Abstract
Background
Postoperative atrial fibrillation (POAF) is a common complication following off-pump coronary artery bypass grafting (OPCABG). Insulin resistance (IR) contributes to the pathogenesis of POAF. IR-related markers, including the triglyceride-glucose (TyG) index and the triglyceride-glucose-body mass index (TyG-BMI), have demonstrated prognostic value for cardiovascular events. However, their associations with POAF specifically after OPCABG remain unclear. This study aimed to evaluate the associations of TyG and TyG-BMI with POAF and their incremental value beyond conventional clinical risk factors.
Methods
We conducted a single-center retrospective cohort study of adult patients who underwent isolated OPCABG between December 2020 and December 2021. Of 757 otherwise eligible patients, 93 without the preoperative fasting glucose and/or triglyceride measurements required to calculate the indices were excluded, leaving 664 participants in the primary analysis. Multivariable logistic regression, restricted cubic spline (RCS), discrimination and reclassification analyses, and bootstrap internal validation were performed.
Results
Of the 664 participants, 180 (27.1%) developed POAF. In the fully adjusted model, each standard deviation increase in TyG and TyG-BMI was associated with higher odds of POAF [TyG: odds ratio (OR), 1.55; 95% confidence interval (CI): 1.28–1.89; TyG-BMI: OR, 1.60; 95% CI: 1.31–1.96; both P<0.001]. RCS analysis showed an approximately linear association for TyG (P for nonlinearity =0.16) and a nonlinear association for TyG-BMI (P for nonlinearity =0.02). Adding TyG or TyG-BMI to the baseline model increased the C-statistic from 0.682 to 0.708 or 0.707, respectively, although the changes were not statistically significant (P=0.07 and P=0.08). Both indices improved the integrated discrimination improvement (IDI) and continuous net reclassification improvement (NRI) (all P<0.001). Associations were stronger among patients with diabetes (P for interaction =0.02 for TyG and 0.005 for TyG-BMI).
Conclusions
Higher preoperative TyG and TyG-BMI values were independently associated with POAF after OPCABG. These routinely available indices provided modest incremental risk information beyond conventional clinical risk factors. External validation is required before their use for clinical decision-making.
Keywords: Insulin resistance (IR), triglyceride-glucose index (TyG index), triglyceride-glucose-body mass index (TyG-BMI), postoperative atrial fibrillation (POAF), off-pump coronary artery bypass grafting (OPCABG)
Highlight box.
Key findings
• Higher preoperative triglyceride-glucose (TyG) and TyG-body mass index (TyG-BMI) values were independently associated with postoperative atrial fibrillation (POAF) after off-pump coronary artery bypass grafting. Associations were stronger among patients with diabetes. Stand-alone discrimination was modest; adding either index did not significantly increase the C-statistic, although integrated discrimination improvement and continuous net reclassification improvement improved.
What is known and what is new?
• Insulin resistance is implicated in atrial remodeling, and TyG-related indices are inexpensive surrogate markers.
• In this retrospective cohort, the adjusted association was approximately linear for TyG and nonlinear for TyG-BMI. Bootstrap analysis supported coefficient stability but indicated some optimism in model performance.
What is the implication, and what should change now?
• TyG and TyG-BMI may complement conventional preoperative risk assessment, particularly in patients with diabetes, but should not be used as stand-alone prediction or treatment-selection tools. External validation is needed before clinical implementation.
Introduction
Postoperative atrial fibrillation (POAF) is a common complication following off-pump coronary artery bypass grafting (OPCABG), with a reported incidence of 20–40% (1-4). POAF is associated with significant clinical consequences, including prolonged intensive care unit (ICU) stay, a three-fold increased risk of stroke, and increased hospitalization costs (5,6). The pathogenesis of POAF involves autonomic imbalance, systemic inflammation, and oxidative stress, processes that can be exacerbated by metabolic dysregulation (1,7). Notably, obesity and metabolic syndrome are established independent risk factors for POAF (8), highlighting the potential contribution of metabolic disturbances to its development.
Clinical risk scores such as CHA2DS2-VASc have been evaluated for predicting POAF after OPCABG (9), but further refinement of risk stratification remains warranted. Other biomarkers, including N-terminal pro-B-type natriuretic peptide (NT-proBNP) and high-sensitivity cardiac troponin T (hs-cTnT), as well as echocardiographic parameters such as left atrial diameter (LAD), have been investigated for risk stratification. However, these markers often reflect end-stage myocardial injury rather than the upstream metabolic mechanisms involved in arrhythmogenesis. Recent evidence suggests that insulin resistance (IR) plays a pivotal role in the pathogenesis of POAF and other cardiovascular disorders. The triglyceride-glucose (TyG) index, a surrogate marker of IR, has been associated with thrombotic events in patients with atrial fibrillation (10), supporting further investigation of its potential relevance to POAF. Obesity and type 2 diabetes mellitus (T2DM) are major contributors to cardiovascular risk, and body mass index (BMI), a key measure of obesity, is intrinsically linked to IR (11). The TyG-BMI index, which integrates TyG and BMI, has demonstrated strong performance in assessing IR (12,13). This index has been associated with major adverse cardiovascular and cerebrovascular events (MACCEs) in elderly and female patients undergoing percutaneous coronary intervention (14), and with coronary lesion complexity in women with acute coronary syndrome (15).
We hypothesized that preoperative TyG and TyG-BMI could serve as useful risk markers for POAF in patients undergoing OPCABG. To test this hypothesis, we conducted a retrospective cohort study to evaluate their associations with POAF and their incremental value beyond conventional clinical risk factors. Our goal was to improve preoperative risk stratification for POAF, thereby facilitating targeted preventive measures to reduce healthcare costs and postoperative complications. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1640/rc).
