Infection is the major complication of cyclophosphamide therapy in patients with lupus nephritis. The objectives of this study were to report and compare the rate of infection between children with lupus nephritis who had received intravenous pulse cyclophosphamide (IVCY) and those who had received oral cyclophosphamide (OCY) and to determine the risk factors for infection during treatment with cyclophosphamide in these groups. Records of nine patients who had received IVCY from the beginning [pure intravenous cyclophosphamide (PIVCY) group], 11 patients who had received prior oral cyclophosphamide and later switched to IVCY [combined intravenous cyclophosphamide (CIVCY) group] and 41 patients who had received OCY were reviewed. Infection occurred in 21 of 61 patients (34%). In the PIVCY group, four episodes of infection occurred in three of nine patients (33%). In the CIVCY group, six episodes of infection occurred in four of 11 patients (36%). In the OCY group, 18 episodes of infection occurred in 14 of 41 patients (34%). The rate of infection between these groups was not different (P=0.99). None of the following parameters were risk factors for infection: cumulative dose of cyclophosphamide, leukopenia and neutropenia. On the contrary, white blood cell (WBC) count and polymorphonuclear cell (PMN) count were significantly less in the no-infection group (P=<0.001, P<0.001, respectively), with odds ratios for leukopenia (WBCs <4,000 mm(3)) and neutropenia (PMNs <1,500 mm(3)) between the infection and the no-infection group equal to 0.18 (95%CI 0.05-0.63) and 0 (95%CI 0-0.19), respectively. Most of the patients who had infection received prednisolone at a dosage of more than 0.5 mg/kg per day (67% of the PIVCY group, 50% of the CIVCY group and 83% of the OCY group). Fatal infections occurred in two patients who had concomitant active systemic lupus erythematosus (SLE). Although lymphopenia (lymphocyte count <1,500/mm(3)) was not the risk factor for infection, it was observed that six of seven patients with herpes zoster had lymphopenia. Herpes zoster seemed to occur more frequently in the OCY group (15%) than in the whole IVCY group (5%), but there was no statistical difference (P=0.41). We conclude that the rate of infection in the IVCY and OCY group was not different. Infection is likely to occur in patients receiving a concomitant high dose of prednisolone. The occurrence of fatal infection in patients with active disease should be noted. No single risk factor was detected in this study.