Hepatic Fat in Participants With and Without Incident Diabetes in the Diabetes Prevention Program Outcome Study

Abstract

Context: There is little information about fatty liver in prediabetes as it transitions to early diabetes.

Objective: This study is aimed at evaluating the prevalence and determinants of fatty liver in the Diabetes Prevention Program (DPP).

Methods: We measured liver fat as liver attenuation (LA) in Hounsfield units (HU) in 1876 participants at ~14 years following randomization into the DPP, which tested the effects of lifestyle or metformin interventions versus standard care to prevent diabetes. LA was compared among intervention groups and in those with versus without diabetes, and associations with baseline and follow-up measurements of anthropometric and metabolic covariates were assessed.

Results: There were no differences in liver fat between treatment groups at 14 years of follow-up. Participants with diabetes had lower LA (mean ± SD: 46 ± 16 vs 51 ± 14 HU; P < 0.001) and a greater prevalence of fatty liver (LA < 40 HU) (34% vs 17%; P < 0.001). Severity of metabolic abnormalities at the time of LA evaluation was associated with lower LA categories in a graded manner and more strongly in those with diabetes. Averaged annual fasting insulin (an index of insulin resistance [OR, 95% CI 1.76, 1.41-2.20]) waist circumference (1.63, 1.17-2.26), and triglyceride (1.42, 1.13-1.78), but not glucose, were independently associated with LA < 40 HU prevalence.

Conclusion: Fatty liver is common in the early phases of diabetes development. The association of LA with insulin resistance, waist circumference, and triglyceride levels emphasizes the importance of these markers for hepatic steatosis in this population and that assessment of hepatic fat in early diabetes development is warranted.

Trial registration: ClinicalTrials.gov NCT00004992.

Keywords: Lipid metabolism; diabetes development; hepatic fat; imaging; lifestyle; metformin; prediction and prevention of type 2 diabetes; weight regulation and obesity.

© The Author(s) 2021. Published by Oxford University Press on behalf of the Endocrine Society.

Figures

Figure 1.

CONSORT flow diagram.

Figure 2.

Mean (black dot) LA in HU and 95% CI (shaded bar) (Panel A) and prevalence of LA

Figure 3.

Empirical probability density functions of LA in the groups with and without diabetes. The total area under each curve is 1, and the proportion of participants with LA measurements between any 2 chosen values is the area under the curve between those points.

Figure 4.

Association of baseline covariates with LA levels and prevalence of LA

Figure 5.

Association of postrandomization covariates (average of all repeated measurements obtained after randomization until the time of the scan) with LA levels and prevalence of LA

Figure 6.

Adjusted models of mean LA differences with 95% CI and odds ratios with 95% CI for LA

Figure 7.

Effects of adding adjustment for baseline and mean follow-up values for triglyceride, waist circumference, FI, and HbA1c (Model 2-5) to model 1 adjusted for age, race, sex, and treatment group, comparing mean LA (Panel A) and prevalence of LA categories (Panel B) in the diabetes vs nondiabetes groups. A final model includes adjustment for all the covariates (Model 6). Effects are expressed as the ratio of the geometric mean and 95% CI of LA (Panel A) between diabetes vs nondiabetes from a linear model with logLA as outcome. The odds ratios (95% CI) between diabetes vs nondiabetes of having LA 50 (Panel B). The x-axis is reversed in Panel A so that values to the right of the null reference line indicate increasing hepatic fat (consistent with Panel B).

Figure 8.

Relationships among LA, diabetes status, and percent change from randomization to the year of the scan in weight (Panel A) and waist circumference (Panel B). Figures present expected mean LA and associated 95% confidence interval for different percent changes in waist and weight estimated from a linear model. Analyses were adjusted for demographic factors, interventions, and baseline and time-averaged FI, triglyceride, HbA1c, and year 1 adiponectin.