Perivascular fat attenuation index on coronary computed tomography angiography predicts postoperative atrial fibrillation after off-pump coronary artery bypass grafting Authors: Xiaodong Chen, Haoyan Li, Haoyu Hu, Dong Wang, Wenlong Yan, Pei Nie, Sumin Yang Published: 2026 ABSTRACT AbstractBackgroundPostoperative atrial fibrillation (POAF) is a prevalent complication following cardiac surgery, which has been associated with inflammatory activity in epicardial adipose tissue. There is a lack of effective means for its prediction and prevention. The fat attenuation index (FAI) serves as a non-invasive imaging biomarker capable of quantifying pericoronary adipose tissue inflammation.MethodsIn this retrospective study, patients who underwent off-pump coronary artery bypass grafting (OPCABG) and had preoperative coronary computed tomography angiography (CCTA) were included. The pericoronary FAI was measured from CCTA images. POAF was identified based on ECG monitoring within 7 days after surgery. A multivariate logistic regression model incorporating variables significant in univariate analysis and clinically relevant factors was constructed. Model performance with and without FAI was compared using the area under the curve (AUC) with DeLong’s test.ResultsA total of 140 patients were finally included, including 41 patients (29.2%) in the POAF group and 99 patients (70.8%) in the NO-POAF group Multivariate logistic regression identified hypertension and a higher FAI around the right coronary artery as independent predictors of POAF, while history of myocardial infarction showed a trend toward significance. The inclusion of FAI (RCA) significantly improved model discrimination (AUC increased from 0.711 for the clinical model to 0.752; DeLong p = 0.049).ConclusionsThe pericoronary FAI, particularly around the RCA, provides incremental predictive information and may help identify patients at elevated POAF risk; however, its precise clinical utility requires further validation.Graphical abstract Perivascular fat attenuation index on coronary computed tomography angiography predicts postoperative atrial fibrillation after off‑pump coronary artery bypass grafting. FAI = fat attenuation index; RCA = right coronary artery; POAF = postoperative atrial fibrillation; OPCABG = off‑pump coronary artery bypass grafting Supplementary InformationThe online version contains supplementary material available at 10.1186/s12872-026-06110-2.Keywords: Fat attenuation index, Postoperative atrial fibrillation, Coronary computed tomography angiography, Off-pump coronary artery bypass graft, Atherosclerotic coronary artery disease Introduction Atherosclerotic coronary artery disease is the most prevalent disorder of the coronary system in clinical practice. Patients with severe coronary artery disease often require surgical revascularization, with off-pump coronary artery bypass grafting (OPCABG)being one of the most predominant revascularization procedures [1, 2]. The introduction of OPCABG has circumvented the adverse effects associated with cardiopulmonary bypass, thereby reducing the degree of intervention on the patient’s cardiovascular system during surgery. Numerous studies have demonstrated that OPCABG significantly reduces the incidence of short-term stroke, early mortality, and the risk of acute kidney injury [3–5]. However, this technique is considerably more challenging and is associated with a relatively higher rate of repeat revascularization [6–8], alongside the persistence of various postoperative complications [9]. Postoperative atrial fibrillation (POAF) is one of the most common complications following cardiac surgery [10], representing a distinct category of atrial fibrillation. It is defined as the new onset of atrial fibrillation in patients with no prior history of arrhythmia before undergoing cardiac surgery. It affects approximately 20–40% [11] of patients after cardiac surgery, typically manifesting between the second and fourth postoperative days. Although the incidence of new-onset atrial fibrillation after coronary artery bypass grafting is relatively low, it still occurs in approximately 20% of patients [12]. Studies have established its association with adverse outcomes, including hemodynamic instability, increased risk of stroke [13, 14], prolonged hospital and