Association of triglyceride–glucose index with coronary severity and mortality in patients on dialysis with coronary artery disease

The present study utilized data derived from the Coronary Revascularization in Patients On Dialysis in China-Retrospective (CRUISE-R) study (ClinicalTrials.gov NCT05841082). The CRUISE-R study was a multi-center, observational registry in China that aimed to investigate clinical characteristics, medical care, and prognostic factors of patients on dialysis with CAD. The CRUISE-R study retrospectively evaluated 455,617 cardiac catheterizations conducted between January 2015 and June 2021 in China. Several exclusion criteria were applied, including patients who did not receive dialysis therapy or had undergone dialysis treatment for less than 3 months (n = 453,421), those without any coronary stenosis exceeding 50% (n = 328), and individuals admitted for reasons other than coronary angiography (such as surgical interventions, valve diseases, or kidney transplants) (n = 87). For patients who were readmitted to the hospital, only data from their initial admission were considered for inclusion in our analysis. Subsequent readmissions were documented as adverse events, specifically as "readmission." (n = 532). As a result, a total of 1,249 patients with obstructive CAD on dialysis were enrolled in the CRUISE-R study. The study was conducted under the Declaration of Helsinki and received approval from the China–Japan Friendship Hospital Ethics Committee, with a waiver of informed consent.

For the present analysis, we further excluded 31 patients suspected of familial hypertriglyceridemia (TG ≥ 5.65 mmol/L), and 157 patients who lacked the necessary data required for TyG index calculation. Ultimately, the current analysis included 1061 participants (Fig. 1). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.

The data were collected from electronic medical records at each participating center by qualified study coordinators. Baseline demographic and clinical information included age, gender, systolic blood pressure, diastolic blood pressure, heart rate, hypertension, diabetes mellitus, current smokers, atrial fibrillation, cerebrovascular disease, valvular disease, peripheral arterial disease, and index presentation. Dialysis details encompassed information about dialysis modality, duration of dialysis (vintage), and dialysis cause. Laboratory measurements, such as hemoglobin, serum creatinine, TG, TC, low-density lipoprotein cholesterol (LDL-C), and HDL-C, were also obtained. The formula used to calculate the TyG index is as follows: ln [(fasting TG (mg/dl) × glucose (mg/dl)/2], using glucose and TG levels obtained within 24 h of admission [22]. In addition, details regarding coronary angiography were recorded, including the access method used, the extent of disease, the presence of moderate or severe calcification, and treatment by percutaneous coronary intervention (PCI). Furthermore, information on medication usage, such as dual antiplatelet therapy, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker, β-blocker, calcium-channel blocker, and statin usage, was documented.

In this study, coronary extent and severity were assessed by the multivessel disease and Gensini score (GS) [31]. Multivessel disease is characterized by the presence of noteworthy stenosis, representing ≥ 50% of two or more main coronary arteries and/or the left main coronary artery. The GS is an assessment tool utilized in coronary angiography to evaluate the extent and severity of stenosis in coronary artery segments based on location and degree of narrowing [31]. The GS is scored as follows: 1 for 1–25% stenosis, 2 for 26–50% stenosis, 4 for 51–75% stenosis, 8 for 76–90% stenosis, 16 for 91–99% stenosis, and 32 for total occlusion. This score is then multiplied by a factor that considers the significance of the lesion’s position within the coronary arterial tree. Lesions located in the left main coronary artery were assigned a weight of 5, while the proximal left anterior descending coronary artery (LAD) or proximal left circumflex coronary artery (LCX) were assigned a weight of 2.5. Lesions located in the mid-region of the LAD were assigned a weight of 1.5, as were those located in the proximal and mid-region of the right coronary artery (RCA). Finally, lesions in the distal region of the LAD and the mid-distal regions of the LCX and RCA were assigned a weight of 1. The summation of the individual scores of each coronary artery segment produces the GS. High GS corresponds to the top tertile of scores, indicating severe coronary stenosis.

