Associations of Skeletal Muscle Mass, Muscle Fat Infiltration, Mitochondrial Energetics, and Cardiorespiratory Fitness With Liver Fat Among Older Adults

Abstract

Background

Muscle mass loss may be associated with liver fat accumulation, yet scientific consensus is lacking and evidence in older adults is scant. It is unclear which muscle characteristics might contribute to this association in older adults.

Methods

We associated comprehensive muscle-related phenotypes including muscle mass normalized to body weight (D₃-creatine dilution), muscle fat infiltration (magnetic resonance imaging), carbohydrate-supported muscle mitochondrial maximal oxidative phosphorylation (respirometry), and cardiorespiratory fitness (VO₂ peak) with liver fat among older adults. Linear regression models adjusted for age, gender, technician (respirometry only), daily minutes of moderate-to-vigorous physical activity, and prediabetes/diabetes status tested main effects and interactions of each independent variable with waist circumference (high: women—≥88 cm, men—≥102 cm) and gender.

Results

Among older adults aged 75 (interquartile range: 73, 79 years; 59.8% women), muscle mass and liver fat were not associated overall (N = 362) but were positively associated among participants with a high waist circumference (β: 25.2; 95% confidence intervals [95% CI]: 11.7, 40.4; p = .0002; N = 160). Muscle fat infiltration and liver fat were positively associated (β: 15.2; 95% CI: 6.8, 24.3; p = .0003; N = 378). Carbohydrate-supported maximum oxidative phosphorylation (before adjustment) and VO₂ peak (after adjustment; β: −12.9; 95% CI: −20.3, −4.8; p = .003; N = 361) were inversely associated with liver fat; adjustment attenuated the estimate for maximum oxidative phosphorylation although the point estimate remained negative (β: −4.0; 95% CI: −11.6, 4.2; p = .32; N = 321).

Conclusions

Skeletal muscle-related characteristics are metabolically relevant factors linked to liver fat in older adults. Future research should confirm our results to determine whether trials targeting mechanisms common to liver and muscle fat accumulation are warranted.

Graphical Abstract

Graphical Abstract.

Older adults are at an elevated risk of multiple metabolic derangements including liver fat accumulation. Excess liver fat is a component of non-alcoholic fatty liver disease (NAFLD), which is associated with an increased risk of cardiovascular events and mortality among persons of advanced age (1). Not only do older adults suffer from a higher prevalence of NAFLD (40% among the aging U.S. population) than their younger counterparts (25% in the general population (1,2)), but they are subject to a more aggressive disease prognosis once hepatic steatosis develops (3). Given the absence of ubiquitous clinical screening procedures for hepatic steatosis and the difficulty of accessing liver tissue for diagnosing NAFLD, it often goes undetected until more advanced stages of disease, such as advanced fibrosis, which can be irreversible and associated with poor outcomes (4). Thus, easily obtainable clinical measurements that are related to liver fat accumulation, are amenable to routine monitoring, and can serve as targets of lifestyle interventions to revert liver fat in older adults are urgently needed.

One such potential target that may provide insight into liver health is skeletal muscle mass. Although skeletal muscle mass and liver fat are inversely correlated in some studies—albeit most have not specifically focused on older adults—there is a lack of consensus in the extant literature about whether muscle loss is implicated in the pathophysiology of liver fat accumulation (5,6). In addition, no studies have comprehensively examined whether muscle characteristics such as myosteatosis, muscle mitochondrial energetics, and cardiorespiratory fitness (the whole body manifestation of muscle mass and mitochondrial energetics) (7,8) may better reflect overall muscle health than muscle mass per se. Thus, it is tempting to speculate that ectopic lipid deposition and low oxidative capacity in muscle could drive muscle insulin resistance and downstream liver fat accumulation (6,9–11). Another potential reason for discordance of prior results is the variability in the rigor of measurements of muscle-related phenotypes and liver fat. Furthermore, despite the fact that numerous muscle characteristics are frequently compromised among older adults (7,12,13), it is unclear whether prior evidence in younger populations that link muscle phenotypes to liver fat can be generalized to older adults.

Given the lack of consensus regarding the connection between muscle mass and liver fat, the scarcity of such evidence focused on the older adult population, and a dearth of studies that have utilized rigorous measurement methods for both muscle and liver, our primary objective was to evaluate whether skeletal muscle mass was associated with liver fat using gold-standard measurements of both tissues in a sample of older adults in the Study of Muscle, Mobility and Aging (SOMMA) aged 75 (interquartile range [IQR]: 73, 79 years)—substantially older than earlier investigations (mean age ~45–60 years) (14,15). Secondarily, to explore beyond the link between muscle mass and liver fat and generate hypotheses about common mechanisms of pathobiology connecting the 2 tissues in the aging population, we interrogated the associations of muscle fat infiltration, muscle mitochondrial energetics, and VO₂ peak with liver fat among SOMMA older adults. Recognizing that there are sex differences in both skeletal muscle mass (16) and liver fat (17) and that adiposity is a key risk factor for liver fat accumulation (1), which may compromise the quality of skeletal muscle mass (5,14), we tested for effect modification by gender and waist circumference (a proxy for visceral adiposity) (18) for all associations.

