[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 18.206.168.65. Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Figure.
Relationship Between Labor Induction and Autism Spectrum Disorders
Relationship Between Labor Induction and Autism Spectrum Disorders

Cumulative risk of being diagnosed with autism spectrum disorders as a function of age. The shaded areas represent the pointwise 95% CI on the failure proportion.

Table 1.  
Maternal and Pregnancy Characteristics During the Study Period (1992-2005)
Maternal and Pregnancy Characteristics During the Study Period (1992-2005)
Table 2.  
Association Between Labor Induction and Offspring Autism Spectrum Disordera
Association Between Labor Induction and Offspring Autism Spectrum Disordera
Table 3.  
Sensitivity Analysis of Association Between Labor Induction and Offspring Autism Spectrum Disordera
Sensitivity Analysis of Association Between Labor Induction and Offspring Autism Spectrum Disordera
1.
Autism and Developmental Disabilities Monitoring Network Surveillance Year 2008 Principal Investigators; Centers for Disease Control and Prevention.  Prevalence of autism spectrum disorders—Autism and Developmental Disabilities Monitoring Network, 14 sites, United States, 2008.  MMWR Surveill Summ. 2012;61(3):1-19.PubMedGoogle Scholar
2.
Gardener  H, Spiegelman  D, Buka  SL.  Perinatal and neonatal risk factors for autism: a comprehensive meta-analysis.  Pediatrics. 2011;128(2):344-355.PubMedGoogle ScholarCrossref
3.
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.PubMedGoogle ScholarCrossref
4.
Juul-Dam  N, Townsend  J, Courchesne  E.  Prenatal, perinatal, and neonatal factors in autism, pervasive developmental disorder-not otherwise specified, and the general population.  Pediatrics. 2001;107(4):E63.PubMedGoogle ScholarCrossref
5.
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.PubMedGoogle ScholarCrossref
6.
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.PubMedGoogle ScholarCrossref
7.
Stavropoulos  KK, Carver  LJ.  Research review: social motivation and oxytocin in autism—implications for joint attention development and intervention.  J Child Psychol Psychiatry. 2013;54(6):603-618.PubMedGoogle ScholarCrossref
8.
Wahl  RU.  Could oxytocin administration during labor contribute to autism and related behavioral disorders? a look at the literature.  Med Hypotheses. 2004;63(3):456-460.PubMedGoogle ScholarCrossref
9.
Vintzileos  AM, Ananth  CV.  Induction or augmentation of labor and autism.  JAMA Pediatr. 2014;168(2):190.PubMedGoogle ScholarCrossref
10.
Stitely  ML.  Induction or augmentation of labor and autism.  JAMA Pediatr. 2014;168(2):189-190.PubMedGoogle ScholarCrossref
11.
Saade  GR, Sibai  BM, Silver  R.  Induction or augmentation of labor and autism.  JAMA Pediatr. 2014;168(2):190-191.PubMedGoogle ScholarCrossref
12.
Barnabei  VM, Rasmusson  RL, Bett  GC.  Autism and induced labor: is calcium a potential mechanistic link?  Am J Obstet Gynecol. 2014;210(5):494-495.PubMedGoogle ScholarCrossref
13.
D’Onofrio  BM, Lahey  BB, Turkheimer  E, Lichtenstein  P.  Critical need for family-based, quasi-experimental designs in integrating genetic and social science research.  Am J Public Health. 2013;103(suppl 1):S46-S55.PubMedGoogle ScholarCrossref
14.
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.PubMedGoogle ScholarCrossref
15.
Statistics Sweden.  Multi-Generation Register 2009—A Description of Contents and Quality. Örebro, Sweden: Statistics Sweden; 2010:13.
16.
National Board of Health and Welfare.  The Swedish Medical Birth Register—A Summary of Content and Quality. Stockholm, Sweden: National Board of Health and Welfare; 2003.
17.
WHO.  International Classification of Diseases. Geneva, Switzerland: World Health Organization WHO; 1992.
18.
National Board of Health and Welfare.  Quality and Content in the Patient Register—Inpatient Care Discharges 1964-2007 and Outpatient (Specialized Care) Visits 1997-2007 [in Swedish]. Stockholm, Sweden: National Board of Health and Welfare; 2009:125-15. Report: 2009-125-15.
19.
Cnattingius  S, Ericson  A, Gunnarskog  J, Källén  B.  A quality study of a medical birth registry.  Scand J Soc Med. 1990;18(2):143-148.PubMedGoogle Scholar
20.
Frost  F, Starzyk  P, George  S, McLaughlin  JF.  Birth complication reporting: the effect of birth certificate design.  Am J Public Health. 1984;74(5):505-506.PubMedGoogle ScholarCrossref
21.
Hultman  CM, Sandin  S, Levine  SZ, Lichtenstein  P, Reichenberg  A.  Advancing paternal age and risk of autism: new evidence from a population-based study and a meta-analysis of epidemiological studies.  Mol Psychiatry. 2011;16(12):1203-1212.PubMedGoogle ScholarCrossref
22.
Sullivan  PF, Magnusson  C, Reichenberg  A,  et al.  Family history of schizophrenia and bipolar disorder as risk factors for autismfamily history of psychosis as risk factor for ASD.  Arch Gen Psychiatry. 2012;69(11):1099-1103. PubMedGoogle Scholar
23.
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.PubMedGoogle ScholarCrossref
24.
Academy of Medical Sciences Working Group.  Identifying the Environmental Causes of Disease: How Should We Decide What to Believe and When to Take Action? London, England: Academy of Medical Sciences; 2007.
25.
D’Onofrio  BM, Class  QA, Lahey  BB, Larsson  H.  Testing the developmental origins of health and disease hypothesis for psychopathology using family-based quasi-experimental designs.  Child Dev Perspect. 2014;8(3):151-157.PubMedGoogle ScholarCrossref
26.
Donovan  SJ, Susser  E.  Advent of sibling designs.  Int J Epidemiol. 2011;40(2):345-349.PubMedGoogle ScholarCrossref
Original Investigation
September 6, 2016

