Cardiac remodelling in non‐alcoholic fatty liver disease in the general population

Non‐alcoholic fatty liver disease (NAFLD) is associated with increased risk for cardiovascular disease. Our study investigates the contribution of NAFLD to changes in cardiac structure and function in a general population.


| INTRODUC TI ON
Obesity is a major global disease burden with a steady increasing prevalence over the past decades in both men and women. 1 As a result, there is a likewise increase in obese related conditions like non-alcoholic fatty liver disease (NAFLD) and cardiovascular disease (CVD). 2,3Besides obesity, both NAFLD and CVD share multiple risk factors like diabetes and dyslipidaemia.However, there is a growing evidence that NAFLD might contribute to CVD progression as an isolated risk factor.Larger past cohort studies suggest an independent association of hepatic steatosis with various CVD like aortic valve sclerosis, 4 carotid arteriosclerosis 5 and left ventricular remodelling and dysfunction [6][7][8] which counts as a precursor of heart failure. 9wever, it was shown that both obesity 10,11 and NAFLD 7,8,12 are associated with subclinical cardiac remodelling so the individual role of NAFLD as a risk factor still remains controversial.Previous cohort studies usually used qualitative liver fat content (LFC) determination using ultrasound or qualitative computed tomography imaging. 7,13ile this is the most practical approach for larger studies, in the daily clinical practice it remains unclear if the use of quantitative liver fat allows further insights in their interactions.Liver biopsy is the gold standard for quantitative assessment of liver fat and disease stage but as an invasive procedure it is not suitable for larger general population studies.During the past decade, magnetic resonance imaging (MRI) has proven as a noninvasive way of determining LFC using proton density fat fraction (PDFF) mapping showing very good correlation with the histopathology. 14Likewise cardiac magnetic resonance imaging (CMR) emerged as the reference standard for assessment of cardiac structure and function with less inter-observer variability and dependency of anatomic conditions compared to echocardiography. 15erefore, the purpose of our study was to examine the relationship between NAFLD and left cardiac structure and function in a large general population cohort using whole-body MRI for quantitative LFC and cardiac measurements.

| Study population and general measurements
The gave written informed consent for the study and MR imaging.The complete study design has been described in detail before. 16,17ily alcohol consumption was determined by beverage specific alcohol consumption (beer, wine, distilled spirits) on the last weekend and last weekday before the examination.Mean daily alcohol consumption was calculated using beverage specific pure ethanol volume proportions.Smoking status of the participants was divided into categories of current, former and never smokers.Individuals who participated in physical training during summer and/or winter less than 1 h/week were classified as being physically inactive.Waist circumference was measured to the nearest 0.1 cm suing an inelastic tape midway between the lower rip margin and the iliac crest.Body mass index (BMI) was calculated by dividing the weight in kilograms (kg) by height in metres squared (m 2 ).Hypertension was defined as systolic blood pressure ≥ 140 mmHg and/or diastolic blood pressure ≥ 90 mmHg or the use of antihypertensive medication.Type 2 diabetes was defined as self-reported history or use of hypoglycaemic medication (ATC-codes A10A, A10B, A10X).Plasma serum levels of glucose, Haemoglobin A 1c (HbA1C), N-terminal prohormone of brain natriuretic peptide (NT-proBNP), total cholesterol, and high and low density lipoprotein were measured on non-fasting blood samples.Treated hypertension was defined as use of antihypertensive medication (ATC-codes C01D, C02, C03, C04, C07, C08, C09).
Lipid-lowering medication was defined as treatment with statins, fibrates or other lipid-lowering medications (ATC-codes C10).
Exclusion criteria were missing image data or variables, bad image quality, excessive alcohol consumption (>30 g/day for men, >20 g/day for women), history of myocardial infarction or heart surgery and severe heart failure (left ventricular ejection fraction [LVEF] <30%; Figure 1).