Method

Study Participants

Older adults aged ≥70 years were recruited for SOMMA (https://www.sommastudy.com/) at the University of Pittsburgh and Wake Forest University School of Medicine between April 2019 and December 2021, as previously described (19). Briefly, individuals were eligible to participate if they were willing and able to complete a skeletal muscle biopsy and undergo magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS). Individuals were excluded if they reported an inability to walk one-quarter of a mile or climb a flight of stairs, had body mass index (BMI) ≥ 40 kg/m², had an active malignancy, dementia, or any medical contraindication to biopsy or MR. Variables included in the present analysis were measured over 3 days of assessments at the baseline SOMMA visit: Day 1 (400 m walk plus most other in-person assessments); Day 2 (cardiopulmonary exercise testing [CPET] and MR); and Day 3 (muscle biopsy). Participants self-reported their demographics (gender, race, and ethnicity), clinical data (health history including smoking, alcohol use, and medications), and lifestyle behaviors (self-reported diet and objective measures of physical activity). A binary variable (yes/no) was constructed for prediabetes or diabetes status based upon self-reported diabetes diagnosis or a hemoglobin A1c ≥ 5.7% measured at each study site’s local clinical laboratory (local Quest Diagnostic Laboratory in Pittsburgh). Participants were asked to fast for 12 hours prior to the Day 3 baseline assessment for biospecimen collection (including blood, urine, and muscle).

Magnetic Resonance Imaging

A whole-body MRI (Siemens Medical System 3T-Prisma [Pittsburgh] or Skyra [Wake Forest]) was used to assess body composition including quadriceps and total thigh muscle volume, muscle fat infiltration (mean fat fraction in the muscle tissue of the right and left anterior thighs), liver fat by proton density fat fraction (the ratio of the density of protons from triglycerides and the total density of protons from triglycerides and water expressed as a percentage (20)), and intraabdominal (visceral) and abdominal subcutaneous adipose tissue (AMRA Medical, Linköping, Sweden). For 3 participants with missing muscle fat infiltration data for the right leg, the left thigh fat infiltration was used instead of the mean of both legs. To compare the characteristics of participants with and without hepatic steatosis, continuous liver fat was dichotomized according to the threshold of ≥5.6% (21). However, liver fat was used as a continuous variable in all statistical models.

Muscle Mass and Muscle Mitochondrial Energetics

The deuterium-labeled creatine (D₃-creatine) dilution method was used to measure whole body muscle mass (D₃Cr muscle mass) through assessment of creatine pool size, as previously described (19). D₃Cr muscle mass was expressed as a percentage of total body weight in kg (ie, D₃Cr muscle mass/wgt).

A skeletal muscle biopsy was performed in the morning while the participant was fasted (12 hours) and had avoided strenuous physical activity for the prior 48 hours (19). Mitochondrial respiration was measured ex vivo in permeabilized muscle fibers using an Oxygraph O2K instrument (Oroboros Inc., Innsbruck, Austria), as previously described (19). In line with the hypothesis that greater carbohydrate-supported OXPHOS in peripheral tissues prevents liver fat accumulation (22), we focused on carbohydrate-supported maximal complex I and II-supported oxidative phosphorylation (Max OXPHOS, pmol/(s*mg)). Fatty acid-supported OXPHOS with and without carbohydrate-supported OXPHOS was included in post hoc correlation analyses with liver fat. Steady-state O₂ flux was normalized to muscle fiber bundle wet weight using Datlab 7.4 software.

As an alternate measure of muscle mitochondrial energetics, we included maximum adenosine triphosphate production (ATP_max, mM/second) measured in vivo by ³¹Phosphorus (³¹P) MRS of the quadriceps muscle following an acute bout of knee extensor exercise, as described elsewhere (19). A 3 Tesla MR magnet (Siemen’s Medical System—Prisma [Pittsburgh] or Skyra [Wake Forest]) collected the phosphorous spectra through the quadriceps. jMRUI software v6.0 was used to calculate ATP_max (19).

Cardiorespiratory Fitness by CPET

A standardized treadmill CPET was used to measure cardiorespiratory fitness (VO₂ peak). A modified Balke or manual protocol was used to measure oxygen consumption and carbon dioxide production during maximal effort exercise (19). VO₂ peak was normalized to body weight (mL/kg/min).

Anthropometrics

Body weight was measured in kg using a digital scale. Two to three measurements of waist circumference were taken using standardized procedures; the 2 closest measurements were averaged. Waist circumference was dichotomized as high (≥88 cm for women or ≥102 cm for men) or normal (<88 cm for women or <102 cm for men) (23).

