maternal smoking, and alcohol intake during pregnancy.
Table 1
study characteristics of the 68 studies included in the systematic literature review.
Study Characteristics | Number of studies (%) | Outcome measure |
| | Psychometric instruments / tests | Diagnostic codes |
Total | 68 (100%) | 62 (91.1%) | 6 (8.8%) |
Countrya Scandinavian USA UK Others | 24 (35.2%) 14 (20.5%) 13 (19.1%) 24 (35.2%) | 19 (30.6%) 14 (22.5%) 13 (20.9%) 23 (37%) | 5 (83.3%) - - 1 (16.7%) (France) |
Study design Cohort Case-control Cross-sectional | 58 (85.2%) 7 (10.2%) 3 (4.4%) | 52 (83.8%) 7 (11.2%) 3 (4.8%) | 6 (100%) - - |
Type of data collection Primary Secondary | 53 (77.9%) 15 (22.0%) | 53 (85.8%) 9 (14.5%) | - 6 (100%) |
Exposureb Antiepileptics Other psychotropics Analgesics | 38 (55.8%) 22 (32.3%) 9 (13.2%) | 34 (54.8%) 20 (32.2%) 9 (14.5%) | 4 (66.6%) 2 (33.33%) - |
Outcome of interestc Cognition IQ Cognitive development Language Education | 32 (47.0%) 17 (25.0%) 18 (26.4%) 11 (16.1%) | 28 (45.0%) 16 (25.8%) 17 (27.0%) 9 (15.0%) | 4 (66.6%) 1 (16.6%) 1 (16.6%) 2 (33.3%) |
Source of outcome measure Psychologist Researchers Diagnostic codes Teachers Parents Computerized test Nurse Not specified | 23 (34.0%) 18 (26.0%) 6 (9.0%) 6 (9.0%) 6 (9.0%) 2 (3.0%) 1 (1.0%) 6 (9.0%) | 23 (34.0%) 18 (26.0%) - 6 (9.0%) 6 (9.0%) 2 (3.0%) 1 (1.0%) 6 (9.0%) | - - 6 (100%) - - - - - |
Psychometric propertiesd Validity Reliability Not mentioned | 25 (33.8%) 15 (22%) 38 (55.8%) | 23 (37.0%) 15 (24.1%) | 2 (33.3%) - 4 (66.6%) |
Confounder control Adjustment Matching Not specified | 50 (73.5%) 11 (16.1%) 7 (10.2%) | 44 (71.0%) 11 (17.7%) 7 (11.2%) | 6 (100%) |
a Some studies were multinational; thus, the numbers add up to more than 100%. 6 studies in the UK and the USA and 1 study in the USA and Brazil. b one study includes antiepileptics and psychotropics. c Outcomes are not mutually exclusive. d Nine studies reported both the validity and reliability of psychometric instruments. |
Detailed characteristics and results for the included studies are presented in the Supplemental Table 1. An overview of the confounders assessed can be found in the supplemental text. Below, the main results are presented.
Antiepileptics
Study characteristics
Antiepileptics were the most studied medication group (n = 38 studies, median sample size of 176 (interquartile range (IQR): 96–538)) exposed children (7, 11, 18–53). Of the studies on antiepileptics, 24 papers assessed IQ as a measure of child cognition, and seven studies assessed general cognitive development. Language development and educational attainment were assessed in six papers each. In these studies, developmental outcomes were assessed from the age of one to 19 years. Psychometric instruments were administered in 33 studies, whereas the remaining five studies employed diagnostic outcome measures. In 22 studies, the assessments were performed by psychologists or researchers using psychometric instruments.
Study findings
IQ and cognitive development
Studies assessing exposure to antiepileptics and IQ (n = 24 studies) as a measure for cognition were unanimous in their conclusion on negative effects after valproate and antiepileptic polytherapy exposure on IQ (n = 17 studies) (7, 18, 23–25, 27, 29, 31, 32, 34, 35, 37–39, 44, 48, 50). Findings regarding effects on IQ following exposure to other antiepileptics such as carbamazepine, lamotrigine, and topiramate were inconsistent (7, 18, 23–25, 27, 29, 31, 33–35, 37–39, 44, 48–53). Seven studies assessed antiepileptic exposure and general cognitive development outcomes. Five studies reported that valproate was associated with negative developmental outcomes when compared to other antiepileptics or the unexposed population (36, 41, 42, 45, 46). Two studies reported no association between lamotrigine and levetiracetam exposure and general cognitive development (19, 26). Please refer to Supplemental Table 1 for effect estimates.
