Diagnostic significance of cancer antigen 125/carcinoembryonic antigen ratio combined with IOTA-ADNEX model in discriminating between primary ovarian cancer and ovarian metastases Authors: Yaoqin Wang, Qianyi Zhang, Shixin Huang, Ruoxuan Lin, Lina Tang, Wenting Xie Published: 2026 ABSTRACT BackgroundOvarian metastases (OMs) are relatively common and exhibit unfavorable biological characteristics. The present study sought to evaluate the diagnostic value of the International Ovarian Tumor Analysis (IOTA) Assessment of Different NEoplasias in the adneXa (ADNEX) model, the cancer antigen 125 (CA125)/carcinoembryonic antigen (CEA) ratio, and their combination in distinguishing OM from primary ovarian cancer (OC).MethodsThis retrospective study enrolled patients with histopathologically diagnosed OM and primary OC. OM was classified into three types based on ultrasound (US) findings. The risk of each mass was calculated using the ADNEX model. The two groups’ clinical data, US characteristics, and ADNEX model risk were recorded and compared. The area under the receiver operating characteristic (ROC) curve (AUC) was employed to evaluate the diagnostic performance.ResultsOM originated most frequently from the intestinal tract (58.3%). In total, 52.8% (38/72) of OM cases were identified simultaneously with primary tumors, whereas 47.2% (34/72) were diagnosed after the primary tumor. About 22.2% (16/72) of OM cases were type A, 37.5% (27/72) were type B, and 40.3% (29/72) were type C. The ADNEX model, CA125/CEA ratio, and the combined CA125/CEA ratio of ADNEX were compared between the two groups. The diagnostic performance of the ADNEX combined CA125/CEA ratio was significantly superior to that of the model alone {AUC = 0.886 [95% confidence interval (CI): 0.836–0.936], sensitivity =0.736 (95% CI: 0.619–0.833), accuracy =0.843 (95% CI: 0.788–0.895)}.ConclusionsBoth ADNEX and the CA125/CEA ratio alone can effectively distinguish OM from primary OC. Adding the CA125/CEA ratio to the IOTA-ADNEX model may improve its performance in differentiating between OM and primary OC. Introduction Ovarian metastases (OMs) originate from other malignant tumors and account for 5–20% of ovarian cancers (OCs) (1). OM can arise from various sites, including the gastrointestinal tract, breast, biliary tract, lymphoma, or other gynecologic cancers (2,3). Metastases frequently develop approximately 3 years following the diagnosis of the primary tumor (4). The treatment strategy and prognosis of OM differ from those of primary OC (5,6). Compared with primary OC, the management of OM depends on the characteristics and severity of primary cancer; surgical resection may or may not be considered (7). Patients with OM have a poorer prognosis than those with primary OC. The 5-year survival rate of non-genital tract metastases and primary OC has been reported as 24.0% and 40.0%, respectively (8,9). Therefore, accurate preoperative diagnosis of OM and primary OC is essential, as it facilitates multidisciplinary consultations and enables optimal therapeutic interventions. The assessment of potential metastases associated with ovarian masses depends on the results of serum markers and pelvic imaging techniques (10-12). Previous studies have reported that tumor markers, such as cancer antigen 125 (CA125) and carcinoembryonic antigen (CEA), are significant in distinguishing between OM and primary OC (13). However, the findings of these serum markers alone have some limitations. Research is underway to explore the role of the CA125/CEA ratio in OM and primary OC (14,15). Additionally, a cut-off value of 25 has been identified as a promising biomarker in clinical practice, demonstrating high accuracy. Ultrasound (US) has gained widespread utilization for evaluating women with pelvic tumors due to its accessibility (16). Assessment of Different NEoplasias in the adneXa (ADNEX), a multi-classification model developed by the International Ovarian Tumor Analysis (IOTA), serves as a reliable estimation tool for assessing the risk of pelvic metastatic lesions (17). Previous research has demonstrated that the ADNEX model possesses diagnostic utility in distinguishing between metastatic and primary OC, achieving a moderate diagnostic performance (18,19). The differentiation between primary OC and OM based exclusively on US imaging presents a significant challenge that merits further attention. The CA125/CEA ratio, which provides complementary information beyond the CA125 value within ADNEX (14), may enhance this differentiation. However, the diagnostic value of the CA125/CEA ratio combined with the ADNEX model for the two diseases requires additional investigation. Consequently, the current study explored the diagnostic value of the ADNEX model, the CA125/CEA ratio, and their combination in distinguishing between OM and primary OC. