From July 1975 to February 1985, 300 patients with extremity soft tissue sarcomas (STS) were entered into treatment protocols at the National Cancer Institute. Using the Cox proportional hazards model, the STS were analyzed for independent prognostic variables that had significant association with disease-free and overall survival. The effects of adjuvant treatment were also included in the analysis. Percent tumor necrosis was the basis for separating intermediate and high-grade STS (less than 15% necrosis = grade 2; greater than 15% necrosis = grade 3). The most important prognostic variable was histologic grade, which was highly associated with relative risk of metastasis and fatal outcome (P less than .0001). Histologic type of STS was not a significant prognostic variable in predicting metastases. Patients with tumors located on the proximal or middle part of the extremity had a fivefold increased risk of death (P = .02) compared with patients who had STS located more distally. Mitotic activity (P = .0005) and vascular invasion (P = .0034) were also associated with malignant behavior, but histologic grade was the most significant predictor of outcome. Histologic grade of malignancy was based upon a combination of tumor typing and visual estimation of extent of tumor necrosis; using these two features, one can reliably separate low-grade (i.e., grade 1) from high-grade STS and also stratify the more aggressive (i.e., grade 2 and 3) tumors. Over the 9.5-year period of study, the tumor-related mortality was as follows: grade 1 (1/38), 3%; grade 2 (19/142), 13%; and grade 3 (45/120), 38%.