Diagnostic performances of the Ovarian Adnexal Reporting and Data System, the Risk of Ovarian Malignancy Algorithm, and the Copenhagen Index in the preoperative prediction of ovarian cancer: a prospective cohort study.

Ovarian cancer (OC) is one of the 3 most common malignancies of the female genital system and one of the most dangerous gynecological tumors. Worldwide, approximately 314,000 new cases of OC were reported, and 207,000 women died from OC in 2020 [ 1 ]. Since clinical symptoms are not specific in the early stages, most women with OC are diagnosed at a late stage, making it challenging for management. The Risk of Ovarian Malignancy Algorithm (ROMA) and Copenhagen Index (CPH-I) are calculated based on the essential biomarkers cancer antigen 125 (CA125) and human epididymal protein 4 (HE4). Many studies have shown that the values of these 2 indicators are similar. In a previous study, the area under the curve (AUC) of ROMA and CPH-I were 0.954 and 0.960, respectively [ 2 ]. Morphological evaluation by ultrasound is important for assessing ovarian tumors. To standardize the terms used to describe ovarian lesions, the Ovarian Adnexal Reporting and Data System (O-RADS) was published in 2018 [ 3 ]. A study by Basha et al. [ 4 ] comparing the malignancy predictive value of O-RADS showed sensitivity (Se) and specificity (Sp) of up to 96.6% and 92.8%, respectively; the AUC of O-RADS was 0.98. Currently, OC stratification strategies based on biomarkers and ultrasound play a fundamental role and are highly valuable. Song et al. [ 5 ] analyzed 3,042 patients with ovarian tumors, and the results showed that a chart combining CPH-I and ROMA with clinical and ultrasound indicators had better predictive values, and CPH-I had the highest predictive value. Several recent studies have investigated the predictive value of the O-RADS in combination with CA125 or HE4 for OC. Wang et al. [ 6 ] showed that the Se, Sp, and accuracy of O-RADS combined with CA125 and HE4 in premenopausal and postmenopausal women were higher than those of O-RADS alone (p<0.05), which enhances the effectiveness of OC diagnosis and is of high value in clinical practice. Our study aimed to assess the diagnostic performance of ROMA, CPH-I, and O-RADS alone and combined O-RADS (O-RADS + CA125 and O-RADS + HE4) for the preoperative prediction of OC.

Of the 462 patients included in the cohort, 381 had benign ovarian tumors, 11 had borderline tumors, and 50 had OC. The main characteristics of the participants are presented in Table 1 . There were significant differences in age and menopausal status between the OC and benign tumor groups (p<0.05). The OC incidence rate was 64% in the menopausal group. Values are presented as number (%). SD, standard deviation. According to the histopathological classification of WHO described in Table 2 , of the 381 cases of benign ovarian tumors, serous cysts, and dermal cysts accounted for the majority (48.6% and 33.9%, respectively). In the borderline ovarian tumor group, borderline mucosal tumors accounted for the highest proportion (9 cases, 81.8%). Of the 50 patients with OC, 35 (57.4%) had serous adenocarcinoma, 3 (6.0%) had mucosal adenocarcinoma, and 4 (8.0%) had dysgerminoma. Table 3 shows the median CA125, HE4, ROMA, and CPH-I values. The median concentrations of CA125 and HE4 in the OC group were significantly higher than those in the benign ovarian tumor group (p<0.05). Similarly, the median CPH-I and ROMA values in the OC group were significantly higher than those in the benign tumor group (p<0.05). Median (Q25%–Q75%). CA125, cancer antigen 125; CPH-I, Copenhagen Index; HE4, human epididymal protein 4; ROMA, Risk of Ovarian Malignancy Algorithm. As shown in Table S1 , 100% of the O-RADS 2 cases were benign, and the incidences of malignancy in the O-RADS groups were 7.7%, 39.7%, and 96.4% for O-RADS 3, O-RADS 4, and O-RADS 5, respectively. No cases were classified as O-RADS 1 (incomplete evaluation). Table 4 , Figs. 1 , 2 , and Fig. S1 shows the predicted ovarian tumor malignancy values for ROMA, CPH-I, O-RADS, O-RADS + HE4, and O-RADS + CA125. Accordingly, the value of O-RADS was higher than that of ROMA and CPH-I, with an AUC of 0.949 (95% CI=0.924–0.968), 0.880 (95% CI=0.846–0.909), and 0.890 (95% CI=0.857–0.918), respectively. When combining O-RADS with HE4 and CA125, the ovarian tumor malignancy predictive value of O-RADS + CA125 was higher than that of O-RADS + HE4 and individual O-RADS, with an AUC of 0.969 (95% CI=0.949–0.983), 0.964 (95% CI=0.942–0.979), and 0.949 (95% CI=0.924–0.968), respectively. Thus, the O-RADS plus CA125 combination model had the highest predictive value among the malignancy predictive values of the investigated models. Values are presented as number (95% confidence interval). AUC, area under the curve; CA125, cancer antigen 125; CPH-I, Copenhagen Index; HE4, human epididymal protein 4; NPV, negative predictive value; O-RADS, Ovarian Adnexal Reporting and Data System; PPV, positive predictive value; ROMA, Risk of Ovarian Malignancy Algorithm; Se, sensitivities; Sp, specificities. CA125, cancer antigen 125; CPH-I, Copenhagen Index; HE4, human epididymal protein 4; O-RADS, Ovarian Adnexal Reporting and Data System; ROMA, Risk of Ovarian Malignancy Algorithm. CA125, cancer antigen 125; CPH-I, Copenhagen Index; HE4, human epididymal protein 4; O-RADS, Ovarian Adnexal Reporting and Data System; ROMA, Risk of Ovarian Malignancy Algorithm. Table S2 presents the logistic regression PI performance of combined ORADS at the optimal cut-off. Accordingly, the optimal cut-off of PI values for O-RADS + CA125 and O-RADS + HE4 are 0.045 and 0.474, respectively.

Our study was conducted to assess the malignancy predictive capability of various algorithms for ovarian tumors. In this study on 462 cases of ovarian tumors, the ROMA, CPH-I, O-RADS alone and combined O-RADS models were all good at predicting the risk of ovarian tumor malignancy. The average age of patients with OC was 52.5±13.9 years, while those in the benign tumor group had an average age of 37.25±14.6 years. This is similar to the results of Tran et al. [ 9 ], which showed 49.3±15.9 years and 36.0±14.9 years, and of Cao et al. [ 10 ] which showed 53.4±11.4 years and 37.7±10.5 years, respectively. In Vietnam and other developing countries, women with OC are often diagnosed in the late stages because of nonspecific symptoms, silent progression, and lack of early screening modalities. Ovarian tumors are divided into 3 major histopathological groups based on the origin of the tumor. Subgroup analysis showed that serous adenocarcinoma was the histological type with the highest proportion (70%) in the ovarian cancer group, followed by dysgerminoma (8%), mucinous adenocarcinoma (6%), malignant Brenner tumor (6%) and others. According to Gupta et al. [ 11 ] research (2019) on 212 ovarian tumors showed that the rate of classification by the origin of ovarian tumors (epithelial cells [71.7%], germ cells [22.2%], sex-cord tumor [3.8%] and other [2.3%]). In our study, the median concentrations of CA125 and HE4 in the OC patient group were significantly higher than those in the benign ovarian tumor group (p<0.05). Similarly, the median CPH-I and ROMA values in the OC group were significantly higher than those in the benign tumor group (p<0.05). In our study, 100% of the O-RADS 2 cases were benign tumors; the higher the O-RADS classification, the higher the incidence of OC. The cancer rates in the O-RADS groups 3, 4, and 5 were 7.7%, 39.7%, and 96.4%, respectively. Results from other groups also showed similar results: the higher the O-RADS, the higher the rate of OC [ 4 10 12 13 14 15 16 17 18 19 ]. According to the O-RADS classification, the cancer rates in each classification group are expected to be O-RADS 2 <1%, O-RADS 3: 1%–10%, O-RADS 4: 10%–50% and O-RADS 5 ≥50% [ 3 ]. In our study, the O-RADS had an excellent OC predictive value with an AUC=0.949 (95% CI=0.924–0.968, p3, Se of 88.5%, and Sp of 89.0%. When comparing the malignant predictive value of O-RADS in some studies, O-RADS had an excellent value in predicting malignant ovarian tumors, such as the studies by Jha et al. [ 17 ] and Cao et al. [ 10 ], where the AUCs were 0.820 and 0.960, respectively (p<0.01). Compared to ROMA and CPH-I in the prediction of ovarian tumor malignancy, O-RADS had better predictive values, with AUCs of 0.949, 0.880, and 0.890 (p<0.05), respectively. Currently, there are a number of studies comparing ROMA and CPH-I in predicting the malignancy risk of ovarian tumors, such as the study by Wang et al. [ 20 ], which showed that the 2 indicators have nearly equal predictive values, with AUCs of 0.810 and 0.807, respectively [ 7 ]. In a study by Tran et al. [ 7 ], both CPH-I and ROMA indicators had good predictive values for ovarian tumor malignancy, with AUCs of 0.898 and 