The primary outcome for follow-up was all-cause death, with the secondary outcome being cardiovascular death. Cardiovascular death was defined as death resulting from acute myocardial infarction, heart failure, sudden cardiac death, stroke, cardiovascular procedures, or cardiovascular hemorrhage. Survival and clinical assessment data were gathered by trained nurses through outpatient clinic visits and telephone interviews. The follow-up duration was extended until June 30th, 2022, and additional data related to readmitted patients were extracted from their medical records. For subjects unreachable for telephone interviews, survival status was determined based on the most recent approved timepoint, including their latest outpatient clinic visit or the last day of any hospital admission.

The patients were divided into three categories based on the tertiles of the TyG index. Mean values with standard deviation or median values with interquartile range (25th–75th percentile) were used for continuous variables and analyzed using the ANOVA test or the Kruskal–Wallis H test, as appropriate. Categorical variables were summarized as frequency and percentages and analyzed using the Chi-square test or Fisher exact test where applicable.

Univariable and multivariable logistic regression models were employed to explore the potential association of the TyG index (modeled as continuous variables and tertiles) with the presence of multivessel disease and high GS. In Model 1, no adjustments were made, while in Model 2, we incorporated age and gender as covariates. Model 3 was a comprehensive adjustment model. The candidate variables included age, sex, systolic blood pressure, diastolic blood pressure, heart rate, hypertension, diabetes mellitus, current smoker, atrial fibrillation, cerebrovascular disease, valvular disease, peripheral arterial disease, dialysis modality, dialysis vintage, cause of dialysis, insulin therapy, index presentation, hemoglobin, TC, LDL-C, and HDL-C. Confounders that were significant in the univariate model, or of clinical importance, were included in Model 3. The results of the logistic regression model were reported as the odds ratio (OR) accompanied by a corresponding 95% confidence interval (CI). Multiple imputation was employed to estimate the missing values. Cumulative survival curves were performed using Kaplan–Meier method and compared with the log-rank test. The associations between the TyG index and all-cause death and cardiovascular death were assessed using univariable and multivariable Cox proportional hazard models, with hazard ratios (HRs) and corresponding 95% CIs reported. Proportional hazards assumptions were validated using Schoenfeld residuals. The candidate variables are listed in Table 1 (except for serum creatinine), with statistically significant or clinically relevant confounders included in the multivariable Cox analysis. To avoid collinearity and potential interactions, TG and glucose, which are components of the TyG index, were not included in the multivariable regression model. The variables included in the multivariable models were assessed for multicollinearity through the examination of variance inflation factor (VIF) values, and no indication of collinearity was discovered, as the VIF values were all below 5. Furthermore, we conducted restricted cubic spline analyses to detect potential nonlinear relationships between the TyG index and the extent and severity of CAD, as well as follow-up outcomes. The restricted cubic spline model was also adjusted for confounding factors that were included in multivariable logistic and Cox regression model. Subgroup analysis considered age, gender, diabetes mellitus, smoking, dialysis modality, index presentation, PCI treatment, high GS, and multivessel. The incremental predictive performance for follow-up outcomes after introducing the TyG index to the baseline risk model with fully adjusted variables was evaluated by calculating the C statistic, continuous net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Several sensitivity analyses were performed. First, we performed a complete-case analysis to assess the association of the TyG index with coronary severity and mortality by excluding all individuals with missing values. Second, we excluded patients who received insulin therapy and repeated the analyses. Considering the important role of microinflammation in the atherosclerotic processes of ESRD patients on dialysis [32, 33], we conducted an exploratory analysis in the subgroup of 631 patients with available C-reactive protein (CRP) data. We aimed to assess whether CRP, as a marker of microinflammation, could potentially elucidate the link between the TyG index and adverse cardiovascular events. Spearman correlation analysis was conducted to examine the relationship between the TyG index and CRP levels. Subsequently, we adjusted our multivariable models for coronary severity and mortality by including CRP as a covariate. For all statistical analyses, two-sided P values were used, with significance considered at a value of < 0.05. We utilized SPSS 23.0 (IBM SPSS 23 Inc) and R 3.6.1 (R Development Core Team, Vienna, Austria) for data analysis.

