ASD indicates autism spectrum disorder. Each sibling is followed from birth to end of follow-up (death, emigration, or end of follow-up) or ASD diagnosis. Analysis methods A and B agree for the clear majority of all sibling pairs. If S2 was not observed with ASD, the concordance or discordance status would be the same for both methods. In the Figure, for the family with 2 siblings where S1 is diagnosed with ASD in 1993 and S2 in 1998, the sibling pair (S1, S2) will be discordant in 1993 because S2 is censored at the ASD diagnosis of S1. However, the pair (S2, S1) will be concordant in 1998. The 2 pairs are 2 candidates representing the family. For calculating heritability, 1 of these representative sibling pairs was randomly selected. As a consequence, the algorithm led to a loss of about half of the concordant pairs compared with results under the assumptions and methods applied in the alternate method (analysis method B),3 in which calculating heritability typically does not consider sibling pairs as both discordant or concordant depending on which sibling is considered dependent, but instead follows them as a pair.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Sandin S, Lichtenstein P, Kuja-Halkola R, Hultman C, Larsson H, Reichenberg A. The Heritability of Autism Spectrum Disorder. JAMA. 2017;318(12):1182–1184. doi:10.1001/jama.2017.12141
Studies have found that autism spectrum disorder (ASD) aggregates in families, and twin studies estimate the proportion of the phenotype variance due to genetic factors (heritability) to be about 90%.1
In a previous study,2 ASD heritability was estimated to be 0.50, and shared familial environmental influences to be 0.04. To define presence or absence of ASD, the study used a data set created to take into account time-to-event effects in the data, which may have reduced the heritability estimates (detailed explanation in Figure).
Using the same underlying data as in the previous study,2 we used an alternate method (used by previous studies in the field) to define concordant and discordant sibling pairs to calculate the heritability of ASD.
A population-based cohort of children born in Sweden 1982 through 2006, with follow-up for ASD through December 31, 2009, were included. The study population and case-ascertainment methods are described in detail elsewhere.2 The study was approved, with a waiver of informed consent, by the ethics committee at the Karolinska Institutet.
Liability-threshold models were fitted using monozygotic or dizygotic twins, full siblings, and paternal and maternal half siblings to decompose the variance in liability to ASD into factors for additive genetic effect (inherited additive effects of different alleles), nonadditive (dominant) genetic factors (interaction effects between alleles at the same locus), shared environmental factors (nongenetic influences contributing to similarity within sibling pairs), and nonshared environmental factors (making siblings dissimilar).3 From each family, 1 sibling pair was randomly included. For each pair, ASD status was defined as the presence or absence of ASD at any time point during follow-up. Differences in probability of being diagnosed depended on birth cohort, due to differing time of follow-up and changes in diagnostic practices, and were handled by adjustment for birth year. Models obtained by excluding 1 or more of the 4 genetic and environmental parameters were compared using likelihood ratio tests. The heritability was calculated as the variance associated with the genetic term(s) divided by the total variance. A 2-sided P value of less than .05 was the threshold for statistical significance. Models were fitted using OpenMx (OpenMx Project), version 2.6.9, and R (R Foundation), version 3.3.3.
The study included 37 570 twin pairs, 2 642 064 full sibling pairs, and 432 281 maternal and 445 531 paternal half-sibling pairs. Of these, 14 516 children were diagnosed with ASD. The model including additive and nonadditive genetic, shared and nonshared environmental parameters was chosen as the full model under which nested submodels were tested. The best-fitting model included only additive genetic and nonshared environmental parameters (Table). Using this model, the ASD heritability was estimated as 0.83 (95% CI, 0.79-0.87) and the nonshared environmental influence was estimated as 0.17 (95% CI, 0.13-0.21). In the full model, the shared environment variance was estimated as 0.04 (95% CI, 0.00-0.14); nonshared environment, 0.16 (95% CI, 0.05-0.30); nonadditive genetic, 0.10 (95% CI, 0.00-0.38); and additive genetic (heritability), 0.69 (95% CI, 0.40-0.86). Using only twins, the heritability was estimated as 0.87 (95% CI, 0.68-0.96).
In a reanalysis of a previous study of the familial risk of ASD, the heritability was estimated to be 83%, suggesting that genetic factors may explain most of the risk for ASD. This estimate is slightly lower than the approximately 90% estimate reported in earlier twin studies1 and higher than the 38% (95% CI, 14%-67%) estimate reported in a California twin study,4 but was estimated with higher precision. Like earlier twin studies, shared environmental factors contributed minimally to the risk of ASD.
Twin and family methods for calculating heritability require several, often untestable assumptions.5,6 Because ASD is rare, estimates of heritability rely on few families with more than 1 affected child, and, coupled with the time trends in ASD prevalence, the heritability estimates are sensitive to the choice of methods. The method initially chosen in the previous study2 led to a lower estimate of heritability of ASD. The current estimate, using traditional methods for defining ASD discordance and concordance, more accurately captures the role of the genetic factors in ASD. However, in both analyses, the heritability of ASD was high and the risk of ASD increased with increasing genetic relatedness.
Accepted for Publication: August 9, 2017.
Corresponding Author: Sven Sandin, PhD, Department of Psychiatry, Icahn School of Medicine at Mount Sinai, 17 E 102 St, New York, NY 10029 (email@example.com).
Author Contributions: Drs Sandin and Kuja-Halkola 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.
Concept and design: All authors.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Sandin, Hultman, Reichenberg.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Sandin, Kuja-Halkola.
Obtained funding: Lichtenstein, Hultman, Larsson, Reichenberg.
Administrative, technical, or material support: Lichtenstein, Hultman.
Supervision: Hultman, Larsson, Reichenberg.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Larson reported serving as a speaker for Eli-Lilly and Shire and receiving research grants from Shire. No other disclosures were reported.
Funding/Support: This study was supported by the Beatrice and Samuel A. Seaver Foundation.
Role of the Funder/Sponsor: The sponsor 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.