The Association of Non-Alcoholic Fatty Liver Disease and Cardiac Structure and Function – Framingham Heart Study

Abstract

Background & Aims:

Non-alcoholic fatty liver disease confers increased risk for cardiovascular disease, including heart failure, for reasons that remain unclear. Possible pathways could involve an association of liver fat with cardiac structural or functional abnormalities even after accounting for body size.

Methods:

We analyzed N=2,356 Framingham Heart Study participants (age 52±12 years, 52% women) who underwent echocardiography and standardized computed tomography (CT) measures of liver fat.

Results:

In cross-sectional multivariable regression models adjusted for age, sex, cohort, and cardiovascular risk factors, liver fat was positively associated with left ventricular (LV) mass (β=1.45; 95% confidence interval (CI): 0.01,2.88), LV wall thickness (β=0.01; 95% CI: 0.00,0.02), mass volume ratio (β=0.02; 95% CI 0.01,0.03), mitral peak velocity (E) (β=0.83; 95% CI 0.31,1.36), and LV filling pressure (E/e’ ratio) (β=0.16; 95% CI 0.09,0.23); and inversely associated with global systolic longitudinal strain (β=0.20, 95% CI 0.07,0.33), diastolic annular velocity (e’) (β=−0.12; 95% CI −0.22,−0.03), and E/A ratio (β=−0.01; 95% CI-0.02,−0.00). After additional adjustment for body mass index (BMI), statistical significance was attenuated for all associations except for that of greater liver fat with increased LV filling pressure, a possible precursor to heart failure (β=0.11; 95% CI 0.03,0.18).

Conclusion:

Increased liver fat was associated with multiple subclinical cardiac dysfunction measures, with most of associations mediated by obesity. Interestingly, the association of liver fat and LV filling pressure was only partially mediated by BMI, suggesting a possible direct effect of liver fat on LV filling pressure. Further confirmatory studies are needed.

LAY SUMMARY

It is unclear if increased liver fat is associated with subclinical cardiovascular disease after accounting for obesity. In our study, increased liver fat was linked with multiple echocardiographic markers of subclinical cardiac dysfunction, with most of associations mediated by obesity. The association of liver fat and LV filling pressure was only partially mediated by BMI, suggesting that increased liver fat may directly affect LV filling pressure, a clinical precursor to heart failure.

Non-alcoholic fatty liver disease (NAFLD) and cardiovascular diseases, particularly heart failure (HF), are obesity-related conditions^1,2 that have increased in prevalence along with the obesity epidemic over the last two decades^3–5. Those with NAFLD can progress to develop end-stage liver disease, though cardiovascular disease (CVD) remains the leading cause of morbidity⁶ and mortality⁷. Growing literature suggests that individuals with NAFLD manifest myocardial functional and structural changes, leading to cardiac remodeling and increased risk of HF^8–13. NAFLD and HF share many common risk factors, including diabetes mellitus, obesity, hypertension, and the metabolic syndrome^14–20. Mechanisms that link NAFLD to HF, beyond shared risk factors, have largely been unexplored. In NAFLD, circulating free fatty acids and triglycerides accumulate in liver and myocardial cells with resultant increased myocardial fatty acid oxidation. Fatty acid oxidation is less efficient than glucose metabolism, and this metabolic inefficiency may contribute to the development of HF²¹. Additionally, inflammation and oxidative stress induced by NAFLD may contribute to cardiac insulin resistance²² and cardiac fibrosis²³, leading to altered cardiac structure and HF. However, obesity may contribute to HF via similar mechanisms.

Although multiple studies have observed associations with NAFLD and left ventricular (LV) dysfunction^12,24–41, right ventricular (RV) dysfunction^42,43 and LV hypertrophy⁴⁴, some results are conflicting^45,46, most include small sample sizes or narrow populations, or did not adequately account for potential confounding variables. Additionally, prior studies have not explored the possible mediation effect of general adiposity, as measured by body mass index (BMI), on the association between NAFLD and subclinical CVD.

