Front Oncol · 2026;16:1849730 · EndoMethy-I Trial · China

EndoMethy-I: CDO1/CELF4/NEFM Methylation Panel for EC

Development and validation of hypermethylated gene markers in cervical cytological samples for detecting endometrial cancer (EndoMethy-I trial)
40 exploratory → 347 training → 149 validation · 3-gene panel Se 94.6% / Sp 92.8% · CDO1 AUC 0.93 · decision-tree strategies
Chen X, Liu L, Jin X, Cheng Y, Wu H, You Y, Liou Y, Liu P, Lang J, Li L
Peking Union Medical College Hospital · Front Oncol 2026;16:1849730 · doi:10.3389/fonc.2026.1849730
496
train 347 + valid 149
94.6%
3-gene Se
92.8%
3-gene Sp
0.93
CDO1 AUC

1Background & Objective

  • Problem: no guideline-recommended biomarkers for EC screening; biopsy invasive & costly.
  • Objective: develop & validate hypermethylated-gene panel on cervical cytology (EndoMethy-I).

2Study Design — Three Cohorts

40
exploratory (21 EC)
347
training
149
validation
  • Exploratory: 40 paired endometrial tissue + cervical scrapings → 11 candidate genes (array).
  • Split: suspected-EC women randomized 7:3 into training/validation.
  • Panel: CDO1 + CELF4 + NEFM hypermethylation from cervical cytology.
  • Model: decision tree with methylation + bleeding + TVS endometrial thickness.
  • Reference: endometrial histopathology (gold standard).
11
candidate genes screened
3
gene panel
7:3
train:valid split
3-gene Se
94.6%
3-gene Sp
92.8%
Se 94.6%
valid · panel
Sp 92.8%
valid · panel
0.93
CDO1 AUC · train
  • Why cervical cells: non-invasive, ample material, mirrors endometrial methylation.
  • Young women: missed cotest cases were Lynch-related (31.8/32.2 y) — genetics-first pathway needed.
  • Inclusion: suspected EC symptoms or high-risk factors (obesity, diabetes, Lynch).
  • Sample: cervical scrapings collected for cytology-based methylation assay.
  • Statistics: Lasso logistic for gene selection; ROC/AUC; decision tree.
  • Miss analysis: strategy-level missed EC cases & hysteroscopy rates compared.
Single-center prospective observational study; three cohorts for discovery → training → validation; methylation-only keeps hysteroscopy at 25.0%.

3Gene Discovery — AUC Ranking

CDO1
AUC 0.93
NEFM
0.91
CELF4
0.89
other 8 genes
<0.89
  • 11-gene screen: top-3 (CDO1/NEFM/CELF4) taken forward to the panel.

4Validation — 3-Gene Panel

94.6%
Se (85.1–98.9)
92.8%
Sp (89.3–95.5)
TVS+
Sp ↑ · Se ↓
  • Excellent accuracy: CDO1/CELF4/NEFM panel — Se 94.6%, Sp 92.8% in validation.
  • TVS addition: endometrial thickness improves specificity slightly, at cost of very low sensitivity.

5Decision-Tree Strategies

Decision tree comparing clinical practice and diagnostic strategies
Fig. Decision tree — clinical practice vs 7 methylation-based strategies (missed EC cases & hysteroscopy rates).
25.0%
hysteroscopy · methylation only
5 missed
methylation-only
~50%
hysteroscopy · cotest
2 missed
cotest (young Lynch)
  • Methylation-only: low hysteroscopy rate (25%), 5 missed cases.
  • Cotest strategies: fewest missed (2, Lynch-related young women) at ~50% hysteroscopy.
  • No-marker practice: unacceptable high missed-diagnosis rates.
Clinical Significance