Applicability of ultrasound muscle thickness measurements for predicting fat-free mass in elderly population

Abstract

Objective: This study aimed to examine the applicability of ultrasound muscle thickness (MT) measurements for predicting whole body fat-free mass (FFM) in elderly individuals. Design and setting: Crosssectional study of 77 healthy elderly individuals. Methods: MTs at nine sites of the body and FFM were determined using B-mode ultrasound and dual-energy x-ray absorptiometry (DXA), respectively, in 44 women and 33 men aged 52 to 78 yrs. Stepwise multiple regression analysis produced two equations for predicting DXA-based FFM with sex (dummy: woman = 0 and man = 1) and either MTs at the anterior and posterior of thigh and lower leg (Eq1) or the product of MT and limb length (MT×LL) at thigh anterior and posterior, lower leg posterior, and upper arm anterior (Eq2) as independent variables. Results: The R2 and SEE for each of the two equations were 0.929 and 2.5 kg for Eq1 and 0.955 and 2.0 kg for Eq2. The estimated FFM from each of Eq1 (44.4 ± 8.9 kg) and Eq2 (44.4 ± 9.0 kg) did not significantly differ from that of the DXA-based FFM (44.4 ± 9.2 kg), without systematic error. However, the absolute value of the difference between the DXA-based and estimated FFM was significantly greater with Eq1 (2.0 ± 1.5 kg) than with Eq2 (1.5 ± 1.3 kg). Conclusion: The current results indicate that ultrasound MT measurement is useful to predict FFM in the elderly, and its accuracy is improved by using the product of MT and limb length as an independent variable.

Introduction

An assessment of the total body fat-free mass (FFM) for elderly individuals is important, because it is considered to be an indicator of the status of sarcopenia, bone mineral content, or morbidity in the corresponding age group (1). Hydrodensitometry has been accepted as the criterion method for assessing body composition, but dual-energy X-ray absorptiometry (DXA) is gradually replacing it (2). DXA is safe, with a low radiation exposure, and requires minimal participation cooperation, and is fast and easy to use (2), whereas there is spatial limitation in the view point of convenient assessment for survey of a large sample.

Another approach which has a high potentiality for assessing FFM in the elderly population is the use of measures of muscle thickness (MT) as independent variables, determined by brightness mode (B-mode) ultrasonography. B-mode ultrasonography has the same advantage as computerized tomography or magnetic resonance imaging (MRI) in visualizing fat and muscle tissues without compression (3). B-mode ultrasound has been successfully used to evaluate MT in not only young but also middle-aged and elderly populations (4, 5, 6, 7). In addition, the measurements of B-mode ultrasound-based MTs can be conducted easily in clinical and field surveys for a large sample with no hazardous effect (8, 9, 10), and the measured values can be significant predictors of limb muscle volume 3, 2, 12 and total muscle mass (13). These characteristics support the applicability of B-mode MT measurements for predicting FFM. Indeed, Abe et al. (14) have developed an ultrasound-based prediction for FFM. In their study, however, the ages of the subjects ranged from 18 to 51 years. It is known that age-related loss of total muscle mass accelerates after the end of the fifth decade (15). The influence of aging on MT has a site-related difference (5). In both men and women, the age-related loss of MT has been shown to be greater in the abdomen and thigh anterior than in other body parts. These findings suggest that even if ultrasound-based MTs are applicable to predict FFM, it is necessary to develop the prediction equation for elderly individuals.

The present study determined MTs and FFM in elderly individuals of both sexes using B-mode ultrasonography and DXA, respectively, and tried to develop ultrasound-based prediction equations for DXA-based FFM. The purpose of the current study was to examine the applicability of the ultrasound-based MTs for predicting DXA-based FFM in elderly individuals.

