Joint assessment of metabolic and functional indices for coronary burden: a real-world study in patients undergoing angiography for chest pain

Figures

• 

  Figure 1. Restricted cubic spline (RCS) analyses of TyG index and LVEF in relation to CHD risk. (A) Dose-response association between the TyG index and CHD risk estimated using an RCS logistic regression model. (B) Non-linear association between left ventricular ejection fraction (LVEF) and CHD risk based on an RCS model. Histograms depict the distribution of TyG and LVEF in the study population. P values for overall association and non-linearity are shown in each panel. Adjustment was made for all covariates, including age, sex, and BMI; smoking status and alcohol consumption; and comorbidities of hypertension (HTN), hyperlipidaemia (HLP), and type 2 diabetes mellitus (T2DM).

• 

  Figure 2. Subgroup analyses of the associations between TyG, LVEF, and CHD. Forest plots present odds ratios (ORs) with 95% confidence intervals (CIs) for CHD across predefined subgroups stratified by sex, age, BMI category, smoking status, alcohol consumption, type 2 diabetes mellitus (T2DM), and hypertension (HTN). (A) Association between the TyG index and CHD within each subgroup. (B) Association between left ventricular ejection fraction (LVEF) and CHD within each subgroup. The vertical reference line indicates OR = 1.0, and estimates are displayed on a logarithmic scale.

• 

  Figure 3. ROC curve analysis for CHD discrimination using TyG, LVEF, the base clinical model, and the TyG-LVEF-augmented model. Receiver operating characteristic (ROC) curves were constructed to compare the discriminatory performance of the TyG index alone, LVEF alone, a base clinical model, and an augmented model for identifying CHD. The base clinical model included age, sex, BMI, smoking status, alcohol consumption, and hypertension (HTN); the augmented model additionally included TyG and LVEF. The area under the ROC curve (AUC) with 95% confidence intervals (CIs) is reported for each curve. The diagonal dashed line denotes no-discrimination (AUC=0.50). The AUC increased from 0.655 for the base model to 0.694 for the augmented model (DeLong P < 0.001).

• 

  Figure 4. Decision-curve analysis of the base and TyG-LVEF-augmented models for identifying angiographically confirmed CHD. Net-benefit curves were calculated using individual predicted probabilities from 20 repeats of stratified 10-fold cross-validation. The base clinical model included age, sex, BMI, smoking status, alcohol consumption, and hypertension; the augmented model additionally included TyG and LVEF. The augmented model provided greater net benefit than the base model between threshold probabilities of 0.20 and 0.75. Treat-all and treat-none strategies are shown for reference. This internally validated analysis remains exploratory and does not establish external transportability.

• 

  Figure 5. Spearman correlation analyses between metabolic/cardiac functional indicators and coronary lesion severity assessed by the gensini score. (A) TyG index showed a weak positive correlation with Gensini score (r = 0.079, P = 0.025). (B) LVEF displayed an inverse correlation with Gensini score (r = −0.25, P < 0.001). Scatter plots include fitted trend lines with 95% confidence bands; marginal density plots depict the distributions of TyG/LVEF and Gensini scores. Adjustment was made for all covariates, including age, sex, and BMI; smoking status and alcohol consumption; and comorbidities of hypertension (HTN), hyperlipidaemia (HLP), and type 2 diabetes mellitus (T2DM).

• 

  Figure 6. Distribution of coronary lesion severity (gensini score) across nine joint TyG–LVEF categories defined by tertiles (TyG: T1–T3; LVEF: T1–T3). Boxplots are ordered by the median Gensini score from low to high. Overall differences among the nine groups were assessed using the Kruskal–Wallis test (P = 3.1 × 10⁻⁹). Boxes represent the interquartile range (IQR) with the median shown as the central line; whiskers indicate the data spread beyond the IQR. Adjustment was made for all covariates, including age, sex, and BMI; smoking status and alcohol consumption; and comorbidities of hypertension (HTN), hyperlipidaemia (HLP), and type 2 diabetes mellitus (T2DM).