intensive care unit stays, and elevated healthcare costs. Epicardial adipose tissue has been demonstrated to be a source of pro-inflammatory paracrine effects that increase the risk and severity of AF [15]. Studies have indicated that patients who develop POAF after coronary artery bypass grafting exhibit distinct proteomic and gene expression profiles in their epicardial adipose tissue compared to those who do not develop POAF. Proteins involved in inflammatory processes and ion channel function are particularly associated with the pathogenesis of POAF [16]. Nevertheless, there remains a lack of clinical imaging modalities for assessing peri-atrial fat metabolism in these patients.When POAF occurs, the treatment primarily relies on medications such as amiodarone, beta-blockers, and digoxin [17]. However, the use of these drugs often exposes patients to their associated side effects. Therefore, effective preventive measures are crucial in reducing the incidence of POAF. Fat attenuation index (FAI) has emerged as a novel, non-invasive imaging biomarker capable of effectively detecting inflammatory activity in pericoronary adipose tissue [18, 19]. Research indicates that the FAI value ≥-70.1 Hounsfield Units (HU) is significantly associated with an increased risk of cardiovascular mortality [20]. However, while FAI has predominantly been utilized for predicting major adverse cardiovascular events [20], its potential for quantifying inflammatory responses in epicardial adipose tissue to forecast the occurrence of POAF in cardiac surgery patients remains insufficiently investigated. Considering the thin atrial wall and lack of standardized protocols, direct quantification of peri-atrial fat on routine coronary computed tomography angiography (CCTA) is technically challenging. In contrast, pericoronary FAI offers a practical and validated alternative. Coronary arteries provide consistent anatomical landmarks, enabling standardized measurements using automated software. This study aims to explore whether preoperative FAI values in patients undergoing OPCABG surgery can predict the incidence of POAF. Methods Study design and patient enrollment The study population consisted of adult patients who underwent OPCABG at the Affiliated Hospital of Qingdao University between December 2023 and February 2025. We specifically focused on patients undergoing OPCABG to minimize the confounding effects of cardiopulmonary bypass, which is a known trigger for systemic inflammation and may independently influence POAF risk [21]. We retrospectively collected the patients’ clinical characteristics from the electronic medical record system. The study was ethically approved by the Ethics Committee of the Affiliated Hospital of Qingdao University, and written informed consent was waived due to the retrospective nature of the study. (Ethics Approval Number: Grant QYFYQZLL30197). Some patients did not undergo CCTA due to financial constraints or impaired renal function. All patients undergoing CABG receive coronary angiography, and the surgical revascularization strategy is primarily based on angiographic findings. Inclusion and exclusion criteria Inclusion criteria (a) Patients who received Off-pump coronary artery bypass graft in the Affiliated Hospital of Qingdao University. (b) Accepted clinically referred CCTA evaluation. (c) Patients were more than 18 years old. Exclusion criteria (a) Previous history of arrhythmia (including atrial fibrillation or ventricular tachycardia). (b) Emergency cardiac surgery. (c) Poor image quality precluding FAI or plaque analysis due to inadequate contrast opacification. (d) Prior cardiac surgery (reoperation). (e) Concomitant valve replacement or repair surgery. (f) Severe renal dysfunction (serum creatinine level exceeding twice the upper limit of normal). (g) Presence of an implantable cardiac pacemaker. Data collection Baseline demographic characteristics and clinical data were recorded, including (a) gender; (b) age; (c) body mass index(BMI); (d) diabetes mellitus; (e) hypertension; (f) dyslipidemia; (g) current smoking; (h) history of myocardial infarction (History of myocardial infarction was defined as a documented clinical diagnosis of MI occurring more than 30 days prior to surgery); (i) troponin T; (j)preoperative medication(Beta-blockers, Statins