Aside from the thresholds, 3 components of liability variance are included, a genetic component, an environmental component shared by twins, and a random environmental component.
The parameter A represents the proportion of liability variance due to the genetic component; the parameter C represents the proportion of liability variance due to common twin environment; and the parameter E represents the proportion of liability variance due to random environment ie, unshared by twins.
When considering the sexes separately, we define 6 total parameters, A m and A f , C m and C f , and E m and E f , corresponding to the values of A, C, and E for males and females, respectively. Male values were estimated from the monozygotic and dizygotic concordance data in males, while female values were estimated from the monozygotic and dizygotic concordance data in females.
This correlation is denoted r mf. Because all monozygotic twin pairs are identical for sex, it is not possible to estimate separate correlations for A m and A f or C m and C f. Liability correlation and hence variance components was estimated from the twin concordance data. Models were fit to data using maximum likelihood with a grid search algorithm as implemented in Excel Microsoft, Redmond, Washington , and hypotheses about model parameters were based on likelihood ratio tests.
Best-fitting models were determined by the Akaike Information Criterion. Models with a low Akaike Information Criterion are the most parsimonious. Details are provided in the eAppendix. A total of pairs fulfilled our initial eligibility criteria. We were unable to establish contact with pairs, primarily because the families had lost contact with the regional center. Of the remaining pairs, declined participation and 10 pairs were ineligible.
Of the pairs whose contact information was sent to the Stanford team, 90 pairs declined participation in the study. Of the remaining pairs, we completed the assessments for pairs. Among the twin individuals, met criteria for ASD of whom also met criteria for strict autism , with the remaining failing to meet criteria for an ASD.
For 10 twin pairs, neither of the twins met study criteria for the broader definition of ASD; these twin pairs were excluded from the genetic analyses. To consider potential bias or differences among the twin pairs who were included in the genetic analysis vs those who were not, we compared the 2 groups on a number of demographic criteria eTable 1. The 2 groups were comparable for most variables, except that the age of the twins included in the genetic analyses was on average slightly older, birth weight of males was somewhat higher, and their mothers and fathers were, on average, slightly more educated and more likely to be white and less likely to be African American.
We also examined clinical differences by examining the distribution of DDS categories between the 2 groups eTable 2. We also looked to see if there was any correspondence between the 6 DDS diagnostic categories and our research diagnoses of strict autism vs ASD eTable 2. Of the twin pairs included in the final analysis, 54 were monozygotic 45 male and 9 female and were dizygotic 45 male, 13 female, and 80 sex-discordant. The monozygotic twins were slightly older and had slightly shorter gestation periods Table 1.
In addition, the mothers of the dizygotic twins were older than the mothers of the monozygotic twins Table 1 , consistent with the known increase in dizygotic twinning with maternal age, 22 and more likely to be white and non-Hispanic.
The probandwise concordance rate for 54 female dizygotic co-twins of male probands was 3. For ASD pairs , concordance estimates for both monozygotic and dizygotic twin pairs were generally higher. Probandwise concordance for female dizygotic co-twins of 76 male probands was 5. Again, these dizygotic concordance rates are higher than previously reported and have a significant impact on the heritability analysis. We also considered possible ascertainment bias in terms of concordance.
Using the DDS categories to define affected, we examined the proportion of twin pairs concordant among the included pairs vs the nonincluded pairs, stratified by sex eTable 1. As described in the eAppendix , twin concordance rates were used to obtain parameter estimates for the genetic models; results are presented in Table 3 and likelihood ratio tests of different models, in Table 4.
In all models, both genetic and shared environmental components were significant. The best-fitting models for both strict autism and ASD suggest that a large proportion of the variance in liability is due to shared environmental factors in addition to genetic heritability.
For strict autism, we could conclude that heritabilities in males and females were equal and that the shared environmental components in males and females were also equal. For the broader ASD phenotype, again we could conclude that heritabilities and shared twin environmental components were equal in males and females.
