Prediction of anastomotic leakage after pancreatic head resections by dynamic magnetic resonance imaging (dMRI).

PURPOSE: The texture of the pancreatic tissue is a main risk factor for leakage after pancreaticojejunostomy and can be differentiated using dynamic contrast enhanced magnetic resonance imaging (dMRI). In order to identify risk factors and to assess the role of pancreatic dMRI, a cohort of patients was retrospectively reviewed. PATIENTS AND METHODS: One hundred seven consecutive patients were identified in the departmental database and examined by means of a standardized dMRI protocol using a 1.5-T MRI system. Signal intensity (SI) measurements (aorta, body of the pancreas, muscle tissue) were performed in the axial T1-weighted sequences before and after 25 and 60 s after i.v. application of gadolinium-diethylenetriaminepentaacetic acid. For all patients with a standardized contrast medium curve in the aorta (n = 72), a muscle-normalized signal intensity curve (SIC) with SI(ratio) was calculated. SI(ratio)s were classified in two groups: rapid increase (SI(ratio) >or= 1.1, early arterial value > portal-venous value, "soft" pancreas) and delayed increase (SI(ratio) <1.1, "firm" or "hard" pancreas). All patients received pancreatic head resection with a duct-to-mucosa pancreaticojejunostomy. The dMRI data was correlated with prospectively acquired clinical data.
RESULTS: Leakage of the pancreaticojejunostomy occurred more frequently (12/37 vs. two of 35, 32% vs. 6%, p = 0.006) in patients with a rapid increase and an SI(ratio) >or= 1.1 ("soft" pancreas, n = 37) compared to those with delayed perfusion (SI(ratio) <1.1, "hard" pancreas, n = 35). The more severe type B and C anastomotic leakages occurred only in the group of patients with SI(ratio) >or= 1.1. Patients with a rapid increase had significantly better preoperative American Society of Anesthesiologists staging, lower carbohydrate antigen 19-9 values, and smaller tumor sizes. Most of them had not only benign tumors but also longer postoperative hospital stay, in comparison to patients with delayed perfusion (SI(ratio) <1.1). Multivariate analysis revealed SI(ratio) of >or=1.1 to be the only preoperative parameter predicting leakage significantly with an odds ratio of 7.9.
CONCLUSION: dMRI with SI(ratio) calculation provided reliable information for the prediction of pancreatic texture. Patients with a SI(ratio) >or= 1.1 had a 7.9-fold increased risk of anastomotic leakage and a prolonged hospital stay. SIC with measurements of SI(ratio) in dMRI could therefore define patients at risk for anastomotic leakage.