Few studies have examined the curtailment of reproduction (ie, stoppage) after the diagnosis of a child with autism spectrum disorder (ASD).
To examine stoppage in a large, population-based cohort of families in which a child has received a diagnosis of ASD.
Design, Setting, and Participants
Individuals with ASD born from January 1, 1990, through December 31, 2003, were identified in the California Department of Developmental Services records, which were then linked to state birth certificates to identify full sibs and half-sibs and to obtain information on birth order and demographics. A total of 19 710 case families in which the first birth occurred within the study period was identified. These families included 39 361 individuals (sibs and half-sibs). Control individuals were randomly sampled from birth certificates and matched 2:1 to cases by sex, birth year, and maternal age, self-reported race/ethnicity, and county of birth after removal of children receiving services from the California Department of Developmental Services. Using similar linkage methods as for case families, 36 215 pure control families (including 75 724 total individuals) were identified that had no individuals with an ASD diagnosis.
History of affected children.
Main Outcomes and Measures
Stoppage was investigated by comparing the reproductive behaviors of parents after the birth of a child with ASD vs an unaffected child using a survival analysis framework for time to next birth and adjusting for demographic variables.
For the first few years after the birth of a child with ASD, the parents’ reproductive behavior was similar to that of control parents. However, birth rates differed in subsequent years; overall, families whose first child had ASD had a second child at a rate of 0.668 (95% CI, 0.635-0.701) that of control families, adjusted for birth year, birth weight, maternal age, and self-reported maternal race/ethnicity. Results were similar when a later-born child was the first affected child in the family. Reproductive curtailment was slightly stronger among women who changed partners (relative rate for second-born children, 0.553 [95% CI, 0.498-0.614]).
Conclusions and Relevance
These results provide the first quantitative assessment and convincing statistical evidence of reproductive stoppage related to ASD. These findings have implications for recurrence risk estimation and genetic counseling.
Autism spectrum disorder (ASD) is one of the most common complex neurodevelopmental disorders, with a recent estimated prevalence of 11.3 per 1000 population.1 Males are much more likely to be affected than females (a ratio of approximately 4:1). Symptoms of ASD occur before 3 years of age, but diagnoses often occur at 3 years or older. Early twin studies based on very small samples estimated very high heritability,2-6 primarily owing to high monozygotic and very low dizygotic twin concordance rates. However, more recent twin studies have seen a higher dizygotic concordance rate, leading to a more moderate estimate of genetic heritability.7
Characterizing the sib recurrence risk in families has been the basis of genetic epidemiologic inference and may also have important implications for genetic counseling.8,9 A number of studies have estimated the recurrence risk for autism or ASD from epidemiologic surveys,10-12 volunteer registries,13,14 or population-based cohorts.9,15 However, few if any studies have focused on the effects of stoppage,16 the phenomenon in which parents who already have a child affected with ASD may tend to curtail their reproduction after symptoms appear and/or an affected child receives the diagnosis. Stoppage can bias recurrence risk estimates negatively if not properly addressed in the analysis. A few studies estimating recurrence risk have avoided stoppage bias by studying only sibs born after an affected child.9,10,15 However, to our knowledge, these studies have not compared these estimates with those using all sibs in the family and have not attempted to quantify stoppage. A large-scale investigation into the reproductive performance of parents with children affected by ASD has not yet been conducted, and so the effect of stoppage on recurrence estimates is not fully understood.
Here we present an analysis of stoppage based on the largest population-based sample of affected full sibships and maternal half-sibships ever assembled and compare it with a sample of control sibships. Records from the California Department of Developmental Services (DDS) database, which contains information on most of the individuals in California receiving services for an ASD diagnosis, were linked with state birth certificates.
The study was approved by the State of California Committee for the Protection of Human Subjects. Informed consent was waived.
Data on ASD were derived from California DDS records. The DDS manages a system of 21 regional centers throughout California, which are responsible for coordinating and providing assessments and services for persons with developmental disabilities (including autism and mental retardation).7,17 The ascertainment of children with a presumptive ASD diagnosis in California by the DDS is estimated to range from 85% to 90%.7 To identify nuclear families including full sibs and half-sibs, the DDS client records were linked by the staff of the California Center for Autism and Developmental Disabilities Research and Epidemiology to California birth certificate files, as described below and depicted in Figure 1.