The shared environment component was estimated to be larger than the genetic heritability component. The poor fit of a pure heritability model can be attributed primarily to the high dizygotic twin concordance relative to the monozygotic twin concordance and population prevalence. To our knowledge, this study is the largest population-based twin study of autism that used contemporary standards for the diagnosis of autism.
All twin individuals underwent a thorough diagnostic and cognitive examination that included a structured interview and observation that allowed differentiation between autism and other delays of development. Although genetic factors also play an important role, they are of substantially lower magnitude than estimates from prior twin studies of autism.
Nearly identical estimates emerged for ASD, suggesting that ASD presents the same liability spectrum as strict autism. The California population, as represented in our twin sample, is highly diverse regarding ethnicity, socioeconomic status, and other demographic factors. Hence, our results should be readily generalizable, especially as compared with previous twin studies, which were based exclusively on individuals from Northern Europe.
However, comparison of demographic and clinical characteristics of twins who did and did not participate showed only modest differences, primarily in parental education eTable 1. Furthermore, concordance rates were not influenced by a range of potential confounding factors.
One possible concern is that a higher proportion of concordant dizygotic pairs participated than concordant monozygotic pairs. Examination of proband status among the monozygotic vs dizygotic pairs suggests this is not the case. Among 28 concordant monozygotic pairs, 24 had 2 probands, while 4 had 1 proband. By contrast, for the same-sex dizygotic pairs, among 13 concordant pairs, 8 had 2 probands and 5 had 1 proband.
Thus, if anything, there might have been some underascertainment of concordant dizygotic pairs. The ACE model we used has several inherent assumptions. Similarly, a critical assumption in the model is that the shared twin environmental effect is the same for monozygotic and dizygotic twins.
If, in fact, monozygotic twins share the relevant environment to a greater degree than the dizygotic twins, some of the effect included in the parameter A would actually be environmental rather than genetic; again, A may actually overestimate the true genetic heritability. Another potential limitation is the validity of the assumptions regarding prevalence rates of autism and ASD. We therefore compared heritability and shared twin environment estimates obtained by varying assumptions about prevalence.
As shown in the eAppendix , we examined the impact of both doubling and halving our prevalence assumptions on derived parameter estimates. Hence, our conclusions regarding the relative importance of genetic and shared twin environment are quite robust to prevalence assumptions. Our study provides evidence that the rate of concordance in dizygotic twins may have been seriously underestimated in previous studies and the influence of genetic factors on the susceptibility to develop autism, overestimated.
Because of the reported high heritability of autism, a major focus of research in autism has been on finding the underlying genetic causes, with less emphasis on potential environmental triggers or causes. The finding of significant influence of the shared environment, experiences that are common to both twin individuals, may be important for future research paradigms.
Because the prenatal environment and early postnatal environment are shared between twin individuals, we hypothesize that at least some of the environmental factors impacting susceptibility to autism exert their effect during this critical period of life.
Nongenetic risk factors that may index environmental influences include parental age, 24 low birth weight, 25 multiple births, 26 and maternal infections during pregnancy. Submitted for Publication: April 13, ; final revision received April 27, ; accepted April 27, Published Online: July 4, Author Contributions: Dr Hallmayer had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Role of the Sponsors: None of the foundations had a role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript. Our website uses cookies to enhance your experience.
By continuing to use our site, or clicking "Continue," you are agreeing to our Cookie Policy Continue. Table 1. View Large Download. Table 2. Table 3. In reality, the distributions of quantitative trait scores for the clinical and non-clinical samples exhibited substantial overlap extending well into the non-pathological range of the distribution; in this sense, there was no clear truncation of the distribution on which to base an absolute threshold to simulate.
