Association of Cesarean Delivery With Risk of Neurodevelopmental and Psychiatric Disorders in the Offspring

1Cesarean delivery can effectively reduce maternal and neonatal mortality and morbidity in the presence of complications, such as antepartum hemorrhage, fetal distress, abnormal fetal presentation, and hypertensive disease. During the past 3 decades, worldwide cesarean delivery rates have seen a more than 3-fold increase, from approximately 6% in 1990 to 21% in 2015, with substantial variations among and within countries. There is no precise information on appropriate cesarean delivery rates at the population level, although the international health care community considers 10% to 15% to be an optimal rate. Public health concerns have been raised because access to medically indicated cesarean delivery may be difficult in low-resource settings and unnecessary cesarean deliveries may be performed in high-resource settings.,,

2Despite being a life-saving procedure in the presence of complications, no evidence, to our knowledge, indicates that cesarean delivery, if not indicated, is beneficial for the offspring. On the contrary, previous studies have reported negative health outcomes in offspring born via cesarean delivery, including obesity,, allergy, asthma,, type 1 diabetes, and acute lymphoblastic leukemia. Cesarean delivery also has a potential association with early brain development. Previous studies have reported worse child cognitive development and higher rates of autism spectrum disorders (ASD) associated with cesarean delivery. Whether cesarean delivery is associated with increased risks of other neurodevelopmental and psychiatric disorders is unclear.

3Previous studies,, have discussed the hypothesized biological mechanisms that may explain the associations between cesarean delivery and negative health outcomes in the offspring. For example, cesarean delivery might alter immune development through the perturbation of bacterial colonization, disturbing immune and sensory activation through lack of stress response or modifying epigenetic regulation in DNA methylation.,, Moreover, these adverse effects might vary depending on whether a cesarean delivery is performed before the onset of labor (ie, elective cesarean delivery) or after (ie, emergency cesarean delivery). Elective cesarean delivery can be scheduled by obstetricians in the presence of medical indications (eg, breech presentation, cephalopelvic disproportion) or requested by the mother because of culture preferences, previous negative birth experience, or fear of birth,, which may indicate a genetic vulnerability to factors associated with psychological changes, such as stress, and associated in turn with neurodevelopmental or psychiatric illness in the offspring. However, emergency cesarean delivery occurs in more traumatic situations (eg, severe fetal distress, preeclampsia) that are associated with multiple negative birth outcomes. Elective and emergency cesarean deliveries are characterized by different factors and thus may be differentially associated with neurodevelopmental and psychiatric outcomes.

4To address these gaps in the literature, we conducted a systematic review and meta-analysis of observational studies to quantify the extent of the association between cesarean delivery and a range of neurodevelopmental and psychiatric outcomes compared with vaginal delivery. Moreover, we investigated whether type of cesarean delivery (elective or emergency) or assisted vaginal delivery (ie, involving use of vacuum or forceps), compared with unassisted vaginal delivery, were differentially associated with neurodevelopmental and psychiatric outcomes in the offspring.

5We reported this systematic review and meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) and Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guidelines. The study was preregistered with International Prospective Register of Systematic Reviews (PROSPERO identifier CRD42018108298) before data extraction and analyses. We searched Ovid MEDLINE, Embase, Web of Science, and PsycINFO from inception to December 19, 2018, without language, date, or location restrictions. The search strategy was developed in association with information specialists at the Karolinska Institutet Library. Search terms included all main mental disorders in the Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) (DSM-5). The search strategy for each database is detailed in eTable 1 in the Supplement. We checked reference lists of relevant reviews for additional studies.