Cases were identified by DDS eligibility for ASD or, for children eligible for services based on another condition, a DDS code indicating comorbid or suspected ASD. A recent twin study,7 derived from the same DDS registry, performed direct assessment-based diagnoses of individuals using the Autism Diagnostic Interview–Revised and Autism Diagnostic Observation Schedule and compared the diagnoses with those provided in the DDS files. The authors found high correspondence between the DDS-based diagnoses and ASD as defined using commonly accepted research criteria combining both autism instruments,18 with a sensitivity of 94.6% and a specificity of 84.6%. Index cases were defined as all individuals with a qualifying DDS diagnosis who were born in California from January 1, 1990, through December 31, 2003.
Linkage to Birth Certificates
Full sibs and half-sibs of cases were identified by linking DDS records to California birth certificates from January 1, 1990, through December 31, 2005, to identify all children who matched an index case for at least 1 parent. Cases were first matched to birth certificates by the first and last name, birth date, birthplace, mother’s and father’s names, and maternal Social Security number in later years. California birth certificate files were then searched to find other individuals whose maternal and/or paternal information matched that of the index case. Matching information varied by birth year. From 1990 through 1996, both parents’ last names and birth dates and mother’s first and maiden names were available. In 1997, both parents’ Social Security numbers became available. After 1997, both parents’ middle names and the father’s first name became available.
Criteria included an exact Social Security number match and a near-exact name match. Manual review of families with impossible relationships (eg, individual A was a full sib of individual B and B was a full sib of individual C, but A and C were half-sibs) reconciled 25 of 151; unresolved families (0.17% of total) were excluded. Most of these impossible relationships derived from families with children born before 1997, when less precise matching information was available.
Children whose information matched both parents were declared full sibs, and those whose information matched only one parent but not the other were declared half-sibs. Sometimes insufficient information was available to determine whether children were full sibs or half-sibs; manual review resolved 11 of these families and the rest were excluded. Owing to data availability as described above, more maternal than paternal half-sibs were unambiguously defined. Because birth order information was available only for mothers, only maternal half-sibs were included.
Analyses were based on complete sibship information and birth order. Families were reconstructed using maternal birth order, recorded as the number of previous live births a mother had on each birth certificate. Only families whose oldest child was born in the linkage period (ie, after 1990) were included. In addition, only families for whom all children in the birth order were identified were retained. Also, all families with multiple births (twins, triplets, etc) were excluded from analysis (their numbers were few and would need to be analyzed separately; eg, reproductive decisions of parents of twins likely differ from those of parents of singletons). After all inclusions and exclusions, a total of 19 710 case families including 39 361 individuals were identified in which the first birth occurred within the study period.
To analyze stoppage, a comparison group representing typical reproductive behavior is required. Therefore, 2 control individuals were matched to each index case based on sex, birth year, and maternal age, self-reported race/ethnicity, and residence at delivery (counties grouped by DDS Regional Center areas). Before matching, potential controls who had been served by DDS with any diagnosis were excluded. Identical procedures to those described above were used to match controls to birth certificates and identify full sibs and half-sibs. Once families were constructed, all families with at least 1 child with ASD were considered case families, including those ascertained as controls. These families technically fall under case and control family categories; as expected, we found few such families given the prevalence of ASD. Pure control families were then defined as families with no affected individuals. A total of 36 215 pure control families were identified, including 75 724 individuals.
Stoppage was investigated by comparing the reproductive behavior of parents who have had a child with ASD with that of parents who have not. Because several years typically pass before a child is suspected of having or is diagnosed as having an ASD, parental reproductive behavior is expected to be typical for the first few years after the birth of an affected child. If stoppage exists, a curtailment of further reproduction would be expected to start around 3 years after the birth of the affected child, about the time when ASD may be suspected or diagnosed. Hence, we modeled, separately for each birth order, the time to the mother’s next full-sib child under a survival framework, comparing case with pure control families. Time was right censored for individuals who had not had another birth before the end of 2005, because that was the latest year of birth certificate matching, or when another child with a different father was born (ie, maternal half-sib). A similar analysis was performed for time to the next maternal half-sib child, examining the reproductive behavior of the mothers when they switched partners. Time was right censored for individuals if they had not had another birth before the end of 2005 or if the mother had a child with the same father as before (ie, full sibs instead of half-sibs). We calculated nonparametric Kaplan-Meier survival curves.