Language
Six studies investigated exposure to antiepileptics and language outcomes. Four studies found valproate to be associated with a greater risk of language delays when compared to other antiepileptic drugs (AED) and to the unexposed population (21, 22, 36, 40, 42). One study showed an association between exposure to lamotrigine and speech delay (26).
Education
Educational outcomes were assessed on different parameters such as learning disabilities (LD), school performance, and special educational support. Prenatal exposure to valproate was associated with the risk of LD, poorer school performance, and provision of special educational support (n = 4 studies) (11, 27, 28, 47). One study found that children prenatally exposed to antiepileptic polytherapy had an increased risk of not receiving a final grade in the last year of compulsory school (43). Another study found no association between prenatal antiepileptic exposure and receiving a grade point average less than 2 above 9th grade (30).
Other psychotropics
Study characteristics
There were 22 studies that reported neurodevelopmental outcomes after exposure to psychotropics other than antiepileptics (median sample size 130 (IQR: 35-3982) exposed children) (30, 54–74). Of these, antidepressants were the most frequently investigated psychotropic medication group (n = 20 studies). The most studied developmental outcome was cognition including IQ and general cognitive development (n = 13 studies). Language outcomes were assessed in nine studies, and educational attainment in four studies. Neurodevelopmental outcomes after other psychotropic exposure were assessed in children from the age of 6 months to 19 years.
Study findings
IQ and cognitive development
13 studies assessed the effects of exposure to other psychotropics on cognition. Eight studies found no association between antidepressant exposure and cognitive abilities in children (59, 61, 64, 65, 68, 71–74). However, three studies on antidepressant exposure (55, 60, 69) and one on benzodiazepines and z-hypnotics (56) found the control group to have higher cognitive abilities than the exposed group.
Language
Among the nine studies that assessed the association between other psychotropics exposure and language outcomes, six studies reported an association between antidepressant and anxiolytic exposure and lower language skills (57, 61–63, 66, 67, 70). Two studies did not find any association (73, 74).
Education
Two studies found an association between antidepressant exposure and special education support (54, 58), while the other two studies found no association (30, 63).
Analgesics
Study characteristics
Neurodevelopmental outcomes following analgesic exposure were assessed in nine studies, with a median sample size of 446 (IQR: 1034–3727) exposed children (75–83). Six studies assessed prenatal exposure to acetaminophen/paracetamol, while the remaining studies investigated exposure to opioid analgesics (n = 2 studies) and aspirin (n = 1 study). Cognition (n = 6 studies) was the predominant outcome in this category, with four studies on IQ and two on general cognitive development. Outcomes were assessed in children from the age of two to 11 years. Only psychometric instruments were employed. Five studies used psychometric instruments administered by psychologists and researchers, three studies used parental, nurse, and teacher questionnaires, and one study did not specify the test administrator.
Study findings
IQ and cognitive development
All six studies found no association between analgesic exposure and cognitive outcomes (77–79, 81–83).
Language
One study found an association between exposure to acetaminophen and language delay in girls but not in boys (80). Another study found no differences between opioid analgesics exposed and unexposed children (76).
Education
One study on analgesic exposure and educational outcomes found the exposed children to score lower on literacy and numeracy tests (75).
Psychometric properties of the developmental outcome measures
Only 29 of the 62 eligible studies (46.7%) commented on the validity and/or reliability of the instruments. Of these, validity was mentioned in 23 studies, reliability was mentioned in 15 studies, and nine studies reported both validity and reliability (Fig. 3). Further, out of the 29 papers that discussed psychometric properties, only two papers explicitly mentioned all the subclassifications of validity and reliability (25, 45). Studies generally reported psychometric properties from normative samples originally used to validate the instrument (19 out of 29 studies) rather than the actual study sample. Only 11 studies out of 29 reported on validity (n = 6 studies) and reliability (n = 9 studies) for the study sample (Supplemental Fig. 2). An overview of reporting of the psychometric properties in the eligible studies is provided in Supplemental Table 2.
