Background: Previous studies have demonstrated that mammographic breast density increases following the initiation of estrogen replacement therapy (ERT). The effect, if any, that this increase in density has on the specificity (related to false-positive readings) and the sensitivity (related to false-negative readings) of screening mammography is unknown.

Purpose: Using a retrospective cohort study design, we assessed the effects of ERT on the specificity and the sensitivity of screening mammography.

Methods: Participants (n = 8779) were postmenopausal women, aged 50 years or older, who were enrolled in a health maintenance organization located in western Washington state and who entered a breast cancer screening program between January 1988 and June 1993. Two-view mammography was performed as part of a comprehensive breast cancer screening visit. Menopausal status, as well as demographic and risk-factor information, was recorded via self-administered questionnaires. Hormonal replacement therapy type and use were determined from questionnaire data and from an automated review of pharmacy records. Individuals diagnosed with breast cancer within 12 months of their first screening-program mammograms were identified through use of a regional cancer registry. Risk ratios (RRs) plus 95% confidence intervals (CIs) of false-positive as well as false-negative examinations among current and former ERT users (with never users as the reference group) were calculated. Reported P values are two-sided.

Results: The specificity of mammographic screening was lower for current users of ERT than for never users or former users. Defining a positive mammographic reading as any non-normal reading (either suspicious for cancer or indeterminate), the adjusted RR (95% CI) of a false-positive reading for current users versus never users was 1.33 (1.15-1.54) (P < .001); for former users versus never users, the RR (95% CI) was 1.00 (0.87-1.15). The adjusted mammographic specificities (95% CIs) for never users, former users, and current users of ERT were 86% (84%-88%), 86% (84%-87%), and 82% (80%-84%), respectively. Defining a positive reading more rigorously (i.e., as suspicious for cancer only), the adjusted RRs (95% CIs) of false-positive readings for current users and former users (versus never users) were 1.71 (1.37-2.14) (P < .001) and 1.16 (0.93-1.45), respectively. Sensitivity was also lower in women currently receiving ERT. The unadjusted RR (95% CI) of a false-negative reading for current users versus never users was 5.23 (1.09-25.02) (P = .04); for former users versus never users, the RR (95% CI) was 1.06 (0.10-10.87). The unadjusted mammographic sensitivities (95% CI) for never users, former users, and current users of ERT were 94% (80%-99%), 94% (69%-99%), and 69% (38%-91%), respectively.

Conclusions and implications: Current use of ERT is associated with lower specificity and lower sensitivity of screening mammography. Lower specificity could increase the cost of breast cancer screening, and lower sensitivity may decrease its effectiveness.