Fatty liver disease, heart rate and cardiac remodelling: Evidence from the UK Biobank

Growing evidence supports an association between fatty liver disease (FLD) and cardiac dysfunction and remodelling, leading to cardiovascular disease and heart failure. Herein, we investigated the independent contribution of FLD to cardiac dysfunction and remodelling in participants from the UK Biobank with cardiac magnetic resonance (CMR) data available.


| INTRODUC TI ON
Fatty liver disease (FLD) has become the leading cause of chronic liver disease worldwide, affecting approximately 25% of the global population. 1 FLD is even more prevalent in individuals with additional metabolic risk factors, namely obesity and type 2 diabetes.
Indeed, more than two thirds of individuals with type 2 diabetes have FLD and this rate raises up to 90% in those with morbid obesity undergoing weight-loss surgery. 2 A growing body of evidence shows that FLD is a strong risk factor for cardiovascular disease (CVD). Nearly one in two individuals with FLD have CVD which remains the first cause of death in these individuals. 3 There has been a debate whether the association between FLD and CVD is mediated by residual confounding cardiovascular risk factors associated with metabolic syndrome. However, more recent data suggest that FLD is not only associated with an increased risk of CVD but also independently increases the risk of cardiovascular events, including stroke, congestive heart failure and cardiovascular mortality. [4][5][6][7] Cardiac remodelling is the earliest response to cardiac injury and cardiac load, eventually resulting in the development of heart failure. 3,8 Growing evidence supports the coexistence of FLD, cardiac dysfunction and remodelling. 3,9 Multiple and complex mechanisms may increase the risk of cardiac remodelling in individuals with FLD, including insulin resistance, systemic inflammation, oxidative stress, as well as activation of the renin-angiotensin system and sympathetic nervous system. 3,5 Specifically, sympathetic nervous system activation is an early sign of cardiac injury and cardiac load in heart failure. However, the pathophysiological mechanism by which FLD affects cardiac function and remodelling remains to be elucidated. 10 Indeed, as part of cardiac remodelling, diastolic dysfunction is frequently present among individuals with FLD. Furthermore, diastolic dysfunction contributes to the onset of heart failure with preserved ejection fraction. 3 Within this context, in a cross-sectional study of approximately 2700 middle-aged adults, FLD detected by computed tomography was associated with features of subclinical left ventricular (LV) remodelling and early LV diastolic and systolic dysfunction as assessed by echocardiography. 11 However, most studies on FLD and cardiac function included cohorts with small sample-size where cardiac structure and function were measured using echocardiography instead of cardiac magnetic resonance (CMR), the gold standard technique for the non-invasive assessment of functional and volumetric cardiac parameters. 12,13 Therefore, in this work, we aim at investigating the independent contribution of FLD to cardiac dysfunction and structural remodelling in more than 18 000 participants from the UK Biobank study in whom heart structure and function were measured by CMR.

| UK Biobank
The UK Biobank is a large prospective cohort study recruiting more than 500 000 participants (age 40-69 years) between 2006 and 2010 from 22 assessment centres throughout the UK. 14 Clinical and laboratory data were collected using highly standardized pro-

| Study cohort
First, we selected unrelated participants of European ancestry based on our quality control pipeline which has been described in detail previously. 15 We added participants who self-reported as being 'Irish' or 'Any other white background' to the subset of white British ancestry. Next, we excluded individuals (1) with more than 10 putative third-degree relatives, (2) with a mismatch between their self-reported and genetically inferred sex, (3) having putative sex chromosome aneuploidy, (4) who had withdrawn consent, and (5) were identified by the UK Biobank as outliers based on heterozygosity and missingness. Then, we excluded individuals reporting chronic viral hepatitis (International Classification of Diseases 10th edition [ICD-10] B18-B19) and valvular heart diseases (ICD-10 I34.0, I35.0) from hospital admissions and death registry. Finally, we included participants with liver magnetic resonance imaging (MRI) proton density fat fraction (PDFF) and CMR data available. A total K E Y W O R D S cardiac magnetic resonance, cardiovascular disease, heart failure, non-alcoholic fatty liver disease, sympathetic nervous system

