| Abstract: |
Grading neuroendocrine tumors using Ki-67 proliferation index (PI) is essential for prognostic assessment and therapeutic decision-making. However, the absence of standardized guidelines has led to methodological inconsistencies across pathology practices. This study aimed to establish more standardized approaches by evaluating grading methodologies and their impact on clinical outcomes using a large multisite data set. We analyzed 734 tissue sections from 325 patients, applying hotspot analysis (HSA) and whole-slide analysis (WSA) to determine Ki-67 PI and World Health Organization grade across primary tumors, regional metastases, and distant metastases. Ki-67 PI was quantified using digital image analysis, with WSA capturing the entire tumor proliferation profile and HSA focusing on the highest proliferating region. A patient-wise analysis was performed to determine the highest grade site per patient, and each case was assigned dual World Health Organization grades based on HSA and WSA. To evaluate the generalizability of our findings, we analyzed an external validation cohort of 74 patients, which was processed with independent image analysis software to ensure reproducibility. The analysis revealed that grading based solely on the primary tumor failed to predict clinical outcomes, as the highest grade site varied among the primary tumor (26.1%), regional metastases (39.1%), and distant metastases (34.8%). Within G2 tumors, survival outcomes differed significantly based on grading methodology, with diffuse G2 tumors (homogeneous Ki-67 distribution) demonstrating significantly worse survival compared with focal G2 tumors (84 vs 136 months; P < .01). Cox proportional hazards regression identified the maximum WSA Ki-67 PI as the sole independent predictor of overall survival, whereas TNM stage and tumor location (pancreatic vs jejunoileal) were not statistically significant. The external validation cohort reinforced these findings, confirming that diffuse G2 tumors exhibited significantly worse progression-free survival than focal G2 tumors. These findings emphasize the necessity of integrating both HSA and WSA for grading neuroendocrine tumors, as well as evaluating all available disease sites to ensure accurate prognostication. Incorporating digital image analysis into grading workflows can provide a more standardized, reproducible approach, improving clinical decision-making and patient outcomes. © 2025 The Authors |
