The investigators diagnosed nonalcoholic fatty liver based on abdominal ultrasonography and ruled out excessive alcohol consumption and other etiologies of liver disease according to the Chinese Liver Disease Association.

The investigators used Cox proportional hazard models to estimate hazard ratios (HRs) and 95% CIs for the associations of weight change during smoking cessation with the incidence of T2DM. The investigators did not find any evidence of violation of the proportional hazard assumption. In multivariable models, the investigators constructed three nested models for analysis: (i) model 1 adjusted for age, and body mass index; (ii) model 2, model 1 plus drinking status; (iii) model 3, model 2 plus aspartate aminotransferase, total cholesterol, triacylglycerol, creatinine, albumin, and systolic blood pressure.

The investigators used Cox proportional hazard models to estimate hazard ratios (HRs) and 95% CIs for the associations of weight change during smoking cessation with the incidence of T2DM. The investigators did not find any evidence of violation of the proportional hazard assumption. In multivariable models, the investigators constructed three nested models for analysis: (i) model 1 adjusted for age, and body mass index; (ii) model 2, model 1 plus drinking status; (iii) model 3, model 2 plus aspartate aminotransferase, total cholesterol, triacylglycerol, creatinine, albumin, and systolic blood pressure.

The investigators conducted a mediation analysis to assess whether the association between smoking status and NAFLD was mediated by smoking relapse.