Association of Labor Induction With Offspring Risk of Autism Spectrum Disorders

Author Affiliations
  • 1Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts
  • 2Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
  • 3Department of Psychological and Brain Sciences, Indiana University, Bloomington
  • 4Department of Obstetrics and Gynecology, Massachusetts General Hospital, Harvard Medical School, Boston
  • 5Astrid Lindgren Children’s Hospital, Karolinska University Hospital, Stockholm, Sweden
  • 6Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts
  • 7Department of Anesthesia, Critical Care, and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts
 

Copyright 2016 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.

JAMA Pediatr. 2016;170(9):e160965. doi:10.1001/jamapediatrics.2016.0965
Abstract

Importance  Induction of labor is a frequently performed obstetrical intervention. It would thus be of great concern if reported associations between labor induction and offspring risk of autism spectrum disorders (ASD) reflected causal influence.

Objective  To assess the associations of labor induction with ASD, comparing differentially exposed relatives (siblings and cousins discordant for induction).

Design, Setting, and Participants  Follow-up of all live births in Sweden between 1992 and 2005, defined in the Medical Birth Register. The register was linked to population registers of familial relations, inpatient and outpatient visits, and education records. Diagnoses of ASD were from 2001 through 2013, and data were analyzed in the 2015-2016 year.

Exposures  Induction of labor.

Main Outcomes and Measures  Autism spectrum disorders identified by diagnoses from inpatient and outpatient records between 2001 and 2013. Hazard ratios (HRs) quantified the association between labor induction and offspring ASD. In addition to considering a wide range of measured confounders, comparison of exposure-discordant births to the same woman allowed additional control for all unmeasured factors shared by siblings.

