one decimal place in the range 10 to 2000 Ω and the accuracy to within 1 % .... for the experienced observer was 0.08% for muscle and. 0.06% for AT, and that of ...
Clinical Science (1999) 96, 647ñ657 (Printed in Great Britain)
Predicting composition of leg sections with anthropometry and bioelectrical impedance analysis, using magnetic resonance imaging as reference N. J. FULLER, C. R. HARDINGHAM*, M. GRAVES*, N. SCREATON*, A. K. DIXON*, L. C. WARD† and M. ELIA MRC Dunn Clinical Nutrition Centre, Hills Road, Cambridge CB2 2DH, U.K., *Department of Radiology, Addenbrooke’s Hospital, Cambridge CB2 2QQ, U.K., and †Department of Biochemistry, University of Queensland, Brisbane, Queensland 4072, Australia
A
B
S
T
R
A
C
T
Magnetic resonance imaging (MRI) was used to evaluate and compare with anthropometry a fundamental bioelectrical impedance analysis (BIA) method for predicting muscle and adipose tissue composition in the lower limb. Healthy volunteers (eight men and eight women), aged 41 to 62 years, with mean (S.D.) body mass indices of 28.6 (5.4) kg/m2 and 25.1 (5.4) kg/m2 respectively, were subjected to MRI leg scans, from which 20-cm sections of thigh and 10-cm sections of lower leg (calf) were analysed for muscle and adipose tissue content, using specifically developed software. Muscle and adipose tissue were also predicted from anthropometric measurements of circumferences and skinfold thicknesses, and by use of fundamental BIA equations involving section impedance at 50 kHz and tissue-specific resistivities. Anthropometric assessments of circumferences, cross-sectional areas and volumes for total constituent tissues matched closely MRI estimates. Muscle volume was substantially overestimated (bias : thigh, k40 % ; calf, k18 %) and adipose tissue underestimated (bias : thigh, 43 % ; calf, 8 %) by anthropometry, in contrast to generally better predictions by the fundamental BIA approach for muscle (bias : thigh, k12 % ; calf, 5 %) and adipose tissue (bias : thigh, 17 % ; calf, k28 %). However, both methods demonstrated considerable individual variability (95 % limits of agreement 20–77 %). In general, there was similar reproducibility for anthropometric and fundamental BIA methods in the thigh (inter-observer residual coefficient of variation for muscle 3.5 % versus 3.8 %), but the latter was better in the calf (inter-observer residual coefficient of variation for muscle 8.2 % versus 4.5 %). This study suggests that the fundamental BIA method has advantages over anthropometry for measuring lower limb tissue composition in healthy individuals.
INTRODUCTION The ability to measure skeletal muscle mass or volume accurately and precisely is potentially of major benefit to a number of biomedical disciplines ranging from fun-
damental research to the monitoring and clinical management of muscle wasting (for a comprehensive review see [1]). Such measurements have obvious relevance for growth and development, ageing, sports and exercise and nutritional interventions [2–6]. However, there is a
Key words : adipose tissue, muscle. Abbreviations : AT, adipose tissue ; BIA, bioelectrical impedance analysis ; BMI, body mass index ; MRI, magnetic resonance imaging ; SFT, skinfold thickness. Correspondence : Dr N. J. Fuller.
# 1999 The Biochemical Society and the Medical Research Society
647
648
N. J. Fuller and others
surprising paucity of appropriate reference data concerning muscle mass, which appears to be limited to a few studies of cadavers. In the largest of these, just 25 male cadavers, aged 50 to 94 years, were subjected to anthropometric measurements for prediction of muscle mass as determined by anatomical dissection [7]. However, such data may not be truly representative nor has it been used for definitive validation of other indirect methods which are subject to further limitations [8]. Although relatively inexpensive, anthropometry is based on uncertain assumptions concerning adipose tissue (AT) distribution along the length and circumference of the limbs and mean cross-sectional areas of bone (e.g. [9]). Scanning techniques such as magnetic resonance imaging (MRI) and computed tomography, which have been used to assess muscle mass [3,5,6,10], are expensive, may not be readily accessible and analysis of results is labour intensive [8]. A simple alternative approach to the prediction of limb muscle mass is that of a novel fundamental bioelectrical impedance analysis (BIA) method based on measured impedance and assumed specific resistivities of all constituent tissues [11]. Until now, application of this technique has been limited to the measurement of crosssectional areas of muscle and AT in sections of upper arm [11]. Furthermore, no attempt has yet been made to (a) assess the reproducibility of the fundamental BIA method, (b) compare the predictive power of this method relative to that of anthropometry, or (c) consider the influence of skin on the results obtained. In addition, the fundamental BIA method has not been applied to the estimation of muscle or AT in sections of thigh or lower leg (essentially calf muscle), where muscle orientation and the overall shape of the limb segments are different from the upper arm. Therefore, the purpose of this study was to evaluate the validity and reproducibility of muscle and AT volume in thigh and lower leg sections predicted using this fundamental BIA method, compared with classic anthropometry, against estimates obtained by the established reference method, MRI, which has improved software based on region growing with adaptive thresholding that enables more sophisticated visualization of tissues in cross-sectional slices.
