The determination of tumor markers in urine samples has been proposed as an effective diagnostic tool in bladder cancer. The aim of the present investigation was to validate in urine samples the assay of the CYFRA21.1 cytokeratin-related marker, the serum concentrations of which showed promising diagnostic utility in patients with bladder cancer. First-voided urine samples were collected from patients with different malignancies. CYFRA21.1 was assayed with a commercially available enzyme immunoassay (Boehringer Mannheim). Different centrifugation patterns, the use of different buffers and nonionic detergents, and pH variations were evaluated. We demonstrated that: (a) cells and cell debris contain a large amount of CYFRA21.1 and must be eliminated by centrifugation; (b) storage at -20 degrees C causes amorphous precipitate, which may aspecifically bind CYFRA21.1; (c) the latter behavior may be prevented by diluting fresh urine samples with phosphate buffer with nonionic detergent added; (d) pH variations within the range 4.9-8.2 do not significantly affect CYFRA21.1 assay results. Provided that samples are diluted with buffer containing nonionic detergent, the CYFRA21.1 assay showed good precision and accuracy characteristic in urine samples. We therefore propose a standard protocol for the collection of urine samples for CYFRA21.1 assay. In a preliminary clinical evaluation, CYFRA21.1 concentrations in 16 patients with primary bladder cancer were higher than in healthy subjects. In the urine collected in the follow-up of patients treated for bladder cancer, CYFRA21.1 tended to be higher in relapsed patients than in those without evidence of disease. These preliminary data induced us to extend the clinical trial to establish the actual role of this assay in routine use.