Age at natural menopause and breast cancer risk: the effect of errors in recall

Many case-control studies of breast cancer have reported estimates of relative risk of approximately 1.9 for menopause after 54 years of age when compared with menopause before 45 years of age. These estimates may be biased towards unity because of errors in recall of a woman's age at menopause. This paper investigates the magnitude of error associated with self-classification of women into categories of age at menopause by using personal interview and medical record data of 67 control women from a case-control study of breast cancer conducted in two retirement communities near Los Angeles in 1977-1978. These estimates of "misclassification" error are combined with various sets of "true" relative risks to yield "observable" relative risks. It is demonstrated that the relative risks associated with various age at menopause groups are likely to range from 1-3.4 rather than 1-1.9 as has been reported in the literature.