The bidirectional relationship between NAFLD and type 2 diabetes: A prospective population-based cohort study

Background and aims To explore the bidirectional relationship between NAFLD and type 2 diabetes and the possible directions of the main effect. Methods and results 30 633 participants from the Jinchang cohort were enrolled. Firstly, cox proportional hazards regression model was used to assess the unidirectional causality between NAFLD and prediabetes and type 2 diabetes. Secondly, cross-lag path analysis model was conducted to estimate the bidirectional relationship between NAFLD and prediabetes and type 2 diabetes, and to determine the direction of the main effects. Finally, potential effect modifications were also considered by age, sex, hyperlipidemia, and overweight/obesity. We found that NAFLD increased the risk of prediabetes and type 2 diabetes with adjusted HR (95%CI) of 1.355(95%CI: 1.255–1.462) and 1.898(95%CI: 1.415–2.545), respectively. Prediabetes and type 2 diabetes also increased the risk of NAFLD, with adjusted HR (95%CI) of 1.245(95%CI: 1.115–1.392) and 1.592(95%CI: 1.373–1.846), respectively. Cross-lag path analysis showed that NAFLD significantly affected the incidence of prediabetes (β = 0.285, P < 0.001), while the effect on type 2 diabetes was not statistically significant. The effect of prediabetes and type 2 diabetes on the risk of NAFLD was weak, and the path coefficients were 0.076 and 0.037, respectively. Stratified analyses showed similar results. Conclusion This study provides evidence that there was a bidirectional causal association between NAFLD and type 2 diabetes, and the progression from NAFLD through prediabetes to type 2 diabetes may be the main pathway.