key: cord-0761921-piuis711
authors: Conte, Caterina; Esposito, Antonio; De Lorenzo, Rebecca; Di Filippo, Luigi; Palmisano, Anna; Vignale, Davide; Leone, Riccardo; Nicoletti, Valeria; Ruggeri, Annalisa; Gallone, Guglielmo; Secchi, Antonio; Bosi, Emanuele; Tresoldi, Moreno; Castagna, Antonella; Landoni, Giovanni; Zangrillo, Alberto; De Cobelli, Francesco; Ciceri, Fabio; Camici, Paolo; Rovere-Querini, Patrizia
title: Epicardial adipose tissue characteristics, obesity and clinical outcomes in COVID-19: a post-hoc analysis of a prospective cohort study
date: 2021-05-03
journal: Nutr Metab Cardiovasc Dis
DOI: 10.1016/j.numecd.2021.04.020
sha: 1f9d0cf85c96c3889f447b5ea38c0ab258f85d19
doc_id: 761921
cord_uid: piuis711

Background and aims Obesity-related cardiometabolic risk factors associate with COVID-19 severity and outcomes. Epicardial adipose tissue (EAT) is associated with cardiometabolic disturbances, is a source of proinflammatory cytokines and a marker of visceral adiposity. We investigated the relation between EAT characteristics and outcomes in COVID-19 patients. Methods and results This post-hoc analysis of a large prospective investigation included all adult patients (≥18 years) admitted to San Raffaele University Hospital in Milan, Italy, from February 25th to April 19th, 2020 with confirmed SARS-CoV-2 infection who underwent a chest computed tomography (CT) scan for COVID-19 pneumonia and had anthropometric data available for analyses. EAT volume and attenuation (EAT-At, a marker of EAT inflammation) were measured on CT scan. Primary outcome was critical illness, defined as admission to intensive care unit (ICU), invasive ventilation or death. Cox regression and regression tree analyses were used to assess the relationship between clinical variables, EAT characteristics and critical illness. One-hundred and ninety-two patients were included (median [25th-75th percentile] age 60 years [53-70], 76% men). Co-morbidities included overweight/obesity (70%), arterial hypertension (40%), and diabetes (16%). At multivariable Cox regression analysis, EAT-At (HR 1.12 [1.04-1.21]) independently predicted critical illness, while increasing PaO2/FiO2 was protective (HR 0.996 [95% CI 0.993; 1.00]). CRP, plasma glucose on admission, EAT-At and PaO2/FiO2 identified five risk groups that significantly differed with respect to time to death or admission to ICU (log-rank p<0.0001). Conclusion Increased EAT attenuation, a marker of EAT inflammation, but not obesity or EAT volume, predicts critical COVID-19. Trial registration NCT04318366.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of Coronavirus disease 2019 (COVID-19) [1] , drives a profound response in susceptible individuals, leading to hyperinflammation, cytokine storm syndrome and acute cardiac injury that associate with disease severity and outcome [2, 3] . Cardiometabolic risk factors such as hypertension, diabetes mellitus and history of ischaemic heart disease (IHD) are highly prevalent among patients with COVID-19, and are associated with increased odds of developing severe disease [4, 5] . Despite a high prevalence of obesity among COVID-19 patients [6, 7] and the close association of obesity with cardiometabolic risk factors, not all data point towards an association between body mass index (BMI) and critical illness [7, 8] , suggesting that the relationship between obesity and COVID-19 severity is more complex.

Recent observational data indicate that visceral adipose tissue rather than total body mass might play a role in determining the severity of COVID-19 [9, 10] . Epicardial adipose tissue (EAT), i.e. the visceral adipose tissue of the heart situated between the myocardium and the visceral layer of the pericardium, is a potential source of inflammatory mediators, including interleukin (IL)-1β, IL-6, and tumour necrosis factor (TNF)-α [11] . In turn, EAT inflammation can act in a paracrine manner to influence the structure and function of neighboring myocardial tissue [11] . The role of EAT as an independent cardiovascular risk factor that associates with fatal and non-fatal cardiovascular events in the general population is well-established [12] . Recently, EAT has also been proposed as a pathophysiological driver of myocardial inflammation in COVID-19 [13] , and preliminary data seem to confirm this hypothesis [14] . We aimed to investigate the relationship between EAT volume and attenuation (EAT-At), which reflects overall EAT inflammation and predicts all-cause and cardiac mortality [15] , with disease severity in patients hospitalised for COVID-19.

