In this study we further analyzed the findings of our previously published study comparing morphological features in a large cohort of autistic children and matched-paired controls. Here, we sought to determine the sensitivity and specificity indices based on these morphological features in ASD, in comparison with normal control subjects (Fig. a–d). This is the first large scale study to explicitly examine the predictive value of morphological features in patients with autism. Several interesting findings emerged from this study.
ROC analysis indicated that the higher prevalence of dysmorphic features in ASD as measured on the whole scale of the morphological features as well as well as on the subscales of major and minor abnormalities is a powerful predictor, which showed extraordinarily high specificity and sensitivity for detecting ASD.
By employing recursive partitioning, we have identified specific morphological features whose expression may be useful diagnostically in discriminating ASD and control subjects, as shown in Fig. . These new findings provide a proof of principle and may have the potential to become the basis for the development of diagnostic or prognostic tests. Of the morphological measures used in this investigation, asymmetry of the face, multiple hair whorls and prominent forehead were most specific to ASD patients. The topographical distribution of dysmorphology in our study is consistent with the literature. Other clinical researchers also proposed to use dysmorphology as a tool to delineate heterogeneity in autism by looking for biologically based phenotypes found in consistent proportions of ASD individuals (Miles et al. 2008
Asymmetry of the face in ASD has been recently documented in a recent 3D morphology study (Hammond et al. 2008
). However, asymmetry of the face has also been found in patients with schizophrenia (Weinberg et al. 2007
). As we confirmed the higher rate of morphological features in autistic patients as compared to normal controls, we are faced with new challenging questions.
First, why do autistic patients have higher rates of morphological features? Apparently, a common genetic vulnerability for developing autism is reflected in morphological features (Rzhetsky et al. 2007
). Several developmental genes have recently been identified that play a paramount role in shaping body structures. Moreover, new insights into craniofacial morphogenesis have indicated that a rapidly increasing number of genes are known to regulate cerebrocraniofacial development (LaMantia 1999
). It can be speculated that the genes that determine the craniofacial morphology overlap with candidate genes for autistic disorders. Alternatively, the observed correlations between specific morphological features and autism might be the effects of temporal exposure of different anlage to as yet undefined factors that impact growth and development.
Although our findings indicated that morphological features could have a predictive value for the diagnosis of autism, an intriguing question concerns their specificity concerning ASD subtypes and other neurodevelopmental psychiatric disorders. Although we predicted that morphological features should be able to classify autistic patients into different subgroups, our data did not support this hypothesis. One reason could be that patients were not selected on the basis of these different subtypes and therefore that not all subtypes were equally prevalent in our study population. Additionally, characterization based on DSM IV subtypes may not relate to underlying differences in etiology. Larger studies that are designed specifically to assess morphological features in different autistic subpopulations are needed to specifically investigate this issue. Another question concerns whether morphological features found in autism differ from those found in other disorders. In a recent meta-analysis, a higher prevalence of morphological features was also established in schizophrenia (Weinberg et al. 2007
). Do morphological features seen in autism have a different etiology than those in schizophrenia, or do disorders associated with morphological features share a common etiological basis with schizophrenia and autism? Some evidence for such an overlap comes from the observation that individuals with ASD may also be at greater risk for developing schizophrenia (Murphy and Owen 1996
; Esterberg et al. 2008
). Emerging studies have described the possible links between the two disorders by means of the genetic overlapping (Carroll and Owen 2009
; Gejman et al. 2011
). Findings indicating overlapping markers could provide important clues regarding the underlying genetic bases of these disorders.
This study had some limitations that should be borne in mind when interpreting the results of this study. There have been several approaches to delineate more homogenous subgroups within autism, and those attempts have also been influenced by diagnostic bias as shown by the Simons Simplex Collection analysis. A huge limitation originates from the complex behavioral phenotype of ASDs. Due to the multifactorial nature of the disease, each individual aberration has a modest effect, and the gene–gene interaction and/or gene-environment interaction may attribute to the observed phenotype. Currently, we do not have a coherent understanding of the relationship of genotype and phenotype in ASDs (State and Levitt 2011
). Moreover, robust diagnostic specificity is often lacking for endophenotypes and reflects the fact that different disorders may share genes, and also share partially overlapping neural substrate dysfunction and clinical features (Braff et al. 2007
Second, as morphologic examination requires in-person examination, it is generally not possible for the raters to be blind to diagnosis. Although we made every attempt to ensure that the assessments were carried out blindly to diagnosis, we acknowledge that blinding may not have been complete. However, to prevent observer bias, 11 % of controls and 10 % of patients were scored independently by two observers, resulting in very high kappa scores. Additionally there were no prior hypotheses as of which morphological abnormalities should be associated with autism, and the finding that some morphological abnormalities were not associated at all with the diagnosis, or were even more frequent in controls than in patients suggests that there was no general rater bias.
Third, we used typically developing children as a comparison group in this study. Future research may extend the findings of this study by investigating non-ASD neurodevelopmental disorders such as schizophrenia, ADHD and bipolar disorders.
Fourth, in order to have a homogenous sample we limited our study population to Caucasian patients and controls; because ethnicity can influence the prevalence of morphological abnormalities. Future studies are needed to establish similar norms for other ethnic groups. Likewise, we restricted ourselves to non-mentally retarded, high functioning ASD patients. Therefore, we cannot generalize our findings to mentally retarded ASD patients.
Despite these limitations, the present study provides evidence that morphological features are significantly increased in the patients with autism and that some unknown prenatal biological mechanism is likely responsible for producing these anomalies which may yield further knowledge about the developmental origins of the disease. If independently replicated, the findings have potential utility for early detection of ASD.