Background: We compare methodological approaches for evaluating gene-environment interaction using a planned study of pediatric leukemia as a practical example. We considered three design options: a full case-control study (Option I), a case-only study (Option II), and a partial case-control study (Option III), in which information on controls is limited to environmental exposure only.
Procedure: For each design option we determined its ability to measure the main effects of environmental factor E and genetic factor G, and the interaction between E and G. Using the leukemia study example, we calculated sample sizes required to detect and odds ratio (OR) of 2.0 for E alone, an OR of 10 for G alone and an interaction G x E of 3.
Results: Option I allows measuring both main effects and interaction, but requires a total sample size of 1,500 cases and 1,500 controls. Option II allows measuring only interaction, but requires just 121 cases. Option III allows calculating the main effect of E, and interaction, but not the main effect of G, and requires a total of 156 cases and 133 controls.
Conclusions: In this case, the partial case-control study (Option III) appears to be more efficient with respect to its ability to answer the research questions for the amount of resources required. The design options considered in this example are not limited to observational epidemiology and may be applicable in studies of pharmacogenomics, survivorship, and other areas of pediatric ALL research.