6We included observational studies that allowed estimation of the associations between obstetric mode of delivery (cesarean vs vaginal delivery) and neurodevelopmental and psychiatric disorders in the offspring. Studies were included regardless of the method of exposure ascertainment (ie, self-report or birth records). To maximize the quality of the included work, we only included studies if the outcome diagnoses were assessed through structured interviews or using standardized diagnostic criteria (eg, International Classification of Diseases, DSM, or equivalent). Self-reported or caregiver-reported outcomes were excluded. We further excluded review articles, book chapters, conference abstracts, and dissertations. Articles in languages other than English were translated by a native speaker or using Google Translate. Two researchers (T.Z. and L.S.-C. or A.V.-P.) independently screened and selected the articles first on the basis of titles and abstracts and then by examining the full texts; discrepancies were resolved through discussion with a senior researcher (A.S.).

7Two researchers (T.Z. and L.S.-C. or A.V.-P.) independently extracted the data using predesigned Excel spreadsheets (Microsoft Corp). Disagreements were resolved through discussion. Extracted variables included country; data source; study design; age of the offspring; exposure; method of exposure ascertainment; diagnosis; diagnostic instruments; covariates controlled for by adjustment or matching; sample size; whether the study reported elective and emergency cesarean delivery or assisted and unassisted vaginal delivery separately and study participants in each group; and risk estimates (eg, odds ratios [ORs], hazard ratios). If the original study reported data on several outcomes, information on each outcome was retrieved separately. If studies reported no effect size for association of interest, we calculated ORs from raw data. When duplicate data were identified, we extracted data from the largest sample size for relevant outcomes (characteristics of the articles excluded because of duplicate data are presented in eTable 2 in the Supplement).

8We assessed the methodologic quality of each study using the Newcastle-Ottawa Scale. Two researchers (T.Z. and L.S.-C. or A.V.-P.) independently assessed and scored each study according to the preestablished criteria. We judged the study quality to be high if the score was at least 7 points (of a possible 9) or otherwise to be low.

9First, we examined the association between birth by cesarean delivery vs vaginal delivery and each outcome in the offspring separately using random effects models. We reported the results obtained after pooling each individual study’s most-adjusted estimate as ORs with 95% CIs. Second, we used the same approach to explore the associations of elective cesarean delivery, emergency cesarean delivery, and assisted vaginal delivery compared with unassisted vaginal delivery with the risk of neurodevelopmental and psychiatric disorders in the offspring. If the original studies reported the results from the analyses of both population data and sibling comparisons,,,,,, we retrieved the former to reduce heterogeneity related to potential differences in study populations. For 8 studies,,,,,,, that did not report ORs but reported hazard ratios, we calculated crude ORs from the data reported in the article using a 2 × 2 table.

10Statistical heterogeneity among studies was assessed using Q and I2 statistics. For the Q statistic, a 2-sided P <.10 was considered as representative of statistically significant heterogeneity, and I2 values of 25%, 50%, and 75% were regarded as low, moderate, and high heterogeneity, respectively. If 10 or more original studies were included in the meta-analysis for an outcome, univariate meta-regressions on publication year, cesarean delivery use proportion, study design, country income status, and exposure ascertainment were applied to assess whether they were responsible for the heterogeneity. We assessed the presence of potential publication bias for each outcome using Egger tests and visualized by funnel plots. If the Egger tests revealed a potential publication bias, we used the contour-enhanced funnel plot and Duval and Tweedie nonparametric trim and fill method to further test the data.

11Furthermore, we undertook subgroup analyses for any differences in the association between birth by cesarean delivery vs vaginal delivery and each outcome for the following 6 moderators: country status (high- vs middle-income countries, according to the World Bank), proportion of cesarean deliveries in the study (lower vs higher than 15% as a proxy indicator for unnecessary cesarean delivery), exposure ascertainment (medical records vs self-report), offspring sex, whether risk measures were adjusted for parental psychiatry history, and study quality (high vs low). The results of the subgroup analyses were reported if at least 3 original studies were present in each stratum. We also performed influence analyses (so-called leave-1-out analysis) by iteratively removing 1 study at a time to confirm that the findings were not influenced by any single study. Finally, we performed sensitivity analyses using the least-adjusted or crude estimates from each study to assess whether the pooled effect size was sensitive to adjustment strategy.

