Session Item

In 2021, the ESGO-ESTRO-ESP endometrial cancer (EC) guideline was updated with risk classification based on clinicopathological factors and the molecular classification. We aim to strengthen the evidence-base for a prognostic and therapeutic framework in women with stage I-III EC. We hereto analyzed large, completely documented cohorts to facilitate risk stratification and support decisions on adjuvant treatment.

Data from the PORTEC-1/-2/-3 randomized trials (n=714/427/660) and an independent prospective Dutch clinical cohort from Enschede (n=270) were pooled for analysis. Competing-risk models for vaginal, pelvic, distant, and overall recurrence and cancer-specific survival were built with established major risk factors (stage, histotype, grade, substantial lymphovascular space invasion, molecular classification) and corrected for age, and time and cohort effects. All models were developed in duplo: with and without the molecular classification. Benefits of adjuvant treatment will be estimated based on a systematic literature review and incorporated in the models within the next months. Model performance was evaluated by quantifying the time-dependent area under the receiver operating characteristic curves (AUC). Internal validation was performed according to a leave-one-cohort-out cross-validation method, where each cohort served as a validation set once for models developed without that cohort.

In total, 2071 women with EC with a median follow-up of 10.0 years (interquartile range 6.9-12.4 years) were available for analyses (table 1). Competing risk analyses confirmed the prognostic relevance of all established clinicopathological risk factors and the EC molecular class for overall recurrence (figure 1). Performance of the prediction models including only clinicopathological factors was good (AUCs from 0.73 [95%CI 0.68-0.79] to 0.81 [95%CI 0.76-0.85]) and improved by adding the molecular classifier (AUC increased to 0.77 [95%CI 0.72-0.82] to 0.84 [95%CI 0.82-0.86]).

The data from multiple large prospective cohorts of molecularly classified endometrial cancers with robust long-term follow-up data yielded clinically relevant prediction models for vaginal, pelvic, distant, and overall recurrence and cancer-specific survival. These evidence-based models accurately estimate absolute risks and may support personalized predictions of the benefit of different adjuvant treatment strategies. The models will form the backbone of a mobile and web-based prediction tool that facilitates risk stratification and shared-decision making.