The low power of gene-environment (G {604} E) interaction studies is of major concern in genetic-epidemiolgic research. Past research involving binary outcomes has focussed mainly on the development of efficient study designs to address this problem. This thesis explores an alternative strategy that uses quantitative 'surrogates' of the 'clinical' binary outcome to improve power to detect G {604} E interactions. Efficiency of the quantitative 'surrogate' outcome X versus the binary outcome Y is assessed for three hypothetical models of the relationship between the outcomes, and their relationships to genetic susceptibility, exposure, and other risk factors. In the first scenario, X is a risk factor of disease, and a mediator for the effect of G {604} E interaction. In the second scenario, X is considered a marker of disease outcome. Finally, repeated measures of the disease marker X are used to define alternative binary and quantitative outcomes. Simulations are used to estimate the power to detect G {604} E interaction in models using these alternative outcomes