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

With the use of data from the Nurses' Health Study, the authors confirm the recent argument by Pike et al. (Am J Epidemiol 1998;147:718-21) that various ways of assigning ages at menopause to women with simple hysterectomy produce relative risks for the effect of hormone replacement therapy (HRT) that are biased downward. Different methods of including women with simple hysterectomy consistently produced relative risks for the association between a 1-year increase in duration of HRT use and breast cancer that were underestimates compared with the relative risk obtained when these women were excluded from analysis. The authors used longitudinal data on recalled age at menopause to demonstrate another source of error related to age at menopause: error in recall of age at menopause that increases with time since menopause.