BACKGROUND The efficacy of intravesical gemcitabine was evaluated compared with repeated administration of bacille Calmette-Guérin (BCG) after BCG failure in high-risk, non-muscle-invasive bladder cancer (BC). METHODS In this multicenter, prospective, randomized, phase 2 trial, eligible patients were those with high-risk non-muscle-invasive BC, failing 1 course of BCG therapy. All patients were randomly allocated to Group A, receiving intravesical gemcitabine (at a dose of 2000 mg/50 mL) twice weekly for 6 consecutive weeks and then weekly for 3 consecutive weeks at 3, 6, and 12 months, or Group B, receiving intravesical BCG (Connaught strain, 81 mg/50 mL) over a 6-week induction course and each week for 3 weeks at 3, 6, and 12 months. Outcome measures were recurrence rate, time to first recurrence, and progression rate. Treatment-related complications were also evaluated. RESULTS Eighty participants were enrolled, 40 for each group 52.5% in Group A developed disease recurrence versus 87.5% of those in Group B (P = .002). There was no statistically significant difference in mean time to the first recurrence (Group A, 3.9 months; Group B, 3.1 months; P = .09). Kaplan-Meier analysis of 2-year recurrence-free survival showed significant differences between Group A and B (19% and 3%, respectively, P < .008). Seven of 21 (33%) patients in Group A and 13 of 35 (37.5%) patients in Group B had disease progression and underwent radical cystectomy (P = .12). Both intravesical administrations were generally well tolerated. CONCLUSIONS Gemcitabine might represent a second-line treatment option after BCG failure in high-risk non-muscle-invasive BC patients.