Gastroenteropancreatic neuroendocrine tumors (GEP-NETs) are a heterogeneous group of neoplasms that originate from neuroendocrine cells of the gastrointestinal tract and pancreas, and they account for more than two thirds of all neuroendocrine tumors worldwide [1]. Their incidence has increased markedly over the past three decades, largely due to improvements in diagnostic imaging and heightened clinical awareness [2]. Although many GEP-NETs follow an indolent course, a substantial proportion exhibit aggressive behavior, early metastatic spread, and significant variability in long term outcomes, which complicates prognostication and therapeutic decision making [3].

Accurate risk stratification is essential for guiding management. Current frameworks incorporate disease stage, tumor differentiation, and the Ki 67 proliferation index, which is the central quantitative biomarker in the World Health Organization classification of neuroendocrine neoplasia [4,5]. Ki 67 is routinely used to distinguish G1, G2, and G3 tumors and is often a determinant of systemic therapy selection. However, despite its central role, the prognostic value of Ki 67 in well differentiated GEP-NETs is not uniform across studies. Several analyses have shown strong associations between higher Ki 67 and mortality [6,7], whereas others report weaker or inconsistent correlations, particularly in real world cohorts where reporting practices and cutoffs vary substantially [8].

There are also methodological challenges in how Ki 67 is modeled. Many studies rely exclusively on categorical thresholds that may obscure nonlinear or threshold based effects. Others analyze Ki 67 as a linear continuous variable, an approach that oversimplifies the complex biology of proliferation in well differentiated tumors. Few studies have applied spline based modeling or have compared multiple analytic strategies within the same dataset. Moreover, some retrospective datasets truncate or cap high Ki 67 values, a practice that introduces bias and may suppress the true prognostic contribution of high grade disease [9].

Real world data with long term follow up provide an important opportunity to reassess prognostic markers using rigorous and transparent statistical methods. In particular, mature single center cohorts allow detailed examination of proliferation indices across the full biological spectrum of GEP-NETs. In this context, we performed a comprehensive survival analysis of a large, well curated GEP NET cohort with nearly 100 months of median follow up. Our aims were to evaluate the prognostic impact of age, tumor stage, WHO grade, and Ki 67; to compare linear, categorical, and spline based modeling of the proliferation index; and to establish a reproducible analytical framework aligned with contemporary recommendations for biomarker evaluation in neuroendocrine oncology.

Our findings provide new evidence regarding the prognostic structure of GEP-NETs and highlight the importance of appropriate modeling strategies for Ki 67. These results have the potential to inform both clinical practice and future research on precision risk stratification in neuroendocrine tumors.