Abstract
Background
This study evaluated the diagnostic performance of CDO1 and ZSCAN12 methylation in paired endometrial (Em) and exfoliated cervical (Cx) samples for detecting endometrial cancer (EC) and endometrial atypical hyperplasia (EAH).
Methods
We analysed 127 histologically confirmed women (43 EC, 13 EAH, 71 benign). Methylation levels were measured using bisulfite-conversion real-time PCR. Agreement, robustness, and diagnostic accuracy were assessed.
Results
Methylation levels increased stepwise from benign to EC across all sample types (all p < 0.001). For EC detection, CDO1m_Em and ZSCAN12m_Em achieved AUCs of 0.858 and 0.832, with ORs of 21.0 and 19.7, respectively. Bootstrap validation confirmed robustness. For the same sample source (Em or Cx), the clinical performance of CDO1m and ZSCAN12m was comparable (all p > 0.05). However, significant differences were noted between Em and Cx samples for ZSCAN12m’s negative predictive value (NPV) in EC detection (p = 0.045), and for CDO1m’s sensitivity, NPV, and negative likelihood ratio (nLR) in EAH/EC detection (p = 0.029, 0.026, and 0.031, respectively). The Em-based two-gene combination showed 97.5% sensitivity for EC (39 of 40 evaluable cases), failing to detect only 1 of 40 cases.
Conclusions
CDO1 and ZSCAN12 methylation demonstrated robust diagnostic performance for EC and EAH. The Em-based two-gene combination showed excellent sensitivity. Most diagnostic metrics were comparable between sample sources, with Em showing advantages specifically in NPV-related measures. Given that these differences may be assay-dependent and that most performance metrics showed no significant differences, cervical sampling remains a promising less invasive option but warrants cautious interpretation and further validation.
Keywords: Endometrial cancer, methylation, diagnostic accuracy, bootstrap validation, endometrial atypical hyperplasia
HIGHLIGHTS
CDO1 and ZSCAN12 methylation levels increase stepwise from benign to EAH to EC across endometrial and cervical samples.
The endometrial-based two-gene combination (CDO1m_Em/ZSCAN12m_Em) detects EC with 97.5% sensitivity, missing only 1 of 40 cases.
Most diagnostic metrics showed no significant differences between sample sources, with Em showing advantages specifically in NPV-related measures; cervical sampling remains promising but requires cautious interpretation.
Background
Endometrial cancer (EC) is one of the most common gynaecological malignancies worldwide [1,2], with an increasing incidence partly attributed to rising obesity rates and an aging population [3]. Early detection of EC and its precursor, endometrial atypical hyperplasia (EAH), is crucial for improving patient outcomes and reducing the need for aggressive surgical interventions [4,5]. Currently, the gold standard for diagnosis involves endometrial biopsy or curettage followed by histopathological examination [6]. However, these procedures are invasive, often painful, and may be associated with complications such as bleeding, infection, and uterine perforation. Consequently, there is a growing interest in developing less invasive or non-invasive diagnostic tools that can reliably detect EC and EAH from alternative sample sources.
Transvaginal ultrasonography (TVUS) is commonly used as a first-line screening tool, primarily measuring endometrial thickness (ET) [7]. While a thickened endometrium (TedE, e.g. >11 mm in premenopausal or >5 mm in postmenopausal women) raises suspicion for malignancy, TVUS has limited specificity, often leading to unnecessary invasive procedures [8]. Therefore, more accurate and non-invasive biomarkers are urgently needed to improve early diagnosis and reduce the burden of invasive sampling.
DNA methylation has emerged as a promising biomarker for cancer detection [9,10]. Aberrant promoter hypermethylation of tumour suppressor genes occurs frequently in EC and can be detected in various biological specimens, including exfoliated cells from the endometrium and cervix [11]. Among the many methylation markers, CDO1 (cysteine dioxygenase type 1) and ZSCAN12 (zinc finger and SCAN domain containing 12) have shown potential for EC detection in preliminary studies [12–14]. However, whether methylation levels of these two genes from exfoliated cervical (Cx) samples perform comparably to those from paired endometrial (Em) samples remains largely unexplored. Moreover, the diagnostic performance of these markers for distinguishing EAH from benign lesions or EC has not been systematically compared in paired sample settings. Understanding the agreement between Em and Cx samples is essential for validating cervical sampling as a viable, less invasive alternative for EC and EAH detection.
To address these gaps, this study was designed to: (1) evaluate the correlation and agreement of CDO1 methylation (CDO1m) and ZSCAN12 methylation (ZSCAN12m) levels between paired Em and Cx samples; (2) assess the diagnostic accuracy and robustness of single and dual methylation markers for detecting EC and EAH; and (3) compare the clinical performance of Em-based versus Cx-based methylation assays, providing evidence for the cautious use of cervical samples in clinical practice.
Methods
Study design
This case–control study enrolled women who underwent endometrial curettage or surgical treatment at the hospital due to abnormal or dysfunctional uterine bleeding, abnormal transvaginal ultrasonography findings, or abnormal palpation findings between October 2024 and October 2025. The study was approved by the Research and Clinical Trials Ethics Committee (approval number: EC-20240913-01) and complied with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants. Inclusion criteria were as follows: (1) patients were aged 24–86 years; (2) patients without systemic diseases or other malignant tumours except for EC. Exclusion criteria were as follows: (1) clinical information was incomplete or non-traceable; (2) patients who are unwilling to sign the informed consent. Importantly, none of the enrolled women had undergone prior hysteroscopy, endometrial curettage, or other intrauterine procedures before study entry and methylation sampling. The Em and Cx brush samples were collected immediately prior to the definitive diagnostic or therapeutic procedures (curettage or surgery), ensuring that the methylation measurements reflect the native endometrial cell population at the time of diagnosis.
A total of 127 subjects were included in this study. Clinical data for each subject, including age, endometrial thickness measured by ultrasonography, histological type of lesion or tumour, and International Federation of Gynaecology, Obstetrics (FIGO) stage, and histopathological diagnosis were obtained from hospital records.
Specimen collection
Prior to endometrial curettage or surgery, samples from different sources were collected in two sequential steps by standardized trained professional gynaecologists. (1) Exfoliated cervical cells (Cx) were obtained using a Rovers Cervex-Brush (Rovers Medical Devices, Oss, the Netherlands). Briefly, the brush was inserted into the cervical os and rotated clockwise five times (360° per rotation) to ensure adequate sampling of both the ectocervix and endocervical canal. The brush head was then immediately rinsed into a vial containing PreservCyt solution (Hologic, Bedford, MA, USA) with vigorous rotation for 30 s, after which the vial was tightly capped. (2) Endometrial cells (Em) were collected with the patient in the lithotomy position. After disinfection of the vulva and vagina, the intrauterine brush [15] was gently inserted to the level of the uterine fundus, rotated four to five revolutions in the same direction, and then withdrawn. The brush was immersed into a vial containing PreservCyt solution. All samples were stored at 4 °C, and DNA extraction was completed within seven days. The extracted DNA was preserved at −20 °C.
