A comparison of urinary nuclear matrix protein‐22 and bladder tumour antigen tests with voided urinary cytology in detecting and following bladder cancer:the prognostic value of false‐positive results

Objectives To evaluate the diagnostic and prognostic value of the nuclear matrix protein‐22 (NMP22) and bladder tumour antigen (BTAstat) tests compared with voided urinary cytology (VUC) in detecting and following bladder cancer, assessing particularly the prognostic value of false‐positive test results in patients followed up for bladder cancer. Patients and methods From 739 patients suspected of having bladder cancer, voided urine samples for the NMP22 and BTAstat tests, and for VUC and urine analysis, were collected before cystoscopy. All patients underwent transurethral resection of bladder lesions or mapping, and were followed for a mean (range) of 27.3 (3–65) months. Results In the 406 patients with bladder cancer, the overall sensitivity was 85% for NMP22, 70% for BTAstat and 62% for VUC. For histological grades 1–3 the sensitivity in detecting transitional cell carcinoma was 82%, 89% and 94% for NMP22, 53%, 76% and 90% for BTAstat, and 38%, 68% and 90% for VUC, respectively. Although the sensitivity in detecting invasive carcinoma was > 85% for all the tests, NMP22 and BTAstat were statistically more sensitive than VUC for superficial tumours. The optimal threshold value for NMP22, calculated using the receiver operating characteristics curve, was 8.25 U/mL. The specificity was 68% for NMP22, 67% for BTAstat, and 96% for VUC. The specificity of VUC remained > 87% and was independent of benign histological findings. In contrast, in patients with no apparent genitourinary disease on histology, NMP22 and BTAstat had significantly higher specificity (94% and 92%, respectively; P = 0.003) than in the group with chronic cystitis (52% for both tests). Forty patients having no bladder cancer at biopsy had a recurrence after a mean (range) follow‐up of 7.7 (3–15) months; all had a previous history of bladder cancer. According to subsequent recurrence, the prognostic positive and negative predictive values were 18% and 91% for NMP22, 13% and 88% for BTAstat, and 79% and 91% for VUC. Both false‐positive VUC and NMP22 tests predicted recurrence (log‐rank test, P