Anesthesia, sex and miscarriage history may influence the association between cesarean delivery and autism spectrum disorder

Autism spectrum disorder (ASD) is a neurodevelopmental disorder in which the exact mechanism remains unclear and is characterized by impairment in social communication, alongside restricted interest and repetitive behaviors [1]. ASD has been of general concern globally, with a substantial increase in prevalence in recent decades [2]. Although ASD is highly genetic [3], there is evidence suggesting the involvement of numerous prenatal and perinatal factors in its development [3‐7]. Among all gestational and obstetric risk factors of ASD, accumulated evidence suggests an association between cesarean section (CS) and ASD which is worrisome given the dramatically increasing trend in CS implementation in numerous countries [8, 9]. A recent meta-analysis demonstrated that a 23% increase in the risk of ASD was associated with CS compared to vaginal delivery [10]. Even though previous studies have reported that CS is not a risk factor for ASD [11, 12], most studies have reached a consistent conclusion [13, 14]. The incidence of CS is increasing globally, with an evaluated annual increase of 4% [15]. Surprisingly, the incidence of CS has increased rapidly in China over the past decade, with a current national rate of approximately 40% [16]. It is becoming increasingly important to fully understand the association between CS and ASD.

The CS is a surgical procedure of delivering a baby to prevent a risk to the health of the mother or baby accompanied by vaginal delivery. Emergency CS is usually performed in an urgent situation due to complications, such as fetal distress which could occur spontaneously [17]. On the contrary, indicated CS (planned or elective CS) is scheduled beforehand and is mostly necessary for obstetric or medical indications such as multiple gestations, suspected macrosomia, or pre-existing obstetric conditions [18]. Nevertheless, in some countries, CS is implemented in a minority of the women, only on maternal request without medical indications (voluntary CS) [19]. Although CS is a relatively safe method compared to vaginal delivery, it has previously been linked to adverse health outcomes in newborns [20‐22]. CS may directly and indirectly affect neurodevelopment, as well as cognitive outcomes [23]. Indirect influence could occur via established correlations between CS and negative outcomes on child health, that is associated with neurodevelopment, including disturbed gut microbiota and changes in stress response, among others [24, 25]. The existing evidence suggests that there is a significant correlation between CS and neurodevelopmental disorders, including ASD and attention deficit hyperactivity disorder (ADHD) [10, 26].

Although the association between CS and ASD has been established in several previous studies, the effect size is heterogeneous when allowing for the adjustment for different confounding factors [27, 28]. Data from previous studies have reported the risks of ASD in children following indicated CS and emergency CS were nearly identical compared with vaginal delivery; however, voluntary CS was not included by existing studies [14, 29]. Contradictory research has suggested that indicated and emergency CS are related to different indications and thus may be associated with differential neurodevelopment process [30]. Alternatively, the observed association may be due to the inclusion of more complete types of surgery indications in the analysis, considering that other confounding factors were controlled.

Emerging evidence has suggested an association between the types of anesthesia and risk of ASD development [14, 31]. CSs are usually performed under general anesthesia (GA) or regional anesthesia (RA); however, RA is more frequently used because GA increases the complications of CS [32, 33]. Exposing the premature brains of the fetuses to obstetric anesthesia can result in histopathological damage, which can affect brain development [34, 35]. The exposure of mothers to GA during labor was significantly associated with higher risk of ASD [14, 31]. On the contrary, inconsistent results have shown that the early exposure of children to anesthesia in the uterus, first 2 years of life, or afterwards is not associated with a higher risk of ASD [36]. Given the contradiction among the conclusions of various research, the susceptibility of a fetus to ASD subjected to obstetric anesthesia during CS requires further exploration.

There could be variation in the observed correlation based on concomitant confounding factors or confounding by indications; this indicates that ASD may be related to the indications and concomitant confounding factors of ASD, rather than the surgical procedure [10]. Considering that different types of CS are accompanied by heterogeneous and complex confounding factors, understanding the risk of ASD among different types of CS is crucial for the early detection and prevention of ASD. Therefore, this study aimed to explore the correlation between different types of CS and ASD, while taking various prenatal and perinatal confounders into account, based on data from a large retrospective case-control study that included comprehensive clinical data on the pregnancy and birth of these children.

