The reproducibility of RIAs of circulating sex hormones has been evaluated as part of recent epidemiological investigations, but none seem to have addressed the reproducibility or validity of RIAs for urinary hormones or their metabolites. As part of a case-control study of breast cancer in Asian-American women, 12-h overnight urine samples were obtained, and a methodological study was conducted to identify laboratories capable of assaying urinary hormones. For the reproducibility component of this study, two laboratories with extensive experience in hormone assays measured urinary estrone, estradiol, estriol, pregnanediol glucuronide, and estrone glucuronide using samples from 15 women (5 midfollicular, 5 midluteal, and 5 postmenopausal). Variance estimates from these measurements were used to calculate the laboratory variability (coefficient of variation) and to assess the magnitude of the biological variability among the women in relation to the total variability (intraclass correlation coefficient). For the validity component, urinary estrone, estradiol, and estriol levels were measured in the same samples by gas chromatography-mass spectroscopy in the laboratory of Dr. Herman Adlercreutz (University of Helsinki, Helsinki, Finland). We found that the degree of assay reproducibility differed between the laboratories, but that laboratory variability was usually low compared with the range of hormone values among women, particularly for the estrogens. Values for estrone and estradiol were well correlated among all of the laboratories. For estriol, the RIAs tended to overestimate levels compared with gas chromatography-mass spectroscopy. In one laboratory, assays for pregnanediol glucuronide and estrone glucuronide were consistently reproduced; in the other, the reproducibility of the RIA for pregnanediol glucuronide was problematic, and estrone glucuronide was not measured. Despite some limitations, urinary hormones and their metabolites can be reliably measured by current RIAs in large investigations attempting to link hormone level to disease risk and may be particularly advantageous for studies of postmenopausal women, where serum concentrations of estrone and estradiol are low and assay measurements are not as dependable.