Introduction

As the most prevalent cardiac rhythm disorder, atrial fibrillation (AF) dramatically interferes with patients’ quality of life and is associated with a 10–25% increase in all-cause mortality and more than 8 million disability-adjusted life years [1,2,3]. AF affects approximately 60 million individuals worldwide, with an incidence that escalates significantly with age as a reflection of comorbidities, cardiovascular risk factors, and metabolic disorders, such as diabetes [1, 4,5,6]. In this regard, multiple studies have assessed the association between AF and diabetes, with diabetic patients showing an approximate 35% increased risk of developing AF in comparison to nondiabetic ones [7,8,9]. In line with this data, different reports supported insulin resistance (IR), the main feature of diabetes, as the main factor strictly intertwined with AF pathogenesis [10, 11]. However, aside from diabetes, IR represents a metabolic substrate associated with several cardiovascular disorders, obesity, and inflammation, all risk factors for AF development [12,13,14,15]. Indeed, in nondiabetic individuals, studies have provided evidence of a positive correlation between IR and AF risk [16]. For this reason, several tools, like the homeostasis model assessment index for insulin resistance (HOMA-IR), the gold standard for assessing IR, have been used to examine this relationship in the community [17]. Analogously, different studies used a simpler and more low-cost estimating tool called the triglyceride–glucose (TyG) index, which has a power of convenience compared to the HOMA-IR [18]. Moreover, this index has been investigated in several cardiovascular conditions and diseases [18,19,20,21].

The TyG index calculation could be performed using the routinely measured parameters, and its application in clinical practice has been widely proposed as it has an enormous diagnostic and prognostic power to assess many IR-related disorders. Based on this premise, we conducted a systematic review and meta-analysis to analyze the data available, establishing the association between the TyG index and AF.

Methods

This systematic review and meta-analysis was conducted in accordance with the “preferred reporting items for systematic reviews and meta-analyses” (PRISMA) guidelines and was registered in the PROSPERO registry by registration number (CRD42023456875) [22].

Systematic search strategy

Articles included in this review were identified via comprehensive systematic searches of electronic databases of PubMed, Embase, Scopus, and Web of Science from inception to September 2023. MeSH terms and keywords, including “Atrial Fibrillation” OR “Auricular Fibrillation” OR “Familial Atrial Fibrillation” AND “triglyceride-glucose index” OR “TyG” were combined to elicit original studies on TyG index and AF association. The full search strings applied in each database are included in Additional file 1: Table S1. Two reviewers (SD, AT), using EndNote 9 software (Tomson Reuters, New York, USA), independently evaluated each article, and they also reviewed the full text and eliminated any duplicates. The inclusion–exclusion criteria were followed for selecting studies. The third author (AA) served as the moderator of consensus sessions to resolve any disagreements that might arise between reviewers.

Inclusion and exclusion criteria

Inclusion criteria

The inclusion criteria of the original studies were: (1) studies with an observational study design; (2) studies investigating the AF and the TyG index correlation, in addition to those assessing the effect of the TyG index on incident AF; (3) individuals older than 18 years of age of any ethnicity and both sexes and (4) studies in which electrocardiogram (ECG) findings confirmed AF diagnosis. The TyG index is calculated from fasting plasma glucose (FPG) and triglyceride (TG):

$${\text{TyG}} = \ln \left( {{\text{TG}}\left( {\frac{{{\text{mg}}}}{{{\text{dL}}}}} \right) \times \frac{{{\text{FPG}}\left( {\frac{{{\text{mg}}}}{{{\text{dL}}}}} \right)}}{2}} \right).$$

Exclusion criteria

The exclusion criteria were: (1) preclinical studies (in vivo in animals or in vitro in cells); (2) interventional studies, book chapters, reviews, and case reports, (3) studies with incomplete data or an unclear method of the TyG index calculation, and (4) conference abstracts and preprints.

