World over, an estimated 3 to 7 per every 1000 pregnancies are affected by epilepsy [1] . Having uncontrolled epilepsy while pregnant increases both maternal and fetal risks of pregnancy-related complications such as decreased oxygen supply to the fetus, slowdown of fetal heart rate, and premature delivery [2] . Clinicians are thus encouraged to treat epilepsy during pregnancy, given the treatment benefits to both the mother and the fetus. The primary treatment for epilepsy is pharmacotherapy with antiseizure medications (ASM) [3] . Commonly prescribed antiseizure drugs include: valproic acid/sodium valproate, lamotrigine, carbamazepine, levetiracetam, oxcarbazepine, and phenytoin [4] . Concerns about the safety of ASM during pregnancy have been raised due to evidence indicating risks of unfavorable birth outcomes [5] , putting physicians in a position of having to weigh the benefits derived from the ASM against potentials for teratogenic effects, before deciding to prescribe or not.

Autism spectrum disorder (ASD), is a childhood neurological disorder characterized by repetitive behaviors and limited interest in the affected children, coupled with social and communication difficulties. Key diagnostic characteristics include: impaired social interactions, stereotyped patterns of behavior and impaired communication ability. In order for a diagnosis to be established, one of these key characteristics must occur before the age of 3 [6] . This disorder results from the interactions among genetic and environmental risk factors [7] . In this context, prenatal exposure to antiseizure medications is a modifiable environmental risk factor.

Studies, both in animal models and humans have demonstrated that exposure to ASM in utero carries an increased risk for the development of ASD [8] - [13] . However, critical analysis of the relative risk of ASD from exposure to the different ASM, which trimester of exposure or dosage presents the highest risk among women with epilepsy is still unclear. [10] [13] [14] [15] This systematic review and meta-analysis thus sought to examine and synthesize evidence from all relevant studies that have characterized and evaluated the association between fetal exposures to ASM in women with epilepsy, with particular attention to the relative risk of ASD among children exposed to the different ASM in utero, and which trimester of pregnancy and/or dosage of exposure carries the most increased risk.

Our meta-analysis found that prenatal exposure to antiseizure medications is associated with an increased risk of ASD in children, with valproic acid presenting the highest risk as a monotherapy. Carbamazepine and lamotrigine exposures showed an association with ASD, but not as strongly as valproic acid. This drug has also been associated with other neurodevelopmental disorders such as attention deficit hyperactivity syndrome (ADHD) [22] and major congenital

malformations [23] - [28] . This finding thus supports the hypothesis that not all antiseizure medications are associated with greater risk to ASD, and so women with epilepsy wishing to get pregnant should be counseled to make an informed decision on what antiseizure medications to use.

According to the dose of ASM and risk of ASD, the analysis was not statistically significant. There is, however, a positive association between the mean dose of exposure to valproic acid in pregnancy and neurodevelopmental disorders [29] . Other adverse outcomes associated with the use of valproic acid are birth defects and neurocognitive impairments [30] [31] [32] [33] [34] . The higher the dose, the worse the outcome. According to a prospective cohort study by Adab et al. [35] from the United Kingdom, the cognitive and verbal skills of children exposed to low-dose valproic acid during pregnancy seemed to be compromised, raising further queries about the safe use of valproic acid in pregnancy. It is thus imperative that women in whom valproic acid is the only effective seizure control medicine be assessed to see if a safe dose exists, that ensures seizures are controlled without causing harm to the baby.

When assessed by trimester of exposure, Bjørk et al. [18] , Wood et al. [13] , Stadelmaier et al. [11] , and Wiggs et al. [12] all reported that exposure to ASM (specially valproate) during first trimester increased the risk of ASD. However, they did not objectively compare the odds across all the three trimesters. On the other hand, Christensen et al. [10] , objectively compared the three trimesters and found a HR of 2.9 [95% CI: 1.6 - 5.1] in first trimester and 3.1 [95% CI: 0.8 - 12.2] in second and third trimesters, although the confidence intervals of the second and third trimester exposures vary widely, probably because the sample size used was small (67 children), and so did not have enough power to detect a meaningful association. The findings of Christensen et al. are supported by Coste et al., who examined a large cohort of French children exposed to ASM in utero, and discovered that exposures in the second and third trimesters were instead more associated with neurodevelopmental disorders [36] . Moreover, Bjørk et al. [15] further found that exposure to valproate in the second and third trimesters without first trimester exposure were still associated with an increased risk of autism in children with a HR 1.94 (1.04 - 3.60). Furthermore, Mezzacappa et al., reported that the risk of ASD in children exposed to antidepressant drugs (some of which are ASM), in utero has no statistical significance by trimester of exposure [37] . A child’s brain development occurs across all the three trimesters of pregnancy; neural tubes divide and form the basic region of the brain in the first semester [38] , more complex synaptic connection occurs in the second trimester [39] , while myelin—a sheath that ensures efficient cell-cell communication—is formed in the third trimester [40] . Exposure to teratogenic antiseizure medications at any of these stages of the brain could have an effect on brain development at that level, hence presenting a risk for ASD. It is therefore plausible to say that the trimester of exposure does not have a significant risk for developing ASD.

