Postoperative atrial fibrillation and atrial epicardial fat: Is there a link?

Highlights • Detailed local EAT analysis does not differ between POAF and non-POAF.• General rather than local effects of EAT play a role in the onset of early POAF.• The dominance of surgical induced factors may obscure the potential role of EAT.


Introduction
The onset of atrial fibrillation (AF) after cardiac surgery is associated with an increased risk of stroke, higher hospitalization costs and decreased short-and long-term survival [1,2]. In addition, the occurrence of postoperative AF (POAF), although often considered to be a transient phenomenon, is a predictor of later recurrences of AF [3]. For example, patients with a single episode of early POAF have an eightfold increased risk in developing AF later on [2]. Although the exact pathophysiological mechanisms underlying this arrhythmia are still incompletely understood, it is believed that POAF is the result of an interplay between acute factors related to the early postoperative period, and pre-existent substrate related to e.g. heart disease and ageing of the heart [4].
Epicardial adipose tissue (EAT) [5], defined as the volume of adipose tissue between the myocardium and the visceral pericardium, can lead to structural remodelling and AF substrate development [6]. However, the association between EAT and AF after cardiac surgery remains unclear. Given the transient character of early postoperative AF (POAF), and the role of EAT early in the development of the AF substrate, it is interesting to evaluate the role of EAT in the occurrence of AF after cardiac surgery. Therefore, this study aims to investigate the association between EAT and early POAF in patients undergoing cardiac surgery. We hypothesize that patients with higher EAT volumes are more likely to develop POAF after cardiac surgery than patients with low EAT volumes.

Study design and setting
This retrospective study was approved by the Institutional Review Board (IRB) and Ethics Committee (METC 2018-0448), and analysed anonymously in accordance with IRB guidelines. The study complies with the ethical principles of the Helsinki Declaration.
Consecutive patients undergoing coronary artery bypass graft (CABG), aorta ascendens surgery or surgical aortic valve replacement (AVR) at the Maastricht University Medical Centre+ (the Netherlands) from the 1st of November 2009 until the 31st of December 2019 were potential candidates. We included patients who had undergone a preoperative coronary computed tomographic angiography (CCTA) prior to the surgical procedure. Patients with a history of AF were included in the study to compare baseline characteristics with non-AF patients, but excluded from the primary analysis comparing EAT between POAF and SR to reduce confounding.

Data selection
POAF was defined as an (a)symptomatic episode of supraventricular arrhythmia of at least 30 s detected by continuous rhythm monitoring or a 12-lead ECG [7]. Postoperative continuous monitoring was used at least three and four days in patients who underwent CABG or aortic (valve) surgery respectively. In the case of symptoms such as palpitations a 12-lead ECG was obtained.

Computed tomographic imaging protocol
All CT scans were conducted using a standardized protocol, using a second generation dual source CT scanner (Somatom Definition Flash, Siemens Medical Solutions, Erlangen, Germany; slice collimation 128x0.6 mm) or a third generation dual source CT-scanner (Somatom Force, Siemens Healthineers, Forchheim, Germany; slice collimation 192x0.6 mm). Tube voltage of scan protocols varied between 80 and 120 kV. The contrast management (CM) injection protocol consisted of 300 mg l/ml Iopromide (Ultravist, Bayer Healthcare, Berlin, Germany), prewarmed to body temperature (37 C; 99 F), injected using a dual-head CT power injector (Stellant, Bayer).
Depending on the heart rate, a different scan protocol was used. In patients with a stable heart rate < 60 bpm, a prospectively ECGtriggered "high pitch" spiral protocol was used ("Flash"-technique). A prospectively triggered "adaptive sequence" protocol (prospective sequential data acquisition) was used in patients with a stable heart rate between 60 and 90 bpm. A retrospectively gated helical protocol was used in patients with an irregular heart rhythm or with a heart rate ≥ 90 bpm. To reduce confounding factors in the analysis, all patients who were scanned with an older generation CT-scanner (Brilliance 64, Philips, Healthcare, Best, Netherlands) were excluded, as differences in scan and CM injection parameters (e.g. tube voltage [kV], total injected CM volume, flow rate) have an effect on the attenuation of EAT as measured by Hounsfield Units (HU).

