Summary. We have examined the outcome for children treated on two consecutive United Kingdom Children's Cancer Study Group studies of localized B-cell non-Hodgkin's lymphoma (NHL). The first study (NHL 8501; 1985–1989) included cyclophosphamide in the treatment regimen at a total cumulative dose of 4 g/m2 whereas the regimen in the succeeding study (NHL 9001; 1990–1996) did not include cyclophosphamide. Ninety children with confirmed B-cell NHL were treated in the two studies (NHL 8501, n = 33 and NHL9001, n = 57). With a median follow-up of 7·5 years, overall survival for localized B-cell NHL did not differ between the two regimens with observed 3-year survivals of 94%[95% confidence interval (CI) 80–98%] and 89% (95% CI 79–95%) respectively (P = 0·47). There was also no difference in the event-free survival between children treated on regimen NHL 8501 and NHL 9001 [91% (95% CI 76–97%) vs 84% (95% CI 73–92%) after 3 years; P = 0·34]. Although the difference in the number of failed remissions between NHL 8501 and 9001 (0/33 vs 6/57) approached statistical significance (P = 0·08, Fisher's exact test), there was no overall statistical difference between the treatment failures on either regimen (P = 0·34). Substantial long-term survival can be achieved for many children with localized B-cell NHL without the use of cyclophosphamide. Further studies are needed to identify whether all clinical or histopathological subgroups will benefit equally from the omission of cyclophosphamide.