Integrated typing system of endometrial cancer based on Chinese population - PKU typing: a retrospective, multicentre study

Endometrial cancer (EC) is a heterogeneous malignancy in which clinicopathological, molecular, and immune characteristics each provide prognostic information but are insufficient when used independently. We aimed to identify prognostic determinants across these domains and develop an integrated, clinically applicable prognostic risk classification system. This multicenter retrospective study included 200 EC patients from Peking University People's Hospital(Center 1) and 200 patients from Beijing obstetrics and Gynecology Hospital Affiliated to Capital Medical University(Center 2). Clinicopathological variables, mismatch repair (MMR) status, p53 status, and stromal CD3⁺ and CD8⁺ lymphocyte densities were analyzed. Cox regression was used to establish prognostic models. A comprehensive model integrating clinicopathological, molecular, and immune features (PKU-typing) was constructed using a nomogram and validated internally and externally. FIGO stage and histological grade were independent prognostic clinicopathological factors. Stromal CD3⁺ and CD8⁺ infiltration levels correlated with prognosis and defined three immune clusters with distinct outcomes. The PKU-typing model demonstrated strong predictive accuracy in the training cohort (C-index 0.899) and remained robust in external validation (C-index 0.765). Compared with single-dimension models, PKU-typing significantly improved risk classification (NRI and IDI across all comparisons). By integrating clinicopathological, molecular, and immune characteristics, the PKU-typing risk classification system provides a biologically informed, practical framework for individualized risk stratification in EC. Its strong performance and reliance on accessible testing methods support its broad applicability in clinical practice. We developed and externally validated an integrated prognostic classification (PKU-typing) for endometrial cancer. The model integrates clinicopathological features, molecular subtypes, and immune features. PKU-typing demonstrated robust discrimination in independent cohorts (C-index 0.899 and 0.765). PKU-typing improved prognostic accuracy compared with single-domain models. The model relies on routinely available immunohistochemistry assays, enhancing clinical feasibility.