Methods
Study design and population
We conducted a single-center retrospective observational cohort study at General Hospital of Northern Theater Command, Shenyang, China. Adult patients who underwent CABG, including on-pump CABG and OPCABG, between December 2020 and December 2021 were screened for eligibility. Preoperative data were collected from the electronic medical record system of General Hospital of Northern Theater Command at the time of the patient’s admission, while postoperative outcomes were tracked through continuous follow-up until the patient’s discharge. Patients were excluded for (I) emergency or salvage surgery; (II) on-pump CABG surgery; (III) any prior cardiac surgery; (IV) pre-existing atrial fibrillation, as determined by 7-day Holter monitoring prior to surgery (16); or (V) missing essential clinical data. After applying these clinical exclusion criteria, 757 patients comprised the eligible source cohort. Because the present analysis focused on TyG and TyG-BMI, 93 patients without the preoperative fasting glucose and/or triglyceride measurements required to calculate these indices were excluded. The primary analysis therefore included 664 participants. Missing exposure values were not imputed.
The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of General Hospital of Northern Theater Command [approval No. Y(2025)323]. Informed consent was waived in this retrospective study.
Data collection and definitions
Patient data were collected from the electronic medical record system of General Hospital of Northern Theater Command and recorded in a standardised format. The clinical definitions employed in this study adhered to the established criteria. Hypertension was defined as a blood pressure reading of ≥140/90 mmHg or the current use of antihypertensive medication. The condition of diabetes mellitus was defined by the use of glucose-lowering medication or by meeting any of the following criteria: fasting plasma glucose ≥7.0 mmol/L, 2-hour oral glucose tolerance test ≥11.1 mmol/L, random plasma glucose ≥11.1 mmol/L in the presence of classic hyperglycemic symptoms, or HbA1c ≥6.5%.
BMI was calculated by dividing weight in kilograms by height in metres squared. The IR-related indices were calculated using triglycerides (TG) and fasting blood glucose (FBG) as follows: TyG index = ln [TG (mg/dL) × FBG (mg/dL) ÷ 2]; and TyG-BMI index = TyG × BMI (kg/m2).
Definition of POAF
POAF was defined as any atrial tachyarrhythmia lasting at least 30 seconds that occurred after surgery (17). POAF was identified via continuous rhythm surveillance using a 7-day single-lead Holter monitor (16). Any episode meeting this duration criterion and confirmed as atrial fibrillation or atrial flutter was classified as POAF.
Statistical analysis
Continuous variables are presented as mean ± standard deviation for approximately normally distributed variables or as median with interquartile range for skewed distributions. Between-group comparisons were performed using Welch’s t-test for normally distributed variables and the Mann-Whitney U test for non-normally distributed variables. Categorical variables are summarized as counts and percentages and were compared using Chi-squared tests or Fisher’s exact test, as appropriate.
Univariable logistic regression analyses were performed to describe crude associations between candidate variables and POAF (Table S1). Covariates for multivariable adjustment were prespecified according to clinical relevance and prior literature rather than selected solely according to the P values in Table 1 or the univariable analysis. The prespecified variables were age, sex, hypertension, diabetes mellitus, current smoking, current drinking, creatinine, heart failure, preoperative beta-blocker use, statin use, LAD, and left ventricular ejection fraction (LVEF). BMI was additionally included in models of TyG, but not in models of TyG-BMI because BMI is a component of TyG-BMI.
Table 1. Baseline characteristics of participants by postoperative atrial fibrillation.
| Variables | Available (n) | Non-POAF (n=484) | POAF (n=180) | P value |
|---|---|---|---|---|
| Clinical characteristics | ||||
| Age, years | 664 (484/180) | 63.00 (57.75–68.00) | 65.00 (60.75–71.00) | <0.001 |
| Male | 664 (484/180) | 342 (70.66) | 137 (76.11) | 0.16 |
| BMI, kg/m2 | 664 (484/180) | 24.32 (22.72–26.00) | 25.12 (23.43–26.48) | 0.002 |
| Heart rate, bpm | 664 (484/180) | 77.36±12.30 | 76.22±12.39 | 0.29 |
| SBP, mmHg | 664 (484/180) | 136.00 (122.75–150.25) | 138.00 (124.00–152.00) | 0.19 |
| DBP, mmHg | 664 (484/180) | 78.56±12.59 | 79.00±12.70 | 0.69 |
| Preoperative IABP | 664 (484/180) | 1 (0.21) | 1 (0.56) | 0.47 |
| NYHA functional class | 660 (480/180) | 0.50 | ||
| I | 7 (1.46) | 1 (0.56) | ||
| II | 432 (90.00) | 167 (92.78) | ||
| III | 38 (7.92) | 12 (6.67) | ||
| IV | 3 (0.63) | 0 (0.00) | ||
| CHA2DS2-VASc score | 664 (484/180) | 3.00 (2.00–4.00) | 3.00 (2.00–4.00) | 0.74 |
| Medical history | ||||
| Hypertension | 664 (484/180) | 288 (59.50) | 104 (57.78) | 0.69 |
| Diabetes mellitus | 664 (484/180) | 195 (40.29) | 80 (44.44) | 0.33 |
| Current smoking | 664 (484/180) | 185 (38.22) | 72 (40.00) | 0.68 |
| Current drinking | 664 (484/180) | 48 (9.92) | 30 (16.67) | 0.02 |
| Previous PCI | 664 (484/180) | 67 (13.84) | 24 (13.33) | 0.87 |
| Previous stroke | 664 (484/180) | 31 (6.40) | 14 (7.78) | 0.53 |