DNA methylation determination
DNA methylation analysis was carried out in a certified laboratory by personnel blinded to all clinical information. CDO1 and ZSCAN12 methylation were assessed using the same specimens collected in PreservCyt solution. Genomic DNA (gDNA) was extracted from samples using a DNA isolation and purification kit (Hoomya, Changsha, China) following the manufacturer’s protocol. DNA concentration was measured with a NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific, DE, USA). Briefly, 200 ng of gDNA was bisulfite-converted using a DNA methylation pretreatment reagent (Hoomya, Changsha, China). CDO1 and ZSCAN12 methylation were then detected using human CDO1 and ZSCAN12 gene methylation detection kits (Hoomya, Changsha, China), with GAPDH as an internal control, on an ABI 7500 real-time PCR System platform (Life Technology, Foster City, CA, USA). Methylation levels of CDO1 and ZSCAN12 were expressed as ΔCp values calculated as follows: ΔCpCDO1 = CpCDO1 − CpGAPDH and ΔCpZSCAN12 = CpZSCAN12 − CpGAPDH. A lower ΔCp value corresponds to a higher methylation level, and vice versa.
Statistical analyses
Sample size was calculated using PASS 15 Power Analysis and Sample Size Software (NCSS, LLC. Kaysville, UT, USA) based on the area under the receiver operating characteristic curve (AUC) and its 95% confidence interval (CI). According to a recent review, the area under the curve (AUC) of most single-gene methylation markers for diagnosing endometrial cancer exceeds 0.75 [11]. Given that a higher AUC indicates better diagnostic performance, the required sample size decreases as the AUC increases. A two-sided 95% CI width (δ) of 0.2 was set. Under these parameters, a sample of 39 subjects (33.3%) from the positive population (EC) and 78 subjects (66.7%) from the negative population (non-EC) was required to achieve a two-sided 95% CI width of 0.2 when the sample AUC was 0.750.
All analyses were performed using R software (version 4.5.1, R Foundation for Statistical Computing, Vienna, Austria). Two-sided tests were conducted, and a P-value < 0.05 was considered statistically significant. Normality of continuous variables was assessed using the Shapiro–Wilk test (stats::shapiro.test()). Variables following a normal distribution were presented as mean (standard deviation, SD), while non-normally distributed variables were presented as median (interquartile range, IQR). Categorical variables were reported as frequencies (%).
Spearman’s rank correlation coefficient for DNA methylation ΔCp values of the same gene from different sample sources was calculated using the stats::cor.test() function with method = ‘spearman’. Correlation scatter plots and Bland-Altman plots for agreement analysis were generated using ggplot2::ggplot(). The trend in DNA methylation across endometrial lesion severity was assessed using Cuzick’s test (PMCMRplus::cuzickTest()). The distribution of DNA methylation levels across different endometrial pathologies was visualized using ggpubr::ggviolin(), and intergroup differences were assessed using the Wilcoxon test. Receiver operating characteristic (ROC) curves were plotted using pROC::ggroc(). The optimal cutoff value was determined based on the maximum Youden index, with the selection of positive cases determined by the differences between EBL and EAH, and between EAH and EC. The cutoff value for CDO1m_Em was 5.59; ΔCp ≤ 5.59 was interpreted as hypermethylation (positive), while ΔCp > 5.59 was interpreted as hypomethylation (negative). The cutoff value for CDO1m_Cx was 9.28; ΔCp ≤ 9.28 was defined as hypermethylation (positive), and ΔCp > 9.28 as hypomethylation (negative). The cutoff value for ZSCAN12m_Em was 5.28; ΔCp ≤ 5.28 indicated hypermethylation (positive), and ΔCp > 5.28 indicated hypomethylation (negative). The cutoff value for ZSCAN12m_Cx was 7.60; ΔCp ≤ 7.60 was considered hypermethylation (positive), whereas ΔCp > 7.60 was considered hypomethylation (negative). The robustness of the AUC was assessed using the Bootstrap method, and the results were visualized with ggplot2::ggplot().
Agreement between binary DNA methylation variables was evaluated using vcd::Kappa() to calculate the kappa coefficient with 95%CI, and a forest plot of kappa values was generated using ggplot2::ggplot(). On ultrasonography, proliferative-phase endometrial thickness > 11 mm (or > 5 mm in postmenopausal women) was defined as thickened endometrium (TedE), which was considered a positive test result [8]. To improve sensitivity, combination of two genes (Gene A/Gene B) was interpreted using a “positive by trust” principle: positivity for either gene was classified as positive, and negativity for both genes was classified as negative. Univariate logistic regression analysis was performed using stats::glm() to calculate odds ratios (ORs) for DNA methylation markers. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (pLR), and negative likelihood ratio (nLR) with their 95%CIs were calculated using DTComPair::acc.paired(). Differences in sensitivity and specificity between different tests were assessed using DTComPair::sesp.mcnemar(). Differences in PPV and NPV were evaluated using DTComPair::pv.gs(). Differences in PLR and NLR were tested using DTComPair::dlr.regtest().
Results
Subject characteristics and agreement between sample types
This study included 127 histologically confirmed subjects: 71 with endometrial benign lesion (EBL), 13 with endometrial atypical hyperplasia (EAH), and 43 with endometrial cancer (EC). Age was normally distributed (Shapiro-Wilk test, p = 0.1482), with a mean of 48.5 ± 12.0 years. Endometrial thickness measurements were missing for 23 subjects (Supplementary Table 1). DNA methylation detection failed for seven specimens due to insufficient DNA quantity (2 Em and 6 Cx). ΔCp values deviated from normality (Shapiro-Wilk test, all p < 0.001). The median ΔCp values for CDO1m_Em and CDO1m_Cx were 6.27 (IQR: 3.42–10.97) and 16.12 (IQR: 6.91–19.96), respectively; for ZSCAN12m_Em and ZSCAN12m_Cx, the median ΔCp values were 5.70 (IQR: 1.74–9.53) and 13.55 (IQR: 5.85–20.00), respectively. Other subject characteristics are shown in Table 1. FIGO stage was missing for 16 of the 43 EC cases (37.2%). This missingness was non-random and primarily reflects cases where complete surgical staging information was not available at the time of data extraction, often due to referral from other institutions or cases managed without complete staging surgery (e.g. due to comorbidities). This may result in underrepresentation of early-stage disease in our cohort and should be considered when interpreting the stage distribution. For the dual-gene combination analyses, three of the 43 EC cases were excluded due to DNA methylation detection failures (1 Em sample and 2 Cx samples failed), resulting in 40 evaluable EC cases for the paired comparisons.
