Can we measure prior Postmenopausal Estrogen/Progestin use? The Postmenopausal Estrogen/Progestin Interventions Trial

The objective of this study was to assess: 1) the accuracy of a single self-report question about postmenopausal estrogen use; and 2) the performance and repeated measures agreement of a standardized hormone use interview. Women (n = 863) in the Postmenopausal Estrogen/Progestin Interventions Trial (PEPI) completed a self-report baseline questionnaire (BQ) at enrollment and a History of Hormone Use interview (HHU) at the 3-month follow-up visit (HHU3mos). A subsample of 101 women completed a second HHU interview 3 years later (HHU3yrs). As determined by the HHU3mos, 479 (56%) of women had ever used postmenopausal estrogen and 261 (30%) had ever used postmenopausal progestin. The mean number of years since last estrogen or progestin use was 2.2 and 1.3 years, respectively. Overall, there was 95% agreement between self-reported estrogen use on the BQ and the HHU3mos (kappa = 0.91). Using the HHU3mos as the criterion standard, the BQ misclassified 2.3% of women as false positives and 6.3% as false negatives. The average duration of estrogen use in the false-negative classifications was 1.9 years (range: 1-9 years). On the HHU3mos, 39.7% of participants could not recall at least one of the specific details of estrogen use (preparation, dose, route, or starting or stopping year); similar patterns of recall were found for progestin use. Factors associated with discordant reporting of ever-use of ERT (BQ vs. HHU3mos) or incomplete reporting of estrogen/progestin use on the HHU3mos were: route of administration, recency of hormone use, duration of hormone use, and race. Age, years since menopause, education, income, and hysterectomy status were not related to discordant and/or incomplete reporting. Agreement between the HHU3mos and HHU3yrs for ever-use of estrogen was 85.2% (kappa = 0.71). in sum, a single self-report question was adequate to ascertain ever-use of postmenopausal estrogen. When a structured interview form was used, details of postmenopausal estrogen and progestin use were not well remembered. Some features of hormone use and participant characteristics were associated with completeness of recalled hormone use, which suggests the potential for differential misclassification.