| Heart failure | 664 (484/180) | 27 (5.58) | 16 (8.89) | 0.12 |
| Laboratory tests | ||||
| Creatinine, μmol/L | 663 (484/179) | 73.70 (63.35–87.91) | 74.81 (64.78–90.75) | 0.39 |
| Glucose, mmol/L | 664 (484/180) | 5.80 (5.10–7.03) | 6.21 (5.36–8.28) | <0.001 |
| Hemoglobin, g/L | 654 (479/175) | 130.00 (120.00–139.00) | 132.00 (123.50–140.00) | 0.09 |
| HbA1c, % | 540 (388/152) | 6.29 (5.72–7.45) | 6.46 (5.73–7.64) | 0.29 |
| CRP, mg/L | 468 (344/124) | 1.20 (0.50–4.71) | 1.33 (0.50–3.36) | 0.53 |
| FT3, pg/mL | 650 (474/176) | 2.45 (2.25–2.82) | 2.58 (2.30–2.87) | 0.04 |
| FT4, ng/dL | 650 (474/176) | 1.10 (0.96–1.27) | 1.10 (0.98–1.31) | 0.52 |
| TSH, mIU/L | 650 (474/176) | 2.23 (1.32–3.71) | 2.25 (1.34–3.55) | 0.89 |
| WBC, ×109/L | 654 (479/175) | 6.50 (5.40–7.90) | 6.60 (5.60–7.70) | 0.61 |
| Neutrophils, ×109/L | 654 (479/175) | 4.07 (3.27–5.09) | 4.19 (3.43–5.13) | 0.33 |
| Neutrophils, % | 654 (479/175) | 62.90 (57.50–68.60) | 63.50 (58.40–68.90) | 0.31 |
| Lymphocytes, ×109/L | 654 (479/175) | 1.73 (1.41–2.17) | 1.67 (1.40–2.08) | 0.43 |
| Monocytes, ×109/L | 654 (479/175) | 0.43 (0.35–0.54) | 0.44 (0.38–0.54) | 0.57 |
| Platelets, ×109/L | 653 (479/174) | 211.00 (180.00–247.00) | 208.00 (178.25–243.75) | 0.53 |
| Total cholesterol, mmol/L | 664 (484/180) | 3.88 (3.26–4.61) | 3.81 (3.23–4.47) | 0.33 |
| Triglycerides, mmol/L | 664 (484/180) | 1.42 (1.09–1.88) | 1.52 (1.21–1.96) | 0.046 |
| HDL-C, mmol/L | 664 (484/180) | 0.95 (0.80–1.12) | 0.91 (0.77–1.08) | 0.04 |
| LDL-C, mmol/L | 664 (484/180) | 2.27 (1.78–2.84) | 2.17 (1.74–2.70) | 0.26 |
| Albumin, g/L | 452 (333/119) | 37.40 (35.80–39.20) | 37.70 (35.75–39.55) | 0.82 |
| Total bilirubin, μmol/L | 662 (482/180) | 8.45 (6.40–11.10) | 9.35 (7.07–12.00) | 0.02 |
| NT-proBNP, pg/mL | 662 (482/180) | 252.80 (93.68–619.75) | 286.55 (116.90–703.17) | 0.13 |
| hs-TnT, ng/L | 664 (484/180) | 15.00 (9.00–33.00) | 15.50 (10.00–40.25) | 0.33 |
| Preoperative medication | ||||
| Statin | 664 (484/180) | 199 (41.12) | 74 (41.11) | >0.99 |
| Beta-blocker | 664 (484/180) | 450 (92.98) | 162 (90.00) | 0.21 |
| ACE-I/ARB | 664 (484/180) | 81 (16.74) | 34 (18.89) | 0.52 |
| Calcium channel blocker | 664 (484/180) | 273 (56.40) | 106 (58.89) | 0.57 |
| Echocardiographic parameters | ||||
| LVEF, fraction | 664 (484/180) | 0.56 (0.52–0.58) | 0.55 (0.52–0.58) | 0.24 |
| Fractional shortening, fraction | 664 (484/180) | 0.29 (0.27–0.30) | 0.29 (0.27–0.30) | 0.18 |
| Left atrial diameter, mm | 664 (484/180) | 36.00 (34.00–38.00) | 37.00 (35.00–40.00) | <0.001 |
| SPAP, mmHg | 664 (484/180) | 35.00 (34.00–38.00) | 36.00 (34.00–38.00) | 0.25 |
| LVEDV, mL | 664 (484/180) | 97.00 (83.00–110.25) | 100.50 (87.00–123.25) | 0.007 |
| Insulin-resistance-related indices | ||||
| TyG index | 664 (484/180) | 8.86 (8.53–9.18) | 9.06 (8.65–9.48) | <0.001 |
| TyG-BMI index | 664 (484/180) | 217.46 (198.81–235.37) | 224.28 (206.81–245.52) | <0.001 |
Data are presented as mean ± standard deviation, median (interquartile range), or n (%), as appropriate. Available (n) is shown as total (non-POAF/POAF). Welch’s t-test was used for variables summarized as mean ± standard deviation; the Mann-Whitney U test was used for variables summarized as median (interquartile range); and the Chi-squared test or Fisher’s exact test was used for categorical variables. Percentages were calculated using available observations. NYHA functional class was unrecorded for four participants, all in the non-POAF group; NYHA percentages and the group comparison use the 660 participants with recorded classes I–IV. ACE-I, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; BMI, body mass index; CRP, C-reactive protein; DBP, diastolic blood pressure; FT3, free triiodothyronine; FT4, free thyroxine; HbA1c, glycated hemoglobin; HDL-C, high-density lipoprotein cholesterol; hs-TnT, high-sensitivity troponin T; IABP, intra-aortic balloon pump; LDL-C, low-density lipoprotein cholesterol; LVEDV, left ventricular end-diastolic volume; LVEF, left ventricular ejection fraction; NT-proBNP, N-terminal pro-B-type natriuretic peptide; NYHA, New York Heart Association; PCI, percutaneous coronary intervention; POAF, postoperative atrial fibrillation; SBP, systolic blood pressure; SPAP, systolic pulmonary artery pressure; TSH, thyroid-stimulating hormone; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose-body mass index; WBC, white blood cell.
We examined the associations of preoperative TyG and TyG-BMI with POAF using logistic regression. For each index, four models were constructed: Model 1 was unadjusted; Model 2 was adjusted for age and sex; Model 3 was additionally adjusted for hypertension, diabetes mellitus, current smoking, current drinking, creatinine, heart failure, preoperative beta-blocker use, and statin use, with BMI additionally included only in the TyG model; and Model 4 further included LAD and LVEF. Results are reported as odds ratios (ORs) with 95% confidence intervals (CIs). Each index was analyzed per one standard deviation (SD) increase and as a categorical variable using tertiles and the optimal cut-off derived from receiver operating characteristic (ROC) analysis. Models 1 and 2 included 664 participants, whereas Models 3 and 4 included 663 complete cases because creatinine was missing for one participant with POAF. No missing covariate values were imputed.