Methods

Subjects

A group of 44 women aged 53 to 77 yrs (65.5 ± 6.6 yrs) and 33 men aged 52 to 78 yrs (63.9 ± 7.9 yrs) voluntarily participated in this study. The descriptive data of height, body mass, and body mass index, and limb length for the subjects are summarized in Table 1. None of the subjects was or had been an athlete. In addition, no participant was osteoporosis, or on extreme diets or was using any major medications, such as chemotherapy, cardiac, respiratory, or antipsychotic drugs. Moreover, none were using sticks or other walking aids and all were functionally independent in daily life. This study was approved by the Ethics Committee on Human Research of Waseda University and was consistent with their requirements for human experimentation. The subjects were fully informed of the purpose and risks of the experiment and gave their written informed consent.

Table 1.

Physical characteristics of subjects

Variables	Total (n = 77)	Men (n = 33)	Women (n = 44)
Age, yrs	64.8 ± 7.2	63.9 ± 7.9	65.5 ± 6.6
Height, cm	159.3 ± 7.8	166.5 ± 5.5	153.8 ± 3.8
Body mass, kg	57.4 ± 9.5	65.5 ± 7.0	51.2 ± 5.6
Body mass index,	22.5 ± 2.5	23.7 ± 2.2	21.7 ± 2.3
kg/m2
Limb length, cm
Forearm	21.6 ±1.5	22.7 ± 1.3	20.7 ± 0.9
Upper arm	30.0 ± 1.8	31.4 ± 1.5	28.9 ± 1.2
Lower leg	35.4 ±2.3	37.0 ± 2.0	34.2 ± 1.7
Thigh	36.1 ± 2.2	37.8 ± 1.9	34.8 ± 1.5

Values are means ± SDs. Forearm length, distance between the styloid process and the head of the radius; Upper arm length, distance between the lateral epicondyle of the humerus and the acromial process of the scapula; Lower leg length, distance between the lateral malleous of the fibula and the lateral condyle of the tibia; Thigh length, distance between the lateral condyle of the femur and the greater trochanter

DXA measurements

A whole-body DXA scanner (Hologic Delphi A-QDR, USA) was used to determine fat mass, bone mineral content, and bone-free lean tissue mass. The radiation dose was less than 0.03 mrem (= 0.3 mSv). Total body image acquisition and analysis were conducted following the manufacturer's instructions. Fat mass, bone mineral content and bone-free lean tissue mass were calculated using software provided by the manufacturer. The sum of bone mineral content and bone-free lean tissue mass was referred to as FFM. A radiation technologist performed DXA measurement and data analysis. The DXA measurement has been shown to be an accurate method for measurement of FFM and fat mass for the elderly 16, 17.

MT measurements

A real time B-mode ultrasound apparatus (SSD-900, Aloka Co., Tokyo) was used to obtain cross-sectional images at nine sites on the right side of the body. The position of the subjects during the ultrasonographic measurements and the site selected for the measurements were the same as those described in a previous study (14). During the measurements, the subjects remained in a standing position with the legs and arms straight and muscles relaxed. Firstly, The measurement sites were precisely located and marked at the anterior and posterior surfaces 60% distal to the lateral epicondyle of the humerus from the acromial process of the scapular, and at the anterior and posterior surface in the middle of the thigh length (the distance from the greater trochanter of the femur to articular cleft between the femur and tibial condyles), and at the anterior and posterior surface in the proximal 30% of the lower leg length (the distance from the articular cleft to lateral malleolus). The length of the segment was determined using a measuring tape. The anatomical landmarks for the measurement sites are presented in Figure 1.

Figure 1.

