Bias in case-control studies.