In total, 950 ASD cases and 764 subjects (as controls) were included in this study. The median age of the ASD cases and controls were 4.0 and 4.8 years, respectively. Majority of the ASD cases were male (792/158), whereas the sex distribution in the controls was balanced (407/357). Table 1 shows the association between prenatal, perinatal, and infant characteristics and ASD risk, respectively. CS was correlated with the risk of ASD (p <0.001). Apart from CS, other variables that were associated with the risk of ASD included: parity number (p <0.001), history of miscarriage (p <0.001) and infant’s weight at birth (p = 0.001). Behavioral characteristics of the samples are presented in Table 2.

Univariate conditional logistic regression analyses revealed that CS was significantly associated with ASD (adjusted odds ratio [aOR] = 1.606; 95% confidence interval [CI] = 1.311–1.969). This association was equally observed between CS and the parity number, and past history of miscarriage. Further multivariate conditional logistic regression models revealed an association between the risk of CS delivery (vs. vaginal delivery) and ASD. After adjusting for prenatal, perinatal or neonatal confounding factors that were associated with ASD (p < 0.05) in this study (Table 3), the results similarly indicated a significant association between CS and ASD (aOR = 1.573; 95% CI = 1.241–1.992). While parity number was not a risk factor for ASD as demonstrated by univariate conditional logistic regression analyses, multiple regression analyses interestingly identified the interaction term of CS and history of miscarriage (β = − 0.68, Wald χ2 = 7.5, df = 1, p = 0.006) as an influencing factor for ASD.

This study demonstrated that the correlation between CS and ASD is exclusive to CSs performed under GA, with an evident heterogeneity in different surgical indications and sex. In addition, CS may interact with miscarriage history to influence the development of ASD.

According to this study, delivery by CS was associated with a higher risk of ASD in children, even after adjusting the potential confounding factors. The result was supported by previous well-established observations that increased risk of ASD was associated with delivery by CS [13, 27, 29]. Further sex stratification analysis revealed a sex-biased ratio in the risk of developing ASD among births by CS, irrespective of the different surgical indications and method of anesthesia. The risk of ASD among females which was twice higher in children delivered by CS than in those delivered vaginally was consistent with several previous reports [14, 31]. Evidently, females were more sensitive to the long-term effects of CS than males. Therefore, further research is required to investigate the mechanisms underlying the observed sex-related distinctions in the risk of ASD among births by CS.

The results supported the original hypotheses that emergency CS is associated with a higher risk of ASD than voluntary CS and indicated CS, and the risk of ASD following indicated CS is slightly higher than the risk following voluntary CS. Consistent results were observed in previous studies that both indicated CS (elective CS or planned CS) and emergency CS were associated with an increased risk of ASD, contrary to this study [14, 29, 31]. These discrepancies may be attributed to the heterogeneity of subgroups relative to surgical indications employed in different studies, contrary to this study. Previous analysis included emergency CS and indicated CS (elective CS or planned CS), excluding voluntary CS. Considering this result, emergency CS may confer additional risk of ASD beyond risks associated with factors leading to CS among indicated CS and voluntary CS. This indicates that CS may exclusively exert an inherent risk on ASD, and there might be an induced risk from complex rather than unequal confounding factors from different CS surgical indications (not CS exclusively). Existing evidence supports these hypotheses; CS, whether planned or emergency, appears to be correlated with a complex mix of confounding factors equally related to ASD risk [39]. A sibling design study revealed that neither emergency CS nor elective CS was correlated with ASD after adjusting for familial confounding factors, indicating that the correlation may not have exclusively originated from CS; in addition, it suggests that potential unknown familial factors might explain the association between CS and ASD, which strongly supports our viewpoints. Concomitant prenatal or perinatal risk factors among children delivery by CS can vary with different surgery indications [30]. This indicates that underlying maternal or fetal indications for CS may be important risk factors of ASD.

In males, emergency CS were associated with an increased risk of ASD in comparison with vaginal delivery, which was consistent with the conclusion drawn from the result of the total study population. On the contrary, no association was observed between emergency CS and ASD in females; instead, both voluntary CS and indicated CS significantly increased the risk of ASD. The rationale behind these discrepancies is yet to be accurately clarified as previous stratified analysis regarding surgical indications is limited. It was speculated that different surgical indications may exert distinct impact on the risk of ASD between males and females. A complex interaction of surgical indications and sex plays a pivotal role in the development of ASD. In addition to the ‘interaction’ hypothesis, another explanation regarding these discrepancies is that the small sample size of female subjects made it difficult to draw accurate conclusions. Thus, the contribution of different surgical indications to the unequal risk of development of ASD among females relative to males require further exploration.