0.882, respectively [ 20 ]. In a study by Gong et al. [ 21 ] in differentiating malignant from benign ovarian masses, the AUCs of HE4, ROMA, and CPH-I increased to 0.946, 0.947, and 0.943, respectively, which were greater than that of CA125 (0.888). When combined with CA125 and HE4 to predict ovarian tumor malignancy, O-RADS + CA125 had higher malignancy predictive values than O-RADS + HE4 or O-RADS alone. The O-RADS + CA125 combination model had the highest predictive value with an AUC of 0.969 (95% CI=0.949–0.983, p<0.01). The O-RADS + HE4 model had an AUC of 0.964 (95% CI=0.949–0.983, p<0.01). Compared with the study by Xie et al. [ 19 ], the O-RADS had good values in the prediction of malignancy with an AUC of 0.804 (95% CI=0.758–0.849) and a Se and Sp of 94.42% and 66.30%, respectively. When combining O-RADS and CA125 to predict malignancy risk, the AUC of this combined model was 0.891 (95% CI=0.858–0.924). In a study by Wu et al. [ 22 ], combining O-RADS and CA125, the diagnostic accuracy for classifying ovarian tumors significantly increased compared with using O-RADS alone (AUCs: 0.97 vs. 0.95, p<0.001 for pre-menopausal and AUCs: 0.93 vs. 0.85, p<0.001 for post-menopausal subjects). In addition, Wang et al. [ 6 ] compared the sensitivities and specificities of O-RADS and O-RADS in combination with CA125 and HE4 in the prediction of ovarian tumor malignancy in premenopausal women. The sensitivities were 92.2% and 94.8%, the specificities were 91.8% and 93.4%, and the accuracies were 91.9% and 93.8%, respectively. In postmenopausal women, the sensitivities of O-RADS and O-RADS in combination with serum CA125 and HE4 were 94.8% and 95.8%, the specificities were 83.9% and 93.6%, and the accuracies were 90.5% and 95.6%, respectively. Based on the collected data, we developed a logistic regression model combining O-RADS with CA125 and HE4, as detailed in the research methodology section. The optimal cut-off threshold for PI is reported in Table S2 offering fundamental metrics for this integrated model. However, it’s essential to note that this study is preliminary and based on a restricted sample size. We believe that there will be many differences between studies in different groups of subjects. Therefore, the PI value derived from the O-RADS combined model serves solely as a reference point for researchers. From the results of our study and some recently published work, it can be assumed that the combination of O-RADS and CA125 helps improve the accuracy of OC prediction. It is a combination of morphological assessment using ultrasonography and a biomarker. According to the recommendations of the International Society of Ultrasound in Obstetrics and Gynecology, transvaginal ultrasound is the first step in the management of pelvic tumors. It is also very popular and widely available in many different countries, health systems, and healthcare facilities. Therefore, executing a strategy for standardized ultrasound skill training for doctors according to the diagnostic criteria of O-RADS will help achieve a better prediction of ovarian tumor malignancy, especially when combined with CA125, the biomarker recommended by the Food and Drug Administration in clinical practice [ 23 ]. Based on our literature search up to May 2023, this is the first study from Vietnam to compare the O-RADS, CPH-I, and ROMA values in OC risk stratification. This study was conducted at 2 gynecological cancer centers in Central Vietnam, and the diagnostic procedures for ovarian tumors were standardized according to the recommendations of international professional societies. The results of this study could contribute to the international literature on the applicability and predictive validity of the O-RADS and the combination of the O-RADS and CA125, helping to increase the value of OC prediction. However, due to the limited number of OC cases, the authors were not able to analyze the value of O-RADS, CPH-I, and ROMA according to each histopathological group. Further studies involving a larger number of OC cases would strengthen the evidence for the use of O-RADS in predicting ovarian malignancy in women presenting with ovarian tumors(s). Incidences of malignancy in the O-RADS 4 and O-RADS 5 groups were 39.7% and 96.4%. The ROMA, CPH-I, O-RADS, O-RADS + CA125, and O-RADS + HE4 models demonstrated good predictive values for ovarian cancer; the combination of O-RADS and CA125 yielded the highest values.

