ENDORISK-2: a personalized Bayesian network for preoperative risk stratification in endometrial cancer, integrating molecular classification and preoperative myometrial invasion assessment

  • Marike S. Lombaers
  • , Casper Reijnen*
  • , Ally Sprik
  • , Petra Bretová
  • , Marcel Grube
  • , Stephanie Vrede
  • , Hege F. Berg
  • , Jasmin Asberger
  • , Eva Colas
  • , Jitka Hausnerova
  • , Jutta Huvila
  • , Antonio Gil-Moreno
  • , Xavier Matias-Guiu
  • , Michiel Simons
  • , Marc P.L.M. Snijders
  • , Nicole C.M. Visser
  • , Stefan Kommoss
  • , Vit Weinberger
  • , Frederic Amant
  • , Peter Bronsert
  • Ingfrid S. Haldorsen, Martin Koskas, Camilla Krakstad, Heidi V.N. Küsters-Vandevelde, Gemma Mancebo, Louis J.M. van der Putten, Irene de La Calle, Peter J.F. Lucas, Arjen Hommersom, Johanna M.A. Pijnenborg
*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Background: ENDORISK is a Bayesian network that can assist in preoperative risk estimation of lymph node
metastasis (LNM) risk in endometrial cancer (EC) with consistent performance in external validations. To reflect
state of the art care, ENDORISK was optimized by integrating molecular classification and preoperative
assessment of myometrial invasion (MI).
Methods: Variables for POLE, MSI, and preoperative assessment of MI, either by expert transvaginal ultrasound or
pelvic magnetic resonance imaging (MRI), were added to develop ENDORISK-2. The p53 biomarker, part of the
molecular classification, was already included in ENDORISK. External validation of ENDORISK-2 for LNM prediction was performed in two independent cohorts from: Brno (CZ), (n = 581) and Tübingen (DE), (n = 247).
Findings: ENDORISK-2 yielded AUCs of 0⋅85 (95 % CI 0⋅80–0⋅90) (CZ) and 0⋅86 (95 % CI 0⋅77–0⋅96) (DE) for
predicting LNM. In patients with low-grade histology, 83 % (CZ) and 89 % (DE) were estimated having less than
10 % risk of LNM, with false negative rates (FNR) of 4⋅3 % (CZ) and 2⋅2 % (DE). The previously defined set of
minimally required variables, i.e.: preoperative tumor grade, three of the four immunohistochemical (IHC)
markers, and one clinical marker, could be interchanged with the new variables, with comparable validation
metrics, including AUC values of 0⋅79–0⋅87 for LNM prediction.
Interpretation. Incorporation of molecular data and preoperative MI improved the flexibility of ENDORISK with
comparable diagnostic accuracy for estimating LNM as when based on low-cost immunohistochemical biomarkers. In addition, the high diagnostic accuracy in patients with low-grade EC demonstrates how ENDORISK-2
could aid clinicians in identifying patients in whom surgical lymph node assessment may safely be omitted. These
results underline its power for clinical use in both high and low resource countries.
Original languageEnglish
Article number116058
Number of pages10
JournalEuropean Journal of Cancer
Volume231
Early online date1 Oct 2025
DOIs
Publication statusPublished - 9 Dec 2025

Fingerprint

Dive into the research topics of 'ENDORISK-2: a personalized Bayesian network for preoperative risk stratification in endometrial cancer, integrating molecular classification and preoperative myometrial invasion assessment'. Together they form a unique fingerprint.

Cite this