The births in this study represent nearly one-third of all births in the United States over this 10-year period. The increased utilization of SBIMs over time in our cohort suggests increased provider comfort with prescribing these medications during pregnancy. If this comfort comes with higher doses and/or longer duration, this could result in disproportionate elevations of ASD diagnosis in the SBIM-treated group. Also, it is notable that the average age at which ASD is diagnosed in the USA is about 4 years of age [73], thus, a significant number of our 18–72-month-old offspring might not have an established diagnosis yet (being censored). Furthermore, ASD diagnosis is often documented in medical records outside the EPIC database. As a result, our data might underestimate the potential impact of maternal SBIM use on developing ASD in the offspring.
There have been previous concerns of long-term neurodevelopmental outcomes for children exposed in utero to antipsychotics and antidepressants [74]. A national birth cohort study assessed the association of antipsychotic drug exposure during the second half of pregnancy with risk of neurodevelopmental disorders [75]. The unadjusted results were consistent with an approximate 2-fold increased risk. After adjusting for over 30 covariates, the risks were no longer significant, with the potential exception of aripiprazole (which is a strong sterol biosynthesis inhibitor) [41, 50]. The authors concluded that a potential signal for aripiprazole required replication in other data before causality can be assumed. We believe that our study provides this replication specifically for ASDs.
Not all SBIMs reported the same hazard ratios for ASD in our study. We believe that this is the function of their mechanism of action, half-life, metabolism differences, drug interactions, blood-brain barrier penetration and magnitude of sterol inhibition. For example, while statins all target HMGCo-A reductase, they differ in their degradation pathways [58]. In addition, the post-lanosterol inhibitors also vary in their mechanisms: fluoxetine is an emopamil binding protein (EBP) and dehydrocholesterol reductase 24 (DHCR24) inhibitor [40], while cariprazine is a primary DHCR7 inhibitor [50]. These findings suggest that disruption at any point along the sterol biosynthesis pathway may pose developmental risks to the fetus.
Another concerning observation from our studies involves polypharmacy, which has become increasingly common [76, 77]. This is also true for pregnant women [78,79,80,81,82,83]. A recent review found that two or more medications were prescribed up to62.4% of pregnant women [84]. In addition, our recent analysis of 1312 de-identified serum samples in pregnant women revealed that 302 samples had elevated 7-DHC levels, and 43 of these samples contained measurable amounts of at least one SBIM [51]. Taking multiple medications further elevated 7-DHC levels, indicating a potential additive or synergistic effect. We believe that this is likely the biological underpinning of the increased hazard ratios for offspring ASD when pregnant women are prescribed two or more SBIMs concurrently.
The public health implications of our findings are substantial [85]. The medications we examined had approximately 400 million prescriptions in 2022, and 9 of the 15 medications we studied were in the top 25 most prescribed medications in the US [86]. Furthermore, our current study shows that SBIM utilization during pregnancy is sharply rising in the US. Statins had long been contraindicated for use during pregnancy, but this warning was removed by FDA in 2021 [87]. The percentage of pregnant mothers receiving at least one SBIM had already more than tripled between 2014 and the removal of the warning in 2021 (Supplemental Material 7), so the change in the FDA warning labeling may be considered more a symptom of increasing comfort with these medications rather than a policy response to updated safety information, which does not appear to be available. Most of the SBIM package labels we evaluated come with significant warnings regarding pregnancy [88], although these warnings are deeply buried within the documents, and the prescribing physicians and patients are rarely aware of them. Thus, it appears that several widely used medications could have unintentionally contributed to a rise of ASD, without prior awareness of this risk.
Both confounders and overadjustment were major concerns in our study, complicating the interpretation of our findings. Namely, statins are typically not prescribed to those who don’t have hyperlipidemia, and psychotropic medications are rarely utilized when a patient does not have a mental disorder. As a result, the genetic factors (e.g. familial predisposition to ASD and/or mental disorders) and maternal medication use are challenging to separate. However, increased risks for ASD were comparable across multiple classes of medications studied - including antipsychotics, antidepressants, anxiolytics, and medications for cardiac and metabolic conditions. Despite differences in clinical indication, chemical structure, metabolism and pharmacological class, however, they share a common effect: inhibition of the sterol biosynthesis pathway [89]. Together with the increased signal in polypharmacy, this is a strong argument that the medications with sterol inhibiting effects or side effects that we studied are significant contributors to the increased risk of ASD seen in the current study.
