Flow cytometry and nuclear morphometry were compared with traditional pathologic grading techniques for predicting the course of malignant fibrous histiocytoma of the extremities. Clinical, pathologic, and flow/morphometric variables from 53 cases were tested by Cox regression for prediction of distant recurrence and mortality. Tumor grading based on extent of tumor necrosis was a significant predictor for both disease-free survival (p = .014) and overall survival (p = .003). The fraction of nuclei in the S + G2M segment of DNA histograms was significant for disease-free survival (p = .007), and remained significant (p = .033) in a joint Cox model with necrosis-based grade (p = .004 for the bivariate model). Relative risk for recurrence varied nearly 10x between the 10th and 90th percentiles of grade and (S + G2M)1/2. Overall survival was predicted by a nuclear shape feature termed "R" (p = .000008), the casewise difference (residual) between expected and observed nuclear perimeter as a function of average Feret diameter. In a bivariate Cox model, relative risk of mortality varied 35x between the 10th and 90th percentiles of grade and R. Cytometric and morphometric data contain information about recurrence-free and overall survival beyond that available from more usual clinical and pathologic features. It seems likely that nuclear morphometry, in particular, will prove to be a useful aid for estimating the prognosis of patients with malignant fibrous histiocytoma of the extremities.