Between June 20, 1977 and February 5, 1983, the Gynecologic Oncology Group entered 1180 women with clinical stage I or II (occult) endometrial carcinoma into a surgical-pathological staging study. Eight hundred ninety-five patients with endometrioid or adenosquamous carcinoma were evaluable for this study which relates surgical-pathological parameters and postoperative treatment to recurrence-free interval and recurrence site. Proportional hazards modeling of time to recurrence was performed. For patients without metastasis determined by surgical-pathological staging the greatest determinant of recurrence was grade 3 histology adenocarcinoma grade 3, relative risk (RR) = 15; adenosquamous carcinoma grade 3, RR = 8.1; all adenocanthomas, RR = 1.0). Of 48 patients with histologically documented aortic node metastases, 47 had one or more of the following features: (1) grossly positive pelvic nodes, (2) grossly positive adnexal metastasis, or (3) outer one-third myometrial invasion. Pelvic radiation was administered to 48.0% and vaginal brachytherapy alone to 10.2% of patients postoperatively; 41.8% received no adjuvant radiation therapy. None of three recurrences in the vaginal implant group were vaginal or pelvic; 7.4% (7 of 95) of recurrences in the pelvic radiation therapy (RT) group were vaginal and 16.8% were pelvic; 18.2% (8 of 44) of recurrences in the no adjuvant radiation group were vaginal and 31.8% pelvic. Because of the high degree of selection bias no valid comparisons can be made of recurrence-free interval in these groups. The 5-year recurrence-free interval for patients with negative surgical-pathological risk factors (other than grade and myoinvasion) was 92.7%; involvement of the isthmus/cervix 69.8%; positive pelvic cytology 56.0%; vascular space invasion 55.0%; pelvic node or adnexal metastases 57.8%; and aortic node metastases or gross laparotomy findings 41.2%. It is not clear that cervix invasion per se diminishes survival, because it is more often associated with poor tumor differentiation (34.7% versus 24.0%, grade 3) and deep myoinvasion (47.0 vs 18.6%) than cases without cervix invasion. The relapse rate among cervix-positive and -negative cases with grade 3 lesions and deep myoinvasion is not dramatically different (48.8% vs 39.8%). The proportion of failures which were vaginal/pelvic (34.6% for the surgery only group compared to 12.5% of the RT group) appears to favor the use of adjuvant radiation for patients with more than one-third myoinvasion and grade 2 or 3 tumor. There were 97 patients in the study group with malignant cytology of which 29.1% had regional/distant failure, which compares to 10.5% of the cytology-negative patients.(ABSTRACT TRUNCATED AT 400 WORDS)