METHODS Characteristics of the 16 healthy subjects (eight men and eight women) who volunteered for the study were as follows. Men : age range 43–62 (median 50) years, mean weight 89.6 (S.D. 17.1) kg, height 1.77 (S.D. 0.04) m and body mass index (BMI) 28.6 (S.D. 5.4) kg\m# ; women : age range 41–60 (median 48) years, mean weight 70.0 (S.D. 16.2) kg, height 1.67 (S.D. 0.06) m and BMI 25.1 (S.D. 5.4) kg\m#. For consistency, each subject fasted # 1999 The Biochemical Society and the Medical Research Society
overnight, all measurements were started in the early morning and all were performed with the subject in the supine position ; this position was adopted for about 10 min before performing each technique in order to standardize for the effects of possible fluid shifts. Body weight (kg), measured using a Sauter Type E1210 electronic scales (Todd Scales, Studlands Park Industrial Estate, Newmarket, Suffolk, U.K.), and standing height (m), measured against a wall-mounted stadiometer, were used to determine BMI (kg\m#).
MRI Contiguous 10-mm axial images of both legs were acquired with a body coil in a 0.5 T magnetic resonance imaging system (Signa, GE Medical Systems, Milwaukee, WI, U.S.A.). A dual-echo fast-spin echo sequence was performed, with parameters as follows : ETL l 8 TE\TR l 17, 102\3000 FOV l 48i36 cm matrix l 256i192i2NEX superior and inferior presaturation pulses. Images were rescaled to a maximum signal value of 255 grey levels. Image segmentation was based on adaptive thresholding and region growing with thresholds calculated from the image grey level histogram. The analysis was implemented using a purposedesigned computer program in Microsoft Visual Cjj for Windows 95. For each image, the grey level histogram was calculated and smoothed using a local second-order polynomial, which utilized the method of the sum of least squares. The background peak was set at grey level zero. After removal of the background peak, the two highest maxima were taken to be the muscle and AT mode grey levels. Three thresholds were then calculated, midway between the three pairs of peaks. There was no opportunity for modification by the operator. For each image, the grey level histogram was displayed with the peaks marked. The user was also provided with two windows representing the current image, one showing the grey level image and the other showing the segmentation image, i.e. the regions which had already been segmented labelled by a simple colour scheme. The operator defined and labelled the regions interactively, marking seed points with the computer mouse. The program stored the segmentation images to enable retrieval if required. Once the image had been segmented, the user was able to produce a text file containing the number of pixels in each slice allocated to each label, along with pixel dimensions. Muscle and AT volumes were assessed, in the left leg, for a section of thigh, the limits of which extended 10 cm either side of a point located two-fifths of the distance between knee joint space and anterior superior iliac spine, and for a section of lower leg that extended 5 cm either side of a point located two-fifths of the distance between knee joint space and ankle joint space. To ensure that exactly the same upper and lower section limits were
Assessing muscle mass in the lower limb
located reproducibly for each method, absolute distances from the knee joint space were defined from anthropometric measurements and used consistently. The following levels were designated : thigh A, the point or slice located at the lower extremity of the section, nearest the knee joint space and 10 cm from thigh B ; thigh B, twofifths of the distance from knee joint space to anterior superior iliac spine ; thigh C, located at the upper extremity of the section, nearest the anterior superior iliac spine and 10 cm from thigh B ; calf A, the point or slice located at the upper extremity of the section, nearest the knee joint space and 5 cm from calf B ; calf B, twofifths of the distance from knee joint space to ankle joint space ; calf C, located at the lower extremity of the section, nearest the ankle joint and 5 cm from calf B. Although 1-cm contiguous slices were available for analysis, sufficiently accurate results were obtained by extrapolation between cross-sectional tissue areas of 1cm MRI slices at 5-cm intervals because these were found to be virtually identical (mean difference 0.4 % for muscle and 0.9 % for AT). The volume of tissue in between these slices was calculated by extrapolation, assuming a uniformly tapering truncated cone between slices. Therefore, tissue volume of the thigh section was calculated from five slices and the calf section from three slices. The sizes and positions of these sections were determined to try to encompass mostly muscle and AT over bone and to avoid other tissues such as tendon or joints that might confound the accuracy of measurements.
Anthropometry Circumferences and skinfold thicknesses (SFTs) of limb sections were measured using a standard tape measure and skinfold callipers (Holtain Ltd, Dyfed, Wales) at the mid-point and extremities of each section in order to reproduce the same points as identified for the MRI assessments (see above). Due to difficulties associated with obtaining consistent measurements of SFT between individuals around the perimeter of the sections, these were limited to SFT measurements in the mid-line at the anterior and posterior aspects of the thigh and the posterior aspect of the calf. Limb cross-sectional areas for all tissues were estimated from circumferences (cm), assuming circularity : Limb cross-sectional area (cm#) l circumference#\4π Measurements of circumference and SFT (cm) were used to estimate AT cross-sectional area (subcutaneous AT only) : AT cross-sectional area (cm#) l circumferenceiSFT\2 Muscle cross-sectional areas, also assumed to be circular, were estimated by the difference between crosssectional areas of the whole limb and AT with an assumed
cross-section of bone with its constituent marrow (6 cm# ; this study and [10,12]) : Muscle cross-sectional area (cm#) l (circumference#\4π)k(circumferenceiSFT\2)k6. Volumes of each section and constituent muscle and AT were calculated assuming a uniformly tapering truncated cone between measurement positions (i.e. the product of mean cross-sectional area and length).