Data extraction and quality assessment

Two authors (SD, AT), after thoroughly conducting full-text screening, independently input the following information into a pre-piloted, standardized Excel spreadsheet: author and publication year, nationality, design of study, definition of case and control groups, TyG index, sex ratio, age, sample size, and body mass index (BMI). We also extracted the hazard ratio (HR) with the highest number of factors for adjustment to compare the highest TyG index groups to the lowest TyG group. Moreover, when the TyG index was used in analyses as a continuous variable, we extracted the HRs reflecting the risk per one-unit increase of the TyG. We extracted the area under the curve (AUC) from studies that reported AUC for the diagnosis of AF or outcome prediction. The extraction was performed using a sheet in Microsoft Excel 2016.

For evaluation of the qualities of the included studies, the Newcastle–Ottawa Scale (NOS) was employed [23]. Selection, comparability, and outcome are the three main domains for quality assessment of cohort studies. Based on this scale, a score of 3 or 4 in the selection domain, 1 or 2 in the comparability domain, and 2 or 3 in the outcome domain is considered “good” quality. Two independent authors (SD and AT) assessed the qualities, and in cases of disagreement, a third author (AA) resolved the issue.

Statistical analysis

R version 4.3.0 (R Core Team [2020]. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria) was used to conduct statistical analysis, and the packages for analyses were “meta” and “metafor”.

In cases of reporting median and interquartile range (IQR) or median and range for any of the variables needed, we calculated mean and standard deviation through the methods suggested by Luo et al. and Wan et al. [24, 25]. To evaluate the relationship between the TyG and the AF incidence, we employed the Hedges’ g standardized mean differences (SMD) with their matching 95% confidence intervals (CIs) by random-effect meta-analysis, as a general indicator [26]. Since the TyG index has a logarithmic scale without any unit, we used mean difference (MD) for the meta-analysis as a sensitivity analysis. The statistical heterogeneity was assessed using the Q-test and I2. It was determined that I2 values of 25, 50, and 75%, respectively, indicated minimal, moderate, and high heterogeneity. To assess the publication bias, Egger's test was conducted [27]. Meta-regression based on publication year, male percentage, mean age, and sample size was performed to assess their effect on overall heterogeneity. Subgroup analysis was also performed for different populations, when possible. A two-sided P less than 0.05 was used to indicate statistical significance.

Results

Study selection and baseline characteristics

We identified two hundred and three records from an initial screening of the four databases analyzed and described in the methods. As reported in Fig. 1, 80 studies were immediately removed because they were duplicated. Then, a further 89 studies were excluded by title/abstract screening, and another 25 were excluded during the full-text assessment for reasons mentioned in Fig. 1. At the end of the screening, nine studies fully met the inclusion criteria and were included in the analysis [28,29,30,31,32,33,34,35,36]. Studies were conducted mostly in China [28, 29, 32,33,34,35,36], followed by the United States [30] and Sweden [31]. The design of the included studies were prospective cohort [30, 31, 33, 34], retrospective cohort [28, 29], case–control [35, 36], and cross-sectional [32]. In these studies, 886 patients with AF were compared against 5285 without AF, with a mean age of 63.46 ± 10.47 years and 56.81 ± 10.54 years, respectively. The male percentage was also 65.46% in the AF population and 58.47% in non-AF cases. Table 1 summarizes each study’s characteristics, including patient groups, population, country, design, sample size, mean age and BMI, sex/gender percentage (male %), left ventricular ejection fraction (LVEF) percentage, TyG index, and main findings. The studies analyzed assessed TyG levels with new-onset AF in population-based cohorts, type 2 diabetes mellitus (T2DM), or CAD patients. Further, some of these studies evaluated the TyG index in AF patients' adverse events. As illustrated in Additional file 1: Table S2, the included studies had NOS 7–8 and “good” quality based on the criteria.