Moderate-to-Vigorous Physical Activity by Actigraphy

Participants wore the Actigraph GT9X (Actigraph, Pensacola, FL) on their nondominant wrist for up to 7 days, as previously described (19). The data were scored using the ActiLife Software and used to calculate daily sedentary and active time in each 24-hour interval (midnight to 23:59). Activity levels were defined and calculated using previously published cutpoints based upon activity level counts per minute for the 3 axes captured by the Actigraph. Cutpoints were <2 860 counts per minute for sedentary activity, 2 860–3 940 counts per minute for light activity, and ≥3 941 counts per minute for moderate-to-vigorous physical activity (MVPA) (24).

Statistical Analysis

We used descriptive statistics [mean ± SD or median IQR appropriate for normally and nonnormally distributed continuous variables, respectively; and N (%) for categorical variables] to describe the demographic and clinical characteristics of the study sample. The demographic and clinical characteristics of participants with and without NAFLD were compared using t tests for normally distributed continuous variables and Wilcoxon rank sum tests for nonnormally distributed continuous variables; and using the Chi-squared or Fisher exact test for categorical variables (Fisher’s test was employed when any participant category had an n < 5). We restricted to participants with liver fat data and excluded the Wake Forest study site where data were unavailable due to logistical data collection issues (Participant Flow Diagram with additional details of participant inclusion and exclusion shown in Supplementary Figure 1). To evaluate the representativeness of our SOMMA subsample, we compared the characteristics of SOMMA participants included and excluded from our analysis.

We used general linear regression models to evaluate associations of an SD increment in each independent variable (primary: D₃Cr muscle mass/wgt; exploratory: muscle fat infiltration, carbohydrate-supported ex vivo and in vivo mitochondrial bioenergetics, and cardiorespiratory fitness) with the outcome of liver fat. We evaluated linearity and model assumptions through inspection of the shape of each association and residual distributions, which led us to log-transform liver fat. To facilitate interpretability, we back-transformed model β estimates and 95% confidence intervals; this provided a proportional interpretation (ie, the relative difference in liver fat associated with each SD increment in the independent variable). Model 1 was unadjusted; Model 2 was adjusted for age and gender (and technician when Max OXPHOS was the independent variable); and Model 3 was further adjusted for mean daily minutes of MVPA and prediabetes or diabetes status (Supplementary Figure 2). Given that self-reported use of antihyperglycemic agents (binary—yes/no) overlapped almost completely with prediabetes/diabetes status and only 58 participants in our sample reported using any such medication, we opted not to include medication use in statistical models.

We tested for effect modification between each independent variable with gender and waist circumference (a proxy for visceral adiposity (18)) for the outcome of liver fat by adding an interaction term between the independent variable and effect modifier to Model 3. We did not test for effect modification by visceral adipose tissue given that liver fat—the outcome variable—is a component of visceral adipose. We stratified results when the interaction term had a p value < .1. All other results were deemed to be statistically significant at an α level of p < .05.

In post hoc analyses, we examined whether fatty acid-supported OXPHOS with or without carbohydrate-supported OXPHOS was correlated with log-transformed liver fat using the Spearman correlation coefficient. To gain further clinical insight with respect to NAFLD (liver fat percentage ≥5.6% or not), we included this as a binary outcome in exploratory analyses using logistic regression models with the same covariate structure and effect modifiers as the primary analysis. We conducted this supplementary analysis given the clinical relevance of NAFLD to the health of older adults. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Participant Characteristics

We included between 321 and 378 SOMMA participants in the analysis, based on data availability for each independent variable. Median age was 75 (IQR: 73, 79 years), 59.8% were women, and 86.5% identified with a White race (Table 1). Median liver fat was 2.8% (IQR: 2.0, 5.4%; range 1.0, 25.3%; N = 378), 93 (24.6%) participants had hepatic steatosis, and mean D₃Cr muscle mass/wgt was 28.7 ± 6.8% (N = 362). SOMMA participants included in our analysis had a slightly lower hemoglobin A1c and VO₂ peak than those excluded (Supplementary Table 1). A smaller proportion of included participants had prediabetes or diabetes (37.3%) as compared to those excluded (47.2%; p = .003).

Table 1.