Thus, we investigated the association between NAFLD and subclinical CVD as defined by abnormal cardiac structure and function via standard and speckle-tracked-based echocardiography in the Framingham Heart Study (FHS). We hypothesized that NAFLD is associated with subclinical CVD, even after accounting for shared risk factors, and that the association is at least partially mediated by adiposity.

Methods

The study sample was derived from the FHS Third Generation and Offspring Cohort participants, as described previously^47,48. A subset of Third Generation and Offspring participants underwent multi-detector computed tomography (CT) that assessed liver fat and visceral adipose tissue (VAT) between 2002 and 2005⁴⁹. Third Generation participants at exam 1 (2002–2005) and Offspring cohort participants at exam 8 (2005–2008) completed a standardized medical history, laboratory evaluation and echocardiography assessment with digital image acquisition and speckle-tracking analyses. We excluded participants with excessive alcohol use (>7 drinks per week for women and >14 drinks per week for men) (n=502), missing liver fat or covariates (n=254), missing echocardiogram (n=5), and overt CVD (prevalent history of angina, acute coronary syndrome, coronary artery disease, or HF) (n=180). Informed consent was provided by all participants prior to attending the examination cycle. This study was approved by the Institutional Review Board of Boston University Medical Center and the Massachusetts General Hospital.

The CT protocol has been described previously to measure liver fat and VAT^49–51. In this protocol, 25 contiguous slices of 5 mm thickness were obtained, and a phantom calibration control (Image Analysis, Lexington, KY, US) was placed under each participant. The average fat attenuation in Hounsfield Units (HU) was measured in three areas of the liver and divided by the calibration phantom (HU) to derive the liver phantom ratio (LPR). With increasing liver fat, the LPR decreases. We defined hepatic steatosis using a cut off of an LPR ≤0.33, as in prior studies^50,51. The LPR was utilized as the indexed standard as the spleen was not visualized on all scans. A liver spleen ratio of 1.1 corresponds to 30% hepatic steatosis as described previously^50,51.

To measure VAT volume, we identified pixels containing fat using an image display window of −195 to −45 HU and a window center of −120 HU on a dedicated workstation (Aquarius 3D Workstation, TeraRecon Inc, San Mateo, California), as described^49,51. We manually traced the separation between the visceral and subcutaneous fat compartments with high intra-reader and inter-reader correlation coefficients of 0.99⁴⁹.

All measurements were attained during a standardized medical history and examination session for participants in exam 1 (Third Generation) or exam 8 (Offspring). Participants were considered current smokers if they smoked at least 1 cigarette daily during the previous year. Alcohol use was recorded as drinks/week or drinks/month as reported on a clinician-administered questionnaire. Plasma glucose, triglycerides, high density lipoprotein (HDL) cholesterol and total cholesterol were measured on fasting morning samples. Diabetes was defined as fasting plasma glucose ≥126 mg/dL or treatment with insulin or hypoglycemic agent. Hypertension was defined as systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥ 90 mm Hg or the use of anti-hypertensive medications. BMI was calculated by dividing the weight in kilograms (kg) by height in meters squared (m²).

All participants in the study underwent routine M-mode, 2-dimensional (2D) and pulse-wave Doppler echocardiography using a Hewlett-Packard 5500 machine (Philips Healthcare, Andover, MA)⁵². The mean time between CT scan and echocardiogram was 9 months. All echocardiograms were evaluated by an experienced sonographer or cardiologist who was blinded to the participant’s demographic and clinical characteristics using a standardized protocol. Cardiac structure measures included: LV end-diastolic diameter and volume, LV end-systolic diameter and volume, LV wall mass and thickness, relative wall thickness (LV wall thickness/LV end-diastole diameter) and LV mass-volume ratio (LV mass/LV end-diastole volume). We calculated LV mass according to a previously validated formula⁵³. We used previously validated formulas to calculate measures of cardiac function, including: Mitral peak (E) velocity (m/s), diastolic annular (e’) velocity (cm/s), LV filling pressure (E/e’ ratio), and the ratio of early (E) to late (A) ventricular filling velocities (E/A ratio). We used the Du Bois formula to estimate body surface area in order to calculate the LV mass index, the LV end diastolic volume index and LV end systolic volume index. We also calculated hemodynamic measures including volume-derived ejection fraction (%) and fractional shortening (%)⁵⁴.