Key points
In Europeans from the large UK Biobank, fatty liver disease is an independent predictor for higher heart rate and cardiac remodelling with reduced ventricular volumes as assessed by cardiac magnetic resonance. Fatty liver disease may be associated with a sympathetic nervous system activation and an early cardiac concentric remodelling as well as hypertension and other well-established cardiovascular risk factors, potentially leading to heart failure. of 18 848 individuals with MRI-PDFF and CMR were included in the final analyses.

| Cardiac and liver magnetic resonance imaging
At the imaging visit, participants underwent cardiac and abdominal MRI at an imaging station covering both cardiac and abdominal examinations. 16 The CMR protocol and image analysis in the UK Biobank have been described in detail elsewhere. 17  and RA maximal volume index respectively. The eccentricity ratio, a measure of concentric remodelling, was calculated by dividing LV mass by LV end-diastolic volume. 18,19 The remodelling index, a CMR marker of advanced LV hypertrophy, was calculated by dividing the cube root of LV end-diastolic volume by the maximal wall thickness across 16 LV myocardial segments described by the American Heart Association (AHA). 20 All cardiac endpoints were obtained from data returned by Bai et al. 21 (return ID 2383).
The abdominal MRI protocol followed the cardiac examination on the 1.5 Tesla scanner. Briefly, a multi-echo-spoiled gradient-echo acquisition was used to quantify liver fat. Data were analysed using LiverMultiScan© Discover software. FLD was defined as steatosis by PDFF >5.5%. 22 Type 2 diabetes was defined by self-reported history or hospital diagnosis of type 2 or unspecified diabetes (ICD-10 E11, E14), treatment with insulin or oral hypoglycemic drugs, serum glucose level ≥ 11.1 mmol/L (200 mg/dL), and/or HbA1c ≥48 mmol/mol (6.5%). 23 Dyslipidemia was defined as self-reported history of high cholesterol, hospital diagnosis of hyperlipidemia (E78.0, E78.1, E78.2, E78.4, E78.5) or use of lipid-lowering drugs. Arterial hypertension was defined as self-reported history or hospital diagnosis of hypertension (ICD-10 I10), or use of anti-hypertensive drugs.
Atrial fibrillation and heart failure were defined as self-reported history or hospital diagnosis of atrial fibrillation (ICD-10 I48) and heart failure (ICD-10 I50), respectively.
Data on smoking, alcohol consumption and physical activity were collected through a touch-screen questionnaire. Smoking status was categorized into two groups: current smoking and never/ former smoking. A detailed explanation of alcohol consumption pipeline has been provided elsewhere. 23 Briefly, frequency of daily alcohol consumption (g/day) was quantified based on the average weekly or monthly alcohol intake. Alcohol grams for each type of drink were derived from the corresponding reference alcohol content (1 unit of alcohol = 8 g of alcohol). Excessive alcohol consumption was defined when daily alcohol intake was ≥30 g and ≥ 20 g for men and women, respectively. 24 Regarding physical activity, participants were categorized into two groups if they underwent physical exercise or not according to the UK physical activity recommendations (i.e. ≥150 min/week and ≥ 75 min/week for moderate and vigorous physical activity, respectively).

| Statistical analyses
All statistical analyses were performed using the software R, ver- Continuous variables were shown as mean ± SD or median (interquartile range [IQR]) as appropriate. Categorical variables were shown as number(percentage). Generalized linear models were used to test the association between FLD and CMR parameters as dependent variables. All multivariable models were adjusted for age, sex, BMI, hypertension, CAD, type 2 diabetes, dyslipidemia, atrial fibrillation, heart failure, smoking, alcohol consumption and physical activity. Additionally, whole-body fat mass and free-fat mass were further included among the covariates instead of BMI.
Sensitivity analyses were performed stratifying by hypertension, type 2 diabetes status, two clinical scores for advanced fibrosis (i.e., 25 and fibrotic NASH (i.e., fibrotic NASH index [FNI]), 26 and excluding from the overall cohort individuals with cardiovascular disease. A value of p < 0.05 was considered statistically significant.
Linear regression models with regularization were used to generate predictive models for heart-related outcomes. For each outcome, a 10-fold nested cross-validation (CV) was performed for model se-