Fundamental BIA Impedance measurements at 50 kHz were obtained from a multi-frequency BIA instrument (Model SFB2, SEAC, Brisbane, Australia), with current-injecting electrodes placed on the base of the fingers and toes (to minimize any adverse effects of inconsistent current distribution [13]) and the centres of receiving electrodes placed in the mid-line at the limits of the sections, measured from the knee joint space exactly as defined for the MRI analysis, on the anterior aspect of the thigh and the posterior aspect of the calf. In order to estimate muscle mass and AT mass within the sections from section length and impedance over that length, a fundamental approach was adopted which took into account specific resistivities at 50 kHz of all constituent tissues. This procedure actually enabled calculation of tissue cross-sectional area, at the mid-point of the section, and volume was calculated using the length of section assuming a uniformly tapering truncated cone. Fundamental BIA equations were derived assuming that tissue resistances are arranged in a parallel configuration and applying the following : ρm l resistivity of human muscle at 50 kHz (1.49 Ω:m [14]) ρat l resistivity of AT at 50 kHz (16 Ω:m [11]) ρn l resistivity of neurovascular tissue at 50 kHz (1.6 Ω:m [11]) ρb l resistivity of bone at 50 kHz ( 100 Ω:m [11]) ρs l resistivity of skin at 50 kHz (5.5 Ω:m ; estimated in this study from the electrical properties of a variety of tissues at different frequencies, e.g. 2.89 Ω:m at 1 MHz [15] or even higher, with these values taken from the comprehensive report of Duck [16]) A l cross-sectional area (cm#) of the whole limb section at the mid-point, either at thigh B or at calf B (calculated from circumference assuming circularity : circumference#\4π). Am l muscle cross-sectional area (cm#) at the same section mid-point Aat l AT cross-sectional area (cm#) An l neurovascular cross-sectional area (cm#), assumed to be 1 % of total cross-sectional area Ab l bone cross-sectional area (cm#), assumed to be 6 cm# As l skin cross-sectional area (cm#) # 1999 The Biochemical Society and the Medical Research Society
649
650
N. J. Fuller and others
L l section length (cmi100) R l impedance at 50 kHz (Ω)
and skin thickness, which was assumed to be 0.16 cm for men and 0.12 cm for women [17]. Volumes of muscle and AT in thigh and calf sections were obtained simply by the product of cross-sectional areas (obtained above) and length, assuming uniformly tapering truncated cones, so that in the equations above area is replaced by volume and L by L# : Full equation (1) :
L Am Aat An As Ab l j j j j R ρm ρat ρn ρs ρb and A l AmjAatjAnjAsjAb so Aat l AkAmkAnkAskAb Aat A Am An As Ab l k k k k ρat ρat ρat ρat ρat ρat
and
substituting in the basic equation for
Vm l
0#
Vm l
Am Am L A An Ab As An Ab As k l k j j j k k k ρm ρat R ρat ρat ρat ρat ρn ρb ρs to give Equation 1 (full equation) :
10
L A An Ab As An Ab As k j j j k k k R ρat ρat ρat ρat ρn ρb ρs
ρm:ρat ρatkρm
1
or Am l 1.643
0
1
L A 0.01A 6.0 A 0.01A 6.0 As i k j j j sk k k R 16.0 16.0 16.0 16.0 1.6 100 5.5
Because the contribution to these equations of the amounts of skin, bone and neurovascular tissue (numerators in above equations) is relatively small compared with the high values of specific resistivities for AT, bone and skin (denominators), simplified versions were also applied : Equation 2 (simplified equation) : Am l
0
L A k R ρat
10
1
0
1
ρm:ρat ρatkρm
or Am l 1.643i
1
0 # 10 L V k R ρat
ρm:ρat ρatkρm
1
Vat l 0.99VkVmkVsk6L
and rearranging
0
ρm:ρat ρatkρm
Simplified equation (2) :
Aat ρat
L A A A A A A A A A l j k mk nk bk s j nj bj s R ρm ρat ρat ρat ρat ρat ρn ρb ρs
Am l
10
L V V V V V V V k j n j b j s k nk bk s R ρat ρat ρat ρat ρn ρb ρs
L A k R 16.0
If it is assumed that bone accounts for a constant crosssectional area (about 6 cm# ; this study and [10,12]) and that neurovascular tissue contributes about 1 % of the total cross-sectional area (A) or less (this study) then Aat l 0.99AkAmkAsk6 where Am may be calculated using either the full equation (equation 1) or simplified version (equation 2). Area of skin (As) was obtained from the product of circumference # 1999 The Biochemical Society and the Medical Research Society
As above, Vm may be calculated using either the full equation (equation 1) or the simplified version (equation 2). Mean cross-sectional area was also estimated from anthropometric measures at A, B and C in order to assess if this might provide better estimates of MRI muscle volume rather than by calculating area from a single circumference taken at the mid-point, ‘ B ’.
Precision A subset of four volunteers (two men and two women, BMI range 21–30 kg\m#) was subjected to repeated measurements by two trained observers for MRI scans (one experienced and one less so) and four trained observers (whose range of experience varied from being recently instructed to some years of practice) for anthropometry (at points thigh B and calf B as above) and impedance (over the section lengths) in thigh and calf sections. Tissue composition was assessed using anthropometric measurements alone and by the fundamental BIA approach which included anthropometric circumference measures.
