Transvaginal ultrasonographic characterization of ovarian masses: comparison of five scoring systems in a multicenter study.

The aim of this work was to test and compare the accuracy of five different morphological scoring systems to identify malignant ovarian masses in a prospective multicenter study. Four of the systems had previously been reported by Granberg, Sassone, De Priest and Lerner and the fifth is newly developed. A total of 330 ovarian neoplasms were collected in three different centers, which adopted the same diagnostic procedures. Of these, 261 masses were benign (mean diameter 50 +/- 26 mm) and 69 were malignant (mean diameter 69 +/- 33 mm) (prevalence 21%). The area under the receiver operating characteristic (ROC) curve for the multicenter score was 0.84. This was significantly better than the areas of the other four scores which ranged from 0.72 to 0.75. The cut-off levels derived from the five ROC curves achieved a sensitivity that ranged from 74% (Sassone score) to 88% (De Priest score > or = 5), and a specificity from 40% (De Priest) to 67% (multicenter); the highest positive predictive value was 41% (multicenter). With a cut-off level of 9, the accuracy of the multicenter score was significantly better than the scores of Granberg and De Priest (McNemar's test p < 0.0001). Similar results were obtained in 207 ovarian masses of < or = 5 cm in mean diameter, and when 19 borderline and 11 stage 1 cancers only were considered. For the clinical purposes of a screening test we also checked a possible cut-off level of > or = 8, which increased the sensitivity to 93% with a drop of specificity to 56%. With the use of the same criteria for the scores of the different authors, the following values were obtained for sensitivity: 96%, 81%, 93% and 90%; and for specificity: 23%, 56%, 28% and 49%. The multicenter score performed well at distinguishing malignant from benign lesions, and was better than the other four traditional scores, for both large and small masses. This was mainly due to the introduction of two criteria that allowed correction for typical dermoids and endohemorrhagic corpora lutea. A completely reliable differentiation of benign from malignant masses cannot be obtained by sonographic imaging alone.