Predicting high-risk pathology across the spectrum of resected and surveilled intraductal papillary mucinous neoplasms: A cohort study.

Abstract

Clinical practice guidelines for intraductal papillary mucinous neoplasms are based on expert opinion because of paucity of clinical evidence. We aim to establish the data-driven correlation between worrisome/high-risk/clinically relevant progression features and high-risk pathology on fine needle aspiration or resection. A prospectively maintained database (1997-2023) of presumed pancreatic cystic neoplasms was queried for intraductal papillary mucinous neoplasm with potentially concerning feature(s) per Fukuoka guidelines. Association and predictive power of specific features was examined via logistic mixed effects modeling and least absolute shrinkage and selection operator regression. Of the 2,686 patients diagnosed with intraductal papillary mucinous neoplasms, 460 (17.1%) had a feature of clinically relevant progression. Median follow-up was 7.1 years (interquartile range 2.99-11.9). Most (n = 365; 79%) were offered pancreatic resection with 230 (63%) undergoing resection. Sixty-nine (15.6%) developed invasive carcinoma. Endoscopic ultrasonography-guided cytology at diagnosis demonstrated a sensitivity of 28.4% (95% confidence interval 18.0%-40.7%) and specificity 98.9% (97.2%-99.7%) for high-risk pathology. Endoscopic ultrasonography-guided cytology after clinically relevant progression demonstrated a specificity of 100% (95% confidence interval 92.1%-100%) and sensitivity 16.1% (5.5%-33.7%).On mixed effects modeling, enhancing nodule (odds ratio 24.6, 95% confidence interval 6.58-91.74), main pancreatic duct involvement (odds ratio 4.77, 95% confidence interval 1.18-14.05), and symptoms (hazard ratio 12.139, 95% confidence interval 1.786-82.48) predicted high-risk pathology; other features, including size or size growth, did not (conditional pseudo-R² = 0.218, marginal = 0.243). On least absolute shrinkage and selection operator analysis, enhancing nodule was the strongest predictor of both high-risk pathology and invasive carcinoma followed by main pancreatic duct dilatation and thick cyst wall. Age, body mass index, cyst size, and rate of size growth all had coefficients converging to zero. Enhancing nodule and any degree of main duct dilatation in an intraductal papillary mucinous neoplasm portend a high risk of malignant pathology, whereas cyst size and growth rate notably did not. These data should aid clinical management and might inform future practice guidelines.