Statistics The strength of relationships between variables was assessed using Pearson’s correlation coefficient (r). The bias and 95 % limits of agreement between the MRI reference method and the alternative prediction techniques (reference minus prediction), and relationships between the size of estimate and difference between methods, were scrutinized [18]. Intra- and inter-observer variability was established, both for the raw measurements and estimates of body composition derived from them, using ANOVA performed on the natural log-transformed data within the
Assessing muscle mass in the lower limb
Data Desk2 statistical program in a similar fashion to that of Fuller et al. [19]. This procedure enables calculation of the residual coefficient of variation, considered to be essentially the same as residual lnS.D.i100, or percentage variability. Where there was no effect of repeated measures (i.e. systematic bias between the first and second measure performed by each observer), the analysis was performed again but without replicate measures.
Ethics approval This study was approved by the local Ethics Committees of Addenbrooke’s Hospital and the Dunn Clinical Nutrition Centre, Cambridge, and informed written consent was obtained from each subject.
RESULTS MRI Table 1 shows mean (S.D.) values obtained using MRI for circumference and cross-sectional areas of total tissue and individual tissues in thigh and calf slices A, B and C, and for estimates of tissue volumes for the sections in between and including these slices. If circularity was assumed then total cross-sectional area calculated from circumference (thigh slice A, 162 cm# ; slice B, 238 cm# ; slice C, 312 cm# : calf slice A, 128 cm# ; slice B, 119 cm# ; slice C, 89 cm#) would be overestimated in all cases compared with summed MRI estimates of component tissues (Table 1).
Table 1
Anthropometric and BIA prediction methods Table 2 presents a comparison of anthropometric estimates of circumference, cross-sectional areas and volumes for total tissues in thigh and calf sections compared with equivalent values derived from MRI. There are only slight discrepancies between MRI and anthropometry for most measures, except the circumference at calf A (anthropometry about 10 % lower) and calf cross-sectional areas (anthropometry, generally about 10 % higher). Table 3 shows how anthropometric estimates of muscle and AT compared with MRI reference measures ; muscle assessed by anthropometry is overestimated in both thigh (substantially) and calf, whereas AT is underestimated. This error is related to the proportion of AT (MRI estimate) in the thigh (r l 0.54, P 0.02) but not in the calf. At neither site was the error related to (absolute) cross-sectional area of AT. The mean (S.D.) impedance at 50 kHz for the 20-cm thigh section was 22.8 (5.3) Ω and that for the 10-cm calf section was 25.2 (7.0) Ω (the SFB2 instrument reads to one decimal place in the range 10 to 2000 Ω and the accuracy to within 1 %, claimed by the manufacturer, has been verified experimentally [20]). Table 4 shows that there was a small mean overestimation of thigh muscle and underestimation of AT, both for cross-sectional area and volume, by the fundamental BIA method compared with the MRI reference measures, whereas mean calf muscle values were slightly underestimated and AT
MRI measurements of thigh and calf sections
Values are means (S.D.). For definition of locations of slices and section, see text. CSA, cross-sectional area. Calf bone l tibia and fibula. Thigh section (20 cm) Slice at A Circumference (cm) Total tissue CSA (cm2) Muscle CSA (cm2) AT CSA (cm2) Bone CSA (cm2) Bone marrow CSA (cm2) Neurovascular CSA (cm2)
Total tissue volume (cm3) Muscle volume (cm3) AT volume (cm3) Bone volume (cm3) Bone marrow volume (cm3) Neurovascular volume (cm3)
44.8 (5.0) 140.5 (32.7) 68.5 (19.8) 63.7 (25.9) 2.1 (1.1) 5.6 (1.3) 0.7 (0.6)
Calf section (10 cm) Slice at B 54.5 (5.3) 206.2 (41.1) 123.5 (30.6) 76.8 (38.8) 4.0 (0.7) 1.8 (0.3) 0.2 (0.2)
Slice at C 62.3 (6.3) 268.5 (52.2) 135.7 (36.1) 126.7 (53.3) 4.5 (0.6) 1.5 (0.4) 0.2 (0.2)
Slice at A
Slice at B
Slice at C
39.9 (3.9) 106.4 (20.6) 71.7 (17.8) 28.5 (13.6) 3.0 (0.8) 3.1 (0.7) 0.1 (0.1)
38.4 (4.2) 94.8 (19.4) 62.7 (15.1) 26.5 (12.0) 3.7 (0.8) 1.8 (0.6) 0.1 (0.1)
33.2 (3.5) 71.5 (15.2) 43.4 (10.7) 23.2 (11.4) 3.4 (0.8) 1.4 (0.4) 0.2 (0.1)
Volume AñB
Volume BñC
Volume AñC
Volume AñB
Volume BñC
Volume AñC
1723.4 (378.7) 962.5 (247.4) 691.9 (322.2) 31.5 (7.1) 33.6 (7.3) 3.7 (2.3)
2387.0 (462.5) 1338.6 (347.6) 989.3 (469.4) 42.7 (6.1) 14.6 (3.2) 1.9 (1.7)
4110.4 (835.5) 2301.1 (586.3) 1681.2 (788.8) 74.3 (12.6) 48.3 (9.2) 5.6 (3.7)
503.2 (99.4) 336.1 (79.9) 137.3 (63.0) 16.8 (4.0) 12.4 (3.0) 0.6 (0.5)
415.9 (85.4) 265.3 (63.7) 124.2 (57.9) 17.7 (4.0) 8.0 (2.1) 0.7 (0.5)
919.0 (183.7) 601.4 (142.3) 261.5 (120.4) 35.5 (7.8) 20.4 (4.9) 1.2 (1.0)
# 1999 The Biochemical Society and the Medical Research Society
651
652
N. J. Fuller and others
Table 2 Comparison between MRI and anthropometric measurements of limb section circumferences, cross-sectional areas and volumes in the thigh and calf
Thigh Circumference (cm) at A at B at C Mean Cross-sectional area (cm2) at A at B at C Mean Volume (l) A to B B to C A to BjB to C A to C direct Calf Circumference (cm) at A at B at C Mean Cross-sectional area (cm2) at A at B at C Mean Volume (l) A to B B to C A to BjB to C A to C direct
MRI Mean (S.D.)