This post-hoc analysis was part of the COVID-BioB study, a large observational investigation performed at San Raffaele University Hospital in Milan, Lombardy Region, Italy. The study protocol complies with the Declaration of Helsinki, was approved by the local Hospital Ethics Committee (protocol no. 34/int/2020) and was registered on ClinicalTrials.gov (NCT04318366). Patients able to provide a signed informed consent were consented prior to data collection. The Ethics Committee waived the requirement for obtaining written informed consent for patients who had died or were unreachable after discharge. Full description of patient management and clinical protocols were previously published [16] .

All patients aged ≥18 years admitted to the Emergency Department (ED) at San Raffaele University Hospital with confirmed SARS-CoV-2 infection since February 25 th , 2020 to April 19, 2020 were consecutively enrolled in the COVID-BioB study. Confirmed infection was defined as positive real-time reverse-transcriptase polymerase chain reaction (RT-PCR) from a nasal and/or throat swab together with signs, symptoms, and radiological findings suggestive of COVID-19 pneumonia. For the purpose of this analysis, only hospitalised patients who underwent a CT scan and had COVID-19 characteristic features on CT scan were included. Patients admitted for other reasons and subsequently diagnosed with superimposed SARS-CoV-2 infection were excluded.

J o u r n a l P r e -p r o o f Data were collected from medical chart review or directly by patient interview and entered in a dedicated electronic case record form (eCRF) specifically developed for the COVID-BioB study. Prior to the analysis, data were cross-checked with medical charts and verified by data managers and clinicians for accuracy. The following variables were extracted for all patients: age, sex, ethnicity, body mass index (BMI, calculated as the ratio of weight in kilograms [kg] divided by height in squared metres), PaO 2 /FiO 2 (calculated as the ratio between the arterial partial pressure of oxygen measured on arterial blood gas analysis and the fraction of inspired oxygen), plasma glucose (mg/dL), estimated glomerular filtration rate (eGFR, as estimated by the CKD-EPI equation and expressed as ml/min/1.73m 2 ), lymphocyte and neutrophil counts (x10 9 /L), and high-sensitivity C-reactive protein (CRP, mg/dL) on admission, peak CRP and high-sensitivity troponin T levels (ng/L) during hospital stay, troponin T levels measured within 3 days of the CT scan, comorbidities (including history of hypertension, diabetes mellitus, dyslipidaemia, IHD, and malignancy) and clinical outcome (discharge, death or invasive ventilation or admission to intensive care unit, ICU). Myocardial injury was defined as high-sensitivity troponin T levels above the 99 th percentile upper reference limit [17] (at our Center, 14 ng/L). 

EAT volume was automatically segmented on a single slice at the level of the left main coronary artery on basal non-contrast scan, as previously described and validated [18, 19] , using a dedicated software (IntelliSpace Portal v.8.0, Philips The Netherlands). Briefly, total thoracic fat was automatically segmented (range between -150 and -70 HU) using the "tissue segmentation" plug-in. Subsequently, an experienced radiologist (D.V., 5 years of experience in chest imaging) manually contoured the pericardium, in order to extract the EAT volume (mm 3 ) and attenuation (HU mean ± standard deviation). CT scan of 50 randomly selected patients were segmented by a second radiologist blinded to the assessment of the first operator (R.L., 4 years of experience in chest imaging), according to the same method as previously described, in order to assess the interobserver reproducibility by calculating the intraclass correlation coefficient (ICC proportional hazards assumption was tested using the Schoenfeld's global test (p=0.063). A classification and regression tree (CART) approach was used to predict critical illness and identify risk groups based on significant predictors [20] . Age, sex, BMI, EAT, EAT-At, PaO 2 /FiO 2 , plasma glucose, absolute lymphocyte and neutrophil count, LDH, eGFR and CRP on admission, history of hypertension, diabetes mellitus, dyslipidaemia, coronary artery disease, and malignancy were included in the regression tree analysis. Time-to-event analyses to test differences across risk groups identified by regression tree analysis were performed using the Kaplan-Meier method and the log-rank test. All statistical tests were two-tailed with p < 0.05 considered significant. The accuracy of the risk prediction model was estimated 

