The correlation between triglyceride glucose index and atrial fibrillation risk: a meta-analysis Authors: Lin Han, Fei-Fei Hou, Bei-Bei Wang, Qiong Wu, Juan Liu Published: 2025 ABSTRACT AbstractBackgroundThe correlation between the triglyceride glucose (TyG) index and atrial fibrillation (AF) risk has been a subject of debate, with existing studies yielding conflicting results. This study aims to explore the effect of the TyG index on the AF risk from a meta-analysis.MethodsA comprehensive literature search was performed in Web of Science, EMBASE, and PubMed from their inception to December 14, 2024. A meta-analysis was conducted on observational studies that reported on the relationship between the TyG index and AF risk. The subgroup analysis and sensitivity analysis were also performed.ResultsIn the meta-analysis, seven studies categorized the TyG index and found that those in the highest category had an elevated AF risk ([hazard ratio]HR = 1.96, 95% [confidence interval]CI 1.23–3.11, I2 = 92%, τ2 = 0.40) when compared to the lowest category. Ten studies treated the TyG index as a continuous measure, revealing that the risk of AF escalated by a factor of 1.94 (HR = 1.94, 95% CI 1.42–2.66, I2 = 89%, τ2 = 0.25) for every 1-unit/standard deviation (SD) increase in the TyG index. Sensitivity analysis showed that, with each study removed sequentially, the pooled AF risk estimate ranged from 1.66 to 2.19 for categorical TyG index and from 1.75 to 2.09 for continuous TyG index, indicating stable and consistent results. Subgroup analyses indicated that individuals with elevated TyG index values (as a continuous variable) faced a higher risk of AF, and this correlation was uniform across various subgroups, irrespective of post-procedure/ablation status and after accounting for potential confounders such as diabetes mellitus and triglyceride levels.ConclusionAn elevated TyG index is independently and consistently associated with a higher risk of AF. These findings support incorporating the TyG index into routine cardiovascular risk assessment to aid in the early identification and prevention of AF.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40001-025-03714-w.Keywords: TyG index, Atrial fibrillation, Meta-analysis, Systematic review Introduction Atrial fibrillation (AF) is a prevalent rapid heart rhythm disorder, affecting approximately 60 million individuals globally in 2019 [1]. The complications arising from AF, such as stroke, thromboembolism, heart failure, and cognitive decline, can significantly diminish quality of life and lead to increased rates of disability and a substantial healthcare burden [1]. The primary pathological basis for AF lies in the structural and electrophysiological abnormalities within the atria. Metabolic disorders have been linked to atrial structural and electrical changes that could trigger AF [2]. Insulin resistance (IR), a condition where tissues are less responsive to insulin, leads to decreased glucose uptake, abnormal lipid levels, and hindered glycogen synthesis [3]. A longitudinal study of individuals without diabetes revealed a positive association between IR and the risk of developing AF, even after controlling for other potential risk factors [4]. Experimental studies have also shown that IR can cause changes in atrial structure and disrupt calcium balance within cells, increasing the risk of AF [5]. The gold standard for evaluating IR is the hyperinsulinemic–euglycemic clamp, but its complexity and associated risks limit its use in clinical environments [6]. The Homeostasis Model Assessment for Insulin Resistance (HOMA-IR) is an alternative approach, yet its clinical application is hindered by the requirement for fasting insulin levels [7]. Despite these limitations, both methods have demonstrated a strong correlation with the TyG index, positioning the TyG index as a viable alternative for identifying IR [6, 7]. Although some studies have reported no strong association between an elevated TyG index and AF risk [8], others have identified the elevated TyG index as an independent risk factor for AF [9, 10]. Given the conflicting results from previous studies, we aimed to synthesize the existing research to provide more robust evidence for clinical practice. Although previous meta-analyses [11, 12] have explored the relationship between the TyG index and AF risk. For instance, Chen et al. [11] conducted a systematic review and meta-analysis, reporting a standardized mean difference (SMD) for the association between the TyG index and AF risk. Another