To address the question of potential confounding due to other factors that might relate to reproductive decisions, a Cox proportional hazards analysis with separate models for each birth order and for full sibs and half-sibs was performed. The primary independent variables were based on dichotomous parameterization of the sibship history of ASD, namely, whether the first, second, or third child was affected (in the analysis of the second-, third-, and fourth-born children, respectively) and whether 2 prior children were affected (in the analysis of the third- and fourth-born children). We refer to these as family history variables. Other model covariates included birth year, birth weight, maternal age, and self-reported maternal race/ethnicity. For time-dependent covariates (birth year, birth weight, and maternal age), values were obtained from the immediately preceding child. In this way, stoppage was assessed while adjusting for demographic factors.
Finally, to assess the effect of stoppage on recurrence risk estimation, the recurrence risks for full and maternal half-sibs without accounting for stoppage were calculated using the singles method.19 Then, to account for stoppage, we calculated the recurrence risks for full sibs and maternal half-sibs born prospectively after an affected individual.
Parental demographic factors are given in Table 1. Parental ages, race/ethnicity, and educational levels were comparable between the case and control families owing to matching. Mean family size was smaller for the case families, likely owing to reproductive stoppage (Table 2). The demographic characteristics of this sample are generally representative of the California ASD population (because most of the children in California with an ASD diagnosis are included in the DDS registries) but differ in certain respects from the general California child population due to associations between some of these factors and the risk for ASD (eg, parental ages).
The correlation between all mothers’ and fathers’ ages was 0.743 (95% CI, 0.739-0.746); for cases, 0.707 (0.699-0.715); and for controls, 0.744 (0.738-0.750). The educational levels of the mothers and fathers were often concordant, so adjustment of only these maternal covariates was included in subsequent analyses.
Kaplan-Meier Estimates of Time to Next Offspring
Kaplan-Meier estimates of the probability of having another child (1 − [usual Kaplan-Meier estimates]), comparing families who previously had an affected child with families who did not, are shown in Figure 2 for full sibs. A pattern consistent with reproductive stoppage is observed. At 3 years and beyond, a difference emerges between families with no ASD history compared with families with a child affected by ASD. In Figure 2A after 3 years of age, the probability of having a second child is higher for parents with no previous child with ASD compared with those for whom the first child has ASD. In Figure 2B, the rate of having a third child is highest for parents without a previous child with ASD; the curve for parents whose second child has ASD mimics closely what was seen in Figure 2A, namely, a decrease in reproduction starting 2 to 3 years after the birth of the second (affected) child. For parents with a first child with ASD, the reduction in the probability of a third child occurs earlier, as expected, because the symptoms/diagnosis of the first child would most likely have preceded the timing decision regarding having a third child. Although the numbers start to decrease, the overall pattern of reduced probability of having a fourth child given a history of affected children is still apparent (Figure 2C). For those whose first affected child is the third, again reproduction appears typical until 2 to 3 years after the birth of the affected child.
Similar curves for mothers who switch partners are shown in Figure 3. The evidence is again consistent with stoppage. In Figure 3A, the probability of a mother having a second child with a new partner (half-sib) is higher for those with no history of a child with ASD compared with mothers with an affected child. In Figure 3B, again a clear difference exists in the probability of having a third child for mothers with a family history of ASD compared with mothers with no history.
Multivariable Cox Proportional Hazards Models
Results of the multivariable Cox proportional hazards model of time to next birth are shown in Table 3. In all analyses, a later birth year decreased the rate of additional offspring. Having a child with a lower birth weight decreased the probability of another child (significantly so for the second- and fourth-born offspring). Increasing maternal age also decreased the probability of further offspring, especially for later birth orders; the same decrease was observed for half-sibs. Several interaction terms involving the family history variables with birth year, birth weight, and maternal age were significant, although of small magnitude. Race/ethnicity and maternal educational level were also significant. All coefficients for family history variables were significant, with effects much larger than other covariates and in the direction shown by the Kaplan-Meier curves. No covariates confounded the relationship of prior affected births with the probability of additional children (stoppage) (ie, the coefficients for family history were very similar when the models were adjusted for covariates or not). Families whose first child had ASD had a second child at a rate of 0.668 (95% CI, 0.635-0.701) that of control families (after adjustment for covariates). Results were similar for later-born children in case families (Table 3). Reproductive curtailment was slightly stronger for women who switched partners; for second-born children, the relative rate was 0.553 (95% CI, 0.498-0.614) for mothers of a first affected child compared with control mothers. Results were similar within each maternal race/ethnicity group in stratified analyses.
