Impact of Preterm Birth Subtype on Risk of Diagnosis of Autism Spectrum Disorders in the Offspring

Autism spectrum disorders (ASD) are a leading cause of lifelong disability with a worldwide prevalence rate of about 100/10,000 children, but rates vary wildly between different regions of the world and are as low as 1.09/10000 in South Korea and as high as 436/10000 in Australia (Zeidan et al., 2022). Although both boys and girls are affected by the condition, boys are more commonly diagnosed than girls by a factor of 4 (Zeidan et al., 2022). Economic analysis suggests that the costs in the United States may exceed 1 trillion dollars by 2025 (Leigh & Du, 2015; Wolff & Piven, 2021). Although the severity of the condition and the extent of the disability varies widely among children, recent economic analyses suggest that the national economic impact of the condition is considerable on patients and societies. Parents of autistic children were far more likely to be unemployed when compared to parents of non-autistic children or those with asthma (Lynch et al., 2023), with mothers taking the brunt of the adverse employment impacts and financial burdens (Liao & Li, 2020). Intellectual disability, a common comorbidity of ASD, further enhanced the financial and emotional strain on the parents (Saunders et al., 2015).

Early studies suggested a strong genetic component to the condition because the twin concordance rate was much higher for monozygotic than dizygotic twins (Bailey et al., 1995; Ritvo et al., 1985). Recurence of the condition amongst siblings and disparities in the prevalence of ASD due to sex and race-ethnicity further suggested that there is a genetic component to the condition. More recent studies, however, have suggested that the heritability (proportion of the disease due to genetics) of the condition is 0.50 (Sandin et al., 2014). This suggests that environmental or non-genetic exposures play an equally important role in the etiology of the condition. Over the past two decades, there has been an increasing appreciation for the role of pregnancy complications and exposures to infectious, pharmacological or toxicological agents as potential risk factors for the condition. Women whose pregnancies were complicated by hyperemesis gravidarum (Fejzo et al., 2019; Getahun et al., 2021), hypothyroidism (Andersen et al., 2018; Getahun et al., 2018; Rotem et al., 2020), preeclampsia (Gardener et al., 2009; Getahun et al., 2017; Wallace et al., 2008), placental abruption (Villamor et al., 2022), infections (Getahun et al., 2023; Zerbo et al., 2015), and exposure to valproic acid (Hernandez-Diaz et al., 2024) were more likely to have their children diagnosed with ASD than women without these conditions. In many of these studies, the risk of ASD disagnosis from the exposures were found to vary due to race-ethnicity and/or sex of the child. This suggests that these factors could modify the risk of environmental exposures as would occur in “2-hit” epidemiological models.

Preterm birth (PTB), the foremost problem in modern obstetrics due to its high prevalence and limited methods of treatment, has also been associated with increased risk for ASD (Agrawal et al., 2018; Haglund & Kallen, 2011; Pritchard et al., 2016). PTB can occur after the spontaneous initiation of contractions that proceed to labor and delivery prior to 37 weeks of gestation (spontaneous PTB, SPTB), or it can happen as an obstetrical intervention where the baby is deliberately delivered prior to 37 completed weeks of gestation (indicated PTB, IPTB) by cesarean section or induction of labor for other pregnancy complications (preeclampsia, placental abruption, fetal distress, or intra-uterine growth restriction) with the aim of providing care to the newborn in the neonatal intensive care unit. Although there is considerable overlap, these different subtypes of PTB, reflect different etiological conditions. SPTB occurs most often due to ascending microbial infections from the lower genital tract through the cervical os. In these infections, microbial growth eventually triggers the production of proinflammatory cytokines, prostaglandins and matrix metalloproteinases that, in turn, lead to premature uterine contractions, cervical ripening and placental detachment. IPTB, however, is often associated with chronic conditions of inadequate placental development/perfusion such as preeclampsia and intrauterine growth restriction. These complications become severe enough that the baby needs to be delivered early in order to avoid worse outcomes should the pregnancy continue any longer.

Rates of IPTB have been increasing over the past 30 years due largely to improvements in fetal heart rate and ultrasound monitoring, while rates of SPTB have been declining due to improvements in prenatal care (Ananth & Vintzileos, 2008; Mensah et al., 2023). The net effect of these changes has been a marked reduction in perinatal mortality for infants (Ananth & Vintzileos, 2008). The extent to which the impact of PTB subtype has on the risk of ASD diagnosis in the offspring and how it could be modified by child sex or race-ethnicity is unclear. Therefore, to test this hypothesis, we performed a retrospective cohort study using electronic health records (EHR) from a large integrated healthcare system.