Fig. 1
figure 1

PRISMA flowchart of the study selection process

Full size image
Table 1 Main characteristics and main findings of the included studies
Full size table

Meta-analysis of TyG levels in patients with AF

Meta-analysis comparing the mean TyG index in patients with and without AF. As shown in Fig. 2, the forest plot, demonstrated a significantly higher TyG index in total AF patients (SMD 1.23, 95% CI 0.71 to 1.75, P < 0.001). This analysis was associated with a high heterogeneity (I2: 98%, 95% CI 97.1% to 98.7%). Next, we conducted a subgroup analysis, and as shown in Fig. 2, no significant difference was observed comparing patients with pure AF vs. non-AF controls (SMD 1.45, 95% CI − 0.41 to 3.31, P = 0.13). Conversely, as shown in Fig. 2, a higher TyG index was observed in the post-ablation population (SMD 1.25, 95% CI 1.07 to 1.43, P < 0.001) that showed late AF recurrence compared to those that did not, and in post-procedural patients with AF (SMD 0.99, 95% CI 0.78 to 1.20, P < 0.001) compared to the controls undergoing the same procedures but without AF [septal myectomy or percutaneous coronary intervention (PCI)] (Fig. 2 and Table 1). Sensitivity analysis by leave-one-out method was performed to assess each study’s effect on the overall effect size. The removal of none of the studies led to no change in overall effect size in terms of significance (Additional file 1: Figure S1). Publication bias was assessed by visual inspection of the funnel plots by trim-and-fill method (Additional file 1: Figure S2). No asymmetry was observed in the funnel plot. Similarly, Egger’s test showed no publication bias (P = 0.634).

Fig. 2
figure 2

Forest plot and subgroup analysis of the association of mean TyG and AF

Full size image

Meta-regression assessing the effect of each variable on overall effect size is summarized in Table 2. Except for publication year which had a significant association with the effect size (P = 0.028), the other variables (mean age, sample size, and male %) had no association. Publication year accounted for 44.40% of the heterogeneity observed. The bubble plots for these meta-regressions are shown in Additional file 1: Figures S3–6. While, in the multivariable meta-regression by these variables, shown in Additional file 1: Table S3, sample size and publication year had significant associations with the effect size (P = 0.036 and P = 0.003), and these variables accounted for 89.01% of the heterogeneity.

Table 2 Univariate meta-regression for meta-analysis of TyG index in patients with AF vs. healthy controls
Full size table

Finally, in the sensitivity analysis by using MD instead of SMD, the same significant result of higher TyG index in patients with AF was obtained (MD 0.72, 95% CI 0.49 to 0.94, P < 0.001). This analysis had a high heterogeneity (I2: 96.6%, 95% CI 94.6% to 97.9%). The forest plot for this analysis is shown in Additional file 1: Figure S7.

Association between the TyG and risk of AF

Nine studies investigated the association between the TyG and the incidence of outcomes [28,29,30,31,32,33,34,35,36]. Table 3 represents the differences in outcomes between groups with low or high TyG index, quartiles (Q), and tertiles (T) of the TyG, and the TyG as a continuous variable. Generally, AF patients presented with a greater TyG index in most cases and across all populations.

