A new technique is demonstrated for the determination of urine albumin combining the commercially available microalbuminuria test [Tina-Quant Albumin in Urine, Roche Diagnostics, Germany] with neural network analysis of reaction kinetic data. We analyzed altogether 102 patient urine samples [27 diabetes patients, 21 with nephrosis or nephritis, 54 with hypertension]. At the diabetes group in 1 of the 27, at the nephrosis/nephritis group 6 in the 21 and at the hypertension cases in 4 of the 54 cases, false negative results, caused by high dose hook effect was detected. Using the Keysys analyzer system (Roche), dilution series were prepared from human albumin in the concentration range from 0 to 40 000 mg/l. The supplier defined measuring range of the assay is up to 400 mg/l, the maximum of the Heidelberger curve is at 1,500 mg/l. A computer program was developed for the off-line downloading of the measured reaction kinetics. The back propagation neural network was able to classify the kinetics into two categories: without or with prozone effect. Using this method routine urine albumin samples can be evaluated safely for the theoretical concentration range.