Background: Adequate management and treatment of ovarian carcinoma requires a complete surgical staging supported by frozen-section examination. To achieve this goal it is necessary a high level of accuracy. Aim: To evaluate the accuracy of frozen-sections in ovarian carcinoma considering the influence of tumor diameter and weight. Patients and methods: Retrospective study of frozen-sections performed in patients with ovarian tumors who underwent surgery. Frozen- and permanent- sections were divided into three categories (benign, borderline and malignant) and stratified by diameter (<10 cm, 10 to 20 cm, >20 cm) and weight (<700 g, 700 a 1400 g, >1400 g). The diagnostic correlation, sensitivity, specificity, predictive values and accuracy of each frozen-section diagnosis were determined. Results: Eight hundred forty two ovarian tumors that underwent frozen-sections between January 1988 and October 1998 were studied. Final diagnosis was 86,7% benign, 2,7% low malignant potential (LMP) and 10,6% malignant. The diagnosis correlation between frozen- and permanent-sections was 98,2%. Misdiagnosis was in epithelial ovarian tumors, particularly in LMP tumors. Sensitivity, specificity, positive- and negative- predictive values and accuracy of the four hundred eighty nine epithelial tumor were 92,6%, 99,2%, 96,7%, 98,2% and 97,9%, respectively. Diagnostic correlation was higher in epithelial ovarian tumors with diameter <10 cm (98,2% v/s 93,8%) and weight <700 g (96,9% v/s 88,9%). Conclusions: Diagnostic correlation with permanent-section examination, sensitivity, specificity and predictive values of frozen-sections are high in ovarian tumors. Accurate diagnosis at frozen sections of epithelial ovarian tumors with diameter >10 cm or weight >700 g (particularly in LMP tumors) is difficult because of the extensive sampling required. Frozen-sections diagnoses are important to determine the type and extent of surgery performed at the initial operation.