Design of Studies for Molecular Genetics
The hope of molecular genetic studies is that they may untangle the issues of genetic subtypes in the future, but the degree to which we will have success identifying these genetic susceptibility factors will in part be determined by the complexity of the genetics and the degree to which we can correctly address these issues in the planning and design of the study. In this section, we will address some key issues that should be considered in the design of molecular genetic studies of ADHD.
When designing an investigation of a phenotype with extensive genetic heterogeneity, provision for a large sample size is imperative. That is, if there are a number of different genes that can influence the phenotype, then a large number of families will be required in order to have the statistical power to detect each gene. Risch and Merikangas (1996) have estimated the number of families that is required to detect a gene of various genotypic risk ratios. At this point for ADHD, we do not know the number of genes that contribute to the phenotype or the impact that each gene has on the expression of the phenotype. It is therefore difficult to specify a priori an appropriate sample size.
Another factor to consider in the design of a genetic study is the penetrance. Given the previous discussion of reduced penetrance and the possibility that an individual could carry ADHD susceptibility genes without expressing the phenotype, then a strategy should be chosen to prevent complications from incomplete penetrance. In other words, some family members may carry the susceptibility genes without expressing the behaviours. For example, a collection of large multi-generational families may be unwise in that the diagnosis of some members of the pedigree may be difficult to specify with certainty; possibly because they are older and no longer symptomatic. Likewise, some individuals may be symptomatic but may not achieve a research criteria (i.e., family members who manifest only five symptoms of inattention, and thus do not exceed the established criteria of six or more of the nine symptoms on this domain). Therefore, it may not be possible to correctly classify all individuals into either an affected or unaffected category. Also critical to the use of large multi-generational pedigrees is the diagnosis of adult members of the pedigree which is known to be problematic (see Chapter 4, by Hay. McStephen, & Levy). Misdiagnosis of a critical member of a pedigree can substantially change the results of the linkage analysis, as was demonstrated by the change in linkage results by orders of magnitude for affective disorder (Egeland et al„ 1987; Kelsoe et al.. 1989).
One strategy would be to include only subjects that can unequivocally be designated as affected, such as an affected sibling pair design. This type of design avoids the difficulties of the requirement to be clear as to which family members definitely do not have the phenotype.
