Mycophenolate mofetil (MMF) has emerged as a new therapeutic option in steroid-dependent nephrotic syndrome (SDNS). We conducted a phase II Bayesian trial of MMF in children with SDNS. Phase II trials, usually single-arm studies, investigate the effect of new treatments. Standard Fleming's procedure relies on observed results (relapse rate during the trial), whereas Bayesian approach combines observed results with prior information (expected relapse rate according to prior studies and clinical experience). All patients were required to have received prior alkylating-agent treatment. Sixty-seven percent of them had also received levamisole. Patients received MMF (1,200 mg/m(2)/day) and prednisone according to a defined schedule [reduction of alternate-day (e.o.d) dose to 50% of pre-MMF dose at 3 months, 25% at 6 months]. Twenty-four children (median age 6.0 years, 2.8-14.4) entered the study and 23 completed it. Bayesian analysis showed that adding four patients would not change significance of results, allowing stopping inclusions. Four patients relapsed during the first 6 months (estimated probability 17.6%, 95% credibility interval: 5.4-35.0%) and two at months 8 and 11.5. In the 19 patients free of relapse during the first 6 months, median (Q1-Q3) prednisone maintenance dose decreased from 25 (10-44) to 9 (7.5-11.2) mg/m(2) e.o.d (p < 0.001) and cumulative dose from 459 (382-689) to 264 (196-306) mg/m(2)/month (p < 0.001) before and on MMF respectively. Pre-MMF patient characteristics and MMF pharmacokinetics did not differ between patients with or without relapse. MMF reduces relapse rate and steroid dose in children with SDNS and should be proposed before cyclosporine and cyclophosphamide.