Table 3 Risk of AF in different groups/levels of the TyG index
Full size table

Population‑based cohorts

Different studies have evaluated the association between TyG and AF risk in the general population. Liu et al. [30], using data from the Atherosclerosis Risk in Communities (ARIC) study, a U-shaped association was found between the TyG index and the incident AF in the cardiovascular disease-free general population in a 24.2-year follow-up period. In their study, the authors analyzed a population of 11,851 nondiabetic participants who were assigned to groups based on the TyG value tertiles at the baseline (T1: TyG < 8.80; T2: 8.80 ≤ TyG ≤ 9.20; and T3: TyG > 9.20), the analysis revealed that T1 and T3 presented with an incidence rate of 0.69 and 1.13 (per 100 person-years), respectively, and showed an increased risk of AF compared to T2, with an incidence rate of 0.83 (P = 0.02). In line with this report, Chen et al. [28] observed that the TyG index was positively associated with AF (OR 2.092, 95% CI 1.412 to 3.100, P < 0.001). In particular, the TyG index was higher in nondiabetic patients with AF compared to individuals without AF (OR 3.065, 95% CI 1.819 to 5.166, P < 0.001) while no differences were observed in the TyG index between diabetic subjects with or without AF (OR 1.286, 95% CI 0.645 to 2.565, P = 0.475). On the other hand, Shi and colleagues [32] demonstrated that the frequency of AF in the diabetes cohort was 6.57%. Every one-SD increase in the TyG value increased the prevalent AF risk by 40.6% in diabetic patients (OR 1.406, 95% CI 1.197 to 1.650, P < 0.001). Of note, the higher quartile (Q4) of the TyG index was associated with 2.120 (95% CI 1.37 to 3.29, P = 0.002) times greater prevalence odds of AF than Q1.

For their part, Zhang et al. [36] in a population of 912 patients with NAFLD, 204 of them with AF, observed that those with the highest Q of the TyG index had a greater risk of AF incidence (33.3%) compared to those in the lowest Q with 16.7% having AF (OR 4.34, 95% CI 2.37 to 7.94, P < 0.001). After adjusting for age, gender, BMI, diabetes, serum creatinine, and total cholesterol), an increase of one unit in the TyG index resulted in a rise in the risk of AF by 4.84 (OR 4.84, 95% CI 2.98 to 7.88, P < 0.001). In contrast with these data, Muhammad and coworkers [31] in a general population-based cohort of 32,917 participants (6950 presented with AF) from Sweden and with a follow-up duration of 16.9 years (mean), failed to observe any difference between subjects in the highest Q (Q4: 4.74–6.70) of the TyG value compared to individuals in the lowest Q (Q1: 3.38–4.38) [HR 0.96, 95% CI 0.89 to 1.04)] (P = 0.14). Additionally, the HR per one-unit increase in the TyG value was the same for cases and controls [HR 0.99, 95% CI 0.89 to 1.11)].

Post-procedural AF

Next, we analyzed studies showing the utility of TyG as a post-procedural AF predictor. In this regard, in the study by Ling et al. [29], the authors observed that the patients who underwent PCI and had the higher TyG index had an incidence of 17.8% new-onset AF, while the low TyG group had 3.1% new-onset AF (P < 0.001). For their part, Wei et al. [34], according to the cut-off point of 7.60, identified by ROC curve analysis, divided a population of 409 patients (348 had no postoperative (septal myectomy) AF (POAF), and 61 were diagnosed with POAF) into two groups based on the TyG value: low (mean: 6.80 ± 0.44) and high (mean: 7.98 ± 0.37) and found that the patients in the high TyG group had an increased incidence of POAF than those in the low group (45.0% vs. 9.7%, P < 0.001).

Post-ablation AF recurrence

We also analyzed the data concerning the association between TyG and post-ablation AF recurrence. In this context, in their study, Tang and colleagues [33] examined a total of 275 nondiabetic participants who underwent first-time radiofrequency catheter ablation (RFCA) for AF stratified into three T based on pre-ablation TyG (T1: < 8.67, T2: 8.68–9.37, and T3: ≥ 9.38). Patients in T3 had a higher rate of late AF recurrence compared to participants in the T1 group (54% vs. 12%, P < 0.001). In another study, Zhang et al. [35] reported that the HR of higher TyG to be 2.021 (95% CI 1.374 to 3.245, P < 0.001), was an effective risk factor for AF recurrence after Cox-maze IV ablation.

