|
Abstract:
|
In planning any type of study, a crucial phase is determining the required sample size to detect the association of interest. In the context of genetic association testing, research on sample size estimation is focused on detecting gene-environment interactions, which are more difficult to detect. An implicit assumption in sample size calculations for case-control association designs is that the case and control exposures are independent. If their environmental exposures are correlated, the distribution of the environment and gene-environment covariates will be affected, requiring an adjustment to the overall sample size to maintain power. We have developed methodology and software that allows for environmental correlation. Its relevance is shown in a genetic study of vitamin D and prostate cancer, where matching subjects by time of blood draw induces an correlation in vitamin D levels.
|
