Standardized measurement of abdominal muscle by computed tomography: association with cardiometabolic risk in the Framingham Heart Study

Objectives To provide a standard for total abdominal muscle mass (TAM) quantification on computed tomography (CT) and investigate its association with cardiovascular risk in a primary prevention setting. Methods We included 3016 Framingham Heart Study participants free of cardiovascular disease (CVD) who underwent abdominal CT between 2002 and 2005. On a single CT slice at the level of L3/L4, we segmented (1) TAM-Area, (2) TAM-Index (= TAM-Area/height) and, (3) TAM-Fraction (= TAM-Area/total cross-sectional CT-area). We tested the association of these muscle mass measures with prevalent and incident cardiometabolic risk factors and incident CVD events during a follow-up of 11.0 ± 2.7 years. Results In this community-based sample (49% women, mean age: 50.0 ± 10.0 years), all muscle quantity measures were significantly associated with prevalent and incident cardiometabolic risk factors and CVD events. However, only TAM-Fraction remained significantly associated with key outcomes (e.g., adj. OR 0.68 [0.55, 0.84] and HR 0.73 [0.57, 0.92] for incident hypertension and CVD events, respectively) after adjustment for age, sex, body mass index, and waist circumference. Moreover, only higher TAM-Fraction was associated with a lower risk (e.g., adj. OR: 0.56 [0.36–0.89] for incident diabetes versus TAM-Area: adj. OR 1.26 [0.79–2.01] and TAM-Index: 1.09 [0.75–1.58]). Conclusion TAM-Fraction on a single CT slice at L3/L4 is a novel body composition marker of cardiometabolic risk in a primary prevention setting that has the potential to improve risk stratification beyond traditional measures of obesity. Key Points • In this analysis of the Framingham Heart Study (n = 3016), TAM-F on a single slice CT was more closely associated with prevalent and incident cardiometabolic risk factors as compared to TAM alone or TAM indexed to body surface area. • TAM-F on a single abdominal CT slice at the level of L3/L4 could serve as a standard measure of muscle mass and improve risk prediction


Introduction
Cardiovascular disease (CVD) is the leading cause of death in the USA and worldwide [1]. Physical inactivity is associated with cardiometabolic risk factors and is highlighted as one of the key modifiable risk factors for CVD in the current primary prevention guidelines [2]. Because reliable quantification of daily routine physical activity is challenging, measures of body composition, including obesity and muscle mass, have been proposed as possible objective surrogate markers because they closely associate with the level of physical activity and cardiorespiratory fitness [3,4].
However, in clinical research, muscle mass has been evaluated predominantly in the context of sarcopenia in patients with chronic diseases, such as chronic lung disease, valvular heart disease, or cancer, demonstrating an association of lower muscle mass with increased morbidity and mortality [5 -8]. For example, among 2115 obese patients with solid tumors of the respiratory and gastrointestinal tract, low muscle mass on CT was associated with a fourfold increase in mortality [9]. These findings have been explained by muscle wasting due to cancer or chronic disease but are considered a separate matter from cardiovascular effects. In the cardiovascular area, high muscle mass has been predominantly used as a surrogate marker for metabolic fitness, associated with improved prognosis. For instance, in patients with coronary artery disease undergoing percutaneous intervention, those with higher muscle mass showed the lowest risk for mortality and MACE, which is interpreted as a positive effect of higher physical activity for CV outcomes [10].
In contrast to a great body of literature on muscle mass in patients with chronic disease [5 -9], limited data are available to describe normal distributions of muscle mass in healthy individuals [11]. Low muscle mass on CT in individuals free of chronic disease has been linked to nonalcoholic fatty liver disease [12], and impaired pulmonary function [13]; however, its association with cardiometabolic risk is incompletely understood [11,14].
Another issue is that no standardized quantitative assessment of TAM exists. The aforementioned include several anatomical locations, including the psoas [5], pectoralis [6], paraspinal [7], and using absolute or relative muscle mass.
To determine whether different measures of abdominal muscle are associated with cardiometabolic risk beyond traditional measures of obesity, we used an established data set of abdominal CT in the community-based Framingham Hearts Study [15,16] to measure muscle mass as an area on a single CT slice at the level of L3/L4, which represents an easy and straight forward measure. Here, we describe normal distributions of muscle mass in the community and assess the associations of various muscle measures with prevalent and incident cardiometabolic risk factors as well as incident CVD events and propose a standard for muscle mass quantification when assessing its association with CVD risk.