Atrial delineation
LA EAT was manually delineated using an image processing program designed for scientific multidimensional images (ImageJ software, U.S. National Institutes of Health, Bethesda, MD, USA) and a pen tablet (Wacom, Kazo, Saitama, Japan) [8]. The regions of interest (ROI) were manually delineated, starting at the basal slice above the mitral valve annulus. In addition to this the left atrial appendage was delineated [8]. Delineation of LA EAT was performed by an independent assessor (C.A. J. van der H.) blinded for patient characteristics and rhythm outcome. In the case of inconclusive quantification, scans were reviewed by a second assessor (B.M.) ( Fig. 1A and D).

Epicardial adipose tissue calculation
Custom-made semi-automated software developed in ImageJ (Adi-poQuant, developed by S.V.) was used to quantify the number of pixels of EAT and myocardial tissue (MYO) based on the HU per ROI. A range within − 30 to − 190 HU was defined as EAT, while MYO was defined as pixels ranging from 10 to 120 HU ( Fig. 1B-C and E-F) [8,9]. Thereafter, total LA EAT and MYO volumes (EAT-V and MYO-V) in millilitre (ml) were calculated using several parameters obtained from the metadata of the CCTA, including pixel size, slice thickness and the spacing between slices. EAT-V and MYO-V were used to calculate the percentage of EAT in the LA myocardial wall (EAT-V/[EAT-V + MYO-V]*100). Consequently, EAT-V and EAT-V indexed to LA-volume (ml) (EAT-LA) were indexed to several clinical parameters that potentially influence the volume of EAT, including body surface area (BSA), body mass index (BMI), MYO-V and left atrial dimension. To determine whether EAT location differed within and between groups, a separate EAT-V analysis at the LA roof was performed. Subsequently, roof EAT-V and non-roof EAT-V (total EAT-V minus roof-EAT-V) were compared ( Fig. 1D-F).

Statistical analysis
Statistical analysis was performed based on rhythm outcome (POAF versus SR) using SPSS version 25.0 (SPSS Inc., Chicago, IL, USA). Normality of distribution of continuous variables was tested using the Shapiro-Wilk test and by visual inspection of histograms. A student's ttest was performed for continuous means in case of normality (mean ± standard deviation) and a Mann-Whitney-U test in case of non-normally distributed data (median and interquartile range). A Pearson's χ 2 test or Fisher's exact test was performed to compare categorical variables between both groups. Univariate analyses were performed using Pearson's r or Spearman's Rho, depending on data distribution, between clinical characteristics and EAT variables. Finally, multivariable regression analysis was performed to identify independent predictors of POAF. For all analyses, differences were considered statistically significant if P < 0.05.

Study population
Of the screened population, 107 patients that underwent cardiac surgery at Maastricht University Medical Centre between 2009 and 2019 had a preoperative CCTA. Eleven patients were excluded due to incomplete or suboptimal quality of the images and 13 patients were excluded due to being scanned with an older scanner, as outlined in the methods. In addition, 12 patients had a history of AF and were therefore excluded from the EAT analysis only. In total, 83 patients were retrospectively included in our analysis.

Procedural characteristics
All procedures were carried out at Maastricht University Medical Centre. Cardiac surgery consisted of surgery of the ascending aorta (with or without aortic valve repair n = 7, with aortic valve repair n = 6), isolated aortic valve surgery (repair n = 4, replacement n = 29) or standalone coronary artery bypass graft (CABG) (on-pump n = 28, off-pump n = 6). A combination of CABG and aortic valve surgery was performed in 9 patients.