To explore potential nonlinear relationships between the continuous indices and POAF, we performed RCS analyses within the fully adjusted logistic regression framework. Four knots were placed at the 5th, 35th, 65th, and 95th percentiles, and the cohort median of each index was used as the reference. Tests for the overall association and nonlinearity were performed.
We evaluated whether adding either index improved the predictive performance of a common baseline clinical model that included age, sex, BMI, hypertension, diabetes mellitus, current smoking, current drinking, creatinine, heart failure, preoperative beta-blocker use, statin use, LAD, and LVEF. TyG or TyG-BMI was added separately as a continuous standardized variable. Changes in the C-statistic, integrated discrimination improvement (IDI), and continuous net reclassification improvement (NRI) were estimated using 2,000 paired nonparametric bootstrap resamples. BMI was retained in the common baseline model so that the incremental value of TyG-BMI was evaluated beyond BMI alone. Collinearity was assessed using variance inflation factors (VIFs). The stability and internal validity of the fully adjusted association models were further examined with 1,000 bootstrap resamples, including coefficient stability, optimism-corrected AUC and Brier score, and the optimism-corrected calibration slope. Diabetes status was evaluated as the primary subgroup of interest using formal product terms. Age (<65 vs. ≥65 years) and sex subgroups were exploratory, and their interaction P values were adjusted using the Benjamini-Hochberg false discovery rate method. All tests were two-sided, and P<0.05 was considered statistically significant.
Data analysis was performed using SPSS version 27.0 (IBM Corp) and R version 4.4.2 (R Foundation for Statistical Computing). Relevant R packages included readxl, ggplot2, rms, Hmisc, pROC, dplyr, and patchwork.
Results
Baseline characteristics
We initially screened 1,275 patients who underwent CABG, including on-pump CABG and OPCABG, at our institution. After excluding 518 patients according to the initial exclusion criteria, including missing essential clinical data, 757 patients comprised the eligible source cohort. Of these, 93 lacked preoperative fasting glucose and/or triglyceride measurements required to calculate TyG and TyG-BMI, leaving 664 participants in the primary analysis (Figure S1). Among the 664 participants, 180 (27.1%) experienced POAF. Table 1 summarizes the baseline characteristics stratified by POAF occurrence.
Patients who developed POAF were older and had higher BMI and a higher prevalence of current drinking than those who did not (all P<0.05; Table 1). The POAF group also had higher glucose, FT3, triglyceride, total bilirubin, LAD, and LVEDV values and lower HDL-C values (all P<0.05). Other baseline characteristics, including sex, hypertension, diabetes mellitus, and LVEF, were comparable between the groups. Both TyG and TyG-BMI were higher in patients who developed POAF than in those who maintained sinus rhythm (both P<0.001; Table 1).
We initially assessed the discriminative capacity of each index for POAF using ROC curve analysis (Figure S2). TyG had an AUC of 0.593 (95% CI: 0.543–0.643), and TyG-BMI had an AUC of 0.599 (95% CI: 0.549–0.649), indicating limited discrimination when used alone. The optimal cut-offs based on the maximum Youden index were 9.131 for TyG (sensitivity, 46.1%; specificity, 72.3%) and 243.833 for TyG-BMI (sensitivity, 27.8%; specificity, 86.6%) (Table S2).
Association between the TyG index and POAF
We further investigated the relationship between TyG and POAF using univariable (Table S1) and multivariable (Table 2) logistic regression analyses. In Model 1, each SD increase in TyG was associated with higher odds of POAF (OR, 1.43; 95% CI: 1.20–1.70; P<0.001). The association remained significant in the fully adjusted Model 4 (OR, 1.55; 95% CI: 1.28–1.89; P<0.001). Findings were consistent when TyG was dichotomized at 9.131 (Model 4 OR, 2.56; 95% CI: 1.71–3.82; P<0.001). When categorized into tertiles, participants in the highest tertile (>9.124) had higher odds of POAF than those in the lowest tertile (<8.700) (Model 4 OR, 2.46; 95% CI: 1.54–3.93; P<0.001), with a significant trend across tertiles (P for trend <0.001).
Table 2. Associations of TyG and TyG-BMI indices with postoperative atrial fibrillation.