Table 1.
Characteristics of cases.
| Characteristic | level | Overall | EBL | EAH | EC |
|---|---|---|---|---|---|
| n | 127 | 71 | 13 | 43 | |
| Age (mean (SD)) | 48.5 (12.0) | 43.1 (9.8) | 49.2 (12.3) | 57.3 (10.0) | |
| ET (median (IQR)) | 11.0 (9.0–15.0) | 11.0 (8.6–14.7) | 12.0 (9.2–16.0) | 11.0 (9.0–16.5) | |
| TedE (%) | No | 34 (26.8) | 28 (39.4) | 1 (7.7) | 5 (11.6) |
| Yes | 70 (55.1) | 39 (54.9) | 9 (69.2) | 22 (51.2) | |
| NA | 23 (18.1) | 4 (5.7) | 3 (23.1) | 16 (37.2) | |
| Menopause (%) | No | 79 (62.2) | 60 (84.5) | 9 (69.2) | 10 (23.3) |
| Yes | 48 (37.8) | 11 (15.5) | 4 (30.8) | 33 (76.7) | |
| Specimen of pathology (%) | Curettage | 85 (66.9) | 69 (97.2) | 7 (53.8) | 9 (20.9) |
| Surgery | 42 (33.1) | 2 (2.8) | 6 (46.2) | 34 (79.1) | |
| Histotype (%) | Adenocarcinoma | 39 (30.7) | / | / | 39 (90.7) |
| Clear cell carcinoma | 2 (1.6) | / | / | 2 (4.7) | |
| Neuroendocrine carcinoma | 1 (0.8) | / | / | 1 (2.3) | |
| Sarcoma | 1 (0.8) | / | / | 1 (2.3) | |
| Non-EC | 84 (66.1) | / | / | / | |
| Differentiation (%) | G1 | 7 (5.5) | / | / | 7 (16.3) |
| G2 | 12 (9.4) | / | / | 12 (27.9) | |
| G3 | 10 (7.9) | / | / | 10 (23.2) | |
| NA | 14 (11.1) | / | / | 14 (32.6) | |
| Non-EC | 84 (66.1) | / | / | / | |
| FIGO (%) | I | 16 (12.6) | / | / | 16 (37.2) |
| II | 4 (3.2) | / | / | 4 (9.3) | |
| III | 7 (5.5) | / | / | 7 (16.3) | |
| NA | 16 (12.6) | / | / | 16 (37.2) | |
| Non-EC | 84 (66.1) | / | / | / | |
| CDO1m_Em (ΔCp, median (IQR)) | 6.27 (3.42–10.97) | 8.98 (6.43–16.1) | 5.30 (4.50–17.67) | 3.03 (1.88–4.56) | |
| CDO1m_Em | ΔCp > 5.59 | 71 (55.9) | 59 (83.1) | 6 (46.2) | 6 (14.0) |
| ΔCp ≤ 5.59 | 54 (42.5) | 11 (15.5) | 7 (53.8) | 36 (83.7) | |
| NA | 2 (1.6) | 1 (1.4) | 0 (0.0) | 1 (2.3) | |
| ZSCAN12m_Em (ΔCp, median (IQR)) | 5.70 (1.74–9.53) | 7.86 (5.54–13.60) | 4.93 (2.39–6.97) | 1.72 (0.95–3.43) | |
| ZSCAN12m_Em | ΔCp > 5.28 | 66 (42.0) | 55 (77.5) | 6 (46.2) | 5 (11.6) |
| ΔCp ≤ 5.28 | 59 (46.4) | 15 (21.1) | 7 (53.8) | 37 (86.1) | |
| NA | 2 (1.6) | 1 (1.4) | 0 (0.0) | 1 (2.3) | |
| CDO1m_Cx (ΔCp, median (IQR)) | 16.12 (6.91–19.96) | 19.06 (12.26–20.41) | 18.31 (13.92–19.96) | 6.71 (3.70–9.11) | |
| CDO1m_Cx | ΔCp > 9.28 | 77 (60.6) | 56 (78.9) | 11 (84.6) | 10 (23.3) |
| ΔCp ≤ 9.28 | 44 (34.7) | 11 (15.5) | 2 (15.4) | 31 (72.0) | |
| NA | 6 (4.7) | 4 (5.6) | 0 (0.0) | 2 (4.7) | |
| ZSCAN12m_Cx (ΔCp, median (IQR)) | 13.55 (5.85–20.00) | 18.64 (10.11–20.57) | 5.85 (5.60–18.86) | 6.30 (4.18–9.87) | |
| ZSCAN12m_Cx | ΔCp > 7.60 | 77 (60.6) | 58 (81.7) | 6 (46.2) | 13 (30.2) |
| ΔCp ≤ 7.60 | 44 (34.7) | 9 (12.7) | 7 (53.8) | 28 (65.1) | |
| NA | 6 (4.7) | 4 (5.6) | 0 (0.0) | 2 (4.7) |
EC: Endometrial cancer; EAH: Endometrial atypical hyperplasia; EBL: Endometrial benign lesion; SD: Standard deviation; IQR: Interquartile range; ET: Endometrial thickness; TedE: Thickened endometrium, ultrasound findings showing a proliferative phase endometrial thickness >11 mm (or >5 mm in postmenopausal women) are considered thickened endometrium; NA: Not Available; FIGO: Federation Internationale Of Gynaecologie And Obstetrigue; Em: Sample source: endometrial cells; Cx: Sample source: exfoliated cervical cells; CDO1m: CDO1 gene methylation; ZSCAN12m: ZSCAN12 gene methylation; ΔCp: difference of cross points.
The agreement between Em and Cx for CDO1m and ZSCAN12m methylation ΔCp values was assessed using Bland–Altman analysis across different pathological groups (Figure 1). Figure 1(A,B) display the Bland–Altman plots for CDO1m and ZSCAN12m in all cases, respectively. The bias (mean difference) between the two sample types was 5.23 for CDO1m and 6.06 for ZSCAN12m, suggesting that Em samples yielded systematically higher methylation levels compared to Cx samples. Subgroup analyses by pathological diagnosis (EC, EAH, and EBL) are shown in Figure 1(C–H). Consistent with this, Spearman correlation analysis (Supplementary Figure 1(A,B)) revealed moderate positive correlations between Em and Cx samples for both genes across all patients (CDO1m: ρ = 0.5, p < 0.001; ZSCAN12m: ρ = 0.5, p < 0.001). Supplementary Figure 1 C–H present the corresponding Spearman correlation analyses for each subgroup.
Figure 1.
