No change of breast cancer mortality has been reported previously after long-term hormone replacement therapy. A conceivable explanation for the apparent discrepancy between incidence and mortality may be selection bias due to lower prevalence of breast cancer in women who receive replacement hormones, compared with nonexposed women. We used a new approach to correct for bias due to this 'healthy drug-user effect,' by adjusting the external, population-based, mortality rates for such cases prevalent during the recruitment period of our cohort. In this cohort of some 23,000 Swedish women, who were prescribed various hormone replacement regimens, breast cancer mortality was analyzed after follow-up to 12 years. External analyses revealed overall standardized mortality ratios for breast cancer rising from 0.71 to 0.81, but not significantly different from unity, after adjustment procedures. In multivariate regression models, excluding prevalent cases in the cohort, women prescribed estradiol, conjugated estrogens, or an estrogen-progestin combination were not at a higher risk relative to those given other and weak estrogens, relative risks being 0.81 and 0.68, respectively. On the basis of the present analytical approach, we conclude that breast cancer mortality does not appear to be changed overall or in subgroups, despite increased incidence.