Comprehensive analysis of the association between triglyceride-glucose index and coronary artery disease severity across different glucose metabolism states: a large-scale cross-sectional study from an Asian cohort – Cardiovascular Diabetology

In our study, we conducted a comprehensive investigation into the relationship between the TyG index and CAD severity across different glucose metabolism states. Four main findings emerged from our investigation. Firstly, we addressed the intense controversy among previous relatively small-sample studies in this field. In the largest sample examined to date, we confirmed the independent association between the TyG index and multi-vessel CAD, regardless of patients’ glucose metabolism state (NGR, pre-DM, or DM), with the only exception when insulin therapy was administered. Secondly, a linear relationship between the TyG index and multi-vessel CAD was observed in the NGR and pre-DM populations, while a nonlinear relationship and threshold saturation effect were noted in the DM non-insulin Rx population. Thirdly, HbA1c exhibited varied mediating roles in the association between the TyG index and multi-vessel CAD, with its mediation becoming more prominent as glucose metabolism abnormalities worsened. Fourthly, the TyG index can provide moderate incremental predictive value for multi-vessel CAD beyond established risk factors, with the most notable effect observed in the NGR population.

Multi-vessel CAD is strongly associated with a poor prognosis in CAD and can increase the complexity of PCI procedures as well as subsequent medication therapy [3]. The exploration of its pathogenesis and predictive indicators has long been a focal point of research. The TyG index, representing IR, is strongly linked to the onset and progression of atherosclerotic cardiovascular diseases, particularly CAD [21]. The investigation into the relationship between TyG and multi-vessel CAD is also receiving increasing attention. Wu et al. found that a higher TyG index increased the risk of arterial stiffness [22]. Thai et al. first reported that an elevated TyG index identified patients at high risk of coronary artery stenoses, correlating with both the number and severity of stenoses [23]. Furthermore, Su et al., Zhang et al., Wang et al., and Xie et al. have successively validated the association between the TyG index and multi-vessel CAD in larger cohorts [16,17,18, 24]. However, two major controversies remain in this field: First, studies have differing conclusions on the relationship between the TyG index and multi-vessel CAD across glucose metabolism subgroups. For instance, Su et al. found this association in the DM group but not in the pre-DM or NGR groups, while Wang et al. found it in the pre-DM group but not in the DM or NGR groups. Second, contradicting the conclusions of the aforementioned studies, a recent study found no association between the TyG index and multi-vessel CAD in new-onset CAD patients, regardless of glucose metabolism state [15]. The authors attribute this conflicting result to the selection of enrolled patients and Neyman bias in previous studies, but their study also suffers from a small sample size. These controversies confound our understanding of the relationship between the TyG index and multi-vessel CAD, necessitating larger and more comprehensive studies to resolve these issues.

Compared to prior studies, our research offers several advantages: Firstly, it boasts the largest sample size, enhancing its persuasive capacity. Secondly, it includes a most extensive array of adjustment factors, encompassing various risk factors, laboratory parameters, and medication utilization. This encompasses novel metrics such as Lp(a), previously overlooked in research. Thirdly, considering the significant impact of insulin on IR, it distinguishes between insulin and common antidiabetic medications, and analyzes them separately by grouping. Fourthly, it performs comprehensive sensitivity analyses in cohorts with potential controversies, including individuals with newly diagnosed CAD and those not prescribed antilipidemic drugs. Drawing from these aforementioned advantages, our study elucidates the association between the TyG index and multi-vessel CAD across the population and within different glucose metabolism subgroups. This finding aligns with the main conclusions drawn from the majority of prior research and meta-analyses [13, 16,17,18, 23, 25]. Contradictory outcomes noted in certain previous studies across different glucose metabolism subgroups could stem from their relatively limited sample sizes or inadequate adjustment for insulin [16, 18]. Moreover, sensitivity analysis results bolster the resilience of findings in the new-onset CAD cohort, indicating that discrepant conclusions in a recent study could be attributed to its smaller sample size (6669 cases vs. 431 cases) [15].

Importantly, given the intricate interplay among IR, glucose metabolism status, and blood glucose levels, we observed significant variations in the dose-response relationship between the TyG index and CAD severity across different glucose metabolism subgroups. Figure 5 shows that the association between the TyG index and CAD severity is significantly stronger in the NGR group, especially in the 4th quartile, compared to other groups. This association largely explains the overall relationship observed in the total population. This finding is plausible, since elevated glucose levels in other groups may interfere with the TyG index’s association with CAD severity. This interference is most pronounced in the DM insulin Rx group, where glucose variability may diminish this association. Furthermore, Fig. 6 indicates that the increased OR is primarily driven by subjects in the top 25th percentile of the NGR group, while the pre-DM group shows this increase only in the top 50th percentile. Conversely, in the DM non-insulin Rx group, the increase is confined to the bottom 50th percentile. Notably, in the DM insulin Rx group, the TyG index demonstrates a nearly U-shaped relationship. When aggregating all subjects, the data suggest a somewhat linear relationship, which is an artifact of differing behaviors across the various groups. Additionally, the confidence intervals for these curves are relatively wide, indicating variability in the data through different subgroups. Given the dose-response differences in TyG index and CAD severity across different glucose metabolism subgroups, the interpretation of the TyG index’s relationship with CAD severity must consider glucose metabolism states [16, 18].

