Existing statistical methods for complex diseases with multivariate phenotypes can be categorized into three types of approaches. The first approach is to conduct a GWAS for each marginal phenotype and then aggregate the results. The major issue with this approach is that it does not make use of the correlation structure among phenotypes. The second approach is to summarize multiple phenotypic traits into a composite score and then conduct a GWAS on the score. This approach, however, may have difficulty in identifying proper summary scores. The third approach involves multiple phenotypic traits simultaneously. Thus, it may gain power as well as avoid the issue of multiple testing. However, it is based on stronger assumptions that may not be satisfied in some practical settings.