We acknowledge several important limitations to our studies. Not all pregnancies prescribed SBIMs result in ASD, and our research does not yet provide insight into all factors that may exacerbate the sterol-inhibiting side effects of these medications. Potential contributors include pharmacogenetic interactions between the medications and single allele pathogenic variants (of mother or child) in sterol biosynthesis genes, found in approximately 3% of the population [90]. Additionally, the role of all possible comorbidities could not be adequately addressed in this dataset. Environmental (e.g. insecticides, air pollution, heavy metals, pesticides) and socioeconomic factors might further interact with all the above-listed factors [91,92,93]. Nutritional factors such as vitamin deficiencies (especially low vitamin D levels) might also increase vulnerability [94]. Furthermore, while we did not have detailed information on the dosage or duration of medication use, previous studies suggest that vulnerability is highest during the first trimester [9]. We acknowledge that our study is not able to define what would be a “safe” dose of these medications for the developing fetal brain. We do not have reliable data on this in our current dataset, and we acknowledge that we have no independent confirmation of ASD diagnosis or potential misclassifications, either. Finally, future studies will have to determine the ASD risk of a child when the mother stopped SBIM use before pregnancy, as SBIMs might lead to epigenetic modifications that can exert an effect even after discontinuation of medication.
We strongly caution against overinterpreting our findings, especially by attempting to extrapolate the developmental findings to adult patients. The medications we evaluated are lifesaving and critically important treatments for adults. While developmental processes (including growth, cell division, and differentiation) have the highest demand for sterols and are therefore most likely to be impacted by sterol inhibition, cholesterol turnover in the adult brain occurs over a period of years [4, 5]. As a result, while sterol inhibition may pose significant risks to developing brains and bodies, these effects are likely negligible for already-formed structures in adults.
In summary, we propose that the use of SBIMs during pregnancy may increase the risk of ASD in offspring through a multistep biochemical cascade. Maternal exposure to SBIMs inhibits DHCR7 (or other sterol biosynthetic enzyme) activity, leading to accumulation of highly reactive oxysterols and reduced availability of cholesterol. These oxysterols can enter the fetal circulation and exert toxic effects on developing neurons and glial cells. As intrinsic sterol biosynthesis becomes established in the fetal brain, pharmacologic inhibition of biosynthetic enzymes similarly disrupts fetal sterol metabolism, further increasing local sterol precursor levels and oxysterol formation. The combined maternal and fetal burden of high sterol precursors and low cholesterol availability is expected to impair cellular homeostasis, perturbing progenitor proliferation, neurite outgrowth, and circuit formation, ultimately contributing to ASD-related (or other) phenotypes later in life. Clearly, the outcome of these processes will depend on the magnitude of sterol biosynthesis inhibition in the fetal brain, which will be dependent on SBIM type, dose, timing, duration of exposure, genetic makeup of the child and mother and many other factors. Notably, our data suggests that the stronger DHCR7 inhibitors are likely to be associated with greater predisposition to ASD (see Fig. 4, R2 = 0.78). Alternatively, or in addition, SBIMs might lead to epigenetic changes that could interfere with normal fetal development. Based on the overall data, we believe it is time to rethink our clinical practices and drug development strategies, including an evaluation of SBIM use to detect potential long-term effects. We recommend 1) creating a comprehensive catalog of all currently used medications known to have sterol-inhibiting side effects; 2) systematically testing all new drugs for their potential to unintentionally inhibit sterol biosynthesis; 3) educating providers and advising women about the effects of any SBIM shown to have potential effects on the pregnancy and/or resulting offspring, as well as examining the biological effects after medication discontinuation (e.g. long-lasting epigenetic effects); 4) discussing the use of safer alternatives whenever discontinuing treatment is not feasible; 5) further studies to determine the effect of these medications when the mother and/or child carries a single-allele pathogenic variant in sterol biosynthesis genes; 6) limiting SBIM polypharmacy during pregnancy; and 7) investing in research focused on this critically overlooked area of drug safety and development.