study [12] reported a meta-analysis, reporting a odds ratio (OR) for the association between the TyG index and AF risk. These studies provided valuable insights, however, they primarily used SMD or OR, which are more common in cross-sectional studies, rather than hazard ratio (HR) that are more commonly used in clinical risk prediction for longitudinal outcomes. This difference in effect measures limits the direct applicability of their findings to clinical risk assessment. Furthermore, regarding the relationship between the TyG index and AF risk in different clinical contexts (such as pre- and post-procedure/ablation status) and after adjusting for multiple potential confounding factors, the exploration in these studies was not deep and comprehensive enough. Therefore, our study aims to further investigate the relationship between the TyG index and AF risk by analyzing the TyG index as both a categorical and continuous variable using HR simultaneously, and by conducting a thorough examination in various clinical contexts and after adjusting for multiple potential confounding factors, to fill these research gaps. Methods Meta-analysis The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were adhered to for the meta-analysis, with the checklist provided in Supplement materials Table S1. The protocol for this study has been registered in the International Prospective Register of Systematic Reviews (PROSPERO 2025 CRD420251141558. Available from https://www.crd.york.ac.uk/PROSPERO/view/CRD420251141558). Two reviewers (F–H and Q-W) conducted the search strategy, study selection, data extraction, and quality assessment independently, with any discrepancies being resolved through discussion (Table 1 and Supplementary Tables S2). Table 1Characteristics of included studies in this meta-analysisReferences (first author, year, Country)Mean age(Y), Male (%), BMI(kg/m2)Study designCategories of TyGOutcomeOR/HR (95%CI)Variables adjusted in AFParticipant characteristicsMain findingsWei,2021,China[10]50.94 ± 12.37,213(52.1),25.28 ± 3.34Retrospective cohortContinuous4.22(2.38–7.47)Age, male, BMI, hypertension, DM,hyperlipemia, left atrial diameter and TyG index409 patients who underwent septal myectomy were includedPOAF was diagnosed in 61 patients during their postoperative courseTyG index (OR: 4.218, 95% CI:2.381–7.473, P < 0.001) were independent risk factors for postoperative AF in patients undergoing septal myectomyChen,2022,China[9]NA,187(52.1),NARetrospective cohortContinuous2.09(1.41–3.10)Age, male sex, BMI, smoking, drinking, hypertension, DM,CAD,FPG,TC,TG,HDL-C,LDL-C, TyG index179 patients with AF, 36 of whom were combined with MD, and 179 age- and gender-matched subjects without AF, of whom included 36 age- and sex-matched diabetic subjectsTyG index (OR = 2.092, 95%CI 1.412–3.100, P < 0.001) were positively associated with AFNA,37(51.4),NA1.29(0.65–2.57)Age, male sex, BMI, smoking, drinking, hypertension, CAD, FPG, TC, TG, HDL-C, LDL-C, TyG index36 patients with AF were combined with DM, and 36 age- and sex-matched diabetic subjectsTyG index (OR = 1.286, 95%CI, 0.645–2.565,P = 0.475) is not associated with AF in diabetic subjectsNA,150(52.4),NA3.07(1.82-5.17)Age, male sex, BMI, smoking,drinking,hypertension,CAD,ACEI/ARB,Beta-blockers,CCBFPG,TC,TG,HDL-C,LDL-C, TyG index143 patients with AF without MD and 36 age- and sex-matched diabetic subjectsTyG index(OR = 3.065, 95%CI, 1.819–5.166, P < 0.001) is associated with AF in non-diabetic subjectsTang,2022,China[17]57.32 ± 9.57,191 (69.4),25.7 ± 2.9Retrospective cohortContinuous2.02(1.41–4.12)Age > 65 years, male gender, hypertension, BMI > 28 kg/m2,hs‑CRP,NT‑proBNP, uric acid, LDL‑C, Hcy, creatinine, TyG index, alcohol intake, persAF,duration of AF history > 5 years, EHRA score,CHA2DS2‑VASc score, APPLE score,DR‑FLASH score, LVEF, LAD, early AF recurrence275 patients with AF who underwent radiofrequency ablation, of which 70 patients were observed to have late recurrence of AF after a follow-up of 26.1 monthsTyG index (HR 2.015 95% CI: 1.408–4.117); p = 0.009) was an independent risk factor for late recurrence of AF after RFCALing,2022,China[18]NA438(79.8)NARetrospective cohortContinuous8.88(1.57–50.27)Age,hypertension,DM,BMI,hemoglobin, white blood cell, neutrophil,platelet,glucose,triglyceride,Total cholesterol,HDL-C,LDL-C,eGFR,TyG index, Left atrium diameter, LVEF, Stent length and Killip class ≥ IIThis retrospective cohort research enrolled 549 STEMI patients that had undergone PCI, with these patients being clustered into the NOAF