Serum hormones have been intensively investigated in association with several chronic diseases, but limited information exists on the reliability of a number of hormone determinations. The one-year reproducibility of dehydroepiandrosterone sulfate (DHEAS), total and free testosterone, total estradiol, insulin, C-peptide, and prolactin was studied in 60 premenopausal and 47 postmenopausal women recruited in Varese province, Italy, 1991-1992. The hormonal determinations were made in blood samples collected twice, one year apart, after 12-h fast, in the same month, day, and hour and for premenopausal women on the same day of the luteal phase of the menstrual cycle. Samples from the first drawing were stored at -80 degrees C. Samples from both drawings were assayed simultaneously and in blind fashion. Total estradiol in postmenopause was not evaluated for limitation in the sensitivity of the laboratory method. The intraclass correlation coefficient in premenopausal women was 0.85 for DHEAS, 0.60 for total testosterone, 0.66 for free testosterone, 0.81 for insulin, 0.83 for C-peptide, 0.40 for prolactin, and 0.06 for total estradiol. In postmenopausal women, the coefficient was 0.90 for DHEAS, 0.88 for total testosterone, 0.71 for free testosterone, 0.67 for insulin, 0.73 for C-peptide, and 0.18 for prolactin. These data indicate that total estradiol measured during the luteal phase has a poor intraindividual reproducibility over time, and these findings may have important implications in studies of hormones in the etiology of chronic disease.