Prognostic factors were evaluated in 109 soft tissue sarcomas of the extremities, walls of the trunk, head, and neck. All lesions were graded according to the systems proposed by the National Cancer Institute (NCI) and the French Federation of Cancer Centers (FNCLCC), and a correlation was found between tumor grade and prognosis. Univariate analysis selected the following variables as unfavorable prognostic factors: invasive tumor margins, extra-compartmental status, deep tumors, tumor diameters over 5 cm, inadequate excision, presence of necrosis, high mitotic count, histologically undifferentiated tumors, and blood vessel invasion. These variables were found to be interdependent. Multivariate analysis selected quality of surgery as the most important variable for predicting local recurrences. The factors selected with regard to overall and metastasis-free survival were tumor size, tumor margins, necrosis, and adequacy of excision. These results permitted classification of patients into four prognostic groups: two with good and two with bad prognosis. Five-year survival for the four groups was 100%, 83%, 53%, and 0%; 5-year metastatic rates were 0%, 12%, 67%, and 100%. Similar groups were obtained when the variables of tumor margins and size were combined with an adaptation of the NCI grading (low-grade tumors/high-grade tumors without necrosis/high-grade tumors with necrosis). Comparative analysis showed that patients with tumors of the same histologic grade or type were not necessarily classed in the same prognostic groups. A better clinicopathologic correlation was obtained using a combination of prognostic factors than with histologic grading or typing alone.