group and sinus rhythm (SR) groupThe TyG index (OR,8.884, 95%CI: 1.570–50.265, P = 0.014) was found to be an independent predictor of NOAFShi,2022,China[19]56.19 ± 10.86,1780 (54.87),25.35 ± 3.44Cross-sectionalQuartiles of TyGContinuousRef1.17 (0.69- 1.96)1.29(0.77–2.14)2.12 (1.30- 3.35)1.41(1.20–1.65)Age, gender, education level, current smoking, drinking status, BMI, WC, Scr, TC, HDL-C, SBP, anti-hypertensive therapy, lipid-lowering therapy, and CVD historyOf the 3244 included subjects with diabetes, 213 (6.57%) had a history of AFIn the quartile analysis, the top quartile showed a 2.120 times risk of prevalent AF compared with the bottom quartileLiu,2023,USA[20]54.04 ± 5.74,5265 (44.4),27.37 ± 5.15Prospectivecohort < 8.80,8.80–9.20, > 9.201.15(1.02–1.29)/1(ref.)/1.18 (1.03–1.37)Age, sex, race, BMI, smoking, drinking, LDL-C,HDL-C, use of cholesterol-lowering medication, history of diabetes, hypertension, and PADOf 11,851 participants, the mean age was 54.0 years; 6586 (55.6%) were female. During a median follow-up of 24.26 years, 1925 incidents of AF cases (0.78/per 100 person-years) occurredboth < 8.80 (aHR = 1.15, 95% CI:1.02–1.29) and > 9.20 levels (aHR = 1.18, 95% CI: 1.03–1.37) of the TyG index were associated with an increased risk of AF compared with the middle TyG index category (8.80–9.20)Muhammad,2023,Sweden[8]NA,NA,NAProspectivecohortQuartiles of TyG;ContinuousRef1.02(0.95–1.09)0.97(0.90–1.04)0.96(0.89–1.04)0.99(0.89–1.11)Age, sex, BMI, systolic blood pressure, cholesterol, smoking status, diabetes, antihypertensive medication, activity and alcoholData for this analysis was used from the Malmö Diet and Cancer Study-Cardiovascular Cohort (MDCS-CV) and the Malmö Preventive Project (MPP)No association was observed with AF (aHR 0.99,95%CI 0.89—1.11, P = 0.142)Zhang Y,2023,China[21]NA,561(61.5)NAretrospective cohort8.39 ± 0.208.78 ± 0.839.09 ± 0.999.57 ± 0.27Continuous variableRef1.20 (0.66–2.17)1.93 (1.07–3.49)4.34 (2.37–7.94)4.84(2.98–7.88)Age, sex, DM, BMI, TC and ScrA retrospective study was conducted on 912 patients diagnosed with NAFLD via ultrasonography. 204 patients with nonvalvular AF and 708 patients without AF (sinus rhythm) were included in the studyTyG was an independent risk factor for AF (OR = 4.84, 95% CI2;.98–7.88, P < 0.001)Zhang J,2023,China[22]NA,300(70.8)NARetrospective cohortContinuous variable2.02(1.37–3.25)Hypertension, diabetes, smoking history, drinking history, BMI, NYHA III ~ IV, left atrial transverse diameterThe data of the patients who underwent valvular surgery with concurrent Cox‑maze IV ablation from June 2017 to May 2022TyG index (HR = 2.021, 95%CI:1.374–3.245, P < 0.001) isrisk factors for AF recurrenceafter Cox‑maze IV ablationPeng,2023,China[23]NA,395(70.4)NARetrospective cohort ≤ 8.45,8.45–8.80, > 8.80Continuous variableRef1.54 (0.25–1.17)2.88 (1.76–4.71)1.24(1.03–1.73)Age, sex,BMI, CRP, DM, Emergency operation, NYHA III–IV, and Left atrial diameterThis study used longitudinal data from the cohort of 542 participants who underwent OPCABG in Beijing Anzhen Hospital since June 2017The highest tertile of VATI (HR = 2.58, 95%CI: 1.12–3.45; p = 0.01) and TyG index (HR = 2.88, 95% CI: 1.76–4.71; p = 0.01) were significantly associated with new-onset POAF compared to the lowest tertileWang,2024,China[24]60.74 ± 11.291421(63.38)25.06 ± 3.33Retrospectivecohort ≤ 8.325,8.325–8.765, > 8.765Continuous variable < 8.692 ≥ 8.692Reference0.87 (0.71–1.05)1.25 (1.03–1.51)1.18 (1.02–1.36)Reference1.40 (1.19–1.64)Age, sex, body mass index, current drinking, diabetes mellitus, hyperlipidemia, coronary heart disease, duration of AF (≥ 24 months), AF type, HbA1c, left atrial diameter, platelet, CHA2DS2-VASc scoreA total of 2242 AF patients who underwent a de novo RFCA between June 2018 to January 2022 at two hospitals in ChinaTyG index was significantly associated with AFrecurrence regardless of whether as continuous variable or categorical variable by optimal cut-of value of ROC (P < 0.05)Yan,2024China[25]63.23 ± 10.73241 (64.27)24.69 ± 3.18Retrospectivecohortthe low cumulative Indexthe medium cumulative indexthe high cumulative indexReference4.95 (1.78–13.78)8.72(3.37–22.54)Age ≥ 75, BMI, Hyperlipemia, persistent AF,duration of AF, lactic dehydrogenase, left atrial diameter, type of AAD, and ablation strategyA total of 576 patients with AF who underwent RFCA at the Second Affiliated Hospital of Xi’an Jiaotong University were included in this studyAfter adjusting for potential confounders, patients in the medium cumulative TyG index group [hazard ratio (HR) = 4.949, 95% CI: 1.778–13.778, P = 0.002] and the high cumulative TyG index group (HR = 8.716, 95% CI: 3.371–22.536, P < 0.001) had a higher risk of AF recurrence than those in the low cumulative TyG index groupAF atrial fibrillation, BMI body mass index, DM diabetes mellitus, CAD coronary artery disease, ACEI/ARB angiotensin-converting enzyme inhibitors/angiotensinII receptor blockers, CCB calcium channel blockers, SBP systolic blood pressure, DBP diastolic blood pressure, FPG fasting plasma glucose, TC total cholesterol, TG triglyceride, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, TyG index triglyceride–glucose index, CRP C-reactive protein, OR odds ratio, CI confidence interval, HR hazard ratio, aHR adjusted hazard ratio, NA not available Characteristics of included studies in this meta-analysis Wei, 2021, China[10] 50.94 ± 12.37, 213(52.1), 25.28 ± 3.34 Age, male, BMI, hypertension, DM, hyperlipemia, left atrial diameter and TyG index 409 patients who underwent septal myectomy were included POAF was diagnosed in 61 patients during their postoperative course Chen, 2022, China [9] NA, 187(52.1), NA NA, 37(51.4), NA NA, 150(52.4), NA 3.07(1.82- 5.17) Tang, 2022, China [17] 57.32 ± 9.57, 191 (69.4), 25.7 ± 2.9 Ling, 2022, China [18] NA 438(79.8) NA Age,hypertension,DM,BMI,hemoglobin, white blood cell, neutrophil,platelet,glucose, triglyceride,Total cholesterol,HDL-C,LDL-C,eGFR,TyG index, Left atrium diameter, LVEF, Stent length and Killip class ≥ II Shi, 2022, China [19] 56.19 ± 10.86, 1780 (54.87), 25.35 ± 3.44 Cross- sectional Quartiles of TyG Continuous Ref 1.17 (0.69- 1.96) 1.29(0.77–2.14) 2.12 (1.30- 3.35) 1.41(1.20–1.65) Liu, 2023, USA [20] 54.04 ± 5.74, 5265 (44.4), 27.37 ± 5.15 Prospective cohort < 8.80, 8.80–9.20, > 9.20 1.15(1.02–1.29)/ 1(ref.)/ 1.18 (1.03–1.37) Muhammad, 2023, Sweden [8] NA, NA, NA Prospective cohort Quartiles of TyG; Continuous Ref 1.02(0.95–1.09) 0.97(0.90–1.04) 0.96(0.89–1.04) 0.99(0.89–1.11) Zhang Y, 2023, China [21] NA, 561(61.5) NA 8.39 ± 0.20 8.78 ± 0.83 9.09 ± 0.99 9.57 ± 0.27 Continuous variable Ref 1.20 (0.66–2.17) 1.93 (1.07–3.49) 4.34 (2.37–7.94) 4.84(2.98–7.88) Zhang J, 2023, China [22] NA, 300(70.8) NA TyG index (HR = 2.021, 95%CI: 1.374–3.245, P < 0.001) is risk factors for AF recurrence after Cox‑maze IV ablation Peng, 2023, China [23] NA, 395(70.4) NA ≤ 8.45, 8.45–8.80, > 8.80 Continuous variable Ref 1.54 (0.25–1.17) 2.88 (1.76–4.71) 1.24(1.03–1.73) Wang, 2024, China [24] 60.74 ± 11.29 1421(63.38) 25.06 ± 3.33 Retrospective cohort ≤ 8.325, 8.325–8.765, > 8.765 Continuous variable < 8.692 ≥ 8.692 Reference 0.87 (0.71–1.05) 1.25 (1.03–1.51) 1.18 (1.02–1.36) Reference 1.40 (1.19–1.64) TyG index was significantly associated with AF recurrence regardless of whether as continuous variable or categorical variable by optimal cut-of value of ROC (P < 0.05) Yan, 2024 China [25] 63.23 ± 10.73 241 (64.27) 24.69 ± 3.18 Retrospective cohort the low cumulative Index the medium cumulative index the high cumulative index Reference 4.95 (1.78–13.78) 8.72(3.37–22.54) AF atrial fibrillation, BMI body mass index, DM diabetes mellitus, CAD coronary artery disease, ACEI/ARB angiotensin-converting enzyme inhibitors/angiotensinII receptor blockers, CCB calcium channel blockers, SBP systolic blood pressure, DBP diastolic blood pressure, FPG fasting plasma glucose, TC total cholesterol, TG triglyceride, HDL-C high-density lipoprotein cholesterol, LDL-C low-density lipoprotein cholesterol, TyG index triglyceride–glucose index, CRP C-reactive protein, OR odds ratio, CI confidence interval, HR hazard ratio, aHR adjusted hazard ratio, NA not available Search strategy Electronic databases such as Web of Science, EMBASE, and PubMed were queried using a combination of the following search terms: "triglyceride-glucose index," "TyG index," "triglyceride glucose index," "triglyceride and glucose index," along with "atrial fibrillation," "atrium fibrillation," "auricular fibrillation," and "cardiac atrial fibrillation." The searches were performed using a mix of Medical Subject Headings (MeSH) and free-text keywords, without any language limitations. To ensure a comprehensive search, we also reviewed the references of related literature. Endnote X9 software by Thomson Reuters in New York, NY, USA, was utilized to manage the collected studies. The initial screening was based on titles and abstracts, followed by a full-text review for further assessment. The literature search was concluded on December 14, 2024. The complete search terms for each database are detailed