Bland-Altman plots of two genes between different sample sources (Em: endometrial cells; Cx: exfoliated cervical cells). (A–B) Bland–Altman plots of CDO1m (A) and ZSCAN12m (B) in all cases; (C–D) Bland–Altman plots of CDO1m (C) and ZSCAN12m (D) in EC cases; (E–F) Bland–Altman plots of CDO1m (E) and ZSCAN12m (F) in EAH cases; (G–H) Bland–Altman plots of CDO1m (G) and ZSCAN12m (H) in EBL cases.
EC: Endometrial cancer; EAH: Endometrial atypical hyperplasia; EBL: Endometrial benign lesion; Em: Sample source: endometrial cells; Cx: Sample source: exfoliated cervical cells; CDO1m, CDO1 gene methylation; ZSCAN12m, ZSCAN12 gene methylation
Diagnostic performance and robustness of methylation markers
Figure 2 illustrates the distribution and diagnostic performance of CDO1m and ZSCAN12m methylation levels in Em and Cx samples across different histological outcomes. The distribution plots (Figure 2(A,D,G,J)) showed a stepwise increase in methylation levels from EBL to EAH to EC (all p < 0.001, Cuzick’s test). CDO1m_Em and CDO1m_Cx levels were differed significantly between EAH and EC patients (all p < 0.01, Mann–Whitney U-test, Figure 2(A,G)). ZSCAN12m_Em and ZSCAN12m_Cx levels were significantly different between EBL and EAH patients (all p < 0.05, Mann–Whitney U-test, Figure 2(D,J)). ROC analysis for detecting EAH/EC (Figure 2(B,E,H,K)) yielded AUCs of 0.822 (95%CI = 0.744–0.900) for ZSCAN12m_Em and 0.751 (95%CI = 0.660–0.842) for ZSCAN12m_Cx. For EC detection (Figure 2(C,F,I,L)), CDO1m_Em and CDO1m_Cx achieved AUCs of 0.858 (95%CI = 0.785–0.932) and 0.815 (95%CI = 0.728–0.901), respectively. ZSCAN12m_Em achieved an AUC of 0.832 (95%CI = 0.755–0.908), while ZSCAN12m_Cx achieved an AUC of 0.738 (95%CI = 0.639–0.837). The ORs for EC associated with hypermethylation (vs. hypomethylation) of CDO1m_Em, CDO1m_Cx, ZSCAN12m_Em, and ZSCAN12m_Cx were 21.0 (95%CI = 7.6–58.2), 15.5 (95%CI = 6.1–39.2), 19.7 (95%CI = 6.8–56.8), and 8.3 (95%CI = 3.5–19.6), respectively (Supplementary Table 2).
Figure 2.
Distribution and ROC plots of two genes (CDO1m and ZSCAN12m) across two sample sources (Em: endometrial cells; Cx: exfoliated cervical cells). (A, D, G, J) Distribution plots of gene methylation ΔCp values across different pathological outcomes for CDO1m_Em (A), ZSCAN12m_Em (D), CDO1m_Cx (G), and ZSCAN12m_Cx (J), respectively; (B, E, H, K) ROC curves for detecting EAH/EC using gene methylation for CDO1m_Em (B), ZSCAN12m_Em (E), CDO1m_Cx (H), and ZSCAN12m_Cx (K), respectively; (C, F, I, L) ROC curves for detecting EC using gene methylation for CDO1m_Em (C), ZSCAN12m_Em (F), CDO1m_Cx (I), and ZSCAN12m_Cx (L), respectively. Comparisons between two groups were analysed by the Mann–Whitney U-test, NS., p ≥ 0.05; *, p < 0.05; **, p < 0.01; ***, p < 0.001.
EC: Endometrial cancer; EAH: Endometrial atypical hyperplasia; EBL: Endometrial benign lesion; Em: Sample source: endometrial cells; Cx: Sample source: exfoliated cervical cells; CDO1m, CDO1 gene methylation; ZSCAN12m: ZSCAN12 gene methylation; ROC: Receiver operator characteristic curve; AUC: Area under the ROC.
Figure 3 evaluates the agreement between Em and Cx samples and the robustness of methylation markers. As shown in Figure 3(A), ZSCAN12m showed a higher kappa value than CDO1m in all cases (0.478 vs. 0.439), whereas CDO1m outperformed ZSCAN12m in EBL (0.428 vs. 0.266). Agreement for both CDO1m and ZSCAN12m between sample types was not statistically significant in EC or EAH (all p > 0.05). Bootstrap analysis further confirmed the robustness of these markers (Figure 3(B,C)). For EC detection (Figure 3(B)), the bootstrap-resampled mean AUC for CDO1m_Em was 0.857 (95%CI = 0.776–0.923), closely matching its original AUC of 0.858. Similarly, for EAH/EC detection (Figure 3(C)), ZSCAN12m_Em yielded a bootstrap mean AUC of 0.821 (95%CI = 0.740–0.893), consistent with the original estimate of 0.822, with narrow confidence intervals supporting the precision of these estimates (Supplementary Figure 2).
Figure 3.
Forest plot of kappa values for agreement between two sample types (Em: endometrial cells; Cx: exfoliated cervical cells) for two genes (CDO1m and ZSCAN12m), and bootstrap AUC plots. (A) Forest plot of kappa values across different pathological outcomes; (B) Bootstrap AUC plots for detecting EC using gene methylation; (C) Bootstrap AUC plots for detecting EAH/EC using gene methylation.
EC: Endometrial cancer; EAH: Endometrial atypical hyperplasia; EBL: Endometrial benign lesion; Em: Sample source: endometrial cells; Cx: Sample source: exfoliated cervical cells; CDO1m: CDO1 gene methylation; ZSCAN12m: ZSCAN12 gene methylation; AUC: Area under the receiver operator characteristic curve.
Clinical performance of single and dual markers
No significant differences were observed between the two sample sources (Em vs. Cx) for CDO1m or ZSCAN12m in terms of sensitivity, specificity, PPV, NPV, pLR, or nLR for detecting EC or EAH/EC (all p > 0.05, Table 2). For EC detection, only the NPV difference between ZSCAN12m_Em and ZSCAN12m_Cx reached statistical significance (92.2% vs. 83.1%; p = 0.045). For EAH/EC detection, significant differences were observed for CDO1m_Em vs. CDO1m_Cx in sensitivity, NPV, and nLR (p = 0.029, 0.026, and 0.031, respectively). DNA methylation markers demonstrated higher specificity than that of ultrasonographic endometrial thickness.
Table 2.