Results  The full cohort included 1 362 950 births, of which 22 077 offspring (1.6%) were diagnosed with ASD by ages 8 years through 21 years. In conventional models of the full cohort, associations between labor induction and offspring ASD were attenuated but remained statistically significant after adjustment for measured potential confounders (HR, 1.19; 95% CI, 1.13-1.24). When comparison was made within siblings whose births were discordant with respect to induction, thus accounting for all environmental and genetic factors shared by siblings, labor induction was no longer associated with offspring ASD (HR, 0.99; 95% CI, 0.88-1.10).

Conclusions and Relevance  In this nationwide sample of live births we observed no association between induction of labor and offspring ASD within sibling comparison. Our findings suggest that concern for ASD should not factor into the clinical decision about whether to induce labor.

Introduction

Autism spectrum disorders (ASD) are a group of permanent developmental disabilities characterized by impairments in social interaction, language development, along with stereotyped and repetitive behaviors; they are estimated to affect approximately 1 in 90 children in the United States.1 Both genetic and environmental factors in early life are thought to be of importance in neurodevelopment.2 Recently, a large, population-based study suggested an independent association between induction and augmentation of labor and risk of offspring ASD.3 The study linked information on 625 042 live births from the North Carolina Detailed Birth Record and the Education Research databases (which included information on approximately 5500 children with a documented exceptionality designation for ASD). After control for confounding variables, the odds ratio for development of ASD following exposure to labor induction and augmentation was 1.27 (95% CI, 1.05-1.52).

Although not the first study to link induction of labor to offspring neurodevelopment,4-6 this population-based study gained widespread media attention and sparked a vigorous debate among clinicians and researchers. Causal speculation has largely focused on the potential role of oxytocin, which is administered to stimulate uterine contractions (in all augmentations and a majority of inductions). Oxytocin is a neurotransmitter involved in social function and cognition,7 and it has been hypothesized that predelivery exposure could predispose to ASD due to a downregulation of oxytocin receptors.8 Alternatively, it may not be the method but the intervention per se that increases risk of offspring ASD, by setting off downstream complications with a negative influence on neurodevelopment (eg, fetal distress and hypoxia, uterine rupture, emergency cesarean delivery, etc). Importantly, many have also argued for noncausal explanations to the findings,9-11 including the possibility that dysfunctional labor and ASD neuropathology share genetic origin (calcium homeostasis).12

Understanding whether induction of labor truly confers increased risk of neuropsychiatric disorders in the offspring is important to clinicians and patients in weighing the risks and benefits of this therapeutic intervention. Herein, we study the association of labor induction with ASD in a nationwide register-based birth cohort, with ability to identify differentially exposed relatives (siblings and cousins who are discordant with respect to whether their births occurred after induction of labor) and to follow up on their diagnosis of ASD. The richness of information available through linkage of Swedish population registers allows for thorough individual confounder adjustment while the identification of differentially exposed relatives (eg, siblings born to same mother, one who was induced for one pregnancy but not the other) allows additional adjustment for unmeasured factors shared in families.13

Box Section Ref ID

Key Points

  • Question Is there a causal link between induction of labor and the risk of autism spectrum disorders in the offspring?

  • Findings This cohort study included all live births in Sweden between 1992 and 2005. When examining siblings whose births were discordant with respect to induction, thus accounting for environmental and genetic factors shared within families, there was no association between induction of labor and offspring risk of autism spectrum disorders.

  • Meaning These results suggest that the previously reported link between induction and autism is unlikely to be causal and should not factor into the clinical decision about whether to induce labor.