The anatomical landmarks for the 9 sites of muscle thickness measurement using ultrasound

The scanning head together with water-soluble transmission gel, which provided acoustic contact without depression of the skin, was placed perpendicular to the tissue interface at each of the marked sites. Distortion of tissue due to excessive compression was eliminated by ensuring that no movement of tissues occurred in the real-time ultrasonic image. MT was determined in accordance with the procedure described in an earlier study (14). The interfaces between subcutaneous adipose tissue and muscle and between muscle and bone were identified from the ultrasonic image, and the distance from the adipose tissue-muscle interface to the muscle-bone interface was measured as representative of MT. The MT at the anterior of lower leg was measured from the fat-muscle interface to the tibial anterior muscle-tibialis posterior muscle interface. The MT at the abdomen was measured from the fat-muscle interface to the muscle-abdominal cavity boundary interface. The measurements were taken with a measure to the nearest 0.5 mm. The accuracy and test-retest repeatability of the MT measurements were certified in prior studies 3, 12. Miyatani et al. (12) reported that MT was a good predictor of limb muscle volume determined by MRI when combined with limb length. Adding to MT, therefore, the products of MT at each of 7 sites of the limbs and the corresponding limb length (MT×LL) were used as independent variables in regression analyses with DXA-based FFM as a dependent variable.

Statistical analysis

We developed two equations for predicting DXA-based FFM by a stepwise multiple regression analysis. One equation (Eq1) was developed by analyzing MT values at the nine sites as independent variables, and the other (Eq2) was developed by analyzing MT×LL at the seven sites as independent variables. For the two cases, sex was coded as a dummy variable: woman = 0 and man = 1, and added as an independent variable. In addition to the test of the significance of the difference between the DXA-based and estimated FFM values, the difference between the two variables (the DXA-based FFM - the estimated FFM) was plotted against the mean FFM of the two methods to examine for systematic error, as described by Bland and Altman (18). The standard error of the estimate (SEE) was calculated to evaluate the accuracy of the estimate with the each equation. For each independent variable selected, the product of the standard regression coefficient in the multiple regression equation and the simple correlation coefficient in the relationship with the DXA-based FFM, expressed as a percentage, was calculated as an index presenting its relative contribution in estimating the DXA-based FFM. A student's paired t-test was used to test the significance of differences between the measured and estimated FFM values and between the two equations in the absolute value of the difference between the DXA-based and estimated FFM values. Descriptive data are presented as means ± SDs. The probability level for statistical significance was set at P < 0.05. All data analyses were conducted using a statistical software (SPSS 19.0 for windows, IBM, Japan).

Results

Descriptive data of MT and DXA measurements are shown in Table 2 and Table 3, respectively. The average value of the total mass obtained by summing up fat mass, bone-free lean tissue mass, and bone mineral content, as determined by DXA, was significantly greater than that of body mass measured by conventional weighing (Table 1), but the difference was small (0.7 ± 0.4 kg) and the two variables were almost identical.

Table 2.

Descriptive data of muscle thickness measurements

Sites	Total (n = 77)	Men (n = 33)	Women (n = 44)
Forearm, cm	1.7 ± 0.3	1.8 ± 0.2	1.5 ± 0.2
Upper arm anterior, cm	2.9 ± 0.4	3.2 ± 0.3	2.7 ± 0.3
Upper arm posterior, cm	2.8 ± 0.5	3.1 ± 0.4	2.5 ± 0.4
Thigh anterior, cm	4.2 ± 0.7	4.7 ± 0.6	3.9 ± 0.6
Thigh posterior, cm	5.4 ± 0.8	6.2 ± 0.6	4.9 ± 0.5
Lower leg anterior, cm	2.8 ± 0.3	3.1 ± 0.2	2.6 ± 0.2
Lower leg posterior, cm	6.3 ± 0.7	6.9 ± 0.5	5.7 ± 0.4
Subscapular, cm	1.8 ± 0.4	2.0 ± 0.4	1.7 ± 0.4
Abdomen, cm	1.0 ± 0.3	1.2 ± 0.3	0.8 ± 0.2

Values are means ± SDs.

Table 3.