Diagnostic performance of single gene methylation.
| Metrics | TedE | CDO1m_Em | ZSCAN12m_Em | P* | CDO1m_Cx | ZSCAN12m_Cx | P& | P# | P$ |
|---|---|---|---|---|---|---|---|---|---|
| For EC | |||||||||
| Sensitivity % (95%CI) n/N | 80.8 (65.6–95.9) 21/26 | 85.0 (73.9–96.1) 34/40 | 87.5 (77.2–97.7) 35/40 | 0.739 | 75.0 (61.6–88.4) 30/40 | 67.5 (53.0–82.0) 27/40 | 0.439 | 0.206 | 0.033 |
| Specificity % (95%CI) n/N | 39.7 (28.5–51.0) 29/73 | 78.8 (69.8–87.7) 63/80 | 73.8 (64.1–83.4) 59/80 | 0.152 | 83.8 (75.7–91.8) 67/80 | 80.0 (71.2–88.8) 64/80 | 0.467 | 0.371 | 0.251 |
| PPV % (95%CI) n/N | 32.3 (21.0–43.7) 21/65 | 66.7 (53.7–79.6) 34/51 | 62.5 (49.8–75.2) 35/56 | 0.289 | 69.8 (56.0–83.5) 30/43 | 62.8 (48.3–77.2) 27/43 | 0.322 | 0.649 | 0.963 |
| NPV % (95%CI) n/N | 85.3 (73.4–97.2) 29/34 | 91.3 (84.6–98.0) 63/69 | 92.2 (85.6–98.8) 59/64 | 0.831 | 87.0 (79.5–94.5) 67/77 | 83.1 (74.7–91.5) 64/77 | 0.371 | 0.266 | 0.045 |
| pLR (95%CI) | 1.34 (1.03–1.75) | 4.00 (2.57–6.22) | 3.33 (2.27–4.90) | 0.292 | 4.62 (2.72–7.83) | 3.38 (2.07–5.50) | 0.323 | 0.651 | 0.963 |
| nLR (95%CI) | 0.48 (0.21–1.12) | 0.19 (0.09–0.40) | 0.17 (0.07–0.39) | 0.832 | 0.30 (0.17–0.52) | 0.41 (0.25–0.64) | 0.372 | 0.279 | 0.063 |
| For EAH/EC | |||||||||
| Sensitivity % (n/N) 95%CI | 83.3 (71.1–95.5) 30/36 | 77.4 (66.1–88.6) 41/53 | 79.2 (68.3–90.2) 42/53 | 0.739 | 60.4 (47.2–73.5) 32/53 | 64.2 (51.2–77.1) 34/53 | 0.655 | 0.029 | 0.088 |
| Specificity % (n/N) 95%CI | 44.4 (32.2–56.7) 28/63 | 85.1 (76.5–93.6) 57/67 | 79.1 (69.4–88.8) 53/67 | 0.157 | 83.6 (74.7–92.5) 56/67 | 86.6 (78.4–94.7) 58/67 | 0.564 | 0.782 | 0.132 |
| PPV % (n/N) 95%CI | 46.2 (34.0–58.3) 30/65 | 80.4 (69.5–91.3) 41/51 | 75.0 (63.6–86.3) 42/56 | 0.203 | 74.4 (61.4–87.5) 32/43 | 79.1 (66.9–91.2) 34/43 | 0.484 | 0.350 | 0.462 |
| NPV % (n/N) 95%CI | 82.4 (69.5–95.2) 28/34 | 82.6 (73.7–91.6) 57/69 | 82.8 (73.6–92.1) 53/64 | 0.957 | 72.7 (62.8–82.7) 56/77 | 75.3 (65.7–85.0) 58/77 | 0.560 | 0.026 | 0.153 |
| pLR (95%CI) | 1.50 (1.15–1.95) | 5.18 (2.87–9.35) | 3.79 (2.33–6.17) | 0.211 | 3.68 (2.05–6.59) | 4.78 (2.52–9.06) | 0.484 | 0.343 | 0.472 |
| nLR (95%CI) | 0.38 (0.17–0.82) | 0.27 (0.16–0.44) | 0.26 (0.13–0.45) | 0.957 | 0.47 (0.33–0.67) | 0.41 (0.28–0.60) | 0.560 | 0.031 | 0.167 |
P*, P-value for comparison between CDO1m_Em and ZSCAN12m_Em (same sample source, Em); P&, P-value for comparison between CDO1m_Cx and ZSCAN12m_Cx (same sample source, Cx); P#, P-value for comparison between CDO1m_Em and CDO1m_Cx (same gene, CDO1); P$, P-value for comparison between ZSCAN12m_Em and ZSCAN12m_Cx (same gene, ZSCAN12).
EC: Endometrial cancer; EAH: Endometrial atypical hyperplasia; TedE: Thickened endometrium, ultrasound findings showing a proliferative phase endometrial thickness >11 mm (or >5 mm in postmenopausal women) are considered thickened endometrium; NA: Not Available; Em: Sample source: endometrial cells; Cx: Sample source: exfoliated cervical cells; CDO1m: CDO1 gene methylation; ZSCAN12m: ZSCAN12 gene methylation; PPV: Positive predictive value; NPV: Negative predictive value; pLR: Positive likelihood ratio; nLR: Negative likelihood ratio.
Table 3 presents the clinical performance of the two-gene combinations. No significant differences were observed between CDO1m_Em/ZSCAN12m_Em and CDO1m_Cx/ZSCAN12m_Cx in sensitivity, specificity, PPV, NPV, pLR, or nLR for detecting either EC or EAH/EC (all p > 0.05). Notably, for EC detection, the sensitivity of the Em-based combination was 97.5% (95%CI = 92.7–100.0) versus 90.0% (95%CI = 80.7–99.3) for the Cx-based combination. Among 40 EC cases, the Em combination missed 1 case, whereas the Cx combination missed 4 cases.
Table 3.
Diagnostic performance of dual-gene methylation markers.
| Sensitivity | Specificity | PPV | NPV | pLR | nLR | |
|---|---|---|---|---|---|---|
| For EC | ||||||
| CDO1m_Em/ ZSCAN12m_Em | 97.5 (92.7–100.0) 39/40 | 71.3 (61.3–81.2) 57/80 | 62.9 (50.9–74.9) 39/62 | 98.3 (94.9–100.0) 57/58 | 3.39 (2.39–4.81) | 0.04 (0.01–0.25) |
| CDO1m_Cx/ ZSCAN12m_Cx | 90.0 (80.7–99.3) 36/40 | 71.3 (61.3–81.2) 57/80 | 61.0 (48.6–73.5) 36/59 | 93.4 (87.2–99.7) 57/61 | 3.13 (2.18–4.49) | 0.14 (0.05–0.36) |
| P | 0.180 | > 0.999 | 0.736 | 0.176 | 0.736 | 0.217 |
| For EAH/EC | ||||||
| CDO1m_Em/ ZSCAN12m_Em | 86.8 (77.7–95.9) 46/53 | 76.1 (65.9–86.3) 51/67 | 74.2 (63.3–85.1) 46/62 | 87.9 (79.5–96.3) 51/58 | 3.63 (2.34–5.65) | 0.17 (0.08–0.35) |
| CDO1m_Cx/ ZSCAN12m_Cx | 81.1 (70.6–91.7) 43/53 | 76.1 (65.9–86.3) 51/67 | 72.9 (61.5–84.2) 43/59 | 83.6 (74.3–92.9) 51/61 | 3.40 (2.17–5.31) | 0.25 (0.14–0.44) |
| P | 0.405 | > 0.999 | 0.817 | 0.413 | 0.817 | 0.417 |
P: P-value for comparison between CDO1m_Em/ZSCAN12m_Em and CDO1m_Cx/ZSCAN12m_Cx for each detection target.
