Assessment of diagnostic utility of multivariate diagnostic models in differential diagnosis of ovarian tumors.

Abstract

Ovarian cancer (OC) diagnosis remains a clinical challenge due to lack of early symptoms and insufficient accuracy of the available diagnostic methods. The purpose of this study was to determine whether osteopontin could be useful in differential diagnosis of ovarian tumors. Serum samples from 92 patients qualified for surgical treatment due to ovarian mass were divided into 2 groups according to the histopathological result: OC including borderline ovarian tumors (n = 39) and benign ovarian tumors (BOTs) (n = 53). CA125, HE4 and osteopontin concentrations were measured in all patients. Areas under the receiver operating characteristic curves (AUC of ROC) were used to compare the discriminative ability of the univariate and multivariate diagnostic models. The addition of osteopontin to ROMA significantly improved the diagnostic performance of the test in 3 of the 5 analyses: 1) in the OC vs BOT group (from AUC of 0.955 to 0.975), 2) in premenopausal women OC vs BOT (from AUC of 0.828 to 0.892) and 3) in the FIGO I-II stage OC vs BOT (from AUC of 0.865 to 0.895). It did not alter the diagnostic performance of multifactor tests in the group of postmenopausal women nor in OC FIGO III-IV stage group. Osteopontin was also the best single marker to differentiate between early stage OC and BOTs (AUC of 0.863). Osteopontin improves the diagnostic performance of a multimarker OC diagnostic test and could be useful in differential diagnosis of ovarian tumors, especially in pre-menopausal women and for early stage OC.