in Supplement materials Table S2. Study selection The inclusion criteria for the study were established according to the PICOS framework, which stands for participants, interventions, comparators, outcomes, and study design. The specifics are as follows (1): Participants: The studies included involved an adult population, defined as individuals aged 18 years or older; (2) Interventions and Comparators: For categorical variables, the studies compared the highest group of TyG index to the lowest. For continuous variables, the studies evaluated the impact of a 1 standard deviation (SD)/unit increase in the TyG index. (3)Outcomes: The studies reported the HR or OR and their corresponding 95% confidence interval (CI) to assess the relationship between the TyG index and the risk of AF. (4) The studies were observational and published as full-length articles. Exclusion criteria were applied to omit certain types of studies, such as review articles, meta-analyses, letters, animal experiments, in vitro studies, duplicate publications, and conference abstracts. Data extraction and quality assessment We extracted the following information from the included studies: the first author’s name, publication year, country/region, mean age, male(%), body mass index(BMI), study design, categories of TyG index, outcome of OR/HR (95%CI), variables adjusted in AF, participant characteristics, and main findings. Each included study was independently assessed for risk of bias using the QUAPAS (Quality Assessment of Prognostic Accuracy Studies)−2, which evaluates five distinct domains: patient selection(D1), index test(D2), reference standard(D3), flow and timing(D4). Each domain was categorized as having a low, unclear risk of bias, or high. A summary plot illustrating the overall assessment of risk of bias across included studies was generated using the Robvis visualization tool. [13] Statistical analysis The relationship between the TyG index and the risk of AF was assessed using the adjusted HR/OR and its corresponding 95% CI as the primary measure. For cross-sectional studies included in the analysis, the OR was treated as equivalent to the HR. When studies presented the TyG index as a continuous variable, the pooled HR was calculated for the incidence of AF per 1 standard deviation (SD) or unit increase in the TyG index. In cases where the TyG index was reported categorically, the pooled HR was determined for patients with the highest category compared to the lowest. The I2 statistic was employed to assess statistical heterogeneity; significant heterogeneity was indicated if I2 exceeded 50% or if the P value was less than 0.1, in which case a random-effects model was used. Otherwise, a fixed-effects model was applied [14]. Sensitivity analysis was conducted to determine the stability and reliability of the results by examining the impact of each individual study on the overall findings. Subgroup analyses were performed based on post-procedure/ablation and adjustments for potential confounding factors such as diabetes mellitus and triglyceride levels. If the meta-analysis included more than 10 studies, publication bias was assessed using funnel plot asymmetry and either Begg's or Egger's tests, with a P-value less than 0.05 suggesting the presence of publication bias [15, 16]. All statistical analyses and the creation of figures were carried out using R software (version 4.3.2). A P-value less than 0.05 or a 95% CI that did not include 1 was considered to indicate statistical significance. Results Meta-analysis Search results and study characteristics The PRISMA flow diagram, depicted in Fig. 1, outlines the comprehensive process of literature search and study selection for this analysis. After the screening process, a total of 12 studies were included in the meta-analysis [8–10, 17–25], with one study involving an American population, another a Swedish population, and the remaining 10 focusing on Chinese populations. The meta-analysis comprised 11 cohort studies and a single cross-sectional study. In these studies, the TyG index was treated as a categorical variable in 7 studies and as a continuous variable in 10 studies (Table 1). The PRISMA flow chart for study selection Risk of bias assessment Risk of bias across twelve studies was assessed using QUADAS-2 and visualized in a traffic-light plot generated by the Robvis tool [13]. Seven studies [10, 19–22, 24, 25] were rated as having unclear patient selection bias because they did not report using random or consecutive sampling. Five studies [9, 10, 