and Aspirin); (k) left ventricular ejection fraction (LVEF); (l) NYHA class; (m) incidence of POAF; (n)occurrence time of POAF. CCTA acquisition CCTA examinations were performed on a 128-slice dual-source CT scanner (Somatom Definition Flash, Siemens Healthcare, Forchheim, Germany) or a 256-slice CT scanner (Revolution CT, GE Healthcare, Milwaukee, Wisconsin, USA). The parameters of the 128-slice dual-source CT scanner were as follows: prospective or retrospective ECG triggered mode covering 35–80% of the R-R interval, 0.28 s gantry rotation time, 2 × 64 × 0.6 mm detector collimation, a slice collimation 2 × 128 × 0.6 mm by z-flying focal spot technique, 120 kV or 100 kV tube voltage, and 400–500 mAs tube current. The protocol of the 256-slice CT scanner included one beat prospective sequential mode, 0.28 s gantry rotation time, 140–160 mm detector width, a slice collimation 2 × 256 × 0.625 mm, 100 kV tube voltage, and a reference tube current of 400–500 mA. Iodinated contrast medium (350 mg I/mL or 370 mg I/ mL) was injected at a volume of 55–65 mL and a speed of 4.5–5.5 mL/s with a saline chaser of 30–40 mL. End-systolic (35–45% of the R-R interval) and mid-diastolic (60–80% of the R-R interval) CCTA phases were reconstructed with the slice thickness/ interval of 0.75 mm/ 0.5–0.625 mm/0.5 mm using a medium smoothtissue convolution kernel (B26f) or a standard algorithm. The optimal CCTA phases were used for analysis. Pericoronary FAI measurement FAI around the three major coronary arteries was measured automatically using a dedicated artificial intelligence-based software system (Shukun AI software; CoronaryDoc, Plus V5.1, ShuKun Technology, Beijing, China). The software defines a region of interest within a radial distance equal to the diameter of the target vessel from the outer vessel wall and calculates the mean CT attenuation (in Hounsfield units) of adipose tissue (− 190 to − 30 HU) within this volume. At the vessel level, 40-mm long proximal segment of three vessels (proximal 10 mm to 50 mm in RCA to avoid effects of aortic wall, proximal 0 mm to 40 mm in LAD, proximal 0 mm to 40 mm in LCX) were traced. To ensure measurement reliability, images from 30 randomly selected patients were independently analyzed by two experienced radiologists (P.N. and W.Y.). Intraclass correlation coefficients (ICCs) were calculated to assess interobserver and intraobserver agreement. The results demonstrated excellent reproducibility for FAI measurements: for RCA, ICCs were 0.93 (interobserver) and 0.95 (intraobserver); for LAD, 0.89 and 0.91; and for LCX, 0.87 and 0.90, respectively. The AI-based software employs a scanner-independent algorithm validated for use across different CT platforms. The excellent reproducibility was achieved on a mixed sample including images from both scanners, further supporting the consistency of FAI measurements across the two platforms. Definitions POAF was defined as newly occurring atrial fibrillation postoperatively in patients without preoperative AF history. All patients were transferred to the cardiac surgery intensive care unit postoperatively and underwent continuous bedside electrocardiographic monitoring (telemetry) for at least 48 h. After transitioning to the general ward, patients continued to receive continuous electrocardiographic monitoring for at least an additional 48 h. If arrhythmia or symptoms such as palpitations occurred during this period, continuous monitoring was extended until postoperative day 7. Nevertheless, we acknowledge that even with extended continuous monitoring in symptomatic patients, brief or entirely asymptomatic episodes of atrial fibrillation may still have been missed, which constitutes a potential methodological limitation of this study. The radiologist performing FAI measurements (P.N.) and the cardiologist determining POAF diagnosis were blinded to each other’s findings. Specifically, FAI analysis was conducted without knowledge of postoperative POAF status, and POAF assessment was performed independently of FAI values. Patients were considered to have no prior history of atrial fibrillation based on a comprehensive review of medical records, including: (a) absence of documented AF diagnosis in outpatient or inpatient records prior to surgery; (b) no history of antiarrhythmic medication use