Baseline Characteristics of Participants With and Without Hepatic Steatosis

	All Participants		NAFLD† (Liver Fat ≥ 5.6%)		No NAFLD (Liver Fat < 5.6%)
	Total N	Mean ± SD, Median (IQR), or N (%)*	N	Mean ± SD, Median (IQR), or N (%)	N	Mean ± SD, Median (IQR), or N (%)	p Value
Demographics
Age (y)	378	75 (73, 79)	93	75 (73, 79)	285	75 (73, 80)	.92
Women	378	226 (59.8)	93	59 (63.4)	285	167 (58.6)	.41
Race	378		93		285		.81
Asian		<5‡		0		<5
Black		45 (11.9)		13 (14.0)		32 (11.2)
Native American/Alaskan native		<5		0		<5
Unknown		<5		<5		<5
White		327 (86.5)		79 (85.0)		248 (87.0)
Hispanic ethnicity	378	<5	93	<5	285	<5	.57
Clinical characteristics
Body mass index (kg/m²)	378	27.8 ± 4.8	93	31 ± 4.4	285	26.7 ± 4.5	<.0001
High waist circumference§	372	167 (44.9)		68 (74.7)		99 (35.2)	<.0001
Prediabetes or diabetes∥	378	141 (37.3)	93	52 (55.9)	285	89 (31.2)	<.0001
Hemoglobin A1c (%)	373	5.5 (5.3, 5.8)	90	5.7 (5.5, 6.4)	283	5.5 (5.2, 5.7)	<.0001
Moderate-to-vigorous physical activity (daily minutes)	378	188 ± 87	93	179 ± 81	285	191 ± 89	.24
Outcome variable
Liver fat (%)	378	2.8 (2.0, 5.4)	93	8.1 (6.4, 12.1)	285	2.4 (1.8, 3.1)	<.0001
Exposure variables
Muscle mass (%)	362	28.7 ± 6.8	89	28.2 ± 6.4	273	28.8 ± 7.0	.48
Muscle fat infiltration (%)	378	7.3 (6.0, 9.1)	93	8.4 (6.9, 10.0)	285	7.1 (5.8, 8.7)	<.0001
Carbohydrate-supported maximum oxidative phosphorylation (pmol/(s*mg))	321	59.7 ± 19.2	78	57.9 ± 17.3	243	60.3 ± 19.8	.34
ATP production (mM/s)	345	0.52 ± 0.14	86	0.51 ± 0.13	259	0.52 ± 0.15	.32
VO2 peak (mg/kg/min)¶	361	20.9 ± 5.4	85	19.1 ± 4.1	276	21.5 ± 5.6	<.0001

Notes: ^*Continuous variables presented as mean ± SD or median (IQR) when normally or nonnormally distributed, respectively. Categorical variables presented as N (%). For continuous variables, p values shown are from t tests for normally distributed variables and from Wilcoxon rank sum tests for nonnormally distributed variables. p Values for categorical variables are from the Chi-squared or Fisher exact test when any category had <5 participants. IQR = interquartile range; ATP = adenosine triphosphate production. Bolded values are statistically significant (p <.05).

^†NAFLD = non-alcoholic fatty liver disease.

^‡To avoid inadvertent participant identification, we present small sizes as n < 5.

^§Normal waist circumference was defined as <88 cm for women and <102 cm for men.

^∥Prediabetes or diabetes was ascertained based on hemoglobin A1c (≥5.7%) or self-report of diabetes.

^¶VO₂ peak was normalized to kg of body weight.

D₃Cr Muscle Mass Is Positively Associated With Liver Fat Only in Older Adults With a High Waist Circumference

Lower muscle mass is associated with a higher prevalence of hepatic steatosis in some, but not all studies (5,6). Therefore, we first asked whether higher muscle mass is associated with lower liver fat in an older study sample using the rigorous D₃Cr muscle mass measurement and MRI proton density fat fraction for liver fat. Neither the crude nor the adjusted associations between D₃Cr muscle mass/wgt and liver fat were statistically significant (fully adjusted β: −0.87, 95% CI: −8.1, 6.9, p = .82; Table 2, Figure 1A, Supplementary Figure 1A). However, the fully adjusted model indicated a positive association between D₃Cr muscle mass/wgt and liver fat among participants with a high waist circumference (p = .001 for interaction; β: 25.2, 95% CI: 11.7, 40.4, p = .0002; N = 160; Figure 1B, Supplementary Figure 3B), but not among those with a normal waist circumference (β: −3.6, 95% CI: −10.9, 4.4, p = .37; N = 196).

Table 2.

Associations of D₃Cr Muscle Mass/Wgt, Muscle Characteristics, and Cardiorespiratory Fitness With Liver Fat

Exposure Variables	N	SD	Liver Fat (%) Back-Transformed β Estimate* (95% CI)	p Value
Primary exposure variable
D3Cr muscle mass/wgt
Model 1†	362	6.8	−4.9 (−11.4, 2.1)	.16
Model 2	362	6.8	−4.7 (−11.8, 3.0)	.22
Model 3	362	6.8	−0.87 (−8.1, 6.9)	.82
Waist circumference interaction‡	357	—	—	.001
Normal	197	6.7	−3.6 (−10.9, 4.4)	.37
High	160	5.6	25.2 (11.7, 40.4)	.0002
Gender interaction	362	—	—	.64
Exploratory exposure variables
Muscle fat infiltration
Model 1	378	2.3	18.2 (10.4, 26.5)	<.0001
Model 2	378	2.3	21.0 (12.3, 30.3)	<.0001
Model 3	378	2.3	15.2 (6.8, 24.3)	.0003
Waist circumference interaction	372	—	—	.41
Gender interaction	378	—	—	.59
Carbohydrate-supported maximum oxidative phosphorylation (pmol/(s*mg))
Model 1	321	19.2	−8.9 (−15.4, −1.9)	.01
Model 2	321	19.2	−6.2 (−13.7, 1.9)	.13
Model 3	321	19.2	−4.0 (−11.6, 4.2)	.32
Waist circumference interaction	315	—	—	.39
Gender interaction	321	—	—	.29
ATP production (mM/s)
Model 1	345	0.14	−5.3 (−11.9, 1.9)	.14
Model 2	345	0.14	−5.3 (−12.0, 2.0)	.15
Model 3	345	0.14	−0.81 (−7.8, 6.8)	.83
Waist circumference interaction	339	—	—	.77
Gender interaction	345	—	—	.16
VO2 peak (mL/kg/min)
Model 1	361	5.4	−14.1 (−19.9, −8.0)	<.0001
Model 2	361	5.4	−17.5 (−23.9, −10.6)	<.0001
Model 3	361	5.4	−12.9 (−20.3, −4.8)	.003
Waist circumference interaction	355	—	—	.20
Gender interaction	361	—	—	.33