12All analyses were performed in Stata, version 15.1 (StataCorp). A 2-sided P <.05 was considered to be statistically significant.

13We identified 6953 articles, of which 549 abstracts were selected for detailed assessment (Figure 1). A total of 61 studies comprising 67 independent samples and 20 607 935 deliveries met our inclusion criteria. The main characteristics of the included studies are presented in Table 1, and the variables adjusted and/or matched in each study are presented in eTable 3 in the Supplement. Of the 61 included studies, 27 studies,,,,,,,,,,,,,,,,,,,,,,,,,, reported on the association of cesarean delivery with a diagnosis of ASD (59 795 cases), 13 studies,,,,,,,,,,,, with attention-deficit/hyperactivity disorder (ADHD) (92 718 cases), 3 studies,, with intellectual disabilities (485 cases), 3 studies,, with tic disorders (6181 cases), 4 studies,,, with eating disorders (4550 cases), 3 studies,, with obsessive-compulsive disorder (OCD; 7295 cases), 5 studies,,,, with major depression or affective psychoses (8561 cases), and 7 studies,,,,,, with nonaffective psychoses (7195 cases). Study quality was defined as high for 18 of 22 cohorts (82%) and 20 of 40 case-control studies (50%). Quality assessment summary scores are reported in Table 1.

14Results of the meta-analysis for the first aim revealed that birth by cesarean delivery was significantly associated with increased odds of the offspring being diagnosed with ASD (OR, 1.33; 95% CI, 1.25-1.41) and ADHD (OR, 1.17; 95% CI, 1.07-1.26) compared with birth by vaginal delivery (Figure 2). The magnitude of the estimates was similar or higher for other neurodevelopmental and psychiatric outcomes, but the associations were not statistically significant (possibly because of the limited number of studies): intellectual disabilities (OR, 1.83; 95% CI, 0.90-3.70), OCD (OR, 1.49; 95% CI, 0.87-2.56), tic disorders (OR, 1.31; 95% CI, 0.98-1.76), and eating disorders (OR, 1.18; 95% CI, 0.96-1.47). The OR for depression/affective psychoses was 1.06 (95% CI, 0.98-1.14) and for nonaffective psychoses was 0.97 (95% CI, 0.78-1.21) (Figure 3).

15For intellectual disabilities, OCD, tic disorders, and eating disorders, leave-1-out analysis revealed statistically significant associations with cesarean delivery births after omitting the following studies: Chen and colleagues (for intellectual disabilities; OR after exclusion, 2.61; 95% CI, 1.51-4.53), Vasconcelos and colleagues (for OCD; OR, 1.13; 95% CI, 1.08-1.19), and Cubo and colleagues (for tic disorders; OR, 1.24; 95% CI, 1.02-1.50) (eTable 4 in the Supplement). For the rest of the disorders, no individual studies influenced the results seen in the main analyses.

16Study heterogeneity was high for ADHD (I2 = 79.2%), intellectual disabilities (I2 = 88.2%), tic disorders (I2 = 75.6%), and eating disorders (I2 = 92.7%); medium for ASD (I2 = 69.5%), OCD (I2 = 67.3%), and nonaffective psychosis (I2 = 82.6%); and low for depression and affective psychoses (I2 = 0.0%). Egger tests and funnel plots suggested a potential publication bias for ASD (eFigure 1 in the Supplement). In addition, a contour-enhanced funnel plot for ASD supported the Egger test, suggesting that the asymmetry was likely caused by publication bias (eFigure 2 in the Supplement). Furthermore, the Duval and Tweedie nonparametric trim and fill method for ASD led to a result similar to that obtained in the main analysis (OR, 1.32; 95% CI, 1.24-1.40).