specifically for AF (e.g., amiodarone, flecainide) ; (c) preoperative 12-lead electrocardiogram (ECG) showing sinus rhythm without evidence of AF or atrial flutter. Statistical methods All statistical analyses were completed using SPSS software version 31.0. Continuous variables that conformed to a normal distribution were expressed as mean ± standard deviation (SD) using the t test. Variables that did not conform to a normal distribution were expressed as medians (quartiles) using the Wilcoxon signed rank sum test. Categorical data were then expressed using the number of cases (%) and the chi-square or Fisher exact test was used to compare categorical data. Variables for inclusion in the multivariate logistic regression model were selected based on a combination of statistical significance (p < 0.10 in univariate analysis) and established clinical relevance to POAF. The latter included age, LVEF, BNP, number of distal anastomoses, and operative time, which were forced into the model regardless of their univariate p-value due to their known association with POAF or surgical complexity. This purposeful selection approach ensured that clinically important predictors were considered regardless of their univariate p-value. Finally, the area under the curve (AUC) was calculated by subject operating characteristic (ROC) curve analysis. A two-sided P < 0.05 was considered a statistically significant difference. Statistical significance of differences between prediction models was calculated using DeLong’s test. Model calibration was assessed using the Hosmer–Lemeshow goodness-of-fit test, with a p value > 0.05 indicating adequate calibration. To assess for potential multicollinearity among the included variables in the final model, variance inflation factors (VIFs) were calculated. A VIF > 5 was considered indicative of significant collinearity. Results Patients’ characteristics During the period from December 2023 to February 2025, a total of 435 patients underwent OPCABG surgery. After excluding 271 patients who did not undergo preoperative CCTA, 164 patients who received preoperative CCTA examinations were included. Further exclusions were then made according to the study criteria. Finally, a total of 140 patients were included, including 41 patients in the POAF group and 99 patients in the NO-POAF group. The flow chart of patient inclusion is shown in Fig. 1. Flow chart of patient enrollment. CCTA = coronary computed tomography angiography; FAI = fat attenuation index; POAF = postoperative atrial fibrillation Univariate analysis for postoperative atrial fibrillation The baseline characteristics of the study population, stratified by the development of POAF, are summarized in Table 1. A total of 140 patients were included (41 POAF, 99 NO‑POAF). Compared with patients who remained in sinus rhythm, those who developed POAF had a higher prevalence of hypertension (78.0% vs. 55.6%, p = 0.013) and a more frequent history of myocardial infarction (41.5% vs. 24.2%, p = 0.042). Preoperative LVEF tended to be lower in the POAF group (53.2 ± 8.3% vs. 55.2 ± 7.8%, p = 0.085), and troponin T levels showed a similar trend (0.11 ± 0.32 vs. 0.05 ± 0.13 ng/mL, p = 0.069). Among the three pericoronary fat attenuation indices, the right coronary artery FAI differed significantly between groups (− 80.24 ± 8.40 HU vs. −83.25 ± 7.76 HU, p = 0.043), whereas the difference for the left anterior descending artery showed a trend (p = 0.080) and that for the left circumflex artery did not reach significance (p = 0.142). Other variables, including age, sex, BMI, diabetes, dyslipidemia, smoking, preoperative medications (beta‑blockers, statins, aspirin, ARNI/ACEI/ARB), BNP, number of distal anastomoses, operative time, and NYHA class, were comparable between the two groups (all p > 0.05). Table 1Differences between patients who developed POAF or notItemsPOAF group(n = 41)No POAF group(n = 99)t/x²/z p Age(years)64.3 ± 8.763.4 ± 8.40.5760.283Male/female33/877/220.1260.722BMI(kg/m²)25.8 ± 3.225.8 ± 3.1-0.0210.492Pulse Pressure(mmHg)52.2 ± 11.752.8 ± 13.5-0.2290.410Hypertension32(78.0%)55(55.6%)6.2350.013Diabetes mellitus16(39.0%)46(46.5%)0.6500.420Hyperlipidemia23(56.1%)63(63.6%)0.6950.404Smoking17(41.5%)42(42.4%)0.0110.917Troponin T(ng/mL)0.11 ± 0.320.05 ± 