Introduction/Background Among patients with adnexal masses, preoperative identification of epithelial ovarian cancer (EOC) or metastatic cancer in the ovary (MCO) is essential for surgical planning. Our aim was to assess the performance of CA125, HE4, and the probability models, Risk of Ovarian Malignancy Algorithm (ROMA) and Copenhagen Index (CPH-I), to preoperatively identify EOC or MCO.
Methodology We performed a single center retrospective study including women who underwent surgery for an ovarian tumor between January 2000 and December 2018. We defined two study groups: one group comprising women with benign pathology and borderline epithelial ovarian tumors and a second group comprising women with EOC or MCO. We computed sensitivity, specificity and predictive values of CA125, HE4, ROMA and CPH-I at different cutoff points. We performed receiver operative curve analysis for tumor markers, CPH-I and ROMA models. We performed subgroup analysis including only premenopausal, postmenopausal women, stage I EOC and women harboring ovarian tumors with inconclusive diagnosis of malignancy by ultrasound features.
Results One thousand seventy-one patients were included, 852 (79.55%) presented benign or borderline epithelial tumors and 219 (20.45%) presented EOC or MCO. Area under the curve (AUC) for HE4 was significantly higher than AUC for CA125 (0.909 vs 0.873). No differences were seen between AUC of ROMA (0.939) and CPH-I (0.936), but they were both higher than HE4 AUC (figure 1). Subgroup analysis showed that in premenopausal women, HE4 performed better than CA125, and was equivalent to ROMA or CPH-I. Considering only ovarian tumors with inconclusive diagnosis by ultrasound, ROMA performed better than CPH-I. None of the tumor markers alone achieved a sensitivity of 90%, however HE4 was highly specific (93.5%). ROMA showed a sensitivity and specificity of 91.1% and 84.6% respectively, while CPH-I showed a sensitivity of 91.1% with 79.2% specificity (table 1).
Conclusion Both ROMA and CPH-I were more favorable than tumor markers alone for differentiating patients harboring EOC or MCO. Probability models can be helpful in order to assess the risk of malignancy of ovarian tumors, especially when expert ultrasound examination is not available or when the diagnosis by ultrasound remains inconclusive.
Disclosures The authors of this abstract have no disclosures.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.