A special feature of gene-environment interaction studies is that in many situations it is scientifically plausible to assume that the genetic and environmental factors are independent in the underlying population.  Under this additional constraint of gene-environment independence, one can derive more efficient estimation techniques than the traditional prospective logistic regression analysis (Piegorsch et al, 1994; Chatterjee and Carroll, 2005).  However, the efficient estimates from the retrospective likelihood may be biased under the violation of the independence assumption. We propose an alternative to relax the constraint of gene-environment independence in an empirical Bayes framework to strike a compromise between efficiency and robustness. We evaluate the new testing strategies in terms of study design, power and Type I error properties. This work is joint with Nilanjan Chatterjee (National Cancer Institute).