Taking stock of critical clues to understanding sex differences in the prevalence and recurrence of autism

In this issue of Autism, new evidence for the equivalency of symptom burden and structure among males and females affected by autism spectrum disorder (ASD) continues to move the spotlight from a notion that the sex ratio in ASD results from the condition being somehow different between the sexes, to the more likely scenario that males and females categorically differ in their susceptibility to reaching a (common) diagnostic threshold in the setting of pathogenic risk. The consensus reached by these studies, that is, of relative equivalency in symptom structure between diagnosed males and females, across a variety of methods of symptom ascertainment, is particularly compelling given the rigorousness and accumulated sample sizes of the respective study designs. Establishing a benchmark of diagnostic equivalency is a critical anchor for ongoing efforts to understand the biology of differential liability to ASD as a function of sex, which is one of the most profound and pervasive modifiers in this remarkably heterogeneous—and remarkably heritable—group of disorders (see Tick et al., 2016; Werling, 2016). Clearly, elucidation of the nature of sex-specific differences in ASD liability—given that the vast majority of the heritable influence appears to be traceable to autosomes (not sex chromosomes)—carries with it one of the most compelling prospects for potent novel therapeutic interventions on the discovery frontier.

It is important to place these findings into the context of what is already known about sex bias in ASD. Five years ago, Constantino and Charman (2012) summarized the accumulated evidence in a commentary on an important epidemiologic observation by Dworzynski et al. (2012) that, at a given absolute threshold for autistic trait burden, boys were more likely than girls to be diagnosed with autism, largely on the basis of comorbid, non-ASD-specific, behavioral abnormalities bringing them to clinical attention (see also Duvekot et al., 2016; Frazier et al., 2014). What had been scientifically established by that time was that subtle sex differences in mean scores for autistic traits observed in the general population could not explain the magnitude of the sex ratio for clinical affectation (see also Øien et al., 2017); that there existed known familial monogenic syndromes (FRX, SHANK1) that exhibited sex-specific reduction in the expression of inherited liability for ASD; that recurrence rates in families were remarkably equivalent for male and female index cases (the absence of the Carter effect for clinical ASD, see Constantino, 2014—this contrasts with an observation of sex-specific variation in quantitative thresholds for the aggregation of subclinical autistic traits, reported by Robinson et al., 2013); that the sex ratio is markedly attenuated (nearly 1:1) for severe ASD syndromes accompanied by intellectual disability and dysmorphism (so-called “complex autism”) and markedly accentuated when exclusively considering ASD syndromes not accompanied by cognitive impairment; and that recurrence ratios were attenuated but not eliminated when using quantitative ratings for affectation standardized by sex as normed thresholds for clinical categorization. It had also been repeatedly observed that a broad array of autistic traits aggregate among the siblings of individuals with ASD, but that the distribution of those traits differed markedly between the sexes—continuously among brothers, dichotomously among sisters—providing evidence for a categorical effect of resilience among females in multiplex ASD-affected families.

