Prognostic inflammatory–immune score-based risk stratification optimizes adjuvant therapy for non-gastric gastrointestinal stromal tumors: a multicenter study

Background Conventional modified National Institutes of Health (mNIH) and Armed Forces Institute of Pathology (AFIP) systems exhibit “anatomical bias” due to their development from predominantly gastric cohorts, resulting in suboptimal prognostic performance in non-gastric gastrointestinal stromal tumors (GISTs). This study aimed to develop an integrated model to refine risk stratification and guide adjuvant imatinib duration in non-gastric GIST. Methods A multicenter retrospective cohort of 471 patients with non-gastric GIST was analyzed. Hematological variables were screened using the Boruta algorithm to construct a Prognostic Inflammatory–Immune Score (PIIS), which was integrated with clinicopathological factors to develop a nomogram. Feature contributions were interpreted using Shapley Additive Explanations (SHAP). Risk re-stratification was conducted to evaluate benefits of different adjuvant imatinib durations. Results The PIIS comprised the platelet-to-albumin ratio (PAR), platelet-to-lymphocyte ratio (PLR), derived neutrophil-to-lymphocyte ratio (dNLR), and lactate dehydrogenase-to-albumin ratio (LAR). Multivariable analysis identified sex, tumor size, mitotic index, Ki-67 index, and PIIS as independent predictors of recurrence-free survival (RFS). The integrated model demonstrated superior discrimination, with C-indices of 0.839 in the training cohort and 0.795 in the external validation cohort, outperforming mNIH and AFIP. The model revealed substantial heterogeneity within conventional risk categories and stratified mNIH high-risk and AFIP intermediate- and high-risk patients into low-, medium-, and high-risk groups. In model-defined medium- and high-risk subgroups, adjuvant therapy beyond 3 years was associated with improved RFS, whereas no clear additional benefit was observed in the low-risk subgroup. Conclusions This multicenter-validated nomogram may improve recurrence risk prediction in non-gastric GIST and may provide a useful reference for risk-adapted adjuvant therapy planning.