An enzyme linked immunoadsorbent assay for urinary albumin using commercially available reagents is described. The assay range is 2.5-120 micrograms/l. When samples are analysed in two standard dilutions, the assayable albumin concentration range is 2.5-240 mg/l, covering the clinical range from normoalbuminuria to overt clinical nephropathy. Intra-assay variation was 2.1% and interassay variation 8.3%. Recovery of added albumin to urine was 95%-106% and dilution of urine was linear. The correlation to urinary albumin determined by immunodiffusion was excellent (n = 80, r = 0.99). Intraindividual variation of the 24 h urine albumin excretion of different days was high in patients with incipient diabetic nephropathy (51.5%) and was only slightly reduced by taking the variation of creatinine excretion into account (39.5%). No correlation was found between albumin excretion, and HbA1c or urine glucose excretion, indicating that minor metabolic variations are not responsible for the huge intraindividual day-to-day variations of UalbV. The study shows that more than one UalbV measurement must be done before classifying patients into groups with or without incipient diabetic nephropathy.