Predictive model and novel classification of pancreatic fistula following minimally invasive enucleation: a prospective cohort study (CSPAC-MIEN-2)

Background: Minimally invasive enucleation (MIEN) of pancreatic tumors differs from standard surgical procedures in terms of wound management strategies, the risk of main pancreatic duct (MPD) injury, and the characteristics of postoperative pancreatic fistula (POPF). Currently, no accurate predictive model or grading system exists for POPF following MIEN. Methods: Data from 173 patients who underwent MIEN at our surgical team between June 2019 and November 2022 were prospectively collected. The patients were categorized into a training group ( n = 122) and a validation group ( n = 51). The external validation group consisted of 56 patients. Results: Of the 173 patients, 88 (50.9%) had non/biochemical POPF, while 85 (49.1%) had grade B POPF. None of the patients suffered grade C POPF. After conducting multivariate logistic regression analysis, preoperative γ-glutamyl transferase [odds ratio (OR) = 1.062], imaging evaluated anatomical relationship between tumor and MPD (clinging to/compressing MPD: OR = 4.080; ecasing MPD: OR = 9.037), intraoperative MPD management (suture repair, or reconstruction with stenting: OR = 65.513; exposed but not injured: OR = 3.854), pancreatic texture (soft: OR = 3.923), and pancreatic wound management (re-serosalization: OR = 0.123) were identified as independent risk factors for clinically relevant POPF (CR-POPF) after MIEN. Based on these risk factors, a predictive model for CR-POPF after MIEN was developed. The model demonstrated favorable consistency [receiver operator characteristic (ROC) = 0.925 in the training group, ROC = 0.840 in the internal validation group, and ROC = 0.839 in the external validation group]. However, due to the prolonged duration and mild symptoms of POPF after MIEN, the International Study Group of Pancreatic Surgery (ISGPS) grading system is inadequate for accurately grading post-MIEN POPF. Therefore, a modified ISGPS classification was proposed, which further subdivides ISGPS grade B POPF based on different treatment modalities, providing an improved ability to distinguish patients. Conclusions: This study developed a personalized prediction model for accurately forecasting the occurrence of CR-POPF after MIEN. Additionally, the proposed modified ISGPS classification provides new insights into more targeted and precise grading for POPF following MIEN.