Preoperative discrimination between benign and malignant ovarian tumors by multivariate analysis

To obtain a precise preoperative diagnosis of benign and malignant ovarian tumors, multivariate analysis was performed using a combination of tumor markers, diagnostic imagings, clinical findings and epidemiological data. Twenty-eight factors were selected for a discriminant analysis, which included 10 factors for tumor markers, 8 for diagnostic imagings, 4 for clinical features and 6 for epidemiological data. On 235 patients (malignant: 100, benign: 135), the usefulness of the discriminant analysis was evaluated retrospectively. As a results, an accuracy of preoperative discrimination was 82% for a discriminant analysis of tumor markers alone, 91% for diagnostic imagings, 84% clinical findings and 65% for epidemiological data. Twenty-one factors were selected by chi-square test or T-test. When 21 factors were used, the sensitivity and the specificity for the diagnosis of malignant ovarian tumor were 98% and 88% respectively. The results indicate that the discriminant analysis is valuable for preoperative discrimination of patient with ovarian tumors.