Discriminative value of CA-125, HE4, Risk of Malignancy Index II (RMI-II) and Risk of Malignancy Algorithm (ROMA) in the differential diagnosis of pelvic masses: conclusions from a referral Centre in Portugal.

The major purpose of this article was to compare the discriminative value of different algorithms and serum biomarkers in the differential diagnosis of adnexal masses. We performed a retrospective study with 247 women with adnexal neoplasia, submitted to surgical treatment and with a histological diagnosis. The evaluation of the area under the curve (AUC) for isolated CA-125 and HE4, and for ROMA and RMI-II, showed a better specificity of HE4 and RMI-II in premenopausal women. In the postmenopausal group, ROMA and RMI-II were the algorithms with a better performance. Impact Statement What is already known on this subject? CA-125 remains the most commonly used biomarker used to predict the behaviour of an adnexal mass, but it has a low sensitivity for stage I tumours. Other isolated serum markers have emerged more recently, such as HE4, as well as more complex algorithms, such as RMI or ROMA. It remains unclear which is the best marker/algorithm to predict the behaviour of an adnexal mass. What do the results of this study add? Our findings showed that ROMA is a suitable marker for postmenopausal women, with no advantage found in the premenopausal women when compared with an isolated HE4. What are the implications of these findings for clinical practice and/or further research? The different algorithms of the preoperative discrimination of ovarian neoplasia appear to have different AUC, SN and SP in the pre- or the postmenopausal patients. For the premenopausal women, the use of ROMA does not seem to have any advantage over the isolated use of HE4, which does not lose specificity even when the borderline tumours are considered for discrimination. In the postmenopausal women, ROMA is a valid algorithm with a good sensitivity. The RMI-II showed a good performance in both groups, although it depends on the ultrasound findings and has an important interobserver variability. This information allows a more targeted selection of markers and algorithms to be requested prior to surgery of ovarian neoplasms regarding the menopausal status of each patient.