[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Download PDF
Table 1.  
Perinatal and Sociodemographic Characteristics Related to Autism Spectrum Disorder (ASD) and Mode of Delivery Among Those Born in Sweden From 1982 Through 2010
Perinatal and Sociodemographic Characteristics Related to Autism Spectrum Disorder (ASD) and Mode of Delivery Among Those Born in Sweden From 1982 Through 2010
Table 2.  
Association Between Mode of Delivery and Autism Spectrum Disorder (ASD) With and Without Intellectual Disability (ID) in the Offspring
Association Between Mode of Delivery and Autism Spectrum Disorder (ASD) With and Without Intellectual Disability (ID) in the Offspring
Table 3.  
Association Between Mode of Delivery and Autism Spectrum Disorder (ASD) Among Sibling Pairs
Association Between Mode of Delivery and Autism Spectrum Disorder (ASD) Among Sibling Pairs
1.
Hollander  E, Kolevzon  A, Coyle  JT. Textbook of Autism Spectrum Disorders. Arlington, VA: American Psychiatric Publishing, Inc; 2010.
2.
Elsabbagh  M, Divan  G, Koh  YJ,  et al.  Global prevalence of autism and other pervasive developmental disorders. Autism Res. 2012;5(3):160-179.
PubMedArticle
3.
Developmental Disabilities Monitoring Network Surveillance Year 2010 Principal Investigators; Centers for Disease Control and Prevention (CDC).  Prevalence of autism spectrum disorder among children aged 8 years: Autism and Developmental Disabilities Monitoring Network, 11 sites, United States, 2010. MMWR Surveill Summ. 2014;63(2):1-21.
PubMed
4.
Hansen  SN, Schendel  DE, Parner  ET.  Explaining the increase in the prevalence of autism spectrum disorders: the proportion attributable to changes in reporting practices. JAMA Pediatr. 2015;169(1):56-62.
PubMedArticle
5.
Sandin  S, Lichtenstein  P, Kuja-Halkola  R, Larsson  H, Hultman  CM, Reichenberg  A.  The familial risk of autism. JAMA. 2014;311(17):1770-1777.
PubMedArticle
6.
Hultman  CM, Sparén  P, Cnattingius  S.  Perinatal risk factors for infantile autism. Epidemiology. 2002;13(4):417-423.
PubMedArticle
7.
Curran  E, O’Neill  S, Cryan  J,  et al.  Birth by caesarean section and development of autism spectrum disorder and attention-deficit/hyperactivity disorder: a systematic review and meta-analysis. J Child Psychol Psychiatry. 2015;56(5):500-508.
PubMedArticle
8.
Mackay  DF, Smith  GC, Dobbie  R, Cooper  SA, Pell  JP.  Obstetric factors and different causes of special educational need: retrospective cohort study of 407,503 schoolchildren. BJOG. 2013;120(3):297-307.
PubMedArticle
9.
Cryan  JF, Dinan  TG.  Mind-altering microorganisms: the impact of the gut microbiota on brain and behaviour. Nat Rev Neurosci. 2012;13(10):701-712.
PubMedArticle
10.
Cho  CE, Norman  M.  Cesarean section and development of the immune system in the offspring. Am J Obstet Gynecol. 2013;208(4):249-254.
PubMedArticle
11.
Chien  LN, Lin  HC, Shao  YH, Chiou  ST, Chiou  HY.  Risk of autism associated with general anesthesia during cesarean delivery: a population-based birth-cohort analysis. J Autism Dev Disord. 2015;45(4):932-942.
PubMedArticle
12.
Betrán  AP, Merialdi  M, Lauer  JA,  et al.  Rates of caesarean section: analysis of global, regional and national estimates. Paediatr Perinat Epidemiol. 2007;21(2):98-113.
PubMedArticle
13.
Declercq  E, Young  R, Cabral  H, Ecker  J.  Is a rising cesarean delivery rate inevitable? trends in industrialized countries, 1987 to 2007. Birth. 2011;38(2):99-104.
PubMedArticle
14.
Tully  KP, Ball  HL.  Misrecognition of need: women’s experiences of and explanations for undergoing cesarean delivery. Soc Sci Med. 2013;85:103-111.
PubMedArticle
15.
Almqvist  C, Cnattingius  S, Lichtenstein  P, Lundholm  C.  The impact of birth mode of delivery on childhood asthma and allergic diseases: a sibling study. Clin Exp Allergy. 2012;42(9):1369-1376.
PubMedArticle
16.
Khashan  AS, Kenny  LC, Lundholm  C, Kearney  PM, Gong  T, Almqvist  C.  Mode of obstetrical delivery and type 1 diabetes: a sibling design study. Pediatrics. 2014;134(3):e806-e813.
PubMedArticle
17.
Ludvigsson  JF, Otterblad-Olausson  P, Pettersson  BU, Ekbom  A.  The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol. 2009;24(11):659-667.