EC: Endometrial cancer; EAH: Endometrial atypical hyperplasia; Em: Sample source: endometrial cells; Cx: Sample source: exfoliated cervical cells; CDO1m: CDO1 gene methylation; ZSCAN12m: ZSCAN12 gene methylation; PPV: Positive predictive value; NPV: Negative predictive value; pLR: Positive likelihood ratio; nLR: Negative likelihood ratio.
Discussion
In this comparative study of 127 women with histologically confirmed endometrial lesions, we systematically evaluated the diagnostic performance of CDO1m and ZSCAN12m in paired endometrial (Em) and exfoliated cervical (Cx) samples for detecting EC and EAH. Our findings demonstrate that both genes exhibit stepwise increases in methylation levels across the pathological spectrum from EBL to EAH to EC, supporting their potential as biomarkers for endometrial lesion progression. The robust diagnostic performance of these markers, particularly the Em-based two-gene combination, suggests clinical utility for EC detection. However, notable differences in certain diagnostic indicators between sample sources suggest that while cervical sampling shows promise as a less invasive alternative, it may not be directly interchangeable with endometrial sampling at the individual patient level, particularly when ruling out disease.
Comparison with existing literature
Our findings align with previous studies demonstrating that DNA methylation biomarkers can effectively discriminate EC from benign lesions [16,17]. The AUCs we observed for CDO1m_Em (0.858) and ZSCAN12m_Em (0.822 for EAH/EC detection) are comparable to or exceed those reported for other single-gene methylation markers in EC detection [18–22]. The ORs for EC associated with hypermethylation of CDO1m_Em (21.0) and ZSCAN12m_Em (19.7) are among the highest reported, suggesting strong associations between promoter hypermethylation and EC risk.
The stepwise increase in methylation levels from benign to EC observed in our study is consistent with the progressive accumulation of epigenetic alterations during endometrial carcinogenesis [23–26]. This pattern supports the potential utility of these markers not only for diagnostic purposes but also for risk stratification of patients with precursor lesions.
Agreement and differences between sample sources
Bland-Altman analysis revealed differences between Em and Cx samples, with Em samples showing consistently higher methylation levels (bias: 5.23 for CDO1m, 6.06 for ZSCAN12m). This finding is expected given the anatomical proximity of endometrial cells to the target lesion versus exfoliated cervical cells, which may contain fewer malignant cells or diluted methylation signals. Despite these absolute differences, Spearman correlation analysis showed moderate positive correlations (ρ = 0.5 for both genes), indicating that relative methylation levels are reasonably consistent between sample sources.
An important consideration when interpreting the Em versus Cx performance differences is the specific CpG regions targeted by the methylation assay. Due to proprietary restrictions, we cannot disclose the exact genomic coordinates. Several published assays measure methylation at these same genes using different CpG windows [18,21,27,28]. Critically, assays optimized for cervicovaginal specimens, such as the WID-qEC test, were developed through extensive epigenome-wide analyses to identify CpG regions that perform optimally in cervical samples [17,28]. The CpG selection is therefore not interchangeable with assays designed primarily for endometrial samples. The observed performance gap between Em and Cx samples in our study may thus reflect, at least in part, an assay property rather than an intrinsic limitation of cervical sampling. This suggests that our Em versus Cx performance differences should not be generalized to all methylation-based assays. Due to proprietary constraints, direct comparison of our CpG targets with published assays is limited, which should be considered when interpreting performance comparisons.
The kappa values for agreement between Em and Cx samples were modest (0.478 for ZSCAN12m, 0.439 for CDO1m across all cases), with no statistically significant agreement in EC or EAH subgroups. This suggests that while the two sample types correlate moderately, they are not directly interchangeable for clinical decision-making at the individual patient level.
Although most diagnostic metrics showed no significant differences between Em and Cx samples, notable exceptions warrant discussion. For EC detection, ZSCAN12m_Em demonstrated significantly higher NPV than ZSCAN12m_Cx (92.2% vs. 83.1%; p = 0.045). For EAH/EC detection, CDO1m_Em showed significantly higher sensitivity, NPV, and nLR compared to CDO1m_Cx (p = 0.029, 0.026, and 0.031, respectively). These differences suggest that Em samples may be more reliable for ruling out disease, particularly in the detection of precursor lesions where methylation signals may be weaker. The nLR differences (lower nLR for Em samples) indicate that a negative test result from an Em sample is more informative for ruling out disease than a negative result from a Cx sample.
When considering the overall diagnostic performance, it is important to note that the majority of diagnostic metrics—including sensitivity, specificity, PPV, pLR, and AUC—showed no statistically significant differences between Em and Cx samples for either gene in either disease setting. The statistically significant differences were limited to NPV-related measures for certain comparisons, specifically ZSCAN12m NPV for EC detection and CDO1m sensitivity, NPV, and nLR for EAH/EC detection. These findings indicate that Em and Cx samples perform comparably on most metrics, with Em showing advantages specifically in measures relevant to ruling out disease.
It is also important to acknowledge that the intrauterine brush used for Em sampling, while less invasive than formal curettage, still requires uterine instrumentation and shares some of the discomfort, patient anxiety, and contraindications (e.g. cervical stenosis, acute pelvic infection, pregnancy) associated with endometrial biopsy procedures. In contrast, cervical sampling is inherently less invasive, better tolerated by patients, and can be performed by trained personnel without the need for intrauterine manipulation. This represents a significant practical advantage of Cx-based testing that must be weighed against its potential diagnostic limitations. The optimal clinical strategy may therefore involve risk stratification: Cx-based methylation testing could serve as an initial triage test to identify patients who warrant definitive endometrial sampling, while Em-based testing might be reserved for cases where Cx results are equivocal or where clinical suspicion is high.
Diagnostic performance and clinical utility
Bootstrap validation confirmed the robustness of our AUC estimates, with mean bootstrap AUCs closely matching original estimates (0.857 vs. 0.858 for CDO1m_Em; 0.821 vs. 0.822 for ZSCAN12m_Em). The narrow confidence intervals support the precision of these estimates, suggesting that our sample size, while modest, was adequate for the primary analyses.