0.131.4930.069ARNI/ACEI/ARB20(48.8%)60(60.6%)1.6560.198Statins31(75.6%)75(75.8%)0.0010.985Beta-blockers40(97.6%)93(93.9%)0.8010.371Aspirin34(82.9%)84(84.8%)0.0810.776History of myocardial infarction17(41.5%)24(24.2%)4.1520.042BNP(pg/mL)341(156,657)446(209,749)-0.6590.510Number of distal anastomoses3(3,3)3(3,3)-0.0850.932Operative time (min)231.2 ± 25.1225.7 ± 25.21.1970.231LVEF(%)53.2 ± 8.355.2 ± 7.8-1.3770.085FAI(hu) LAD-79.90 ± 8.73-82.40 ± 7.161.7610.080 LCX-75.34 ± 8.40-77.48 ± 7.551.4780.142 RCA-80.24 ± 8.40-83.25 ± 7.762.0380.043NYHA Ⅰ-Ⅳ Ⅰ5(12.2%)9(9.1%)1.7530.625 Ⅱ13(31.7%)43(43.4%) Ⅲ20(48.8%)40(40.4%) Ⅳ3(7.3%)7(7.1%) Differences between patients who developed POAF or not Multiple regression analysis for postoperative atrial fibrillation Variables associated with POAF in univariate analysis (p < 0.10) were entered into a multivariate logistic regression model together with clinically relevant factors (age, LVEF, BNP, number of distal anastomoses, and operative time). The complete results are presented in Table 2. After adjustment, hypertension remained an independent predictor of POAF (OR 3.93, 95% CI 1.53–10.11, p = 0.005). A higher FAI around the right coronary artery was also independently associated with POAF (per 1 HU increase: OR 1.08, 95% CI 1.01–1.15, p = 0.022). History of myocardial infarction showed a trend toward independent predictive value but did not reach statistical significance (OR 2.54, 95% CI 0.99–6.53, p = 0.054). Age, LVEF, troponin T, BNP, number of distal anastomoses, operative time, and FAI around the left anterior descending artery were not independent predictors in the final model (all p > 0.05). The overall model was statistically significant (omnibus test: χ² = 26.11, df = 10, p = 0.004) and demonstrated adequate calibration (Hosmer–Lemeshow goodness‑of‑fit test: χ² = 6.335, df = 8, p = 0.610). All variance inflation factors were below 2.0, indicating no substantial multicollinearity. The results are visualized in the forest plot (Fig. 2). Table 2The final POAF prediction model was constructed by entering variables that were significant in univariate analysis (p < 0.10), along with other clinically relevant factors considered a priori (e.g., age, left ventricular ejection fraction), into a multivariate logistic regression framework. This approach aligns with the purposeful variable selection strategy outlined in the Statistical Methods sectionItemsBOR95%CI p Age0.0291.030(0.978,1.084)0.266Troponin T0.3611.434(0.192,10.715)0.725LVEF-0.0510.950(0.898,1.006)0.079FAI(LAD)-0.0020.998(0.938,1.061)0.944FAI(RCA)0.0771.080(1.011,1.153)0.022BNP0.0011.000(0.972,1.029)0.989Number of distal anastomoses-0.1710.843(0.356,1.999)0.698Operative time0.0081.008(0.989,1.027)0.412Hypertension1.3683.928(1.526,10.108)0.005History of MI0.9312.537(0.986,6.531)0.054 The final POAF prediction model was constructed by entering variables that were significant in univariate analysis (p < 0.10), along with other clinically relevant factors considered a priori (e.g., age, left ventricular ejection fraction), into a multivariate logistic regression framework. This approach aligns with the purposeful variable selection strategy outlined in the Statistical Methods section Forest plot of multivariate logistic regression analysis for predicting postoperative atrial fibrillation. The model included all variables listed in Table 2. Horizontal lines represent 95% confidence intervals. The vertical dashed line indicates an odds ratio of 1.0. Variables with p < 0.05 are indicated by filled markers. *p* values: Hypertension p = 0.005; FAI RCA p = 0.022. Abbreviations as in Table 2 Sensitivity analysis for potential confounding by RCA territory disease burden To further evaluate whether the association between FAI (RCA) and POAF was merely a reflection of greater disease burden in the RCA territory rather than a marker of local perivascular inflammation, a sensitivity analysis was performed. The multivariate model was re-estimated with the additional inclusion of RCA territory grafting status, a clinically meaningful surrogate for severe RCA stenosis requiring surgical revascularization. After this adjustment, FAI (RCA) remained an independent predictor of POAF (OR 1.08, 95% CI 1.01–1.15, p = 0.022), whereas RCA grafting itself was not significantly associated with POAF (OR 0.59, 95% CI 0.02–22.34, p = 