Comparison of subcutaneous, visceral, liver and muscle fat depots in relation to prevalent and incident diabetes

AIMS: To examine the association between fat distribution and diabetes. MATERIALS AND METHODS: We included UK Biobank participants who underwent an imaging visit and did not have type 1 diabetes at the time of that visit. Exposures were four MRI-measured fat depots, including abdominal subcutaneous adipose tissue (ASAT, in litres), visceral adipose tissue (VAT, in litres), liver fat (LF, in percentages) and muscle fat infiltration (MFI, in percentages). Logistic regression was used to estimate the odds ratios for prevalent diabetes, identified through self-reported diabetes diagnosis or non-metformin glucose-lowering medication use, or diagnosis records in primary care and hospitalizations before the imaging date, while Cox proportional hazards regression was used to estimate the hazard ratios (HRs) for incident diabetes, captured by hospitalizations only. RESULTS: Of the 44 857 participants (mean age: 66 years, 52% women) included, 2428 had prevalent diabetes. Of the remaining 42 429 who were free of diabetes at the imaging visit, 252 were hospitalized with diabetes over a median follow-up of 3.6 years. In age-, sex- and ethnicity-adjusted models, ASAT, VAT, LF and MFI were all associated with a higher likelihood of having or developing diabetes. Those associations were attenuated after additional adjustment of body mass index (BMI) or other fat depots, and generally followed the ranking of VAT>LF/MFI>ASAT. The HRs for incident diabetes, per standard deviation increase, adjusted for age, sex, ethnicity and BMI, were 2.44 (95% CI: 1.91-3.11) for VAT, 1.75 (95% CI: 1.56-1.97) for LF, 1.14 (95% CI: 0.98-1.34) for MFI, and 0.80 (95% CI: 0.63-1.02) for ASAT. CONCLUSIONS: Visceral fat showed a stronger association with diabetes than did other fat depots, highlighting the central role of visceral fat in the development of diabetes.