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2026-0759/rc) (20). Methods Study population Among 409 consecutive surgical patients with histologically diagnosed OM or primary OC from January 2017 to January 2024 at Fujian Cancer Hospital, 233 patients were excluded due to missing preoperative CEA or CA125 values, absence of or poor-quality preoperative US, or prior treatment received before US. The remaining 176 patients constituted the final retrospective analytic cohort. Patient data, including age at diagnosis of ovarian masses, menopausal status, laterality, primary tumor, histological subtype, the timing of metastasis (synchronous or metachronous), and tumor biomarkers (CEA and CA125), were extracted from the medical records. For bilateral tumors, the lesion with greater dimensions or more morphologically intricate features underwent subsequent analysis. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Fujian Cancer Hospital (approval No. K2023-232-01). The requirement for informed consent was waived because of the retrospective design of the study. US protocol Ovarian masses were detected using high-end US equipment. The frequency range for abdominal US probes was 3.5–5.0 MHz, whereas that for vaginal US probes was 5.0–9.0 MHz. Experienced sonographers performed all US examinations. The following US characteristics of each adnexal mass were recorded: largest diameter of the lesion (mm), largest diameter of the solid component (mm), number of papillary projections, number of cyst locules, ascites, and acoustic shadows. The ADNEX model was acquired as part of the built-in mobile software. The ADNEX model comprised three clinical predictors, including age (years), type of center (oncology center or non-oncology center), and CA125 level, and six US predictors, including the largest diameter of the lesion (mm), the largest diameter of the solid tissue (mm), over 10 cyst locules in the tumor (yes/no), the number of papillary projections (0, 1, 2, 3, or >3), acoustic shadows (yes or no), and ascites (yes or no). These predictors were input into the model, and then the risk predictive value was calculated as a percentage. The conditional risk of OM in the ADNEX model was calculated using the formula: risk (metastatic)/[risk (metastatic) + risk (primary OC)] (21,22). All US findings of OM were classified into the following three types (23): type A: tumors mimicking primary epithelial ovarian tumor; multilocular cyst with solid components; type B: tumors present as homogeneous or heterogeneous purely solid masses; type C: tumors consisted of solid masses with round or oval cysts. Statistical analysis The software SPSS 21.0 (IBM Corp., Armonk, NY, USA) was used for data analysis. Normally distributed continuous variables were expressed as mean ± standard deviation (SD) and compared using Student’s t-test. Non-normally distributed continuous data were expressed as the median and interquartile range (IQR) and compared using the Mann-Whitney U test. Categorical data were expressed as frequencies and percentages and compared using the Chi-squared test. Patients were divided into the OM and primary OC groups. The CA125/CEA ratio was calculated for each patient. A receiver operating characteristic (ROC) curve was constructed to evaluate the diagnostic performance of the ratio. The literature-based cutoff of 25 was chosen for primary analysis due to its higher specificity compared to the ROC-derived optimal cutoff, as well as its consistency with prior literature. To assess potential redundancy between the ADNEX risk score and the CA125/CEA ratio, we calculated the variance inflation factor (VIF). A VIF value below 5 was considered indicative of no substantial multicollinearity. To achieve a combined diagnosis, we performed a joint ROC analysis based on multivariable logistic regression. We first constructed the regression model with the ADNEX risk score and the CA125/CEA ratio. The continuously predicted probabilities from the model were taken as a diagnostic indicator, and the gold standard was used as the grouping variable to plot the ROC curve. The optimal cutoff value for the combined index was determined using the Youden index. The area under the ROC curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated to evaluate the diagnostic performance, all with their corresponding 