| Clinical characteristics
We examined a total of 18 848 individuals of whom 3681 (19.5%) fulfilled the criteria for FLD (PDFF>5.5%). Individuals with FLD were in average younger, more often male and had higher BMI, glucose, systolic blood pressure and circulating triglycerides (p < 0.001) ( Table 1).
Consistently, FLD was associated with higher risk of dyslipidemia, hypertension and type 2 diabetes, after adjusting the analyses for age, sex and BMI (p < 0.001).

| FLD and cardiac structure and function
FLD was independently associated with a higher average heart rate (p < 0.001) ( Table 2 and Table S1). Further, FLD was associated with higher cardiac remodelling (i.e., higher eccentricity ratio and lower remodelling index), lower LV and RV end-systolic and enddiastolic volumes, and lower LV and RV stroke volumes (p < 0.001).

| Independent predictors of cardiac structure and function
To understand the prediction capability and the relationship between the covariates, we used three different penalized regression models. In these analyses, we included FLD and major cardiometabolic risk factors, namely age, sex, BMI, hypertension, CAD, type 2 diabetes, dyslipidemia, atrial fibrillation, heart failure, smoking, alcohol consumption and physical activity. Standardized effects from the best model performed in the nested cross-validation step were used as a proxy of their importance in predicting each of the main cardiac outcomes (Figure 1 and Table S7). The strongest positive predictor for average heart rate was FLD followed by age, hypertension and type 2 diabetes. The strongest negative predictors were male sex and the presence of CAD. Male sex was the strongest positive predictor for eccentricity ratio followed FLD, age, hypertension and BMI. Moreover, FLD had a null effect in predicting LV myocardial mass index which was mostly influenced by sex, hypertension and BMI. For LV volumes, FLD was the strongest negative predictor along with age, while male sex was the strongest positive predictor.
As a sensitivity analysis, we used different regularizations (L1 in LASSO, L2 in Ridge and both in Elastic Net) which did not change the contribution of each covariate in predicting the heart outcomes as indicated by a similar MSE on test set ( Figure 1). Moreover, LASSO did not shrink any coefficient to be exactly zero, suggesting all risk factors examined in our analyses contribute to the generalizability of the predictive models.

| DISCUSS ION
In this work, we find that, as for hypertension and other wellestablished cardiovascular risk factors, FLD is associated with increased cardiac remodelling and reduced LV volumes. Further, our findings show that FLD is a strong independent predictor of higher heart rate and eccentricity ratio, and of lower LV volumes.
Among all the variables related to cardiac function, FLD was the strongest positive predictor for heart rate. One hypothesis is that this may be caused by an obesity-related sympathetic nervous system activation, directly leading to an increase in heart rate. 27 Sympathetic nervous system activation is an early sign of heart failure, 10 but it is also present in those with hypertension and type 2 diabetes, two diseases often clustering with FLD which may partially explain this finding. 28 However, our analyses showed that FLD had the strongest effect on heart rate independently of BMI, type 2 diabetes and hypertension.
Alternatively, the increase in heart rate may be due to changes in the hepatic splanchnic circulation 29 (Figure 2). This circulation is responsible for a substantial amount of the cardiac preload. In this model, intracellular lipid droplet accumulation and ballooning would directly restrict the hepatic sinusoid volume impeding the outflow of venous blood to the central vein and therefore to the inferior vena cava, causing a reduction in the preload. 30 Therefore, a sympathetic TA B L E 1 Clinical characteristics of UK Biobank participants with CMR data stratified by FLD status (N = 18 848

