Potential for bias in case-control studies of oral contraceptives and breast cancer.