Sex steroid concentrations in urine samples from post-menopausal women have been associated with risk of various chronic diseases. The basic requirement for the assessment of risk in such large-scale epidemiological studies is that subjects be ranked accurately by their average, long-term hormone levels. We examined the reproducibility over time of measurements of urinary testosterone (T), 5alpha-androstane-3alpha, 17beta-diol (ADIOL), estrone (E1), estradiol (E2), 2-hydroxy estrone and 2-hydroxy estradiol, (2(OH)-E), 16alpha-hydroxyestrone (16alpha(OH)-E1) and the ratio of 2(OH)-E and 16alpha(OH)-E1, in a representative sub-sample of post-menopausal women (n = 43) participating in an ongoing prospective cohort study. Women collected three first morning urine voids on different occasions, with average time difference between the first and the third urine sample of 5.1 years. T, ADIOL, E1 and E2 were measured by radio immunoassay after enzymatic hydrolysis, solid-phase extraction and HPLC purification of the samples, while 2(OH)-E and 16alpha(OH)-E1 were assayed by solid-phase enzyme immunoassay after enzymatic hydrolysis. Intra-class correlation co-efficients (ICCs) over time were very good for T (r = 0.85), acceptable for E2, E1 and ADIOL (r > 0.55), but low for 2(OH)-E, 16alpha(OH)-E1 and their ratio (r < 0.46). The adjustment for creatinine concentrations did not increase these correlations.