Two Million Births Link Maternal Sleep Disorders to Child Neurodevelopment, With Caveats

When a pregnant woman lies awake at 3 a.m., the question of whether her sleeplessness matters for her baby’s developing brain has long hovered between folklore and science. A new commentary published in the Journal of Clinical Sleep Medicine by Po-Yang Tsou of Harvard Medical School and Ignacio E. Tapia of the University of Miami Health System takes a hard look at one of the largest attempts yet to answer that question, and reaches a conclusion that is as much about the limits of big data as it is about the power of it. The commentary, published on 2 October 2026, examines a nationwide South Korean cohort study of nearly two million births that found a measurable association between maternal sleep disorders and long-term neurodevelopmental disorders in children, while carefully dissecting what such an association can and cannot prove.

The underlying study, conducted by Oh and Song, drew on South Korea’s National Health Insurance Service database and analyzed 1,993,494 live births between 2008 and 2012. Maternal sleep disorders were identified using ICD-10 diagnostic codes, and children were followed through 2023 for up to fifteen years, with researchers tracking long-term neurodevelopmental disorders defined by the ICD-10 F70 to F95 diagnostic range. Approximately 1.7 percent of the mothers in the cohort carried a coded sleep disorder diagnosis, a figure the commentary emphasizes reflects clinically recognized disease rather than true prevalence, which is known to be considerably higher. After propensity-score matching at a one-to-five ratio, children of mothers with a coded sleep disorder showed a higher risk of any long-term neurodevelopmental disorder, with a hazard ratio of 1.36 and a 95 percent confidence interval of 1.32 to 1.41. The cumulative incidence of these disorders was 13.3 percent in the exposed group compared with 9.9 percent in the matched comparison group.

What makes the signal striking is its consistency. The association held across intellectual, developmental, and behavioral diagnoses alike, and it was stronger when the maternal sleep disorder had been diagnosed before pregnancy rather than during it, with hazard ratios of 1.31 versus 1.18. The finding persisted in analyses restricted to singleton births and to first-time mothers, and it remained stable across strata of maternal psychiatric and obstetric comorbidity. As Tsou and Tapia argue, a signal this consistent is not easily explained away by chance or by any single confounder. Smaller studies have hinted at this link for years, but none were powered to confirm it, and previous nationwide analyses had addressed only single conditions such as maternal sleep apnea. The new work, the commentary concludes, establishes the association as a robust population-level signal.

The biological plausibility behind the association is not in dispute. Sleep disturbance is common in pregnancy, with insomnia, restless legs syndrome, and sleep-disordered breathing each emerging or worsening as gestation advances. Several mechanisms offer plausible routes by which disrupted maternal sleep could alter fetal neurodevelopment: intermittent hypoxia, in which oxygen levels repeatedly dip; glucocorticoid signaling, the hormonal cascade of stress physiology; and systemic inflammation. Animal work adds weight to these pathways. In a 2022 study published in PLoS Biology, researchers showed that a feature of maternal sleep apnea during gestation causes autism-relevant neuronal and behavioral phenotypes in offspring, tying the hypoxic burden of apnea directly to developmental outcomes in a controlled model.

Yet the commentary’s central argument is that the very design enabling this scale simultaneously limits what the data can tell us. The first and most fundamental limitation is intrinsic to administrative databases: a maternal sleep disorder is defined here by a diagnostic code, not a phenotype. Even the apnea-hypopnea index, the standard laboratory measure of sleep apnea severity, would capture only severity rather than the full clinical picture, because obstructive sleep apnea ranges from a sleepy, symptomatic presentation to a minimally symptomatic one. Insomnia is equally diverse. The exposure is also heavily heterogeneous: insomnia accounts for roughly two-thirds of the diagnosed group and likely skews the overall results, while sleep apnea, with only 388 coded cases, and restless legs syndrome, with 453, are far less common and mechanistically distinct. Rarer subtypes are too sparse to interpret at all, with just 17 cases of parasomnia in the entire cohort.

