Among nephrotic children with frequent relapses at risk for cumulative steroid toxicity, identification of children who may be at high risk for subsequent relapse is very important in making the decision to introduce cytotoxic drugs. We examined the clinical course of 467 relapses in 121 steroid-sensitive nephrotic children to elucidate the risk factors for subsequent relapse, using the Cox proportional-hazards regression model. Gender, age at onset, duration of illness from onset, prednisolone dosage at the most-recent relapse, and regimens of initial steroid therapy at onset were not associated with risk. Relapse within the 1st year was a powerful independent predictor of subsequent relapse irrespective of the duration of illness. The hazard ratio of patients with more than one relapse within the 1st year increased to 1.72-2.12 compared with those without a relapse within the 1st year. The remission period just before the most-recent relapse was also a significant predictor. The risk for patients with a 1-year or longer remission period decreased to 0.57. Patients treated with cyclophosphamide for 12 weeks had a significantly longer remission than those treated with prednisolone alone. Our results suggest that early relapse after onset and/or a short remission period just before recent relapse are independent risk factors for subsequent relapse. Cytotoxic therapy has serious adverse effects and its effect may be limited. Our results may be helpful in deciding on the suitability of cytotoxic drugs.