Periconceptional maternal sleep disturbances and risk of congenital defects: A meta-analysis

Search strategy

We performed a comprehensive systematic review and meta-analysis to investigate the association between maternal PSDs and CDs, adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [Supplementary Material].^[26] The protocol was not registered in PROSPERO; however, the review followed a predefined methodological framework, with clearly specified inclusion criteria, outcomes, and statistical methods established before analysis. We systematically searched PubMed and Medline for eligible studies published from inception through February 2025, without language restrictions to ensure comprehensive inclusion. The search strategy combined controlled vocabulary (MeSH terms) and free-text terms using Boolean operators (Sleep OR sleep disorder OR night sleep OR sleep deprivation OR insomnia OR sleep cycle OR sleep apnea) AND (CHD OR cardiac birth defect OR fetal heart anomaly OR congenital heart malformation) AND (pregnancy OR prepregnancy OR preconception OR periconception OR conception OR prenatal OR early pregnancy). The full search syntax is provided in Supplementary Table S1. Additionally, we manually reviewed bibliographies and citations of eligible studies to identify potentially relevant articles not captured through database queries.

Eligibility criteria

Two independent reviewers (SA and AHK) screened studies based on inclusion criteria: (1) Study design: Observational studies (case-control, cross-sectional, and cohort) involving pregnant women or mother-infant dyads without any interventions; (2) Exposure: Maternal sleep disturbances assessed during the periconceptional or early gestational period; (3) Outcome: Clearly defined CDs diagnosed using consistent and validates methods across groups; (4) Assessment: Sleep disturbances evaluated using validated instruments or diagnostic criteria; (5) Data: Studies reporting effect estimates such as odds ratios (ORs), relative risks (RRs), or hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). We included studies that reported any type of sleep disturbance, without specifying the particular sleep issue. Exclusion criteria included case reports, letters, conference abstracts, editorials, animal studies, and secondary findings such as literature reviews.

Data extraction

Duplicates were removed using the EndNote citation management software.^[27] Two authors (SA and AHK) independently extracted data using a pre-designed data extraction form. Discrepancies were resolved through consensus or consultation with a third reviewer (PK). Extracted variables included the first author’s name, publication year, study title, country, study design, participant characteristics and sample size, type and method of sleep assessment, type and diagnostic criteria for CDs, and adjusted or unadjusted ORs, RRs, or HRs with 95% CIs. When multiple effect estimates were reported, preference was given to adjusted values accounting for confounding factors.

Quality assessment

Study quality was assessed using a Newcastle-Ottawa Scale (NOS) for observational studies.^[28] Two reviewers (SA and AHK) independently scored each study across three domains: Selection (0–4 points), Comparability (0–2 points), and Outcome/Exposure (0–3 points). Discrepancies were resolved through discussion or with a third reviewer (PK). Studies were characterized as bad (0–3), fair (4–6), or good quality (7–9) based on total scores.

Certainty of evidence

The certainty of evidence for the primary outcome (overall CDs) was assessed using the grading of recommendations assessment, development, and evaluation (GRADE) framework, considering risk of bias, inconsistency, indirectness, imprecision, and publication bias.^[29] As all included studies were observational, the certainty of evidence was initially rated as low and subsequently evaluated according to predefined GRADE criteria.

Sensitivity analyses

To assess the robustness of the meta-analysis findings, we conducted three types of sensitivity analyses: (1) Leave-oneout analysis to assess the influence of individual studies on the pooled effect; (2) Study quality-based exclusion to assess the influence of lower-quality evidence by including studies rated as “good” quality in meta-analysis; and (3) Subgroup analysis by study design, i.e., separately analyze case–control versus cohort studies.

Statistical analysis

We synthesized the association between maternal PSDs and CDs using pooled ORs, RRs, or HRs with corresponding 95% CIs. CDs are rare outcomes, occurring in <1% of live births.^[30] For rare outcomes, ORs closely approximate RRs; therefore, they were considered comparable and pooled together, in accordance with established epidemiological guidance.^[31] Heterogeneity across studies was assessed using Cochran’s Q-statistic (significance threshold: p < 0.10) and quantified with the I² statistic. An I² > 50% was considered indicative of moderate-to-high heterogeneity. In cases of substantial heterogeneity, a random-effect model using the DerSimonian-Laird method was employed. Knapp-Hartung adjustments were used to improve the accuracy of CIs under small-sample conditions.