PubMedArticle
18.
The National Board of Health and Welfare (Socialstyrelsen). In English – the Swedish Medical Birth Register.http://www.socialstyrelsen.se/register/halsodataregister/medicinskafodelseregistret/inenglish. Accessed January 14, 2015.
19.
Donovan  SJ, Susser  E.  Commentary: advent of sibling designs. Int J Epidemiol. 2011;40(2):345-349.
PubMedArticle
20.
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Arlington, VA: American Psychiatric Association; 2013.
21.
Sullivan  PF, Magnusson  C, Reichenberg  A,  et al.  Family history of schizophrenia and bipolar disorder as risk factors for autism. Arch Gen Psychiatry. 2012;69(11):1099-1103.
PubMedArticle
22.
Rai  D, Lee  BK, Dalman  C, Golding  J, Lewis  G, Magnusson  C.  Parental depression, maternal antidepressant use during pregnancy, and risk of autism spectrum disorders: population based case-control study. BMJ. 2013;346:f2059.
PubMedArticle
23.
Lee  BK, Magnusson  C, Gardner  RM,  et al.  Maternal hospitalization with infection during pregnancy and risk of autism spectrum disorders. Brain Behav Immun. 2015;44:100-105.
PubMedArticle
24.
Ludvigsson  JF, Andersson  E, Ekbom  A,  et al.  External review and validation of the Swedish National Inpatient Register. BMC Public Health. 2011;11:450.
PubMedArticle
25.
Langridge  AT, Glasson  EJ, Nassar  N,  et al.  Maternal conditions and perinatal characteristics associated with autism spectrum disorder and intellectual disability. PLoS One. 2013;8(1):e50963.
PubMedArticle
26.
Idring  S, Rai  D, Dal  H,  et al.  Autism spectrum disorders in the Stockholm Youth Cohort: design, prevalence and validity. PLoS One. 2012;7(7):e41280.
PubMedArticle
27.
Guinchat  V, Thorsen  P, Laurent  C, Cans  C, Bodeau  N, Cohen  D.  Pre-, peri- and neonatal risk factors for autism. Acta Obstet Gynecol Scand. 2012;91(3):287-300.
PubMedArticle
28.
Dodds  L, Fell  DB, Shea  S, Armson  BA, Allen  AC, Bryson  S.  The role of prenatal, obstetric and neonatal factors in the development of autism. J Autism Dev Disord. 2011;41(7):891-902.
PubMedArticle
29.
Lahey  BB, D’Onofrio  BM.  All in the family: comparing siblings to test causal hypotheses regarding environmental influences on behavior. Curr Dir Psychol Sci. 2010;19(5):319-323.
PubMedArticle
30.
Obel  C, Olsen  J, Henriksen  TB,  et al.  Is maternal smoking during pregnancy a risk factor for hyperkinetic disorder? findings from a sibling design. Int J Epidemiol. 2011;40(2):338-345.
PubMedArticle
31.
De Rubeis  S, He  X, Goldberg  AP,  et al; DDD Study; Homozygosity Mapping Collaborative for Autism; UK10K Consortium.  Synaptic, transcriptional and chromatin genes disrupted in autism. Nature. 2014;515(7526):209-215.
PubMedArticle
32.
Iossifov  I, O’Roak  BJ, Sanders  SJ,  et al.  The contribution of de novo coding mutations to autism spectrum disorder. Nature. 2014;515(7526):216-221.
PubMedArticle
33.
Glasson  EJ, Bower  C, Petterson  B, de Klerk  N, Chaney  G, Hallmayer  JF.  Perinatal factors and the development of autism: a population study. Arch Gen Psychiatry. 2004;61(6):618-627.
PubMedArticle
34.
Gregory  SG, Anthopolos  R, Osgood  CE, Grotegut  CA, Miranda  ML.  Association of autism with induced or augmented childbirth in North Carolina Birth Record (1990-1998) and Education Research (1997-2007) databases. JAMA Pediatr. 2013;167(10):959-966.
PubMedArticle
35.
Burstyn  I, Sithole  F, Zwaigenbaum  L.  Autism spectrum disorders, maternal characteristics and obstetric complications among singletons born in Alberta, Canada. Chronic Dis Can. 2010;30(4):125-134.
PubMed
36.
Lavender  T, Hofmeyr  GJ, Neilson  JP, Kingdon  C, Gyte  GML.  Caesarean section for non-medical reasons at term. Cochrane Database Syst Rev. 2012;3:CD004660.
PubMed
37.
Frisell  T, Öberg  S, Kuja-Halkola  R, Sjölander  A.  Sibling comparison designs: bias from non-shared confounders and measurement error. Epidemiology. 2012;23(5):713-720.
PubMedArticle
38.
Ludvigsson  JF, Reichenberg  A, Hultman  CM, Murray  JA.  A nationwide study of the association between celiac disease and the risk of autistic spectrum disorders. JAMA Psychiatry. 2013;70(11):1224-1230.
PubMedArticle
Original Investigation
September 2015

