Objectives To determine the efficacy of a three-fold reduced dose (RD, 27 mg) of intravesical bacille CalmetteGuérin (BCG) against the standard dose (81 mg) in patients with superficial bladder cancer, assessing recurrence, progression and differences in toxicity. Patients and methods Five hundred patients with superficial bladder cancer (Ta, T1, Tis) were enrolled and randomly assigned to be treated after transurethral resection of all visible lesions with intravesical BCG Connaught strain (weeklyrsix and thereafter fortnightlyrsix) either with the standard or RD instillation. Results All but one of the 500 patients were evaluable for efficacy and toxicity (252 in the standard arm and 247 in the RD arm). The median follow-up was 69 months (maximum 104); 71 (28%) patients in the standard arm and 76 (31%) in the RD arm developed recurrences; the median time to recurrence has not yet been attained, but at 5 years the mean (SD) percentage of recurrence-free patients was 70.5 (3.12) and 70.4 (3.1) for the standard and RD arms, respectively. In patients presenting with multifocal tumours, the standard dose was more effective against recurrences than the RD (P=0.0151). In those with G3 and high-risk tumours overall, the superiority of the standard dose was marginal (P=0.060 and P=0.082). Twenty-nine (11.5%) tumours in the standard arm and 33 (13.3%) in the RD arm progressed to invasive disease; the median time to progression has not yet been attained, but the percentage of progression-free patients at 5 years was 88.8 (2.23) and 86.9 (2.31) for the standard and RD arms, respectively. The standard dose was more effective than the RD against progression only in patients with multifocal disease (P=0.048). Twelve (4.8%) cystectomies were performed in the standard and 15 (6.1%) in the RD arm. Currently, 106 (21.2%) patients have died, but only 38 (7.6%) from bladder cancer, i.e. 20 (7.9%) in the standard and 18 (7.5%) in RD arm. Overall the disease-specific death rate was lower for those patients who completed the scheduled treatment. The cause-specific survival at 5 years did not differ between the arms (P=0.76) but there was a trend toward better cause-specific survival for patients with multifocal tumours in the standard arm. Toxicity differed between the arms, significantly more patients having no toxicity in the RD arm, and fewer having delayed instillations or withdrawing. However, severe systemic toxicity occurred even in patients treated with the RD, in a similar proportion to those receiving the standard dose. Conclusion Overall, the RD gave similar results for recurrence and progression but with significantly less toxicity. However, patients with multifocal tumours fared better with the standard dose and there was a trend towards better recurrence rates in patients with highrisk tumours. We recommend continuing to use the standard dose for high-risk tumours, while we consider the reduced dose safe and effective for intermediaterisk lesions and for maintenance schedules.