We retrospectively analyzed 304 patients who underwent CABG at the Affiliated Hospital of Qingdao University from January 2018 to December 2022. Telephone follow-up was conducted in November to December 2024. The median follow-up duration was 44 months (IQR 31–62 months), during which 82 patients experienced MACE.

Inclusion criteria: age ≥18 years, preoperative CCTA during hospitalization, complete medical records, and ability to complete follow-up.

Exclusion criteria: concomitant surgery (valve surgery, surgical ablation, or congenital heart surgery); prior PCI or CABG; severe cardiomyopathy (LVEF <40% or severe left ventricular dilation [LVEDD >65 mm]); severe pulmonary, hepatic (Child-Pugh B/C), or renal disease (eGFR <30 mL/min/1.73 m²); malignancy; severely elevated triglycerides (TG ≥5.65 mmol/L) or suspected familial hypertriglyceridemia; or missing key data.

This was a non-interventional, retrospective study approved by the institutional ethics committee and conducted in accordance with the Declaration of Helsinki.

Clinical data were obtained from the hospital’s electronic medical record system, including demographic characteristics, medical history, comorbidities, and medication use. Venous blood samples were collected after overnight fasting (from 10:00 PM the previous night) at approximately 6:00 AM on the day following hospital admission, prior to surgery. Fasting plasma glucose (FPG), serum creatinine (SCr), and lipid profiles were measured according to standard hospital laboratory procedures.

Multivessel disease was defined as two- or three-vessel coronary artery involvement, and left main disease as ≥50% stenosis in the left main coronary artery, as assessed by preoperative coronary computed tomography angiography (CCTA). Hypertension was defined as systolic blood pressure (SBP) ≥ 140 mmHg, diastolic blood pressure (DBP) ≥ 90 mmHg, or use of antihypertensive. Diabetes was defined as random glucose ≥ 11.1 mmol/L, FPG ≥7.0 mmol/L, or use of hypoglycemic medications. Hyperlipidemia was defined as the use of total lipid-lowering agents or a total cholesterol ≥ 240 mg/dL or use of lipid-lowing therapy (20).

The TyG index was calculated as Ln [fasting TG (mg/dL) × FPG (mg/dL)/2], as originally described by Guerrero-Romero et al. (2010) (21). Epicardial adipose tissue (EAT) volume was quantified from coronary computed tomography angiography (CCTA) images using dedicated post-processing software (AWS). Two radiologists independently performed the measurements, with any discrepancies resolved by a senior radiologist. CCTA was conducted on a 128-slice dual-source cardiac CT scanner with acquisition parameters of 100 kV, 601 mA, and a slice thickness of 0.66 mm. Imaging covered the region from the aortic root to the cardiac apex. Three-dimensional reconstruction was performed and EAT volume was automatically calculated by the software.

MACE included all-cause mortality, repeat coronary revascularization, heart failure, severe arrhythmias, and stroke. Repeat revascularization was defined as subsequent percutaneous coronary intervention (PCI) or CABG of the index lesion. Heart failure was defined as impaired cardiac function due to myocardial ischemia, including cases where new-onset atrial fibrillation led to heart failure exacerbation requiring hospitalization. Severe arrhythmias included sustained ventricular tachycardia, high-grade atrioventricular block, and clinically significant sinus bradycardia (heart rate <40 bpm or requiring intervention). Stroke was defined as ischemic stroke due to interruption of cerebral blood flow; no hemorrhagic stroke events occurred in this cohort.

Statistical analyses were performed using SPSS 29.0 and R 4.4.2. A two-tailed p-value < 0.05 was considered statistically significant.

Receiver operating characteristic (ROC) curve was used to determine optimal cut-off values for the TyG index and EAT volume in predicting MACE. The optimal cut-offs were 8.65 for TyG (AUC = 0.624, 95% CI: 0.554–0.694, p < 0.05) and 116.2 for EAT (AUC = 0.586, 95% CI: 0.510–0.661, p < 0.05). Based on these cut-offs, patients were stratified into four groups: Group 1, low TyG and low EAT (TyG ≤ 8.65, EAT ≤ 116.2); Group 2, low TyG and high EAT (TyG ≤ 8.65, EAT > 116.2); Group 3, high TyG and low EAT (TyG > 8.65, EAT ≤ 116.2); and Group 4, high TyG and high EAT (TyG > 8.65, EAT > 116.2).

Baseline characteristics were compared across groups. Normality of continuous variables was evaluated using the Shapiro–Wilk test in SPSS, and all variables were determined to be non-normally distributed. Consequently, continuous variables were summarized as median (IQR) and compared using appropriate non-parametric tests (Mann–Whitney U or Kruskal–Wallis). Categorical variables were reported as counts (percentages) and compared using the chi-square test. Because no variable met the assumption of normality, assessment of variance homogeneity was not applicable. Potential outliers were assessed through boxplot visualization, and as no observations were indicative of measurement or data-entry errors, all data points were retained for analysis. Kaplan–Meier curves were generated to compare cumulative event rates among groups, with log-rank tests used for survival differences.

