We then use the multivariate JLS method to identify and replicate an association between the apical gene set, and the SLC9A3 complex within the set, and lung disease severity in CF; this finding would have been missed by the traditional multivariate gene-set tests that do not account for (heterogeneous) GxG and GxE interactions. Through extensive simulation analyses, we demonstrate that the proposed JLS method has good type 1 error control with improved power compared to other testing options, and, in contrast to the recently proposed distribution16 and LRT joint testing methods,17 can be easily implemented in the context of gene-set and pathway analyses.