Anthropometry Mean (S.D.)
44.8 (5.0) 54.5 (5.3) 62.3 (6.3) 53.9 (5.4)
43.9 (5.4) 53.2 (5.7) 60.3 (6.1) 52.4 (5.6)
140.5 (32.7) 206.2 (41.1) 268.5 (52.2) 205.1 (41.3)
145.8 (25.0) 214.7 (35.5) 282.9 (55.0) 214.5 (36.4)
1.72 (0.38) 2.39 (0.46) 4.11 (0.84) 4.11 (0.84)
1.80 (0.29) 2.49 (0.43) 4.29 (0.72) 4.29 (0.76)
39.9 (3.9) 38.4 (4.2) 33.2 (3.5) 37.2 (3.5)
36.8 (3.5) 36.4 (3.6) 31.8 (3.1) 35.0 (3.3)
106.4 (20.6) 94.8 (19.4) 71.5 (15.2) 90.9 (18.1)
108.5 (20.7) 106.3 (20.6) 81.0 (15.9) 98.6 (18.5)
0.50 (0.10) 0.42 (0.09) 0.91 (0.18) 0.91 (0.18)
values overestimated. These discrepancies were not related to the size of measurement or proportion of AT in these limb sections. There was no significant difference in correlation coefficients obtained between MRI and anthropometric estimates of muscle mass compared with those between MRI and impedance index (L#\impedance) for thigh (r l 0.59 and 0.68 respectively) or calf (r l 0.83 and 0.90 respectively). Table 5 shows that the fundamental BIA equations predict MRI estimates of thigh and calf sections better than anthropometry. Table 6 shows that a single circumference measurement at point B does not accurately reflect the mean circumference (taken at A, B and C). Therefore, these limb sections cannot be assumed to be uniformly tapering truncated cones. The simplified BIA equation (equation 2) provided # 1999 The Biochemical Society and the Medical Research Society
0.54 (0.10) 0.47 (0.09) 1.00 (0.19) 0.95 (0.18)
Bias
95 % Limits of agreement
1.0 1.3 2.0 1.4 k5.2 k8.5 k14.4 k9.4 k0.08 k0.10 k0.18 k0.17
3.2 2.0 1.5 2.2 k2.1 k11.4 k9.5 k7.7 k0.03 k0.05 k0.09 k0.03
k2.1 to 4.1 k3.5 to 6.1 k2.3 to 6.3 k2.2 to 5.1 k46.5 to 36.1 k67.0 to 49.9 k51.1 to 22.3 k50.0 to 31.3 k0.59 to 0.44 k0.52 to 0.32 k1.09 to 0.73 k0.85 to 0.50
0.7 to 5.6 k2.8 to 6.8 k3.7 to 6.6 k0.5 to 4.9 k16.6 to 12.5 k29.5 to 6.7 k26.4 to 7.4 k21.8 to 6.5 k0.11 to 0.04 k0.13 to 0.03 k0.24 to 0.06 k0.17 to 0.10
estimates of limb section muscle [thigh 2.59 (S.D. 0.65) litres ; calf 0.57 (S.D. 0.13) litres] that were only about 2 % higher than the full BIA equation (equation 1), and the estimates of AT [thigh 1.41 (S.D. 0.61) litres ; calf 0.36 (S.D. 0.15) litres] were less than 5 % lower. Substituting into the full BIA equation alternative values for skin resistivity or the use of the mean cross-sectional area at A, B and C instead of that at the mid-point (B) affected the results to the same extent or less (2 % for muscle and 5 % for AT) than the simplified equation.
Reproducibility The relative (percentage) variability for repeated MRI analysis was practically trivial. Intra-observer variability for the experienced observer was 0.08 % for muscle and 0.06 % for AT, and that of the less experienced observer 0.63 % and 0.35 % respectively. Inter-observer variability
Assessing muscle mass in the lower limb
Table 3 Comparison between MRI and anthropometric measurements of muscle and AT crosssectional areas and volumes in thigh and calf sections
Thigh Muscle cross-sectional area (cm2) at A at B at C Mean Muscle volume (l) A to B B to C A to BjB to C A to C direct AT cross-sectional area (cm2) at A at B at C AT volume (l) A to B B to C A to BjB to C A to C direct Calf Muscle cross-sectional area (cm2) at A at B at C Mean Muscle volume (l) A to B B to C A to BjB to C A to C direct AT cross-sectional area (cm2) at A at B at C AT volume (l) A to B B to C A to BjB to C A to C direct
MRI Mean (S.D.)
