BACKGROUND: Tumour recurrence following transurethral resection (TUR) for Ta and T1 bladder cancer is a major clinical problem. Intravesical administration of mitomycin C (MMC) or bacillus Calmette-Guerin (BCG) has proven prophylactic activity but both are associated with local and systemic side-effects. A systematic review was carried out to compare the efficacy of these two agents. OBJECTIVES: To undertake a systematic review and meta-analysis comparing intravesical mitomycin C and Bacillus Calmette-Guerin in terms of tumour recurrence, disease progression and overall survival in Ta and T1 bladder cancer. Treatment-related toxicities would also be evaluated. SEARCH STRATEGY: A comprehensive search of MEDLINE, EMBASE, Healthstar, Cochrane Controlled Trials Register, Cancerlit, and DARE was performed, and hand searching of relevant journals undertaken. SELECTION CRITERIA: Trials in any language were included in the meta-analysis if they were properly randomised, included medium to high risk patients with Ta or T1 bladder cancer and compared intravesical MMC versus BCG. DATA COLLECTION AND ANALYSIS: Trial eligibility, methodological quality and data extraction were assessed independently by two reviewers. Time to event analysis was evaluated using log hazard ratios, with a sensitivity analysis for subgroups according to patient's risk of recurrence. MAIN RESULTS: Twenty-five articles were identified but only seven were considered eligible. This represented 1901 evaluable patients in total, 820 randomised to MMC and 1081 to BCG. Six trials had sufficient data for meta-analysis and included 1527 patients, 693 in the mitomycin arm and 834 in the BCG arm. The weighted mean log hazard ratio (variance) for tumour recurrence for the six trials was - 0.022 (0.005). This indicated no significant difference between MMC and BCG (p = 0.76). However, the meta-analysis indicated evidence of significant heterogeneity between trials (p = 0.001). A subgroup analysis of three trials that included only high risk Ta and T1 patients indicated no heterogeneity (p = 0.25) and a log hazard ratio (variance) for recurrence of -0.371 ( 0.012). With MMC used as the control in the meta-analysis, a negative ratio is in favour of BCG and, in this case, is highly significant (p = 0.0008). The seventh trial, in abstract form only, used BCG in low doses for two arms of the trial (27 mg and 13.5mg) compared to a standard dose of mitomycin C (30mg), and reported a significantly reduced recurrent rate with BCG (27mg) compared to mitomycin C (p = 0.001). Only two trials included sufficient data to analyse disease progression and survival, representing a total of 681 patients; 338 randomised to BCG and 343 to MMC. There was no significant difference between MMC and BCG for disease progression (log hazard ratio + variance: 0.044 + 0.04, p = 0.16) or survival (-0.112 + 0.03, p = 0.50). Local toxicities (dysuria, cystitis, frequency, and haematuria) were associated with both MMC (30%) and BCG (44%). Systemic toxicities, such as chills, fever and malaise, were observed with both MMC and BCG (12% and 19%, respectively) although skin rash was more common with MMC. REVIEWER'S CONCLUSIONS: The data from the present meta-analysis indicate that tumour recurrence was significantly reduced with intravesical BCG compared to MMC only in the subgroup of patients at high risk of tumour recurrence. However, there was no difference in terms of disease progression or survival, and the decision to use either agent might be based on adverse events and cost.