Related Experiment Video
Updated: Mar 14, 2026
Technical Detail for Robot Assisted Pancreaticoduodenectomy
Published on: September 28, 2019
Re-evaluating routinely collected clinical and laboratory parameters in the preoperative risk assessment of
Camila Hidalgo Salinas1, Mahip Grewal2, Vishnu Jayaprakash2
1Department of Surgery, NYU Grossman School of Medicine, New York, NY, USA; Centre for Evidence-Based Medicine, Nuffield Department of Primary Care Health Sciences, University of Oxford, Radcliffe Observatory Quarter, Woodstock Road, Oxford, OX2 6GG, United Kingdom; Department for Continuing Education, University of Oxford, Rewley House, 1 Wellington Square, Oxford, OX1 2JA, United Kingdom.
Background:
Accurate preoperative malignancy risk assessment in intraductal papillary mucinous neoplasm (IPMN) is essential to balance timely intervention for high-grade dysplasia or invasive cancer (HGD/IC) against avoiding unnecessary or premature surgery in low-grade IPMN. This study aimed to externally validate the 2023 International Association of Pancreatology (IAP)/Kyoto guidelines and develop a combined prediction model incorporating routinely collected clinical data and laboratory parameters.
Methods:
We conducted a retrospective cohort study of 194 patients who underwent resection for IPMN between 2012 and 2024. We evaluated the predictive performance of the current IAP/Kyoto criteria ("Kyoto model"), developed a clinical model using routinely available laboratory and clinical variables, and integrated both into a combined model. Model performance was assessed using discrimination and calibration metrics, with internal validation via bootstrapping and five-fold cross-validation.
Results:
The Kyoto model demonstrated modest discrimination (AUC 0.62). The clinical model, incorporating neutrophil-to-lymphocyte ratio (NLR), smoking history, blood glucose, CA19-9, and alkaline phosphatase, achieved an optimism-corrected AUC of 0.76. Compared to the Kyoto model, the combined model (AUC 0.77) significantly improved discrimination and calibration (p < 0.001). At a predicted probability threshold of >0.75, the combined model achieved a 90% specificity and 91% positive predictive value for HGD/IC, identifying a high-risk subgroup suitable for surgical intervention.
Conclusions:
Integrating routinely collected clinical and laboratory parameters with guideline-based imaging features shows promise to enhance preoperative identification of high-risk IPMN in patients already being considered for surgical resection. The combined model offers a practical, high-specificity tool to support surgical decision-making in this selected population, though its performance metrics should not be extrapolated to unselected surveillance cohorts. External validation is required before broader clinical implementation.
