OBJECTIVE: To assess the value of nuclear matrix protein-22 (NMP22), compared with urinary cytology, in predicting the recurrence of bladder cancer that is not transitional cell carcinoma (non-TCC). PATIENTS AND METHODS: We tested the sensitivity, specificity and the predictive accuracy of NMP22 in the context of non-TCC bladder cancer recurrence, and compared it to the performance of urinary cytology. The study group comprised 2687 patients with history of non-muscle-invasive bladder cancer from 10 centres across four continents. RESULTS: The mean patient age was 64.8 years and 75.4% were men; of all patients, 513 (19.1%) had positive urinary cytology, 906 (33.7%) had a positive NMP22 test (>or=10 units/mL) and 80 (3.0%) had non-TCC recurrence. Most of these, i.e. 60 (75%), were stage >or=T2. The sensitivity and specificity of urinary cytology were, respectively, 20.0% and 94.8%, vs 77.5% and 81.8% for NMP22 of >or=10 units/mL. The predictive accuracy of urinary cytology was 57.5%, vs 87.1% for NMP22 >or= 10 units/mL. A combined model that included dichotomized NMP22 and urinary cytology was 85.3% accurate. CONCLUSION: The ability of a NMP22 level of >or=10 units/mL to predict non-TCC recurrence was better than that of urinary cytology, suggesting that NMP22 might have a role in the surveillance of patients at risk of non-TCC recurrence.
