Because of increased interest in the assay of albumin in urine and the sensitivity required to quantify concentrations associated with (a) increased risk of developing end-stage renal disease and cardiovascular disease among people with diabetes and (b) renal damage caused by exposure to nephrotoxic substances, we conducted a pilot study of the variation of these measurements within and among five laboratories that use various immunoassays. These assays included two different enzyme immunoassays, two different immunoturbidimetric assays, a fluorescent immunoassay, and a zone immunoelectrophoresis assay. The results indicate considerable variation both within and among laboratories for measurements at or near the normal range. Variability is equally attributable to the precision of individual immunoassays and to the variation of the mean values obtained by each laboratory. Individual laboratory CVs ranged from 5.8% to 18.2% for mid- and high-concentration samples treated with preservative and from 8.4% to 23.6% for mid- and high-concentration samples containing no preservative. The relative bias of individual laboratory means ranged from -56.4% to 20.5% for the two preserved materials and from -32.6% to 0.8% for the two materials containing no preservative. To reduce the chance of misdiagnosing the risk associated with above-normal albumin concentrations in urine, we need to address the problems contributing to imprecision and inaccuracy, particularly laboratory-to-laboratory variability.