Dairy consumption and incident prediabetes: prospective associations and network models in the large population-based Lifelines study

Background: Evidence on associations between dairy consumption and incident prediabetes is inconsistent. One potential explanation for heterogeneity is that health behavior and food intake may co-vary with consumption of various high-fat and low-fat dairy types.
Objective: To investigate the associations of total dairy and dairy types with incident prediabetes, and to assess how dairy intake is linked with metabolic risk factors, lifestyle behaviors and foods, as potential explanations for these associations.
Methods: A total of 74,132 participants from the prospective population-based Lifelines study were included (mean age 45.5±12.3 years, 59.7% female). Dairy intake was measured at baseline using a validated food frequency questionnaire. Prediabetes at follow-up was defined based on the WHO/IEC criteria as fasting plasma glucose of 110-125 mg/dl or A1C levels of 6.0-6.5%. Associations were analyzed using Poisson regression models adjusted for social demographics, lifestyle behaviors, family history of diabetes and food group intake. Interconnections were assessed with mixed graphical model (MGM) networks. 
Results: At a mean follow-up of 4.1±1.1 years, 2,746 participants developed prediabetes (3.7%). In regression analyses, neutral associations were found for most dairy types. Regular milk and low-fat milk intake were associated with a higher risk of prediabetes in the top vs. bottom quartiles (RR 1.17, 95%CI 1.05-1.30, ptrend=0.04 and 1.18, 95%CI 1.06-1.31, ptrend=0.01). Strong but non-significant effect estimates for high-fat yogurt in relation to prediabetes were found (RRservings/day 0.80, 95%CI 0.64-1.01, p=0.06), The network analysis showed that low-fat milk clustered with energy-dense foods including bread, meat, and high-fat cheese, however, high-fat yogurt had no clear link with lifestyle risk factors and food intake.
Conclusions: In this large cohort of Dutch adults, low-fat milk intake was associated with higher prediabetes risk. Heterogeneous associations by dairy type and fat content might partly be attributed to confounding by behaviors and food intake related to dairy intake.
Keywords (5-10): dairy, milk, yogurt, cheese, diet, prediabetes, type 2 diabetes, epidemiology, network analysis, mixed graphical model