Between February 25 th , 2020 and April 19, 2020, 652 patients with confirmed SARS-CoV-2 infection were admitted to San Raffaele University Hospital, of which 212 underwent a CT scan and had clinical and anthropometric data available for analyses (Supplemental Figure 1 ).

Twenty patients with CT images not suitable for analysis (contrast-enhanced CT images and/or post-surgical fibrotic mediastinal alterations) were excluded. A total of 192 patients were included in the present analysis, whose characteristics are presented in Table 1 . Median time from admission to CT scan was 3 (1.0; 6.5) days. The median time of observation was 44 (14; 54) days. The median age was 60 years (53; 70). Most patients were men (76%), of Non-Hispanic ethnicity (88%). Overall, nearly 70% of patients had overweight or obesity. On admission to the ED, 59% met the criteria [21] for acute respiratory distress syndrome (ARDS). The most common comorbidity was arterial hypertension (40%), followed by diabetes mellitus (16%) and dyslipidaemia (11%). Twenty-six (13.5%) patients had died after a median of 21 (14; 35) days in the hospital, 166 (86.5%) had been discharged (median length of hospital stay 19 [12; 33] days), and 23% had been admitted to ICU.

Median EAT volume in the whole group was 2510 (1561; 3539) mm 3 Critically ill patients had worse respiratory function, had higher markers of systemic inflammation, lower lymphocyte count and higher plasma glucose on admission (Supplemental Table 1 ).

Both BMI and EAT volume were similar between the two groups, while EAT-At was significantly greater in patients with critical illness (Supplemental Table 1 ). Higher TnT values (either peak values or values measured within 3 days of the CT scan), along with occurrence of myocardial injury were more frequently reported among critically ill patients (Supplemental Table 1 ). Table 2 ).

In the CART analysis, CRP, plasma glucose on admission, EAT-At and PaO 2 /FiO 2 emerged as most important predictors of critical illness, allowing patient stratification into five risk groups ( Figure 2 ). The first node was a CRP above or below 217 mg/dL. The probability of J o u r n a l P r e -p r o o f critical illness was highest in patients with CRP ≥ 217 mg/dL (87%, Group 5) . For those with a CRP < 217 mg/dL, there was an additional node, i.e. plasma glucose on admission above or below 128. The probability of critical illness was greater for those with a plasma glucose ≥ 128 (46%, Group 3). In those below this value, a EAT-At of -96.3 HU identified two additional subgroups: the probability of death or ICU admission was lowest (5%) in the subgroup with lower EAT-At (Group 1). A higher adipose tissue attenuation (less negative HU) is suggestive of EAT inflammation. Figure 3 shows the difference between a patient with EAT-At < -96.3 HU (less inflamed EAT, Figure 3A ) and one with EAT-At ≥ -96.3 HU (more inflamed EAT, Figure 3B ). For those with lower CRP (<217 mg/dL) and lower plasma glucose on admission (<128 mg/dL) but a EAT-At greater or equal to -96. 3 56.5%, 60% and 89.5% from group one to five, respectively; p<0.001) (Supplemental Figure   3 ). Finally, we found a significant-although weakcorrelation between EAT-At and troponin T levels (Spearman's rho 0.19, p=0.049, Supplemental Figure 5 ).

