GWAS have provided us with a tool to identify new genes for multifactorial disorders like ASD and ADHD. These studies require large samples to be investigated, with thousands of cases and equal numbers of controls to be analyzed. For this, international collaborations are being formed; one of the largest ones for psychiatric disorders being the Psychiatric GWAS Consortium [
74]. One of the three main objectives of the Psychiatric GWAS Consortium is “cross-disorder analyses, including analyses of combinations of disorders and of phenotypes observed in two or more disorders … based on recommendations of an expert committee. Because data are insufficient to determine what common cross-disorder etiological factors might exist, alternative phenotypes should be explored. GWAS analyses have produced surprising cross-disorder associations, such as those found for cancers and inflammatory bowel diseases, which could also exist for psychiatric disorders given the many common symptoms” (p. 548 [
73]). No such cross-disorder analyses have been conducted for ADHD in combination with ASD. Therefore, we will propose several steps for future research that may facilitate the detection of pleiotropic etiological factors for ADHD and ASD.
A first step in future family genetic studies on the shared etiology of ADHD and ASD would be to assess ADHD and ASD in both parents and (multiple) off-spring. If parents and children are assessed using comparable phenotypic measures, a more direct examination can be made for inter-generational transmission of ADHD, ASD, and ADHD + ASD symptoms. It is of crucial importance to include not only categorical measures of ADHD and ASD into this design, but also to make use of quantitative phenotypic measures. The latter does more justice to the quantitative nature of both disorders; individuals with subthreshold symptoms are otherwise grouped together with individuals with no symptoms. Two types of families would be suitable for this approach: families in which the disorders are separately present in different family members, and families in which members are affected with both disorders, simultaneously. Both types of families may provide insight into pleiotropic genes and genes that are uniquely associated with one of both disorders: the former through simultaneously regressing gene effects on the ADHD and ASD phenotypes within families, the latter through regressing gene effects on the ADHD and ASD phenotypes within individuals.
A second (or parallel) step would be to assess candidate endophenotypes within a family-based design in which both ADHD and ASD are present in one or more family members. Endophenotypes, such as neuroimaging or neuropsychological functions, are defined as heritable vulnerability traits that form a link between genes and observable symptoms [
39]. The first generation of ADHD and ASD GWAS studies using the ASD or ADHD clinical phenotypes made it clear that psychiatric diagnostic phenotypes may not be optimally suited for gene discovery. Compared to phenotypes, endophenotypes may improve statistical power by (a) being more heritable than phenotypes, (b) being more easily quantifiable than dichotomous
DSM diagnostic categories, (c) being more reliable and objective than phenotypic measures, (d) being more biologically relevant, and (e) being more useful in creating genetically homogeneous subgroups of patients. Candidate endophenotypes for ADHD + ASD would most likely show the following characteristics: (a) more impaired in ADHD + ASD probands compared to ADHD or ASD probands, (b) more impaired in parents and siblings of ADHD + ASD-probands compared to those of ADHD or ASD probands, (c) stronger correlation between siblings in ADHD + ASD families compared to those in ADHD or ASD families, (d) cross-correlating with both ADHD and ASD in family members. In extensive review on candidate ADHD + ASD endophenotypes (Rommelse et al. submitted), we concluded that executive functioning, response variability, social cognition, motor coordination, language and intelligence are the most promising candidate neuropsychological endophenotypes. Candidate structural brain endophenotypic measurements include overall brain size, corpus callosum and cerebellum size, and structural integrity of the fronto-striatal circuitry. Too little data existed on functional brain imaging measures to propose candidate endophenotypes in this domain. Body parameters that may be useful for future studies are finger length ratio, height and weight and, possibly, fatty acid abnormalities. By incorporating these candidate endophenotypes into molecular genetic studies, both statistical power as well as insight into gene-behavior pathways may be gained.
However, the endophenotype approach is not without its shortcomings. For example, it is not well established that endophenotypes are indeed more reliable and objective than clinical phenotypes [
11]. Furthermore, a putative endophenotype might not be less ‘genetically complex’ or more heritable than the phenotype itself [
11], thereby not automatically resulting in an increased power to detect genetic effects. For example, though useful [
78] and potentially more powerful than studies of clinical phenotypes, quite substantial sample sizes are still needed when endophenotypes at the level of neurocognitive performance are used for genetic studies [
11]. However, structural and functional neuroimaging measures do appear to be able to reduce necessary sample sizes for molecular genetic analyses considerably [
58,
62]. Therefore, despite possible shortcomings, we would nevertheless recommend incorporating candidate endophenotypic measures in family-genetic research for ADHD + ASD.
Another recommendation is to analyze multiple endophenotypes at the same time. This could be achieved by combining related traits into a single outcome using principal component analyses or factor analyses. Hereby common underlying constructs are extracted based on covariance between measures while the number of variables is reduced. Main advantages over the use of multiple correlated endophenotypes in univariate analyses are an increased reliability and decreased risk of false positive results. A potential disadvantage of this approach may be the loss of information by removing variance that is specific to each of the endophenotypes. Alternatively, multivariate multi-level models may be used. Multivariate multi-level models flesh out the relationships between numerous intermediate phenotypes [
82]. Using this technique, multiple explanatory genetic variables can be simultaneously included in the model (multivariately) and regressed on variables measured at the endophenotype and phenotype level. As most cases of ADHD and ASD are probably caused by many genes of small individual effect, incorporating the relationships between multiple genes, endophenotypes and the clinical phenotypes in the model are even more important. The multivariate endophenotype approach has already proved useful in various other domains of endophenotypic research [
13,
71,
97] and will most likely prove fruitful in the search for common etiological factors for ADHD + ASD.
Last but not least, the effect of environmental variables on endophenotypic and phenotypic functioning should not be overlooked. A candidate endophenotype can be associated with a diagnosis, be heritable and be reliably assessed, but nevertheless be under (strong) influence of environmental factors. Gene–environment interactions (mainly related to early/obstetric variables) on the phenotype are increasingly reported in ADHD and ASD literature [
12,
60,
68,
90,
91; see also elsewhere in this issue] and may also exist for endophenotypes. Therefore, gene–environment interactions are of vital importance to take into consideration when studying a common etiological basis for both disorders.
In sum, properly designed and powered candidate gene and linkage studies are missing for ADHD + ASD. Recent GWAS for clinical ADHD and ASD phenotypes have identified a number of viable pleiotropy candidates. GWAS currently seem to be the most appropriate approaches to learn about the role of common genetic variants in the combined disorders. We propose that future studies examining shared familial etiological factors for ADHD and ASD (a) use a family-based design in which the same measurements are obtained from all family members, (b) incorporate a coherent set of candidate endophenotypic measures that are most likely shared risk factors for ADHD and ASD, (c) apply multivariate multi-level models for statistical analysis, in order to examine the relationships between the various endophenotypes and phenotypes, and (d) incorporate environmental measures into the design.