Ketone Body Metabolism and Sodium Glucose Cotransporter 2 Inhibitor Response in Heart Failure With Ischemic Cause: A Precision Medicine Approach Using Metabolomics Integration With Global Burden of Disease Population Data Across Diverse Genetic Backgrounds

Figures

• 

  Figure 1. Distribution of baseline ketone body metabolites across genetic ancestry groups. Violin plots with embedded box plots displaying the distribution of baseline ketone body metabolites across 5 genetic ancestry groups in patients with heart failure of ischemic cause initiating SGLT2 inhibitor therapy (N=3847). A, (β‐HB) concentrations. B, Acetoacetate concentrations. C, Total ketone body concentrations (β‐HB+acetoacetate+ acetone). Statistical analysis: 1‐way ANOVA with post hoc pairwise comparisons using Bonferroni correction. β‐HB indicates β‐Hydroxybutyrate; and SGLT2, sodium glucose cotransporter 2.

• 

  Figure 2. Kaplan–Meier survival analysis by baseline β‐hydroxybutyrate tertiles. Kaplan–Meier curves for the primary composite end point (cardiovascular death, heart failure hospitalization, or ≥40% decline in eGFR) stratified by baseline β‐hydroxybutyrate tertiles over 48 mo of follow‐up (N=3847). Tertile definitions: T1 (lowest), 0.28 to 1.67 mmol/L; T2 (middle), 1.68 to 2.41 mmol/L; T3 (highest), 2.42 to 4.89 mmol/L. Shaded regions represent 95% CIs. The vertical dashed line indicates the 36‐mo primary end point assessment time. Event rates at 36 mo were 21.3% (T1), 30.7% (T2), and 43.9% (T3). HR indicates hazard ratio.

• 

  Figure 3. Global burden of ischemic heart disease by world region (GBD 2021). Regional variation in ischemic heart disease burden based on GBD 2021 estimates across 15 world regions. A, Age‐standardized DALY rates per 100 000 population with 95% uncertainty intervals. The red dashed line indicates the global average (2212.2 per 100 000). Central Asia demonstrated the highest burden (4864.5 per 100 000), ∼5.8‐fold higher than Western Europe (843.8 per 100 000). B, Total deaths in millions by region (global total: 9.0 million). C, Total prevalent cases in millions by region (global total: 254.3 million). DALY indicates disability‐adjusted life year; and GBD, Global Burden of Disease.

• 

  Figure 4. Metabolic pathway enrichment analysis. Pathway enrichment analysis identifying metabolic pathways significantly associated with SGLT2 inhibitor nonresponse (false discovery rate <0.05). A, Bubble plot displaying pathway impact scores vs statistical significance (−log10 P value). Bubble size is proportional to the number of metabolite hits per pathway. Colors indicate regulatory direction in nonresponders: red (upregulated), blue (downregulated), teal (mixed regulation). Horizontal dashed lines indicate significance thresholds at P=0.05 and P=0.001. B, Horizontal bar plot showing the 14 significantly enriched pathways ranked by statistical significance, with pathway impact scores annotated. FDR indicates false discovery rate; KEGG, Kyoto Encyclopedia of Genes and Genomes; SGLT2, sodium glucose cotransporter 2; and TCA, tricarboxylic acid.

• 

  Figure 5. Machine learning feature importance analysis. Feature importance analysis from the XGBoost prediction model for SGLT2 inhibitor nonresponse (N=3847; 5‐fold cross‐validation). A, SHAP beeswarm plot displaying the contribution of each feature to model predictions. Each point represents a patient; horizontal position indicates the SHAP value (impact on model output); color indicates feature value (blue=low, red=high). Positive SHAP values indicate increased predicted probability of nonresponse. B, Relative feature importance scores for the top 20 predictors, colored by metabolite category. β‐hydroxybutyrate demonstrated the highest importance (100.0), followed by acetoacetate (78.3) and C4‐carnitine (65.2). C, Donut chart showing aggregate contribution of each feature category to the prediction model. AUC‐ROC indicates area under the receiver operating characteristic curve; eGFR, estimated glomerular filtration rate; HMG, 3‐hydroxy‐3‐methylglutaryl‐coenzyme A synthase 2; SGLT2, sodium glucose cotransporter 2; SHAP, Shapley Additive Explanations; TCA, tricarboxylic acid; and XGBoost, Extreme Gradient Boosting.

• 

  Figure 6. Mendelian randomization analysis of β‐hydroxybutyrate and cardiovascular outcomes. Two‐sample MR scatter plots examining the causal relationship between genetically predicted β‐hydroxybutyrate levels and cardiovascular outcomes. A, Heart failure hospitalization. B, Cardiovascular death. HF indicates heart failure; IVW, inverse‐variance weighted; MR, Mendelian randomization; OR, odds ratio; and SNV, single‐nucleotide variant.

• 

  Figure 7. Subgroup analysis: association of baseline β‐hydroxybutyrate tertiles with primary composite end point. Forest plot displaying HR and 95% CIs for the association between baseline β‐HB tertiles and the primary composite end point across prespecified clinical subgroups. The analysis compares Tertile 3 (highest β‐HB, 2.42–4.89 mmol/L) vs Tertile 2 (middle, 1.68–2.41 mmol/L) as the reference group. β‐HB indicates β‐hydroxybutyrate; eGFR, estimated glomerular filtration rate; HbA1c, glycated hemoglobin; HR, hazard ratio; LVEF, left ventricular ejection fraction; and NT‐proBNP, N‐terminal pro‐B‐type natriuretic peptide.