With a growing number of disease- and trait-associated genetic variants detected and replicated across genome-wide association studies (GWAS), scientists are increasingly noting the influence of individual genetic variants on multiple seemingly unrelated traits. Cross-phenotype association tests, applied on two or more traits, usually test the null hypothesis of no association of a genetic variant with any trait. Rejection of the null can be due to association between the genetic variant and a single trait, with no indication if the variant influences >1 trait. In this talk, I will present a new statistical approach to discover genetic variants shared between two phenotypic groups using GWAS summary statistics. It can be applied to studies with overlapping samples. We compute an approximate asymptotic p-value of association with both outcomes, and is computationally efficient to be implemented genome-wide. Our simulations show that this test maintains appropriate type I error at genome-wide error levels even when moderate sample size or effect size differences in the two groups exist. I will show an application to GWAS of orofacial clefts, which are typically categorized as cleft lip with or without cleft palate (CL/P) and cleft palate alone (CP) based on epidemiologic and embryologic patterns. Historically, most studies of CL/P and CP have been analyzed separately. Our approach of jointly analyzing the two cleft subtypes identified 8 loci in or near candidate genes, including the well-recognized FOXE1 gene, that appear to influence risk of both cleft subtypes.