OBJECTIVE: To prospectively evaluate in a blinded fashion the accuracy of frozen section in endometrial cancer. METHODS: Sixty patients with endometrial cancer or complex atypical hyperplasia were consecutively enrolled. Intraoperatively, a frozen section was obtained, processed, and stored for interpretation by blinded pathologists. Final pathologic diagnosis was conducted in the usual fashion with the pathologists blinded to frozen results. Histologic grade and myometrial invasion on frozen section was correlated with final pathology. RESULTS: Median age was 61 years (range, 39-82 years). Fifty-seven percent of patients were white, and mean body mass index was 40 mg/kg2. Depth of invasion on frozen correlated with final pathology in 67% (95% confidence interval [CI] 55-79%). Twenty-eight percent (95% CI 17-39%) of patients were upstaged from frozen to final. Patients with no invasion on frozen were upstaged in 46% (95% CI 28-64%). Histologic grade on frozen correlated with final pathology in 58% (95% CI 46-70%); 38% (95% CI 26-50%) of patients were upgraded by final grade. Patients with frozen grade 1 histology or less were upgraded in 61% (95% CI 45-77%). Clinically relevant upstaging occurred in 11 patients (18%) (95% CI 8-28%). CONCLUSION: Frozen section for histologic grade and depth of myometrial invasion in endometrial cancer correlates poorly with final pathology. Because a large number of patients are potentially understaged with the use of frozen section with a subsequent risk of over and under treatment, we recommend consideration of comprehensive surgical staging for all patients with endometrial cancer. LEVEL OF EVIDENCE: II-2.