The impact of diets rich in low-fat or full-fat dairy on glucose tolerance and its determinants: a randomized controlled trial

Abstract

Background: Dairy foods, particularly yogurt, and plasma biomarkers of dairy fat intake are consistently inversely associated with incident type 2 diabetes. Yet, few trials assessing the impact of dairy on glucose homeostasis include fermented or full-fat dairy foods.

Objectives: We aimed to compare the effects of diets rich in low-fat or full-fat milk, yogurt, and cheese on glucose tolerance and its determinants, with those of a limited dairy diet.

Methods: In this parallel-design randomized controlled trial, 72 participants with metabolic syndrome completed a 4-wk wash-in period, limiting dairy intake to ≤3 servings/wk of nonfat milk. Participants were then randomly assigned to either continue the limited dairy diet, or switch to a diet containing 3.3 servings/d of either low-fat or full-fat dairy for 12 wk. Outcome measures included glucose tolerance (area under the curve glucose during an oral-glucose-tolerance test), insulin sensitivity, pancreatic β-cell function, systemic inflammation, liver-fat content, and body weight and composition.

Results: In the per-protocol analysis (n = 67), we observed no intervention effect on glucose tolerance (P = 0.340). Both the low-fat and full-fat dairy diets decreased the Matsuda insulin sensitivity index (ISI) (means ± SDs -0.47 ± 1.07 and -0.25 ± 0.91, respectively) and as compared with the limited dairy group (0.00 ± 0.92) (P = 0.012 overall). Body weight also changed differentially (P = 0.006 overall), increasing on full-fat dairy (+1.0 kg; -0.2, 1.8 kg) compared with the limited dairy diet (-0.4 kg; -2.5, 0.7 kg), whereas the low-fat dairy diet (+0.3 kg; -1.1, 1.9 kg) was not significantly different from the other interventions. Intervention effects on the Matsuda ISI remained after adjusting for changes in adiposity. No intervention effects were detected for liver fat content or systemic inflammation. Findings in intent-to-treat analyses (n = 72) were consistent.

Conclusions: Contrary to our hypothesis, neither dairy diet improved glucose tolerance in individuals with metabolic syndrome. Both dairy diets decreased insulin sensitivity through mechanisms largely unrelated to changes in key determinants of insulin sensitivity.This trial was registered at clinicaltrials.gov as NCT02663544.

Keywords: adiposity; cardiometabolic disease; dairy; diabetes; glucose tolerance; humans; inflammation; insulin sensitivity; liver fat; metabolic syndrome.

© The Author(s) 2020. Published by Oxford University Press on behalf of the American Society for Nutrition.

Figures

FIGURE 1

Consolidated Standards of Reporting Trials flow diagram.

FIGURE 2

Impact of diets limited in dairy (n = 22, A, D) or rich in low-fat dairy (n = 24, B, E) or full-fat dairy (n = 21, C, F) on plasma glucose (A–C) and insulin (D–F) concentrations in 3-h frequently sampled oral-glucose-tolerance tests. Circles and squares represent means at baseline (visit 1) and after the 12-wk dietary intervention (visit 2), respectively. Error bars represent SEMs.

FIGURE 3

Changes in measures of glucose homeostasis, insulin sensitivity, and pancreatic β-cell function during the limited dairy diet (n = 22), low-fat dairy diet (n = 24) and full-fat dairy diet (n = 21) (per-protocol analysis, n = 67). Glucose homeostasis and its determinants were assessed by (A) glucose tolerance, i.e., the glucose AUC, (B) the Matsuda ISI, (C) the HOMA-IR index, (D) fasting plasma insulin, (E) the insulinogenic index, (F) glucose sensitivity, (G) the oral DI, (H) fasting plasma glucose, and (I) HbA1c. Outcome variables are represented as the change variable calculated as the value at follow-up minus the value at baseline. Boxes represent IQRs, and whiskers 5th and 95th percentiles, with outliers represented by a solid dot. The medians are represented by horizontal bars across the boxes and the means are represented by crosses. The P values for the time × diet interactions from the overall RM-ANOVA, adjusted for baseline glucose AUC, fasting glucose, and HbA1c, as appropriate, are displayed at the top of each box plot. Bars indicate significant differences between diet groups in post hoc testing (independent-samples t tests or RM-ANOVA with 2 diet groups), again adjusted for baseline AUC glucose, fasting glucose, and HbA1c, as appropriate (*P < 0.05 after adjustment for multiple testing according to Bonferroni). HbA1c, glycated hemoglobin; ISI, insulin sensitivity index; oral DI, oral disposition index; RM-ANOVA, repeated-measures analysis of variance.

FIGURE 4

Changes in liver fat content (A) and measures of low-grade chronic systemic inflammation (B–D) in the limited dairy diet (n = 20 and n = 19, respectively), low-fat dairy diet (n = 22 and n = 20, respectively), and the full-fat dairy diet (n = 19 and n = 20, respectively) (per-protocol analysis, n = 61 for liver fat, n = 59 for biomarkers of inflammation). Outcome variables are represented as the change variable calculated as the value at follow-up minus the value at baseline. Boxes represent IQRs, and whiskers 5th and 95th percentiles, with outliers represented by a solid dot. The medians are represented by horizontal bars across the boxes and the means are represented by crosses. The P values for the time × diet interactions from the overall repeated-measures ANOVA are displayed at the top of each box plot. CRP, C-reactive protein.

FIGURE 5

Changes in measures of body weight, anthropometrics, and body composition in the limited dairy diet (n = 21), low-fat dairy diet (n = 24), and the full-fat dairy diet (n = 21) (per-protocol analysis, n = 66). (A) Body weight, (B) total fat mass, (C) lean mass, (D) waist circumference, (E) hip circumference, (F) waist-to-hip ratio, (G) trunk fat mass, (H) peripheral fat mass, and (I) visceral fat area. Outcome variables are represented as the change variable calculated as the value at follow-up minus the value at baseline. Boxes represent IQRs, and whiskers 5th and 95th percentiles, with outliers represented by a solid dot. The medians are represented by horizontal bars across the boxes and the means are represented by crosses. The P values for the time × diet interaction from the overall RM-ANOVA are displayed at the top of each box plot. Bars indicate significant differences between diet groups in post hoc (independent-samples t tests or RM-ANOVA with 2 diet groups) testing (*P < 0.05 after adjustment for multiple testing according to Bonferroni). RM-ANOVA, repeated-measures analysis of variance.