Endometrial cancer in G20 countries, 1990-2050: trends, projections, and machine learning-based risk drivers.

OBJECTIVE: This study aimed to evaluate long-term trends and future trajectories of endometrial cancer burden and population-level factors associated with its incidence across G20 countries. METHODS: Using Global Burden of Disease 2023 data, we assessed incidence, prevalence, mortality, and disability-adjusted life years (DALYs) among women in G20 countries from 1990 to 2023. Trends were analyzed by age and country using estimated annual percentage change, Joinpoint regression, cluster analysis, and decomposition. Future burden was projected using autoregressive integrated moving average, exponential smoothing, and Bayesian age-period-cohort models. Summary exposure values for 85 risk factors were incorporated into machine-learning models interpreted using SHapley Additive exPlanations (SHAP). RESULTS: Substantial cross-national heterogeneity was observed in 2023. Although absolute case numbers increased in most countries, age-standardized mortality and DALYs generally stabilized or declined. Population ageing and growth primarily drove rising case numbers, whereas epidemiological improvements offset mortality. Projections indicated continued growth in absolute burden. Machine-learning analyses identified nonlinear associations of incidence with age structure, sociodemographic development, and lifestyle-related factors. CONCLUSION: Endometrial cancer burden in G20 countries is shifting from a mortality-dominated pattern toward increasing chronic health loss, largely because of population ageing and growth. Prevention and control strategies should extend beyond mortality reduction to emphasize risk stratification, early detection, survivorship care, and long-term management. Machine-learning and SHAP findings provide exploratory evidence of population-level exposure profiles associated with incidence and should be considered hypothesis-generating rather than causal.