The role of the geometric mean in case-control studies