Commentary: Composite cognitive and functional measures for early stage Alzheimer's disease trials.

In this commentary I consider the issues raised in Schneider and Goldberg's review of composite cognitive and functional measures. I find much to agree with in their commentary and especially their concerns regarding satisfactory psychometric validation of composite measures. I endorse also their provision for analysis by cognitive domain, backed by the use of statistical methods for grouping test variables. The authors helpfully mention the possibility that treatment effects may be peculiar to specific domains of cognitive function. I develop this view and argue for exploratory studies of new therapeutic interventions to include broad assessments of the cognitive domains known to be compromised in early Alzheimer's disease. I suggest that the results of exploratory studies be used to help identify target domains for confirmatory studies. Finally, I note that computerized cognitive composite assessments have often been validated in the fashion that the authors recommend for composite measures.