Methods
Data Source and Study Cohort

All residents in Sweden are assigned a unique civic registration number, through which individuals can be tracked and linked in national population registries of health and demographics.14 The Multi-Generation Register15 links all Swedish residents to their parents, adoptive or biological, thereby allowing for the identification of full and half siblings, as well as more complex family structures. The Medical Birth Register16 contains medical records from antenatal visits and the delivery and pediatric examination of the newborn for 96% to 99% of all live births in Sweden since 1973. Women are routinely enrolled in antenatal care during weeks 8 through 12, and the midwife uses a standardized form to record information such as the mother’s weight and height, sociodemographics (eg, age and cohabitation status), reproductive history, use of tobacco smoke, and current and previous illnesses. Standardized charts are also used at the time of delivery to record information including length of gestation, fetal presentation, onset, and mode of delivery. After the delivery, all relevant diagnoses and procedures until discharge are recorded using the World Health Organization’s International Classification of Diseases (ICD).17 This study also made use of the National Patient Register,18 which includes data from all psychiatric inpatient visits since 1973, all hospital admissions since 1987, and all specialized outpatient care since 2001. Finally, the Education Register at Statistics Sweden allowed assessment of the mothers’ highest attained level of education. The institutional review boards at the Karolinska Institutet and Indiana University approved the analysis of deidentified data and waived informed consent.

Labor induction was introduced as a yes-no indicator on the standardized delivery charts during 1991. Since 1997, it has been further possible to document through procedure coding on the discharge record. Check-box indicators on charts are generally considered to have high reliability because they are less likely missed or prone to error than manual recording of codes.19,20 Cases of ASD were identified based on the recording of ICD-10 codes (F84: Pervasive Developmental Disorders) in National Patient Register records from 2001. These diagnoses are set by specialized (ie, not general practice) physicians. Case reviews have shown high agreement with Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) criteria.21,22 Further face validity may be drawn from finding prevalence estimates of ASD based on this type of identification consistent with a large-scale detailed assessment from the same period.23

To allow exposure (from 1991) and outcome (up to 2013) identification, the study base consisted of the birth cohorts between 1992 and 2005. All live births in the period were followed up with respect to neuropsychiatric diagnosis (event), emigration from Sweden, or death (right censoring) through the end of 2013 (ages 8-21 years). Missing information predominantly occurred for maternal early pregnancy body mass index (BMI, calculated as weight in kilograms divided by height in meters squared, 16%), and smoking status (6%). Deletion of individuals with missing information on any of the final covariates of interest (mother’s country of origin, education, BMI, and smoking) yielded a sample of 1 117 220 (82%) available for complete case analysis. Among these, we further identified 694 612 siblings (bound by mothers) and 323 436 cousins (bound by sisters).

Statistical Analysis

First, we assessed the association between labor induction and ASD graphically, plotting for induced and not induced, respectively, the cumulative risk of ASD with age using Kaplan-Meier estimation. The influence of induction was then modeled in Cox proportional hazard regression, with age as the underlying time scale and allowing censoring due to emigration or death. To account for clustering arising from the inclusion of more than one offspring per woman (siblings), a robust standard error estimation was used. With follow-up beginning in 2001, some in the earlier-born cohorts may be subject to left censoring (with incident diagnosis occurring before 2001). Because the majority of (and all severe) affected patients are expected to be frequent consumers of psychiatric and medical care, they are very likely identified as cases during follow-up, just not capturing the true incident age at diagnosis. To account for this, and any concern for birth cohort effects (due to a concomitant rising prevalence of induction and diagnoses of neuropsychiatric disorders with time), all analyses were adjusted for birth year. All analyses were performed using SAS statistical software version 9.4 (SAS Institute Inc).

To evaluate the influence of covariates on the association between induction and offspring ASD, we performed complete case analysis following an a priori defined modeling strategy to sequentially increase the degree of confounder adjustment. The baseline model including birth year, parity, and maternal age was expanded with measured stable maternal covariates (not likely to vary between consecutive births) such as educational attainment and country of origin. Covariates specific to each birth were further added, including smoking status and BMI in early pregnancy, multiple gestation, gestational diabetes, gestational hypertension, preeclampsia, chorioamnionitis, urogenital infection, premature rupture of membranes, and prolonged or high-risk pregnancy. These covariates were selected based on known or plausible association with both labor induction and offspring ASD. After fitting each of the population-based models, we tested the proportional hazards assumption explicitly by evaluating the scaled Schoenfeld residuals for nonzero slope and found no evidence that the induction parameter violated the assumption in any of these models. In a final model, we then included a fixed effect to allow the underlying hazard to vary between mothers, making the contrast within siblings only, while maintaining adjustment for individual-level covariates (unique to each birth).

