Personalizing Postoperative Surveillance: A Risk-Stratified Predictive Model for Recurrence of Gastrointestinal Neuroendocrine Neoplasms.

BACKGROUND: Current Ki-67-based prognostic tools have limitations in stratifying recurrence risk for gastrointestinal neuroendocrine neoplasms (GI-NENs) after primary tumor resection (PTR), necessitating personalized predictive models to optimize surveillance. METHODS: Data from 316 GI-NEN patients undergoing PTR were analyzed. Variables selected via LASSO regression were incorporated into Cox and Random Survival Forest (RSF) models. A nomogram was constructed based on superior model, and cut-off scores stratified patients into three recurrence risk groups. RESULTS: The RSF model showed superior accuracy (C-index: 0.854) with good discrimination and calibration. A nomogram integrating eight key predictors stratified patients into low-risk (n = 241), intermediate-risk (n = 57), and high-risk (n = 18) groups. The Kaplan-Meier curves among them were significantly distinct. High-risk group had a median recurrence-free survival of 11.4 months, with nearly all experiencing recurrence or progression within 26 months. Compared with traditional NET grading, this model reclassified some G2 patients into intermediateand even high-risk groups and down-stratified G3 patients with favorable prognoses into the intermediate-risk group. CONCLUSIONS: This study developed and validated a risk-stratified model that effectively identifies postoperative recurrence risk in GI-NENs, enabling tailored surveillance. High-risk patients require intensive follow-up every 2-3 months and early adjuvant intervention. Low-risk patients can safely extend to annually. Most intermediate-risk patients may recieve follow-up every 6-12 months, though G3 patients near the high-risk threshold may still require surveillance every 3-6 months. This model facilitates personalized follow-up to improve long-term outcomes.