Association Between Obstetric Mode of Delivery and Autism Spectrum DisorderA Population-Based Sibling Design Study

Author Affiliations
  • 1The Irish Centre for Fetal and Neonatal Translational Research (INFANT), Department of Obstetrics and Gynaecology, Cork University Maternity Hospital, Cork, Ireland
  • 2Division of Public Health Epidemiology, Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden
  • 3Department of Epidemiology and Public Health, University College Cork, Cork, Ireland
  • 4Alimentary Pharmabiotic Centre, Department of Anatomy and Neuroscience, University College Cork, Cork, Ireland
  • 5Alimentary Pharmabiotic Centre, Department of Psychiatry, University College Cork, Cork, Ireland
JAMA Psychiatry. 2015;72(9):935-942. doi:10.1001/jamapsychiatry.2015.0846
Abstract

Importance  Because the rates of cesarean section (CS) are increasing worldwide, it is becoming increasingly important to understand the long-term effects that mode of delivery may have on child development.

Objective  To investigate the association between obstetric mode of delivery and autism spectrum disorder (ASD).

Design, Setting, and Participants  Perinatal factors and ASD diagnoses based on the International Classification of Diseases, Ninth Revision (ICD-9),and the International Statistical Classification of Diseases, 10th Revision (ICD-10),were identified from the Swedish Medical Birth Register and the Swedish National Patient Register. We conducted stratified Cox proportional hazards regression analysis to examine the effect of mode of delivery on ASD. We then used conditional logistic regression to perform a sibling design study, which consisted of sibling pairs discordant on ASD status. Analyses were adjusted for year of birth (ie, partially adjusted) and then fully adjusted for various perinatal and sociodemographic factors. The population-based cohort study consisted of all singleton live births in Sweden from January 1, 1982, through December 31, 2010. Children were followed up until first diagnosis of ASD, death, migration, or December 31, 2011 (end of study period), whichever came first. The full cohort consisted of 2 697 315 children and 28 290 cases of ASD. Sibling control analysis consisted of 13 411 sibling pairs.

Exposures  Obstetric mode of delivery defined as unassisted vaginal delivery (VD), assisted VD, elective CS, and emergency CS (defined by before or after onset of labor).

Main Outcomes and Measures  The ASD status as defined using codes from the ICD-9 (code 299) and ICD-10 (code F84).

Results  In adjusted Cox proportional hazards regression analysis, elective CS (hazard ratio, 1.21; 95% CI, 1.15-1.27) and emergency CS (hazard ratio, 1.15; 95% CI, 1.10-1.20) were associated with ASD when compared with unassisted VD. In the sibling control analysis, elective CS was not associated with ASD in partially (odds ratio [OR], 0.97; 95% CI, 0.85-1.11) or fully adjusted (OR, 0.89; 95% CI, 0.76-1.04) models. Emergency CS was significantly associated with ASD in partially adjusted analysis (OR, 1.20; 95% CI, 1.06-1.36), but this effect disappeared in the fully adjusted model (OR, 0.97; 95% CI, 0.85-1.11).

Conclusions and Relevance  This study confirms previous findings that children born by CS are approximately 20% more likely to be diagnosed as having ASD. However, the association did not persist when using sibling controls, implying that this association is due to familial confounding by genetic and/or environmental factors.

Introduction

Autism spectrum disorder (ASD) is a disorder characterized by impairment in social interaction and communication with the presence of restricted interest and repetitive behaviors.1 Autism spectrum disorder is thought to affect an estimated 0.62% of children worldwide,2 although recent estimates in the United States have been closer to 1.5%.3 The prevalence of ASD has increased drastically since 1980, yet changes in diagnostic criteria might only explain 60% of this increase.4

Although ASD is highly heritable,5 it has previously been linked to numerous perinatal factors,6 possibly including birth by cesarean section (CS).7 There are several possible mechanisms underlying this association, including early-term (ie, 37-39 weeks’ gestation) birth,8 exposure to altered microbiota,9 changes in stress response,10 and type of anesthesia.11 The observed association may alternatively be due to residual confounding or confounding by indication, meaning ASD could be associated with the indication for CS rather than the CS itself,7 or to an unknown genetic factor that is associated with increased risk of CS and ASD. The prevalence of birth by CS is estimated to be 15% worldwide and is more than 20% in developed countries.12 Because the prevalence of ASD is increasing worldwide, for a variety of reasons,13,14 it is becoming increasingly important to fully understand any possible long-term effects of birth by CS.

Birth by CS has previously been linked to several long-term outcomes, including asthma, allergic rhinitis, diabetes mellitus, and gastrointestinal disease.10 Notably, recent population-based studies that used siblings as controls found that these reported associations could be explained by genetic factors or family environment for asthma15 and type 1 diabetes.16 Thus, it is important to consider residual confounding from environmental and genetic factors with appropriate methods.

The objective of this investigation was to assess the effect of mode of delivery, specifically birth by CS, on ASD using data from a large Swedish population-based registry. To our knowledge, this is the largest study on this subject to date7 and the only one to use the combination of adjusted cohort and sibling control study designs. Using this combination and taking familial environmental and genetic confounding factors into consideration, we aimed to robustly assess any association between mode of delivery and ASD.

