Non-branched microcysts of the pancreas on MR imaging of patients with pancreatic tumors who had pancreatectomy may predict the presence of pancreatic intraepithelial neoplasia (PanIN): a preliminary study.

Abstract

To evaluate whether pancreatic parenchymal abnormalities on magnetic resonance imaging (MRI) are associated with pancreatic intraepithelial neoplasia (PanIN) on histology. Retrospective study approved by institutional review board. One hundred patients (48 men, 52 women; mean age, 53.2 ± 16.29 [SD]) underwent MRI before pancreatectomy for pancreatic tumors analyzed by two independent observers blinded to histopathological results for the presence of non-communicating microcysts and pancreatic atrophy (global or focal) beside tumors. MRI findings were compared to histopathological findings of resected specimens. Interobserver agreement was calculated. The association between parenchymal abnormalities and presence of PanIN was assessed by uni- and multivariate analyses. PanIN was present in 65/100 patients (65%). The presence of microcysts on MRI had a sensitivity of 52.3% (34/65 [95%CI, 51.92-52.70%]), a specificity of 77.1% (27/35 [95%CI, 76.70-77.59]), and accuracy of 61% (61/100 95%CI [50.7-70.6]) for the diagnosis of PanIN while global atrophy had a sensitivity of 24.6% (16/6 [95%CI, 24.28-24.95]) and a specificity of 97.1% (34/35 [95%CI, 96.97-97.32%]). In multivariate analysis, the presence of microcysts (OR, 3.37 [95%CI, 1.3-8.76]) (p = 0.0127) and global atrophy (OR, 9.79 [95%CI, 1.21-79.129]) (p = 0.0324) were identified as independent predictors of the presence of PanIN. The combination of these two findings was observed in 10/65 PanIN patients and not in patients without PanIN (p = 0.013 with an OR of infinity [95%CI, 1.3-infinity]) and was not discriminant for PanIN-3 and lower grade (p = 0.22). Interobserver agreement for the presence of microcysts was excellent (kappa = 0.92), and for the presence of global atrophy, it was good (kappa = 0.73). The presence of non-communicating microcysts on pre-operative MRI can be a significant predictor of PanIN in patients with pancreatic tumors. • In patients with pancreatic tumors who had partial pancreatectomy, MR non-communicating pancreatic microcysts have a 52.3% sensitivity, a 77.1% specificity, and a 61% accuracy for the presence of PanIN with univariate and with an odds ratio of 3.37 with multivariate analyses. • The association of global atrophy and non-communicating microcysts increases the predictive risk of PanIN.