Epicardial adipose tissue measurements
EAT-V and MYO-V were calculated for all patients, with exclusion of the patients known with a history of AF (n = 12). An example of a patient with a relatively low and high EAT-V on CCTA is shown in Fig. 2. When comparing the volume of EAT between those with POAF and SR, no significant difference was seen (0.68 ml POAF vs. 0.67 ml SR, P = 0.431). Moreover, the percentage of EAT in the LA wall ( differed between both groups (all P > 0.05). Also EAT-V indexed to LA (ml) (EAT-LA) indexed to BSA, BMI, LAVI and MYO did not differ between groups either (Table 3). Furthermore, the percentages of EAT of the total LA wall, roof and non-roof did not statistically differ between POAF and SR (all P > 0.05), but patients with POAF did have a higher percentage of EAT in the roof compared to non-roof (23.28 ± 12.65 vs. 18.23 ± 8.91, P = 0.020) (Tables 3-4).
Following univariate analyses, EAT-V variables were positively correlated with MYO-V and the percentage in the LA wall, but not with POAF. A weak positive trend was seen for age and EAT-V normalized to LA volume normalized to BSA, but this was not significant (r = 0.270, P = 0.053). After multivariable logistics regression analysis (with the exclusion of patients with pre-operative AF) using age, LAVI (ml/m 2 ), the total percentage of EAT and the percentage of EAT in the roof, only age and LAVI were independent predictors for developing POAF (odd's ratio (OR) = 1.  (Table S2).

Discussion
In this study we quantified EAT-V in the left atrial wall. We found that no local EAT-V variables were associated with POAF following cardiac surgery, while advanced age and LAVI were independent

Mechanisms underlying EAT and AF
Although previous studies have highlighted the role of EAT in AF, the exact mechanisms of the underlying pathophysiological process remain unclear [10]. In this process, two main potential arrhythmogenic mechanisms can be distinguished. First, EAT directly influences atrial electrophysiological properties by fatty infiltration into the contiguous epicardial layer of the myocardium [11], resulting in shortening of the atrial refractoriness, thereby favouring local conduction block and micro re-entry circuits [12,13]. Secondly, paracrine secretion of factors such as adipokines can influence the direct adjacent myocardium, thereby leading to oxidative stress, autonomic-and diastolic dysfunction and adipocyte-related atrial gene expression [14,15,16]. Eventually, a combination of these factors can induce myocardial inflammation and fibrosis, resulting in slow but progressive structural remodelling and stabilization and perpetuation of AF [17]. On the contrary, Antonopoulos et al. reported a potential protective effect of atrial EAT in relation to oxidative stress [18]. During AF oxidative stress leads to an alteration in gene expression in EAT, such as the adipokine adiponectin. In the human heart, adiponectin in turn decreases myocardial oxidative stress via endocrine or paracrine effects, thereby representing a novel defence mechanism against myocardial oxidative stress. In addition to its role in inflammation and oxidative stress, EAT is involved in lipid and energy homeostasis by secreting free fatty acids [17]. Compared to other visceral depots, it has lower glucose utilization and also has brown fat properties that are hypothesized to protect the myocardium against hypothermia by producing free fatty acids [17]. Non-invasive imaging techniques can be used to measure adipose tissue as a surrogate marker of inflammation [19]. As the attenuation of inflamed adipocytes increases, inflamed cells can be distinguished from non-inflamed cells by CT-scan analyses based on the HU. Furthermore, PET-scan analysis could be used to provide real-time visualization of metabolic and inflammatory activity by the uptake of radioactive tracers within adipose tissue, thus providing more information on the metabolic characteristics of adipose tissue in a certain region of interest [22]. Unfortunately, these kinds of analyses have only been described on peri-coronary adipose tissue and not yet on fatty tissue on the atrial level. Subsequently, these analyses have not yet been validated for atrial adipose tissue and therefore not yet applicable in our population.

