Statistical approaches to complex phenotypes: evaluating neuropsychological endophenotypes for attention-deficit/hyperactivity disorder.

There is renewed interest among psychiatric geneticists in endophenotypes, constructs posited to be more directly and strongly influenced by candidate genes than manifest disorders. Researchers have proposed various criteria for the selection of endophenotypes useful in finding genes that underlie psychiatric disorders. These criteria include the endophenotype's psychometric properties, its relationship to the disorder in the population and within families, its expression in probands' unaffected relatives, its heritability and common genetic influences with the disorder, its association with candidate genes that underlie the disorder, and its mediation and moderation of association between the candidate gene and the disorder. In this article, analytic methods for evaluating the validity and utility of putative endophenotypes consistent with these proposed criteria are reviewed. The use of such analyses is illustrated with data on childhood attention-deficit/hyperactivity disorder and time to complete Trails A and B from both a candidate gene study of clinically referred children and a study of non-referred twins. It is demonstrated that both putative endophenotypes show association with the dopamine D4 receptor gene and meet most (but not all) of the criteria proposed for their validity and utility.