In this study, we validated measurements of free testosterone (fT) and free estradiol (fE(2)) concentrations calculated from total serum concentrations of testosterone (T), estradiol (E(2)), and sex hormone-binding globulin (SHBG), measured by direct, commercial radioimmunoassays, by comparison with reference measurements obtained by dialysis plus an in-house radioimmunoassay after extraction and chromatographic purification. The study was conducted in serum samples from 19 postmenopausal women who were part of an ongoing prospective cohort study. We also performed sensitivity analyses to examine the robustness of the theoretical calculations. Sensitivity analyses showed that in this population, competitive binding of dihydrotestosterone and total T could be ignored in the calculation of fE(2), and competitive binding by dihydrotestosterone does not need to be taken into account for calculation of fT. Furthermore, variations in albumin and SHBG concentrations had negligible effects on fT and fE(2) calculations. Values of fT and fE(2), calculated from total T and E(2) concentrations obtained by the same in-house radioimmunoassay used for the dialysis method, correlated highly with the measurements by dialysis (Pearson's coefficients of correlation above 0.97). When calculating fT and fE(2) using total T and total E(2) concentrations obtained by different direct radioimmunoassays, almost all kits gave good correlations with the reference method for fT (Pearson's r > 0.83), but only a few gave good correlations for fE(2) (Diagnostic System Laboratories and DiaSorin; r > 0.80). The direct radioimmunoassays giving the best correlation for fT and fE(2) with the dialysis method were those that best measured total concentrations of T and E(2). Furthermore, mean values of fT and fE(2) corresponded well to mean values by the reference method if SHBG measurements were also well calibrated. We conclude that in postmenopausal women, theoretical calculations are valid for the determination of fT and fE(2) concentrations and can give reliable estimation of cancer risk in epidemiological studies when the total concentrations of T, E(2), and SHBG are measured accurately.