OBJECTIVE We compare the individual and combined sensitivity of urinary CYFRA 21-1, urinary bladder cancer antigen, tissue polypeptide antigen and NMP22 to detect bladder cancer, evaluate the false-positive rates for different pathological conditions, and assess differential sensitivity regarding histological and clinical characteristics of disease. METHODS A total of 267 subjects entered the study. Sensitivities of the tests were evaluated in 111 patients with active bladder cancer and 76 with no evidence of disease. False-positive rates were evaluated in 80 symptomatic and asymptomatic controls, including patients with benign urological conditions and nonbladder malignancies, and healthy subjects. CYFRA 21-1 was determined by electrochemoluminescent immunoassay in the Elecsys 2010, urinary bladder cancer antigen was quantified by enzyme linked immunosorbent assay (IDL Biotech), tissue polypeptide antigen was measured by the Prolifigen TPA-IRMA and NMP22 was assayed by enzyme linked immunosorbent assay (Matritech). Cutoffs were obtained by the 95% percentile in patients with no evidence of disease, which gave a 95% specificity for all biomarkers. Differences in sensitivity of urinary biomarkers regarding stage, grade, tumor size, pattern of growth, focality and recurrence were evaluated. RESULTS At a specificity of 95% cutoffs were 5.4 ng./ml. for CYFRA 21-1, 15.5 microg./l. for urinary bladder cancer antigen, 760.8 U./l. for tissue polypeptide antigen and 14.6 U./ml. for NMP22. Using these cutoffs sensitivities were 75.7% for NMP22, 83.8% for CYFRA 21-1, 73.9% for urinary bladder cancer antigen quantitative and 80.2% for tissue polypeptide antigen. The additional determination of cytokeratins increased the sensitivity of NMP22. Cytokeratins did not appear to be specific for bladder cancer, and false-positives rates were between 20% for urinary bladder cancer antigen and 36% for tissue polypeptide antigen for benign urological conditions, and between 40% and 52%, respectively, for nonbladder malignancies. NMP22 showed lower false-positives rates, mainly for benign diseases. Urinary tumor markers appeared to be associated with some of the most relevant histological and clinical parameters of bladder cancer. CONCLUSIONS Our preliminary evaluation showed the tests to be potential noninvasive adjuncts to help determine the need for cystoscopy. The combination of 2 tumor markers, NMP22 and 1 cytokeratin (CYFRA 21-1 or urinary bladder cancer antigen), seemed to be the most effective. Further comparative studies are needed to assess the promising diagnostic role of these markers.