Development and validation of prognostic biomarkers to enhance clinical triage in high-risk endometrial carcinoma

Document Type

Conference Proceeding

Publication Date

5-1-2026

Publication Title

Gynecol Oncol

Keywords

biological marker, carbohydrate mixture plus electrolytes, cohort analysis, conference abstract, controlled study, drug therapy, endometrium cancer, endometrium carcinoma, female, histology, human, human tissue, major clinical study, male, nested cross validation, patient triage, protein fingerprinting, proteomics, RNA sequencing, survival rate

Abstract

Objectives: Proteomic profiling was used to refine risk stratification beyond clinical factors and TCGA molecular subtypes in high-risk endometrial cancer (HR-EC). Methods: Patients with stage I–III HR-EC (uterine serous, grade 3 endometrioid, clear cell, or mixed) and evaluable archival tumor samples were included. Proteomic features were screened by univariate Cox analysis and refined with LASSO Cox regression, nested cross-validation, and stability selection in a randomly selected training set, stratified by progression status. A 16-protein prognostic signature was evaluated with pre-op factors (age + histology), intra-op factors (age + histology + stage) ± TCGA molecular subtypes using Kaplan–Meier and Cox regression analyses, time-dependent ROC and C-index. Cross-platform concordance analysis was performed using RNA-seq expression. Results: In the training set (n = 182), categorized levels of the 16-protein signature was associated with worse PFS (Fig. 1A) with a 5-year survival rate of 86.8% vs. 30.2% for patients with low vs. high proteomic signature (p < 0.0001), respectively. The protein signature retained independent prognostic value after adjustment for age, stage, and histology (adjusted HR 1.64, 95% CI 1.12–2.42). Fig. 1B illustrates that the signature demonstrated excellent discrimination (AUC 0.87, C-index 0.80) outperforming pre-op factors (AUC 0.63, C-index 0.59), intra-op factors (AUC 0.75, C-index 0.66) and TCGA molecular subtype (AUC 0.59, C-index 0.59). AUC increased to 0.87 with pre-op factors + the protein signature ± TGCA molecular subtypes, to 0.89 with intra-op factors + the protein signature, and to 0.90 with intra-op factors + the molecular subtype + TCGA molecular subtypes. In the testing set (n = 92), AUC was 0.66 for pre-op factors, 0.76 for intra-op factors, 0.70 for the protein signature and 0.55 for TCGA molecular subtypes. Fig. 1B shows the top predictive performances in this set with intra-op factors + the protein signature (AUC 0.80) or intra-op factors + the protein signature + TCGA molecular subtypes (AUC 0.83). High vs. low levels of proteomic signature continued to be associated with worse PFS in this set (Fig. 1C). In the independent (proteomics only) cohort (N = 73), the top predictive performance was with intra-op factors + the protein signature (AUC 0.81, Fig. 1B), and high vs. low levels of the protein signature was associated with worse PFS (Fig. 1D). Cross-platform concordance analysis in HR-EC cases from TCGA (N = 272) showed 12/16 candidates in the protein signature had HRs in the same direction across proteomic and transcriptomic datasets, with several candidates consistently associated with both PFS and OS (p < 0.001). Conclusions: A 16-protein signature stratifies HR-EC patients by progression risk independent of clinical factors. The protein signature with age, histology and stage demonstrated robust performance across cohorts and platforms, supporting further development for HR-EC risk stratification. [Formula presented]

Volume

208

First Page

S46

Last Page

S47

Find It @ Sladen

Share

COinS