To explore the influence of potential bias, we performed a series of sensitivity analyses. First, we repeated all analyses restricted to the cohort born from 1999 through 2005, in whom left censoring was less likely. Because the selection to the sibling comparison could affect generalizability (representativeness of the population), we assessed whether the cohort estimates in each of the first 3 models were different when the sample was restricted to individuals who had at least 1 sibling in the cohort. We also compared occurrence of ASD in maternal first cousins (offspring of sisters) differentially exposed to labor induction, avoiding the requirement of at least 2 births to the same woman. The comparison accounts for all factors shared by children of sisters, including some genetic and maternal environmental factors. The use of cousins in the fixed-effects contrast further allowed us to assess all 4 models in first-born individuals only, to completely exclude any confounding influence of birth order. We also performed all analyses after excluding the 6% delivered through elective cesarean delivery, which restricts the contrast to reflect those with and without indication to induce (comparison being spontaneous start of labor). Last, because a majority of missingness was due to mother’s BMI, we performed a complete case analysis without consideration of BMI (including 94% of the study base).

Finally, it is important to note that the comparison of relatives would rely solely on the pairs (of siblings or cousins) that are discordant with respect to exposure status. In our sample, 15.2% of all maternal sibling pairs and 18.2% of all maternal cousin pairs were discordant for labor induction. With an overall induction prevalence of 12% in this sample, a random match of unrelated individuals should produce on average 21% discordance for this obstetric intervention. The lower discordance seen among relatives could be due to familial factors that make relatives more similar (concordant). In siblings, this could also be counteracted by a potential influence of birth order (making siblings different).

Results

Of the 1 362 950 individuals included in the cohort, 22 077 were diagnosed with ASD during follow-up. Overall, 11% of all live births in Sweden between 1992 and 2005 were preceded by labor induction (with a slight increase over time). Table 1 shows the maternal and pregnancy characteristics, stratified by induction status. Comparing distributions, the induced deliveries were more likely to occur in later years of the cohort and to women who were primiparous, of older age, and of higher BMI than in the general population. Mother’s educational status, country of origin, and smoking in early pregnancy did not differ substantially across exposure groups (induced and not induced). Labor induction however occurred more commonly in association with a number of pregnancy complications, including gestational diabetes, gestational hypertension, and preeclampsia. Twenty-three percent of all induced pregnancies were postterm (≥42 weeks of gestation), 15% had preeclampsia, and 7% had intrauterine growth restriction (Table 1).

The occurrence of ASD in the complete case sample and its relation to labor induction is shown graphically in the Figure, in which the cumulative risks of ASD are plotted as a function of age, stratified by exposure to induction. An exponential increase in the cumulative risk of ASD reflects the increased rate (slope) of discovery/diagnosis with age. By the age of 20 years, a little more than 2.5% of the study population had been diagnosed with ASD (3.5% among the induced and 2.5% among the noninduced).

The main analysis exploring the association between labor induction and ASD is presented in Table 2. In the baseline model, labor induction was statistically significantly associated with ASD in the full cohort (hazard ratio [HR], 1.32; 95% CI, 1.27-1.38). Adjustment for stable maternal characteristics including maternal educational level and country of origin (model 2) did not substantially change the risk estimate (HR, 1.31; 95% CI, 1.26-1.37). After adjustment for all measured factors including stable maternal characteristics and birth-specific characteristics (model 3), the association was still statistically significant, albeit somewhat attenuated (HR, 1.19; 95% CI, 1.13-1.24). However, when further adjustment was made using fixed-effects models—comparing discordant siblings to each other to account for all factors they share (model 4)—labor induction was no longer associated with ASD (HR, 0.99; 95% CI, 0.88-1.10).