Publication bias was evaluated visually through funnel plot symmetry and quantitatively using Egger’s linear regression test (p < 0.10). All analyses were performed using STATA version 11.0 (Stata Corp LP, College Station, TX, USA).

Study characteristics

The study selection process is detailed in the PRISMA flowchart [Figure 1]. We identified 553 potentially eligible articles from PubMed and Medline. After removing 245 duplicates and excluding 289 irrelevant records based on title and abstract screening, 19 articles were assessed for full-text review. Of these, 14 were excluded – 10 for inappropriate study population or design and 4 for reporting outcomes unrelated to CDs. A total of five observational studies met the eligibility criteria, comprising 6,201 cases and 9,328,416 controls [Table 1]. Of these, three studies evaluated the association between maternal PSDs and overall CDs,^[20,21,23] while two specifically addressed NTDs and CHD.^[10,24]

Close

Figure 1:

Preferred reporting items for systematic reviews and meta-analyses diagram of literature screening and studies selection.

Export to PPT

The studies spanned from 2015 to 2024 and were geographically distributed between China (n = 2) and the United States (n = 3). Two studies were case-control,^[10,24] while three were cohort.^[20,21,23]

Findings from meta-analysis

Pooled analysis revealed a significant positive association between maternal PSDs and the risk of CDs in offspring [Figure 2]. The overall OR was 2.16 (95% CI: 1.57–2.98; p = 0.003), indicating that maternal PSDs were associated with nearly two-fold increased odds of CDs. The prediction interval ranged from 1.19 to 3.93, reinforcing the potential clinical relevance of the findings across various populations.

Moderate heterogeneity was observed across the pooled study (I² = 50.6%, p = 0.09), suggesting some variability in effect sizes across studies, likely attributed to differences in design, exposure measurement, or population characteristics.

Close

Figure 2:

Forest plot of the meta-analysis for selected studies. CI: Confidence interval, IV: Independent variable, OR: Odds ratio, df: Degree of freedom, SE: Standard error.

Export to PPT

Risk of bias assessment

As summarized in the table 2, one study was rated as high quality, and the remaining four were rated as fair based on the NOS. No study was classified as poor, indicating a general low risk of serious bias. Visual inspection of the funnel plot [Figure 3] and Egger’s linear regression test (t = 2.37, p = 0.10) provided no statistically significant evidence of publication bias.

Close

Figure 3:

Begg’s funnel plot for publication bias test.

Export to PPT

Based on the GRADE framework, the overall certainty of evidence for the association between maternal PSDs and CDs was rated as low to moderate [Table 3].

Sensitivity analysis

A leave-one-out analysis demonstrated no single study exerted a disproportionate effect on the pooled estimate. The effect remained stable (OR = 2.16, 95% CI: 1.93–2.42; I2 = 0%, p = 0.91) [Figure 4]. Exclusion of Bourjeily et al.,^[23] reduced the heterogeneity from moderate (I² = 50.6%) to low (I² = 0%), confirming this study as a potential source of variability [Supplementary Figure S1]. A quality-based sensitivity analysis using the only “good”-rated study yielded a comparable OR of 2.49 (95% CI: 1.62–3.82), supporting the consistency and reliability of the overall findings.^[10]

Close

Figure 4:

Sensitivity analysis using leave-one-out method. CI: Confidence interval, IV: Independent variable, OR: Odds ratio, df: Degree of freedom, SE: Standard error.

Export to PPT

Subgroup analyses

Subgroup analyses by study design yielded consistent directionality of effects [Figure 5]. Among the three cohort studies, the pooled OR was 1.88 (95% CI: 1.36–2.60, p = 0.014), indicating a strong, consistent association. The two case–control studies did not reach statistical significance (p = 0.13), possibly due to a smaller sample size and differences in exposure/outcome classification.

Close

Figure 5:

Subgroup analysis (a) Case–control studies, and (b) Cohort studies. CI: Confidence interval, IV: Independent variable, OR: Odds ratio, df: Degree of freedom, SE: Standard error.

