All 50 patients had a history of recurrent non-muscle-invasive bladder tumours and were undergoing outpatient surveillance cystoscopy, including 18 after BCG therapy. A suspected lesion or abnormal area of mucosa was seen on WLI. The bladder was then evaluated by NBI cystoscopy. Each lesion(s) was biopsied and resected for histological evaluation. In all, 26 patients had a tumour and 24 had benign histology.
Table 1 shows individual surgeon variability in NBI findings to identify recurrent bladder tumours. H.H., M.D., and G.D. detected each patient with recurrent tumour, definitely identifying lesions as malignant or suspicious enough to warrant resection. J.T. missed three cases with cancer, because she did not consider the lesions in question significant to warrant a biopsy. G.D. and M.D. assessed six cases and seven cases, respectively, having malignant disease as benign on WLI, but changed their interpretations to cancer based on NBI cystoscopy. For H.H., M.D., G.D., and J.T., respectively, the sensitivity of detecting histological tumour using NBI was 92%, 77%, 77%, 69%; the specificity was 79%, 58%, 79%, 67%; positive predictive value was 83%, 67%, 80%, 69%; and the negative predictive value was 90%, 76%, 70%, 67%. M.D. had a lower specificity because she considered more lesions positive for tumour than other urologists, leading to biopsy of benign lesions. J.T. performed ≈10% worse than the experienced urologists, but overall, there was no significant differences between expert urologists using NBI (P = 0.4), or between experienced urologists and novice (P = 0.1) by Pearson’s chi-square test or analysis of variance (two-sided).