Imaging features to distinguish malignant and benign branch-duct type intraductal papillary mucinous neoplasms of the pancreas: a meta-analysis.

Abstract

To systematically determine the imaging findings for distinguishing malignant and benign branch-duct type intraductal papillary mucinous neoplasms (BD-IPMNs), including mixed type, and their diagnostic value through meta-analysis of published studies. Management of BD-IPMNs, including mixed type, largely relies on imaging findings. The current knowledge on imaging findings to distinguish malignant and benign BD-IPMNs has weak evidence and is mostly from scattered individual retrospective studies. Thorough literature search in Ovid-MEDLINE and EMBASE databases was conducted to identify studies where findings of computed tomography, magnetic resonance imaging, and endoscopic ultrasonography of BD-IPMNs with or without main pancreatic duct (MPD) dilatation were correlated with surgical/pathological findings. Review of 1128 article candidates, including full-text review of 102 articles, identified 23 eligible articles with a total of 1373 patients for meta-analysis. Dichotomous data regarding distinction between malignant and benign BD-IPMNs were pooled using random effects model to obtain the diagnostic odds ratios (DORs) and their 95% confidence intervals (CIs) of various individual imaging findings for diagnosing malignant BD-IPMN. Presence of mural nodules revealed the highest pooled DOR (95% CI) of 6.0 (4.1-8.8) followed by MPD dilatation [3.4 (2.3-5.2)], thick septum/wall [unadjusted, 3.3 (1.5-6.9); publication bias-adjusted, 2.3 (0.9-5.5)], and cyst size greater than 3 cm [2.3 (1.5-3.5)]. Multilocularity and multiplicity of the cystic lesions did not reveal statistically significant association with malignancy. Presence of mural nodules should be regarded highly suspicious for malignancy warranting a surgical excision whereas cyst size greater than 3 cm, MPD dilatation (5-9 mm), or thick septum/wall may better be managed by careful observation and/or further evaluation.