EAT and POAF
Since EAT is involved in the early phase of structural remodelling leading to AF, but has also shown to have cardioprotective effects, it is interesting to investigate the role of these early remodelling processes of EAT in the onset of POAF, especially given the transient occurrence of early POAF. In the postoperative setting, Opolski et al. [20] studied independent predictors of POAF after CABG in 108 patients. They found that LA EAT, measured by a comprehensive CCTA analysis using semiautomated software, is a significant predictor of POAF (OR: 1.21). It must be noted that 7% of the included study population had paroxysmal AF, while we excluded patients with diagnosed AF from the EAT analysis as it is known that EAT is associated with the occurrence and severity of AF by the release of adipokines leading to various alterations of the structural and functional properties of the atrial myocardium [17]. Furthermore, in the study by Drossos et al. a strong association between pericardial adipose tissue and POAF was reported following CABG [21]. However, the authors analysed a mixture of epicardial and pericardial fat and performed whole heart instead of selective atrial analysis. However, in the pathogenesis of AF, the arrhythmic effects of EAT seem local rather than systemic. Furthermore, our study used 2D source data for analysis, while the study of Drossos et al. performed their analysis based on a 3D reconstruction, which may have altered the morphological composition of EAT [21]. Kogo et al. analysed the association between EAT and POAF based on local EAT analysis, and did not find an association [22]. Also we did not find a significant correlation between EAT and POAF incidence. It seems that local EAT analysis does not allow to find an association between the quantification of EAT and the onset of POAF, while whole heart analysis  of EAT does suggest that there is an association. As such, it might be that analysis of local EAT is not sensitive enough to find an association between EAT and the onset of POAF, but EAT analysis on the level of the whole heart is. On the other hand, it is known that in the acute postoperative phase, surgical induced factors (such as sympathetic activation, oxidative stress and inflammation) and hemodynamic and biochemical instability are known to elicit POAF also in patients with less severe AF substrates [4,5]. As we performed our EAT analysis in a relatively low-risk population without a history of AF and a LAVI of 38 ± 12 ml/m 2 , the dominance of these acute factors might obscure the effects of a predisposing substrate in this low risk population.

Limitations
This study has some limitations. First, the retrospective character of the study combined with a small number of patients included in this analysis limits drawing valid conclusions. Patients with mitral valve disease were excluded since the underlying aetiology of AF is probably different as mitral valve disease causes significant left atrial (LA) dilatation as the result of mitral regurgitation [23]. However, despite this, predictors of POAF such as LA dimension and age were also found to be predictive for POAF in the present study. Secondly, while manual CTscan analysis is considered a valid technique for measuring EAT and semi-automated quantification in custom made software allows for manual delineation of all regions of interest [9], no 'gold standard' overcoming human error in the analysis of EAT is yet available to compare our results with. Not only is manual delineation using semiautomated software time consuming (±45 min per scan), it can also be difficult to determine the exact anatomic borders of the ROI, such as the caudal and cranial part of the LA, as well as the entry points of the pulmonary veins into the LA. Finally, since both scan and CM injection parameters are patient specific and may differ between scanners, potential bias based on the heterogeneity in the attenuation of EAT as measured in HU cannot be excluded in this study. Although probably less important than LA EAT, as AF most often derives from the PV ostia and the posterior LA, right atrial EAT could not adequately be analysed in the present study due to the lack of contrast in the right atrium at the time of scanning.

Conclusion
This study found that advanced age and LAVI were independent predictors of POAF and that patients with POAF have significantly more EAT in the roof compared to non-roof. However, local EAT analysis was not associated with the occurrence of early POAF following cardiac surgery. This might suggest that early postoperative triggers dominate the ignition of early POAF, which potentially obscure the association of POAF with EAT, or that general rather than local effects of EAT play a role in the onset of POAF. Furthermore, perhaps not only the quantification of LA EAT is interesting to analyze, but also its metabolic and inflammatory properties. Therefore, it would be interesting if larger and prospective future studies combined PET-CT scan volume with inflammatory and/or metabolic activity analyses to compare not only the quantity of EAT between POAF and non-POAF, but also the metabolic and inflammatory characteristics of EAT, both systemic and local.

Grant support
None.