Characteristics of the mothers and their pregnancies who gave birth to a child that was later diagnosed with ASD and those whose child was not diagnosed are listed in Table 1. Women whose children were diagnosed with ASD were significantly more likely to: be older, have higher income, be nulliparous, have a PTB, have smoked during pregnancy, or come from non-White or Hispanic race-ethnicity. They were also more likely to have reported drug use during pregnancy, have had gestational or pre-gestational diabetes, and to have given birth to a male infant. Although there was a slight excess in the proportion of women with late or no prenatal care, the results did not reach statistical significance.

As expected, children from women whose pregnancies ended in PTB were at significantly greater risk of being diagnosed with ASD (adjusted HR: 1.96, 95% CI: 1.62, 2.36). Survival curves demonstrate a more rapid accumulation of ASD cases in children born preterm than at term (Fig. 2, Panel A). A series of sensitivity analyses demonstrated that this effect persisted after excluding women diagnosed with ASD, antepartum depression, psychosocial disorders in pregnancy, who were prescribed anti-depressants prior to or during pregnancy, or who had no or late (3rd trimester) initiation of prenatal care (Table S1). Additional adjustments for possible confounding by preeclampsia/eclampsia, neonatal sepsis, maternal comorbidities (chronic hypertension, autoimmune disease, and thyroid conditions) as well as year of delivery and maternal BMI did not affect the observed association (Table S1). It was estimated that an unknown, unmeasured confounding variable would have to be rather large (E-value = 3.33) in order to explain the observed associations between PTB and risk of ASD.

Afbeelding vergroten

Increased risk of ASD was present for children born of both spontaneous and indicated subtypes (Table 2). However, the association appeared to be much stronger for IPTB given the higher point estimates for both the hazards and incidence rates (Table 2) and a more rapid accumulation of ASD cases for pregnancies that ended in IPTB than SPTB, which was still much higher than term birth (Fig. 2, Panel B).

We found a getational age-dependent relationship between gestational age at delivery for all types of PTB and an increased risk of ASD (Table 3). The incidence rates, however, were higher at each gestational age of delivery for IPTBs than they were for SPTBs, suggesting a higher level of risk (Table 3).

Further analysis and stratification of the data by sex suggested statistically significant sex by PTB subtype interaction (P = 0.003). Both SPTB and IPTB increased the risk of ASD for both boys and girls. However, the association appeared to be stronger for girls than for boys, as suggested by the larger hazard ratios for each PTB subtype (Table 4). Examination of the incidence rate ratios for IPTB vs. SPTB suggested a similar trend. Girls had a 1.40-fold increased incidence rate, but for boys, it was only 1.33-fold higher for IPTB vs. SPTB. Boys had high incidence rates of ASD for both types of PTB, suggesting that they are at increased absolute risk for the diagnosis (Table 4).

Significant interactions between PTB and race-ethnicity were also found (P < 0.001), suggesting that the impact of PTB on ASD risk differs by child’s race-ethnicity. Both SPTB and IPTB subtypes were found to increase the risk of ASD diagnosis for children of non-Hispanic White, Hispanic, and Other/Multiple race-ethnicity (Table 5). No association was found for any PTB subtype for non-Hispanic Black and children of unknown race-ethnicity. Asians/Pacific Islanders were at increased risk of ASD diagnosis if they were born from an IPTB, but not an SPTB (Table 5). Comparisons of incidence rates suggest that the incidence for ASD was higher for any PTB, IPTB and SPTB than it was for no PTB, regardless of race-ethnicity. Similarly, the incidence of ASD was higher for pregnancies born from IPTBs than it was for SPTBs for all race-ethnicities (Table 5).

In this large retrospective cohort study, we found PTB is a strong risk factor for the diagnosis of ASD in the offspring that persists after adjustment for confounding variables and after exclusion of patients with mental health disorders. This sugges that PTB is not simply a biomarker for maternal mental illness that is passed onto her offspring in the form of ASD. Any unknown, unmeasured confounding variable(s) would have to be particularly large (E-value = 3.33) to account for the observed findings (Table S1).