17For the second aim, 20 studies,,,,,,,,,,,,,,,,,,, (10 620 715 deliveries) and 18 studies,,,,,,,,,,,,,,,,, (10 597 234 deliveries) assessed the risk of ASD, ADHD, tic disorders, and affective and nonaffective psychoses in offspring born via elective and emergency cesarean delivery, each compared with offspring born via unassisted vaginal delivery. Similar to the results of the main analyses, both elective and emergency cesarean deliveries were associated with increased odds of ADHD and ASD; for these analyses, heterogeneity was low (I2 range, 10.7%-33.6%) except for ASD when assessing the association with elective cesarean delivery (I2 = 87.9%) (eFigure 3 in the Supplement). Elective and emergency cesarean deliveries were not associated with tic disorders or affective and nonaffective psychoses. In the 23 studies,,,,,,,,,,,,,,,,,,,,,,, (8 415 429 deliveries) that assessed the associations with assisted vs unassisted vaginal delivery, we did not observe any association with any of the outcomes of interest except for tic disorders (OR, 1.28; 95% CI, 1.17-1.41; I2 = 0.0%) (eFigure 4 in the Supplement).

18Results of our subgroup analyses for study design revealed that, compared with offspring born via vaginal delivery, the odds of offspring born via cesarean delivery being diagnosed with ASD or ADHD were significantly higher when pooling the results of case-control studies than when pooling those of cohort studies (Table 2). In studies that indicated unnecessary use of cesarean delivery (proportions ≥15%), we observed higher odds of ASD compared with studies with cesarean delivery proportions less than 15% (Table 2). Study quality scores and exposure assessment did not significantly affect the results (Table 2 and eTable 5 in the Supplement). Subgroup analyses on offspring sex and parental psychiatric history were not performed because of too few (<3) studies on each outcome with data available. Owing to data availability, we only performed meta-regression analyses for ASD and ADHD. Country income level and exposure ascertainment were associated with between-study variance for ADHD, but residual heterogeneity still remained significant (eTable 6 in the Supplement). None of the 5 factors examined (ie, publication year, proportion of cesarean deliveries, study design, country income status, and exposure ascertainment) were associated with heterogeneity or reduced residual heterogeneity. For our sensitivity analyses, we replicated all the analyses using crude or least-adjusted estimates from the same studies (eFigure 5 in the Supplement). No significant differences between the models were detected, although heterogeneity increased in the model including the least-adjusted estimates.

19This systematic review and meta-analysis explored the association of cesarean delivery with a wide range of neurodevelopmental and psychiatric outcomes. Compared with vaginal delivery, cesarean delivery was associated with increased risk of several neurodevelopmental and psychiatric disorders. The estimates for ASD (33% increased odds) and ADHD (17% increased odds) were statistically significant, but other disorders, such as learning disabilities, tic disorders, OCD, and eating disorders, presented with similar or higher odds, although these were not statistically significant, possibly because of the modest number of studies. In the leave-1-out analysis, associations with intellectual disabilities, OCD, and tic disorders became statistically significant after the exclusion of single outlier studies. By contrast, the associations were close to the null and nonsignificant for depression and affective and nonaffective psychotic disorders, indicating some specificity. The association between cesarean delivery and offspring psychopathologic disorders contrasts with findings for assisted vaginal delivery (eg, forceps delivery), which was associated only with increased odds of tic disorders, whereas the association was close to the null with the remaining outcomes.

20Our findings are consistent with previous meta-analyses conducted with a limited range of psychiatric outcomes, such as ASD or schizophrenia, and considerably expand the evidence by including all available data on a broad range of neurodevelopmental and psychiatric disorders. This inclusion was in part possible because of increased power afforded by the publication of large, nationwide registry-based studies on psychiatric disorders. For example, Curran and colleagues could not confirm an association between cesarean delivery and ADHD because of the availability of only 2 suitable studies at the time. We were able to include 13 studies on ADHD, of which 7 used data from population-based registries,,,,, or national birth cohorts.