In this cohort of patients hospitalised for COVID-19, EAT inflammation (EAT-At) and poor respiratory function (PaO 2 /FiO 2 ) independently predicted critical illness. These variables, along with plasma glucose on admission and systemic inflammation (CRP) emerged as the most important predictors of critical illness in the CART analysis, which allowed to identify five risk groups with elevated sensitivity and specificity. It is not surprising that other variables compared to the Cox regression analysis turned out to be significant predictors of critical COVID-19 in the CART analysis, as this machine learning approach tests all values relative to a single variable against others to identify the "nodes" that allow splitting patients into risk groups, selecting the optimal sequence of classifications as defined by a hierarchy of prognostic factors and associated cut-points in each subgroup of patients. The five risk groups could readily be used for patient stratification in clinical practice. Patients with the lowest EAT attenuation (more negative values, less inflamed EAT) were at the lowest risk of critical illness (Group 1). Patients with greater systemic inflammation (CRP ≥217 mg/dL) were at the highest risk (Group 5). Of note, despite EAT-At not being a "node" for risk Groups 3 and 5, patients in these groups had significantly higher EAT-At as compared with Group 1 (lowest risk), suggesting a role of EAT inflammation in the development of critical illness even in these patients.

Increased EAT attenuation on chest CT is a marker of EAT inflammation. Visceral adipose tissue inflammation is the initial and key event triggering whole body metabolic abnormalities [22] , and may trigger coronary microvascular inflammation and dysfunction [23] . EAT inflammation could be triggered by SARS-CoV-2 infection. SARS-CoV-2 enters cells via the angiotensin converting enzyme 2 (ACE2) receptor, which is highly expressed in the cardiovascular system, including EAT [24, 25] . Loss of ACE2 induces EAT inflammation, cardiac insulin resistance and alterations in cardiac metabolism in murine models of diet-induced obesity [26] . Binding of SARS-CoV-2 to ACE2 reduces surface ACE2 expression [25] , possibly leading to EAT inflammation [13] . We found that EAT inflammation weakly -although significantly -correlated with systemic inflammation. EATderived inflammatory mediators might have systemic effects [27, 28] , but it is more likely that during COVID-19 EAT acts locally, as a 'fuel for cardiac inflammation' [29] , enhancing the release of cytokines such as leptin, IL-1, IL-6 and TNF-α and free fatty acids that may induce lipotoxicity [11, 30] . However, both local (e.g., EAT) and systemic inflammation may induce myocardial injury [17] . When systemic inflammation prevails, as in Group 5, myocardial injury might be secondary to an imbalance between myocardial oxygen demand and supply due to increased oxygen consumption during exaggerated systemic inflammation and reduced oxygenation due to respiratory insufficiency [31] . Inflammation-mediated myocardial injury is increasingly reported in patients with troponin increase during COVID-19 [32] . In these patients, increased EAT-At might be a marker of inflammation-mediated J o u r n a l P r e -p r o o f myocardial injury. In summary, EAT inflammation could be either a trigger for myocardial inflammation/injury or a marker of myocardial injury induced by systemic inflammation.

We found a significant, positive correlation between EAT-At and troponin T (Spearman's rho 0.19, p=0.049), i.e. the more the inflammation (less negative EAT-At), the higher the troponin values (Supplemental Figure 5 ). This correlation, however, was weak, probably because of the interplay between local and systemic inflammation contributing to myocardial injury [17] . In fact, the proportion of patients with myocardial injury increased across risk groups, which were identified by both EAT-At and CRP. Few data are available on the relationship between EAT inflammation and troponin T levels, a marker of myocardial injury. Iacobellis and colleagues found an inverse relationship (r = −0.45; P < 0.05) between EAT-At and troponin T levels in 41 COVID-19 patients, while Grodecki and colleagues found no correlation between EAT-At and troponin T in a cohort of 109 COVID-19 patients [14] . It is possible that these discrepancies are due to the smaller sample size of previous studies and, as already mentioned, to systemic factors involved in the pathophysiology of myocardial injury. Of note, the proportion of patients with myocardial injury was high, ranging from 26.3% to 89.5% in the five risk groups, despite a low prevalence of CAD in our cohort. This is in agreement with the report of Iacobellis and colleagues, who found that the degree of EAT-At in patients with critical COVID-19 was similar to that observed in patients with CAD, despite few patients having CAD and/or high coronary calcium content [33] . It should be highlighted, however, that the main aim of our study was to investigate the relationship between EAT characteristics and disease severity in patients hospitalised for COVID-19, and the assessment of the relationship between EAT-At and myocardial injury as assessed by troponin T levels was only exploratory.

