Comment on "Ascertainment adjustment in complex diseases".

Glidden and Liang [2002] have raised important issues regarding ascertainment adjustment in the framework of variance-components modeling for complex genetic traits. While the structure of the authors’ logistic variance-component model is simple, ascertainment issues arising with this model are likely analogous to ascertainment issues in more complex variance-component models commonly used for the analysis of either genetic disease data [Duggirala et al., 1997; Burton et al., 1999] or quantitative trait data [Amos, 1994; Almasy and Blangero, 1998]. Therefore, the results of Glidden and Liang [2002] are of importance both to investigators who design gene mapping studies, and to analysts who use variancecomponent methods to study genetic trait data. The authors first demonstrate that, if the ascertainment scheme is correctly modeled, ascertainment-adjusted parameter estimates from their logistic variancecomponent model for analyzing disease data reflect the true values of the populationbased parameter values rather than the sample-based parameter values. These results are analogous to those of Epstein et al. [2002], who used a similar logistic variancecomponent model initially proposed by Burton et al. [2000]. de Andrade and Amos [2000] showed similar results for the traditional linear variance-component method that assumed major gene, polygene, and environmental effects. In their example, de Andrade and Amos [2000] selected families in which one sibling had a trait value more extreme than 90% of the population, and properly accounted for ascertainment by dividing the unconditional likelihood by the likelihood that the selected