A retrospective review of 46 cases of Ewing's (43) and extraosseous Ewing's (3) sarcoma was performed to examine for prognostic factors. Follow-up ranged from 27 to 135 months with a mean of 77.6 months, 86% greater than 36 months. Nine (20%) patients presented with distant metastases, 98% received multiagent chemotherapy, and 98% received radiation therapy. Overall actuarial survival and local control at 5 years were .52 +/- .08 and .78 +/- .07, respectively. The freedom from relapse or treatment related mortality at 5 years was .46 +/- .08. In 12 (26%) patients, surgical resection of the involved bone or soft tissue was part of the initial treatment plan. Ninety-two percent (11/12) of these patients also received radiation therapy. In addition to surgical resection, tumor size (less than 500 cc vs. greater than or equal to 500 cc), primary site (central vs. other), and stage were also analyzed for prognostic significance, and where appropriate, were included in Cox multivariate analyses. Considering all patients, the 5-year actuarial survival was .92 +/- .08 vs. .37 +/- .09 for patients receiving and not receiving surgical resection, respectively (p = .001 by logrank, p = .02 by Cox). To make the groups more comparable, 8 patients with local failure and 5 patients with non-evaluable primary sites were excluded. After these exclusions, the presence or absence of surgical resection had decreased significance: .92 +/- .08 vs. .59 +/- .11, respectively; p = .01 by logrank, p = .07 by Cox. Only primary site remained statistically significant: .86 +/- .08 for noncentral sites vs. .38 +/- .16 for central sites, p less than .0002 by logrank, p = .0004 by Cox. Surgical resection when added to local irradiation has prognostic significance in this retrospective review. The reason for this is not clear, but may be related to a decrease in local failure after resection.