A speckle-tracking software package (2D Cardiac Performance Analysis, TomTec Imaging Systems, Unterschleissheim, Germany) was used to perform previously validated speckle-tracking-based analyses of LV myocardial deformation⁵⁵. Global systolic longitudinal strain is a measure of LV mechanical function, as previously described⁵⁶.

Descriptive analyses were performed to compare the baseline demographic, metabolic, and echocardiographic characteristics between participants with and without hepatic steatosis. We present descriptive summaries as means ± standard deviations for continuous variables and as proportions for categorical variables. Age-, cohort-, and sex-adjusted partial Pearson correlations were calculated to assess the correlation between LPR and multiple echocardiographic measures of cardiac structure and function. We performed multivariable linear regression analyses to examine the associations between liver fat (as a continuous or dichotomous measure) and echocardiographic measures of cardiac structure and function. Covariates in the multivariable models were selected a priori based on prior studies suggesting an association with the exposure and outcome. The initial model (base model) included adjustment for age, sex, cohort, smoking status, and alcohol intake. Additional multivariable models included additional adjustment for HF risk factors: diabetes, systolic blood pressure, anti-hypertensive medication use, lipid lowering therapy use, total cholesterol, HDL cholesterol, triglycerides, and fasting glucose. Separate models further adjusted for VAT or BMI, respectively. We evaluated for effect modification by sex. For echocardiographic measures that were associated with liver fat, we performed mediation analyses to explore BMI as a possible mediator. SAS version 9.3 was used to perform the primary analyses and R version 3.5.3 with the ‘mediation’ package was used to perform the mediation analyses⁵⁷. A 2-tailed p-value <0.05 was considered statistically significant. No adjustments were made for multiple testing, since our analyses were hypothesis generating.

Results

The baseline demographic and metabolic characteristics of the 2,356 participants (mean age 52±12 years, 52% women) by hepatic steatosis presence are presented in Table 1. The prevalence of hepatic steatosis in the study sample was 16.3%. Participants with hepatic steatosis had a higher proportion of men, hypertension, diabetes, higher fasting glucose level, higher triglycerides, lower high density lipoprotein levels (HDL), and higher lipid-lowering medication use as compared to those without hepatic steatosis. Cardiac structural and functional characteristics by hepatic steatosis status are summarized in Table 2. Compared to those without hepatic steatosis, participants with hepatic steatosis had, on average, a greater LV wall mass index, greater LV wall thickness, higher regional wall thickness, and a higher mass volume ratio. Those with hepatic steatosis also had, on average, lower diastolic annular velocity (e’), higher LV filling pressure (E/e’ ratio) and worse global systolic longitudinal strain.

Table 1.

Demographic and Metabolic Characteristics of FHS Study Sample by Presence of Hepatic Steatosis

