Metabolomic signatures of hepatic steatosis reveal heterogeneity in cardiometabolic risk and responses to lifestyle interventions in type 2 diabetes: a post hoc analysis.

Figures

• 

  Fig. 1. Modelling of the metabolomics-based liver fat. A Correlation between MRI-measured liver fat (liver fat) and model-predicted liver fat (pliver fat, blue scatter plot) at baseline. The blue and red histograms represent the distribution of liver fat and pliver fat, respectively. R2cor indicates optimism-corrected R2 estimated using bootstrap resampling. B Correlation between liver fat and metabolomics-based liver fat (mliver fat) (blue scatter plot) at baseline. The blue and green histograms represent the distribution of liver fat and mliver fat, respectively. R2cor indicates optimism-corrected R2 estimated using bootstrap resampling. C Distribution of all metabolites included in the construction of the mliver fat model, colored by super meta pathways. D Distribution of metabolites consistently selected across all 10 folds in the model, colored by super meta pathways. E Distribution of liver fat at baseline. F Distribution of mliver fat at baseline. G Distribution of Δliver fat at baseline. H Distribution of mliver fat according to different MRI-measured liver fat content. I Distribution of Δliver fat according to different MRI-measured liver fat content

• 

  Fig. 2. Validation of metabolomics-based liver fat model using after-intervention data. A Correlation between MRI-measured liver fat (liver fat) and model-predicted liver fat (pliver fat) (blue scatter plot) after intervention. The blue and red histograms represent the distribution of liver fat and pliver fat, respectively. R2cor indicates optimism-corrected R2 estimated using bootstrap resampling. B Correlation between liver fat and metabolomics-based liver fat (mliver fat) (blue scatter plot) after intervention. The blue and green histograms represent the distribution of liver fat and mliver fat, respectively. R2cor indicates optimism-corrected R2 estimated using bootstrap resampling. C Distribution of liver fat after intervention. D Distribution of mliver fat after intervention. E Distribution of Δliver fat after intervention. F Distribution of mliver fat according to different MRI-measured liver fat content. G Distribution of Δliver fat according to different MRI-measured liver fat content

• 

  Fig. 3. Metabolite differences between Δliver fat tertiles T3 and T1. A The differences in metabolite levels between T3 and T1 of Δliver fat at baseline. The x-axis represents the regression coefficient, and positive values indicate higher metabolite levels in T3 vs. T1; negative values indicate lower levels in T3 vs. T1. The y-axis shows the log10-transformed P values after Benjamini-Hochberg (BH) correction. Metabolites with P > 0.05 are shown in grey, while those with P ≤ 0.05 are colored according to their respective super meta-pathways. B The differences in metabolite levels between T3 and T1 of Δliver fat at week 12 (after intervention). The x-axis represents the regression coefficient, and positive values indicate higher metabolite levels in T3 vs. T1; negative values indicate lower levels in T3 vs. T1. The y-axis shows the log10-transformed P values after BH correction. Metabolites with P > 0.05 are shown in grey, while those with P ≤ 0.05 are colored according to their respective super meta-pathways. C Overlap of significantly different metabolites between T3 and T1 of Δliver fat at baseline and week 12 (after intervention). D The outermost ring represents baseline differences…

• 

  Fig. 4. Relationship between baseline metabolomics-based liver fat and cardiometabolic traits. A Linear associations of baseline metabolites which were consistently selected across all 10 folds in the model with cardiometabolic traits at baseline. B Linear associations of baseline MRI-measured liver fat (liver fat), metabolomics-based liver fat (mliver fat), and Δliver fat with cardiometabolic traits. Colors indicate the regression coefficients, and all traits were standardized using z-scores. *P < 0.05, **P < 0.01, and ***P < 0.001 after Benjamini-Hochberg (BH) correction. Abbreviations: ADA, adaptation index; BMI, body mass index; DBP, diastolic blood pressure; DI, disposition index; FIB-4, Fibrosis-4 Index; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin; HDL, high density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; ISI, insulin sensitivity index; LDL, low density lipoprotein; OGIS, oral glucose insulin sensitivity; PPG, post-load plasma glucose; SBP, systolic blood pressure; TG, triglyceride

• 

  Fig. 5. Relationship between after-intervention metabolomics-based liver fat and cardiometabolic traits. A Linear associations of after-intervention metabolites which were consistently selected across all 10 folds in the model with cardiometabolic traits after intervention. B Linear associations of after-intervention MRI-measured liver fat (liver fat), metabolomics-based liver fat (mliver fat), and Δliver fat with cardiometabolic traits. Colors indicate the regression coefficients, and all traits were standardized using z-scores. *P < 0.05, **P < 0.01, and ***P < 0.001 after Benjamini-Hochberg (BH) correction. Abbreviations: ADA, adaptation index; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; DBP, diastolic blood pressure; DI, disposition index; FIB-4, Fibrosis-4 Index; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin; HDL, high density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; ISI, insulin sensitivity index; LDL, low density lipoprotein; OGIS, oral glucose insulin sensitivity; PPG, post-load plasma glucose; SBP, systolic blood pressure; TG, triglyceride

• 

  Fig. 6. The effect of diet intervention on cardiometabolic traits across baseline Δliver fat. Linear model was used to estimate the changes of cardiometabolic traits from baseline to 12 weeks and 48 weeks between diet intervention and control group stratified by baseline Δliver fat. P values indicate treatment × Δliver fat tertile (T3) interaction. Abbreviations: BMI, body mass index; DI, disposition index; HbA1c, glycated hemoglobin; ISI, insulin sensitivity index