0.776) (Supplementary Table S1). These findings suggest that the predictive information provided by FAI (RCA) is not confounded by the presence of severe RCA stenosis, supporting the interpretation that this imaging biomarker captures peri-coronary inflammatory activity beyond that conveyed by anatomical disease burden alone. POAF prediction model based on multivariate To evaluate the incremental predictive value of FAI (RCA), two prediction models were constructed and compared using receiver operating characteristic (ROC) analysis (Fig. 3). The clinical model, which incorporated age, hypertension, history of myocardial infarction, LVEF, troponin T, BNP, number of distal anastomoses, and operative time, yielded an area under the curve (AUC) of 0.711 (95% CI 0.618–0.805). When FAI (RCA) was added to this model, the AUC increased to 0.752 (95% CI 0.662–0.843). A DeLong test confirmed that the improvement in discrimination was statistically significant (p = 0.049). These findings indicate that the addition of pericoronary fat attenuation assessment provides supplementary risk information beyond that obtained from established clinical and perioperative variables alone. Receiver operating characteristic curves for predicting postoperative atrial fibrillation. The clinical model (blue) incorporates age, hypertension, history of myocardial infarction, LVEF, troponin T, BNP, number of distal anastomoses, and operative time. The clinical + FAI model (red) additionally includes the fat attenuation index of the right coronary artery. AUC = area under the curve; CI = confidence interval; DeLong test: p = 0.049 Discussion The primary finding of this study is that pericoronary adipose tissue attenuation around the RCA, measured as FAI, independently predicts POAF after OPCABG. Moreover, adding FAI (RCA) to a clinical prediction model significantly improved its ability to identify high-risk patients. A sensitivity analysis further demonstrated that this association was not confounded by RCA territory disease burden, as FAI (RCA) remained independently predictive after adjusting for RCA grafting. It is important to distinguish FAI from epicardial adipose tissue volume. FAI measures mean CT attenuation (-190 to -30 HU), reflecting adipocyte size, lipid content, and local inflammation—with inflamed tissue showing higher (less negative) values due to microvascular leakage and fibrosis. Epicardial fat volume represents total fat mass and correlates with chronic metabolic overload and structural burden. These parameters capture distinct biological processes: FAI is a dynamic marker of inflammatory activity, while fat volume reflects stable structural remodeling. Previous studies show they are poorly correlated and provide independent prognostic information [20, 22]. We focused on FAI because perioperative inflammatory activation—rather than chronic fat accumulation—may be more relevant to POAF pathogenesis. The absence of epicardial fat volume quantification in our cohort represents an opportunity for future research. Univariate analysis initially identified hypertension, history of MI, and FAI (RCA) as factors associated with POAF. This aligns with established knowledge, as both hypertension [23, 24] and prior MI [25–29] are well-recognized risk factors for atrial fibrillation, primarily due to their role in promoting left atrial remodeling, increasing left ventricular filling pressures [30, 31], and creating a pro-arrhythmic substrate [32, 33]. The significant difference in FAI (RCA) between the POAF and non-POAF groups is a notable finding. A higher FAI value (less negative) is generally interpreted as a marker of increased inflammation and adverse adipocyte remodeling in the pericoronary fat. In the multivariate analysis, hypertension and FAI (RCA) emerged as independent predictors, while history of MI showed a trend (p = 0.054). Notably, several variables that are established POAF risk factors in other settings—including age, BNP, number of distal anastomoses, and operative time—were not independently predictive in our cohort. This may reflect the relatively narrow range of these variables in an OPCABG population with preserved cardiac function, as well as our limited sample size. The preoperative use of medicine was comparable between groups and therefore was not entered into the