Both SPTB and IPTBs are associated with increased risk of ASD in a dose-dependent manner; however, the relative and absolute risk may be higher for IPTBs than they are for SPTBs. These findings are consistent with previous studies where any type of PTB (Buchmayer et al., 2009; Curran et al., 2018; D'Onofrio et al., 2013) as well as the pregnancy complications that are often comorbid with PTB such as preeclampsia (Buchmayer et al., 2009; Wallace et al., 2008), hypertension in pregnancy (Curran et al., 2018; Wallace et al., 2008), small for gestational age (Lampi et al., 2012; Moore et al., 2012), maternal hypothyroidism (Andersen et al., 2018; Ge, 2020; Getahun et al., 2018) and placental abruption (Villamor et al., 2022) are associated with ASD diagnosis in the offspring. Our finding of a dose-dependent relationship between the degree of PTB and the risk of ASD diagnosis in the offspring is consistent with previous studies where very high rates of ASD were found in very early PTBs (< 32 weeks) (Chen et al., 2019). By comparing SPTB with IPTB, we found that IPTB may be a stronger predictor for ASD diagnosis in the offspring. This occurred at all degrees of PTB with the dose-dependent association between reduced gestational length and the incidence of ASD in the offspring, demonstrating parallel trends with higher rates for the IPTBs (Table 3).

The reason for a stronger association with IPTB than SPTB is that conditions often resulting in IPTB may have more opportunity to adversely affect neurodevelopment. Conditions such as preeclampsia, placental abruption, and inadequate fetal growth are likely to emerge during placentation and continue throughout the pregnancy and have recently been associated with an increased risk of ASD in the offspring (Villamor et al., 2022). Most SPTBs, by contrast, are associated with ascending infections from the lower genital tract, where the pathogens spread to the maternal–fetal interface, grow, and reach a threshold that activates the labor mechanism. This reflects a sudden revocation of the immunological privileges that the fetus has enjoyed up until that time, and the woman’s body pushes out the fetus before the infection can spread further (Peltier, 2003). Further research is needed to determine which biomarkers may be responsible for the observed differences between IPTB and SPTB on the risk of ASD.

One potential biomarker for pregnancies at higher risk for later child ASD may be interleukin (IL)−6, which is produced at higher levels in women with pregnancies complicated by preeclampsia (Kauma et al., 1995; Silver et al., 1993), placental abruption, intrauterine growth restriction (Silver et al., 1993), thyroid dysfunction (Oztas et al., 2015), and gestational diabetes (Jafarzade et al., 2023) -pregnancy complications that have all been previously associated with increased risk of ASD in the offspring (Andersen et al., 2018; Buchmayer et al., 2009; Ge et al. 2020; Getahun et al., 2018; Wallace et al., 2008; Xiang et al., 2015). IL-6 is also detected at higher levels in cord blood (Galazios et al., 2002) as well as the maternal (Oztas et al., 2015) circulation in cases complicated by PTB. Although IL-6 has been conclusively shown to not be in the causal pathway for infection-mediated PTB (Sadowsky et al., 2006; Yoshimura & Hirsch, 2003), prenatal exposure to this cytokine does lead to ASD-like conditions in the offspring (Smith et al., 2007). Additional studies have suggested that the placenta is the source of IL-6 in animal models of ASD (Hsiao & Patterson, 2011), where it may function by altering myelination (Han et al., 2021), synaptogenesis (Ohki et al., 2024; Russo et al., 2018) and differentiation of neurons (Wu et al., 2017) that are disrupted in ASD. Given that IL-6 is associated with pregnancy complications that can lead to both SPTB and IPTB, higher levels of production of this cytokine could be a point of convergence between these different etiologies. Further studies are needed to determine if the observed differences between the risk of ASD from IPTB and SPTB are mediated through relative differences in IL-6 production.

Our finding that the risk of ASD was greater for females delivering preterm than males delivering preterm is unlikely to reflect differences in neuronal fragility. Estradiol is thought to have neuroprotective effects on the developing brain (Pansiot et al., 2017), and female fetuses have higher amniotic fluid concentrations of estradiol than male fetuses do (Robinson et al., 1977). Furthermore, amniotic fluid estradiol concentrations are increased in pregnancies that are complicated by PTB (Mazor et al., 1994) that may be part of a compensatory mechanism. These higher levels of estradiol would be expected to protect the developing brain from the neurological injury that results from conditions that result in PTB or the PTB process itself. Although there is strong evidence of IL-6 as a biomarker of PTB, most studies have suggested that male placentas produce more IL-6 than female ones do in a setting of infection (Osman et al., 2024), preeclampsia (Muralimanoharan et al., 2013), or environmental toxins (Miller et al., 2010). Therefore, additional research is needed to determine what biological factors could be responsible for sex differences in ASD risk in the setting of PTB.