The high sensitivity of the Em-based two-gene combination (97.5%) for EC detection is particularly noteworthy, missing only 1 of 40 EC cases. In contrast, the Cx-based combination showed lower sensitivity (90.0%), missing four cases. Previous reports showed similar results: the sensitivity of the two-gene panel (CDO1m/CELF4m) was 85.7% [27]; the two-gene panel (ZSCAN12m/GYPCm) achieved a sensitivity of 90.9% [28]; and the three-gene panel (CDH13 + CDO1 + ZIC1) demonstrated a sensitivity of 93% [21]. The numerical difference in sensitivity between the Em-based and Cx-based dual-gene combinations for EC detection (97.5% vs. 90.0%, missing 1 vs. 4 of 40 cases) did not reach statistical significance (p = 0.180). However, the clinical consequence of missing four versus one case is substantial and warrants cautious interpretation. This underscores the importance of considering both statistical and clinical significance when evaluating diagnostic tests, particularly in settings where false-negative results could delay cancer diagnosis.
The higher specificity of DNA methylation markers compared to TVUS-measured ET addresses a major limitation of current screening approaches. TVUS has high sensitivity but low specificity, leading to many unnecessary invasive procedures [29]. Methylation markers could potentially reduce the number of unnecessary biopsies by providing a more specific second-line test following abnormal TVUS findings [30,31].
Strengths and limitations
Strengths of this study include: (1) the paired design allowing direct comparison of Em and Cx samples from the same patients; (2) histologically confirmed diagnoses with a spectrum of endometrial pathologies; (3) comprehensive statistical approach including agreement analysis, ROC analysis, bootstrap validation, and head-to-head comparison of diagnostic metrics; and (4) clinically relevant two-gene combination analysis.
Limitations include: (1) the relatively small sample size, particularly for the EAH group (n = 13), which may have limited power for subgroup analyses; (2) the single-center design, which may limit generalizability; (3) while bootstrap validation confirmed the internal robustness of our AUC estimates, this validation was performed on the same dataset used to derive the cutoffs. The confidence intervals derived from bootstrap resampling may therefore underestimate the variability that would be observed in an independent external validation cohort. External validation in a separate population is needed to confirm the generalizability of our findings; (4) the failure of DNA methylation detection for 7 specimens due to insufficient DNA quantity, highlighting potential technical challenges with Cx samples; and (5) the lack of longitudinal follow-up data to assess the prognostic value of these markers. Another important consideration is the high prevalence of EC in our cohort (34%), which is substantially higher than the ∼3–10% prevalence typically observed in women presenting with postmenopausal bleeding in routine clinical practice. This reflects our case-control study design, which was optimized to achieve adequate statistical power for the paired comparisons of Em and Cx samples rather than to represent a general screening population. While sensitivity, specificity, and AUC are relatively robust to changes in disease prevalence, predictive values are prevalence-dependent. Therefore, the PPV and NPV reported in this study should be interpreted with caution and may not directly generalize to low-prevalence screening populations. External validation in cohorts with lower EC prevalence is warranted to confirm the clinical utility of these markers in routine practice.
Conclusion
In conclusion, CDO1 and ZSCAN12 methylation levels increase progressively with endometrial lesion severity, demonstrating robust diagnostic performance for EC and EAH detection in both Em and Cx samples. The Em-based two-gene combination showed excellent sensitivity (97.5%) for EC detection, failing to detect only 1 of 40 cases, suggesting potential clinical utility as a rule-out test. Bootstrap validation confirmed the robustness of these findings, with narrow confidence intervals supporting estimate precision. Importantly, most diagnostic metrics were comparable between Em and Cx samples, with statistically significant differences limited to NPV-related measures. This finding, combined with the possibility that the observed differences may be assay-dependent, suggests that the evidence does not support a general recommendation against cervical sampling. Rather, cervical sampling remains a promising less invasive option for EC screening, particularly in settings where endometrial sampling is not feasible (e.g. cervical stenosis, patient refusal) or as a triage test to identify patients requiring definitive sampling. However, the modest Em-Cx agreement at the individual patient level and the numerically lower sensitivity of the Cx-based combination (90.0% vs. 97.5% for Em, p = 0.180) warrant cautious interpretation and highlight the need for assay optimization for cervicovaginal specimens. Future multicenter studies with larger sample sizes, particularly including more EAH cases, and external validation cohorts are warranted to confirm these findings and establish optimal clinical algorithms for methylation-based EC detection using different sample sources.
Acknowledgements
Supplementary Material
Acknowledgments
The authors thank the participating subjects for their cooperation and support and Hunan Hoomya Gene Technology Co., Ltd. for their technical support.
Acknowledgments
Rong Wang, Xing Fan, and Hong Tao conceived and designed the study. Fang Yu and Peng Gan wrote the first draft of the manuscript. Xing Fan, Fang Yu, Saiping Mao, Zhengjiao Tong, and Juxiang Xiong acquired patient samples and clinical information. Peng Gan and Hong Tao acquired data and performed the data analysis. Rong Wang supervised the study. All authors read and approved the final manuscript.
Funding Statement
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Ethics approval and consent to participate
This study was approved by the Research and Clinical Trial Ethics Committee, Changsha Hospital for Maternal & Child Health Care (No: 20240913-01) and informed consent was signed by the patients.