| Exposure | Model 1 | Model 2 | Model 3 | Model 4 | ||||
|---|---|---|---|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | |
| TyG index (per SD increase) | 1.43 (1.20–1.70) | <0.001 | 1.55 (1.30–1.86) | <0.001 | 1.56 (1.29–1.90) | <0.001 | 1.55 (1.28–1.89) | <0.001 |
| <9.131 | Reference | Reference | Reference | Reference | ||||
| ≥9.131 | 2.19 (1.53–3.12) | <0.001 | 2.51 (1.73–3.63) | <0.001 | 2.51 (1.69–3.73) | <0.001 | 2.56 (1.71–3.82) | <0.001 |
| Tertile 1 (<8.700) | Reference | Reference | Reference | Reference | ||||
| Tertile 2 (8.700–9.124) | 0.85 (0.54–1.33) | 0.47 | 0.91 (0.58–1.44) | 0.69 | 1.00 (0.62–1.60) | >0.99 | 0.99 (0.61–1.59) | 0.97 |
| Tertile 3 (>9.124) | 1.98 (1.31–2.99) | 0.001 | 2.35 (1.53–3.60) | <0.001 | 2.45 (1.54–3.89) | <0.001 | 2.46 (1.54–3.93) | <0.001 |
| P for trend | <0.001 | <0.001 | <0.001 | <0.001 | ||||
| TyG-BMI index (per SD increase) | 1.54 (1.28–1.84) | <0.001 | 1.66 (1.37–2.01) | <0.001 | 1.71 (1.41–2.09) | <0.001 | 1.60 (1.31–1.96) | <0.001 |
| <243.833 | Reference | Reference | Reference | Reference | ||||
| ≥243.833 | 2.41 (1.59–3.67) | <0.001 | 2.97 (1.91–4.62) | <0.001 | 3.10 (1.96–4.89) | <0.001 | 2.78 (1.74–4.42) | <0.001 |
| Tertile 1 (<206.845) | Reference | Reference | Reference | Reference | ||||
| Tertile 2 (206.845–232.205) | 1.36 (0.88–2.12) | 0.17 | 1.39 (0.89–2.18) | 0.15 | 1.41 (0.89–2.24) | 0.14 | 1.38 (0.87–2.20) | 0.17 |
| Tertile 3 (>232.205) | 2.01 (1.31–3.07) | 0.001 | 2.28 (1.47–3.55) | <0.001 | 2.46 (1.56–3.86) | <0.001 | 2.13 (1.34–3.38) | 0.001 |
| P for trend | 0.001 | <0.001 | <0.001 | 0.001 | ||||
Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: additionally adjusted for hypertension, diabetes mellitus, current smoking, current drinking, creatinine, heart failure, preoperative beta-blocker use, and statin use; BMI was additionally included only in the TyG model. Model 4: Model 3 plus left atrial diameter and LVEF. TyG and TyG-BMI were standardized using the cohort-specific SD. Cutoffs were selected by the maximum Youden index. Tertile boundaries were calculated separately for each index. Models 1–2 included 664 participants (180 events); Models 3–4 included 663 participants (179 events) because creatinine was missing for one participant with POAF. BMI, body mass index; CI, confidence interval; LVEF, left ventricular ejection fraction; OR, odds ratio; POAF, postoperative atrial fibrillation; SD, standard deviation; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose-body mass index.
RCS analysis showed an approximately linear increase in POAF odds with increasing TyG (P for overall association <0.001; P for nonlinearity =0.16; Figure 1A). The median TyG value of 8.89 was used as the reference. These findings support a positive association between TyG and POAF without clear evidence of nonlinearity.
Figure 1.
Restricted cubic spline associations of the TyG and TyG-BMI indices with postoperative atrial fibrillation. (A) TyG index. The median TyG value of 8.89 was used as the reference (OR =1); P for overall association <0.001 and P for nonlinearity =0.16. (B) TyG-BMI index. The median TyG-BMI value of 218.75 was used as the reference (OR =1); P for overall association <0.001 and P for nonlinearity =0.02. Four knots were placed at the 5th, 35th, 65th, and 95th percentiles. Curves show adjusted ORs with 95% CIs. Both models were adjusted for age, sex, hypertension, diabetes mellitus, current smoking, current drinking, creatinine, heart failure, preoperative beta-blocker use, statin use, LAD, and LVEF; the TyG model was additionally adjusted for BMI. BMI, body mass index; CI, confidence interval; LAD, left atrial diameter; LVEF, left ventricular ejection fraction; OR, odds ratio; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose-body mass index.
Association between the TyG-BMI index and POAF
A parallel analysis was conducted for TyG-BMI (Table 2). Each SD increase in TyG-BMI was associated with higher odds of POAF in Model 1 (OR, 1.54; 95% CI: 1.28–1.84; P<0.001) and in the fully adjusted Model 4 (OR, 1.60; 95% CI: 1.31–1.96; P<0.001). Findings were consistent when TyG-BMI was dichotomized at 243.833 (Model 4 OR, 2.78; 95% CI: 1.74–4.42; P<0.001). Participants in the highest tertile (>232.205) had higher odds of POAF than those in the lowest tertile (<206.845) (Model 4 OR, 2.13; 95% CI: 1.34–3.38; P=0.001), with a significant trend across tertiles (P for trend =0.001).
The RCS curve for TyG-BMI indicated an overall association with POAF (P<0.001) and evidence of nonlinearity (P for nonlinearity =0.02; Figure 1B). With the median value of 218.75 as the reference, POAF odds were relatively stable across lower values and increased more steeply at higher TyG-BMI values. This pattern should be interpreted as a nonlinear dose-response association rather than as proof of a single clinical threshold.
Incremental predictive value of TyG and TyG-BMI
To evaluate incremental predictive value, we added TyG or TyG-BMI separately to the prespecified baseline clinical model (Table 3). The baseline model had a C-statistic of 0.682 (95% CI: 0.635–0.726). Addition of TyG increased the C-statistic to 0.708 (95% CI: 0.663–0.750), and addition of TyG-BMI increased it to 0.707 (95% CI: 0.662–0.750); neither paired increase reached statistical significance (P=0.07 and P=0.08, respectively).
Table 3. Incremental predictive value of TyG and TyG-BMI indices for postoperative atrial fibrillation.
| Model | C-statistic (95% CI) | P value | IDI (95% CI) | P value | Continuous NRI (95% CI) | P value |
|---|---|---|---|---|---|---|
| Baseline risk model | 0.682 (0.635–0.726) | – | Reference | – | Reference | – |
| + TyG index | 0.708 (0.663–0.750) | 0.07 | 0.033 (0.018–0.048) | <0.001 | 0.359 (0.191–0.521) | <0.001 |
| + TyG-BMI index | 0.707 (0.662–0.750) | 0.08 | 0.034 (0.020–0.050) | <0.001 | 0.346 (0.178–0.514) | <0.001 |
The baseline risk model included age, sex, BMI, hypertension, diabetes mellitus, current smoking, current drinking, creatinine, heart failure, preoperative beta-blocker use, statin use, left atrial diameter, and LVEF. TyG or TyG-BMI was added separately as a continuous standardized variable. All models used the same 663 complete cases (179 POAF events). C-statistic confidence intervals and paired comparisons, IDI, and continuous NRI were estimated using 2,000 paired nonparametric bootstrap resamples of participants. P values compare each expanded model with the baseline model. BMI was retained in the common baseline model so that the incremental value of TyG-BMI was evaluated beyond BMI alone. BMI, body mass index; CI, confidence interval; IDI, integrated discrimination improvement; LVEF, left ventricular ejection fraction; NRI, net reclassification improvement; POAF, postoperative atrial fibrillation; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose-body mass index.