	Hepatic steatosis N=384	No hepatic steatosis N=1972	Overall N=2356
Age (years)	53 ± 12	52 ± 12	52 ± 12
Women n, (%)	177 (46.1%)	1046 (53.0%)	1223 (51.9%)
Smoking n, (%)	42 (10.9%)	211 (10.7%)	253 (10.7%)
Drinks per Week	3.1 ± 3.9	3.0 ± 3.4	3.0 ± 3.5
Systolic Blood Pressure (mm Hg)	127 ± 15	120 ± 15	121 ± 16
Diastolic Blood Pressure (mm Hg)	78 ± 10	75 ± 9	75 ± 9
Hypertension n, (%)	187 (48.7%)	525 (26.6%)	712 (30.2%)
Use of anti-hypertensive meds n, (%)	136 (35.4%)	373 (18.9%)	509 (21.6%)
Use of lipid lowering meds n, (%)	97 (25.3%)	352 (17.8%)	449 (19.1%)
Total Cholesterol (mg/dl)	193 ± 37	191 ± 34	191 ± 35
HDL Cholesterol (mg/dl)	47 ± 14	56 ± 16	54 ± 16
Triglycerides (mg/dl)	168 ± 112	108 ± 62	117 ± 76
Fasting Glucose (mg/dl)	109 ± 29	98 ± 19	100 ± 21
Diabetes n, (%)	62 (16.1%)	85 (4.3%)	147 (6.2%)
BMI (kg/m2)	31.6 ± 6.0	26.8 ± 4.9	27.6 ± 5.4
Estimated body surface area (m2)	2.0 ± 0.2	1.9 ± 0.2	1.9 ± 0.2
VAT volume (cm3)	2540 ± 1008	1523 ± 885	1689 ± 981

HDL, high density lipoprotein; BMI, body mass index; VAT, visceral adiposity tissue

Hepatic steatosis was defined as a liver phantom ratio ≤ 0.33 on computed tomography examination

Continuous variables expressed as mean (sd), categorical variables as n, (%)

Table 2.

Echocardiographic Characteristics of FHS Study Sample by Presence of Hepatic Steatosis

	Hepatic steatosis N=384	No hepatic steatosis N=1972	Overall N=2356
Cardiac Structure
LV wall mass index (g/m2)	87 ± 17	85 ± 18	85 ± 17
LV wall thickness (cm)	2.0 ± 0.3	1.8 ± 0.3	1.9 ± 0.3
LV end-diastole diameter (cm)	4.9 ± 0.4	4.9 ± 0.4	4.9 ± 0.4
LV end-systole diameter (cm)	3.1 ± 0.4	3.1 ± 0.4	3.1 ± 0.4
LV end-diastolic volume index (mL/m2)	58 ± 10	61 ± 10	60 ± 10
LV end-systolic volume index (mL/m2)	40 ± 12	39 ± 11	39 ± 11
Regional wall thickness	0.62 (0.56–0.69)	0.58 (0.53–0.64)	0.58 (0.54–0.65)
Mass volume ratio	1.50 (1.33–1.65)	1.37 (1.23–1.53)	1.38 (1.25–1.55)
LV Systolic Function
Global systolic longitudinal strain (%)	−19 ± 3	−20 ± 3	−20 ± 3
Ejection Fraction (%)	66 ± 5	66 ± 6	66 ± 6
Fractional shortening (%)	37 ± 4	37 ± 4	37 ± 4
LV Diastolic Function
Mitral peak velocity (E) (cm/s)	67 ± 13	66 ± 12	67 ± 13
Diastolic annular velocity (e’) (cm/s)	10 ± 2	11 ± 3	11 ± 3
LV filling pressure (E/e’)	7.0 (6–8)	6.0 (5–7)	6.0 (5–7)
Early-to-late ventricular filling velocity (E/A ratio)	1.06 (0.88–1.22)	1.16 (0.93–1.41)	1/14 (0.91–1.38)

LV, left ventricular

Continuous variables expressed as mean ± sd

^*Hepatic steatosis was defined as a liver phantom ratio ≤ 0.33 on computed tomography examination

We observed significant correlations between LPR and a majority of the echocardiographic measures of cardiac function and structure as summarized in Table 3. Results were generally consistent with the trends noted above. Increasing liver fat was weakly negatively correlated with a lower E/A ratio and reduced diastolic annular velocity (e’ velocity). We observed weak, but statistically significant, positive correlations between liver fat and a higher mitral peak velocity (E), LV wall thickness, LV end-diastolic volume index, LV end-systolic volume index, mass volume ratio, regional wall thickness, global systolic longitudinal strain and LV filling pressure (E/e’ ratio).

