Detection of gene-environment interaction by case-only studies

Background: The detection of gene-environment interaction can provide important clues not only for resolving biological mechanisms underlying diseases, but also for disease prevention. The newly introduced case-only study was compared with traditional case-control study in terms of statistical power to detect significant gene-environment interaction. Methods: Odds ratios for interaction were calculated in the framework of case-control study and case-only study separately, by an unconditonal logistic model. Hypothetical data with 200 cases and 200 or 400 controls and real published data derived from four cancer case-control studies of genotype and smoking were used for the comparisons. Results: Although odds ratio estimates for interaction were the same, 95% confidence intervals were narrower in case-only studies than in case-control studies. Similarly there were no substantial differences in point estimates for interaction in four real cancer case-control studies between the two study designs, but the confidence intervals were narrower with the case-only study. Conclusions: Although the case-only study does not provide odds ratios for exposure or genotype alone, it is very useful for the detection of interaction, especially for screening purposes.