The lack of association between obesity or EAT volume with disease severity is also a novel finding. Previous cross-sectional reports suggested obesity as a risk factor for COVID-19 severity, mainly based on a higher prevalence of obesity among patients with severe and critical COVID-19 [6] . The prevalence of obesity was 21% in our cohort, which is nearly twice as high as in the general Italian population (10.9%), and 2.5-fold that in the Lombardy Region (7.9%) [34] . Neither BMI nor obesity (BMI ≥ 30 kg/m 2 ) predicted critical illness.

Moreover, neither BMI nor obesity did differ across the five risk groups identified based on CRP, plasma glucose on admission, EAT-At and PaO 2 /FiO 2 (Figure 4 ). Having obesity increases the odds of a positive SARS-CoV-2 test (OR 1.41) [35] . However, previous studies demonstrated that obesity is associated with increased risk of ARDS, but lower risk of mortality (the so called "obesity paradox") [36] . The mechanisms responsible for this "obesity paradox" are yet to be clarified. Similar to BMI, EAT volume -which moderately correlated with BMI -neither predicted critical illness nor differed among groups. This finding is in contrast with those from a smaller study by Grodecki and colleagues in which EAT volume was independently associated with adverse outcomes in patients with COVID-19 in logistic regression models, despite a moderate correlation of BMI, which did not differ between patients with or without clinical deterioration [14] . Consistent with that study, we found that EAT-At was a significant predictor of worse outcomes. However, in our study neither the time-varying analyses nor the CART analysis support a role for obesity or EAT volume in determining the severity of COVID-19. Our findings are in line with recently published data on patients hospitalised with COVID-19 in New York City [7] , and with reports of no difference in EAT thickness between patients with or without severe COVID-19 [10, 33] . parameters that may be easily obtained upon hospital admission. A strength of this risk classification is the measurement of EAT-At on non ECG-gated non-contrast CT scans done for the assessment of lung involvement in COVID-19, a reproducible method that does not involve additional radiation exposure for patients. Image analysis took approximately three minutes per patient, and can be done using commercial softwares. The finding that EAT inflammation predicts critical illness in COVID-19 could also inspire downstream research about potential therapeutic implications, as agents commonly used in the management of diabetes mellitus and dyslipidaemia such as pioglitazone and some statins were shown to reduce EAT inflammation [37, 38] . Of note, statin use has been associated with reduced inhospital mortality in COVID-19 patients with diabetes [39] .

The results of our investigation should be viewed as hypothesis-generating in light of the study design and its limitations. The observational nature of the study does not allow to establish a causative role of EAT inflammation in COVID-19 critical illness. Autopsies or endomyocardial biopsies were not performed, therefore we could not compare CT findings with myo-pericardial histology. Patients with pre-exhisting cardiovascular disease are more likely to have myocardial injury [40] . A limitation of the present analysis is that we relied on medical history to rule out CAD rather than collecting echocardiography parameters or measuring coronary artery calcium (CAC) score, the latter being also an independent predictor of in-hospital mortality in COVID-19 [41] . Our results should be confirmed in larger multicenter studies, and EAT parameters complemented with the measurement of echocardiography parameters, CAC and cytokines such as IL-1β, IL-6, and TNF-α. Strengths of the present study include comprehensive clinical, biochemical and imaging evaluation, the J o u r n a l P r e -p r o o f study conduction at a single site, which allowed for homogenous treatment protocols, and the use of multivariable, time-varying analyses to assess the relationship of clinical, biochemical and EAT characteristics with critical illness in COVID-19 patients.