Study cohort
We included participants from the Framingham Heart Study, including the Offspring and 3 rd -generation cohorts, who underwent routine thoracic and abdominal multidetector computed tomography (MDCT) between 2002 and 2005 [15]. Of the 3,529 participants in the MDCT sub-study, 3,253 were free of artifacts and evaluable for TAM-Area, TAM-Index, and TAM-Fraction. We excluded 237 participants who had prevalent CVD, resulting in a final sample size of 3016 individuals. Each patient gave informed consent and the study protocol was approved by the Institutional Review Board.

Definition of cardiometabolic risk factors and CVD events
The definition of cardiometabolic risk factors and CVD events in the Framingham Heart Study have been described elsewhere in detail [17]. In short, arterial hypertension was defined as systolic blood pressure ≥ 140 mm Hg or diastolic blood pressure ≥ 90 mm Hg or antihypertensive medication, and diabetes as a fasting plasma glucose level ≥ 126 mg/dL or treatment with either insulin or a hypoglycemic agent. Metabolic syndrome was determined according to the definition of Modified National Cholesterol Education Program Adult Treatment Panel III guidelines [18]. Body mass index (BMI), as a measure of general adiposity, was defined as the weight (kilograms) divided by height squared (meters), and waist circumference (WC), a measure of central adiposity, was measured at the level of the umbilicus.
The physical activity index, a composite score, was calculated by adding the weighted time spent in each activity as described previously [19].
We defined incident CVD as fatal or nonfatal myocardial infarction, fatal or nonfatal stroke, angina, atherothrombotic infarction of the brain, intermittent claudication, heart failure, or CVD death.

Computed tomography and quantification of muscle mass
The technical imaging details have been extensively described elsewhere [15,20]. In short, participants underwent 8-slice MDCT imaging of the abdomen and 25 continuous 5-mm thick slices were recorded, covering 125 mm above the level of S1 vertebral level. We used the open-access software "3D Slicer," version 4.8.1 [21], to quantify muscle area on a single slice at the level of L3/L4 based on previous reports [22]. This software allows for easy segmentation processes and offers macro-based solutions for fast imaging post-processing (see Fig. 1).
The abdominal slice at the level L3/L4 was identified manually and each slice was checked for artifacts. Only slices without any artifacts were used for analysis. Muscle was defined based on signal intensity (−50 to 150 HU).
We report three different CT measures of muscle mass, including (A) TAM area (TAM-Area), giving the area of muscle on the single L3/L4 slice in a square centimeter, (B) TAM-Index, indexing TAM-Area to the individual's height squared, and (C) TAM-Fraction, which is the percentage of TAM in relation to the entire cross-sectional area at the level of L3/L4.

Statistical analysis
We report mean and standard deviation (SD) and total numbers and percentages for descriptive statistics.
Linear, logistic, and Cox proportional hazards regression were performed to assess the associations of measures of muscle mass (TAM-Area, TAM-Index, and TAM-Fraction) and the following: (A) prevalent cardiometabolic risk factors, including systolic blood pressure (SBP), fasting glucose, cholesterol levels, triglyceride levels, Framingham Risk Score, hypertension, diabetes, and metabolic syndrome; (B) incident cardiometabolic risk factors (hypertension, diabetes, metabolic syndrome), and C) incident CVD. Cross-sectional comparisons were performed at the baseline exam at the time of the CT. For models predicting incident cardiometabolic risk factors, participants who had a prevalent cardiometabolic risk factor of interest at or before baseline were excluded.
To allow for better comparison between different measures of muscle mass, beta coefficients (adj. beta), odds ratios (adj. OR), and adjusted hazard ratios (adj. HR) are reported per a 1-SD increase for each muscle mass measurement.
We furthermore report age-and sex-adjusted Pearson correlation coefficients to describe the association between measures of muscle mass with anthropometric measurements, including BMI and WC, and established risk factors, including the Framingham Risk Score. A two-sided alpha level of 0.05 or less was considered statistically significant. All statistical analyses were performed using SAS v9.4.

Study population
A total of 3016 participants (1088 Offspring and 1928 thirdgeneration cohorts) free of CVD were analyzed. Table 1 lists the baseline characteristics of the entire study cohort. The mean age was 50.0 ± 10.0 years and 1472 (48.8%) were women. Prevalence of arterial hypertension, diabetes, and metabolic syndrome was 26.3%, 4.9%, and 24.3%, respectively. The mean BMI was 27.4 kg/m 2 (SD = 4.8) and a total of 25.4% were labeled as obese based on a BMI of ≥ 30 kg/m 2 .