Notes: ^*The back-transformed beta estimate represents the relative percentage difference in liver fat associated with a 1-SD increment in each independent variable. Interaction models are fully adjusted. Stratified estimates are shown when the p value for interaction was < .1. ATP = adenosine triphosphate production; VO₂ peak = peak oxygen uptake. Bolded values are statistically significant (p <.05).

^†Model 1. Crude; Model 2. Adjusted for age and gender (model with maximum oxidative phosphorylation further adjusted for respirometry technician); Model 3. Further adjusted for daily minutes of MVPA and prediabetes/diabetes status.

^‡Normal waist circumference was defined as <88 cm for women and <102 cm for men.

Figure 1.

Associations of D₃Cr muscle mass and muscle fat infiltration with liver fat among Study of Muscle, Mobility and Aging older adults. Panel (A) shows the association between each SD increment in muscle mass/wgt and log-transformed liver fat adjusted for age, gender, daily minutes of moderate-to-vigorous physical activity, and prediabetes/diabetes status (N = 362). Panel (B) shows the association in panel (A) stratified by waist circumference (N = 357). Panel (C) shows the fully adjusted association of muscle fat infiltration with liver fat (N = 378). A high waist circumference was defined as ≥88 cm for women and ≥102 cm for men.

Skeletal Muscle Characteristics Including Muscle Fat Infiltration and Mitochondrial Bioenergetics Are Associated With Liver Fat

Recent reports indicate that the skeletal muscle characteristic of estimated muscle fat infiltration is associated with liver fat in participants with advanced fatty liver disease and class III obesity (10). Through exploration of this association in an older sample, we uncovered a positive association according to the fully adjusted model (15.2% higher liver fat per SD increment in muscle fat infiltration; 95% CI: 6.8, 24.3, p = .0003; N = 378; Figure 1C, Supplementary Figure 3C) that did not vary by gender or waist circumference.

Lower mitochondrial energetics is a salient and modifiable skeletal muscle characteristic that is related to cardiometabolic disease (8), yet its association with liver fat is unclear. We showed for the first time an inverse association between ex vivo carbohydrate-supported muscle mitochondrial bioenergetics (Max OXPHOS) with liver fat according to the crude model (β: −8.9; 95% CI: −15.4, −1.9; p = .01) but was attenuated to nonsignificance after adjustment for age, gender, and technician and was further attenuated in magnitude in the fully adjusted model (β: −4.0; 95% CI: −11.6, 4.2; p = .32; N = 321; Figure 2A, Supplementary Figure 3D). Our post hoc exploration revealed a null crude correlation between fatty acid-supported OXPHOS and liver fat (r = −0.006, p = .92; data not shown). This was also true for the combination of fatty acid and complex I or complexes I and II carbohydrate-supported OXPHOS when correlated with liver fat (r = 0.07, p = .26–.27; data not shown). ATP_max, a secondary measure of muscle mitochondrial bioenergetics measured in vivo, was not associated with liver fat in either crude or adjusted models in the overall sample (fully adjusted β: −0.81; 95% CI: −7.8, 6.8; p = .83; Supplementary Figure 3E) nor within any subgroups.

Figure 2.

Associations of muscle mitochondrial energetics and cardiorespiratory fitness with liver fat among Study of Muscle, Mobility and Aging older adults. Panel (A) shows the association between carbohydrate-supported maximum oxidative phosphorylation (ie, muscle mitochondrial energetics) and log-transformed liver fat (N = 321). Panel (B) shows the association of cardiorespiratory fitness measured as VO₂ peak with liver fat (N = 361). Models were adjusted for age, gender, daily minutes of moderate-to-vigorous physical activity, and prediabetes/diabetes status. Max OXPHOS = maximum oxidative phosphorylation; VO₂ peak = peak oxygen uptake.