95% confidence intervals (CIs). The AUCs were compared using the DeLong test for correlated ROC curves. A P value <0.05 was considered statistically significant. Results Clinical and ultrasonic features A total of 176 patients were included in the study, comprising 72 OM patients with a mean age at diagnosis of 48.6±10.3 years and 104 primary OC patients with a mean age at diagnosis of 54.6±9.6 years. The baseline characteristics of patients in the OM and primary OC groups are summarized in Table 1. Both groups exhibited statistically significant differences in terms of age, laterality, CA125, CEA, and CA125/CEA ratio. The positive rate of bilateral cases was higher in the OM group than in the primary OC group (OM =50%; OC =26%; P=0.001). The median CA125 level was significantly lower in the OM group than in the primary OC group [OC =282.5 (IQR, 101.3–854.2); OM =61.5 (IQR, 21.1–183.8); P<0.001]. The median CEA level was significantly higher in the OM group than in the primary OC group [OC =1.4 (IQR, 0.9–2.4); OM =6.5 (IQR, 1.9–28); P<0.001]. The positive rate of tumors was higher in the OM group than in the primary OC group when the CA125/CEA ratio was <25 (OM =75%; OC =17.3%; P<0.001). There were no significant differences in menopausal status between the two groups. Table 1Clinical and ultrasonic features of primary OC and OMCharacteristicsPrimary OC (n=104)OM (n=72)P valueAge at diagnosis (years)54.6±9.648.6±10.3<0.001Menopausal status0.566 Premenopausal46 (44.2)35 (48.6) Postmenopausal58 (55.8)37 (51.4)Laterality0.001 Unilateral77 (74.0)36 (50.0) Bilateral27 (26.0)36 (50.0)CA125 (U/mL) Median (IQR)282.5 (101.3–854.2)61.5 (21.1–183.8)<0.001 <3510 (9.6)27 (37.5) ≥3594 (90.4)45 (62.5)CEA (ng/mL) Median (IQR)1.4 (0.9–2.4)6.5 (1.9–28)<0.001 <4.791 (87.5)30 (41.7) ≥4.713 (12.5)42 (58.3)CA125/CEA ratio<0.001 ≤2518 (17.3)54 (75.0) >2586 (82.7)18 (25.0)Largest diameter of lesion (mm)101.7±44.4108.9±48.60.315Largest diameter of the solid component (mm)77.3±38.687.7±51.50.129Number of papillary projections<0.001 None86 (82.7)69 (95.8) 11 (1.0)2 (2.8) 22 (1.9)0 (0.0) 30 (0.0)1 (1.4) >315 (14.4)0 (0.0)>10 cyst locules0.340 No92 (88.5)55 (76.4) Yes12 (11.5)17 (23.6)Ascites0.808 No64 (61.5)43 (59.7) Yes40 (38.5)29 (40.3)Acoustic shadows– No104 (100.0)72 (100.0) Yes0 (0.0)0 (0.0)Data are presented as mean ± SD or n (%), unless otherwise specified. CA125, carcinoma antigen 125; CEA, carcinoembryonic antigen; IQR, interquartile range; OC, ovarian cancer; OM, ovarian metastasis; SD, standard deviation. Clinical and ultrasonic features of primary OC and OM Data are presented as mean ± SD or n (%), unless otherwise specified. CA125, carcinoma antigen 125; CEA, carcinoembryonic antigen; IQR, interquartile range; OC, ovarian cancer; OM, ovarian metastasis; SD, standard deviation. The two groups showed statistically significant differences in the number of papillary projections. Approximately 82.7% of the OC and 95.8% of the OM cases had no papillary projections. No statistically significant differences were found between the two groups regarding the largest diameter of the lesion, the largest diameter of the solid component, >10 cyst locules, and ascites. The largest diameter of the lesion was 108.9±48.6 and 101.7±44.4 mm for the OM and primary OC groups, respectively. Neither OM nor primary OC had an acoustic shadow. OM findings As shown in Table 2, the primary sites of OM included the stomach, intestinal tract, bile ducts, gallbladder, breast, cervix, vagina, lung, and bladder. Most metastases originated from the intestinal tract and cervix [58.3% (42/72) and 16.6% (12/72), respectively]. About 52.8% (38/72) of OM cases were simultaneously identified with primary tumors, whereas 47.2% (34/72) were identified after primary tumor diagnosis. The interval time ranged from 2 to 168 months. Among the OM cases of intestinal origin, 27.8% were synchronous, and 30.6% were metachronous. The interval time ranged from 5 to 90 months. Metastases were identified 2 months after the primary tumor in one patient with a history of vagina squamous cell carcinoma. One mass originating from the breast had the longest interval (168 months). Table 2The primary site of OMsPrimary siteSynchronous, n (%)Metachronous, n (%)Total, n (%)Interval time (months)Stomach2 (2.8)5 (6.9)7 (9.7)11–19Intestinal tract20 (27.8)22 (30.6)42 (58.3)5–90Bile ducts1 (1.4)1 (1.4)2 (2.8)46Gallbladder1 (1.4)0 (0.0)1 (1.4)0Breast1 (1.4)3 (4.1)4 (5.6)36–168Cervix11 (15.2)1 (1.4)12 (16.6)26Vagina0 (0.0)1 (1.4)1 (1.4)2Lung2 (2.8)0 (0.0)2 (2.8)0Bladder0 (0.0)1 (1.4)1 (1.4)70Total38 (52.8)34 (47.2)72 (100.0)–OM, ovarian metastasis. The primary site of OMs OM, ovarian metastasis. As shown in Table 3, approximately 22.2% (16/72) of OM cases were type A, 37.5% (27/72) were type B, and 40.3% (29/72) were type C. Metastases originating from the intestine mainly were solid with round or oval cysts (type C; 26.4%). Representative US types of OMs are shown in Figure 1. Table 3The US types of OMsPrimary siteType A, n (%)Type B, n (%)Type C, n (%)Stomach2 (2.8)4 (5.6)1 (1.4)Intestinal tract7 (9.7)16 (22.2)19 (26.4)Bile ducts2 (2.8)0 (0.0)0 (0.0)Gallbladder1 (1.4)0 (0.0)0 (0.0)Breast0 (0.0)2 (2.8)2 (2.8)Cervix3 (4.1)3 (4.1)6 (8.3)Vagina1 (1.4)0 (0.0)0 (0.0)Lung0 (0.0)1 (1.4)1 (1.4)Bladder0 (0.0)1 (1.4)0 (0.0)Total16 (22.2)27 (37.5)29 (40.3)OM, ovarian metastasis; US, ultrasound. The US types of OMs OM, ovarian metastasis; US, ultrasound. Grayscale and color Doppler US images of OM. (A-C) Representative type A case: a 50-year-old woman was diagnosed with intrahepatic cholangiocarcinoma at the time of surgery for OM. US images show a large multilocular-solid mass with smooth margins and good inner US penetration. (D-F) Representative type B case: a pelvic mass was found 17 months after the diagnosis of colorectal carcinoma in a 42-year-old woman with a history of colorectal carcinoma. US findings show a homogeneous, oval, well-defined pure solid mass in the left pelvis. (G-I) Representative type C case: a pelvic mass was found 36 months after the diagnosis of breast carcinoma in a 42-year-old woman with a history of breast invasive ductal carcinoma. US images show bilateral solid masses with round or oval cysts and clear boundaries. OM, ovarian metastasis; US, ultrasound. A subgroup analysis within the OM cohort was performed, comparing the diagnostic performance of the CA125/CEA ratio between gastrointestinal-origin (n=49) and non-gastrointestinal-origin (n=23) OMs. The median (IQR) of the CA125/CEA ratio in gastrointestinal and non-gastrointestinal cases was 4.47 (1.63–17.2) and 11.8 (6.2–96.3), respectively. The two groups had a statistically significant difference in the CA125/CEA ratio (P=0.005). Table 4 depicts the performance of the CA125/CEA ratio in differentiating between gastrointestinal-originated and non-gastrointestinal-originated cancers. The AUC, accuracy, sensitivity, and specificity were 0.708 (95% CI: 0.581–0.834), 0.602 (95% CI: 0.487–0.712), 0.469 (95% CI: 0.337–0.606), and 0.869 (95% CI: 0.679–0.955), respectively. Table 4The performance of the CA125/CEA ratio in differentiating between gastrointestinal-originated cancers and non-gastrointestinal-originated cancersMethodAUC (95% CI)Accuracy (95% CI)Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)CA125/CEA ratio0.708 (0.581–0.834)0.602 (0.487–0.712)0.469 (0.337–0.606)0.869 (0.679–0.955)0.885 (0.719–0.958)0.435 (0.361–0.511)AUC, area under the receiver operating characteristic curve; CA125, carcinoma antigen 125; CEA, carcinoembryonic antigen; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value. The performance of the CA125/CEA ratio in differentiating between gastrointestinal-originated cancers and non-gastrointestinal-originated cancers AUC, area under the receiver operating characteristic curve; CA125, carcinoma antigen 125; CEA, carcinoembryonic antigen; CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value. Performance of IOTA-ADNEX combined with the CA125/CEA ratio The VIF between the ADNEX risk score and the CA125/CEA ratio was 1.115, indicating no substantial multicollinearity. ROC curve analysis of the ADNEX model, CA125/CEA ratio, and their combination in the study population is shown in Figure 2. The diagnostic performance of the three strategies for differentiating OM from primary OC is presented in Table 5. ROC analysis identified an optimal cutoff of 22.87 for the CA125/CEA ratio. When applying the literature-based cutoff of 25 to our data, the sensitivity was 0.750 (95% CI: 0.634–0.844) and the specificity was 0.827 (95% CI: 0.740–0.894). The combined model achieved a significantly higher AUC (0.886; 95% CI: 0.836–0.936) compared to the ADNEX model alone or the CA125/CEA ratio alone (both P<0.05; pairwise comparisons in Table 5), whereas no significant difference was observed between the latter two (P=0.360). Based on the 2×2 contingency table (Table S1), the sensitivity, specificity, and accuracy of the combined model were 0.736 (95% CI: 0.619–0.833), 0.904 (95% CI: 0.830–0.953), and 0.843 (95% CI: 0.788–0.895), respectively. ROC curves for ADNEX, CA125/CEA ratio, and ADNEX and CA125/CEA ratio combination demonstrate efficacy in distinguishing