When we ignored stoppage (ie, including children born before and after the first affected child), the recurrence risk estimate was 8.7% (1324 of 15 153) for full sibs and 3.2% (81 of 2528) for maternal half-sibs. Estimating recurrence risks prospectively by including only the full sibs or half-sibs born after the first affected child yielded a recurrence risk of 10.1% (667 of 6621) for full sibs and 4.8% (31 of 644) for maternal half-sibs and demonstrated the downward bias when not adjusting for stoppage.
Other investigators16 have previously predicted that sib recurrence estimates are influenced by reproductive stoppage, but few have examined this phenomenon directly. We have provided very compelling evidence that parents of children with ASD reduce their reproduction after the first signs or diagnosis of ASD in an affected child. For the first few years after the birth of the affected child, reproductive rates were similar to those of controls. However, after 2 years, a significant reduction in the likelihood of subsequent children in case compared with control families emerges, whether the first child affected with ASD in the family was the first, second, or third born. In all scenarios, reproduction was similar to that of controls before the birth of the first affected child and within 2 years after, but then reduced afterward. Reproductive curtailment appeared to be comparable or slightly stronger for mothers who switched partners (maternal half-sibs). The stoppage phenomenon was robust and not explained by other potential confounding factors.
Our study has several limitations. First, case diagnoses were derived from the DDS rather than structured interviews; however, a prior study of structured interviews of individuals from the DDS showed reasonably high sensitivity and specificity between the diagnoses.7 The largest potential impact would be that some children included as having ASD in our analysis may have had a non-ASD developmental problem (ie, too mildly affected to meet ASD criteria). If anything, parents of mildly affected children would be less likely to curtail reproduction than parents of children diagnosed as having ASD, leading to an underestimated stoppage effect. However, the stoppage rate is unlikely to deviate greatly from our estimate because the proportion of children misclassified is small. Second, cases and controls were not initially matched on birth order, which resulted in a slight difference in the proportion of case vs control families meeting the first-born (after 1990) inclusion criterion (case probands were more often first-born children than control probands). However, the proportional difference was modest and unlikely to have affected our results substantially because analyses were stratified on birth order and controlled for birth year and other potentially confounding covariates. Third, missingness caused by our exclusion criteria (eg, excluding multiple-birth families) was modest and similar for case and control families, thus unlikely to affect our results. Fourth, our analyses included only live births and not pregnancies. A history of miscarriages or stillbirths may have affected rates and timing of live births and also influenced parental reproductive decisions; however, this history alone is unlikely to explain our observations because no difference was observed in reproduction for parents of a child with ASD compared with parents without in the first 2 to 3 years after the birth of that child. Fifth, our assessment of stoppage for mothers who switched partners may be confounded by additional factors, including time (eg, to develop another relationship) and thus age and additional persons’ views on childbearing or possibilities of recurrence. In our analyses, we found that women with an affected child appeared to be even less likely to have another child with a different partner than did women who did not switch partners or who had no history of affected children.
Stoppage likely represents a number of parental concerns. For example, parents of an affected child may believe that subsequent children may be at a higher risk of developing the same condition. Having an affected child may also create an unanticipated parental burden or stress, leading to the decision to reduce further procreation, although evidence suggests that it does not lead to an increased rate of divorce or separation.20
These results are, to our knowledge, the first to quantify reproductive stoppage in families affected by ASD by using a large, population-based sample of California families. Overall, we observed a reproduction reduction of approximately one-third; this reduction was not influenced by potential confounders and was consistent across various demographic subgroups (eg, racial/ethnic groups). We also demonstrated the stoppage effect on the sib recurrence risk estimates. Accurate recurrence risk estimates comparing full sibs and twins are critical for proper etiologic inference in genetic epidemiologic analysis; we will elaborate further on these in a subsequent publication.