Cardiorespiratory Fitness Was Associated Inversely With Liver Fat

Cardiorespiratory fitness (ie, VO₂ peak) is the whole-body manifestation of muscle mass and muscle mitochondrial energetics (8) and a biomarker of health status that declines in aged populations and in individuals with NAFLD—albeit according to studies with indirect scores estimating liver fat (25,26). Thus, we hypothesized that in older adults, a higher VO₂ peak would be associated with lower liver fat when using gold-standard MRI proton density fat fraction. Indeed, each SD increment in VO₂ peak (mL/kg/min) was associated with a ~13% lower liver fat in the fully adjusted model (Figure 2B, Supplementary Figure 3F; fully adjusted β: −12.9; 95% CI: −20.3, −4.8; p = .003; N = 361) and did not vary across strata of gender or waist circumference.

Skeletal Muscle-Related Characteristics Are Associated With the Odds of NAFLD

The results using dichotomous NAFLD as the outcome were similar in directionality to the primary results. Specifying NAFLD dichotomously additionally indicated a trend toward a negative association between D₃Cr muscle mass/wgt and the odds of NAFLD among participants with a normal waist circumference; however, the result did not reach statistical significance (OR: 0.65, 95% CI: 0.39, 1.1, p = .11; N = 197; Supplementary Table 2). The positive association between D₃Cr muscle mass/wgt and the odds of NAFLD among participants with a high waist circumference mirrored the results of the primary analysis (OR: 2.1, 95% CI: 1.4, 3.2, p = .0004; N = 160).

Contrary to the primary results, supplemental results indicated a potential interaction between ATP_max and gender for the odds of NAFLD (p = .09). However, neither of the stratified gender-specific estimates were statistically significant. This may indicate a spurious interaction owing to a sparse data distribution at the high end of ATP_max, which almost always corresponded with a low liver fat; Supplementary Figure 3E). Thus, the interaction should be interpreted with caution and confirmed in future studies with adequate power.

Discussion

Using state-of-the-art measurements, we asked whether deeply phenotyped characteristics of skeletal muscle and cardiorespiratory fitness were associated with liver fat in older adults. We did not find evidence that D₃Cr muscle mass/wgt per se was associated with liver fat in our overall sample; however, the association was positive among older adults with a high waist circumference—a surrogate for visceral adipose volume (27). As an adjunct to these findings, our supplementary results suggest that a higher D₃Cr muscle mass/wgt may be associated with a lower odds of the clinical NAFLD phenotype among older adults without abdominal obesity (ie, with a normal waist circumference), although the lack of statistical significance necessitates confirmation of the trend in future studies. Using robust imaging techniques, we found that muscle fat infiltration was associated with liver fat in an aged population in whom maintenance of muscle health is a major concern (28). Importantly, our investigation illuminates a potential link between higher carbohydrate-supported ex vivo muscle mitochondrial energetics and lower liver fat, albeit no association between in vivo mitochondrial energetics and liver fat was observed. Finally, our study uncovered a robust inverse association between VO₂ peak and liver fat in this aged study sample.

Our results that capitalized on precise measurements of both muscle mass and liver fat highlight the need to consider abdominal adiposity as a modifying factor that may alter the association between the 2 tissues in older adults (5,14). We found that muscle mass and liver fat were directly associated in older adults with a high waist circumference and who may have greater visceral adipose depots (29), but trended toward being inversely associated among those without abdominal obesity (as hypothesized). This aligns with prior literature demonstrating that visceral adiposity confers metabolic risk even at higher relative quantities of lean mass (30). This general principle is likely magnified among older adults in whom adiposity as a proportion of total body mass, and in particular central adiposity, is exacerbated with aging and relates to functional decline (31). This may be because a higher waist circumference and therefore visceral fat could theoretically drive muscle fat infiltration and loss of muscle integrity. Although the point estimate suggests a 25% relatively higher liver fat per SD increment in muscle mass in older adults with a high waist circumference, reducing muscle mass would not be a treatment target. Instead, the focus should be on reducing abdominal adiposity, which might improve muscle metabolism. Indeed, muscle insulin resistance is a frequent comorbidity of central adiposity (32). Coupled with increased availability of free fatty acids or glucose from central adipose tissue, dysfunctional skeletal muscle unable to oxidize these substrates appropriately could drive de novo hepatic lipogenesis even in older adults with high relative muscle mass. In contrast, although our point estimate was small in magnitude, a higher relative muscle mass may be inversely associated with liver fat among older adults with a normal waist circumference independent of reported physical activity levels due to the maintenance of insulin sensitivity; whether molecular muscle characteristics such as mitochondrial energetics explain this trend (33) should be elucidated in prospective studies.

One reason for the lack of consensus in the literature regarding the association between muscle mass and liver fat may be due to the use of indirect muscle and liver fat (5,34,35) measurements or those that are complicated by changes in body water commonly present in chronic diseases and aging (eg, bioelectrical impedance analysis (35) or dual energy x-ray absorptiometry in the case of muscle mass (5)). The measurement methods we used circumvented these challenges. Despite having a comparable distribution of muscle mass (36) and liver fat (37,38) to contemporary population-based cohorts of older adults, our results may also be discordant with some prior research due to differences in participant demographics. Much of the available data describing the muscle-liver axis has been collected in younger (mean age ~45–60 years) study samples (5,35) comprised predominantly of participants with an Asian lineage (5,35), who may have noteworthy differences in regional body composition as compared to non-Hispanic White individuals (which represent the majority of SOMMA participants) (39). Finally, given that SOMMA was not expressly designed to recruit participants across the spectrum of NAFLD severity, study participants had relatively low levels of liver fat, although this is not necessarily an indicator of NAFLD progression (40). We emphasize that our results should be interpreted within the context of these cross-study differences.

