Improving predictive accuracy for surgical indication of the current management guideline for intraductal papillary mucinous neoplasm of the pancreas via cutoff adjustments.

objectiveTo improve the predictive accuracy of international guidelines for malignant pancreatic intraductal papillary mucinous neoplasm (IPMN) via cutoff adjustments. MATERIALS AND

methodsWe retrospectively analyzed surgically confirmed IPMN patients at a single center between 2011 and 2023. Preoperative MRI were reviewed based on international guidelines. We calculated sensitivity and specificity of high-risk stigmata (HRS). Receiver operating characteristic analyses for enhancing mural nodule (EMN) size and main pancreatic duct (MPD) diameter identified cutoffs achieving specificities of 85%, 90%, and 95%. Revised HRS criteria (HRS-85/90/95) and 6-point scales were established based on these cutoffs, and compared with the original HRS-based 6-point scale using DeLong's test.

resultsOf 418 patients (mean age, 66.6 ± 8.2 years; 263 men) analyzed, 173 had malignant IPMN. The original HRS showed 82.7% (95% confidence interval [CI]: 76.2-88.0%) sensitivity and 66.9% (95% CI: 60.7-72.8%) specificity. Obstructive jaundice and suspicious/positive cytology were highly specific (98.8-99.6%), but EMN ≥ 5 mm and MPD ≥ 10 mm were less specific (80.4%). Feature-level thresholds achieving specificities of ≥85%, 90%, and 95% were 9.5/12.0/17.5 mm for EMN and 11.0/12.0/13.5 mm for MPD. Optimism-corrected sensitivities and specificities of HRS-85, 90, and 95 were 74.4%/76.7%, 65.2%/82.4%, and 53.1%/89.9% (95% CIs: 68.2-80.9%/71.0-80.2%, 58.4-72.8%/78.0-86.5%, and 45.7-60.1%/85.7-93.5%), respectively. The 6-point scale area under the curve (AUC) improved from 0.771 to 0.795-0.808 (apparent AUC differences, 0.027-0.041; Holm-Bonferroni adjusted p = 0.046).

conclusionCurrent HRS criteria carry a high false-positive rate for malignancy. Increasing EMN and MPD cutoffs improves specificity and AUC, though sensitivity decreases. Thresholds should be optimized according to the clinical context. KEY POINTS: Question Current surgical indications for pancreatic IPMN show high false-positive rate for malignancy; therefore, improving predictive accuracy for malignant IPMN is warranted. Findings Revised HRS criteria with increased EMN and MPD cutoffs revealed higher specificity at the expense of sensitivity, and significantly improved 6-point scale AUC. Clinical relevance Adjusting cutoff values for EMN size and MPD diameter improves specificity and AUC in predicting malignancy for IPMN, potentially guiding more targeted surgical interventions. However, the associated reduction in sensitivity necessitates a balanced, individualized assessment in clinical practice.