Our study had a few limitations that could affect the interpretation of the findings: 1) There could have been a possibility of publication bias, however we could not draw a funnel plot due to the small number of studies analyzed, because, in general, funnel plot asymmetry should only be tested in meta-analyses with at least 10 studies included [41] . 2) Heterogeneity among the combined studies was high (I² = 87%). This could also be seen by a simple eyeballing of the forest plot, Figure 3. The most probable causes of heterogeneity could be the fact that majority of the studies did not use classic cohort or case-control designs, but rather used population registries in a retrospective approach. Also, the different studies used different methods of ASD diagnosis. Despite the existence of relatively standard tools for ASD diagnosis, much of the assessment questions in these tools are answered based on the assessor’s clinical judgment and not a definite laboratory or radiological exam. This scenario creates assessor bias that results in misdiagnosis of cases, hence variability among studies. For instance, Stadelmaier et al. [11] discovered 3 false-positive ASD diagnoses by rescreening children using a different diagnostic tool, and wood et al. [13] reported the possibility of including children with an intellectual disability who are not necessarily autistic as ASD since their clinical presentations tend to mimic ASD. The use of electroencephalogram (EEG) has been evaluated and shown to be promising for ASD diagnosis [42] [43] , while expression of SHANK 3; a genetic biomarker has shown a very strong association with ASD [44] [45] [46] Furthermore, big data analysis of the gut microbiome has shown an association between the composition of the gut microbiome and ASD [47] [48] . Studies along these lines should be accelerated to ensure a more non-subjective diagnosistic tool for ASD.

In summary, this study found out that prenatal exposure to antiseizure medications significantly increased the risk of developing ASD in children. The relative risk (RR) of developing ASD in children born to women with epilepsy exposed to valproic acid monotherapy during pregnancy was twice higher than that of carbamazepine and lamotrigine. Though many studies reported an increased risk for ASD from ASM exposure in the first trimester of pregnancy, the quality of evidence was not robust enough to make an independent conclusion. Similarly, there was no statistically significant association between the dosage of ASM used and the risk of ASD. It is therefore important to prescribe antiseizure medications that have been proven to have a low risk of ASD, and consider counseling women with epilepsy intending to get pregnant in order to customize for them the safest regimen available.

The author’s Z.A.M and E.T conducted literature search, data extraction, data analysis and manuscript drafting. A.O.J and Z.L conducted data analysis and reviewed the manuscript. J.F conceived the study and supervised the entire process of the study. All authors have read and agreed to the published version of the manuscript.

The data used to support the findings of this study are included within the article.

The authors declare no conflicts of interest regarding the publication of this paper.

Mohamed, Z.A., Thokerunga, E., Jimale, A.O., Liu, Z.C. and Fan, J.Y. (2023) Risk of Autism Spectrum Disorder According to the Dose and Trimester of Exposure to Antiseizure Medications: A Systematic Review and Meta- Analysis. Open Journal of Psychiatry, 13, 106-121. https://doi.org/10.4236/ojpsych.2023.132011

Thresholds for converting the Newcastle-Ottawa scales to AHRQ standards (good, fair, and poor): Good quality: 3 or 4 stars in selection domain AND 1 or 2 stars in comparability domain AND 2 or 3 stars in outcome/exposure domain; Fair quality: 2 stars in selection domain AND 1 or 2 stars in comparability domain AND 2 or 3 stars in outcome/exposure domain; Poor quality: 0 or 1 star in selection domain OR 0 stars in comparability domain OR 0 or 1 stars in outcome/exposure domain IRR; Risk of Bias regarded as the following: High when agreement is below 50%; medium when agreement is between 51% - 80% and low when agreement is greater than 80%.