PURPOSE The ultimate goal of the Third Intergroup Rhabdomyosarcoma Study (IRS-III, 1984 to 1991) was to improve treatment outcome in children with rhabdomyosarcoma through clinical trials comparing risk-based protocols of surgery and multiagent chemotherapy, with or without irradiation. PATIENTS AND METHODS One thousand sixty-two previously untreated, eligible patients who were entered onto the study after surgery were randomized or assigned to treatment by clinical group (I through IV), histology (unfavorable or favorable), and site of the primary tumor. Initial responses, progression-free survival (PFS), and survival (S) were the end points used in comparisons between randomized groups and between patients treated in IRS-III and IRS-II (1978 to 1984). RESULTS The overall outcome of therapy in IRS-III was significantly better than in IRS-II (5-year PFS, 65% +/- 2% v 55% +/- 2%; P < .001 by stratified testing). Patients with group I favorable-histology tumors fared as well on a 1-year regimen of vincristine and dactinomycin (VA), as did a comparable group treated with VA plus cyclophosphamide (C) (5-year PFS, 83% +/- 3% v 76% +/- 4%; P = .18). Results for patients with group II favorable-histology tumors, excluding orbit, head, and paratesticular sites, were inconclusive regarding the benefit from addition of doxorubicin (ADR) to VA. Patients with group III tumors, excluding those in special pelvic, orbit, and other selected nonparameningeal head sites, fared much better on the more intensive regimens of IRS-III than on pulsed VAC or VAC-VADRC in IRS-II (5-year PFS estimates, 62% +/- 3% v 52% +/- 3%; P < .01); however, there were no significant differences in outcome among the groups treated in IRS-III. Patients with metastatic disease at diagnosis (clinical group IV) did not benefit significantly from the more complex therapies evaluated in IRS-III. CONCLUSION Intensification of therapy for most patients in IRS-III, using a risk-based study design, significantly improved treatment outcome overall. The largest gain from this strategy was realized in patients with gross residual tumor after biopsy (clinical group III). It was also possible to decrease therapy for selected patient subsets without compromising survival.