Methods
Study Population

The study cohort consisted of all singleton live births in Sweden from January 1, 1982, through December 31, 2010, using data from the Swedish Medical Birth Register, the Swedish National Patient Register, and the Swedish Multi-Generation Register. Data from these registers can be linked using the personal identification number given to each Swedish resident. More than 99% of all births in Sweden are recorded in the Swedish Medical Birth Register.17,18

We started follow-up from the first birthday (left censored), consistent with a previous study5 on ASD in this population. Before 1987, no code for autism was available in the register;5 therefore those persons who turned 1 year of age before 1987 began follow-up on January 1, 1987. Children were followed up until first diagnosis of ASD, death, migration, or December 31, 2011 (end of study period), whichever came first.

Sibling design studies have been used in investigations of prenatal or perinatal risk factors and were developed as a way to control for shared genetic factors and familial environment.19 For the sibling study, we included all sibling pairs that included the first 2 children for each mother, with one diagnosed as having ASD and one not. Because conditional logistic regression does not adjust for differential follow-up time, follow-up time during the study period was longer for the control sibling (ie, the control sibling died, emigrated, or left the study at an older age than when their sibling was diagnosed as having ASD). The case sibling could be the younger or older sibling. This approach is consistent with a previous sibling control study16 that used this data set. Ethical approval was obtained from the research ethics committee at Karolinska Institutet. Informed consent was waived by the ethics committee.

Exposure

Mode of delivery was divided into 4 categories: unassisted vaginal delivery (VD), assisted VD, elective CS, and emergency CS. Assisted VD was defined as VD with the use of forceps or vacuum extraction. Unassisted VD included spontaneous and induced VD. Elective CS and emergency CS were defined as CS before and after onset of labor, respectively. Although the Swedish Medical Birth Register began in 1973, variables indicating emergency and elective CS are available from 1982, which marks the beginning of our investigation.

Outcome

Information on ASD status and date of first diagnosis was obtained from the Swedish National Patient Register. In line with the DSM-5, we included all pervasive developmental disorders as cases of ASD,20 including codes from the International Classification of Diseases, Ninth Revision (ICD-9) (code 299), and the International Statistical Classification of Diseases, 10th Revision (ICD-10) (code F84).2123 The Swedish National Patient Register includes inpatient data from 1973 and outpatient data from 2001,24 with good outpatient coverage from 2006. Children in Sweden undergo a mandatory developmental assessment at 4 years of age, and children with suspected developmental disorders are referred for further assessment by a child psychiatry unit.5 The procedures for this assessment are standardized across Sweden. To account for the possibility of increased diagnosis using this method, we conducted a sensitivity analysis on children born on or before 2007 to allow everyone to be followed up at least until 4 years of age. Previous results have indicated that risk factors may differ for ASD with and without intellectual disability (ID).25 Therefore, we looked at ASD overall and at ASD with and without ID (defined as an IQ <70 and functional impairment26).

Statistical Analysis

We performed Cox proportional hazards regression analysis to estimate hazard ratios (HRs) and 95% CIs. We assessed the proportional hazards assumption using graphical displays of the empirical score process. Although originally violated, when analyses were adjusted for year of birth and stratified by every fourth birth year, the assumption was no longer violated. Therefore, all analyses were partially adjusted in this way. In fully adjusted models, we also included infant sex,27 maternal age,27 gestational age,27 maternal and paternal citizenship,27 small for gestational age,27 large for gestational age, 5-minute Apgar score,27 parity,27 social welfare status, family disposable income,28 and maternal and paternal depression,22 bipolar disorder, and nonaffective disorder21 in the total population (eAppendix 1 in the Supplement). We used robust CIs, with analysis clustered on maternal ID, to account for increased likelihood of ASD diagnosis within families. Sensitivity analyses are detailed in eAppendix 2 in the Supplement.

For the sibling study, we conducted conditional logistic regression using nonaffected siblings as controls, matched on maternal ID. Estimates for the association between mode of delivery and ASD were obtained using pairs discordant on ASD and mode of delivery, although pairs concordant on mode of delivery were included in the analysis because they contribute to the covariate estimates. The conditional logistic regression analysis included adjustment for the same covariates as the Cox proportional hazards regression model with the exception of maternal country of birth, which was the same for each sibling pair. We conducted a sensitivity analysis that included only full siblings (ie, same father and mother).

All analyses were conducted using SAS statistical software, version 9.4 (SAS Institute Inc).