When considering, however, that ascertainment of clinical cases is based on a general threshold of impairment, another reasonable approach to establishing mathematical expectations would be to consider a range of less stringent latent thresholds. Again, our observed correlation coefficients fell well below these values. Also, there was no effect of erosion of twin—twin correlation when considering exclusively the non-clinical sample which, in essence, reflected more truncation in relation to the pathological range than the clinical sample did in relation to the non-pathological range.
We note also that a previous study of supra normal intelligence, restricted to MZ twins above the 85th percentile for IQ exhibited no significant erosion of the identical twin correlation in comparison to that observed for IQ in the general population Haworth et al.
Body mass index reflects a counterexample in which differential heritability at the tail of the distribution has been documented and is believed to reflect distinct environmental influences that come on line in that range Tsang et al.
Although these findings await replication in a sample comprising an even larger number of identical twins in the clinical range of affectation, an overarching interpretation of these results is that the factors responsible for variation in severity within the clinical range diverge from those that are responsible for the heritability of the condition itself. The inference is that when genetic liability for ASD exceeds a clinical threshold, it engenders a marked increase in vulnerability to the effects of non-shared environmental influences, which were observed to exert minimal effects on individuals whose ASD trait burden fell below the threshold for diagnosis.
Recently, Hegarty et al. The latter dominated our observations and could encompass epigenetic changes, random developmental in utero or environmental extrauterine perturbations, or the effects of somatic mutations, to which affected individuals might be more vulnerable than unaffected individuals.
We note that in the latter case, a relatively high and influential somatic mutation rate would need to be invoked to explain the continuous nature of the distribution of accentuated twin—twin differences observed in our clinical sample. Overall, these findings constitute an identical twin demonstration of a remarkable observation originally reported by Spiker et al.
Very recently, Gazestani et al. In contrast, when variation at the pathologic extreme disrupts this process, a direct consequence may be vulnerability to such perturbations and an increase in variance as a signature of atypicality—this has been observed in the pleiotropic effects of other deleterious influences on developmental phenotypes, including patterns of variation in gene expression across individuals affected by discrete monogenic syndromes engendering autism Nishimura et al.
In this study, the notable absence of age effects on the substantial non-shared environmental factors influencing severity ratings of identical twins, coupled with knowledge from prior studies of the very high longitudinal stability of severity measurements within an affected individual from early childhood through adulthood Wagner et al. This interpretation must be considered cautiously since our data were cross-sectional in nature and would need to be followed by demonstration of early effects in a prospective longitudinal context.
Future studies should explore whether this general phenomenon may contribute to the formidable—but as yet largely unexplained—influence of non-shared environmental factors observed in genetic epidemiologic studies of many psychiatric disorders.
Our sample size of identical twins in which one or both was affected by clinical ASD was necessarily limited in size and clinically-ascertained. To our knowledge, there does not yet exist an epidemiologic sample collection large enough from which greater numbers of quantitatively-characterized identical twins with clinical-level ASD affectation can be derived.
Our trait correlations for clinically affected MZ twins were substantially lower than those obtained from severity measurements in comparably-affected identical twins ascertained in the TEDS UK study via population screening Colvert et al. Although it is possible that clinical ascertainment in this study could have resulted in selection for more discordance than is representative of the population of MZ twins, the categorical probandwise concordance for the sample was over 0.
Furthermore, equivalent continuous distributions of twin—twin differences, observed in our two independent clinical samples, mitigated concerns about the effects of specific ascertainment strategies, and suggested that an increase in sample size would be unlikely to yield fundamentally different results. Another potential limitation is that only a subset of the identical twins in this study had zygosity confirmed by molecular genetic analysis; however, the high probandwise concordance again, on the order of 0.
Some of the observed non-shared environmental influence on clinical twin severity ratings could have occurred on the basis of measurement error, but this was not observed in the epidemiologic sample and a recent analysis of test-retest reliability for maternal ratings of clinically-affected subjects using the SRS-2 Wagner et al. For the most extreme differences observed in the present cohort, ratings provided by mothers were corroborated by ratings derived from expert clinician observations, making it unlikely that rater contrast effects disproportionately influenced the measurements of clinically-affected MZ twins.