TA B L E 1 (Continued)
nervous system activation would serve as a compensatory mechanism restoring the cardiac preload and the increase in heart rate may be considered as an epiphenomenon of these events. At the same time, it is also possible the sympathetic nervous system activation may be compensating for a smaller heart with less volume capacity to maintain the blood flow per minute. Finally, male sex and CAD were found to be negative predictors for heart rate, most likely due to a higher resting heart rate in women and the use of beta-blockers in individuals after CAD.
We observe that FLD is an independent predictor of lower LV volumes. Furthermore, we find that FLD is a strong independent predictor of higher cardiac remodelling, namely higher eccentricity ratio and lower remodelling index. Notably, eccentricity ratio, i.e., a ratio between LV mass and LV end-diastolic volume, is a measure of concentric remodelling that has been proposed as a useful indicator of LV adaptation and future CVD in CMR. 18,19,31,32 Similarly, the newly developed remodelling index is inversely correlated with LV hypertrophy (i.e., the lower index, the higher myocardial injury) and was found to predict the occurrence of adverse cardiovascular events in individuals with hypertension. 20 Previous work from the CARDIA study has similarly shown a reduction in LV enddiastolic volume measured with echocardiography in individuals with FLD. In this study, unlike our results, an increase in LA volume index was found in individuals with FLD. However, it is important to bear in mind that in the CARDIA study these results were no longer significant when adjusted for adiposity and heart failure risk factors. Moreover, these inconsistencies may be explained by different nature of these study in terms of sample size, ethnicity, prevalence of diabetes, obesity and cardiovascular disease. 11 The UK Biobank study participants were middle-aged and presumably healthy. Therefore, our results may reflect an early phase

F I G U R E 2
Contribution of fatty liver disease to cardiac dysfunction and structural remodelling. Fatty liver disease is an independent predictor of higher heart rate, higher cardiac remodelling, and lower left ventricular volumes. The positive association between fatty liver disease and heart rate may be related to a sympathetic nervous system activation secondary to sinusoidal compression by enlarged fatty hepatocytes, resulting in a decrease in sinusoidal flow to the central vein and therefore to the inferior vena cava, ultimately leading to a decrease in preload.
the general population and find that FLD is associated with reduced LV and RV cavity, reduced LA and RA maximal volume index, an increase in the eccentricity ratio as well as a decrease in the remodelling index, altogether suggesting an impact on total cardiac structure and remodelling in an early phase of FLD. Several different mechanisms have been proposed which may impact cardiac remodelling in FLD besides activation of the renin-angiotensin and sympathetic nervous system. Cardiac accumulation of ectopic fatty tissue including myocardial fat and adipose tissue surrounding the heart is a central link between FLD and CVD. 34 The epicardial adipose tissue shares microcirculation with the myocardium and it has been shown that when there is an accumulation of epicardial fat, induced by obesity or FLD, the biological characteristics of epicardial fat alters and synthesize proinflammatory cytokines that activate the inflammation and profibrotic pathways. 35 This may lead to structural changes and remodelling of the myocardium and ultimately result in the development of atrial fibrillation and heart failure with preserved ejection fraction.
The strength of our study is that cardiac function and structure were measured by CMR, the gold standard technology to assess these measurements, and the large sample size from the general population. This is the first time that all these measurements are done simultaneously and allowed us to generate a model explaining the reason for FLD to be associated with heart rate. Limitations of this study are (a) its cross-sectional nature, (b) the inclusion of participants of only European ancestry, and that (c) we did not use echocardiography, the golden standard for the non-invasive assessment of diastolic dysfunction, but data from CMR. Moreover, it is fair to say that comparison between measurements from echocardiography and CMR are difficult to make.
In conclusion, we find that FLD is an independent predictor of higher heart rate and remodelling associated with lower LV volumes.
Further longitudinal and molecular studies are warranted to understand the mechanisms behind this association.

ACK N O WLE D G E M ENTS
We thank the staff and the participants of the UK Biobank study.
This research has been conducted using the UK Biobank resource (application #37142).

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no financial or other relationships with drug manufacturers that could lead to a conflict of interest.

E TH I C S S TATEM ENT
The UK Biobank study has been approved by the National Research