Results
Descriptive Statistics

There were 2 941 629 live births in Sweden from January 1, 1982, through December 31, 2010. After excluding 77 209 multiple births, 17 968 children who died or emigrated before 1 year of age, 87 whose first diagnosis of ASD was before 1 year of age, and 149 050 children with unknown mode of delivery, 2 697 315 remained in the final cohort. A total of 2 161 148 children (80.1%) were born via nonassisted VD, 164 305 (6.1%) via elective CS, 175 803 (6.5%) via emergency CS, and 196 058 (7.3%) via assisted VD (eFigure in the Supplement). A total of 28 290 children (1.0%) were diagnosed as having ASD (Table 1). Median age at diagnosis was 12 years (interquartile range [IQR], 7-16 years), with 11 years (IQR, 7-16 years) in boys and 14 years (IQR, 9-17 years) in girls.

Cox Proportional Hazards Regression Analysis

In partially adjusted analyses (ie, adjusted for year of birth), elective CS (HR, 1.39; 95% CI, 1.33-1.45), emergency CS (HR, 1.40; 95% CI, 1.34-1.46), and assisted VD (HR, 1.18; 95% CI, 1.13-1.23) were all associated with ASD when compared with unassisted VD (Table 2). After full adjustment, the adjusted HRs (aHRs) were reduced, but the association remained significant for elective CS (aHR, 1.21; 95% CI, 1.15-1.27) and emergency CS (aHR, 1.15; 95% CI, 1.10-1.20) but not for assisted VD (aHR, 1.03; 95% CI, 0.98-1.07). Sensitivity analyses that excluded children who were preterm, female, small for gestational age, large for gestational age, born after 2007, and had mothers who had ever been diagnosed as having depression did not materially change the results (eTable 1 in the Supplement). Inclusion of children diagnosed before 1 year of age also had no effect. Exclusion of children who had been born before 1990 and 2000 similarly did not materially change the results (Table 2). Inclusion of parental educational level after exclusion of those born before 1990 had no effect on the results (eTable 2 in the Supplement). Sensitivity analyses on birth order, CS order, and family size had no effect on the results (eTable 3 in the Supplement). Similarly, inclusion of induction of labor also did not change the results (eTable 4 in the Supplement). Sensitivity analyses on maternal age, obesity, diabetes, and hypertension did not have an effect on the results (eTable 5 in the Supplement). Likewise, analyses regarding breech presentation did not have an effect (eTables 6-8 in the Supplement).

When cases of ASD were stratified by the presence of ID, assisted VD was not significantly associated with ASD with ID (aHR, 1.11; 95% CI, 1.00-1.22) or without ID (aHR, 1.01; 95% CI, 0.96-1.06) (Table 2). Elective CS was significantly associated with ASD with (aHR, 1.26; 95% CI, 1.14-1.39) and without ID (aHR, 1.19; 95% CI, 1.12-1.26). Similarly, emergency CS was associated with ASD with (aHR, 1.25; 95% CI, 1.14-1.37) and without ID (aHR, 1.12; 95% CI, 1.06-1.18). Exclusion of children born before 1990 and 2000 did not change results materially.

Sibling Control

In the sibling control study, there were 13 411 sibling pairs discordant on ASD; 2555 pairs were also discordant on mode of delivery (with one sibling born by unassisted VD). In the partially adjusted analysis, there was no association between assisted VD (odds ratio [OR], 1.04; 95% CI, 0.91-1.15) or elective CS (OR, 0.97; 95% CI, 0.85-1.11) and ASD (Table 3). After full adjustment, assisted VD (OR, 0.91; 95% CI, 0.82-1.02) and elective CS (OR, 0.89; 95% CI, 0.76-1.04) were not associated with ASD. Conversely, although emergency CS was associated with ASD in partially adjusted analysis (OR, 1.20; 95% CI, 1.06-1.36), this association did not persist after full adjustment (OR, 0.97; 95% CI, 0.85-1.11). Because of the lack of overall ASD association, we did not further analyze the effect of mode of delivery on ASD stratified by ID status. Exclusion of children born before 1990 and 2000 did not materially change the results. Sensitivity analysis with only the 11 634 full biological siblings (ie, shared both father and mother) did not change results (eTable 9 in Supplement).

Discussion

The current study combines 2 approaches to assess the effect of obstetric mode of delivery on ASD: a conventional cohort analysis, which we used to test the association in a large population, and a sibling control analysis, which we used to control for familial confounding and assess causality. In the conventional cohort analysis, children born by elective CS were 21% more likely to be diagnosed as having ASD after controlling for known confounders, similar to previous findings regarding birth by CS and ASD.7 However, in the sibling control analysis, no association was found between mode of delivery and ASD.

Although the traditional cohort analysis revealed birth by CS to be associated with ASD, it is not necessarily a cause because the association could be due to residual confounding. Because randomizing mode of delivery would be unethical, large population-based cohort studies are the most robust way to assess an association. In such cases, the use of siblings as controls is seen as a way to test the causality of an association when randomized experiments are not feasible.29 This use of siblings as controls adjusts for shared factors that relate to genetics and family environment that may not be measured or fully understood.30 Therefore, because the association between birth by CS and ASD did not persist in the sibling control analysis, we can conclude that there is no causal association. It is more likely that birth by CS is related to some unknown genetic or environmental factor that leads to increased risk of both CS and ASD.