In the 5 years since that time, there have been some very significant new advances, notably: (1) although under-identification of affected females and the phenomenon of symptomatic “camouflage” in school-aged girls (Dean et al., 2016) remain significant concerns for the delivery of appropriate clinical care and attention to affected females, ascertainment of early diagnosis in the later-born infant siblings of affected individuals (i.e. in systematic, unbiased, prospective designs) reveals the same pronounced 3:1 sex ratio that is universally observed later in childhood and adolescence (Constantino, 2016; Messinger et al., 2015); (2) in stark contrast to the case for intellectual disability, only a tiny proportion of molecular genetic risk for ASD is resolvable to genes on sex chromosomes; (3) a relatively higher mutational burden is observed in affected females than affected males, supporting a “female protective model” in autism and related neurodevelopmental disorders (Jacquemont et al., 2014); (4) circulating prenatal testosterone is an unlikely culprit for sex differences (Eriksson et al., 2016); (5) there is no gene on the X chromosome harboring a singular polymorphic that protects females from ASD in dose-dependent fashion (Gockley et al., 2015), however; (6) X-linked genes controlling secondary sex characteristics may collectively exert some portion of the influence on sex-specific differential liability to ASD (Mitra et al., 2016)—this latter finding utilized a very large accumulated single nucleotide polymorphism dataset and additionally revealed that there was (a) no difference in polygenic load between males and females affected by ASD, (b) no sex difference in autosomal loci conferring polygenic risk, and (c) no preponderance of common variation in brain genes with expression levels known to be influenced by androgen or differentially developmentally expressed in boys versus girls. These foundational new findings are critically informing a new generation of studies focused on comprehensive explorations of normative sexual dimorphisms that may provide critical clues to differential liability. For example, McCarthy and Wright (2016) described early developmental dimorphisms that may move males closer to neurodevelopmental vulnerability thresholds that can consequently be more easily breached by genetic or environmental perturbations during critical periods of brain development. Other recent studies have focused on sexual dimorphisms in brain circuitry that may accentuate (Lai et al., 2013) or become disrupted (Alaerts et al., 2016) in ASD.

So it is within this now-formidable (and continuously evolving) body of scientific evidence relevant to the sex ratio in ASD that these new contributions are best understood. Frazier and Hardan (2016) re-examine parent-report symptom burden and factor structure in the Simons Simplex Collection database to determine, essentially, whether it requires more severe impairment to be a female diagnosed with ASD than a male with ASD. They conclude that sex differences in autism prevalence are not likely to be accounted for by the manner in which symptoms are ascertained or by true differences in symptom levels between males and females. Similarly, Grove et al. (2016) examined self-report ratings on the Autism Spectrum Quotient Short Form from 550 adults with confirmed diagnoses of ASD (a majority of the sample with above average IQ) and observed a highly comparable factor structure and nearly equivalent levels of severity for males and females affected by the disorder. Sutherland and colleagues solicited parental observations regarding the types of behaviors encompassed by repetitive behaviors or restricted interests in ASD-affected children and found that in the setting of equivalent symptom burden there were nuanced differences in particular interests that fell along traditional gender lines (dinosaurs and transportation favored by boys, animals and art by girls). Hull and colleagues conducted a systematic meta-analysis of studies comparing behavioral and cognitive characteristics between the sexes not only within ASD but (importantly) among typical controls and observed that sex differences in core ASD symptoms that emerged from the meta-analysis closely mirrored sexual dimorphisms among typical controls. Furthermore, although the magnitude of these between-sex differences in core ASD traits was modest, they observed that sex differences in other behavioral and cognitive characteristics can be considerably more pronounced in ASD—this raises the important issue of complete phenotypic characterization (i.e. beyond core ASD symptoms) in future research characterizing individuals with ASD syndromes of diverse genetic origin (see Constantino and Charman, 2016). Finally, Halladay capitalized upon clinician impressions of their ASD-affected patients and identified perspectives on the course of adolescent development—a period of pronounced normal sexual dimorphism—that highlights the way in which sex may interact with the autistic phenotype over the life course to differentially impact adaptation, even if the syndrome itself is fundamentally equivalent between sexes at the time of diagnosis or when exclusively considering core autistic symptomatology.

The overarching conclusion, as articulated by Frazier and Hardan, is that neurobiological studies are needed, all the more urgently given what we now know, “to uncover the female protective or male risk factors that lead to sex discrepancy.” It is therefore with ever-stronger scientific justification—and an invigorated sense of the pivotal importance of sexual dimorphism in discovering new opportunity for higher-impact intervention—that the pathway to translational advances in ASD must incorporate consideration of these important principles. Future research must adopt designs that will yield a next set of clues to a definitive elucidation of the biology of sex bias, not only for those factors that differentially influence ASD prevalence but for those (possibly overlapping) factors that operate in the many neurodevelopmental disorders of childhood which manifest marked differences in prevalence or severity on the basis of sex.