Both indices improved reclassification measures. For TyG, the IDI was 0.033 (95% CI: 0.018–0.048; P<0.001) and the continuous NRI was 0.359 (95% CI: 0.191–0.521; P<0.001). For TyG-BMI, the IDI was 0.034 (95% CI: 0.020–0.050; P<0.001) and the continuous NRI was 0.346 (95% CI: 0.178–0.514; P<0.001) (Table 3).
Collinearity was low in both fully adjusted association models (maximum VIF, 1.32 for the TyG model and 1.29 for the TyG-BMI model; Table S3). In 1,000 bootstrap resamples, the exposure estimates remained directionally consistent in all successful resamples, with median ORs of 1.57 for TyG and 1.63 for TyG-BMI. The optimism-corrected AUCs were 0.676 and 0.668, and the optimism-corrected calibration slopes were 0.829 and 0.827, respectively. These findings support coefficient stability but also indicate some optimism in apparent model performance.
Subgroup analyses
Diabetes status was assessed as the primary subgroup of interest because of its close relationship with IR. Age and sex subgroup analyses were performed as exploratory analyses (Table 4 and Table S4). Associations were estimated per SD increase in each index, and effect modification was assessed using formal interaction terms.
Table 4. Associations of TyG and TyG-BMI indices with postoperative atrial fibrillation according to diabetes status.
| Exposure | Diabetes status | Participants/events | Adjusted OR (95% CI) | P value | P for interaction |
|---|---|---|---|---|---|
| TyG index (per SD) | No diabetes | 388/99 | 1.22 (0.93–1.60) | 0.16 | 0.02 |
| Diabetes | 275/80 | 2.20 (1.57–3.08) | <0.001 | ||
| TyG-BMI index (per SD) | No diabetes | 388/99 | 1.21 (0.92–1.60) | 0.18 | 0.005 |
| Diabetes | 275/80 | 2.29 (1.63–3.22) | <0.001 |
ORs are reported per 1-SD increase in the index. Models were adjusted for age, sex, hypertension, current smoking, current drinking, creatinine, heart failure, preoperative beta-blocker use, statin use, left atrial diameter, and LVEF; the TyG model was additionally adjusted for BMI. Diabetes mellitus, the stratification variable, was omitted from within-stratum models. Interaction P values were obtained from models containing the index, diabetes status, and their product term. Analyses used complete cases without imputation. BMI, body mass index; CI, confidence interval; LVEF, left ventricular ejection fraction; OR, odds ratio; SD, standard deviation; TyG, triglyceride-glucose; TyG-BMI, triglyceride-glucose-body mass index.
Among participants with diabetes, TyG (OR, 2.20; 95% CI: 1.57–3.08; P<0.001) and TyG-BMI (OR, 2.29; 95% CI: 1.63–3.22; P<0.001) were strongly associated with POAF. Corresponding associations were weaker and not statistically significant among participants without diabetes (TyG: OR, 1.22; 95% CI: 0.93–1.60; P=0.16; TyG-BMI: OR, 1.21; 95% CI: 0.92–1.60; P=0.18). The interaction P values were 0.02 for TyG and 0.005 for TyG-BMI (Table 4).
In exploratory age-stratified analyses, both indices were positively associated with POAF in participants aged <65 years and those aged ≥65 years. However, the interactions with age were not statistically significant (P for interaction =0.13 for TyG and 0.37 for TyG-BMI; Table S4).
Both indices were also positively associated with POAF in women and men. There was no evidence of interaction by sex (P for interaction =0.16 for TyG and 0.94 for TyG-BMI; Table S4). None of the four exploratory age or sex interaction tests remained significant after false discovery rate adjustment.
Collectively, these results suggest that diabetes may modify the associations of TyG and TyG-BMI with POAF. In contrast, the data did not support effect modification by age or sex, and these exploratory subgroup findings should be interpreted cautiously.
Discussion
This study demonstrates that higher preoperative TyG and TyG-BMI values are independently associated with POAF in patients undergoing OPCABG. The associations were more pronounced among patients with diabetes. A further finding was the nonlinear association between TyG-BMI and POAF, with risk increasing more steeply at higher values. The two indices also improved reclassification measures when added to conventional clinical factors, although the changes in the C-statistic were not statistically significant.
IR represents a central pathophysiological feature of metabolic disorders and a fundamental mechanism in type 2 diabetes (18). Abnormalities in glucose and lipid metabolism driven by IR are strongly associated with adverse cardiovascular outcomes (19). Recent evidence continues to support the clinical relevance of IR surrogates such as the TyG index. For example, one study identified the TyG index as a reliable predictor of all-cause mortality in patients with cardiometabolic multimorbidity, highlighting its utility in guiding management strategies for metabolically compromised patients (20). Another investigation demonstrated that individuals in the highest TyG quartile had significantly greater odds of aortic valve calcification, suggesting that the TyG index may serve as a marker for this pathology (21). In populations with cardiovascular-kidney-metabolic syndrome stages 0-3, TyG-BMI has been associated with all-cause and cardiovascular mortality, whereas TyG was associated with cardiovascular mortality but not all-cause mortality (22). Furthermore, TyG-BMI has shown nonlinear associations with all-cause mortality and heart failure rehospitalization in patients with coronary heart disease and heart failure (23).
Accumulating evidence indicates that IR indices, particularly the TyG index and TyG-BMI, are associated with diverse adverse cardiovascular outcomes, including atrial fibrillation (AF). Previous research has established a strong connection between IR and AF incidence, with the TyG index being extensively investigated in AF populations (24). For instance, one study demonstrated that both TyG and TyG-BMI served as independent predictors of AF recurrence following radiofrequency ablation, with TyG-BMI providing additional incremental value, particularly among diabetic patients (25). Furthermore, a recent investigation suggested that an elevated TyG index effectively predicts new-onset POAF after OPCABG (26), a finding consistent with our results. The TyG index has also been correlated with thromboembolic risk in AF patients (10). Thus, the relationship between IR and AF is both biologically plausible and supported by accumulating scientific literature. Our study strengthens this evidence base by specifically demonstrating this association in the context of OPCABG, a surgical population that has been relatively understudied in this regard.