There have been recent developments in understanding genetic risk factors for ASD,31,32 but it is likely that, even among inherited cases, genetic and environmental factors interact.27 Therefore, although birth by CS may not have a causal association with ASD, future research into environmental risk factors for ASD is warranted. However, although CS may not have a causal association with ASD, it is not without health risks for the mother and fetus and should not be undertaken without due regard for those risks.

Previous Studies

A systematic review and meta-analysis7 on mode of delivery and ASD reported a 23% increased risk of ASD in relation to CS, although studies have been limited. Only one investigation in recent years, conducted in Australia by Glasson et al,33 has made use of a sibling design. Glasson et al33 detected no significant effect of birth by elective CS on ASD overall but found a significant effect separately on autism, Asperger syndrome, and pervasive developmental disorder not otherwise specified. However, the study did not provide adjusted estimates beyond the use of sibling controls and was much smaller than the current study (n = 473 sibling pairs). Three previous cohort studies25,34,35 have reported an adjusted estimate of the association between CS and ASD; however, all were subject to some limitation. Burstyn et al35 and Gregory et al34 did not provide separate estimates for emergency and elective CS, making it difficult to distinguish between the effect of CS and possible confounding by indication because an association with emergency CS may be more related to complications during labor rather than the CS itself. Langridge et al25 distinguished between emergency and elective CS and also stratified by ID status similar to this study. However, the present study accounts for known confounders that were not included in study by Langridge et al,25 such as family income and parental psychiatric health.

Strengths and Limitations

Our study is the largest on this subject to date, with 2.6 million births included in our analysis. In addition, because data were prospectively obtained from national registries, they were not subject to recall bias or selection bias. In addition, because of the nature of the data set, we were able to adjust for a variety of known potential confounders. Even after controlling for these variables, we found birth by elective CS to be associated with a 20% increased risk of ASD, which is comparable to previous estimates.7 We were also able to conduct sensitivity analyses by birth cohort to determine whether the association changed in recent years, which was a limitation in some previous studies.34 This approach also allowed us to address possible bias due to changes in ASD reporting practices over time. That we saw no change over time in the traditional cohort analysis or sibling analysis leads us to conclude that changes in ASD definition and reporting did not affect the association. Finally, because of such large numbers and quality of data, we were able to conduct a sibling design study, which helped control for unknown genetic and environmental confounders, after which the association did not persist.

There are also limitations to the present investigation. Although we were able to differentiate between emergency and elective CS based on timing of labor, we did not have access to the indication for CS. In addition, it is possible that parents who stopped having children after their first was diagnosed as having ASD were different from parents with multiple children, one or more of whom were diagnosed as having ASD. Although our sensitivity analysis on only children did not reveal a difference in the overall association, we were unable to include these children in the sibling analysis and so were unable to see whether the association was attenuated in a similar manner when genetic and family environment were controlled for. In addition, because we have mostly inpatient data until 2001, severe cases of ASD were likely overrepresented, and the association between mode of delivery and ASD may be different for milder forms of ASD. However, because sensitivity analysis from 2000 onward revealed no difference in the results, this association likely did not affect our findings. Moreover, although ours is the largest study to date, to our knowledge, on this subject, we reduced the sample size in the sibling control analysis because of strict inclusion criteria, which is reflected in the relatively wider CIs. In addition, modes of delivery were likely not independent between siblings because women who have previously given birth by CS are more likely to give birth by CS,36 and it is possible that such relatedness with the exposure could lead to increased bias in the sibling control design.37 However, the results of our sensitivity analyses based on the cohort analysis, conducted by birth order, CS order, and family size, would imply that the difference in results was in fact due to the comparison to siblings rather than other factors, such as cohort size or clustering of exposure. There are several known limitations to sibling control analyses,19,37 including an inability to distinguish whether the confounding is due to genetics, family environment, or both.19 For example, we were unable to measure any epigenetic changes related to CS that differed among siblings and may have been associated with ASD. In addition, in the presence of measurement error, sibling control analysis could lead to an artificial decrease in the association.37 A limitation of the use of anonymized registry data is that we were not able to validate cases. However, the positive predictive value of psychiatric disorders in this registry has previously been estimated to be 85% to 95%,24 and a review of the medical records for 88 cases of ASD found a positive predictive value of 94.3%.38

Conclusions

This study confirms previous results that children born by CS are approximately 20% more likely to be diagnosed as having ASD after adjustment for known confounders. However, the lack of association in analyses using sibling controls implies that this association is not causal and therefore is due to unknown genetic or environmental factors.

Back to top
Article Information

Submitted for Publication: February 13, 2015; final revision received April 14, 2015; accepted April 16, 2015.

Corresponding Author: Ali S. Khashan, PhD, The Irish Centre for Fetal and Neonatal Translational Research, Department of Obstetrics and Gynaecology, Cork University Maternity Hospital, Fifth Floor, Wilton Road, Wilton County, Cork, Ireland (a.khashan@ucc.ie).