| Magnetic resonance imaging and image analysis
Whole-body MRI was performed on a 1.5T magnetic resonance system (Magnetom Avanto; Siemens Medical Systems, Erlangen, Germany) using a whole-body-surface coil system.The independent readers were blinded for clinical covariates.The complete imaging protocol was described in detail before. 18alysis of LFC has been published before. 19In short LFC was determined by offline reconstructions of a PDFF map.Mean PDFF values were determined by a single radiologist with 11 years experience of abdominal MR imaging using operator-defined regions of interest placed at the centre of the liver with Osirix (v3.8.1; Pixmec Sarl, Bernex, Switzerland).NAFLD was defined as LFC higher than 5.1% and cutoffs for mild, moderate and severe NAFLD were 5.1%, 14% and 28%, respectively, based on histopathologic grading from a previous study. 20rdiac assessment of morphological and functional left ventricular and left atrial parameters was ascertained using semi-automatic software (QMass MR 7.2, MEDIS, Leiden, The Netherlands) by two readers with 3 and 5 years of CMR imaging experience using steadystate free precession cine imaging.The detailed image analysis for LV myocardial mass (LVM), LV end-systolic volume (LVESV), LV enddiastolic volume (LVEDV), LVEF and left atrial end-diastolic volume (LAEDV) has been described before. 21LV concentricity index (LVC) was calculated as LVM/LVEDV.LV stroke volume (LVSV) was calculated as LVED-LVESV.Left ventricular end-diastolic wall thickness was calculated as the average of all measured segments.Volumetric measurements for LV and LA and LVM were further indexed to body height 2.7 . 22Global longitudinal strain (GLS), global transversal strain, global circumferential strain and global radial strain were analysed using semi-automatic software (2D Cardiac Performance Analysis Software Version 1.0, TomTec Imaging Systems GmBH, Germany) by two independent readers using short axis, 2-chamber and 4-chamber steady-state free precession cine imaging.

| Statistical analysis
Descriptive statistics were reported stratified by presence of NAFLD as median, 25th and 75th percentile (continuous data) or as absolute numbers and percentages (categorical data).NAFLD was associated with markers from CMR by linear regression models adjusted for age, showing inclusion and exclusion of study participants.sex, height 2.7 , antihypertensive medication intake, systolic blood pressure, type 2 diabetes, smoking status and study.Adjusting for BMI was waived because liver fat content as a continues variable already shows correlation with those. 19,23All models were repeated stratified by sex.
A p < .05 was considered as statistically significant.All analyses were conducted with Stata 17.0 (Stata Corporation, TX, USA).

| Characteristics of the study population
After exclusion 1096 adults with a mean age of 50.9 years (49.3% female, range 21-82 years) were eligible for the analyses (Figure 1).
The baseline characteristics of the study sample divided by the presence of NAFLD are summarized in Table 1.The prevalence for NAFLD was 35.9%.Participants with NAFLD were more likely male, older, less physically active and had higher BMI and waist circumferences.The presence of NAFLD was accompanied by a worse cardiovascular risk profile including increased percentage of hypertension, type 2 diabetes and unfavourable triglycerides/ HDL-cholesterol ratio.Accordingly, participants with NAFLD took more often medication for hypertension, diabetes and hyperlipidaemia compared to participants without NAFLD.Stratified by sex, women in our study population had less alcohol consumption, less hypertension and diabetes as well as lower liver fat content compared to men (Table S1).
Unadjusted comparisons of cardiac structure and function are summarized in Table 2. Study participants with NAFLD showed on average a higher LVM, LVC, LVED wall thickness and LAEDV.
When stratified by sex, there were significant differences in the associations of NAFLD in men and women with left cardiac geometry and function.
In order to investigate if the grade of liver fat has an influence on our observations we did a further analysis in which we divided the study sample by no, mild and moderate to severe NAFLD with the results summarized in Table 4.We observed similar results to the overall NAFLD study sample in

| DISCUSS ION
In this large population-based MRI study using data from SHIP, NAFLD was significantly associated with higher LVM, LVED wall thickness, LVC and LAEDV and a lower LVEDV after adjusting for demographic and classical cardiac risk factors.Moreover, we observed significant different patterns in men and women.Compared to women, men with NAFLD showed lower LVEDV and LVSV, while women showed a higher LVM, LV cardiac output and LAEDV and a lower GLS.Compared to mild, moderate to severe NAFLD was, in general, strongly associated with higher LVM, LVC and LVED wall thickness.