Anthropometry Mean (S.D.)
Bias
95 % Limits of agreement
68.5 (19.8) 123.5 (30.6) 135.7 (36.1) 109.2 (27.9)
102.8 (24.1) 160.0 (31.8) 220.4 (48.4) 161.1 (31.7)
k34.4 k36.6 k84.8 k51.9
k68.2 to k0.48 k91.9 to 18.8 k158.5 to k11.0 k98.6 to k5.2
0.96 (0.25) 1.34 (0.35) 2.30 (0.59) 2.30 (0.59) 63.7 (25.9) 76.8 (38.8) 126.7 (53.3) 0.69 (0.32) 0.99 (0.47) 1.68 (0.79) 1.68 (0.79)
71.7 (17.9) 62.7 (15.1) 43.4 (10.7) 59.3 (14.0) 0.34 (0.08) 0.27 (0.06) 0.60 (0.14) 0.60 (0.14) 28.5 (13.6) 26.5 (12.0) 23.2 (11.4) 0.14 (0.06) 0.12 (0.06) 0.26 (0.12) 0.26 (0.12)
was 0.11 % for muscle and 0.26 % for AT. The relative (percentage) variability for repeated anthropometric or BIA measures by the same observer was less than that between observers for all basic measurements and estimates of section tissue composition derived from them (Table 7). Limb circumference was found to be the least variable of the basic measures, translating into small variability in whole-section cross-sectional areas. The relatively large variability of SFT measurements had little
1.31 (0.26) 1.90 (0.38) 3.22 (0.63) 3.23 (0.66) 36.9 (19.1) 48.6 (21.0) 56.4 (24.5) 0.43 (0.20) 0.52 (0.22) 0.95 (0.42) 0.93 (0.42)
77.4 (19.7) 75.7 (22.6) 54.5 (16.9) 69.2 (19.2) 0.38 (0.10) 0.33 (0.10) 0.71 (0.20) 0.66 (0.18) 25.1 (11.7) 24.6 (10.9) 20.5 (7.8) 0.12 (0.06) 0.11 (0.04) 0.24 (0.10) 0.23 (0.09)
k0.35 k0.56 k0.92 k0.93
k0.77 to 0.07 k1.16 to 0.04 k1.89 to 0.06 k1.89 to 0.03
26.8 28.2 70.4
k15.9 to 69.5 k28.6 to 85.0 k6.8 to 147.5
0.26 0.46 0.73 0.75
k5.6 k12.9 k11.1 k9.9 k0.05 k0.06 k0.11 k0.06 3.3 1.9 2.7 0.01 0.01 0.02 0.03
k0.22 to 0.75 k0.22 to 1.15 k0.43 to 1.89 k0.43 to 1.93
k23.1 to 11.8 k35.9 to 10.0 k29.6 to 7.3 k25.5 to 5.7 k0.13 to 0.04 k0.15 to 0.03 k0.28 to 0.06 k0.21 to 0.09 k20.0 to 26.7 k18.4 to 22.2 k13.9 to 19.3 k0.09 to 0.12 k0.07 to 0.10 k0.16 to 0.21 k0.15 to 0.21
material effect on estimates of section muscle and AT when combined with circumferences because of the greater influence of the latter measure. Variation in the basic impedance measurements was found to be intermediate between that for SFTs and circumference. This translated into similar intra- and inter-observer variability in thigh muscle and AT but generally less variability in calf composition than that observed using anthropometry alone. # 1999 The Biochemical Society and the Medical Research Society
653
654
N. J. Fuller and others
Table 4 Comparison between MRI and fundamental BIA measurements of muscle and AT cross-sectional areas and volumes in thigh and calf sections
Thigh Muscle CSA at B (cm2) AT CSA at B (cm2) Muscle volume (l) AT volume (l) Calf Muscle CSA at B (cm2) AT CSA at B (cm2) Muscle volume (l) AT volume (l)
MRI Mean (S.D.)
BIA Mean (S.D.)
Bias
95 % Limits of agreement
123.5 (30.6) 76.8 (38.8) 2.30 (0.59) 1.68 (0.79)
128.9 (32.6) 71.4 (30.6) 2.58 (0.65) 1.43 (0.61)
k5.4 5.4 k0.28 0.25
k62.7 to 51.9 k58.4 to 69.3 k1.39 to 0.84 k1.04 to 1.55
62.7 (15.1) 26.5 (12.0) 0.60 (0.14) 0.26 (0.12)
57.1 (13.1) 36.2 (14.9) 0.57 (0.13) 0.36 (0.15)
5.7 k9.7 0.03 k0.10
k8.8 to 20.2 k28.1 to 8.6 k0.09 to 0.15 k0.29 to 0.08
Table 5 Summary of comparisons between MRI estimates and alternative predictions (anthropometry and fundamental BIA) of muscle and AT volume in thigh and calf sections
Anthropometric prediction
Thigh Muscle volume (l) AT volume (l) Calf Muscle volume (l) AT volume (l)
Fundamental BIA prediction
MRI estimate Mean (S.D.)