Published Online: June 24, 2015. doi:10.1001/jamapsychiatry.2015.0846.

Author Contributions: Ms Curran and Dr Khashan had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

Study concept and design: All authors.

Acquisition, analysis, or interpretation of data: Curran, Kearney, Kenny, Cryan, Dinan, Khashan.

Drafting of the manuscript: Curran, Dalman.

Critical revision of the manuscript for important intellectual content: Kearney, Kenny, Cryan, Dinan, Khashan.

Statistical analysis: Curran, Khashan.

Administrative, technical, or material support: Dalman, Khashan.

Study supervision: Kearney, Kenny, Dinan, Khashan.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by grant 12/RC/2272 from The Irish Centre for Fetal and Neonatal Translational Research (Dr Kenny).

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

Additional Contributions: Henrik Dal, MSc, Division of Public Health Epidemiology, Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden, provided data management support and advice. Cecilia Magnusson, PhD, Division of Public Health Epidemiology, Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden, provided advice on the study design.

References
1.
Hollander  E, Kolevzon  A, Coyle  JT. Textbook of Autism Spectrum Disorders. Arlington, VA: American Psychiatric Publishing, Inc; 2010.
2.
Elsabbagh  M, Divan  G, Koh  YJ,  et al.  Global prevalence of autism and other pervasive developmental disorders. Autism Res. 2012;5(3):160-179.
PubMedArticle
3.
Developmental Disabilities Monitoring Network Surveillance Year 2010 Principal Investigators; Centers for Disease Control and Prevention (CDC).  Prevalence of autism spectrum disorder among children aged 8 years: Autism and Developmental Disabilities Monitoring Network, 11 sites, United States, 2010. MMWR Surveill Summ. 2014;63(2):1-21.
PubMed
4.
Hansen  SN, Schendel  DE, Parner  ET.  Explaining the increase in the prevalence of autism spectrum disorders: the proportion attributable to changes in reporting practices. JAMA Pediatr. 2015;169(1):56-62.
PubMedArticle
5.
Sandin  S, Lichtenstein  P, Kuja-Halkola  R, Larsson  H, Hultman  CM, Reichenberg  A.  The familial risk of autism. JAMA. 2014;311(17):1770-1777.
PubMedArticle
6.
Hultman  CM, Sparén  P, Cnattingius  S.  Perinatal risk factors for infantile autism. Epidemiology. 2002;13(4):417-423.
PubMedArticle
7.
Curran  E, O’Neill  S, Cryan  J,  et al.  Birth by caesarean section and development of autism spectrum disorder and attention-deficit/hyperactivity disorder: a systematic review and meta-analysis. J Child Psychol Psychiatry. 2015;56(5):500-508.
PubMedArticle
8.
Mackay  DF, Smith  GC, Dobbie  R, Cooper  SA, Pell  JP.  Obstetric factors and different causes of special educational need: retrospective cohort study of 407,503 schoolchildren. BJOG. 2013;120(3):297-307.
PubMedArticle
9.
Cryan  JF, Dinan  TG.  Mind-altering microorganisms: the impact of the gut microbiota on brain and behaviour. Nat Rev Neurosci. 2012;13(10):701-712.
PubMedArticle
10.
Cho  CE, Norman  M.  Cesarean section and development of the immune system in the offspring. Am J Obstet Gynecol. 2013;208(4):249-254.
PubMedArticle
11.
Chien  LN, Lin  HC, Shao  YH, Chiou  ST, Chiou  HY.  Risk of autism associated with general anesthesia during cesarean delivery: a population-based birth-cohort analysis. J Autism Dev Disord. 2015;45(4):932-942.
PubMedArticle
12.
Betrán  AP, Merialdi  M, Lauer  JA,  et al.  Rates of caesarean section: analysis of global, regional and national estimates. Paediatr Perinat Epidemiol. 2007;21(2):98-113.
PubMedArticle
13.
Declercq  E, Young  R, Cabral  H, Ecker  J.  Is a rising cesarean delivery rate inevitable? trends in industrialized countries, 1987 to 2007. Birth. 2011;38(2):99-104.
PubMedArticle
14.
Tully  KP, Ball  HL.  Misrecognition of need: women’s experiences of and explanations for undergoing cesarean delivery. Soc Sci Med. 2013;85:103-111.
PubMedArticle
15.
Almqvist  C, Cnattingius  S, Lichtenstein  P, Lundholm  C.  The impact of birth mode of delivery on childhood asthma and allergic diseases: a sibling study. Clin Exp Allergy. 2012;42(9):1369-1376.
PubMedArticle
16.
Khashan  AS, Kenny  LC, Lundholm  C, Kearney  PM, Gong  T, Almqvist  C.  Mode of obstetrical delivery and type 1 diabetes: a sibling design study. Pediatrics. 2014;134(3):e806-e813.