Export to PPT

Overview of reviewed evidence

This meta-analysis synthesizes available observational evidence on maternal PSDs and CDs, demonstrating a statistically significant association with an increased pooled risk. The direction of effect was consistent across study designs, although the magnitude varied. Reflecting differences in exposure definitions and covariate adjustment. These findings suggest that sleep disturbances during the periconceptional window may represent an underrecognized risk marker for CDs. However, the observed associations should be interpreted in light of methodological heterogeneity and residual confounding.

Although limited in scope, existing literature supports a potential link between maternal sleep disturbances and CDs [Table 1]. Among the compelling evidence, Li et al.,^[24] reported that PSDs (≥4 days/week) were associated with a fourfold increase in NTDs (OR = 4.1; 95% CI: 1.9–8.8), with the strongest effect noted for spina bifida (OR = 6.4; 95% CI: 2.8–14.5). Zhao et al.,^[10] further linked PSDs to increased risk of both simple (OR = 2.49, 95% CI: 1.62–3.82) and severe CHD (OR = 1.95, 95% CI: 1.27–2.99), while daytime naps appeared protective against simple CHD (OR = 0.62, 95% CI: 0.41–0.93).

Three U.S.-based studies yielded consistent findings. Bourjeily et al.,^[23] reported increased odds of CDs following in-utero exposure to OSA (OR = 1.26, 95% CI: 1.11–1.43), including elevated risk for musculoskeletal anomalies (OR = 1.89, 95% CI: 1.16–3.07). Similarly, Passarella et al.,^[21]and Pressman et al.,^[20] found a comparable risk of CDs in OSA-affected (OR = 2.3, 95% CI: 1.7–3.0) and SDB-affected pregnancies (OR = 1.95, 95% CI: 0.68–5.55), respectively.^[32]

In a descriptive study using the Pittsburgh Sleep Quality Index, OSA was linked to Trisomy 21 risk through observed reductions in maternal alpha-fetoprotein levels during triple screening.^[33] Further, Orabona et al.,^[34] noted that obese women with respiratory distress syndrome due to SDB presented a higher incidence of CDs, supporting the pathophysiological relevance of sleep in early gestation.

Although the periconceptional period was the exposure of interest across all included studies, its operationalization varied. Some studies assessed maternal sleep disturbances during the periconception phase,^[10,24] others during early pregnancy (first trimester),^[20] while studies using administrative or registry-based data often did not specify precise timing within this window.^[21,23] For the purposes of the meta-analysis, these exposure definitions were considered biologically comparable, as they encompass critical stages of fertilization and early organogenesis.^[35] This variability in exposure timing may result in a potential source of clinical heterogeneity, hence to be considered in the interpretation of the pooled estimates.

Biological mechanisms underpinning PSD-related CDs

While mechanistic data remain preliminary, several plausible interconnected pathophysiological pathways have been proposed. Sleep disturbances – especially SDB and PSQ – can induce intermittent hypoxia, oxidative stress, endothelial dysfunction, systemic inflammation, and “hypothalamic-pituitary-adrenal (HPA) axis” activation.^[11,17,36] These changes compromise placental function and fetal nutrient delivery, thereby increasing vulnerability during organogenesis.^[37]

Reduced melatonin levels due to PSDs may lower antioxidant defenses, impairing embryonic cell proliferation and fetal tissue integrity.^[38,39] Elevated maternal cortisol from chronic sleep loss may dysregulate glucocorticoid receptor-mediated gene expression, particularly in neural and cardiac organogenesis.^[36,40] A summary of proposed mechanisms reported across included studies is provided in the table 4.

