A key aspect of quantitative variation in autistic symptomatology involves the questions of whether (or to what extent) the three components of the autistic syndrome (social deficits, communicative deficits, and stereotypic behavior/restricted interests) represent independent symptom clusters (i.e., each with its own quantitative or qualitative architecture), or rather covary (“travel” together) in nature. In this review we focus on autistic social impairment as a core feature of autistic syndromes but recognize that language impairments and stereotypic behaviors may exhibit distributions and patterns of aggregation that diverge from the pattern described here for autistic social impairment. Here we summarize briefly what is understood about the associations between the three symptom domains of the autism triad in order to best contextualize the implications of a quantitative architecture for autistic social impairment.
Although some large-scale general population studies have suggested that the inherited influences on the three domains of symptoms are substantially (20
) or partially (21
) independent of one another, a caveat is that such studies can be confounded by measurement methods that do not ascertain autistic symptoms and traits with enough specificity. For example if children with non-autism-related disorders (such as specific language impairment or conduct disorder) are contributing to elevations in “autistic trait” scores ascertained in large populations, it can lead to overestimates of the extent to which separate lines of inheritance are responsible for the symptoms, since many of these syndromes are likely to have their own causal mechanisms independent from autism.
In contrast, factor, cluster, and latent class analyses of autistic symptoms in family studies have revealed substantial overlap in the three criterion domains for autism delineated in the Diagnostic and Statistical Manual of Mental Disorders 4th
edition (DSM-IV). For example, Spiker and colleagues (22
) studied sibling pairs affected by autism, and found that empirically-derived clusters of symptoms within families differed not by specific symptom sets, but by the degree of impairment that existed (mild, moderate or severe) across all three DSM-IV criterion domains for autism. Their findings were most consistent with a model of autistic symptomatology arising from a single, heritable, continuously distributed deficit which might influence dysfunction in all three symptom domains.
Similarly, Sung, Dawson and colleagues (12
) examined features of the broader autism phenotype in the relatives of autistic probands and found evidence for the primary aggregation of highly heritable social deficits that explained variation in symptomatology across other domains of the autistic syndrome. Constantino and colleagues (23
) applied factor, cluster, and latent class analysis to diagnostic interview and quantitative trait data (using the Social Responsiveness Scale, SRS) from children representing the entire range of autistic social impairment, from minimally affected to severe, and consistently observed a unitary factor structure across data sources and methods of analysis, reinforcing the syndromic nature of autistic impairment. A recent principal components factor analysis of an accumulated Washington University sample of 1,799 boys from 1,799 separate families, including over 300 clinically affected by autism, yields the factor structure depicted in the scree plot in , indicating that a primary factor accounts for greater than 30% of the variance with all other possible factors contributing only minor components of variance. It is important to note that in this observation and in previous studies, the symptoms which load most strongly onto the principal factor represent all three of the DSM-IV criterion domains for autism.
Scree plot of principal components analysis of SRS sample of 1,799 boys from 1,799 separate families, representing the full range of variation in autistic severity from unaffected to severely affected by clinical autistic syndromes,
Subsequent to the original reports summarized above, Gotham, Lord, and colleagues (26
) factor analyzed data from thousands of structured diagnostic observations of children with autism and concluded that the symptoms encompassing social deficiency and communicative deficiency comprised a single empirically-derived factor. Even at face value, a review of typical autistic traits reveals aspects of overlap across symptom domains that provide insight into key unifying constructs. The tendency for the social ability of individuals with ASD to be compromised by an over-focus on details (“missing the forest for the trees”) is highly reminiscent of the preoccupation with detail that characterizes young affected children's unusual (stereotypic) play with toys and older children's restriction in range of interests. It has been hypothesized that an underlying deficit in the assignment of salience (28
) and the consequent absorption with details could underlie the social, stereotypic/behavioral, and even the communicative deficits (e.g., sentence comprehension compromised in the context of preserved decoding of individual words) of the autistic syndrome.
These observations provide a new framework for understanding quantitative variation in autistic symptomatology and challenge the 3-criterion taxonomy for differentiating specific pervasive developmental disorders (PDDs) in DSM-IV, by raising the possibility that each of the common disorders (Autistic Disorder, Asperger Syndrome, PDD-NOS) lie along a continuous distribution of impairment with a more parsimonious underlying factor structure. In such a reconceptualization, variation in autistic severity interacts with variation in other domains of development (general cognition, temperament, proneness to anxiety) to produce specific profiles of individual adaptation. For example, Asperger Syndrome, which is described in DSM-IV as a separate disorder (defined as autistic impairment without substantive language delay, and generally characterized by average to above average intellectual functioning), might be viewed as an autistic syndrome that is compensated by a level of intellectual functioning that is adequate to sustain normal language development, even in the presence of a level of social impairment that is otherwise typically associated with language impairment. In that sense it is the preservation of a relatively high level of general cognition—not the existence of a separate disorder—that uncouples the usual association between social and communicative impairment observed in autistic syndromes.
Molecular genetic studies are now just beginning to add to our knowledge of the factoral structure of autistic syndromes. While many disparate mutations have been associated with the same triad of symptoms observed in autism, quantitative trail loci (QTL) analyses have indicated that common autism susceptibility alleles may preferentially confer risk among specific subsets of autistic patients, for example those with versus without severe language delays (29
). It has also been observed that a given large chromosomal rearrangement or allelic variation associated with autism (i.e. one that occurs more common in ASD than in the general population) might have highly variable phenotypic expression, as observed for 16p11 deletions which are found in patients with autism, patients with mental retardation without autism, and in relatively-typically-developing individuals within autism-affected families (3
). Recognizing that a wide range of quantitative autistic traits may be present in the siblings of children with autism, it is possible to explore whether the association between phenotype and underlying genetic or neurobiologic mechanisms might be more appreciable when considering quantitative autistic trait information from all siblings in affected families rather than restricting analysis to the transmission of categorical disease states. Molecular genetic analyses utilizing quantitative trait information from all children of participating families (not just the fully affected subjects) have indicated that specific linkage signals in two independent multiplex family registries were substantially enhanced when adopting this approach (30
Finally, it is well recognized that some syndromes of specific language impairment (SLI) are inherited independently from autism, however, as discussed above, several studies have reported that language disorders with distinct autistic qualities (pronoun reversal, socially inappropriate phrases) and accompanied by mild-to-moderate levels of social deficiency also aggregate in the unaffected siblings of autistic probands (4
). Thus, although the relationships between the symptom domains of the autism triad remain a subject of active investigation, there is accumulating evidence that they represent correlated behavioral manifestations of an underlying quantitative neurodevelopmental impairment. Ultimately molecular genetic studies and neuroimaging studies are poised to make major contributions to our understanding of how (and in what combinations) the various clinical manifestations of autism arise, and this in turn will inform strategies for intervention for specific sub-groups. It is highly conceivable that the severity distribution that constitutes the autism spectrum disorders is fully continuous with the distribution of sub clinical autistic symptomatology that is appreciable in the general population (as reviewed below). It has recently been demonstrated for autism (32
) and for other complex diseases such as hypertension (33
) that variation within the normal phenotypic range can be caused by variation in the same genes that are responsible for clinical disease states. From an evolutionary standpoint, it is possible that inherited factors that are actually adaptive when phenotypic expression is mild (as may well be the case for autistic traits) may be highly preserved in the population and result in clinical disease states only when severely expressed or in interaction with other genetic or environmental factors.