Decoding labour epidural analgesia and autism

In 2020, the publication of work reporting a 37% relative increase in the risk of autism spectrum disorder (ASD) in the offspring of mothers who had received labour epidural analgesia (LEA)1 led to widespread scrutiny in the lay press and social media, potentially influencing maternal decision-making all over the world. However, it remains imperative to approach this matter critically, appreciating the complexity of association and causation and the need for a balanced interpretation of the available evidence. Currently, within the developed world, LEA stands as the benchmark for managing labour pain. The technique is widely acknowledged as safe and beneficial to both the mother and the foetus and consistently yields high patient satisfaction. The WHO recommends LEA in healthy women depending on a woman's preferences.2 However, the long-term effects, and more specifically the neurodevelopmental effects of LEA on the offspring have recently been a subject of inquiry. ASD is a neurodevelopmental condition characterised by challenges in social communication and the manifestation of restricted, repetitive behaviours or interests. Epidemiologically, ASD is observed in approximately 2 to 3% of the United States' population. The observation of an apparent upward trend in prevalence over recent years has been attributed, at least in part, to heightened ascertainment practices but environmental factors might also bear a degree of responsibility.3 Various factors identified in epidemiological studies have been associated with ASD, yet none exhibit absolute specificity for ASD development. The complex, poorly understood, cause of ASD appears to be intertwined with a convergence of heterogeneous genetic and environmental factors, including maternal health, infectious and pharmacological causes, social influences and environmental exposures. Maternal factors, including gestational hypertension, maternal overweight status before or during pregnancy, pre-eclampsia and maternal age of 35 years or older, have been identified as correlating with elevated rates of ASD in the offspring. Other cohort and case–control studies have reported an association between the incidence of ASD in the offspring and advanced paternal age, medication use during pregnancy, and both short (<12 months) and prolonged (72 months) interpregnancy intervals. Concerning the relationship between medication use during pregnancy and ASD risk, exposure to valproic acid and the utilisation of selective serotonin reuptake inhibitors (SSRIs) during pregnancy have also been associated with an increased risk for ASD. An observational, retrospective cohort study by Qiu et al.1 reported a 37% relative increase in the risk of ASD in the offspring of mothers who had received LEA. In addition, they delineated a discernible dose–response relationship between LEA and the likelihood of ASD.4,5 The study had some significant limitations. There was a lack of data about labour complications (e.g. prolonged labour, foetal distress and malposition), which are known to increase the risk of ASD. These factors are, in turn, associated with increased LEA and are important confounders. Women who delivered by caesarean section (itself an important association with autism) were excluded. Additionally, children not enrolled with the health insurance scheme by 1 year of age were excluded. It is possible that these children were from lower socio-economic groups, who are historically subject to underdiagnosis of autism, and who were also over-represented in the nonepidural cohort. Despite these limitations, the study raised substantial concerns regarding the safety of LEA and implications for public health. Following this initial study, various epidemiological cohort studies have sought to explore the suggested association between LEA and the increased prevalence of ASD in the offspring. In this European Journal of Anaesthesiology issue, Zamstein and co-workers (pp. 282–287) report the results of their retrospective population-based cohort study, which shows that LEA was not independently associated with ASD after adjusting for confounding factors. This conclusion corroborates the findings of other retrospective cohort studies.6–11 Furthermore, several studies employing a sibling-matched retrospective cohort design revealed in the preliminary unmatched analysis an elevated hazard ratio, which dissipated after sibling-matching, culminating in an adjusted hazard ratio that did not exhibit an increase.6,7,9–11 This adds complexity to the debate, underscoring the challenges of isolating specific variables in observational studies. Sibling matching offers a distinct advantage over other observational studies by mitigating bias from residual confounding attributed to unmeasurable variables that persist consistently within families. Similarly, the only prospective observational birth cohort study found no statistically significant correlation between maternal LEA and autistic traits. The assessment was conducted prospectively using a battery of ASD screening tests up to the age of 7 years.12 These important results address the existing gap in the literature by employing a standardised prospective neurobehavioral battery for all participants (and not depending on a clinical diagnosis of ASD) as primary outcome. This holds paramount significance as there is a growing acknowledgement that subclinical autistic traits manifest to diverse extents within the general population, with a clinical diagnosis being established only when individuals exhibit substantial functional impairment and psychological or social distress. Studies employing administrative databases, such as those focused on particular education data and health records, generally tend to underestimate the prevalence of ASD when juxtaposed with investigations utilising more meticulous methodologies for case identification. Meta-analyses conducted on this topic present an intriguing picture.13–16 On one hand, they reveal an association between neuraxial labour analgesia and autism when no confounding variables are considered. However, a critical turning point emerges when confounders are pondered, resulting in a very weak association. Moreover, this association was also based on low-quality evidence.14 A recent systematic review by Fang and co-workers, revealed a notably elevated risk of ASD when the influence of confounding variables was disregarded [hazard ratio 1.3, 95% confidence interval (CI), 1.25 to 1.3521, and hazard ratio 1.24, 95% CI 1.14 to 1.345]. However, upon accounting for potential confounders, the effect diminished, yet remained statistically significant [hazard ratio 1.13, 95% CI, 1.03 to 1.25, and hazard ratio 1.11, 95% CI, 1.06 to 1.165].16 Nevertheless, when only studies using a sibling-matched design were analysed, this meta-analysis concluded that there was no significant association (hazard ratio 1.07, 95% CI, 0.99 to 1.16),13,16 whereas another demonstrated a significant association in sibling-matched designed studies [hazard ratio 1.10, 95% CI, 1.02 to 1.18],13 with, however, no dose–response effect. It was concluded that the small statistically significant association between LEA and ASD may be elucidated by selection bias and residual (unmeasured) confounders, such as genetic factors.16 A final but critical consideration is the role of the biological plausibility of the potential role of LEA in the development of ASD. Qui et al. speculated that the local anaesthetic used in the mother's epidural space may have a neurotoxic effect on the foetal brain, increasing the risk of autism. However, while epidurally administered local anaesthetics can cross the placenta easily because of their low molecular weight, studies in humans indicate negligible systemic absorption with low concentrations in the neonatal bloodstream17,18 and evidence of foetal local anaesthetic metabolism8,19; therefore, any argument for causation is severely weakened. The lack of a consistent dose–response effect further complicates matters, making it challenging to establish a straightforward relationship between neuraxial labour analgesia and autism. It is, therefore, crucial to approach the findings of Qiu et al.1 and related meta-analyses13–16 with caution. The presence of an association without confounders does not imply causation. The intricacies of human biology and the myriad factors influencing neurodevelopment demand a meticulous examination of the evidence. The concern surrounding LEA and autism, initially fuelled by the Qiu article and subsequent media coverage, requires a tempered response. The strength of evidence hinges on the robustness of its underlying data. Although systematically reviewing the available evidence suggests a weak association, the significance of this association is contingent on the consideration of (unmeasured) confounding variables. The inconclusive results of sibling matching, the absence of an apparent dose–response effect and the first negative prospective observational study emphasise the need for continued research before drawing definitive conclusions. As we navigate the delicate balance between scientific inquiry and public perception, it is paramount to communicate clearly. For now, the available evidence should provide reassurance to prospective parents, enabling them to make informed and personalised decisions regarding labour analgesia, free from anxiety and guilt associated with the choice of LEA.