We examined the method of diagnosis for a group of women who developed recurrent endometrial carcinoma after being rendered clinically disease-free by primary therapy. We then used this information to develop a follow-up protocol that maximizes the chances for detecting recurrence while minimizing surveillance costs. In brief, we evaluated all women with clinical stage I endometrial carcinoma who were treated with curative intent during a 7-year period. Medical records were examined to identify patients who had tumor recurrence diagnosed during follow-up in our clinic. Clinical presentation, time to diagnosis, method of diagnosis, and subsequent outcome were analyzed. This information was used to design a surveillance protocol for further clinical testing. Ninety-six percent of 412 women treated between 1985 and 1992 were clinically disease-free after primary surgery with or without adjuvant treatment. Median follow-up is 64 months. Overall, 44 patients (11%) developed recurrent cancer after a median interval of 14.8 months. Complete follow-up data were available for the 39 patients who had their recurrence diagnosed in our clinic. The cumulative percentages of diagnosed recurrences were 51, 82, and 95% at 12, 24, and 36 months, respectively. Sixteen women (41%) had symptoms that led to the identification of recurrent disease. Recurrences in the 23 asymptomatic women (59%) were diagnosed by physical examination in 13, chest radiograph in 1, serum CA-125 level in 6, vaginal cytology in 1, and computed tomography in 2. Only 1 patient with a grade 1 adenocarcinoma had treatment failure. At the time of analysis, 30 patients with recurrent cancer had died of disease, 6 were alive with disease, and 3 were free of disease. A surveillance scheme consisting of an examination, vaginal cytology, and serum CA-125, combined with immediate evaluation of symptomatic women, could be expected to identify 95% of recurrences. Such an approach, performed at 6- to 12-month intervals for 3 years, could be limited to patients with grade 2-3 adenocarcinomas or variant cell types. However, given the high failure rate of salvage therapy, the prompt detection of recurrence may not convey a survival advantage.