The current study recruited a total of 1061 patients (Fig. 1), with a mean age of 61.8 ± 10.5 years, of whom 74.4% were male. The entire cohort exhibited a median TyG index value of 9.1 (8.5–9.5). There were a total of 24 patients with missing baseline data, which accounted for 0.7% of hemoglobin (n = 7), 0.6% of total cholesterol (n = 6), 1.5% of high-density lipoprotein cholesterol (n = 16), and 1.4% of low-density lipoprotein cholesterol (n = 15). These variables were similar before and after multiple imputation (Additional file 1: Table S1). The baseline characteristics were stratified based on the TyG index tertiles (Table 1). A higher prevalence of hypertension, diabetes mellitus, diabetes as the cause of dialysis, multivessel disease, insulin therapy, and beta-blocker therapy was observed among patients in the third TyG index tertile compared to those in the first tertile. In addition, patients with a higher TyG index exhibited a higher heart rate, TC, LDL-C, and high GS, along with lower diastolic blood pressure and HDL-C.

Patients with multivessel disease exhibited significantly higher TyG index values compared to those with single-vessel disease. (9.1 [8.6–9.6] vs 8.8 [8.3–9.3], P < 0.001) (Fig. 2A).In addition, as the tertile of GS score increased, the TyG index was found to gradually increase (tertile 1 [GS ≤ 38]: 9.0 [8.4, 9.5]; tertile 2 [GS 38–73]: 9.0 [8.5, 9.5]; tertile 3 [GS > 73]: 9.1 [8.7, 9.7]; P = 0.001) (Fig. 2B). Additional information on baseline characteristics, categorized according to the number of narrowed coronary arteries and GS, was provided in Additional file 1: Tables S2 and S3. To investigate the association between the TyG index and as high GS well as multivessel disease, logistic regression analysis was performed in this study. The univariate regression analysis (Table 2, Model 1) revealed a positive association between the TyG index as a continuous variable and high GS (OR 1.40; 95% CI 1.18–1.67). After adjustment for age and sex, the same association was observed. A subsequent full adjustment for the baseline clinical risk factors also revealed a significant elevation in the risk of high GS related to an increasing TyG index (adjusted OR, 1.33; 95% CI 1.10–1.61; P = 0.003) (Table 2, Model 3). Investing the tertiles of the TyG index revealed that the third tertile group had a 1.51-fold greater risk of developing high GS in the multivariate logistic regression model as compared to the reference (first) tertile group (adjusted OR 1.51; 95% CI 1.07–2.12; P = 0.018). In addition to high GS, the TyG index was also found to be associated with multivessel according to the univariate regression analysis (OR 1.59; 95% CI 1.27–1.99). The strength of this association remained significant after accounting for multiple possible confounders (adjusted OR 1.51; 95% CI 1.18–1.94; P = 0.001) (Table 2, Model 3). When comparing the highest tertile of the TyG index to the reference tertile in a fully adjusted model, a significant association with an elevated risk of multivessel was revealed (adjusted OR 1.74; 95% CI 1.09–2.77; P = 0.021). In addition, the findings of the restricted cubic splines are depicted in Fig. 3. The results revealed a discernible dose–response connection between the TyG index and the likelihood of high GS as well as multivessel disease, following thorough adjustment (both tests for nonlinearity revealed P > 0.05).