between primary OC and OM. ADNEX, Assessment of Different NEoplasias in the adneXa; CA125, carcinoma antigen 125; CEA, carcinoembryonic antigen; OC, ovarian cancer; OM, ovarian metastasis; ROC, receiver operating characteristic. Table 5Predictive values of the ADNEX, CA125/CEA ratio, and ADNEX combined CA125/CEA ratio in women with primary OC and OMAssessment methodAUC (95% CI)Accuracy (95% CI)Sensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)ADNEX†,§0.826 (0.764–0.887)0.763 (0.700–0.825)0.694 (0.575–0.797)0.807 (0.719–0.878)0.714 (0.621–0.792)0.792 (0.727–0.846)CA125/CEA ratio‡0.788 (0.717–0.860)0.794 (0.734–0.853)0.750 (0.634–0.844)0.827 (0.740–0.894)0.750 (0.659–0.823)0.827 (0.760–0.878)ADNEX combined CA125/CEA ratio0.886 (0.836–0.936)0.843 (0.788–0.895)0.736 (0.619–0.833)0.904 (0.830–0.953)0.841 (0.743–0.907)0.832 (0.770–0.880)†, indicates a significant difference compared with that of ADNEX combined CA125/CEA ratio, z statistic =2.482, P=0.013. ‡, indicates a significant difference compared with that of ADNEX combined CA125/CEA ratio, z statistic =4.611, P<0.0001. §, indicates a significant difference compared with that of CA125/CEA ratio, z statistic =0.914, P=0.360. ADNEX, Assessment of Different NEoplasias in the adneXa; AUC, area under the receiver operating characteristic curve; CA125, carcinoma antigen 125; CEA, carcinoembryonic antigen; CI, confidence interval; NPV, negative predictive value; OC, ovarian cancer; OM, ovarian metastasis; PPV, positive predictive value. Predictive values of the ADNEX, CA125/CEA ratio, and ADNEX combined CA125/CEA ratio in women with primary OC and OM †, indicates a significant difference compared with that of ADNEX combined CA125/CEA ratio, z statistic =2.482, P=0.013. ‡, indicates a significant difference compared with that of ADNEX combined CA125/CEA ratio, z statistic =4.611, P<0.0001. §, indicates a significant difference compared with that of CA125/CEA ratio, z statistic =0.914, P=0.360. ADNEX, Assessment of Different NEoplasias in the adneXa; AUC, area under the receiver operating characteristic curve; CA125, carcinoma antigen 125; CEA, carcinoembryonic antigen; CI, confidence interval; NPV, negative predictive value; OC, ovarian cancer; OM, ovarian metastasis; PPV, positive predictive value. Discussion To the best of our knowledge, no study has assessed the value of the CA125/CEA ratio combined with the ADNEX model in discriminating OM from primary OC. Herein, we compared the ADNEX model, CA125/CEA ratio, and their combination in patients with primary OC and OM. Previous studies indicated that the ADNEX model is effective in distinguishing various types of ovarian tumors from OM (18,24). Epstein et al. performed a study to assess the ADNEX model in differentiating primary OC from metastatic non-OC, which included 143 cases with pelvic tumor malignancy of unknown primary origin. Their findings indicated that the ADNEX model performed well in differentiating primary OC from metastatic OC, with an AUC of 0.891 (95% CI: 0.794–0.946) (25). Our present study found that the diagnostic AUC value of ADNEX, including CA125 in two groups, was 0.826, which is higher than that reported by Stukan et al. (22). This difference may be explained by the different distribution of tumor types across datasets. Our results are consistent with earlier research indicating that the ADNEX model demonstrates high accuracy and specificity but low sensitivity for detecting metastases in patients with breast cancer (26). A previous study reported that the CA125/CEA ratio demonstrated strong efficacy in discriminating between ovarian neoplasms and OM in multilocular masses with five or more locules and multilocular-solid masses, with an AUC of 0.758 (14). The present study yielded comparable results in primary OC and OM, achieving an AUC of 0.788. Furthermore, the combination of ADNEX and the CA125/CEA ratio exhibited good performance in differentiating OM from primary OC, with an AUC of 0.886. A previous study recommended a CA125/CEA ratio threshold of 25 (27); using this established cutoff in our study maintains consistency with prior literature. Our study found that patients with primary OC had higher CA125 values than their OM counterparts. Conversely, a previous study reported that CEA was significantly higher in the OM group than it was in the primary OC group (P<0.001) (28). In the present study, approximately 75% of OM patients had a CA125/CEA ratio below 25, whereas only 17.3% of primary OC patients fell below this threshold. A specificity of 0.827 at the cutoff of 25 indicates that a ratio exceeding 