Mean (S.D.)
Bias
95 % Limits of agreement
Mean (S.D.)
Bias
95 % Limits of agreement
2.30 (0.59) 1.68 (0.79)
3.22 (0.63) 0.95 (0.42)
k0.92 0.73
k1.89 to 0.06 k0.43 to 1.89
2.58 (0.65) 1.43 (0.61)
k0.28 0.25
k1.39 to 0.84 k1.04 to 1.55
0.60 (0.14) 0.26 (0.12)
0.71 (0.20) 0.24 (0.01)
k0.11 0.02
k0.28 to 0.06 k0.16 to 0.21
0.57 (0.13) 0.36 (0.15)
0.03 k0.10
k0.09 to 0.15 k0.29 to 0.08
Table 6 Comparison of MRI estimates taken at point B with the mean estimate from points A, B and C
Mean value of A, B, and C
Thigh Circumference (cm) Cross-sectional area (cm2) Muscle CSA (cm2) AT CSA (cm2) Calf Circumference (cm) Cross-sectional area (cm2) Muscle CSA (cm2) AT CSA (cm2)
MRI estimate (S.D.) at B
Bias
95 % Limits of agreement
54.5 (5.3) 206.2 (41.1) 123.5 (30.6) 76.8 (38.8)
k0.6 k1.1 k14.3 13.2
k2.0 to 0.8 k12.0 to 9.9 k24.0 to k4.6 k0.0 to 24.5
38.4 (4.2) 94.8 (19.4) 62.7 (15.1) 26.5 (12.0)
k1.2 k3.9 k3.4 k0.5
k5.0 to 2.6 k8.9 to 1.1 k10.3 to 3.4 k4.9 to 4.0
DISCUSSION Anthropometry and MRI compared Anthropometry generally underestimated limb circum# 1999 The Biochemical Society and the Medical Research Society
ference by between 2 and 8 % (2.0–3.2 % in the thigh and 4.2–7.8 % in the lower leg). This may be due to some inconsistency in locating exactly the same measurement positions when applying the different methods, and is especially important when measuring tapering limb sections. It may also be due to compressibility of limb tissues by the tape measure (a 2-cm wide tape was used to minimize this source of error) and to the possible lack of contact between the tape and skin where the limb contour was concave. The small MRI pixel size ( 2 mm) is considered to make negligible contribution towards discrepancies in circumference measurements. Despite the generally lower measures of circumference by anthropometry, cross-sectional area calculated from it is generally greater than that obtained by MRI, further suggesting that the assumption of limb circularity is not correct. Anthropometric estimates of tissue composition in limb sections may be compromised if the following assumptions are made : that the external boundary of the muscle cross-section is circular ; that SFTs are uniform at all points around the limb circumference and that they reflect all AT in the section, including that located within the outer boundaries of muscle ; and that limb com-
Assessing muscle mass in the lower limb
Table 7 Intra- and inter-observer variation for basic anthropometric and BIA measurements and estimates of limb section composition derived from these measurements
CV, coefficient of variation ; NS, not significant. Intra-observer variability
Residual CV (%)
Significance of differences between replicates (P value)
Residual CV (%)
48.2 (1.6) 11.3 (4.0) 23.1 (3.3)
0.6 2.5 1.8
NS NS NS
1.2 27.5 3.0
34.7 (1.4) 9.5 (3.7) 24.5 (5.1)
0.5 8.2 2.1
NS NS NS
1.2 19.6 3.9
184.9 (12.3) 152.5 (7.7) 26.4 (9.5)
1.1 1.7 5.9
NS NS NS
2.3 3.5 8.6
96.0 (7.7) 74.3 (8.9) 15.7 (5.8)
1.1 3.2 8.4
NS NS NS
2.3 8.2 17.4
126.7 (22.9) 47.4 (21.8)
1.8 5.0
NS NS
3.8 8.7
NS
59.9 (12.8) 24.3 (8.0)
2.3 3.8
NS NS
4.5 17.0
NS
Mean (S.D.) Basic measurement Thigh Circumference (cm) Skinfold thickness (mm) Impedance (Ω) Calf Circumference (cm) Skinfold thickness (mm) Impedance (Ω) Estimates derived from anthropometry alone Thigh Whole cross-sectional area (cm2) Muscle cross-sectional area (cm2) AT cross-sectional area (cm2) Calf Whole cross-sectional area (cm2) Muscle cross-sectional area (cm2) AT cross-sectional area (cm2) Estimates derived using the fundamental BIA equation Thigh Muscle cross-sectional area (cm2) AT cross-sectional area (cm2) Calf Muscle cross-sectional area (cm2) AT cross-sectional area (cm2)
Inter-observer variability
position changes uniformly between two measurement points. This study shows that anthropometry substantially overestimates MRI assessments of muscle volume, especially in the thigh (mean 40 % compared with 18 % in the calf) where the extent of discrepancy is influenced by the proportion of fat present, and underestimates AT. Although our findings in the thigh support those of Knapik et al. [10], there is a greater overestimate of muscle by anthropometry in the present study (40 % versus 22 %). This may be attributable to greater fatness in the thigh muscle of our subjects and may reflect greater levels of fatness overall (mean BMIs of the middle-aged men and women in this study were 28.6 kg\m# and 25.1 kg\m# respectively, compared with 25.8 kg\m# and 22.2 kg\m# for the younger adults in the study of Knapik et al. [10]). Discrepancies arising from assumptions about bone cross-sectional area [mean values for bone with marrow in femur, 6.4 (S.D. 1.1) cm# ; and in tibia plus
Significance of differences between observers (P value)
NS 0.0001 NS 0.049 0.004 0.050
NS 0.045 0.0001 0.047 0.001 0.006
0.010
0.049
fibula, 5.6 (S.D. 1.2) cm#] are small and considered trivial in estimating the much larger muscle cross-sectional area.