Another crucial finding is that CS performed under GA were related to a higher risk of ASD than CS under RA or vaginal delivery, following adjustments for potential confounders consistently between males and females in further analysis stratified by to sex. This finding is in accordance with previous research which reported a significantly higher risk among children who were delivered by CS under GA [14, 31]; however, this contradicted other research which reported no association between GA and ASD (Creagh et al. 2016). Nonetheless, this study was designed to investigate the association between exposure to GA and ASD across periods before, during, and after delivery, rather than the effect of GA performed during CS on the risk of ASD. This discrepancy could imply an interaction between the effects of anesthesia and CS on the risk of ASD. Similarly, in the stratification by sex, a stronger association was observed between ASD and females than males; the explanation for this difference may be attributed to the greater susceptibility of females to the long-term effects of GA than males. Increasing evidence from animal models and observational human research has supported the hypothesis that GA during delivery is associated with neurotoxicity, affecting later neurodevelopment [32, 40, 41]. Another possibility is that the reason behind the GA in pregnant women is a risk factor for ASD and may have nothing to do with the CS itself. GA was generally administered for bleeding emergencies and foetal distress [42, 43], and the complications associated with these emergencies may be risk factors for ASD. However, we did not further explore the reasons for the use of GA in pregnant women, so it is impossible to control these confounding factors. Although this study, as well as previous studies, have controlled the confounding factors as much as possible, the possibility of the contribution of certain undocumented confounding factors or GA-related situations to the correlations between GA and risk of ASD cannot be excluded. Certain risks which are inherent to CS cannot be completely excluded, given the observation from a previous study that CS is associated with neurodevelopmental impairment [44]. Thus, whether GA independently or in interaction with CS contributes to the development of ASD remains unclear.

A novel finding in this study is that an interaction between miscarriage history and CS contributes to ASD risk. Only one previous study has reported miscarriage as a specific risk factor of ADHD and ASD [38], which is inconsistent with a recent study, which was similar to ours, that revealed no association between recurrent miscarriage and ASD [14]. In addition, the sample size was insufficient to detect the association between ASD and miscarriage. This study reported that a higher risk of ASD was identified among subjects with miscarriage history compared to those without miscarriage history. Further, stratified analysis suggested that no association was observed between CS and ASD in the subjects with miscarriage history. On the contrary, in subjects without miscarriage history, the risk of ASD following CS was significantly higher compared to that in vaginal deliveries, indicating miscarriage history may exert a greater impact on the risk of ASD than CS; this can be attributed to the covering effect of miscarriage history on CS regarding the risk of ASD among subjects with miscarriage history. One possible explanation of the mechanism underlying the interaction between CS and ASD is that miscarriage may cause disturbed gut microbiota and stress response [45]. Thus, it was speculated that miscarriage history and CS could induce analogous ASD risk factors, even if they are heterogeneous in nature. This demonstrates that miscarriage history and CS may play a synergistic role in the development of ASD by inducing similar risk factors; this strongly supported the view that the observed association may be due to confounding by indication rather than the CS procedure itself. However, to our knowledge, none of the previous publications has revealed an interaction between CS and the miscarriage history; thus, the mechanism for this interaction needs to be further clarified.

This study has some limitations. Limited known prenatal and perinatal risk factors of ASD were detected in this study; therefore, sufficient potential confounders could not be adjusted in multivariate analysis, such as crying, color at birth, convulsions at birth and jaundice, especially APGAR index is not available in many families given the retrospective study. Consequently, the findings could have been influenced by potential unmeasured confounding factors. However, the CS is a major exposure factor and therefore unlikely to be seriously affected by this shortcoming. Next, the imbalance in sex distribution could have affected the accuracy of the research results, although further sex-specific analysis was conducted; the small sample size of female subjects limited the number of strata used in the subgroup analyses. In addition, a sibling analysis was not performed in this study, and we were unable to further adjust for familial confounding factors. All data obtained were self-reported. Therefore, the results may be influenced by potential recall bias and exposure misclassification. Nevertheless, the exposure factor (CS) is a dominant event and is therefore unlikely to be influenced by recall bias; in addition, it was sufficient to investigate most of the hypotheses. It is also worth mentioning that the only diagnostic criterion, DSM-4, was adopted in our study, and the failure to use ADOS is another shortcoming. Lastly, our study does not further explore the possible correlation between the CS and the level of autistic symptoms, and further exploratory studies are needed in the future.

Our current research shows that CS is associated with ASD; and CS performed under GA were related to a higher risk of ASD than CS under RA. The associations between different types of CS and ASD is influenced by sex and miscarriage history.