BACKGROUND Pelvic recurrence is a significant problem following curative resection for rectal cancer. Although treatment options include surgery, chemotherapy, radiotherapy, or any combination of these, the role of surgery remains controversial in management of these patients. OBJECTIVE In this study, we have attempted to define the patient with pelvic recurrence following curative rectal surgery who may benefit from reresection. METHODS A review of the prospective colorectal database at Memorial Sloan Kettering Cancer Center (MSKCC) between 1983 and 1991 identified 25 patients who had pelvic recurrence following a curative resection for rectal cancer and 52 patients who had their initial rectal surgery at an outside institution (OI) and their pelvic recurrence treated at MSKCC. Survival was calculated from time of recurrence by the Kaplan-Meier method, and survival comparisons were made by log-rank analysis. There were no differences between the two groups related to age, gender, type of initial surgery, stage, or use of adjuvant therapy. RESULTS For the MSKCC group, median time to initial recurrence was 18 months, and median survival was 40 months. Recurrence was symptomatic in 17 patients and asymptomatic in 8 patients. Pain and bleeding accounted for more than one-half of symptomatic recurrences. Of the 17 symptomatic recurrences, 11 (65 percent) had relief of preoperative symptoms. There were no clinical or pathologic factors identified of the primary tumor or recurrence that predicted improved survival following salvage therapy. It was not possible to preoperatively determine which patients could undergo curative reresection. For the OI group, median time to recurrence was 13.7 months, and median survival from time of initial recurrence was 31 months. Curative reresection was the only factor that predicted for improved survival compared with noncurative treatment (P = 0.02). A comparison of the two groups revealed that pelvic recurrence was more likely to be reresected for cure in the OI group vs. the MSKCC group (34/51 vs. 9/25; P < 0.02). There was no survival difference between the two groups when comparing curative with noncurative management of these patients. CONCLUSIONS Symptoms from recurrent rectal cancer can be palliated with surgery. The only patients who had a survival benefit were those patients in the OI group whose disease could be completely resected. These differences in reresection rates may be attributable to the presence or absence of available planes for dissection around the recurrence in the OI group, as determined by the method of initial curative resection.