Table 3.

Age-, Sex-, Cohort-Adjusted Pearson’s Correlations Between Liver Phantom Ratio and Markers of Subclinical Myocardial Dysfunction

	More Liver Fat	P-value
Cardiac Structure
LV wall mass index	−0.01	0.66
LV wall thickness	0.16	<0.0001
LV diameter at end-diastole	0.00	0.85
LV diameter at end-systole	0.01	0.74
LV end-diastolic volume index	0.13	<0.001
LV end-systolic volume index	0.08	<0.001
Regional wall thickness	0.11	<0.0001
Mass volume ratio	0.16	<0.0001
LV Systolic Function
Global systolic longitudinal strain	0.13	<0.0001
Fractional shortening	−0.01	0.77
Ejection fraction	0.00	0.83
LV Diastolic Function
Mitral peak velocity (E)	0.07	<0.01
Diastolic annular velocity (e’)	−0.12	<0.0001
LV filling pressure (E/e’)	0.15	<0.0001
E/A ratio	−0.10	<0.0001

LV, left ventricular; LPR, liver phantom ratio

For each additional standard deviation of liver fat (lower LPR), we observed increases in mean LV wall mass (β=1.45; 95% confidence interval (CI) 0.01,2.88), LV wall thickness (β=0.01; 95% CI 0.00,0.02), and mass volume ratio (β=0.02; 95% CI 0.01,0.03) in multivariable models adjusted for demographic and HF risk factors (Table 4). After additionally adjusting for BMI or VAT, mean end-diastolic volume and diameter were negatively associated with liver fat, though the association with LV wall mass, LV wall thickness, and the mass volume ratio were no longer statistically significant (Table 4).

Table 4.

Multivariable-Adjusted Linear Regression Analysis for the Association of Liver Fat with Markers of Subclinical Myocardial Dysfunction