multivariate model. The finding that only RCA-derived FAI was an independent predictor may have an anatomical basis. The RCA courses within the right atrioventricular groove, in close proximity to the right atrial myocardium. Inflammation in the pericoronary fat surrounding the RCA may thus more directly reflect a pro-inflammatory state adjacent to the right atrium, a key substrate for AF initiation, particularly in the postoperative setting [34, 35] .The lack of significant association between LCX FAI and POAF, despite the anatomical proximity of the LCX to the left atrium, warrants further consideration. Several factors may explain this observation. First, while left atrial remodeling is central to the pathogenesis of chronic atrial fibrillation, postoperative AF may have a distinct underlying mechanism. Emerging evidence suggests that POAF is often triggered by acute postoperative inflammation and sympathetic stress [12]. In OPCABG, the right atrium is anatomically closer to the surgical field and may experience greater mechanical irritation than the left atrium, which is located on the posterior heart and relatively shielded. This differential exposure may render the right atrium more susceptible to pro-arrhythmic triggers. Second, the proximal LCX segment (0–40 mm from the ostium) measured in our study may not consistently approximate the left atrial myocardium along its entire course, particularly in patients with dominant left circumflex systems or variable coronary anatomy. Third, the right atrium’s thinner wall and greater density of autonomic ganglia may render it more susceptible to paracrine inflammatory signals from adjacent epicardial fat [36]. These anatomical and pathophysiological distinctions underscore that not all pericoronary fat depots are equivalent in their contribution to POAF, and that RCA FAI may be uniquely positioned to capture the peri-atrial inflammatory environment most relevant to postoperative arrhythmogenesis. Furthermore, our sensitivity analysis demonstrated that the predictive value of FAI (RCA) persisted after adjustment for RCA territory grafting, supporting the interpretation that it captures peri-coronary inflammatory activity independent of RCA stenosis severity. The core of this study lies in the development and comparison of prediction models. The inclusion of FAI (RCA) led to a statistically significant improvement in model discrimination (ΔAUC = 0.041, DeLong p = 0.049). As an objective imaging biomarker readily obtainable from preoperative CCTA, FAI (RCA) adds a new dimension to POAF risk assessment. While this suggests that FAI provides incremental predictive information beyond traditional risk factors, the absolute gain in AUC is moderate and the p-value is borderline, and the clinical impact of this improvement should be interpreted with caution. If validated in future studies, preoperative identification of patients with elevated FAI (RCA) could allow for hypothesis-driven exploration of targeted monitoring and prophylactic strategies, potentially reducing the incidence and associated complications of POAF. However, the definition of an “elevated” FAI value for clinical decision-making requires further investigation, including the establishment of validated cut-offs in larger prospective cohorts [5]. Several limitations must be acknowledged. First, this single-center retrospective study had a limited sample size (n = 140), largely due to excluding patients without preoperative CCTA. This may affect model stability and generalizability. However, our supplementary comparison (Supplementary Table S2) showed that excluded patients were broadly comparable to included patients across major baseline characteristics, suggesting reasonable representativeness. Second, although our multivariate model included 10 covariates, the number of POAF events was only 41, resulting in an events‑per‑variable (EPV) ratio of approximately 4.1. This is below the commonly recommended minimum of 10 (Peduzzi rule), raising the possibility of model overfitting and unstable coefficient estimates (e.g., the wide confidence interval for troponin T). Consequently, the effect sizes and statistical significance should be interpreted with caution, and the model should be considered hypothesis‑generating rather than definitive. Third, although