Measurements of total and free testosterone levels in women have lacked precision and accuracy because of limited assay sensitivity. The paucity of normative data on total and free testosterone levels in healthy women has confounded interpretation of androgen levels in women with human immunodeficiency virus (HIV) infection and other disease states. Therefore, the objectives of this study were to develop sensitive assays for the measurement of the low total and free testosterone levels in women to define the range for these hormones during the normal menstrual cycle and assess the total and free testosterone levels in HIV-infected women. By using a larger volume of serum, increasing the incubation time, and reducing the antibody concentration, the sensitivity of the total testosterone assay was increased to 0.008 nmol/L, and that of the free testosterone assay was increased to 2 pmol/L. The mean percent free testosterone was 1.0 +/- 0.1% of the total testosterone. Serum total and free testosterone levels in the follicular and luteal phases were not significantly different, but both demonstrated a modest preovulatory increase, 3 days before the LH peak. Serum total [0.50 +/- 0.32 (14.60 +/- 9.22) vs. 1.2 +/- 0.7 nmol/L (34.3 +/- 21.0 ng/dL); P < 0.0001] and free testosterone levels (5.56 +/- 2.70 (1.58 +/- 0.80) vs. 12.8 +/- 5.5 pmol/L (3.4 +/- 1.7 pg/mL); P < 0.0001) were significantly lower in HIV-infected women (n = 37) than in healthy women (n = 34). Serum total and free testosterone levels were also significantly lower in HIV-infected women who were menstruating normally. There were no significant differences in serum total and free testosterone levels between those who had lost weight and those who had not. Testosterone levels correlated inversely with plasma HIV ribonucleic acid copy number. Serum FSH, but not LH, levels were significantly higher in HIV-infected women than in controls. Using assays with sufficient sensitivity, we defined the range for total and free testosterone levels during the normal menstrual cycle. Serum total and free testosterone levels are lower in HIV-infected women and correlate inversely with plasma HIV ribonucleic acid levels. The hypothesis that androgen deficiency contributes to wasting in HIV-infected women remains to be tested.