Development of MR quantified pancreatic fat deposition as a cancer risk biomarker.

BACKGROUND: Excess body adiposity is associated with increased risk of pancreatic cancer, and in animal models excess intra-pancreatic fat is a driver of pancreatic carcinogenesis. Within a programme to evaluate pancreatic fat and PC risk in humans, we assessed whether MR-quantified pancreatic fat fraction (PFF) was 'fit for purpose' as an imaging biomarker. METHODS: We determined PFF using MR spectroscopy (MRS) and MR chemical shift imaging (CS-MR), in two groups. In Group I, we determined accuracy of MR-derived PFF with histological digital fat quantification in 12 patients undergoing pancreatic resection. In a second study, we assessed reproducibility in 15 volunteers (Group IIa), and extended to 43 volunteers (Group IIa & IIb) to relate PFF with MR-derived hepatic fat fraction (HFF), body mass index (BMI), and waist circumference (WC) using linear regression models. We assessed intra- and inter-observer, and between imaging modality levels of agreement using Bland-Altman plots. RESULTS: In Group I patients, we found strong levels of agreement between MRS and CS-MR derived PFF and digitally quantified fat on histology (rho: 0.781 and 0.672 respectively). In Group IIa, there was poor reproducibility in initial assessments. We refined our protocols to account for 3D dimensionality of the pancreas, and found substantially improved intra-observer agreements. In Group II, HFF and WC were significantly correlated with PFF (p values < 0.05). INTERPRETATION: Both CS-MR and MRS (after accounting for pancreatic 3D dimensionality) were 'fit for purpose' to determine PFF and might add information on cancer prediction independent from measures of general body adiposity.