Outcome	Model 1: Base Model*		Model 2: Base + HF Risk Factors**		Model 3: Model 2 + BMI		Model 4: Model 2 + VAT
	β (CI)	P value	β (CI)	P value	β (CI)	P value	β (CI)	P value
Cardiac Structure
LV wall mass	4.17 (2.77,5.56)	<0.001	1.45 (0.01,2.88)	0.048	−1.19 (−2.57,0.19)	0.09	−1.36 (−2.82,0.11)	0.07
LV wall thickness	0.03 (0.03,0.04)	<0.001	0.01 (0.00,0.02)	0.003	−0.00 (−0.01,0.01)	0.89	−0.00 (−0.01,0.01)	0.67
LV at end-diastole diameter	0.00 (−0.01,0.02)	0.86	−0.00 (−0.02,0.01)	0.58	−0.02 (−0.04,0.01)	0.003	−0.02 (−0.04,0.01)	0.005
LV at end-systole diameter	0.00 (−0.01,0.02)	0.73	−0.00 (−0.01,0.02)	0.85	−0.01 (−0.02,0.00)	0.16	−0.01 (−0.03,0.00)	0.09
LV end diastolic volume	0.07 (−0.75,0.89)	0.87	−0.26 (−1.13,0.61)	0.56	−1.30 (−2.17,0.44)	0.003	−1.31 (−2.21,0.40)	0.005
LV end systolic volume	0.05 (−0.36,0.47)	0.80	0.02 (−0.41,0.46)	0.92	−0.33 (−0.77,0.12)	0.15	−0.39 (−0.85,0.06)	0.09
Regional wall thickness	0.01 (0.01,0.02)	<0.001	0.00 (−0.00,0.01)	0.06	0.00 (−0.00,0.01)	0.40	0.00 (−0.00,0.01)	0.32
Mass volume ratio	0.04 (0.03,0.05)	<0.001	0.02 (0.01,0.03)	0.001	0.01 (−0.00,0.02)	0.23	0.01 (−0.00,0.02)	0.29
LV Systolic Function
Global systolic longitudinal strain	0.39 (0.27,0.51)	<0.001	0.20 (0.07,0.33)	0.002	0.13 (−0.00,0.26)	0.05	0.09 (−0.04,0.23)	0.17
Ejection fraction	−0.03 (−0.20,0.14)	0.75	−0.09 (−0.27,0.10)	0.36	−0.10 (−0.28,0.09)	0.30	−0.05 (−0.24,0.14)	0.62
Fractional shortening	0.03 (−0.20,0.14)	0.75	−0.09 (0.27,0.10)	0.36	−0.10 (−0.28,0.09)	0.29	−0.05 (−0.24,0.14)	0.62
LV Diastolic Function
Mitral peak velocity (E)	0.81 (0.32,1.31)	0.001	0.83 (0.31, 1.36)	0.002	0.60 (0.07,1.13)	0.03	0.75 (0.20,1.30)	0.008
Diastolic annular velocity (e’)	−0.27 (−0.36,−0.18)	<0.001	−0.12(−0.22,−0.03)	0.009	−0.07 (−0.16,0.03)	0.17	−0.00 (−0.10,0.09)	0.93
LV filling pressure (E/e’)	0.26 (0.19,0.33)	<0.001	0.16 (0.09,0.23)	<0.001	0.11 (0.03,0.18)	0.004	0.08 (0.01,0.16)	0.03
E/A ratio	−0.03 (−0.04,−0.02)	<0.001	−0.01 (−0.02,−0.00)	0.035	−0.00 (−0.02,0.01)	0.46	0.00 (−0.01,0.01)	0.79

LV, left ventricular; HF, heart failure; BMI, body mass index

^*MV model adjusts for age, sex, cohort, smoking, alcohol use; β expressed per SD increase in liver fat

^**Heart failure (HF) risk factors: diabetes, systolic blood pressure, antihypertensive med use, lipid lowering med use, total cholesterol, HDL, triglycerides, and fasting glucose

With regard to cardiac function, we observed a negative association between liver fat and global systolic longitudinal strain (β=0.20; 95% CI 0.07,0.33) in multivariable adjusted models. Increases in mean liver fat was associated with greater mean mitral peak (E) velocity (β=0.83; 95% CI 0.31,1.36), lower diastolic annular velocity (e’ velocity) (β=−0.12; 95% CI −0.22,−0.03), decreased mean E/A ratio (β=−0.01; 95% CI-0.02,−0.00), and increased mean (E/e’ ratio) (β=0.16; 95% CI 0.09,0.23). After additional adjustment for BMI or VAT, most associations between liver fat and measures of cardiac dysfunction were no longer significant. However, liver fat remained positively associated with mitral peak velocity and LV filling pressure in the adjusted BMI model.

Results were generally similar for models where liver fat was defined dichotomously based on LPR ≤ 0.33 (Supplemental Table 1) Results presented in terms of standardizing ed beta coefficients are presented in Supplemental Table 2.

Results of the mediation analyses demonstrate that BMI was a significant mediator in the associations between liver fat and subclinical myocardial dysfunction (Figure 1). BMI fully mediated the associations between liver fat and the myocardial structural measures of LV wall mass, LV wall thickness, and mass volume ratio and the myocardial functional measures of diastolic annular velocity and E/A ratio. BMI partially mediated the relationships between liver fat and mitral peak velocity (percent mediation 28%), global systolic longitudinal strain (percent mediation 36%) and LV filling pressure (percent mediation 33%), respectively.

Figure 1.

Mediation effect of BMI in the association between liver fat and subclinical markers of myocardial dysfunction.