21For the first time, to our knowledge, we were able to examine elective cesarean delivery separately from emergency cesarean delivery in association with neurodevelopmental and psychiatric disorders. This distinction is theoretically important because it may hint at potentially different implicated mechanisms in the association between cesarean delivery and neurodevelopmental and psychiatric disorders. Of interest, the odds of ASD and ADHD in offspring born via elective and emergency cesarean delivery were nearly identical compared with unassisted vaginal delivery. The results for tic disorders and affective and nonaffective psychoses were less conclusive because of the few studies with small sample sizes. Of note, there have been no clear definitions of elective and emergency cesarean delivery in the literature, which may have influenced our results toward the null because of potential nondifferential exposure misclassification. Future studies should carefully define elective and emergency cesarean delivery, for example, using the Robson classification system.

22Subgroup analyses and meta-regressions (only performed for ADHD and ASD) suggested a potential association between higher rates of cesarean delivery and the presence of disorders in the offspring. Although it seems reasonable to assume that cesarean delivery is overused when exceeding a 15% rate of the total deliveries, additional studies are needed to investigate the reasons behind potentially unnecessary cesarean delivery to properly evaluate costs and benefits. Our subgroup analyses by type of study design found that the odds for ASD and ADHD were significantly weaker in cohort studies compared with case-control studies, indicating a potential bias in the main meta-analysis, in which studies of both designs were pooled together. This finding is also reflected in our quality assessment, in which cohort studies were more likely to be classified as high quality; however, subgroup analyses did not detect significant group differences between high- and low-quality studies. Future research should examine how potential bias, such as confounders, in particular indications for cesarean delivery, contribute to the observed associations.

23A strength of our review is the comprehensive scope of the literature search across multiple neurodevelopmental and psychiatric disorders, across 19 countries, in all languages, and with low publication bias. We used 4 different bibliographic databases and conducted the screening, data extraction, and quality assessment in duplicate. The main limitation of our study is the high level of heterogeneity in all outcomes except for depression and affective psychoses. Heterogeneity was decreased when we separated elective and emergency cesarean deliveries. Nevertheless, neither the subgroup nor meta-regression analyses could fully explain the high level of heterogeneity. Several other factors might have contributed to it. First, indication for cesarean delivery likely contributes to the heterogeneity among studies, but we were unable to explore the role of indications because of a lack of data in the original studies. The medical indications in the mothers may play an important role in the observed associations between cesarean delivery and the outcomes. Second, baseline age, age at diagnosis, and follow-up time substantially varied among studies and were not reported in some studies, which precluded us from exploring the role of these factors as contributors to heterogeneity. Third, methodologic differences in adjustment for parental, perinatal, and fetal factors in each study might account for the variance among studies. Studies with both population and sibling analyses suggest that the observed associations are likely attributable to familial confounding because the significant findings from the population-level analyses were attenuated in sibling comparisons.,,,, We therefore acknowledge the existence of unadjusted confounders and that the observed associations cannot be explained by only a single factor but are likely to be multifactorial.

24Our study findings suggest that birth by cesarean delivery is associated with certain neurodevelopmental and psychiatric disorders. The results appear to further add to the known adverse health outcomes associated with cesarean delivery and suggest judicious use of cesarean delivery.,, Statistical heterogeneity was high in the meta-analysis of some specific outcomes despite the use of strict inclusion criteria and our attempts to address the source of heterogeneity in subgroup analyses and meta-regression. This finding might suggest that other confounders, such as indication for cesarean delivery, could contribute to explaining some of the variation across studies. Future research should include further adjustment for potential confounders and consider genetically sensitive designs, such as sibling comparisons or twin and adoption studies. The mechanisms underlying the observed associations remain unknown and require empirical investigation to examine whether cesarean delivery plays a causal role in the development of neurodevelopmental and psychiatric disorders.