Pancreatic fat content by magnetic resonance imaging in subjects with prediabetes, diabetes, and controls from a general population without cardiovascular disease

by Sophia D. Heber, Holger Hetterich, Roberto Lorbeer, Christian Bayerl, Jürgen Machann, Sigrid Auweter, Corinna Storz, Christopher L. Schlett, Konstantin Nikolaou, Maximilian Reiser, Annette Peters, Fabian Bamberg

Despite the relevance of pancreatic fat content in the development of metabolic diseases, its association with impaired glucose metabolism, diabetes, and other adipose tissue compartments remains unclear. Thus, we determined differences in pancreatic fat content by magnetic resonance imaging (MRI) between subjects with prediabetes, diabetes, and normal controls in a cohort from the general population.

Subjects without history of cardiovascular disease with established diabetes or prediabetes as well as normal controls were included and underwent whole-body MRI on a 3T scanner. Pancreatic fat content was quantified by measuring the proton-density fat fraction (PDFF_panc) using a 3D multi-echo GRE sequence (increment: 1.23 ms, 6 echoes) by placing ROIs in the pancreatic head, body, and tail by independent readers. In addition, hepatic fat content as well as abdominal subcutaneous and visceral adipose tissue (SAT and VAT) were measured by multi-echo GRE and 3D 2-point volume-interpolated DIXON MRI, respectively. Univariate and multivariate analyses were employed to determine associations.

A total of 385 subjects were included in the analysis (median age: 57 years, 58.2% males), of them 53 were classified as subjects with diabetes, 95 as prediabetes, and 237 as controls (13.8%, 24.7%, and 61.6%; respectively). The median PDFF_panc was 5.2% [IQR 3.3–9.4], and significantly higher in subjects with prediabetes and diabetes as compared to controls (PDFF_panc: 6.2% [IQR: 3.5–12] vs. 8.6% [IQR: 4.3–17.5] vs. 4.9% [3.1–7.4], p<0.001, respectively). After adjusting for age, gender and BMI the association was attenuated (all p>0.12). While in univariate analysis BMI, PDFF_hepatic, SAT and VAT were associated with PDFF_panc (all p<0.05), only VAT predicted PDFF_panc independently (β: 0.02, 95%-confidence interval: 0.01–0.04, p<0.001).

While pancreatic fat content differs significantly between subjects with prediabetes, diabetes and controls, this association may be confounded by age, gender, and the amount of VAT in this cross-sectional study.