A series of sensitivity analyses were performed to test the robustness of the findings from the main analysis (Table 3). Refitting the models in samples restricted to the later-born cohorts or to individuals with 1 or more siblings respectively produced very similar estimates as in the full population. Comparison of exposure discordant maternal cousins showed attenuation from the fully adjusted cohort model (model 3) although the point estimate did not completely reach null as in the sibling analysis. Restriction to first-born individuals showed slight attenuation of all estimates, but with an intact pattern of statistically significant positive associations in the cohort further attenuated within cousins. The exclusion of elective Cesarean deliveries, while leading to slightly stronger cohort associations, still showed complete attenuation (no association) in the sibling comparison. Lastly, and reassuringly, the complete case analysis excluding only 6% (reintroducing those only missing BMI, and not adjusting for this covariate) was nearly identical to the main complete case analysis (which excluded 18% of the cohort for missing data; Table 3).

Discussion

In this large, population-based study from Sweden, using a family comparison design, we observed no relationship between induction of labor and offspring ASD. Our findings suggest that concern about ASD after induced labor should not factor into the clinical decision about whether to induce labor. The results also provide reassurance to parturients, that undergoing this common obstetrical intervention will not increase their child’s risk of developing this condition.

Consistent with recent prior studies,3,6 we observed a significant crude association between induction of labor and the risk of ASD that persisted also after adjustment for measured maternal factors and pregnancy conditions that were prespecified as potential confounders. However, when we applied a fixed-effects model to compare induction-discordant siblings to each other (ie, siblings born to same mother—in one, the labor was induced; in the other, it was not), this association was no longer present. The method allowed further control for all shared maternal factors (present across all pregnancies) that are unmeasured in registries but appear to confound the association between labor induction and neurodevelopmental disorders in the offspring. These unmeasured characteristics are likely to have also been present in prior studies that observed an association between induction and ASD. Through the use of this rich data and innovative, family-based design, we were able to account for factors that are not possible to capture using traditional approaches.

Exactly what constitutes the unmeasured factors that lead to residual confounding in traditional approaches cannot be directly deduced from our data. The source would have to be a common cause of the exposure (labor induction) and outcome (ASD), and further a factor that is present across all pregnancies to the same woman. This points to genetic or environmental factors that are shared by siblings and confer risk of both labor induction and adverse neurodevelopmental outcome. A previous commentary has for example pointed to genes involved in cellular calcium homeostasis, which may play a role in the initiation and progression of labor, as well as in neurodevelopment.12 A shared environmental factor could, speculatively, involve the characteristics of the health care setting where women and their offspring are treated; if cared for in a higher-intensity medical system it is possible that a woman would be more likely to be induced and her child more likely to be diagnosed with a neurodevelopmental disorder than a woman and child treated in a lower-intensity system. Because Sweden has a decentralized government-funded health care system with universal access, the potential for such differences might, if anything, arise from local variation in health care practice.

Our study has a number of strengths. The source population for the analysis encompasses nearly all births that occur in Sweden, ensuring that the study is free from selection bias. The large size also allows for precise estimates of the association between induction of labor and the neurodevelopmental outcomes. The study benefits from the multiple database linkages, including information from the Medical Birth Register and the Multi-Generation Register that allow for the sibling- and cousin-based designs, and the National Patient Register, which allows for the long-term follow-up of offspring for the development of neurodevelopmental problems. It uses an innovative analytic approach that does not rely solely on measured covariates to account for common causes.24-26 The exposure was identified based on a combination of both codes and check boxes in the delivery records, ensuring that it is captured with both sensitivity and specificity. Likewise, the approach to identify the outcome of ASD has been shown to correlate well with Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) criteria.