Although TyG and TyG-BMI individually demonstrated only modest discriminatory ability, both were significantly associated with POAF. TyG-BMI showed high specificity but low sensitivity at its ROC-derived cut-off, whereas TyG provided a more balanced sensitivity-specificity profile. Accordingly, neither index should be used as a stand-alone diagnostic test. Their principal value may lie in providing additional metabolic information when considered alongside established clinical and echocardiographic factors.
Patients who developed POAF exhibited higher preoperative TyG and TyG-BMI values than those who did not, and the associations persisted after comprehensive adjustment for potential confounders. The RCS analysis for TyG-BMI indicated a nonlinear relationship, with POAF odds increasing more steeply at higher values. Although this pattern may help identify patients with greater metabolic risk, the curve does not establish a validated treatment threshold. Any threshold-based clinical application would require confirmation in independent cohorts.
Adding TyG or TyG-BMI to the baseline clinical model improved IDI and continuous NRI, whereas the increases in the C-statistic did not reach statistical significance. These complementary findings indicate improved separation and reclassification of predicted risks without conclusive evidence of better overall discrimination. The associations were stronger among patients with diabetes, which is biologically plausible because diabetes represents a chronic state of IR and hyperglycemia that may promote atrial electrophysiological remodeling and create a substrate for arrhythmogenesis. The significant interaction terms support diabetes as a potential effect modifier, although this finding requires external confirmation.
In exploratory analyses, the positive associations were observed across age and sex strata, but neither age nor sex significantly modified the associations. The absence of interaction should not be interpreted as proof that the effects are identical across subgroups, particularly given the smaller number of women and the limited power of interaction tests. Larger studies are required for age- and sex-specific evaluation.
From a clinical perspective, TyG and TyG-BMI are practical candidate risk markers because they are derived from routinely measured fasting glucose, triglycerides, and BMI. They may complement standard preoperative assessment and help identify patients who warrant closer postoperative rhythm surveillance. However, given their modest stand-alone discrimination and the lack of external validation, the present results do not support selecting prophylactic drug treatment solely on the basis of either index.
Our study has several limitations. First, as a retrospective observational investigation, it can establish associations but cannot demonstrate that IR directly causes POAF. Second, residual confounding remains possible despite adjustment for prespecified clinical factors. Third, 93 of the 757 otherwise eligible patients lacked the preoperative fasting glucose and/or triglyceride measurements required to calculate TyG and TyG-BMI and were excluded from the primary analysis. Because these measurements were unavailable because they had not been obtained before surgery, rather than being set to zero or imputed, complete-case selection may have introduced selection bias. Fourth, this was a single-center study, and measurement inaccuracies or undocumented clinical factors inherent to retrospective records may have affected the findings. Fifth, the bootstrap analysis provided internal validation only; the optimism-corrected discrimination and calibration estimates indicate that external validation and potential model recalibration are required. Finally, subgroup analyses, particularly those among women, had limited statistical power, and the diabetes interaction finding should be regarded as requiring confirmation.
Conclusions
In conclusion, higher preoperative TyG and TyG-BMI values were independently associated with POAF in patients undergoing OPCABG. Both indices provided modest incremental risk information beyond conventional clinical factors, and their associations were stronger among patients with diabetes. Because discrimination remained moderate and validation was internal, these readily available indices should be considered candidate adjunctive markers rather than stand-alone clinical prediction tools.
Supplementary
The article’s supplementary files as
Acknowledgments
We would like to express our sincere gratitude to all the individuals and departments who contributed to this study. Specifically, we thank the staff of the Department of Cardiovascular Surgery at General Hospital of Northern Theater Command for their support in data collection. We also appreciate the valuable support from the Ethics Committee of the General Hospital of Northern Theater Command in approving this study. Lastly, we would like to express our deepest gratitude to all the patients whose data were included in this research.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of General Hospital of Northern Theater Command [approval No. Y(2025)323]. Informed consent was waived in this retrospective study.
Footnotes
Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1640/rc
Funding: This work was supported by grants from the Liaoning Revitalization Talents Program (No. XLYC2203117) and National Scientific Research Foundation of Liaoning Province (No. 2025-MSLH-726).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1640/coif). J.Z. reports grants from the Liaoning Revitalization Talents Program and National Scientific Research Foundation of Liaoning Province, both paid to his institution. The other authors have no conflicts of interest to declare.