| Prevalence of NAFLD in the study cohort
The prevalence of NAFLD in our study cohort was high with 35.9%.A recent systematic review by Younossi et al. found an average pooled prevalence of 25.10% for Western Europe (20.55%-30.28%). 24However, the majority of included studies use ultrasound for qualitative liver fat quantification which has a significantly lower sensitivity for mild hepatic steasosis. 25So we hypothesize that there might be a relevant amount of false negative individuals in ultrasound studies.

| NAFLD and cardiac geometry and function
Only a small number of studies used the combination of quantitative MRI liver fat content imaging and CMR to study their associations.
A recent analysis from Schlett et al. 26  inconsistencies may be explained by different sample size and composition of the study cohorts.While our overall prevalence of NAFLD was 35.9%, it was 19.5% in the UK Biobank study, which suggests that the UK Biobank study might exhibit a healthy volunteer bias. 29ere are several studies showing significant associations between fatty NAFLD and diastolic dysfunction. 7,12,13While those studies used echocardiography data and therefore are not directly comparable to our study, we observed higher LAEDV in women with NAFLD which might be used as an indirect MRI marker for diastolic dysfunction. 30Since diastolic dysfunction can be a precursor for development of heart failure, 31,32 prospective studies are needed to confirm the clinical relevance of our findings.

| Sex specific cardiac remodelling in NAFLD and obesity
Our results suggest significantly different patterns in cardiac structure and function in the presence of NAFLD between males and females.
In our cohort study, men with NAFLD show a more concentric left ventricular hypertrophy tendency with lower LVEDV and LVSV compared to men without NAFLD, while women with NAFLD showed higher LVM, LV cardiac output and LAEDV and a lower GLS compared to women without NAFLD.The literature suggests sex specific differences for cardiac remodelling in response to different obesity conditions.Rider et al. 33 reported sex specific differences in left ventricular remodelling in subjects with obesity with men showing a concentric hypertrophic remodelling while women had either concentric or eccentric hypertrophic remodelling.However, they only took the BMI into account and no other fat compartments.Another study of van Hout et al. 34 using data from the UK biobank showed a smaller LVEDV in visceral obesity in both men and women while only men showed an inverse association with left ventricular systolic function.
Compared to previous studies, 7 our results show a lower (more negative) GLS in women with NAFLD.Two recent studies indicate a sex specific response of GLS in obesity, determined by BMI, with a lower GLS in women. 35,36A possible explanation could be the different composition of body fat departments in men and women.
Typically, men show a higher percentage of visceral adipose tissue which is associated with a concentric cardiac remodelling. 37Men also show a higher epicardial adipose tissue volume which is asso-

| Influence of NAFLD severity on cardiac remodelling
In the overall study population, we observed stronger associations for higher LVM, LVC and LVED wall thickness when stratified by NAFLD severity using 14% LFC as a cutoff.
A recent study from Dennis et al. 38 has shown that MRI derived PDFF correlates with the histopathologic NAFLD activity score (NAS).In their study, a median PDFF of 15% or higher was associated with a NAS ≥ 5 which indicates more likely the presence of inflammatory non-alcoholic steatohepatitis. 39The literature suggests an inflammatory pathway as one part of the interplay between NAFLD and CVD with NAFLD leading to increased serum levels of inflammatory cytokines like Interleukin-6 and tumour necrosis factor alpha. 40,41 Increased serum levels of interleukin-6 and tumour necrosis factor-a are also observed in patients with heart failure 42 and show adverse cardiac remodelling in mice models. 43,44Therefore our observations might be partially mediated by a more pronounced inflammatory process in subjects with a higher LFC.Further investigation is needed to proof an increased inflammatory activity in NAFLD subjects in our study cohort since cytokines were not part of the laboratory testing.