In addition to addressing major controversies in this field, another significant finding of our study is the notable interaction effect of insulin treatment. Previously, it was believed that the TyG index was developed as a reliable biochemical surrogate for identifying IR in both diabetic and non-diabetic individuals [12]. Unlike other IR indices such as the homeostasis model assessment of insulin resistance (HOMA-IR), quantitative insulin sensitivity check index (QUICKI), and homeostasis model assessment of β-cell function (HOMA-β), the TyG index does not require insulin quantification [26]. Therefore, it has traditionally been considered less influenced by insulin treatment [12, 26, 27]. Our study found that the TyG index levels progressively increased across the NGR, pre-DM, DM non-insulin Rx, and DM insulin Rx groups. From the perspective of the components of the TyG index, blood glucose and TG by definition increase with worsening blood glucose control. This is also consistent with the understanding that greater glucose metabolism abnormalities correspond to more severe IR. However, within the DM insulin Rx group, the positive correlation between TyG and CAD severity was not established, and this difference exhibited a significant interaction effect. This suggests that the TyG index may not be entirely unaffected by exogenous insulin. We hypothesize that the interaction effect related to exogenous insulin may occur through the following pathways: First, high blood glucose can induce insulin resistance by acting on post-receptor signaling pathways, and moderate exogenous insulin supplementation can effectively lower blood glucose levels, thereby reducing the fasting glucose component of the TyG index and improving insulin sensitivity [28]. Second, moderate exogenous insulin can promote the conversion of fatty acids into triglycerides in adipose tissue, reducing free fatty acids in the blood and subsequently lowering plasma triglyceride levels [29]. Third, persistent hyperinsulinemia itself can induce central and peripheral insulin resistance; long-term or inappropriate use of high doses of insulin may lead to weight gain and fat accumulation, which can, in turn, increase IR and elevate the TyG index [30, 31]. Fourth, excessive and inappropriate use of insulin, which may result in hypoglycemia, can further induce insulin resistance by stimulating the secretion of counter-regulatory hormones, mainly glucagon, along with catecholamines, cortisol, and growth hormone [32, 33]. Therefore, the impact of insulin therapy on the TyG index and IR depends on the overall net effect, which is associated with factors such as disease duration, degree of hyperglycemia, and duration of treatment. This dual role of insulin may explain the observed interaction effect, at least regarding the correlation between the TyG index and multi-vessel CAD. Future studies may also need to prudently take into account the potential influence of exogenous insulin on the TyG index.

Furthermore, as the precise mechanism underlying the association between the TyG index and CAD remains incompletely understood, we conducted exploratory analyses through mediation effects in different glucose metabolism groups. The TyG index is currently recognized as a reliable indicator of IR, which could potentially account for this correlation [26, 34]. IR can induce glucose metabolism imbalance, contributing to hyperglycemia, which in turn triggers inflammation and oxidative stress. The molecular pathways linking IR and CAD encompass metabolic adaptability, endothelial dysfunction, abnormalities in coagulation, and dysfunction of smooth muscle cells [14, 31, 35,36,37,38]. A key discovery of our study is that HbA1c serves as a partial mediator in the relationship between the TyG index and multi-vessel CAD. HbA1c is created when hemoglobin from red blood cells binds with glucose. It represents the average blood glucose level over the preceding three months. Previous studies have suggested a link between increased HbA1c levels and the incidence and mortality of cardiovascular diseases [39,40,41,42]. Similarly, HbA1c is distinctly associated with CAD severity, including in both DM and non-DM patients, where higher HbA1c levels correlate with an increased risk of multi-vessel CAD [43,44,45,46,47]. Our study found that as glucose metabolism abnormalities worsen (from pre-DM to DM), the mediating effect proportion of HbA1c gradually increases (from 11.8 to 43.6%), indicating the importance of improving insulin resistance and glycemic control in both pre-DM and DM populations. Meanwhile, in the NGR population, we observed that the indirect effects of HbA1c-mediated associations between TyG and multi-vessel CAD were − 24.6%. On one hand, this suggests that HbA1c within the normal range has a protective effect on blood vessels, consistent with previous research findings: oscillating glucose can have detrimental effects on endothelial function and oxidative stress, and recurrent low blood glucose also contributes to CAD progression [48, 49]. On the other hand, this also suggests that in the NGR population, IR may primarily lead to coronary artery damage and multi-vessel CAD through the aforementioned non-glycemic effects [14, 34].

Strengths and limitations

To the best of our knowledge, this study utilized the largest sample size to date for investigating the relationship between the TyG index and CAD severity, enhancing the reliability and generalizability of the findings. Through a large sample size and extensive subgroup analysis, this study effectively addressed the inconsistencies in previous research regarding different glucose metabolism subgroups and the presence or absence of new-onset coronary artery disease. However, our study has limitations. First, this study is based on a single-center Asian cohort, so extrapolating its conclusions to other racial groups requires further validation. Second, the data were obtained from a PCI cohort, which excluded patients with extremely low CAD complexity not requiring revascularization and those with extremely high CAD complexity undergoing CABG, potentially introducing bias. Third, lifestyle factors beyond smoking, such as diet, exercise, sleep, and stress, significantly impact IR [50]; however, similar to previous studies, our research was unable to fully adjust for these influences. Nevertheless, considering that lifestyle serves as an upstream determinant of IR, the lack of adjustment for these factors may not necessarily affect the conclusions drawn in this study. Fourth, this study is cross-sectional, limiting the ability to infer causality between the TyG index and CAD severity. Fifth, in this study, the incremental predictive value of the TyG index over traditional predictive factors is relatively small. Although there is a statistical difference, its specific clinical application scenarios should be comprehensively considered.

In summary, while this study provides robust and generalizable findings due to its large sample size and extensive subgroup analysis, future research should focus on multi-center, multi-ethnic cohort studies to enhance the generalizability of results, incorporate comprehensive lifestyle factor adjustments, and employ longitudinal designs to establish causal relationships between the TyG index and CAD severity.