Regressive threshold model for familial analysis of complex diseases with variable age of onset

Few models for segregation (or combined segregation-linkage) analysis have been developed to account for variable age of onset. The unified model (UM) can only take into account age at examination. In the logistic hazard model (LHM), Abel and Bonney ([1990] Genet. Epidemiol. 7:391-407) incorporated survival analysis concepts into the regressive logistic model of Bonney ([1986] Am. J. Med. Genet. 18:731-749), but interpretation of familial dependence parameters is difficult. In this article, we extended the regressive threshold model (RTM) proposed by Demenais ([1991] Am. J. Hum. Genet. 49:773-785) to account for a variable age of onset of complex diseases. This model assumes an underlying liability to disease and is more general than the original logistic formulation, since the phenotypes of each individual's antecedents can be adjusted for their own genotypes and covariate effects. The variation of risk with age can be expressed as a general step function, and variants of the model have been proposed by imposing different types of constraints among the time-dependent thresholds. The performances of the three models (UM, LHM, and RTM) were compared in the context of segregation analysis of family data generated with variable age of onset. All analysis models were robust with respect to false conclusion of a major gene, and the best results were obtained under RTM. The power to detect the major gene was higher under LHM than RTM, but the best fit of the estimated cumulative age-dependent penetrance with respect to the true value was obtained under RTM. This new model may thus prove helpful in contributing to identification of genes underlying complex diseases, since it can easily include linked marker loci and linkage disequilibrium. (C) 2002 Wiley-Liss, Inc.