Association of muscle mass with anthropometric measures
There was a significant correlation between all measures of CT muscle mass and BMI and WC ( Table 2, p < 0.0001 for all; Fig. 2). After adjustment for age and sex, TAM-Area and TAM-Index were positively correlated with BMI (r = 0.578 and 0.621, p < .0001 each) and WC (r = 0.470 and 0.420, p < .0001 each). In contrast, TAM-Fraction was negatively correlated with both anthropometric measures of adiposity (r = −0.655 and −0.662, for BMI and WC respectively; p < .0001 for both). Table 3 depicts differences in TAM-Area, TAM-Index, and TAM-Fraction between individuals without and with BMI indicative of obesity. Individuals with obesity had significantly higher TAM-Area (182.9 ± 45.3 versus 156.9 ± 40.8 kg/m 2 , p < .0001) and TAM-Index (62.5 ± 11.4 versus 53.4 ± 10.2 cm 2 /m 2 , p < .0001) but lower TAM-Fraction (0.20 ± 0.04 versus 0.26 ± 0.06, p < .0001) as compared to individuals without obesity. We observed the same pattern when comparing individuals with and without hypertension and diabetes. Patients with hypertension and diabetes had higher TAM-Area and TAM-Index but lower TAM-Fraction when compared to their disease-free counterparts (details, see Table 3).

Association of muscle mass with prevalent cardiometabolic risk factors
TAM-Area, TAM-Index, and TAM-Fraction were significantly associated with prevalent cardiometabolic risk (Tables 4 and 5). However, the directionality of the association was substantially different between the measures. In general, higher TAM-Area and TAM-Index were associated with an altered risk profile (e.g., for TAM-Area: adj. Beta  In contrast, higher TAM-Fraction was associated with significantly more favorable risk profile (e.g., adj. Beta −4.16 [95% CI: −4.86, −3.46], p < .0001 for systolic blood pressure, per 1-SD increase, adjusted for age and sex) and lower  After further adjustment for anthropometric measures of overall and central adiposity (BMI and WC), TAM-Area and TAM-Index were not significantly associated with key risk factors (see Table 2). However, higher TAM-Fraction

Association with incident cardiometabolic risk factors
During 6.4 years of follow-up, a total of 192 (7.0%) incident cases of diabetes were observed, as were 1,005 (36.4%) of hypertension, and 825 (30.5%) of metabolic syndrome.
The association between measures of muscle mass with incident cardiometabolic events is displayed in Table 6.
Similar to the association with prevalent cardiometabolic risk, in age-and sex-adjusted analyses, higher TAM-Area and TAM-Index were associated with a higher risk for developing hypertension, diabetes, and CVD. In contrast, higher TAM-Fraction was protective against all these incident findings. After further adjustment for BMI and WC, only TAM-Fraction was significantly associated with lower risk for incident outcomes ( Figure 2 displays the opposite direction of association between TAM-Area and TAM-Index, compared with TAM-Fraction, with BMI and WC. Figure 3 displays the different directionality between TAM-Area, TAM-Index, and TAM-Fraction with prevalent and incident outcomes.