18, 19, 23] exhibited high risk of bias in the index test domain by employing TyG thresholds or quantiles that lacked pre-specification. All studies were classified as low risk of bias, since the diagnosis of AF relied on electrocardiogram or International Classification of Diseases, Ninth Revision (ICD-9) code. All articles showed high bias risk in the flow and timing dimension because none refrained from post-test interventions. Overall, three studies [10, 19, 22] were classified as high risk of bias, six studies [9, 18, 20, 21, 24, 25] as having some concerns, and the remainder as low risk of bias. (Fig. 2). Traffic-light plot of risk of bias assessment using QUADAS-2 for included studies (n = 12), generated using the robvis visualization tool Relationship between TyG index and the risk of AF In the analysis of the TyG index as a categorical variable across seven studies, it was found that individuals in the highest TyG group had a significantly increased risk of AF compared to those in the lowest TyG group, with a HR of 1.96 (95% CI: 1.23 to 3.11). This analysis also revealed a high degree of heterogeneity, reflected by an I2 statistic of 92% and a tau-squared (τ2) value of 0.40 (Fig. 3A). In addition, when examining the TyG index as a continuous variable in ten studies, the results indicated that for each 1-unit or SD increase in the TyG index, the risk of AF increased by 1.94 times, with an HR of 1.94 (95% CI: 1.42 to 2.66). This analysis also demonstrated a substantial level of heterogeneity, with an I2 statistic of 89% and a τ2 value of 0.25 (Fig. 3B). The forest plot for the association between TyG index and the risk of AF in meta-analysis. A TyG index as a categorical variable and the risk of AF B TyG index as a continuous variable and the risk of AF. The Chen 2022, Chen 2022A, and Chen 2022B are different populations from the same study. TyG triglyceride glucose index, AF atrial fibrillation, HR hazard ratio, CI confidence interval, SD standard deviation Sensitivity analysis, subgroup analyses and publication bias A sensitivity analysis conducted by eliminating each study one by one and re-evaluating the meta-analysis showed that when the TyG index was treated as a categorical variable, the pooled risk estimate for AF varied between 1.66 (95% CI 1.14 to 2.41) and 2.19 (95% CI 1.35 to 3.56) (Fig. 4A). This suggests that the overall impact remains robust even when individual studies are excluded. When the TyG index was considered as a continuous variable, the pooled risk estimate spanned from an HR of 1.75 (95% CI 1.32 to 2.31) to 2.09 (95% CI 1.52 to 2.86) (Fig. 4B). These results indicate that the findings are both dependable and consistent. Subgroup analyses indicated that individuals with elevated TyG index values (as a continuous variable) faced a higher risk of AF, and this correlation was uniform across various subgroups, irrespective of post-procedure/ablation status and after accounting for potential confounders such as diabetes mellitus and triglyceride levels (Fig. 5A–C). Publication bias assessment was not conducted in this research due to the inclusion of fewer than ten studies, which is the typical minimum for such evaluations. Subgroup analyses for the association between the TyG index analyzed as a continuous variable and the risk of AF. A Subgroup analysis according to the post-procedure/ablation. B Subgroup analysis according to the adjust for triglyceride. C Subgroup analysis according to the adjust for diabetes. The Chen 2022, Chen 2022A, and Chen 2022B are different populations from the same study. TyG triglyceride glucose index, AF atrial fibrillation, HR hazard ratio, CI confidence interval, SD standard deviation Sensitivity analysis of the association between TyG index and the risk of AF. The Chen 2022, Chen 2022A, and Chen 2022B are different populations from the same study. TyG triglyceride glucose index, AF atrial fibrillation, HR hazard ratio, CI confidence interval, SD standard deviation Discussion To our understanding, this study represents the first comprehensive meta-analysis to explores the correlation between the TyG index and AF risk by analyzing the TyG index as both a categorical and continuous variable, utilizing HRs to provide a more precise and clinically applicable measure of this relationship. Our findings not only align with previous studies [11, 12] but also provide more detailed risk assessment data. Through extensive sensitivity and subgroup analyses, we further verified the stability and consistency of this relationship. Crucially, our