Our results and those of other impactful research studies strongly point to the need to look beyond muscle mass and more deeply into how specific muscle characteristics relate to liver fat. One such characteristic is muscle fat infiltration. Our results suggest that an SD decrement in muscle fat infiltration is associated with a 21% relatively lower liver fat. Meanwhile, a 30% relative reduction in liver fat is considered to be clinically meaningful with respect to the resolution of histological symptoms of NASH (41). Prospective studies equipped to measure change in liver fat over time as well as NASH symptomatology should confirm the clinical relevance of our results. Nonetheless, our findings are echoed by those of a younger (63 ± 7.5 years) study sample from the population-based UK Biobank study (n = 5 326), in whom muscle fat infiltration was a superior correlate of liver fat than sarcopenia (low hand grip strength and height-adjusted appendicular lean mass) before adjustment for BMI (6). Another recent report similarly found that myosteatosis (a more direct measure of intramuscular fat) is a stronger determinant of early-stage non-alcoholic steatohepatitis (ie, a more advanced stage of NAFLD characterized by inflammation, ballooned hepatocytes, and/or fibrosis) (1) as compared to sarcopenia among middle-aged participants (9). We speculate that lower muscle fat infiltration, particularly in untrained individuals, may reflect greater skeletal muscle oxidative capacity through enhanced mitochondrial function or increased glucose disposal (5), thereby shunting less glucose to the liver and potentially lowering liver fat (42).

Indeed, our study revealed for the first time, to our knowledge, that muscle oxidative capacity—a key muscle characteristic contributing to cardiometabolic risk that is diminished in aging and obesity (20)—may be an important mechanistic component of the muscle-liver fat axis in older adults. Of note, liver mitochondrial energetics are impaired in animal models and in humans with NAFLD (33,43,44), but the association of muscle mitochondrial energetics with liver fat has never been described in humans using standardized ex vivo assays. Our results suggest that higher carbohydrate-supported skeletal muscle mitochondrial respiration (ie, Max OXPHOS) as measured ex vivo in permeabilized fiber bundles is associated with lower liver fat. The attenuation of this result through adjustment for potential confounding variables including prediabetes/diabetes status may point to the diabetes-related hallmark of insulin resistance as a key pathophysiologic factor linking lower mitochondrial respiration in both skeletal muscle and liver (33,45). Lower mitochondrial energetics and skeletal muscle insulin resistance could drive postprandial glucose flux to the liver where increased substrate-driven de novo lipogenesis may contribute to hepatic steatosis (42). In contrast to carbohydrate-supported OXPHOS, fatty acid-supported muscle mitochondrial OXPHOS was not correlated with liver fat. Although speculative, these results support a “reroute hypothesis,” which posits that more optimal carbohydrate-supported OXPHOS in peripheral tissues such as muscle prevents energy storage in hepatic tissues, preventing downstream liver insulin resistance and hepatic steatosis (22). Nonetheless, the magnitude of the unadjusted estimate was small, suggesting a ~9% relatively lower liver fat per SD higher Max OXPHOS. This should be explored further in prospective studies designed for this purpose and in individuals with diabetes specifically.

In contrast to our ex vivo measure of mitochondrial energetics (Max OXPHOS), we did not find a statistically significant association between ATP_max—our secondary measure of mitochondrial energetics taken in vivo—and liver fat in our sample of older adults. This negative result recapitulates the finding that in vivo mitochondrial function measured by ³¹P-MRS of the quadriceps muscle was not different in 17 participants with NAFLD compared with 18 age- and BMI-matched controls (11). This may be due to interindividual variability in blood flow or local substrate levels; or in the delivery, transport, and metabolism of substrates needed to fuel ATP_max that are standardized in the ex vivo respirometry assay (46). Thus, the standardization procedures conducted in the ex vivo assay may be necessary to observe potential linkages between muscle mitochondrial energetics and liver fat.