The study is also subject to certain limitations inherent in its data and design. Similar to the earlier large population-based study reporting an association between induction and ASD3 our exposure information did not include specification of the type of method used. Contrary to this study, we did not have information on labor augmentation. From this follows that our findings pertain to the risks associated with induction per se and not the method or medication used, hence not specifically testing the proposed biological pathway through oxytocin exposure. Although not specifically coded in the delivery charts, given contemporary obstetric practice, it is likely that a majority of women whose deliveries were induced were exposed to oxytocin (for induction, augmentation, or both), but we also note that a proportion of the women with spontaneous start of labor in the comparison group also would have been exposed to oxytocin through augmentation. A more important potential limitation is that the within-family comparison relied on the selection of discordant family members. Although the analysis of discordant relatives allowed for the control for all the factors they share, it also meant that the discordance had to be caused by something other than the shared factors. If for example, this is due to the influence of unmeasured birth-specific confounders or misclassification of the exposure, it will bias the within-relative comparison. However, because induction is a common obstetric intervention and its recording is facilitated by a check-box indicator on the delivery chart, it is likely captured with high fidelity. Concern about confounding from unmeasured factors should further be ameliorated by the combined adjustment for shared factors and by an extensive list of measured birth-specific (individual) confounders (eg, preeclampsia, chorioamnionitis, urogenital infection, premature rupture of membranes, prolonged and high-risk pregnancy) achieved complete attenuation of the within-sibling comparison.

Conclusions

Using a design that incorporates the comparison of exposure-discordant relatives, the findings of this study provide no support for a causal association between induction of labor and offspring development of ASD.

Back to top
Article Information

Corresponding Author: Anna Sara Oberg, PhD, Department of Epidemiology, Harvard T. H. Chan School of Public Health, 677 Huntington Ave, Boston, MA 02115 (oberg@hsph.harvard.edu).

Accepted for Publication: April 3, 2016.

Published Online: July 25, 2016. doi:10.1001/jamapediatrics.2016.0965.

Author Contributions: Drs Oberg and Rickert had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Oberg, Bateman.

Acquisition, analysis, or interpretation of data: Oberg, Ecker, D’Onofrio, Rickert, Hernandez-Diaz, Almqvist, Larsson, Lichtenstein, Bateman.

Drafting of the manuscript: Oberg, Bateman.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Oberg, Rickert.

Obtained funding: Oberg, D’Onofrio, Rickert, Hernandez-Diaz, Ecker, Almqvist, Larsson, Lichtenstein, Bateman.

Administrative, technical, or material support: D’Onofrio, Rickert, Almqvist, Larsson, Lichtenstein.

Study supervision: D’Onofrio, Hernandez-Diaz.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grants 2012-34 (International Postdoctoral grant) and 340-2013-5867 (Swedish Initiative for Research on Microdata in the Social and Medical Sciences [SIMSAM]) from the Swedish Research Council and grants K08HD075831 and R01HD061817 from the National Institutes of Health Eunice Kennedy Shriver National Institute of Child Health & Human Development.

Role of the Funder/Sponsor: The funders 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 decision to submit the manuscript for publication.