AUC for the prediction of AF and its outcomes

In our last analysis, we examined all the studies showing data regarding the AUC value of TyG. In their study, Chen and coworkers [33], compared the nondiabetic AF patients with those without AF and found that the TyG index displayed an AUC value of 0.600 (95% CI 0.542 to 0.659, P = 0.001), an optimal cut-point value of 8.35 with a sensitivity of 65.4%, and a specificity of 52.0%. However, when the TyG index was combined with variables like hypertension and total cholesterol, it showed a higher AUC value of 0.667 (95% CI 0.611 to 0.723, P < 0.001), with a cut-off value of 0.466, 71.5% sensitivity, and 58.1% specificity.

Next, Zhang et al. [35] promoted using AUC of TyG to predict AF recurrence after the Cox-maze IV procedure. These authors found an AUC of 0. (95% CI 0.796 to 0.871, P < 0.001), supporting the high predictive value for AF recurrence of the TyG in this population of patients. Notably, the cut-off value of the TyG index reported in this study was 8.86, and the sensitivity and specificity were 88.6% and 44.7%, respectively.

In line with this notion, Ling et al. [29] demonstrated that an AUC value of the TyG index of 0.758 (95% CI 0.720 to 0.793, P < 0.001) has a predictive value for new-onset AF (NOAF) incidence in ST-segment elevation myocardial infarction (STEMI) patients following PCI.

For their part, Shi et al. [32] showed that the AUC of the TyG index alone for identifying prevalent AF in a diabetic population was 0.631 (95% CI 0.614 to 0.648, P < 0.001). In addition, the authors demonstrated that when TyG was added to conventional cardiovascular risk factors, the AUC improved for the detection of prevalent AF (0.825 vs. 0.812, P = 0.02), an effect also supported by the continuous net reclassification index (0.227, 95% CI 0.088 to 0.365, P = 0.001) and integrated discrimination index (0.007, 95% CI 0.001 to 0.012, P = 0.03).

Further, Wei et al. [34] found a moderate predictive value for the TyG index for postoperative AF in hypertrophic obstructive cardiomyopathy patients who underwent septal myectomy, showing an AUC of 0.723 (95% CI 0.650 to 0.796, P < 0.001). However, after adding TyG to the model based on conventional risk factors, it only numerically, but not significantly increased the prediction ability of postoperative AF, with an AUC of 0.742 (95% CI 0.671 to 0.814) compared to 0.793 (95% CI 0.726 to 0.860) for the conventional risk factor model alone (P = 0.065). The optimal cut-off point for the TyG index was found to be 7.60. At this point, the sensitivity of the TyG index was 44.3%, and the specificity was 90.5%.

Finally, Zhang et al. [36], in contrast with the results shown above, revealed that the AUC for the sole TyG index was 0.615, suggesting a weak predictive ability for AF incidence in NAFLD patients. However, when they combined TyG with traditional risk factors, the predictive value for AF was significantly improved, as evidenced by an AUC of 0.857 (P = 0.001). The sensitivity and specificity of the TyG model were 0.68 and 0.72, respectively. The TyG model had a higher predictive power for AF than traditional risk factors alone.

Discussion

In this study, a systematic review and meta-analysis of 50,921 individuals in nine studies were performed (Table 1) to explore the association between TyG and the incidence of AF. We performed also a subgroup analysis: (1) post-ablation, (2) post-procedure, and (3) AF versus non-AF populations. The highlighted findings of the current study are: (1) AF incidence is higher in the group with high TyG levels compared to the lower TyG group; (2) the TyG index is higher in AF patients, compared to the normal population. AF is associated with higher morbidity and mortality rates due to related complications such as heart failure, cardiomyopathies, and thromboembolic events [37]. Despite various management options for AF, the prevalence of AF is rising due to new diagnostic methods. Therefore, recognition of AF risk factors and predictors may help clinicians in order to identify the patients that have a higher risk of AF, helping early detection of AF. In this context, several are the biomarkers identified and tested for risk stratification of AF, including the C-reactive protein (CRP) [38], the fibroblast growth factor-23 (FGF-23) [39], the high sensitivity troponin I [40], Galectin-3 [41], the N-terminal pro-B-type natriuretic peptide (NT-proBNP) [38, 39, 42,43,44], and several micro-RNAs [45]. However, other biomarkers associated with baseline risk factors, such as diabetes [46] and IR [47], have also been considered due to their bilateral diagnostic and prognostic usage.

