Theoretically, a multiple phenotype approach always reduces the number of tests that are conducted, and also increases power by exploiting correlation across phenotypes. In our study, we discovered that the multi-phenotype approach leads to elevated ORs in Table 4, aiding with risk stratification. In general, if there is correlation among the pollutants, the discovery approach based on conditional associations may yield new results. If there is correlation among the outcomes or different phenotypes, the multi-phenotype approach, in spite of being a test with higher degrees of freedom, will yield a more powerful analysis. For example, six new pollutants were discovered with the multi-phenotype approach in our case study.