This prospective cohort study included women diagnosed with ovarian tumors who were hospitalized at the Departments of Obstetrics and Gynecology, Hue University of Medicine and Pharmacy Hospital, and Hue Central Hospital between May 2020 and December 2022. Ethical approval for this research was obtained from the Ethical Committee for Biomedical Research at Hue University of Medicine and Pharmacy, Hue, Vietnam (Decision No. H2020/184). Written informed consent was obtained from all the study subjects. 1. Preoperative ovarian cysts or pelvic masses found on pelvic imaging (ultrasound, computed tomography scan, or magnetic resonance imaging). 2. Pathological results obtained and evaluated at our center confirmed benign ovarian disease or OC. The exclusion criteria were preoperative radiotherapy and chemotherapy, the presence of other malignant tumors (such as cervical, breast, and colon cancers), and pregnant women. The sample size was calculated according to the formula to estimate the specificity in 2 steps: Step 1: Calculate FP + TN Step 2: Calculate sample size Abbreviations: FP, false positive; TN, true negative; Z: the “Z” value for confidence interval (CI) of 95% ( Z α 2 2 =1.96 with α=0.05); w (errors)=0.03; p sp , the specificity from the study of Wang et al. [ 6 ] is 0.936. p dis , the prevalence rate, according to GLOBOCAN [ 1 ], the prevalence rate of OC in Vietnam ( p dis ) is 2.4 cases/100,000 women=0.000024. N sp , the minimum sample size for specificity. The calculated minimum sample size was 131 patients. A total of 462 patients who met the inclusion criteria were enrolled. The research steps included administrative interviews, medical histories, and clinical examinations conducted by gynecologists. Subsequently, transabdominal pelvic and transvaginal ultrasounds were performed by a gynecologist/radiologist with more than 5 years of experience in the field. All participating radiologists were self-trained and standardized by using the O-RADS classification system ( https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/O-Rads ). Based on the O-RADS guidelines, 6 categories were used for risk classification. These included O-RADS 0 (incomplete evaluation), O-RADS 1 (physiologic category, including a normal premenopausal ovary), O-RADS 2 (almost certainly benign category, < 1% risk of malignancy), O-RADS 3 (low risk of malignancy, 1%–<10%), O-RADS 4 (intermediate risk of malignancy, 10%–50%), and O-RADS 5 (high risk of malignancy, ≥50%) [ 3 ]. CA125 and HE4 serum testing was conducted using a Luminescent Electrochemical Immunoassay machine on a COBAS 6000 system (Roche, Rotkreuz, Switzerland) at Hue University Hospital of Medicine and Pharmacy and by a chemiluminescent microparticle immunoassay machine on an Architect i1000 system (Abbott Diagnostics, Abbott Park, IL, USA) at Hue Central Hospital. The test results were controlled using an Internal Quality Control system with RANDOX standard control samples and programs. CPH-I predicts the risk of preoperative ovarian tumor malignancy according to the following algorithm [ 2 ]: The ROMA index was calculated to predict the risk of ovarian tumor malignancy before surgery using the following algorithm [ 7 ] PI is the Prediction Index, determined as follows: The ORADS stratification system is combined with biomarkers CA125 and HE4 by establishing a logistic regression equation as below: For ORADS + CA125: For ORADS + HE4: Surgical removal or tumor resection was indicated on a case-by-case basis. Histopathological testing was based on the World Health Organization (WHO) 2014 criteria and classifications [ 8 ]. Based on these 2 indicators, the patients were classified into a benign tumor group or a malignant tumor group (borderline and malignant tumors). Finally, the parameters were matched with the histopathological results (including benign, borderline, and malignant ovarian tumors) to calculate the diagnostic values of CPH-I, ROMA, O-RADS, and combined O-RADS. Data analysis was performed using the statistical software SPSS (version 20.0; SPSS, Inc., Chicago, IL, USA), and ROC_AUCs analysis was performed using MedCalc. Classification variables were reported as numeric (percentage) and continuous variables as medians (standard deviation; confidence interval [CI]). Categorical data were compared using the chi-square test. Continuous data were compared using one-way analysis of variance or the Mann-Whitney U test, depending on the data distribution. The p<0.05 was considered statistically significant.