Categorical traits such as cases-control status are often used as response variables in association studies of genetic factors associated with complex diseases. Using categorical variables to summarize likely continuous disease liability may lead to loss of information, thus reduction of power to recover associated genetic factors. On the other hand, a direct study of disease liability is often infeasible because it is an unobservable latent variable. In some diseases, the underlying disease liability is manifested by several phenotypes, and thus the associated genetic factors may be identified by combining the information of multiple phenotypes. Since case-control studies often collect information on secondary phenotypes, the appropriate use of the information of secondary phenotypes may enhance the statistical powers to identify the associated genetic factors with the primary trait. We propose a novel method to address this challenge. Simulation results demonstrate good performance of the methods. The real data analysis of the Alzheimer's disease illustrates the practical utility of the techniques.