Validation of the ezrin, CK20, and Ki-67 as potential predictive markers for BCG instillation therapy of non-muscle-invasive bladder cancer

Objective The aim of our study was to try to validate 3 promising predictive biomarkers in a database based on prospective trials comparing bacillus Calmette-Guerin (BCG) with mitomycin-C and a combination of epirubicin and interferon, respectively. Background The most common form of bladder cancer is non–muscle-invasive tumors treated initially with transurethral resection. Unfortunately more than half recur and some also progress. Consequently, an attempt to prevent poor outcome is frequently made by intravesical instillations either by chemo- or immunotherapy. The response to such treatment is unpredictable, which is why markers predicting outcome would be valuable. Patients and methods Immunohistochemical expression of ezrin, CK20, and Ki-67 was evaluated in a tumor tissue microarray based on 2 nordic multicenter trials comparing treatment with BCG vs. other intravesical adjuvant therapies. Kaplan-Meier analysis, log-rank test, and Cox regression were used to evaluate the effect of the biomarkers on recurrence-, progression-, and treatment failure-free survival. Results Of the 294 available patients immunoreactivity could be assessed in 285 patients for ezrin (97%), 285 patients for CK20 (97%), and 294 patient׳s for Ki-67 (100%). The 3 biomarkers did not predict time to any of the endpoints. Multifocality was the only predictive factor for time to recurrence ( P = 0.029) and progression ( P = 0.031). Ezrin was, however, predictive for treatment failure ( P = 0.029) in a subgroup (BCG treated in one of the trials). In a multivariate analysis among BCG treated, none of the variables correlated to recurrence and only multifocality correlated to progression. Limitations in our study are the retrospective design and those inherent to immunohistochemistry. Conclusions The negative results from this validation study question the ability of the tested biomarkers to predict therapy effect.