Descriptive data of DXA measurements

Variables	Total (n = 77)	Men (n = 33)	Women (n = 44)
FM, kg	13.7 ± 3.7	12.4 ± 3.4	14.6 ± 3.6
%FM, %	23.9 ± 6.3	18.5 ± 3.5	27.9 ± 4.8
LTM, kg	42.6 ± 8.9	51.8 ± 4.7	35.7 ± 2.9
BMC, kg	1.8 ± 0.4	2.2 ± 0.3	1.5 ± 0.3
FFM, kg	44.4 ± 9.2	54.0 ± 4.9	37.2 ± 3.0
Total mass, kg	58.1 ± 9.6	66.4 ± 7.0	51.9 ± 5.7

Values are means ± SDs. FM, fat mass; %FM, FM relative to total mass; LTM, bone-free lean tissue mass; BMC, bone mineral content; FFM, fat-free mass; Total mass, the sum of FM, LTM, and BMC

From the results of multiple regression analysis, sex and MT values at four sites and sex and MT×LL at four sites were selected in Eq1 and Eq2, respectively, as significant predictors for DXA-based FFM (Table 4). The contributions of sex and MT values at four sites in Eq1 were 35.3% for sex, 10.9% for thigh anterior, 17.6% for thigh posterior, 10.4% for lower leg anterior, and 18.7% for lower leg posterior, respectively. Those of sex and MT×LL at the four sites in Eq2 were 25.5% for sex, 9.0% for upper arm anterior, 14.4% for thigh anterior, 20.9% for thigh posterior, and 25.7% for lower leg posterior, respectively.

Table 4.

Multiple regression model for predicting DXA-based FFM from muscle thickness (MT, Eq1) and product of MT and limb length (MT × LL, Eq2)

Eq1, DXA-based FFM (kg) = (X1 × 7.217) + (X2 × 1.985) + (X3 × 2.355) + (X4 × 3.633) + (X5 × 2.670) - 6.759

Eq2, DXA-based FFM (kg) = (X1 × 5.233) + (X6 × 0.006630) + (X7 × 0.05153) + (X8 × 0.05579) + (X9 × 0.07097) + 1.774

Equation	Variables	r2	Regression coefficient	Standard regression coefficient	Contribution %
Eq1	X1, (sex: female = 0, male = 1)	0.821	7.217	0.390	35.5
	X2, MT at thigh anterior	0.513	1.985	0.152	10.9
	X3, MT at thigh posterior	0.719	2.355	0.207	17.6
	X4, MT at lower leg anterior	0.661	3.633	0.128	10.4
	X5, MT at lower leg posterior	0.799	2.670	0.209	18.7
Eq2	X1, (sex: female = 0, male = 1)	0.821	5.233	0.282	25.5
	X6, MT × LL at upper arm anterior	0.711	0.06630	0.107	9.0
	X7, MT × LL at thigh anterior	0.676	0.05153	0.175	14.4
	X8, MT × LL at thigh posterior	0.830	0.05579	0.229	20.9
	X9, MT × LL at lower leg posterior	0.878	0.07097	0.274	25.7

The units of MT and LL are centimeter (cm)

The relations between the DXA-based FFM and estimated FFM values for each of the two equations are shown in Figure 2. The R2 and SEE for each of the two equations were 0.929 and 2.5 kg, respectively, for Eq1 and 0.955 and 2.0 kg, respectively, for Eq2. The average value of the FFM estimated from each of Eq1 (44.4 ± 8.9 kg) and Eq2 (44.4 ± 9.0 kg) did not significantly differ from that of the DXA-based FFM, without a significant systematic error in Bland-Altman plots (Figure 3). However, the absolute value of the difference between the DXA-based FFM and estimated FFM values was significantly greater with the Eq1 (2.0 ± 1.5 kg) than with the Eq2 (1.5 ± 1.3 kg).

Figure 2.

Regression between fat-free mass estimated by the multiple regression equation (Estimated FFM) and that measured by DXA (DXA-based FFM). Open circles are expressed as female, and filled circles as male. The estimated FFM in each of the right and left panels was calculated using Eq1 and Eq2, respectively, listed in Table 4. The dash line is an identical line

Figure 3.