Discussion
In a community-based cohort of participants free of CVD, muscle mass assessed on a single-slice abdominal CT scan associates with cardiometabolic and CVD risk. However, among different CT measures of muscle mass, only for TAM-Fraction (less muscle) was associated with higher cardiometabolic risk, suggesting that further investigations defining a new obesity phenotype to improve risk stratification may primarily focus on TAM-Fraction among the various ways to measure muscle. Physical activity and avoidance of excessive adipose tissue play an important part in cardiometabolic disease prevention. This is highlighted in the 2019 ACC/AHA Guideline on the Primary Prevention of Cardiovascular Disease, which promotes at least 150 min per week of accumulated moderateintensity physical activity or 75 min per week of vigorousintensity physical activity [2]. This recommendation targets primarily the avoidance of obesity primarily as a key risk factor for cardiometabolic risk, similar to the role of coronary artery calcification as a risk factor for CVD [2]. However, standardized quantification of physical activity is challenging as self-reported data lack accuracy [23,24] and functional assessment of cardiorespiratory fitness is not feasible in a population-based setting [25].
Assessment of BMI as a marker of general adiposity is easy and recommended annually in all adults and WC as an indicator for central obesity in individuals at higher cardiometabolic risk [2]. In contrast, no standardized measure is available for muscle mass, another key body composition component besides obesity. Muscle mass has been shown to correlate with the level of physical activity [3]. It can easily be assessed and quantified on computed tomography (CT) and magnetic resonance imaging (MRI), commonly used imaging techniques for various indications. In chronic disease, muscle mass on CT as a surrogate for physical fitness is well established. Low muscle mass-interpreted as a disease-associated loss of function-is associated with worse outcomes. This has been shown for patients with lung disease [26], valvular heart disease [5], coronary artery disease [10], and especially cancer [27].
In contrast to patients with chronic disease, only a few studies report on muscle mass in population-based settings [12,13] and its association with CV risk is unknown.
Several factors may contribute to the fact that muscle mass is not clinically used as a body composition component for cardiometabolic risk assessment, including a lack of a standardized approach to assess and report muscle mass. Different anatomic landmarks have been proposed, including the psoas muscle [5], the pectoralis muscle [6], paraspinal muscles [7], or skeletal muscle mass at the thigh level [8]. This substantially hampers direct comparison between studies. In our study, we used a standardized approach to quantify muscle mass at the level of L3/L4 on a single abdominal CT slice. This has been shown to correlate almost perfectly with volumetric data of adipose tissue in the Framingham Heart Study [22].
Furthermore, some authors report total muscle area (TAM-Area) on a single or multiple tomographic slices [28,29], whereas others index muscle mass to body height or body surface area (TAM-Index) [5,12]. We chose to measure both not-indexed TAM-Area as well as indexed to height because these were reported in previous studies [5,12,28,29]. In our cohort, we provide for the first time descriptive data for TAM-Area and TAM-Index and, in addition, introduce the concept of TAM-Fraction as a novel body composition tool, representing the fraction of muscle mass of the entire tomographic L3/ L4 slice. Comparing these three measures of muscle mass, we observed a different directionality in the associations of these measures with the risk of outcome events. Higher TAM-Area and TAM-Index were associated with a worse cardiometabolic risk profile in age and sex-adjusted models. Similar findings are reported by an analysis of the Multi-Ethnic Study of Atherosclerosis (MESA), where higher TAM-Area, averaged from several single slices between L2 and L5, was associated with more CAC burden. This was in contrast to the authors' hypothesis of a protective effect of higher muscle mass [29] but is similar to our results for higher TAM-Area and TAM-Index indicating higher cardiometabolic risk. This might, in part, be explained by a close positive correlation between TAM-Area and TAM-Index and both BMI and WC.
In contrast, TAM-Fraction is negatively correlated with BMI and WC. Indeed, higher TAM-Fraction indicated a beneficial cardiometabolic risk profile across a wide spectrum of prevalent and incident risk factors. Higher TAM-Fraction, but not TAM-Area or TAM-Index, remained significantly associated with lower prevalence of several cardiometabolic risk factors, including hypertension and metabolic syndrome, and lower risk for new-onset hypertension and diabetes and CVD events even when adjusted for BMI and WC.
While the purpose of this paper was to introduce the concept of TAM-Fraction as a novel body composition tool, representing the fraction of muscle mass of the entire tomographic L3/L4 slice, future studies are warranted to investigate the interplay between TAM-Fraction and more advanced measures of adiposity, including visceral and subcutaneous adipose tissue, with the potential to define a more precise individual obesity phenotype for improved CV risk prediction.

Limitations
Several limitations merit comment. As with any analyses from the Framingham Heart Study, we were not able to assess ethnic/race-related differences due to our predominant white study sample. In addition, CT scans were performed between 2002 and 2005, when demographic characteristics, such as obesity rates, were different compared to the present [30]. This analysis also lacks information on fat infiltration within the skeletal muscle, which may be an important aspect.
Funding Open access funding provided by Medical University of Vienna. The authors state that this work has not received any funding.

Declarations
Guarantor The scientific guarantor of this publication is Dr. Andreas Kammerlander Statistics and biometry Dr. Lyass and Dr. Massaro kindly provided statistical advice for this manuscript.
Informed consent Written informed consent was obtained from all subjects (patients) in this study.
Ethical approval Institutional Review Board approval was obtained.

Methodology
• prospective • observational • performed at one institution Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.