Prenatal exposures such as polycyclic aromatic hydrocarbons and early postnatal environmental exposures are of particular concern because of the heightened susceptibility of the fetus and infant to diverse environmental pollutants. Marked inter-individual variation in response to the same level of exposure was observed in both mothers and their newborns, indicating that susceptibility might be due to genetic factors. With the mother-child pair design, existing methods developed for parent-child trio data or random sample data are either not applicable or not designed to optimally use the information. To take full advantage of this unique design, which provides partial information on genetic transmission and has both maternal and newborn outcome status collected, we developed a likelihood-based method that uses both the maternal and the newborn information together and jointly models gene-environment interactions on maternal and newborn outcomes. Through intensive simulation studies, the proposed method has demonstrated much improved power in detecting gene-environment interactions. The application on a real mother-child pair data from a study conducted in Krakow, Poland, suggested four significant gene-environment interactions after multiple comparisons adjustment.