Our results indicate that differences in the criteria used to define menopausal status affect how women are classified as premenopausal or postmenopausal and the distribution of certain characteristics, such as age, within menopausal status groups. These differences did not translate into marked differences in menopausal status-specific breast cancer incidence or detection rates. Within the age range of this analysis, there was little difference between premenopausal versus postmenopausal rates for breast cancer incidence and detection when menopausal status was based solely on whether a woman's menstrual periods had stopped, suggesting that additional information may be necessary to properly discriminate between premenopausal and postmenopausal women.
In a previous study, Morabia and Flandre assessed the overlap and effects of discrepancies between definitions of postmenopausal status based solely on attained age (i.e., age >45, >50, or >55 years) or solely on time since cessation of menses (i.e., >3 months, >12 months, >24 months, or >10 years) [15
]. In that study it was reported that menopausal status-specific associations between nulliparity, age at first birth, and breast cancer risk were not appreciably altered by the choice of definition for menopausal status. The authors of that analysis found that, in the absence menstrual history information, using a cutoff age of 50 to distinguish premenopausal from postmenopausal women offered the highest specificity for the lowest false positivity compared to menstrual history-based definitions. However, the fact that most studies report a later age at menopause among breast cancer cases than controls means that misclassification bias due to the use of an age-based proxy for menopausal status will be differential by case status.
Our results suggest that stratification by age as a proxy for menopausal status offers modest overlap with definitions using comprehensive self-reported epidemiologic data. While comparison to age-based definitions is dependent on the age structure of the study population, 25% of women aged 50–54 were classified as premenopausal under the complex definition used by the BCSC, and 10% of women aged 40–49 were classified as postmenopausal. By comparison, the Study of Women's Health Across the Nation (SWAN) reported that the median age at natural menopause was 51.4 years, where menopausal status was classified using more detailed information on the duration and reasons for cessation of menses [16
]. Discordance between definitions of menopausal status noted in this analysis (and the impact of such discordance), however, is likely to vary between populations according to the distribution of factors associated with menopausal status and age at menopause (e.g., age, race/ethnicity). Additionally, small to moderate shifts of women between menopausal status groups may have less impact when the outcome of interest is a rare event (e.g., breast cancer), than might be observed with a more common outcome. Thus, studies with different population structures or outcomes may be more impacted by the choice of definition for menopausal status. We also cannot rule out the possibility that similarities in breast cancer rates across menopausal status groups are the result of extensive misclassification across groups. Furthermore, although it was beyond the scope of this analysis, residual confounding due to misclassification of menopausal status is a practical concern and the use of different definitions for menopausal status could contribute to differences in effect estimates adjusted for menopausal status.
Certain caveats must be considered when interpreting these findings. We had no gold-standard for determining menopausal status. While it would be preferable to collect prospective data to more accurately stage menopausal status and to distinguish women in various stages of the menopausal transition using clinical criteria (e.g., the STRAW guidelines [9
]), the limited scope of our data is consistent with the situation faced by most epidemiologic studies, where menopausal status must be classified based on limited cross-sectional or retrospective self-reported data. Thus, the results of these analyses cannot speak to the validity of different definitions for menopausal status, but do address the impact and trade-offs of using differing levels of detail in menopausal status definitions. These trade-offs are important for studies to consider when deciding how to ascertain menopausal status. Depending on the purpose of the study, a simplistic definition may be sufficient, especially if menopausal status is not a main effect or key covariate, and could save time and resources from collecting more detailed data. Collecting cross-sectional or retrospective information on the duration and variability of menstrual cycles (in addition to the time since last period) and history of vasomotor symptoms may be useful for drawing comparisons to STRAW stages and more finely categorizing stages of menopausal transition. However, such data collection assumes adequate recall of potentially complicated menstrual histories and does not capture the experience of women with surgical menopause or the variability between women in symptomology of the menopausal transition.
While including information on the reason for cessation of menses and type of surgical menopause in the definition for postmenopausal status did not impact breast cancer rates for postmenopausal women, such information does allow for the identification of a subgroup of women who have a distinct risk factor profile: women whose menstrual periods have stopped but who may not be truly postmenopausal (i.e., women with surgical / other amenorrhea). The observation that breast cancer incidence and detection rates were lower in this subgroup of women presents some rationale for distinguishing this subgroup from postmenopausal and premenopausal women. Similarly, breast cancer rates in women classified as perimenopausal under the complex definition were distinct from those for women classified as postmenopausal and premenopausal, suggesting some utility in separately classifying perimenopausal women. However, the usefulness of distinguishing perimenopausal and surgical menopause groups must be weighed against the practicality of collecting the information necessary to make such distinctions, and the potential for bias in that information.
Menopausal status is a key main effect, covariate, and/or stratification factor in many epidemiologic studies of breast cancer and other diseases. Still, the complexity of defining menopausal status contributes to the lack of a standardized definition for this factor in the epidemiologic literature. Differences between studies in how menopausal status is defined raise questions about the comparability of findings based on those classifications. Our results indicate that distinctions in how menopausal status is defined contribute to notable differences in terms of how women are classified, but may translate to only slight differences in menopausal status-specific breast cancer incidence and detection rates. However, since there is no standardized approach to classifying menopausal status in epidemiologic studies, differences in such classification must be considered when comparing results across studies.