Among the tools to measure IR in clinical practice, the TyG index has been established to be one of the easier-to-measure and cost‐effective parameters with a diagnostic and prognostic value comparable to other IR markers. In this regard, previous studies have correlated this surrogate marker of IR with several disorders, including COVID-19 [48], cerebrovascular disease or ischemic stroke [21, 49], hypertension [50], metabolic dysfunction-associated fatty liver disease [51], NAFLD [52], heart failure [18], acute myocardial infarction [53], and diabetes. In addition, our meta-analysis demonstrated that in the total AF population, the TyG index has a higher predictive value (SMD 1.23, 95% CI 0.71 to 1.75). Moreover, as shown in our subgroup analysis, there is no statistical difference between AF and non-AF groups (SMD 1.45, 95% CI -0.41 to 3.31). Conversely, in post-ablation (SMD 1.25, 95% CI 1.07 to 1.43) and post-procedure patients (SMD 0.99, 95% CI 0.78 to 1.20), AF was associated with a higher TyG index than their control groups.

The clinical utility of our findings includes considering TyG as an easy-to-use marker of IR in patients with AF. Considering the limitations of other common markers of IR, such as HOMA-IR, such as complexity of measurement or higher cost of measurement, TyG can be a valuable and reliable biomarker of IR [54]. Even in some studies, the TyG index outperformed HOMA-IR in evaluating IR, which makes it a potentially more useful index in the AF population as well [55, 56]. Our findings suggest an association between the TyG index and AF, however, the prognostic impact of TyG on outcomes of AF has not been investigated in these included studies. Meanwhile, the presence of DM and glycemic dysregulation in patients with AF has been shown to increase the risk of adverse outcomes such as cardiovascular mortality, sudden cardiac death, and stroke [57]. All of this highlights the need for further studies aiming at assessing this role.

Strengths and limitations

This study was the first to assess the role of the TyG index in AF with a systematic review and meta-analysis. By pooling the data from these analyses, we found a higher TyG index in the subjects with AF than those without AF. Assessment of several outcomes and providing an insight into diagnostic ability are among the strengths of this study. This study will pave the way for future studies to specifically assess IR in AF pathology and prognosis through measurement of the TyG index. However, there are five main limitations to be disclosed. First, a high heterogeneity was observed in the analysis. Although performing subgroup analysis reduced the overall heterogeneity, these differences might stem from those in clinical settings and designs in each study. So, our study could be interpreted as a demonstrator of a possible relationship between the TyG index and AF, emphasizing the need for further larger studies. The second limitation is the number of included studies in the analyses which were low and limited our findings as well as some of the methods used such as publication bias assessment by funnel plots, Egger’s test, and meta-regression which is mainly performed in cases of higher than ten studies [58]. Third, different AF definitions could be a source of bias in our study, in addition to different follow-up durations for assessment of incident AF could be another limiting factor of our study. Fourth, different cut-off values across the included studies for the TyG index, may result in differences in categorizing individuals into the low or high TyG value groups. Fifth, the observational nature of the included studies prevents us from drawing any causal relationship conclusion. Hence, further larger studies investigating the association between the TyG index, and this high-risk population are warranted.

Conclusion

Based on this study’s findings, the TyG index can be used as an IR marker for AF. Clinicians can take advantage of this index in patients with AF as well as its recurrence after ablation and other high-risk procedures. Further studies and association with other factors primarily influencing IR lifestyle modifications, pharmacological interventions, and other confounding variables (e.g., comorbidities) are needed to confirm the findings of the current study.