Our finding of a higher incidence of ASD in non-White children is consistent with recent studies that have documented large increases for non-White children since 2010, that may be due to improvements in screening, diagnosis, and insurance coverage for minority populations (Gallin et al., 2024). The variation in race-ethnicities on the impact of PTB, where we found both subtypes of PTBs are associated with increased risk of ASD diagnosis in the offspring of only non-Hispanic White, Hispanic, and Other/mixed race/ethnicities may be a function of sample size as we did not employ any form of stratified sampling. Women of non-Hispanic Black or Asian/Pacific Islander race/ethnicities were a much smaller proportion of the patient population. Even with much smaller sample sizes, we observed that IPTB increased the risk of ASD in Asian-Pacific Islanders. Further studies that employ stratified sampling to over-sample individuals of minority race-ethnicites are needed to fully test the hypothesis that race-ethnicity may be a risk-modifier for the impact of PTB on ASD risk. These studies would also need to determine the role of the disparities in prevalence of pregnancy complications (e.g., preeclampsia) that can lead to IPTB, as well as the screening and interventions for them have in any observed racial-ethnic disparities.

Strengths of our study include the large patient population of diverse socio-economic background in the United States with long-term follow-up that allowed for complete adjustment for potential confounders, and stratified analyses by sex and race-ethnicity. Since the KPSC model of health care is based around vertical integration of the health care resources into insurance, all member women and children in the study have access to health care, and implementation of standardized protocols for screening and diagnosing children for ASD across the network ensures that each child is at equal baseline probability for being diagnosed for the condition. Validation of our ASD-related ICD-10 codes minimizes the risk of patient misclassification. In this study, we characterized PTB by its subtypes using NLP algorithm with remarkable performance (96% PPV) in ascertaining PTB subtypes (Xie et al., 2022).

An additional strength of this study was the extensive sensitivity analyses performed to scrutinize the finding that PTB is a strong risk factor for ASD diagnosis. ASD has a strong genetic component, and women with ASD are at greater risk of having an IPTB, cesarean section, and preeclampsia (Sundelin et al., 2018). Women with higher degrees of autistic traits (as measured by the Autism Spectrum Quotient Japanese Version 10) were found to be more likely to deliver preterm, very early preterm, or to give birth to a baby that was small for gestational age (Hosozawa et al., 2024). We removed cases from our study where the mothers had diagnosis of ASD, depression or any psychosocial disorder in pregnancy or who consumed antidepressants and found that the overall results and conclusions were not affected (Table S1).

Limitations include limited information about exposure to environmental factors, such as exposure to ambient air pollution levels that are known risk factors for both PTB and ASD. Although Race-Ethnicity is commonly used to look at distinct groups of people withing a population, more comprehensive measures of social determinants of health may be more appropriate that take into account other factors such as food insecurity, employment and residential neighborhood. These, however, require additional individual-level data that we do not currently have access to.

Subtype of preterm labor is unlikely a cause of ASD but is rather a marker for underlying pregnancy complications that correlate with ASD. The underlying pregnancy complications that result in SPTB (eg. infection); however, are often quite different from those that result in IPTB (eg. preeclampsia, GDM, placental abruption and PPROM). Therefore, we expected that there would be different impacts of IPTB and SPTB on risk of ASD diagnosis in the offspring as was observed in our study. This suggests that future studies on specific pregnancy complications such as preeclampsia need to be performed with mediation analysis to explore whether IPTB or SPTB is in the causal pathway between pregnancy complications and ASD diagnosis.

We also do not have ready access to the results of screening tests that could be useful to determine if there is a dose-dependent relationship between the severity of PTB and severity of ASD and its impact on the quality of life that can vary widely from child to child. Furthermore, we are also unable to account for the possible effects of tocolytics or other medications received during pregnancy to prevent PTB (e.g., progesterone) that could also affect the risk of ASD and would be confounded with PTB. Nonetheless, the strong association observed in our study suggests that any confounding factor would have to be very strong (E-value > 3.0; Table S1) to account for the observed associations in this study.

Our findings suggest that both IPTB and SPTB increase the risk of ASD in a dose-dependent manner but that the association may be stronger for IPTB. Although the association was observed in both girls and boys, the impact of PTB on the diagnosis of ASD may be stronger for girls. Race-ethnicity may also be a risk modifier, with PTB increasing the risk largely in non-Hispanic White and Hispanic populations.