Data Sharing Statement
Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1640/dss
References
- 1. Li R, Zhang J, Ke L, et al. Association of epicardial adipose tissue density with postoperative atrial fibrillation after isolated aortic valve replacement. Int J Cardiol Heart Vasc 2024;54:101481.. 10.1016/j.ijcha.2024.101481 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Maesen B, Nijs J, Maessen J, et al. Post-operative atrial fibrillation: a maze of mechanisms. Europace 2012;14:159-74.. 10.1093/europace/eur208 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Wang Y, He Y, Li R, et al. Gut Microbiota in Patients with Postoperative Atrial Fibrillation Undergoing Off-Pump Coronary Bypass Graft Surgery. J Clin Med 2023;12:1493.. 10.3390/jcm12041493 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Zhang J, Wang Y, Jiang H, et al. Preventive Effect of Berberine on Postoperative Atrial Fibrillation. Circ Arrhythm Electrophysiol 2022;15:e011160.. 10.1161/CIRCEP.122.011160 [DOI] [PubMed] [Google Scholar]
- 5. Lomivorotov VV, Efremov SM, Pokushalov EA, et al. New-Onset Atrial Fibrillation After Cardiac Surgery: Pathophysiology, Prophylaxis, and Treatment. J Cardiothorac Vasc Anesth 2016;30:200-16.. 10.1053/j.jvca.2015.08.003 [DOI] [PubMed] [Google Scholar]
- 6. Todorov H, Janssen I, Honndorf S, et al. Clinical significance and risk factors for new onset and recurring atrial fibrillation following cardiac surgery - a retrospective data analysis. BMC Anesthesiol 2017;17:163.. 10.1186/s12871-017-0455-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Yang Y, Du Z, Fang M, et al. Metabolic signatures in pericardial fluid and serum are associated with new-onset atrial fibrillation after isolated coronary artery bypass grafting. Transl Res 2023;256:30-40.. 10.1016/j.trsl.2023.01.001 [DOI] [PubMed] [Google Scholar]
- 8. Echahidi N, Mohty D, Pibarot P, et al. Obesity and metabolic syndrome are independent risk factors for atrial fibrillation after coronary artery bypass graft surgery. Circulation 2007;116:I213-9.. 10.1161/CIRCULATIONAHA.106.681304 [DOI] [PubMed] [Google Scholar]
- 9. Krishna VR, Patil N, Nileshwar A. Prospective evaluation of the utility of CHA2DS2-VASc score in the prediction of postoperative atrial fibrillation after off-pump coronary artery bypass surgery - An observational study. Ann Card Anaesth 2020;23:122-6.. 10.4103/aca.ACA_161_18 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Rivera-Caravaca JM, Soler-Espejo E, Ramos-Bratos MP, et al. The triglyceride-glucose index, a marker of insulin resistance, as a predictor of thrombotic risk in atrial fibrillation. Cardiovasc Diabetol 2025;24:281.. 10.1186/s12933-025-02809-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Boden G. Obesity, insulin resistance and free fatty acids. Curr Opin Endocrinol Diabetes Obes 2011;18:139-43.. 10.1097/MED.0b013e3283444b09 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Er LK, Wu S, Chou HH, et al. Triglyceride Glucose-Body Mass Index Is a Simple and Clinically Useful Surrogate Marker for Insulin Resistance in Nondiabetic Individuals. PLoS One 2016;11:e0149731.. 10.1371/journal.pone.0149731 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Lim J, Kim J, Koo SH, et al. Comparison of triglyceride glucose index, and related parameters to predict insulin resistance in Korean adults: An analysis of the 2007-2010 Korean National Health and Nutrition Examination Survey. PLoS One 2019;14:e0212963.. 10.1371/journal.pone.0212963 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Cheng Y, Fang Z, Zhang X, et al. Association between triglyceride glucose-body mass index and cardiovascular outcomes in patients undergoing percutaneous coronary intervention: a retrospective study. Cardiovasc Diabetol 2023;22:75.. 10.1186/s12933-023-01794-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Yang X, Li K, Wen J, et al. Association of the triglyceride glucose-body mass index with the extent of coronary artery disease in patients with acute coronary syndromes. Cardiovasc Diabetol 2024;23:24.. 10.1186/s12933-024-02124-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Zhang J, He Y, Yin Z, et al. Circulating neuropeptide Y as a biomarker in postoperative atrial fibrillation cases administered off-pump coronary bypass Graft surgery. Heliyon 2024;10:e31251.. 10.1016/j.heliyon.2024.e31251 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Wang H, Zhang Y, Xin F, et al. Calcium-Induced Autonomic Denervation in Patients With Post-Operative Atrial Fibrillation. J Am Coll Cardiol 2021;77:57-67.. 10.1016/j.jacc.2020.10.049 [DOI] [PubMed] [Google Scholar]
- 18. Sampath Kumar A, Maiya AG, Shastry BA, et al. Exercise and insulin resistance in type 2 diabetes mellitus: A systematic review and meta-analysis. Ann Phys Rehabil Med 2019;62:98-103.. 10.1016/j.rehab.2018.11.001 [DOI] [PubMed] [Google Scholar]
- 19. Gao X, Chen T, Zhou F, et al. The association between different insulin resistance surrogates and all-cause mortality and cardiovascular mortality in patients with metabolic dysfunction-associated steatotic liver disease. Cardiovasc Diabetol 2025;24:200.. 10.1186/s12933-025-02758-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Zhang H, Tu Z, Liu S, et al. Association of different insulin resistance surrogates with all-cause and cardiovascular mortality among the population with cardiometabolic multimorbidity. Cardiovasc Diabetol 2025;24:33.. 10.1186/s12933-025-02576-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Wang P, Zeng Y, Wang L, et al. Association of TyG index with aortic valve calcification in valvular heart disease patients. Postgrad Med J 2024;100:917-24.. 10.1093/postmj/qgae085 [DOI] [PubMed] [Google Scholar]
- 22. Zhang P, Mo D, Zeng W, et al. Association between triglyceride-glucose related indices and all-cause and cardiovascular mortality among the population with cardiovascular-kidney-metabolic syndrome stage 0-3: a cohort study. Cardiovasc Diabetol 2025;24:92.. 10.1186/s12933-025-02642-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Lyu L, Wang X, Xu J, et al. Association between triglyceride glucose-body mass index and long-term adverse outcomes of heart failure patients with coronary heart disease. Cardiovasc Diabetol 2024;23:162.. 10.1186/s12933-024-02213-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Azarboo A, Behnoush AH, Vaziri Z, et al. Assessing the association between triglyceride-glucose index and atrial fibrillation: a systematic review and meta-analysis. Eur J Med Res 2024;29:118.. 10.1186/s40001-024-01716-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Wang Z, He H, Xie Y, et al. Non-insulin-based insulin resistance indexes in predicting atrial fibrillation recurrence following ablation: a retrospective study. Cardiovasc Diabetol 2024;23:87.. 10.1186/s12933-024-02158-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Peng Z, Zhao R, Yang Y, et al. Predictive Value of the CT-Based Visceral Adiposity Tissue Index and Triglyceride-Glucose Index on New-Onset Atrial Fibrillation after Off-Pump Coronary Artery Bypass Graft: Analyses from a Longitudinal Study. Rev Cardiovasc Med 2023;24:338.. 10.31083/j.rcm2411338 [DOI] [PMC free article] [PubMed] [Google Scholar]