Misclassification cuts in both directions, and the commentary is unusually candid about this. A diagnostic code requires that a mother recognize her symptoms, seek care, and receive a diagnosis. But because disrupted sleep is so often normalized in pregnancy, many affected mothers never receive care at all. Undiagnosed but affected mothers sitting in the comparison group dilute the exposure contrast and attenuate associations toward the null, biasing the study toward finding nothing. Meanwhile, more thorough outcome ascertainment among mothers who do carry a coded diagnosis could introduce bias in the opposite direction, inflating the apparent association. The exposure that matters biologically, the authors stress, is not the diagnosis itself but its physiological correlates: severity, intermittent hypoxia or sleep fragmentation, and cumulative burden over the course of pregnancy. Capturing those physiologic measurements, they argue, is the step that would move an observed association toward a mechanism, and ultimately toward a target for intervention.

The study’s authors reasonably urge early identification and management of maternal sleep disorders, but the commentary warns that transitioning from a known association to an effective fetal-protective treatment is far more complex than it sounds. Even where treatment clearly helps, its benefits may align more closely with a patient’s specific physiology than with their formal diagnosis. The commentary points to a multi-trial analysis published in the European Heart Journal in 2026, which found that the cardiovascular benefits of continuous positive airway pressure, or CPAP, were heavily concentrated in participants with the highest baseline cardiovascular risk physiology rather than spread evenly across all treated patients. Comorbidity compounds the problem further: in clinical practice, starting CPAP while leaving coexisting restless legs syndrome untreated seldom succeeds, because the disorders interact and adherence fails. A diagnostic code cannot anticipate any of this.

Treatment options during pregnancy are also narrower than many readers might assume. Dopaminergic agents and opioids, mainstays for restless legs syndrome outside pregnancy, are avoided in expectant mothers, and several hypnotic medications are discouraged. CPAP remains among the few clearly acceptable options for sleep apnea in pregnancy, supported by systematic review evidence, though the commentary notes that the evidence that treating maternal sleep disorders actually benefits the fetus remains, for now, inconclusive. The constructive path forward, Tsou and Tapia argue, lies in study designs that permit cleaner causal comparison: treated-versus-untreated cohorts phenotyped at baseline, sibling and cousin comparisons that hold family background fixed, and dose-response analyses keyed to disorder severity and timing.

The commentary closes with a three-step roadmap for the field. The first step is recognition. The heavily suppressed 1.7 percent coding rate, compared against a much higher true prevalence of sleep disorders in pregnancy, reflects both the constraints of administrative data and a systemic tendency to overlook maternal sleep health altogether; prospective ascertainment is essential to identify this underserved and uncounted population. The second step is mechanism. Pregnancy drives sleep disorders through distinct pathways, with anatomical airway and weight changes contributing to sleep apnea, iron depletion driving restless legs syndrome, and hyperarousal underpinning insomnia, and objective measurements would capture the physiology that a diagnostic code omits. The third step is treatment, constrained by low awareness and the narrow range of therapies available during pregnancy, and complicated by the challenge of isolating treatment effects from confounding variables across years of developmental follow-up.

For now, the takeaway for clinicians and expectant parents alike is a measured one. The largest study of its kind has confirmed that maternal sleep disorders and childhood neurodevelopmental outcomes travel together at the population level, a finding that smaller cohorts could only suggest. But a diagnostic code is a proxy for a physiology that no database currently records, and the leap from association to intervention demands precisely characterized disorders, objective sleep measurements, and multi-year developmental tracking. By establishing the first robust signal at scale, the Korean cohort study has taken what the commentary calls the essential first step. The harder work of turning that signal into protected fetal brains, the authors make clear, has only just begun.

Subject of Research: The association between maternal sleep disorders during pregnancy and long-term neurodevelopmental outcomes in offspring

Article Title: Beyond the diagnostic code: what two million births can and cannot tell us about maternal sleep and offspring neurodevelopment

Article References: Beyond the diagnostic code: what two million births can and cannot tell us about maternal sleep and offspring neurodevelopment. (n.d.). https://doi.org/10.1007/s44470-026-00142-3

Image Credits: AI Generated

DOI: 10.1007/s44470-026-00142-3

Keywords: maternal sleep disorders, pregnancy, neurodevelopmental disorders, sleep apnea, insomnia, restless legs syndrome, administrative database study, propensity-score matching, intermittent hypoxia, CPAP, South Korea national cohort, fetal brain development