Bias in breast cancer analyses due to error in age at menopause.