We assessed the sample size required for detecting gene-environment (G
x E) interactions in a case-control study of complex diseases. The re
sults suggest that large numbers of cases and controls will be require
d to detect G x E interaction for some odds ratio and exposure frequen
cy combinations. These and other results suggest that alternative stud
y designs may be needed to detect G x E interaction particularly with
rare genes or uncommon environmental exposures. (C) 1997 Wiley-Liss, I
nc.