In conclusion, we found that despite being highly prevalent among COVID-19 patients, 

A new coronavirus associated with human respiratory disease in China

COVID-19 and the Heart

SARS-CoV-2: a storm is raging

The Relationship of COVID-19 Severity with Cardiovascular Disease and Its Traditional Risk Factors: A Systematic Review and Meta-Analysis

Comorbid Chronic Diseases and Acute Organ Injuries Are Strongly Correlated with Disease Severity and Mortality among COVID-19 Patients: A Systemic Review and Meta-Analysis

Prevalence of obesity among adult inpatients with COVID-19 in France

Epidemiology, clinical course, and outcomes of critically ill adults with COVID-19 in New York City: a prospective cohort study

Visceral Adiposity and High Intramuscular Fat Deposition Independently Predict Critical Illness in Patients with SARS-CoV-2

Visceral fat shows the strongest association with the need of intensive care in patients with COVID-19

Epicardial Adipose Tissue May Mediate Deleterious Effects of Obesity and Inflammation on the Myocardium

Association of epicardial fat with cardiovascular risk factors and incident myocardial infarction in the general population: the Heinz Nixdorf Recall Study

Does epicardial fat contribute to COVID-19 myocardial inflammation?

Epicardial adipose tissue is associated with extent of pneumonia and adverse outcomes in patients with COVID-19

Non-invasive detection of coronary inflammation using computed tomography and prediction of residual cardiovascular risk (the CRISP CT study): a post-hoc analysis of prospective outcome data

Biobanking for COVID-19 research

Cardiac Troponin for Assessment of Myocardial Injury in COVID-19: JACC Review Topic of the Week

Single-slice epicardial fat area measurement: do we need to measure the total epicardial fat volume?

A single slice measure of epicardial adipose tissue can serve as an indirect measure of total epicardial adipose tissue burden and is associated with obstructive coronary artery disease

Classification and Regression Trees

The Berlin definition of ARDS: an expanded rationale, justification, and supplementary material

The integrative biology of type 2 diabetes

Coronary microvascular dysfunction in hypertrophy and heart failure

Higher ACE2 expression levels in epicardial than subcutaneous stromal cells from patients with cardiovascular disease: Diabetes and obesity as possible enhancer

Angiotensin Converting Enzyme 2: A Double-Edged Sword

ACE2 Deficiency Worsens Epicardial Adipose Tissue Inflammation and Cardiac Dysfunction in Response to Diet-Induced Obesity

Targeting the Adipose Tissue in COVID-19

Obesity Accompanying COVID-19: The Role of Epicardial Fat

Epicardial adipose tissue: fuel for COVID-19-induced cardiac injury?

Local and systemic effects of the multifaceted epicardial adipose tissue depot

Myocardial injury and COVID-19: Possible mechanisms

Cardiac Magnetic Resonance Characterization of Myocarditis-Like Acute Cardiac Syndrome in COVID-19

Epicardial Fat Inflammation in Severe COVID-19

Italian Institute of Health (Istituto Superiore di Sanità I. Overweight and Obesity, national data

Risk factors for SARS-CoV-2 among patients in the Oxford Royal College of General Practitioners Research and Surveillance Centre primary care network: a cross-sectional study

Obesity Paradox" in Acute Respiratory Distress Syndrome: Asystematic Review and Meta-Analysis

Inflammatory genes in epicardial fat contiguous with coronary atherosclerosis in the metabolic syndrome and type 2 diabetes: changes associated with pioglitazone

Statins Reduce Epicardial Adipose Tissue Attenuation Independent of Lipid Lowering: A Potential Pleiotropic Effect

Statin Use and In-Hospital Mortality in Diabetics with COVID-19

Cardiovascular Implications of Fatal Outcomes of Patients With Coronavirus Disease 2019 (COVID-19)

Coronary and total thoracic calcium scores predict mortality and provides pathophysiologic insights in COVID-19 patients

Continuous variables are expressed as median (25 th and 75 th percentile). Categorical variables are expressed as absolute values (%). ARDS, acute respiratory distress syndrome

BMI, body mass index

C-reactive protein; eGFR, estimated glomerular filtration rate, calculated with the CKD-EPI equation

PaO 2 /FiO 2 , arterial partial pressure of oxygen measured on arterial blood gas analysis/fraction of inspired oxygen. *ARDS defined according to the Berlin criteria 22