We and others have shown that oxidative capacity within the muscle (ie, muscle mitochondrial energetics) is related to whole-body oxygen consumption (ie, cardiorespiratory fitness measured as VO₂ peak), muscle mass, and whole-body metabolic health among older adults (8,47). Building on these findings, our exploratory analyses uncovered an inverse association between VO₂ peak and liver fat among our older adult study sample, which was robust to adjustment for confounding variables including prediabetes/diabetes status. This result echoes prior literature conducted in younger study samples, including an interventional study in which results withstood adjustment for fat mass loss, despite VO₂ peak estimation using cycle ergometry (as opposed to CPET—the gold-standard measurement used in our study) (48). Prospective studies should confirm the magnitude of the effect estimate (~13% relatively lower in liver fat per 1-SD increment in VO₂ peak), which is roughly half of what is considered clinically relevant. Future studies should also investigate whether VO₂ peak mediates the association between muscle mitochondrial energetics with liver fat, as an enhanced oxidative capacity of muscle mitochondria could theoretically prevent ectopic fat accumulation in the liver while raising VO₂ peak (47,49). Furthermore, VO₂ peak, mitochondrial energetics, and functional measures such as walk speed are closely related (8), so elucidating the causal pathway by which these variables relate to liver fat will be key to developing interventions that address both liver fat and functional outcomes in older adults. Taken together, these findings support the premise that specific muscle characteristics and associated phenotypes may be key for understanding muscle-liver crosstalk, particularly in the aging population in whom central adiposity may render muscle mass less effective at protecting against liver fat accumulation. Central adiposity appears to be a relevant modifier of the association between skeletal muscle mass and liver fat, but not for associations of muscle-related characteristics and liver fat. These hypothesis-generating results require confirmation in prospective studies.

This study includes several limitations and strengths. We were unable to test the directionality of the muscle-liver association in our cross-sectional study, which is important to consider given that the liver communicates with muscle tissue via circulating hepatokines (50), delivers excess lipid to muscle via lipoproteins, and can contribute to muscle inflammation that in turn drives sarcopenia (5,51). Indeed, people with advanced stages of NAFLD, such as fibrosis and cirrhosis, often have reduced muscle mass and impaired muscle function (52) and these characteristics are associated with poor prognosis (eg, hospitalizations, ascites, and encephalopathy) (53). Directionality can be evaluated in future studies with collection of longitudinal data. Our study was underpowered to test for interaction between Max OXPHOS and diabetes status for the outcome of liver fat, which may be a critical element to include in follow-up research to elucidate mechanisms of muscle-liver crosstalk. Although unlikely, some participants who self-reported as having diabetes may have had type 1 diabetes, in whom mechanisms linking muscle and liver phenotypes may be unique. Our analysis is subject to residual confounding as is the case with virtually all observational epidemiologic studies. The primary strength of this study is the rigor and breadth of our measurement methods for determining muscle characteristics and liver fat content in an aged cohort, which gives rise to hypotheses about mechanisms like skeletal muscle mitochondrial energetics and liver fat.

In conclusion, our results reveal muscle characteristics that are associated with liver fat in older adults, suggesting that muscle and liver health are linked potentially via specific mechanisms and biomarkers. Our finding that both muscle mitochondrial bioenergetics and cardiorespiratory fitness are associated with liver fat accumulation suggests that clinical assessments of fitness, coupled with standard biochemical measures, could provide a scalable approach to evaluate risk factors for NAFLD in the aging population. Future studies with prospective designs should confirm and expand upon these results using both lifestyle and pharmacological interventions. In particular, exercise interventions may serve as a model for studying the complex links amongst muscle characteristics and liver fat in older adults as exercise increases cardiorespiratory fitness, muscle mass (54), and muscle mitochondrial respiration (13) while reducing liver fat. Such study designs will enrich our knowledge of the directionality and mechanistic underpinnings of the muscle-liver axis and ultimately inform optimization of NAFLD treatment and prevention in the aging population (55).

Funding

Study of Muscle, Mobility and Aging (SOMMA) is funded by the National Institute on Aging (NIA) (grant number R01AG059416). Study infrastructure support was funded in part by NIA Claude D. Pepper Older American Independence Centers at University of Pittsburgh (P30AG024827), Wake Forest University School of Medicine (P30AG021332), and the Clinical and Translational Science Institutes which is funded by the National Center for Advancing Translational Science at Wake Forest (UL1TR001420).

Conflict of Interest

S.R.C. and Pe.M.C. are consultants to Bioage Labs. J.L. is an employee of AMRA Medical AB and reports consultancy/speaking honoraria from Eli Lilly and BioMarin. The other authors report no conflict of interest.

Data Availability

The data that support the findings of this study are available from the Study of Muscle, Mobility and Aging (SOMMA). As of August 10, 2023, data can be obtained via https://sommaonline.ucsf.edu

Author Contributions

Pa.M.C., K.D.C., and D.I. contributed substantially to the conception and design of the manuscript. D.I., J.M., Pa.M.C., and K.D.C. wrote the original draft. D.I. conducted the formal analysis and curated the data with input from S.H., Pe.M.C., and J.L. Pa.M.C., Pe.M.C., B.H.G., S.R.C., R.T.H., S.H., J.L., and S.B.K. were involved in SOMMA study investigation. S.R.C., R.T.H., and S.B.K. acquired funding. All study authors, including M.J.J. and D.M., participated in extensive review and editing of the manuscript.

Ethics Approval

All research was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent, and the study was approved by the WIRB-Copernicus Group Institutional Review Board (20180764) for all participating sites.

Participant Informed Consent

All patients provided written informed consent for the anonymous use of their data for research purpose.