Challenges and practices to avoid

Certain pharmacologic and behavioral practices during pregnancy may amplify the risk of CDs in the context of PSDs. Exposure to topiramate in the first trimester is associated with an increased risk of oral clefts (OR = 6.26, 95% CI: 3.13–12.51),^[41] while zolpidem, even though uncommon, has been linked to multiple CDs (OR = 1.83– 2.18).^[42] Serotonin reuptake inhibitors exposure in early pregnancy has also been associated with increased CD risk, though evidence is limited by low exposure prevalence.^[43]

Maternal sleep position is another critical factor.^[44] A meta-analysis conducted in low- and middle-income countries revealed that supine sleeping position after 28-week pregnancy increases the risk of stillbirth and small gestational age compared to left-side sleeping, due to compromised uteroplacental blood flow.^[45]

Maternal depression and anxiety exacerbate sleep loss and adversely affect fetal outcomes through placental gene dysregulation, decreased melatonin receptor expression, and elevated cortisol levels,^[46] whereas longer nocturnal sleep durations appear protective against preterm birth.^[47]

Management strategies for PSDs

“Continuous positive airway pressure” and “cognitive behavioral therapy for insomnia” are effective for OSA and chronic insomnia, respectively. Both reduce the risk of gestational hypertension by 35% and preeclampsia by 30%, improve sleep architecture, and reduce placental hypoxia and endothelial dysfunction.^[11,48] Digital psychoeducational interventions have also demonstrated efficacy in reducing antenatal insomnia in randomized trials.^[49]

Melatonin supplementation, though used sparingly, exhibits cardiovascular, anti-inflammatory, and anti-apoptotic benefits.^[22] It crosses the placenta readily and equilibrates in fetal serum within 150 min,^[50] improves uterine artery function, contributing to better maternal outcomes, improves neonatal survival, and lowers blood pressure in offspring.^[51]

Nutritional and lifestyle factors also play a role. High saturated fat diets and sugar-sweetened beverages are associated with insomnia and shorter sleep duration, while excessive caffeine intake in late pregnancy reduces sleep efficiency.^[52] Regular physical activity improves sleep and alleviates symptoms of sleep disorders,^[53] while mindfulness yoga has been associated with improved maternal emotional well-being, lower stress scores, higher vaginal delivery rates, reduced analgesic use, shorter labor durations, higher Apgar scores, and better neonatal outcomes.^[54]

From a clinical and public health perspective, the findings underscore the importance of recognizing sleep disturbances during the periconceptional period as a potential risk marker for adverse fetal outcomes. Although causality cannot be inferred, early identification and counselling regarding sleep health in women planning pregnancy may offer a low-risk opportunity for risk stratification. Future prospective studies incorporating objective sleep measures and standardized exposure definitions are needed to clarify temporal relationships and inform preventive strategies.

Limitations

Despite demonstrating a significant association between maternal PSDs and CDs, this meta-analysis study has several limitations that must be acknowledged.

First, the limited number of included studies (n = 5) and the heterogeneity in the types of CDs examined may constrain the generalizability of our findings. Although statistical heterogeneity was not observed (p = 0.09), a moderate I² value (50.6%) suggests underlying clinical and methodological variation. Sensitivity analysis using the leave-one-out method identified Bourjeily et al.,^[23] as a major contributor to this variability, as their estimates were not adjusted for residual confounders. Upon excluding this study, heterogeneity dropped to 0% (p = 0.53), indicating its disproportionate influence on pooled estimates [Supplementary Figure S1]. Pooling different effect measures comparability under the rare-disease assumption; although widely accepted, this may introduce minor imprecision.

Second, variability in study design (cohort vs. case-control) and timing of sleep assessments introduces the potential for recall bias, particularly in studies relying on retrospective self-reporting of sleep patterns.

Third, inconsistency in sleep assessment tools and the lack of adjustment for potential residual confounders – including maternal BMI, folic acid supplementation, socioeconomic status, and other lifestyle factors – limited our study from conducting meta-regression or evaluating effect modification. Many of these factors are independently associated with both PSDs and the risk of CDs, and incomplete control may lead to overestimation or attenuation of the pooled effect. Notably, studies with more comprehensive adjustment tended to report effect estimates of similar direction but smaller magnitude, suggesting that the observed association is robust but may partially reflect residual confounding.^[35] While OSA was typically diagnosed through objective clinical criteria, most other sleep disturbances were self-reported, potentially inflating measurement bias.

Finally, although no evidence of publication bias was statistically detected (Egger’s test, p = 0.10), the restriction to English-language publications may have introduced geographical and linguistic bias.

Nonetheless, this meta-analysis provides novel quantitative evidence linking PSDs with increased risk of CDs, a relationship that has been underexplored in prior literature.