25 strongly suggests primary OC, with a low misclassification rate for true OC cases. This higher specificity is clinically preferred for ruling in primary OC and avoiding inappropriate surgical management of metastatic disease. Meanwhile, 68% of OM cases originated from the gastrointestinal tract, consistent with findings from previous studies (29-31). The absence of a primary renal carcinoma in all cases of our series is consistent with the extreme rarity of OM from renal cell carcinoma (32). Subsequently, OM was classified into multilocular cysts with solid components (type A), purely solid masses (type B), and solid masses with cysts (type C) based on US features. US classification revealed that 40.3% of OM manifested as type C and 37.5% as type B. Metastases from the gastrointestinal tract site were mainly type B and type C (both 27.8%). Metastases can occur synchronously or metachronously with the primary tumor. Patients with a history of malignancy may help diagnose OM from primary OC in clinical practice (33). The present study found that 47.2% of OM patients were identified after primary tumor diagnosis, and one patient had a history of breast cancer with the longest interval time. We also found that synchronous metastases occur more frequently in cervical carcinoma, whereas metachronous metastases to the ovaries are more likely to occur in intestinal tract cancers. OM exhibited a higher prevalence of bilateral involvement than primary OC, which is in line with previous reports (34). Beyond the specific distinction between OM and OC, a broader clinical consideration warrants discussion. Although the present study focuses on discrimination between OM and primary OC, clinicians should recognize that complex adnexal masses may also include benign mimics. Palisciano et al. reported a postmenopausal patient with a large adnexal mass showing suspicious multimodal imaging features, including solid components, papillary projections, and ascites, with malignancy suggested by risk assessment. In contrast, final pathology revealed a benign serous cystadenofibroma (35). This case highlights an important limitation of image-based and model-based triage: suspicious morphology does not always indicate malignancy. Therefore, the CA125/CEA ratio and the ADNEX model should be interpreted as adjunctive tools within a broader diagnostic framework that includes clinical history, metastatic risk from known primary tumors, multimodal imaging review, and histopathology when indicated. Our results should not be interpreted as supporting algorithmic replacement of expert clinical judgment, but rather as suggesting that combined biomarker-US assessment may improve stratification within selected malignant-appearing adnexal masses. Nevertheless, several limitations of the present study should be acknowledged. First, the retrospective design, which may have introduced selection bias, as retrospective data collection may lead to incomplete documentation of clinical indications. Second, the relatively small sample size, particularly with respect to specific OM subtypes, which precludes internal validation methods. Although we were able to perform a subgroup analysis comparing gastrointestinal-origin and non-gastrointestinal-origin OM, the limited sample size of the non-gastrointestinal subgroup means that these results should be considered exploratory. Given that CEA is more commonly elevated in gastrointestinal malignancies, the diagnostic performance observed in this study may be influenced by the predominance of gastrointestinal primaries and may not be directly generalizable to OM of non-gastrointestinal origin. Future studies with larger, more balanced cohorts are needed to validate our findings across different primary tumor types. Third, this study was based on a single-center cohort. External validation in independent, preferably multicenter, cohorts is therefore necessary, as patient characteristics, imaging protocols, and laboratory assays may vary across centers. Conclusions In this retrospective cohort of malignant adnexal lesions, the ADNEX model and the CA125/CEA ratio each demonstrated the ability to distinguish primary OC from OM. Combining the CA125/CEA ratio with the IOTA-ADNEX model further enhanced discriminative performance between the two entities. Accordingly, this combined strategy may represent a valuable alternative for differentiating primary OC from OM. Nonetheless, these findings warrant external validation in prospective, unselected populations before clinical implementation. Supplementary The article’s supplementary files as