Mixed-effects Cox models of alcohol dependence in extended families.

The presence of disease is commonly used in genetic studies; however, the time to onset often provides additional information. To apply the popular Cox model for such data, it is desirable to consider the familial correlation, which involves kinship or identity by descent (IBD) information between family members. Recently, such a framework has been developed and implemented in a UNIX-based S-PLUS package called kinship, extending the Cox model with mixed effects and familial relationship. The model is of great potential in joint analysis of family data with genetic and environmental factors. We apply this framework to data from the Collaborative Study on the Genetics of Alcoholism data as part of Genetic Analysis Workshop 14. We use the S-PLUS package, ported into the R environment http://www.r-project.org, for the analysis of microsatellite data on chromosomes 4 and 7. In these analyses, IBD information at those markers is used in addition to the basic Cox model with mixed effects, which provides estimates of the relative contribution of specific genetic markers. D4S1645 had the largest variance and contribution to the log-likelihood on chromosome 4, but the significance of this finding requires further investigation.