The role of surgery in the management of primary gastrointestinal lymphoma remains controversial. We retrospectively reviewed the management and outcome of 107 patients with the diagnosis of gastrointestinal lymphoma treated at the UCLA Medical Center during the period 1956-1990. Sixty-four patients underwent surgical exploration at the UCLA Medical Center; 35 of these underwent resection for cure. Sixteen of these 35 patients received no postoperative adjuvant therapy. Twenty-nine patients underwent palliative or "noncurative" resection. There were five postoperative deaths (mortality rate 8%). The overall morbidity rate was 48%. There were 3 perforations in a total of 53 patients receiving multiagent chemotherapy. Five-year actuarial survival was as follows: 59% for curative resection alone, 5 1 % for curative resection plus adjuvant therapy, and 28% for "noncurative" resection (P C 0.05). Multivariate analysis revealed that stage of disease (P < 0.01) and resection for cure (P < 0.05) were independent predictors of survival. These results suggest that patients undergoing resection for cure have improved survival. The apparent low risk of perforation during chemotherapy, along with the considerable risk of morbidity and mortality associated with operation, suggests that a policy of debulking large tumors prior to chemotherapy is unwarranted.