The median follow-up duration was 21.9 months (12.6–35.6 months). A total of 358 (33.7%) cases of all-cause mortality were observed, with 248 attributed to cardiovascular-related causes. Additional information on baseline characteristics between non-survivors and survivors was provided in Additional file 1: Table S4. The incidence of all-cause death was found to gradually increase as the tertile of TyG index increased: 27.2% (96/353), 34.2 (121/354), and 39.8% (141/354), respectively. Furthermore, the occurrence of cardiovascular death among those with first TyG index tertile, second TyG index tertile, and third TyG index tertile was noted as 17.8% (63/353), 23.7% (84/354), and 28.5% (101/354), respectively. Kaplan–Meier curves exhibited a significantly higher risk of both all-cause death and cardiovascular death in patients with the third TyG tertile group compared to other groups (log-rank test P = 0.002 and 0.003, respectively) (Fig. 4). The association between TyG index and outcomes is summarized in Table 3. When modeling TyG index as a continuous variable, a significantly increased risk of both all-cause death (adjusted HR 1.23, 95% CI 1.06–1.43, P = 0.007) and cardiovascular death (adjusted HR 1.33, 95% CI 1.11–1.59, P = 0.002) was observed in patients with high TyG index after adjustment for potential clinical risk factors (including Gensini score and multivessel). When comparing the third TyG index tertile to the first TyG index tertile, a significantly increased risk of all-cause death and cardiovascular death was identified in the fully adjusted model (adjusted HR 1.43; 95% CI 1.08–1.89; P = 0.012; adjusted HR 1.68; 95% CI 1.20–2.36; P = 0.002, respectively). Furthermore, restricted cubic splines analysis identified a discernible dose–response connection between the TyG index and the risk of all-cause death (P for nonlinearity = 0.225; Fig. 3C) as well as cardiovascular death (P for nonlinearity = 0.622; Fig. 3D). Clinically relevant variables, including age, gender, diabetes mellitus, smoking, dialysis modality, index presentation, PCI treatment, high GS, and multivessel, were used for subgroup analyses (with the TyG index modeled as a continuous variable). Despite variations in effect size and statistical significance, no significant interactions were identified regarding the risk of all-cause death in the selected subgroups (all P values for interaction > 0.05) (Fig. 5). Similarly, these clinically relevant variables showed no significant interaction with the TyG index regarding the risk of cardiovascular death (all P values for interaction > 0.05) (Fig. 6). Furthermore, the risk prediction for all-cause death was increased by adding the TyG index to the baseline risk model with fully adjusted variables (Table 4), with the C-statistic increasing from 0.698 to 0.700 (P = 0.007). Similarly, adding the TyG index into the baseline risk model significantly improved the predictive performance for cardiovascular death, with the C-statistic increasing from 0.725 to 0.729 (P = 0.005). Moreover, the NRI values for all-cause death and cardiovascular death were 0.095 (95% CI 0.009–0.170; P < 0.001) and 0.091 (95% CI 0.004–0.178; P < 0.001), respectively. Similarly, improvements were also observed using the IDI metric, with values of 0.006 (95% CI 0.001–0.021; P < 0.001) for all-cause death, and 0.012 (95% CI 0.001–0.029; P < 0.001) for cardiovascular death.

Several sensitivity analyses were performed. After excluding 24 patients with missing data, we performed a complete-case analysis and identified a significant association of the TyG index with coronary severity and mortality (Additional file 1: Table S5). Furthermore, excluding 355 patients who received insulin therapy did not materially change our findings (Additional file 1: Table S6).

Given the important role of CRP-related microinflammation in the atherosclerotic processes of ESRD patients on dialysis, we conducted an exploratory analysis in the subgroup of 631 (59.5%) patients with available CRP data. The median CRP value was 9.3 mg/L (interquartile range: 3.4–25.0 mg/L). While some of the baseline variables showed significant differences between patients with and without CRP data, most of them were similar between the two groups (Additional file 1: Table S7). The Spearman correlation coefficient between the TyG index and CRP was 0.147 (P < 0.001). Scatter plots illustrated a positive correlation between the TyG index and CRP. After adjusting for CRP in multivariable logistic regression model (Additional file 1: Table S8), we found that the TyG index remained significantly associated with an increased risk of high Gensini score and multivessel disease, independent of CRP. These findings were consistent when examining the follow-up outcomes. In the multivariable Cox regression model (Additional file 1: Table S8), we identified a significant association between the TyG index and all-cause mortality and cardiovascular mortality, independent of CRP.