Submitted for Publication: October 1, 2013; final revision received January 17, 2014; accepted March 3, 2014.
Corresponding Author: Neil Risch, PhD, Institute for Human Genetics, University of California, San Francisco, 513 Parnassus Ave, San Francisco, CA 94143-0794 (email@example.com).
Published Online: June 18, 2014. doi:10.1001/jamapsychiatry.2014.420.
Author Contributions: Drs Hoffmann and Risch had full access to all 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: Hoffmann, Windham, Croen, Grether, Risch.
Acquisition, analysis, or interpretation of data: Hoffmann, Windham, Anderson, Risch.
Drafting of the manuscript: Hoffmann, Windham, Grether, Risch.
Critical revision of the manuscript for important intellectual content: Hoffmann, Windham, Anderson, Croen, Grether.
Statistical analysis: Hoffmann, Anderson, Risch.
Obtained funding: Risch.
Study supervision: Windham, Grether, Risch.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by funds from the Institute for Human Genetics, University of California, San Francisco, and by grant R25 CA112355 from the National Cancer Institute.
Role of the 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 decision to submit the manuscript for publication.
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
A, Le Couteur
et al. Autism as a strongly genetic disorder: evidence from a British twin study. Psychol Med
. 1995;25(1):63-77.PubMedGoogle ScholarCrossref
M. Infantile autism: a genetic study of 21 twin pairs. J Child Psychol Psychiatry
. 1977;18(4):297-321.PubMedGoogle ScholarCrossref
CM. The genetics of autistic disorders and its clinical relevance: a review of the literature. Mol Psychiatry
. 2007;12(1):2-22.PubMedGoogle ScholarCrossref
ML. Evidence for autosomal recessive inheritance in 46 families with multiple incidences of autism. Am J Psychiatry
. 1985;142(2):187-192.PubMedGoogle Scholar
et al. A twin study of autism in Denmark, Finland, Iceland, Norway and Sweden. J Child Psychol Psychiatry
. 1989;30(3):405-416.PubMedGoogle ScholarCrossref
et al. Genetic heritability and shared environmental factors among twin pairs with autism. Arch Gen Psychiatry
. 2011;68(11):1095-1102.PubMedGoogle ScholarCrossref
E. Genetic counseling in autism and pervasive developmental disorders. J Autism Dev Disord
. 1998;28(5):447-456.PubMedGoogle ScholarCrossref
et al. Recurrence risk for autism spectrum disorders: a Baby Siblings Research Consortium study. Pediatrics
. 2011;128(3):e488-e495. doi:10.1542/peds.2010-2825.PubMedGoogle Scholar
PB. Effects of familial risk factors and place of birth on the risk of autism: a nationwide register-based study. J Child Psychol Psychiatry
. 2005;46(9):963-971.PubMedGoogle ScholarCrossref
et al. The UCLA–University of Utah Epidemiologic Survey of Autism: prevalence. Am J Psychiatry
. 1989;146(2):194-199.PubMedGoogle Scholar
M. Sibling risk of pervasive developmental disorder estimated by means of an epidemiologic survey in Nagoya, Japan. J Hum Genet
. 2006;51(6):518-522.PubMedGoogle ScholarCrossref
P. Sibling recurrence and the genetic epidemiology of autism. Am J Psychiatry
. 2010;167(11):1349-1356.PubMedGoogle ScholarCrossref
et al. Autism recurrence in half siblings: strong support for genetic mechanisms of transmission in ASD. Mol Psychiatry
. 2013;18(2):137-138.PubMedGoogle ScholarCrossref
ET. Recurrence of autism spectrum disorders in full- and half-siblings and trends over time: a population-based cohort study. JAMA Pediatr
. 2013;167(10):947-953. PubMedGoogle ScholarCrossref
GC. Risk of autism and increasing maternal and paternal age in a large North American population. Am J Epidemiol
. 2009;170(9):1118-1126.PubMedGoogle ScholarCrossref
et al. Combining information from multiple sources in the diagnosis of autism spectrum disorders. J Am Acad Child Adolesc Psychiatry
. 2006;45(9):1094-1103. PubMedGoogle ScholarCrossref
EA. Relationship status among parents of children with autism spectrum disorders: a population-based study. J Autism Dev Disord
. 2012;42(4):539-548.PubMedGoogle ScholarCrossref