Of the 23 studies that provided information on validity, 14 studies did not specify the type of validity but instead mentioned that the instrument used was considered a gold standard, valid, or had undergone external validation. Regarding reliability, seven studies commented on external reliability, six studies discussed internal reliability, and two studies did not specify the type of reported reliability. An overview of these psychometric instruments, classified by medication group, is given below.
Antiepileptics
Of the 38 papers that focused on antiepileptics, 14 commented on the validity and/or reliability of the psychometric outcome measures employed. Validity and reliability of the normative sample for the assessment used was reported by four (25, 38, 40, 53) and eight studies (25, 29, 33, 36, 38, 40, 51, 52) respectively, while validity and reliability for the study sample was reported by four (29, 32, 34, 45) and three studies (23, 28, 45) respectively. It was most common to report validity among the studies on antiepileptics (n = 10 studies), but only three (25, 29, 45) specified the type of validity that was reported. None of the 10 papers commented on content validity.
Other psychotropics
Of the 22 papers that used psychometric outcome measures, only 11 mentioned either validity and/or reliability of the psychometric instruments. Psychometric properties were reported based on normative samples for eight studies (55, 57, 61, 62, 64, 66–68), and for the study sample by only three studies (54, 56, 59). Although validity (n = 9 studies) was the most common psychometric property to be reported, most studies did not mention the type of validity, and none had comments on content validity. Of the five papers that mentioned reliability, three commented on internal reliability without mentioning the subclassifications (56, 59, 68), while two reported on external reliability (54, 57).
Analgesics
Five of the nine papers that investigated analgesic medications commented on the validity and/or reliability of at least one instrument utilized to measure developmental outcomes. It was slightly more common to report psychometric properties for the study sample (three studies) (80–82) than for the normative sample (n = 2 studies) (76, 78). Validity was reported in four studies and reliability in one study. Three studies specified the type of validity, and none commented on content validity.
Assessment of outcomes in studies using diagnostic codes
Six studies used diagnostic outcome measures to evaluate neurodevelopmental outcomes in children. Only two of these studies addressed to some extent the validity of the diagnostic outcome measure (Table 2). Bjørk et al. referred to positive predictive values of ASD diagnostic codes (18) and performed a sensitivity analysis that required two diagnostic codes to address outcome misclassification. Blotiere et al. acknowledged that the diagnostic codes used were not externally validated (23). The remaining four studies did not acknowledge or report on the validity of the diagnostic codes.
Table 2
Overview of methods used to address the validity of outcome measures in studies using diagnostic codes (n = 6 studies).
Reference | Exposure | Outcomesa | Data source | Method |
Validity of outcome discussed | Restriction to x2 Dx codes | Validated algorithm used | Reporting PPVs | Case validation | Quantitative bias analysis |
Bjørk et al. (2022) | AED | ASD; ID; NDD | Nordic registries | --- | Yes | --- | ASD: PPVs 86–90% | --- | --- |
Daugaard et al. (2020) | AED | ID; ID/DD | Danish registries | --- | --- | --- | --- | --- | --- |
Blotiere et al. (2020) | AED | NDD; MD; DD; CRD | French registries | Acknowledged no validation | --- | --- | --- | --- | --- |
Bech et al. (2018) | AED | MD; DD; ASD; BD | Danish registries | --- | --- | --- | --- | --- | --- |
Viktorin et al. (2017) | SSRI | ID | Swedish registries | --- | --- | --- | --- | --- | --- |
Brown et al. (2016) | SSRI | Language; education | Finnish registries | --- | --- | --- | --- | --- | --- |
a For the specific ICD-10 diagnostic codes, please refer to Supplemental Table 3. |
Abbreviations: AED: antiepileptic drugs, ASD: autism spectrum disorder, BD: behavioural disorder, CRD: communication-related disorder, DD: delayed childhood outcomes, Dx: diagnosis, ID: intellectual disability, MD: mental retardation, NDD: Neurodevelopmental disorders, PPV: positive predictive value, PV: predictive value, x2: restricted to two diagnostic codes (main or sensitivity analysis). |