subgroup analyses uncovered that the relationship between the TyG index and AF risk is robust irrespective of post-procedure/ablation status. This is a significant advantage because it demonstrates the consistency of the TyG index as a risk predictor across different treatment scenarios. post-procedure/ablation status can vary widely among patients with AF, and the fact that our findings hold true regardless of whether patients have undergone ablation procedures indicates that the TyG index is a reliable marker that can be applied broadly in clinical practice. Furthermore, the relationship is independent of diabetes mellitus and triglyceride levels. This independence is crucial because diabetes mellitus and triglyceride levels are well-known risk factors for AF and can often confound the assessment of other risk markers. By showing that the TyG index remains a significant predictor even when accounting for these factors, our study underscores the unique and valuable role of the TyG index as a marker of integrated metabolic health. This finding highlights the potential of the TyG index to provide a more comprehensive and nuanced understanding of the metabolic underpinnings of AF risk. Our findings extend prior evidence linking TyG to coronary artery disease, ischaemic stroke and all-cause mortality [26, 27]. A recent systematic review by Liu et al [28]. similarly reported a positive association between TyG and composite cardiovascular events in the general population. but AF was not specifically analysed. Two previous cohort studies, one from China [17] and the other from Sweden [8], estimated HRs per SD of 1.18 and 1.02 for AF, respectively; our pooled estimate is numerically higher, probably because we restricted the analysis to fully-adjusted models and included data from multicenter registries. Notably, compared with previous studies [11, 12] that included mostly cross-sectional studies, the present study included mostly cohort studies. This significantly reduced the potential recall bias that may be associated with a cross-sectional study design. Cohort studies allow for the observation of outcomes over time, crucial for understanding temporal relationships. They also minimize selection bias by following a specific group of individuals and provide a rich data source for exploring multiple outcomes, enhancing research reliability and robustness. The subgroup analyses in this study underscore the robustness of the TyG index as a predictor of AF risk across varied clinical contexts. Specifically, the analysis accounting for post-procedure/ablation status shows that the TyG index remains a significant predictor of AF risk, regardless of whether individuals have undergone procedures like radiofrequency ablation. This consistency is evident even when combined with other risk factors, highlighting its utility in both pre- and post-procedure settings. For example, Wei et al. [10] demonstrated that the TyG index has a moderate predictive value for postoperative AF in hypertrophic obstructive cardiomyopathy patients, with an AUC of 0.723 (95% CI 0.650 to 0.796, P < 0.001). Similarly, Zhang et al. [21] found that combining the TyG index with traditional risk factors significantly improved the predictive value for AF in patients with non-alcoholic fatty liver disease, with an area under the curve (AUC) of 0.857 (95% CI 0.827–0.887, P < 0.00.1). Adjusting for diabetes mellitus, a known risk factor for AF, the TyG index still shows a significant association with AF risk. This indicates that the TyG index captures metabolic dysfunction beyond diabetes, making it a valuable tool across populations with varying diabetes prevalence. Liu et al. [20] reported a U-shaped association between the TyG index and AF incidence in a general population without known cardiovascular disease, further supporting its broad applicability. Moreover, the subgroup analysis adjusting for triglyceride levels confirms that the TyG index's predictive power is not solely driven by triglycerides but reflects a comprehensive metabolic profile. Shi et al. [19] found the TyG index useful for detecting prevalent AF in a type 2 diabetic population, reinforcing its role as a metabolic marker. In summary, the TyG index consistently predicts AF risk across different clinical scenarios and populations, even when accounting for other metabolic factors. This makes it a potentially valuable tool for cardiovascular risk stratification. Future research should explore the underlying mechanisms and validate the TyG index's utility