Although the significance of various prognostic factors. such as tumor size and mitotic index (MI), has been well established for smooth-muscle tumors of the stomach, the significance of these factors in other sites is less well defined. We studied 1004 patients with gastrointestinal smooth-muscle tumors for whom vital status could be determined. The average MI and tumor size varied significantly among the five major sites examined: esophagus (53 cases), stomach (524 cases), small bowel 252 cases), colon/rectum (108 cases), and omentum/mesentery/peritoneum (67 cases). There was a significant difference in site-specific survival (p = 0.001), with 10-year survival varying between 50% and 70%. Multivariate analysis demonstrated tumor location (p = 0.0320), size (p = 0.0003), MI (p < 0.0001), and patient age (p < 0.0001) to each carry independent prognostic value. The significance of MI was highly site dependent. Separation of survival curves for the stomach, using a threshold for analysis of either 5 or 10 mitotic figures/50 high-power fields, was very good. In contrast, small-bowel tumors showed little separation between survival curves, regardless of whether a threshold of 1, 5, or 10 mitotic figures MF/50 high-power fields was used to distinguish groups. In no site were tumor size and MI alone sufficient to provide an accurate longterm prediction of prognosis. Although tumor location, size, MI, and age have independent value in predicting the prognosis of patients with gastrointestinal smooth-muscle tumors, better methods are still required to accurately predict clinical course.