Variability in the performance of nuclear matrix protein 22 for the detection of bladder cancer

PURPOSE: We assessed variability in the diagnostic performance of NMP22 for detecting recurrence and progression in patients with Ta, T1, and/or CIS transitional cell carcinoma of the bladder in a large international cohort. MATERIALS AND METHODS: NMP22 voided urine levels were measured in 2,871 patients who underwent office cystoscopy for monitoring previous stage Ta, T1 and/or CIS transitional cell carcinoma at 12 participating institutions. RESULTS: Patient characteristics varied considerably among institutions. Overall 1,045 patients (36.4%) had recurrent transitional cell carcinoma (range across institutions 13.6% to 54.3%). Median NMP22 was 5.5 U/ml (range across institutions 2.5 to 18.8). Of the patients 33.5% had grade III tumors (range across institutions 20.6% to 54.0%) and 22.4% had muscle invasive tumors (range across institutions 3.2% to 38.2%). Area under the ROC curve for bladder TCC detection was 0.735 (95% CI 0.715 to 0.755, range across institutions 0.676 to 0.889). The manufacturer recommended cutoff of 10 U/ml detected 57% of cases with a 19% false-positive rate. AUC for grade III and stage T2 or greater disease was 0.806 (95% CI 0.780 to 831) and 0.864 (95% CI 0.839 to 0.890), respectively. For each NMP22 cutoff NMP22 had higher sensitivity for detecting grade III and stage T2 or greater bladder transitional cell carcinoma than for detecting any cancer. No optimal cutoffs for detecting any or aggressive bladder transitional cell carcinoma could be derived based on NMP22 values. CONCLUSIONS: There is a substantial degree of heterogeneity in the diagnostic performance of NMP22 applied to populations from different institutions. There is no clearly defined NMP22 cutoff but there is a continuum of risk for recurrence and progression.