Evaluation of fatty pancreas by proton density fat fraction using 3-T magnetic resonance imaging and its association with pancreatic cancer.

PURPOSE: To evaluate whether pancreatic magnetic resonance imaging-proton density fat fraction (MRI-PDFF) correlates with histological pancreatic fat fraction and its possible usefulness as a biomarker of pancreatic cancer compared with pancreatic index (PI) using computed tomography (CT number of the pancreas divided by that of the spleen). METHOD: We included 55 consecutive patients (24 with pancreatic cancer and 31 controls; median age, 72 years) who preoperatively underwent MRI-PDFF using IDEAL-IQ and unenhanced CT and did not receive preoperative therapy. Histologic pancreatic fat fraction was measured in non-tumorous pancreatic tissues at the resection stump. A board-certified radiologist evaluated MRI-PDFF and PI. Correlations were evaluated among MRI-PDFF, PI, and histologic pancreatic fat fraction; the usefulness of MRI-PDFF as a predictor of pancreatic cancer was assessed. RESULTS: Histologic pancreatic fat fraction significantly correlated with MRI-PDFF and PI (r = 0.802 and -0.534, respectively; P < 0.01). The absolute correlation coefficient was significantly higher for MRI-PDFF than for PI (P < 0.01). Compared with the control group, the pancreatic cancer group had higher MRI-PDFF and histologic pancreatic fat fraction (P < 0.01) but lower PI (P < 0.01). In multivariate analysis, MRI-PDFF was found to be the sole independent risk factor for pancreatic cancer (odds ratio: 1.19; P < 0.01). CONCLUSIONS: Pancreatic fat, which was associated with pancreatic cancer, could be quantified by MRI-PDFF measurement; therefore, MRI-PDFF should be considered as a promising and superior imaging biomarker for estimating the probability of pancreatic cancer than PI.