References
1.
Autism and Developmental Disabilities Monitoring Network Surveillance Year 2008 Principal Investigators; Centers for Disease Control and Prevention.  Prevalence of autism spectrum disorders—Autism and Developmental Disabilities Monitoring Network, 14 sites, United States, 2008.  MMWR Surveill Summ. 2012;61(3):1-19.PubMedGoogle Scholar
2.
Gardener  H, Spiegelman  D, Buka  SL.  Perinatal and neonatal risk factors for autism: a comprehensive meta-analysis.  Pediatrics. 2011;128(2):344-355.PubMedGoogle ScholarCrossref
3.
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.PubMedGoogle ScholarCrossref
4.
Juul-Dam  N, Townsend  J, Courchesne  E.  Prenatal, perinatal, and neonatal factors in autism, pervasive developmental disorder-not otherwise specified, and the general population.  Pediatrics. 2001;107(4):E63.PubMedGoogle ScholarCrossref
5.
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.PubMedGoogle ScholarCrossref
6.
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.PubMedGoogle ScholarCrossref
7.
Stavropoulos  KK, Carver  LJ.  Research review: social motivation and oxytocin in autism—implications for joint attention development and intervention.  J Child Psychol Psychiatry. 2013;54(6):603-618.PubMedGoogle ScholarCrossref
8.
Wahl  RU.  Could oxytocin administration during labor contribute to autism and related behavioral disorders? a look at the literature.  Med Hypotheses. 2004;63(3):456-460.PubMedGoogle ScholarCrossref
9.
Vintzileos  AM, Ananth  CV.  Induction or augmentation of labor and autism.  JAMA Pediatr. 2014;168(2):190.PubMedGoogle ScholarCrossref
10.
Stitely  ML.  Induction or augmentation of labor and autism.  JAMA Pediatr. 2014;168(2):189-190.PubMedGoogle ScholarCrossref
11.
Saade  GR, Sibai  BM, Silver  R.  Induction or augmentation of labor and autism.  JAMA Pediatr. 2014;168(2):190-191.PubMedGoogle ScholarCrossref
12.
Barnabei  VM, Rasmusson  RL, Bett  GC.  Autism and induced labor: is calcium a potential mechanistic link?  Am J Obstet Gynecol. 2014;210(5):494-495.PubMedGoogle ScholarCrossref
13.
D’Onofrio  BM, Lahey  BB, Turkheimer  E, Lichtenstein  P.  Critical need for family-based, quasi-experimental designs in integrating genetic and social science research.  Am J Public Health. 2013;103(suppl 1):S46-S55.PubMedGoogle ScholarCrossref
14.
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.PubMedGoogle ScholarCrossref
15.
Statistics Sweden.  Multi-Generation Register 2009—A Description of Contents and Quality. Örebro, Sweden: Statistics Sweden; 2010:13.
16.
National Board of Health and Welfare.  The Swedish Medical Birth Register—A Summary of Content and Quality. Stockholm, Sweden: National Board of Health and Welfare; 2003.
17.
WHO.  International Classification of Diseases. Geneva, Switzerland: World Health Organization WHO; 1992.
18.
National Board of Health and Welfare.  Quality and Content in the Patient Register—Inpatient Care Discharges 1964-2007 and Outpatient (Specialized Care) Visits 1997-2007 [in Swedish]. Stockholm, Sweden: National Board of Health and Welfare; 2009:125-15. Report: 2009-125-15.
19.
Cnattingius  S, Ericson  A, Gunnarskog  J, Källén  B.  A quality study of a medical birth registry.  Scand J Soc Med. 1990;18(2):143-148.PubMedGoogle Scholar
20.
Frost  F, Starzyk  P, George  S, McLaughlin  JF.  Birth complication reporting: the effect of birth certificate design.  Am J Public Health. 1984;74(5):505-506.PubMedGoogle ScholarCrossref
21.
Hultman  CM, Sandin  S, Levine  SZ, Lichtenstein  P, Reichenberg  A.  Advancing paternal age and risk of autism: new evidence from a population-based study and a meta-analysis of epidemiological studies.  Mol Psychiatry. 2011;16(12):1203-1212.PubMedGoogle ScholarCrossref
22.
Sullivan  PF, Magnusson  C, Reichenberg  A,  et al.  Family history of schizophrenia and bipolar disorder as risk factors for autismfamily history of psychosis as risk factor for ASD.  Arch Gen Psychiatry. 2012;69(11):1099-1103. PubMedGoogle Scholar
23.
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.PubMedGoogle ScholarCrossref
24.
Academy of Medical Sciences Working Group.  Identifying the Environmental Causes of Disease: How Should We Decide What to Believe and When to Take Action? London, England: Academy of Medical Sciences; 2007.
25.
D’Onofrio  BM, Class  QA, Lahey  BB, Larsson  H.  Testing the developmental origins of health and disease hypothesis for psychopathology using family-based quasi-experimental designs.  Child Dev Perspect. 2014;8(3):151-157.PubMedGoogle ScholarCrossref
26.
Donovan  SJ, Susser  E.  Advent of sibling designs.  Int J Epidemiol. 2011;40(2):345-349.PubMedGoogle ScholarCrossref
×