in diverse clinical settings. Our study demonstrated substantial heterogeneity (I2 > 80%). Sensitivity analyses revealed that the heterogeneity remained relatively stable across different modeling approaches for TyG (I2 ranged from 85 to 93% for categorical and 85% to 90% for continuous). However, subgroup analyses provided insights into potential sources of heterogeneity. The I2 value decreased to 53% when triglyceride levels were not adjusted for, compared to 91% when they were adjusted. Similarly, the I2 value decreased to 64% when diabetes mellitus was not adjusted for, compared to 92% when it was adjusted. These findings indicate that both triglyceride levels and diabetes mellitus status significantly impact the observed heterogeneity, but they cannot explain the source of heterogeneity. From a clinical standpoint, this may suggest that simply considering triglyceride levels and diabetes mellitus status is insufficient when assessing cardiovascular risk. This is because they are closely related to insulin resistance, which is a key component of metabolic syndrome and is associated with various cardiovascular risk factors. From an epidemiological perspective, the initially high I2 values (up to 92%) likely reflect methodological differences across studies, such as variations in follow-up duration, methods for calculating the TyG index, and tools for AF diagnosis. This highlights the necessity for future studies to use standardized protocols to reduce heterogeneity and enhance the comparability of results. Additionally, it indicates that the relationship between the TyG index and AF risk may vary across different populations and clinical settings, further emphasizing the importance of validating the TyG index in diverse populations to ensure its generalizability and applicability. IR plays a crucial role in linking elevated TyG levels and AF risk through multiple pathways. Experimental work shows that IR promotes atrial fibrosis via activation of the PI3K/Akt/mTOR pathway, down-regulation of connexin-43, and enhanced oxidative stress [5]. In parallel, IR-driven hyperinsulinaemia increases sympathetic tone, shortens atrial effective refractory period and facilitates ectopic firing from pulmonary-vein sleeves [5]. Moreover, IR is intimately coupled with a pro-thrombotic state characterized by increased platelet reactivity and elevated fibrinogen, which may further perpetuate atrial remodelling. [29] Limitations First, despite rigorous adjustment, residual confounding from unmeasured factors (e.g., sleep apnoea severity [30], physical activity [31], dietary patterns [32]) cannot be excluded. Second, although the heterogeneity was reduced after subgroup analyses, the high initial I2 values (up to 92%) likely reflect underlying methodological differences across studies, such as variations in follow-up duration, TyG calculation protocols, and AF ascertainment tools. Third, the absence of patient-level data precluded time-to-event modelling or non-linear dose–response curves. Finally, publication bias could not formally be assessed because fewer than ten studies were available for each analytic model, yet the comprehensive search strategy and inclusion of grey literature mitigate this concern. Conclusion In summary, this first comprehensive meta-analysis demonstrates that an elevated TyG index—whether analysed categorically or continuously utilizing HRs—is independently associated with an approximate doubling of AF risk across diverse populations. The association is consistent, biologically plausible, and not materially altered by ablation status or adjustment for diabetes and triglyceride levels. Incorporating the TyG index into routine cardiovascular risk assessment may enhance early AF detection and inform targeted preventive strategies. Future prospective studies should clarify optimal thresholds, evaluate therapeutic modulation, and extend validation to under-represented ethnicities. Supplementary Information Supplementary file 1 Acknowledgements None. Author contributions Study design: LH, QW and FH; Data collection: QW, FH, and BW; Data analyses: QW, JLand FH; Manuscript revised: LH, QW and JL; Manuscript writing: All authors; Results interpretations: All authors; Manuscript proofing: All authors. Funding This work was no supported by grants. Data availability Not applicable. Declarations Ethics approval and consent to participate Does not involve ethical approval and consent. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary file 1 Data Availability Statement Not applicable.