Discussion

In our large, community-based sample from the FHS, liver fat was significantly associated with multiple subclinical cardiac structural and functional abnormalities after adjusting for a number of demographic and HF risk factors. Increased liver fat was associated with worse global systolic longitudinal strain, a sensitive prognostic marker of LV systolic dysfunction⁴⁷. Many of the associations we observed were attenuated by obesity; however, liver fat remained associated with higher LV filling pressure (E/e’ ratio), a sensitive marker of diastolic heart dysfunction and potential precursor to clinical HF, in models further adjusted by general or visceral adiposity. Moreover, mediation analyses revealed that BMI appeared to only partially mediate the relation of liver fat with LV filling pressure and diastolic annular velocity; therefore, liver fat may contribute to diastolic dysfunction above and beyond general adiposity.

Robust evidence has shown that the presence and severity of NAFLD is associated with multiple markers of subclinical CVD, including decreased coronary flow, impaired flow-mediated vasodilation, increased carotid intima-media thickness, arterial thickness, and carotid atherosclerotic calcification independent of several cardiometabolic risk factors^58–63. Previous studies have also observed an association between NAFLD and impaired diastolic function after adjusting for HF risk factors similar to our study^{12,29–37,40}. Whereas prior studies were limited by small sample sizes^{30,32,34,37,40}, the use of select populations, including men only³⁶, those with diabetes^30,40, bariatric surgery³², or patients with hypertension³⁴, or were conducted outside of the US³⁷.

Our study adds to the literature by evaluating the association between liver fat and multiple measures of cardiac structure and function in a large, US community-based sample. Additionally, few prior studies measured LV filling pressure, a potential precursor to clinical heart failure, in their investigation of diastolic impairment^12,33,40. We employed novel speckle-tracking techniques that were applied by a limited number of prior studies^12,40 to assess early markers of systolic dysfunction.

Our results are generally similar with the findings from the Coronary Artery Risk Development in Young Adults (CARDIA) study, with a few exceptions. In the CARDIA study, the association with liver fat and LV filling pressure was no longer significant after adjusting for general or visceral adiposity; however, in our study, this association persisted and was only partially mediated by BMI¹². Interestingly, an updated analysis of the CARDIA study found that liver fat was prospectively associated with increased LV filling pressure compared to non-NAFLD patients independent of adiposity during a 30-year follow-up, strengthening the perspective that that liver fat may add additional risk for diastolic heart failure⁶⁴. The prior CARDIA study also observed an association between NAFLD and worse global systolic longitudinal strain after accounting for similar HF risk factors as in our study¹²; however, the association persisted even after further adjusting for VAT in their study. Nearly half of the participants in the CARDIA study were black and the prevalence of diabetes was higher and that of NAFLD was lower compared to our sample. It is possible that differences in the study sample may account for some of the differences in results, although further investigation is needed.

Obesity, is a major contributor to numerous comorbidities, including NAFLD and CVD^65–67. HF with preserved ejection fraction is a complex and heterogeneous clinical entity with multiple underlying pathophysiologic substrates, including obesity⁶⁸. We add to the literature by testing BMI’s mediation effect on the relationship between NAFLD and measures of cardiac structure and function. Our findings support the hypothesis that general adiposity may lie in the causal pathway between fatty liver and early LV diastolic dysfunction. Future studies are needed to determine if early cardiac abnormalities could serve as predictive markers of HF risk in patients with NAFLD. Additional studies are also needed to determine if decreasing general adiposity reduces the risk of HF in patients with NAFLD. Interestingly, our results suggest that obesity alone does not completely account for the association between increased liver fat and LV filling pressure, a notable marker of diastolic dysfunction. Most individuals with NAFLD are also obese, but a significant proportion are lean⁶⁹. Lean individuals with NAFLD have comparable risk for adverse CVD outcomes compared to those with both NAFLD and obesity⁷⁰. The pathophysiologic mechanisms underlying the association between NAFLD and impaired LV filling pressure beyond obesity should be further explored.