Disclosure statement
Peng Gan and Hong Tao are employed by Hunan Hoomya Gene Technology Co., Ltd., which commercializes the CDO1 and ZSCAN12 methylation detection kits used in this research. The other authors declare no competing interests.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
- 1.Siegel RL, Kratzer TB, Giaquinto AN, et al. Cancer statistics, 2025. CA Cancer J Clin. 2025;75(1):10–45. doi: 10.3322/caac.21871. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Han B, Zheng R, Zeng H, et al. Cancer incidence and mortality in China, 2022. J Natl Cancer Cent. 2024;4(1):47–53. doi: 10.1016/j.jncc.2024.01.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Raglan O, Kalliala I, Markozannes G, et al. Risk factors for endometrial cancer: an umbrella review of the literature. Int J Cancer. 2019;145(7):1719–1730. doi: 10.1002/ijc.31961. [DOI] [PubMed] [Google Scholar]
- 4.Chou AJ, Bing RS, Ding DC.. Endometrial atypical hyperplasia and risk of endometrial cancer. Diagnostics. 2024;14(22): 01–21. doi: 10.3390/diagnostics14222471. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Uccella S, Zorzato PC, Dababou S, et al. Conservative management of atypical endometrial hyperplasia and early endometrial cancer in childbearing age women. Medicina. 2022;58(9): 01–16. doi: 10.3390/medicina58091256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Anderson B. Diagnosis of endometrial cancer. Clin Obstet Gynaecol. 1986;13(4):739–750. [PubMed] [Google Scholar]
- 7.Khan J. Diagnostic comparison of transabdominal and transvaginal ultrasound in determining endometrial thickness. Natl J (Wash). 2024;9:214–220. [Google Scholar]
- 8.Koss LG, Schreiber K, Oberlander G, et al. Screening of asymptomatic women for endometrial cancer. Obstet Gynecol. 1981;57(6):681–691. [PubMed] [Google Scholar]
- 9.Wang B, Wang M, Lin Y, et al. Circulating tumor DNA methylation: a promising clinical tool for cancer diagnosis and management. Clin Chem Lab Med. 2024;62(11):2111–2127. doi: 10.1515/cclm-2023-1327. [DOI] [PubMed] [Google Scholar]
- 10.Papanicolau-Sengos A, Aldape K.. DNA methylation profiling: an emerging paradigm for cancer diagnosis. Annu Rev Pathol. 2022;17(1):295–321. doi: 10.1146/annurev-pathol-042220-022304. [DOI] [PubMed] [Google Scholar]
- 11.Zheng H, Yu C, Yang L, et al. Research progress of DNA methylation markers for endometrial carcinoma diagnosis. J Cancer. 2025;16(3):812–820. doi: 10.7150/jca.104214. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wang X, Zheng L, Zhu G, et al. CDO1 and CELF4 methylation assay as the dominant predictor of endometrial cancer: a cohort analysis across pre- and post-menopausal cohorts. Gynecol Oncol. 2026;206:82–92. doi: 10.1016/j.ygyno.2026.01.776. [DOI] [PubMed] [Google Scholar]
- 13.Redl E, Herzog C, Vavourakis C, et al. The cervico-vaginal DNA methylation WID-qEC test: an epigenetic marker associated with ovarian cancer in the absence of endometrial and cervical cancer. Int J Cancer. 2026;158(10):2530–2536. doi: 10.1002/ijc.70354. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Nouwens AJ, Schaafsma M, van Trommel NE, et al. Detection of recurrent endometrial cancer via DNA methylation analysis of cervicovaginal self-samples and urine. Int J Cancer. 2026;158(9):2491–2500. doi: 10.1002/ijc.70358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Han L, Du J, Zhao L, et al. An efficacious endometrial sampler for screening endometrial cancer. Front Oncol. 2019;9:67. doi: 10.3389/fonc.2019.00067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wang L, Dong L, Xu J, et al. Hypermethylated CDO1 and ZNF454 in cytological specimens as screening biomarkers for endometrial cancer. Front Oncol. 2022;12:714663. doi: 10.3389/fonc.2022.714663. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Schreiberhuber L, Herzog C, Vavourakis CD, et al. The WID-qEC test: performance in a hospital-based cohort and feasibility to detect endometrial and cervical cancers. Int J Cancer. 2023;152(6):1269–1274. doi: 10.1002/ijc.34275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Kong LH, Xiao XP, Wan R, et al. The role of DNA methylation in the screening of endometrial cancer in postmenopausal women. Zhonghua Yi Xue Za Zhi. 2023;103(12):907–912. doi: 10.3760/cma.j.cn112137-20220929-02058. [DOI] [PubMed] [Google Scholar]
- 19.Barrett JE, Jones A, Evans I, et al. The WID-EC test for the detection and risk prediction of endometrial cancer. Int J Cancer. 2023;152(9):1977–1988. doi: 10.1002/ijc.34406. [DOI] [PubMed] [Google Scholar]
- 20.Liew PL, Huang RL, Wu TI, et al. Combined genetic mutations and DNA-methylated genes as biomarkers for endometrial cancer detection from cervical scrapings. Clin Epigenetics. 2019;11(1):170. doi: 10.1186/s13148-019-0765-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Wever BMM, van den Helder R, van Splunter AP, et al. DNA methylation testing for endometrial cancer detection in urine, cervicovaginal self-samples and cervical scrapes. Int J Cancer. 2023;153(2):341–351. doi: 10.1002/ijc.34504. [DOI] [PubMed] [Google Scholar]
- 22.Yuan J, Mao Z, Lu Q, et al. Hypermethylated PCDHGB7 as a biomarker for early detection of endometrial cancer in endometrial brush samples and cervical scrapings. Front Mol Biosci. 2021;8:774215. doi: 10.3389/fmolb.2021.774215. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Li M, Xia Z, Wang R, et al. Unveiling DNA methylation: early diagnosis, risk assessment, and therapy for endometrial cancer. Front Oncol. 2024;14:1455255. doi: 10.3389/fonc.2024.1455255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Wang SF, Du CY, Li M, et al. Endometrial cancer detection by DNA methylation analysis in cervical papanicolaou brush samples. Technol Cancer Res Treat. 2024;23:15330338241242637. doi: 10.1177/15330338241242637. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Qi B, Sun Y, Lv Y, et al. Hypermethylated CDO1 and CELF4 in cytological specimens as triage strategy biomarkers in endometrial malignant lesions. Front Oncol. 2023;13:1289366. doi: 10.3389/fonc.2023.1289366. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Inoue F, Sone K, Toyohara Y, et al. Targeting epigenetic regulators for endometrial cancer therapy: its molecular biology and potential clinical applications. Int J Mol Sci. 2021;22(5):2305. doi: 10.3390/ijms22052305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zhao X, Yang Y, Fu Y, et al. DNA methylation detection is a significant biomarker for screening endometrial cancer in premenopausal women with abnormal uterine bleeding. Int J Gynecol Cancer. 2024;34(8):1165–1171. doi: 10.1136/ijgc-2024-005723. [DOI] [PubMed] [Google Scholar]
- 28.Evans I, Reisel D, Jones A, et al. Performance of the WID-qEC test versus sonography to detect uterine cancers in women with abnormal uterine bleeding (EPI-SURE): a prospective, consecutive observational cohort study in the UK. Lancet Oncol. 2023;24(12):1375–1386. doi: 10.1016/S1470-2045(23)00466-7. [DOI] [PubMed] [Google Scholar]
- 29.Long B, Clarke MA, Morillo ADM, et al. Ultrasound detection of endometrial cancer in women with postmenopausal bleeding: systematic review and meta-analysis. Gynecol Oncol. 2020;157(3):624–633. doi: 10.1016/j.ygyno.2020.01.032. [DOI] [PubMed] [Google Scholar]
- 30.den Helder RV, Wever BM, van Trommel JA, et al. DNA methylation markers for endometrial cancer detection in minimally invasive samples: a systematic review. Epigenomics. 2020;12(18):1661–1672. doi: 10.2217/epi-2020-0164. [DOI] [PubMed] [Google Scholar]
- 31.Asaturova A, Zaretsky A, Rogozhina A, et al. Advancements in minimally invasive techniques and biomarkers for the early detection of endometrial cancer: a comprehensive review of novel diagnostic approaches and clinical implications. J Clin Med. 2024;13(24):7538. doi: 10.3390/jcm13247538. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.