Prognostic comparison of the FIGO 2009 and 2023 endometrial cancer staging systems: a large, single-institution retrospective cohort study

Objective The International Federation of Gynecology and Obstetrics (FIGO) 2023 endometrial cancer staging system includes 21 substages and proposes a shift from pure anatomic staging to an approach incorporating histology, grade, lymphovascular invasion, and molecular classification. We sought to evaluate the overall survival prognostic ability of the FIGO 2023 endometrial cancer staging system versus the 2009 system. Methods We collected clinicopathologic characteristics, molecular profiling results, and survival outcomes for all patients with a new diagnosis of endometrial cancer treated at our institution between 2016 and 2020. All cases had pathology review at our institution by a gynecologic pathologist. A FIGO 2009 and 2023 stage was assigned. We applied R 2 ε statistics based on an isotonic proportional hazards model to compare prognostic ability of both staging systems. Results Of 2,067 patients with endometrial cancer, 952 (46%) underwent tumor molecular profiling. When comparing their FIGO 2009 and 2023 stage, 538 (26.0%) patients were upstaged, meaning they were staged higher in the 2023 system, and 100 (4.8%) were downstaged. Five-year overall survival estimates by staging systems were similar for most stages, with the exception of the FIGO 2023 IA3 substage, which demonstrated improved outcomes (100% vs. 70.3% FIGO 2009 IIIA), and the IICm p53abn group, which had poorer prognosis (73% vs. 86.8% FIGO 2009 II). Using isotonic proportional hazards regression models, the 2023 model slightly numerically outperformed the 2009 model for overall survival (R 2 ε 0.546 vs 0.415); however, adjusted 95% confidence intervals were underdeveloped. Conclusion Using a real-world, large, well-annotated cohort of patients with endometrial cancer with expert pathology review, we were able to calculate 5-year progression-free and overall survival rates for each substage. The FIGO 2023 model suggests slightly improved numerical discrimination between stages compared to FIGO 2009; however, small patient numbers in each substage and concerns for overfitting limit this analysis.