Difference between DXA-based and the estimated FFM by multiple regression equation (Estimated FFM) vs. the mean values of FFM determined by the two methods. Open circles are expressed as female, and filled circles as male. The estimated FFM in each of the right and left panels was calculated using Eq1 and Eq2, respectively, listed in Table 4. The dotted horizontal lines are mean differences and 95% CIs

Discussion

The regression analyses produced two prediction equations with sex and either MT values at the four sites (Eq1) or MT×LL at the four sites (Eq2) as significant predictors of the DXA-based FFM (Table 4). The R2 and SEE for each of the two equations were 0.929 and 2.5 kg, respectively, for Eq1 and 0.955 and 2.0 kg, respectively, for Eq2. The observed SEE values are comparable with those reported in previous studies which developed FFM prediction equations for populations including the elderly (>60 yrs) by the use of bioelectrical impedance and anthropometric parameters as independent variables, 1.7 - 3.4 kg [19-24]. In addition, the SEE obtained from each equation was found (Figure 3). Thus, the current results indicate that B-mode MT measurements are applicable for predicting FFM in population studies for the elderly, with a similar accuracy to that observed in prediction equations based on bioelectrical impedance and anthropometric measurements. In the viewpoint of convenience of measurement, nevertheless, bioelectrical impedance and anthropometric measurement have better advantages than the ultrasound measurement. However, a recent study has demonstrated that the prevalence rate of sarcopenia detected by total skeletal muscle mass differed from that of site-specific thigh sarcopenia (25). As described in earlier, the influence of aging on skeletal muscle size has site-related difference (5). In both men and women, the age-related loss has been shown to be greater in the abdomen and thigh anterior than in other parts 6, 26. Bioelectrial impedance and anthropometric methods can not quantify the size of a specific muscle group. Considering these aspects, we may say that ultrasound MT measurement is feasible for estimating total body fat free mass with consideration for the site-related difference in the magnitude of decline in muscle mass with aging.

To the best of our knowledge, available information on the prediction of FFM by using MTs as an independent variable is limited to the report of Abe at al. (14). They used young and middle-aged men and women as subjects, and developed predicting equations with the MT values at the abdomen, lower leg posterior, and thigh anterior for men and those at the abdomen, thigh anterior, and upper arm anterior for women as significant contributors for predicting FFM. In the current study, sex was adopted as a significant predictor. In addition, there was a difference in the procedure used for determining FFM; DXA in the current study and hydrodensitometry in the previous study. With regard to the sites at which MT values were significant predictors, therefore, we cannot directly compare between the current and previous studies. In both Eq1 and Eq2, however, it should be noted that either MT or MT×LL for the lower extremity were mainly selected. In addition, MT values at the four sites in Eq1 and MT×LL values at three sites of the lower extremity in Eq2 contributed 57.6% and 61.0%, respectively, to estimate the DXA-based FFM. This indicates that, for the elderly, MT measurements at the lower extremity play an important role in developing prediction equations of FFM. Janssen et al. (15) have shown that, regardless of sex, age-related loss of total muscle mass is explained by a decrease in lower body muscle mass occurring after the fifth decade. Taking these findings into account, it seems that, as compared to a younger population, the magnitude of total body FFM in the elderly would depend more on that of the lower body, and consequently, the MT values at the corresponding limbs might have been selected as significant predictors for the DXA-based FFM.

In both Eq1 and Eq2, the difference between DXA-based FFM and estimated FFM was not associated with the mean values of the two methods. However, its absolute value was significantly greater in Eq1 than in Eq2. This may be due to the fact that the MT determined here was obtained at a specific site in a body segment and was not a measure of the total mass of the muscle groups located in the corresponding segment. In the present study, the sites for measuring MTs were the same as in previous studies (4, 5, 6, 13, 14). The selection among the upper and lower extremities was based on a finding that the maximal muscle cross-sectional area is obtained at the corresponding level as a result of analysis of only two cadavers (27). In addition, the measurement sites for the abdomen and subscapular are those adopted generally for determining skinfold subcutaneous fat thickness. Thus, whether the MT at the prescribed site is representative of the total muscle mass of the corresponding segment is uncertain. However, adding segment length or body height to the MT improves the analysis as the length factor of the muscle has shown to be a strong predictor of the volume of a specific muscle group 3, 2, 12 or segmental and total body mass (13). For example, Miyatani et al. (12) reported that ultrasonographic muscle thickness measurement was a good predictor of the knee extensor muscle volume when combined with limb length. Furthermore, Sanada et al. (13) showed that, in body segments including the trunk, the products of height and MT, determined at the same sites used here, was highly correlated with the muscle mass of the segment in both sexes. These findings indicate that, as compared to MT only, the combination of MT and either limb length or height becomes a more significant predictor of segmental and total muscle mass. At the same time, it provides a reason for the lower absolute error in Eq2 compared to Eq1, in which the product of MT and limb length was used as an independent variable to predict DXA-based FFM.

