Integrated local and global Ki67 profiling for prognostic stratification and therapeutic guidance in high-grade neuroendocrine neoplasms: a multicenter validation study

This is the authors' abstract. We don't add key points for this paper.

BACKGROUND: The distinction between high-grade neuroendocrine G3 tumors (G3 NETs) and carcinomas (NECs) is a critical diagnostic dilemma with profound therapeutic implications. Conventional Ki67 hotspot analysis often fails to provide robust risk stratification. We aimed to develop and validate a robust, interpretable digital pathology biomarker that overcomes these limitations to improve risk stratification and guide therapeutic decisions. METHODS: We developed the Integrated Local-Global Ki67 (IL-GKi67) score, which synergizes localized hotspot quantification with global intratumoral heterogeneity assessment from Ki67-stained whole-slide images. The score's prognostic and predictive performance was rigorously evaluated in a multicenter cohort (N = 349) with internal and external validation sets. RESULTS: The IL-GKi67 score was a powerful and independent prognostic factor, where high scores correlated with drastically worse overall survival (multicenter validation: HR = 4.169, P = 0.009), remaining significant where the traditional hotspot method failed (P = 0.128). Most importantly, IL-GKi67 functioned as a definitive predictive biomarker. Patients with low IL-GKi67 scores derived a profound survival benefit from adjuvant chemotherapy (mOS: 45.0 months, at 50 months, the restricted mean survival time of chemo subset was 48.9 months (95% CI, 45.36 to 52.02); HR = 0.069, P = 0.002), an effect particularly pronounced in the aggressive NEC subgroup (HR = 0.061, P < 0.001). The score served as a powerful morphological surrogate for TP53 mutation status (P < 0.001) and refined the conventional G3 NET/NEC classification by re-stratifying a substantial portion of patients into more clinically relevant risk groups. CONCLUSIONS: IL-GKi67 is an interpretable and validated digital biomarker that overcomes the limitations of manual Ki67 assessment. By integrating local and global proliferation patterns, it improves risk stratification and provides a practical tool to guide personalized adjuvant chemotherapy decisions for patients with HGNENs.