Agreement between interview information and physician records on history of menopausal estrogen use

Abstract

A case-control study of the association between replacement estrogen use and breast cancer risk was conducted between 1975 and 1980 on Oahu, Hawaii. Data from this study were used to compare menopausal estrogen histories obtained through a personal interview with information from the subject's physician(s) or clinical records. The sample included 344 breast cancer cases, 344 hospital controls, and 344 neighborhood controls. Study participants included Japanese and white women aged 45-74 years who were residents of Oahu. Interviewers asked participants whether they had ever used replacement or menopausal estrogens for 1 month or longer. The month and year of initial use and the duration of use for each preparation were recorded for the users. Questions were also asked about a number of other medications. Information on estrogen use from the users and nonusers was then verified by the physician(s) or clinic(s) named by the study subject. The results showed moderate to substantial agreement between users and physicians on ever/never use of estrogens (kappa = 0.74), duration of estrogen use (intraclass correlation coefficient (rI) = 0.54), and age at initial use of estrogens (rI = 0.57). There was no differential misclassification by the case-control status of the subject. Agreement tended to be better for Japanese subjects, younger subjects, nonsmokers, and those of higher socioeconomic status, as measured by two indicators, college education and home ownership. Agreement was also negatively influenced by the duration of estrogen use and length of recall. These results suggest that women can recall estrogen use with a high degree of accuracy and support the use of a personal interview for obtaining information on replacement estrogens in case-control studies.