Increased liver fat is associated with markers of inflammation and oxidative stress. Liver fat was associated with increased C-reactive protein, urinary isoprostanes, IL-6, intercellular adhesion molecule 1, and P-selectin even after adjustment for BMI or VAT in a prior FHS analysis⁷¹. NAFLD, as a chronic inflammatory condition, may contribute to the overproduction of systemic inflammatory mediators associated with impaired cardiac function^29,41,72,73. Insulin resistance, marked by elevated insulin levels, which is a common feature of NAFLD, may also occur as a result of increased inflammation and oxidative stress. Inflammation may contribute to abnormal myocyte growth and fibrosis and activate the sympathetic nervous system through increased sodium retention, causing impaired cardiac performance^25,74,75. Moreover, emerging literature suggest that NAFLD is associated with abnormal fatty acid oxidation and intestinal dysbiosis, which may alter lipid metabolism and inflammation, contributing to CVD^8,76. Obesity is also associated with increased inflammatory mediator production and reduced levels of adiponectin⁷⁷, a known anti-inflammatory factor, and may therefore be involved in the pro-inflammatory pathway leading to increased HF risk in NAFLD. Increased myocardial steatosis, common in obese patients, may additionally alter the efficiency of cardiac metabolism and lead to myocardial lipotoxicity and cardiac dysfunction in the NAFLD population^78,79. Additional mechanistic studies are needed to deepen our understanding of how NAFLD contributes to subclinical myocardial impairment.

Major strengths of our investigation include using CT imaging to objectively evaluate liver fat and acquisition of comprehensive cardiovascular outcomes data using well-measured covariates. Our study also applies speckle-tracking echocardiography techniques, a sensitive tool for detection of subclinical myocardial dysfunction⁸⁰. Our use of a large community sample size additionally provides more power to find significant differences not otherwise detectable in smaller studies.

The present study has certain limitations. Though measuring liver fat by CT imaging correlates well with histologic assessments of liver fat⁸¹, CT is insensitive to mild liver fat, so we likely underestimated the burden of hepatic steatosis in our sample and biased our results towards the null. Furthermore, hepatic steatosis may diminish as NAFLD progresses into steatohepatitis and fibrosis, reducing our ability to fully capture the entire NAFLD spectrum in our study. We acknowledge that the prevalence of NAFLD in our study is low compared to other studies, however we did have sufficient power to detect clinically meaningful differences between groups. The low prevalence of NAFLD in our cohort likely reflects the lower population prevalence of NAFLD in the early 2000s compared to more contemporary cohorts and the insensitivity of CT scans for detecting mild NAFLD. Additionally, LV filling pressure was not directly measured, however E/e’ ratio was used as a marker of LV filling pressure, which has been done in previous studies. While E/e’ is useful for assessing filling pressure noninvasively, it is less accurate at predicting LV filling pressures in certain clinical settings, including significant LV dilation and mitral regurgitation. Furthermore, our study is limited by the lack of information on other chronic liver diseases, including viral hepatitis, which can contribute to the features of liver fat on CT scan, leading to a misclassification bias. The generalizability of our findings to other ancestries is also not known. The mediation analyses were hypothesis generating; we cannot exclude the possibility that the attenuation of the associations were due to confounders associated with adiposity measures. As our study was exploratory, we opted to not correct for multiple testing; it is possible that the associations we observed are the result of random chance. Confirmatory studies are needed.

In summary, increased liver fat is associated with alterations in subclinical cardiac structure and function that may lead to clinical heart failure in the absence of overt liver disease and after accounting for notable HF risk factors. Many of the associations between NAFLD and early myocardial dysfunction may be related to adiposity; though NAFLD is associated with LV filling pressure after accounting for obesity. Further studies are warranted to gain mechanistic insights and to confirm our findings.