As a limitation of the MT measurements, however, it should be noted that the ultrasonography used here cannot exclude non-contractile tissues such as inter-muscular adipose tissue and connective tissue within the muscle compartment. It is known that aging is associated with increasing inter-muscular adipose tissue within the muscle compartment (28, 29, 30). Therefore, we cannot rule out that the MT determination overestimates the size of pure muscle and influences the accuracy of the predicting equations. It has been shown that mid-thigh low density lean tissue is associated with age and adiposity (28-31). For example, Goodpaster et al. (29) reported that mid-thigh muscle attenuation values with computerized tomography were negatively correlated with total body fat mass determined by DXA. Considering these additional factors, it might be assumed that, if the magnitude of inter-muscular fat in the subjects influenced the accuracy of the prediction equations, the error of the estimate would be associated with either age or total body fat mass. In the current subjects, however, the difference between the DXA-based and estimated FFM values for each of Eq1 and Eq2 were not significantly associated with age: r = 0.047 (n.s.) for Eq1 and r = 0.113 (n.s.) for Eq2. Similarly, the corresponding relationships with fat mass and its percentage to total body mass were also not significant (r = -0.045 - 0.129, n.s.). These results support the assumption that, at least in individuals who are in the fifth to the seventh decade, the aforementioned influence, related to the limitation of MT measurements, on the accuracy of the prediction equations may be negligible. However, the body mass indices for the subjects examined here were less than 30. With regard to this concern, therefore, further investigation using individuals who are classified as obese is needed.

Furthermore, our study had two limitations in relation to the experimental design. First, we did not set a cross-validation group. Therefore, whether the prediction equations developed here can be applied to other populations remain questionable. The findings obtained here simply indicate the applicability of using B-mode ultrasound MT measurements to predict DXA-based FFM in elderly individuals. The second limitation is the accuracy of the DXA-based FFM. As described earlier, it has been shown that body composition variables determined by DXA differ from those with criterion methods. This discrepancy may be due to the difference in DXA-derived variables of body composition in relation to hydration, software version, hardware, and subject population (32). Lhoman et al. (32) reviewed articles published in the latter half of the 1990’s and indicated that, for both the Lunar and Hologic systems of DXA, agreement is found between DXA-based and the multicomponent approaches for percent body fat. In contrast, Schoeller et al. (33), who examined the validity of DXA (QDR 4500A) measurements against criterion methods in a large heterogeneous population, showed that DXA overestimated FFM by 5%. This finding clearly indicates a need for calibration of DXA-based FFM through comparisons with criterion methods. In the results of Schoeller et al. (33), however, a strong linear relationship between the two methods was found. In addition, Tylavsky et al. (34) and Visser et al. (16) reported that the difference between FFM values derived from DXA and criterion methods was not associated with the mean of the two methods. These findings support the assumption that, even if the DXA-based FFM values obtained here differ from those derived with criterion methods, it does not negate the present result that the ultrasound MT measurements are applicable to predict total body FFM in elderly individuals.

In conclusion, the current results indicate that ultrasound muscle thickness measurement is useful to predict fat free mass in the elderly, and its accuracy is improved by using the product of muscle thickness and limb length as an independent variable.