| Metabolic dysfunction-associated steatotic liver disease
This year, a multinational consensus statement revised the nomenclature for fatty liver disease. 45Here, the term metabolic dysfunction-associated steatotic liver disease (MASLD) was chosen to address the underlying disease mechanic and for less stigmatization regarding alcohol consumption.The new definition is based on the presence of hepatic steatosis accompanied by at least one classic cardiometabolic risk factor like obesity, diabetes, hypertension or hyperlipidaemia.According to the criteria proposed by the consensus statement, our data showed 383 out of 391 participants (98.0%) with NAFLD also meeting the criteria for MASLD (Table S2).The difference should not be statistically significant for our present study.

| LI M ITATI O N S
Our study has at least five limitations.First, the study population based on a white Caucasian population sample in Northeast Germany.Second, even though using MRI has major advantages regarding determination of cardiac geometry and functions and liver fat content, the strict exclusion criteria in our study may have led to a rather healthy population so we cannot rule out a selection bias completely.Third, while MRI allows precise quantification of liver fat content to distinguish between healthy controls and NAFLD, we were only able to graduate by liver fat content so we do not have further information regarding possible liver fibrosis.
Fourth, we don't have information about valvular heart disease in our study population so we cannot fully rule out a partial remodelling effect due to these in certain individuals.However, the SHIP cohorts are rather healthy so we assume the total amount of higher grade valvular disease is low.Finally, the cross-sectional study design does not allow to draw conclusions regarding a causal relationship between fatty liver disease and cardiac alterations.Other unknown confounders might influence our results.
Longitudinal studies are needed to study the long-term results and clinical significance of the findings.

| CON CLUS ION
In our study sample, NAFLD was associated with an unfavourable left cardiac remodelling showing an overall hypertrophic like remodelling with different associations between men and women.While men showed a concentric remodelling tendency, women showed higher LVM, LV cardiac output and LAEDV and lower GLS.Larger studies using quantitative liver fat imaging taking different body fat compartments and sex into account are needed to confirm our results.
Study of Health in Pomerania (SHIP) is a population-based study in Northeast Germany.For the present study, we used data gathered between September 2008 and September 2012 from the second follow-up investigation cycle of the first cohort (SHIP-START-2) and from baseline examinations of a second cohort (SHIP-TREND-0), when whole-body MRI was first introduced.Overall 6753 Caucasian individuals aged 20 to 81 years were eligible for this study.All subjects were randomly selected out of the local registration office database without individual overlap.For all participants in both cohorts, computer-assisted interviews were performed at baseline and during follow-up examinations including questions about their medical history, lifestyle behaviours and drug usage.Furthermore, various clinical and laboratory examinations were also ascertained.MRI took place close to time of the interviews and physical examinations.The study was approved by the local institutional review board and complies with the Declaration of Helsinki.All study participants

TA B L E 4
ciated with a higher (more positive) GLS underlining the influence of different fat compartments.More studies taking quantitative LFC into account are needed to confirm whether liver fat as an individual fat compartment has sex specific influence on cardiac remodelling.Associations of liver fat content and NAFLD with CMR measurements.

Table 2
Characteristics of the study population stratified by NAFLD.
when comparing participants TA B L E 1
using data from the KORA-Associations of liver fat and NAFLD with CMR measurements.
28current large study from Jamialahmadi et al.28using MRI data from the UK Biobank with 18 848 individuals is partially in concordance with our results.They also confirmed a higher LVC and lower LVEDV in subjects with higher LFC.Contrary to our results, they observed an inverse association for LVM and LA volume index.TheseTA B L E 3Note: Coefficients are derived from linear regression models adjusted for age, sex, body height 2.7 (except for indexed values), systolic blood pressure, intake of antihypertensive medication, type 2 diabetes mellitus, smoking status and study.Indexed: values are indexed to body height 2.7, CI confidence interval.*p < .05.