PubMedArticle
17.
Ludvigsson  JF, Otterblad-Olausson  P, Pettersson  BU, Ekbom  A.  The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol. 2009;24(11):659-667.
PubMedArticle
18.
The National Board of Health and Welfare (Socialstyrelsen). In English – the Swedish Medical Birth Register.http://www.socialstyrelsen.se/register/halsodataregister/medicinskafodelseregistret/inenglish. Accessed January 14, 2015.
19.
Donovan  SJ, Susser  E.  Commentary: advent of sibling designs. Int J Epidemiol. 2011;40(2):345-349.
PubMedArticle
20.
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Arlington, VA: American Psychiatric Association; 2013.
21.
Sullivan  PF, Magnusson  C, Reichenberg  A,  et al.  Family history of schizophrenia and bipolar disorder as risk factors for autism. Arch Gen Psychiatry. 2012;69(11):1099-1103.
PubMedArticle
22.
Rai  D, Lee  BK, Dalman  C, Golding  J, Lewis  G, Magnusson  C.  Parental depression, maternal antidepressant use during pregnancy, and risk of autism spectrum disorders: population based case-control study. BMJ. 2013;346:f2059.
PubMedArticle
23.
Lee  BK, Magnusson  C, Gardner  RM,  et al.  Maternal hospitalization with infection during pregnancy and risk of autism spectrum disorders. Brain Behav Immun. 2015;44:100-105.
PubMedArticle
24.
Ludvigsson  JF, Andersson  E, Ekbom  A,  et al.  External review and validation of the Swedish National Inpatient Register. BMC Public Health. 2011;11:450.
PubMedArticle
25.
Langridge  AT, Glasson  EJ, Nassar  N,  et al.  Maternal conditions and perinatal characteristics associated with autism spectrum disorder and intellectual disability. PLoS One. 2013;8(1):e50963.
PubMedArticle
26.
Idring  S, Rai  D, Dal  H,  et al.  Autism spectrum disorders in the Stockholm Youth Cohort: design, prevalence and validity. PLoS One. 2012;7(7):e41280.
PubMedArticle
27.
Guinchat  V, Thorsen  P, Laurent  C, Cans  C, Bodeau  N, Cohen  D.  Pre-, peri- and neonatal risk factors for autism. Acta Obstet Gynecol Scand. 2012;91(3):287-300.
PubMedArticle
28.
Dodds  L, Fell  DB, Shea  S, Armson  BA, Allen  AC, Bryson  S.  The role of prenatal, obstetric and neonatal factors in the development of autism. J Autism Dev Disord. 2011;41(7):891-902.
PubMedArticle
29.
Lahey  BB, D’Onofrio  BM.  All in the family: comparing siblings to test causal hypotheses regarding environmental influences on behavior. Curr Dir Psychol Sci. 2010;19(5):319-323.
PubMedArticle
30.
Obel  C, Olsen  J, Henriksen  TB,  et al.  Is maternal smoking during pregnancy a risk factor for hyperkinetic disorder? findings from a sibling design. Int J Epidemiol. 2011;40(2):338-345.
PubMedArticle
31.
De Rubeis  S, He  X, Goldberg  AP,  et al; DDD Study; Homozygosity Mapping Collaborative for Autism; UK10K Consortium.  Synaptic, transcriptional and chromatin genes disrupted in autism. Nature. 2014;515(7526):209-215.
PubMedArticle
32.
Iossifov  I, O’Roak  BJ, Sanders  SJ,  et al.  The contribution of de novo coding mutations to autism spectrum disorder. Nature. 2014;515(7526):216-221.
PubMedArticle
33.
Glasson  EJ, Bower  C, Petterson  B, de Klerk  N, Chaney  G, Hallmayer  JF.  Perinatal factors and the development of autism: a population study. Arch Gen Psychiatry. 2004;61(6):618-627.
PubMedArticle
34.
Gregory  SG, Anthopolos  R, Osgood  CE, Grotegut  CA, Miranda  ML.  Association of autism with induced or augmented childbirth in North Carolina Birth Record (1990-1998) and Education Research (1997-2007) databases. JAMA Pediatr. 2013;167(10):959-966.
PubMedArticle
35.
Burstyn  I, Sithole  F, Zwaigenbaum  L.  Autism spectrum disorders, maternal characteristics and obstetric complications among singletons born in Alberta, Canada. Chronic Dis Can. 2010;30(4):125-134.
PubMed
36.
Lavender  T, Hofmeyr  GJ, Neilson  JP, Kingdon  C, Gyte  GML.  Caesarean section for non-medical reasons at term. Cochrane Database Syst Rev. 2012;3:CD004660.
PubMed
37.
Frisell  T, Öberg  S, Kuja-Halkola  R, Sjölander  A.  Sibling comparison designs: bias from non-shared confounders and measurement error. Epidemiology. 2012;23(5):713-720.
PubMedArticle
38.
Ludvigsson  JF, Reichenberg  A, Hultman  CM, Murray  JA.  A nationwide study of the association between celiac disease and the risk of autistic spectrum disorders. JAMA Psychiatry. 2013;70(11):1224-1230.
PubMedArticle
×