PURPOSE: To define the prognostic factors in adult patients with locally controlled soft tissue sarcoma (STS) and to determine which patients should be considered for adjuvant treatment. PATIENTS AND METHODS: Five hundred forty-six patients with a nonmetastatic and locally controlled STS, collected in a cooperative data base by the French Federation of Cancer Centers (FNCLCC) Sarcoma Group from 1980 and 1989, were studied. Histologic slides of all patients were collegially reviewed. Initial treatment consisted of complete tumor resection with amputation in only 4% of the patients. Adjuvant radiotherapy was administered to 57.9% and adjuvant chemotherapy to 31%. Relationships between tumor characteristics were analyzed, and univariate and multivariate analyses were performed using Cox models for the hazards rate of tumor mortality, development of distant metastasis, and strictly local recurrence. RESULTS: Unfavorable characteristics with an independent prognostic value for tumor mortality were: grade 3 (P = 3 x 10(-10)), male sex (P = 1.5 x 10(-5)), no adjuvant chemotherapy (P = 5.4 x 10(-5)), tumor size > or = 5 cm (P = 3.8 x 10(-3)), and deep location (P = 4.6 x 10(-3)). Unfavorable characteristics for the development of distant metastasis were: grade 3 (P = 4 x 10(-12)), no adjuvant chemotherapy (P = 6.4 x 10(-4)), tumor size > or = 10 cm (P = 9.8 x 10(-4)), and deep location (P = 1.3 x 10(-3)). For the development of local recurrence, the unfavorable characteristics were: no adjuvant radiotherapy (P = 3.6 x 10(-6)), poor surgery (local excision) (P = 2 x 10(-4)), grade 3 (P = 7.6 x 10(-4)), and deep location (P = 10(-2)). Grade, depth, and tumor size were used to define groups of patients according to the metastatic risk. Adjuvant chemotherapy was beneficial in terms of overall survival and metastasis-free survival in grade 3 tumor patients only. Despite worse characteristics concerning tumor depth, tumor-node-metastasis (TNM) and American Joint Committee (AJC)/International Union Against Cancer (UICC) classifications and grade in patients with adjuvant radiotherapy, the latter experienced significantly fewer local recurrences than patients with no radiotherapy. CONCLUSION: Grade, tumor depth, and tumor size could be used to select patients with a high metastatic risk, for which adjuvant chemotherapy could be beneficial.