inflammation is a plausible biological link between elevated FAI and POAF, we did not directly validate this mechanism through histological or molecular analysis of pericoronary adipose tissue. This is further supported by recent evidence that circulating inflammatory biomarkers, such as decreased preoperative interleukin-2 levels, independently predict POAF after cardiac surgery [37], suggesting that imaging and molecular markers may capture complementary aspects of the pro-arrhythmic inflammatory state. Fourth, potential confounders such as preoperative COPD and postoperative complications were not systematically captured; these should be incorporated in future prospective models. Fifth, our POAF detection protocol—continuous telemetry for at least 4 days postoperatively, with extended continuous monitoring in patients who developed arrhythmia or symptoms such as palpitations—may still have missed brief or entirely asymptomatic AF episodes, potentially underestimating the true incidence of POAF. Future studies employing prolonged continuous ambulatory monitoring are needed to provide a more accurate assessment of POAF burden. Sixth, left atrial size was not systematically measured; however, BNP levels were included as a surrogate marker of atrial pressure overload. Sixth, our strict exclusion criteria may limit applicability to older, higher-risk, or more comorbid surgical populations. Prospective, multicenter studies with larger cohorts and integrated biomarker analyses are needed to validate these findings and clarify the underlying mechanisms. From a clinical perspective, the preoperative identification of high POAF risk using FAI could guide targeted prophylactic strategies. For such patients, more rigorous perioperative beta-blocker administration [38, 39], consideration of anti-inflammatory regimens (e.g., colchicine [11]), intensified electrolyte management, and extended postoperative rhythm monitoring [40], and even concomitant Cox‑maze IV [41–44] procedure could be considered.Future research may need to develop additional laboratory and imaging methods based on the underlying mechanisms of atrial fibrillation (e.g., inflammation, exosomes [45]) to better predict its occurrence. In summary, FAI (RCA) provides promising incremental predictive information and may help identify patients at elevated POAF risk, potentially supporting more personalized perioperative management. However, its precise clinical utility and optimal threshold require validation in prospective studies. Supplementary Information Supplementary Material 1. Acknowledgements The authors thank all contributors who made this work possible. Abbreviations Area Under the Curve Body Mass Index Coronary Computed Tomography Angiography Confidence Interval Fat Attenuation Index Hounsfield Units Left Anterior Descending Artery Left Circumflex Artery Left Ventricular Ejection Fraction Myocardial Infarction Off-Pump Coronary Artery Bypass Grafting Percutaneous Coronary Intervention Postoperative Atrial Fibrillation Right Coronary Artery Receiver Operating Characteristic Standard Deviation Authors’ contributions Conceptualization, Sumin Yang, Pei Nie; writing—original draft preparation, Xiaodong Chen, Haoyan Li; writing—review and editing, Haoyu Hu, Dong Wang; Statistical Analysis, Xiaodong Chen; All authors have read and agreed to the published version of the manuscript. Funding The authors state that this work has not received any funding. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request, subject to institutional ethical approval. The data are not publicly available due to privacy or ethical restrictions. Declarations Ethics approval and consent to participate All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments. Institutional Review Board approval was obtained (Ethics Approval Number: Grant QYFYQZLL30197). Written informed consent was waived due to the retrospective nature of the study. 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. Xiaodong Chen, Haoyan Li and Haoyu Hu contributed equally to this work. Contributor Information Pei Nie, Email: niepei@qdu.edu.cn. Sumin Yang, Email: yangsumin@qdu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request, subject to institutional ethical approval. The data are not publicly available due to privacy or ethical restrictions.