Cox proportional hazards regression was used to evaluate associations of TyG index and EAT volume with MACE. Model 1 adjusted for age and sex; Model 2 included variables with P < 0.05 in univariate analysis; Model 3 was further adjusted for all variables in Model 2 plus age, sex, left main disease, triglyceride level, and serum creatinine, based on penalized variable selection and collinearity assessment. To reduce model complexity and minimize the risk of overfitting, multicollinearity among candidate variables was assessed using variance inflation factors (VIFs), and least absolute shrinkage and selection operator (LASSO) regression was applied as a penalized variable selection approach before fitting the multivariable Cox model (22). TyG index and EAT volume were first analyzed as continuous variables, standardized into z-scores to reflect the effect per 1 SD increase and reduce outlier influence. They were also analyzed as categorical variables based on ROC cut-offs. Additive interaction between TyG index and EAT volume was assessed using RERI, AP, and SI (23–25). The SI represents a measure of statistical additive interaction; it should not be interpreted as indicating a mechanistic or biological synergistic effect. Incremental predictive value was evaluated with the C-statistic, net reclassification improvement (NRI), and integrated discrimination improvement (IDI), and model fit was assessed using the AIC, BIC, and the likelihood ratio test.

ROC curve analysis identified an optimal TyG index cut-off of 8.65 for predicting MACE, based on the maximum Youden index. Baseline characteristics stratified by this cut-off are presented in Table 1. Of the 304 patients, those with TyG > 8.65 were more often female, had lower mean age, and exhibited higher BMI, FPG, TG, TC, and LDL-C, but lower HDL-C and EAT volume. They also had a higher prevalence of diabetes, greater use of glucose-lowering drugs, and an increased risk of MACE.

Patients were also stratified by the optimal EAT volume cut-off (Table 2). Compared with those with lower EAT volume, patients with higher EAT volume showed a similar age distribution and a lower proportion of males. They had higher FPG and TG levels and were more likely to experience MACE.

During a median follow-up of 44 months (IQR31–62), 82 patients (27.0%) experienced MACE.5 (1.6%) all-cause mortality, 28 (9.2%) repeat revascularization, 10 (3.3%) of heart failure, 7 (2.3%) severe arrhythmias, and 36 (11.8%) nonfatal stroke. As some patients experienced multiple adverse events during follow-up, the summed counts of individual event categories exceed the total number of patients with MACE.

Kaplan–Meier showed significantly higher cumulative MACE incidence in patients with TyG > 8.65 compared with those ≤ 8.65 and in those with EAT > 116.2 compared with lower volumes (p < 0.0001, Figure 1.).

In univariate Cox regression, both TyG index and EAT volume were significantly associated with MACE; these associations persisted after adjustment for conventional cardiovascular risk factors (Table 3; Figure 2.). Compared with patients with low TyG/low EAT volume, high TyG was associated with a 2.40-fold increased risk of MACE, and high EAT with a 2.92-fold higher risk.

Patients were stratified into four groups based on TyG (cut-off 8.65) and EAT volume (cut-off 116.2 cm³). Kaplan–Meier curves showed the highest cumulative MACE incidence in the high TyG/high EAT group and the lowest in the low TyG/low EAT group (p < 0.0001, Figure 3.).

Univariate Cox regression, high TyG/high EAT was associated with the greatest MACE risk (HR 8.39, 95% CI 3.68–19.10, p < 0.001). This remained significant after adjustment for confounders. In the fully adjusted model, the high TyG/high EAT group had an adjusted HR of 7.62 (95% CI 3.27–17.76) compared with the low TyG/low EAT reference group (Table 4).

As shown in Table 5, the combined effect of elevated TyG and increased EAT volume on MACE risk exceeded the sum of their individual effects. The relative excess risk due to interaction (RERI) was 3.81, the attributable proportion (AP) was 0.50, and the synergy index (SI) was 2.34, all statistically significant (p < 0.05). These values indicate that 50% of the excess MACE risk in patients with both risk factors could be attributed to their interaction, and that the combined presence of high TyG and high EAT more than doubled the expected risk if acting independently.

As shown in Table 6, adding either TyG index or EAT volume individually to the baseline model moderately improved predictive performance. When both TyG and EAT were included simultaneously, the C-statistic increased, indicating improved discrimination for identifying patients at risk of MACE. Consistently, the NRI showed enhanced risk reclassification, and the IDI demonstrated a significant improvement in the model’s overall predictive accuracy. These results indicate that combined use of TyG index and EAT volume provides substantial incremental predictive value for MACE risk assessment.

As shown in Table 7, adding TyG index and EAT volume progressively reduced AIC and BIC, indicating improved model fit. χ² values increased, further supporting better data fitting. All p-values were < 0.05, confirming that the model including both variables outperformed the baseline model.