These and other convergent findings have a number of key implications for ongoing research in autism and related disorders. Studies attempting to relate neural or behavioral signatures to genetic susceptibility for ASD may have substantially higher statistical power when conducted among individuals with total symptom burden below rather than above the clinical threshold for diagnosis.
Symptom severity ratings derived from behavioral observations of individuals clinically-affected by ASD, though stable over the course of development and highly reliable in a psychometric sense, may bear little to no relation to the causal genetic underpinnings of the condition itself, should not be construed as heritable unless proven to be so, and likely reflect, at least in part, the consequences of random events to include neonatal medical events that have disrupted otherwise predictable developmental trajectories, possibly early in life Willfors et al.
Two prior studies involving alternate measures of autism-related symptomatology in concordant ASD-affected siblings yielded findings that are highly congruent with our observations. Mazefsky et al. Similarly, Goin-Kochel et al.
Sibling correlations for parent-reported adaptive functioning in the social and communication domains were on the order of 0. If our findings on the relative non-heritability of severity ratings among clinically-affected cases are representative of most or all autistic syndromes, any observed relationships between biological markers and severity ratings would not be expected to relate to the cause of the condition, rather to modifiers or epiphenomena, and, therefore, should be interpreted cautiously.
Moreover, these data underscore alternate prospects for biomarker discovery in which neural, genetic, and physiologic signatures are linked to endophenotypic contributors to autistic syndromes see Pohl et al.
Finally, these data qualify understanding of the foundational statistic of monozygotic twin concordance in autism. Our findings confirm probandwise categorical concordance statistics consistent with marked heritability for ASD Sandin et al. Identical pairs in which one or both exceed the clinical threshold for affectation exhibit a continuous range of contrasts, the most pronounced of which qualify for categorical discordance, but this occurs rarely.
More often, even when there exist substantial differences between twins, both typically exhibit levels of symptom burden near or above the threshold for a diagnosis. The broad range of quantitative twin—twin differences documented in this study account for the wide variations in previously reported statistics for MZ twin correlation, particularly from small samples in which concordance was defined in relation to an arbitrary threshold.
Based on the current findings, we favor a more standardized parameterization of identical twin discordance that is quantitatively defined, on the basis of 1 at least a 1.
Autism 19 5 — PubMed Google Scholar. JAMA Psychiatry 72 5 — Constantino JN The quantitative nature of autistic social impairment. Pediatr Res R—62R. Constantino JN Recurrence rates in autism spectrum disorders. JAMA — Constantino JN Early behavioral indices of inherited liability to autism. Pediatr Res 85 2 — Western Psychological Services, Torrance. Google Scholar.
Arch Gen Psychiatry — Constantino JN, Todd RD Intergenerational transmission of subthreshold autistic traits in the general population. Biol Psychiatry — J Child Psychol Psychiatry — Am J Psychiatry — Nature — Autism — J Autism Dev Disord 44 8 — Nat Neurosci 22 10 — J Autism Dev Disord 38 6 — Goldsmith HH A Zygosity questionnaire for young twins: a research note.
Behav Genet 21 3 — Pediatrics — Behav Genet — Mol Psychiatry. Wiley, New York, p J Autism Dev Disord 30 3 If heritability provides such limited information, why do researchers study it?
Heritability is of particular interest in understanding traits that are very complex with many contributing factors. Moore DS, Shenk D. The heritability fallacy. Wiley Interdiscip Rev Cogn Sci. Epub Dec 1. PubMed: Tenesa A, Haley CS. The heritability of human disease: estimation, uses and abuses. Nat Rev Genet.
Other chapters in Help Me Understand Genetics. Genetics Home Reference has merged with MedlinePlus. Learn more. The information on this site should not be used as a substitute for professional medical care or advice. Contact a health care provider if you have questions about your health. What is heritability?
0コメント