BIA and MRI compared In comparison to reference MRI, the fundamental BIA method overestimated muscle volume by a mean of 11 % in the thigh and underestimated it by about 5 % in the lower leg with considerable variability between individuals (Tables 4 and 5). This may be due largely to the use of a constant value (1.49 Ω:m [14]) for the resistivity of human muscle in vivo, which has a dominant effect in the equation. Substituting the value for dog muscle (1.18 Ω:m [15] ; previously applied in a similar fashion to the upper arm by Brown et al. [11]) resulted in substantially lower estimations of muscle in the thigh (by 21 %) and calf (by 37 %) than the equivalent MRI values. Other factors that might influence the agreement between methods include (a) inexact location of sections in the # 1999 The Biochemical Society and the Medical Research Society
655
656
N. J. Fuller and others
limbs for circumference measurements and electrode placements, (b) violation of the assumption of circularity of limbs, (c) non-uniformity of cross-sectional areas of limb tissues (i.e. the proportion of the limb crosssectional area represented by particular tissues may not be consistent along the section length) and (d) uncertainty about the resistivity of skin. Extrapolation of mid-point cross-sectional areas (Table 4) to section volumes worsens muscle prediction in the thigh (from 4 % to 11 %) but improves it in the calf (from 9 % to 5 %). Although non-uniformity of tissue cross-sectional areas throughout section length has implications for anthropometric measurements (changes in anatomical distribution of tissues), it also has implications for the assumed uniform impedance of current through the sections (especially muscle). However, theoretical calculations involving impedance based on maximal limb tapering indicate a potential error of no more than 1–2 %. There is some doubt as to what might be the most appropriate value for resistivity of human skin in vivo that should be applied to the fundamental BIA equations. It is generally accepted that, compared with resistivity of muscle and AT, skin resistivity is relatively high or complex at the very least (e.g. [21,22]). Almost all published values for skin appear to be in the transverse direction across the skin. However, the resistivity of skin in the longitudinal plane is of major importance in the present study. This is because the assessment of tissues in limb sections is dependent on the assumption of parallel arrangement of tissue resistances from which the fundamental BIA equation is derived as there may be conduction of current along at least some of the skin layers. (Note that for the measurement of whole-body composition, the use of the tetra-polar arrangement of electrodes and a high-input impedance amplifier eliminates the substantial contribution of skin to wholebody impedance). In addition, the value for skin resistivity is likely to vary (anisotropy) according to location and its state of hydration (e.g. sweating). Although high resistivities have been reported for the stratum corneum [21,23], this relatively thin layer accounts for only a small fraction of total skin thickness. At microwave frequencies skin appears to have electrical properties that are intermediate between AT and muscle [24] or between AT and bone [25,26]. The value used for the fundamental equation in this study (5.5 Ω:m) was based on the electrical properties of a wide variety of tissues at different frequencies and extrapolated to provide a tentative value at 50 kHz of between 5 and 6 Ω:m, which falls somewhere in between existing values for muscle and AT or bone. However, if alternative values for skin resistivity, at the extremes of the possible range of values considered here, were incorporated into the full equation (equation 1), estimates of muscle volume would be decreased by only 1.6 %, for a value of 2.89 Ω:m [20], and increased by 1.6 % if the value for stratum corneum # 1999 The Biochemical Society and the Medical Research Society
( 100 Ω:m) was applied. Use of these equivalent values would increase estimates of AT by 2.7 % or decrease them by 2.8 %. Similarly, if terms relating to bone or neurovascular tissue areas and resistivities were to be eliminated from the full equation (equation 1), the effect on estimates of muscle volume would be less than about 1.1 %. Therefore, as the combined effect of ignoring all terms incorporating areas and resistivities of skin, bone and the neurovascular bundle is relatively small (less than 1.7 %), the simplified version (equation 2) may be used in such groups of subjects as used here without substantially compromising muscle estimates.
Comparisons among methods Although the empirical relationship between BIA (L#\ impedance) and MRI estimates of muscle is no better than that between anthropometry and MRI, the fundamental BIA method provides closer predictions of the MRI estimates of both muscle and AT volume in sections of thigh and calf (this study and [10]). Unlike anthropometry, which only assesses the difference between the whole limb section and subcutaneous AT (with a constant for bone), the fundamental BIA approach has a theoretical advantage in that it accounts for all constituent tissues of the limb section between electrodes, including AT, irrespective of whether or not it is located subcutaneously or deep within the muscle boundaries. Indeed, and in contrast to anthropometric estimates in the thigh, there is no significant relationship between the proportion of AT and the bias between MRI and the fundamental BIA prediction. Although the precision of the raw impedance measurement is better than that for SFT (Table 7), the reproducibility of muscle and AT estimates obtained using the fundamental BIA approach is generally similar to that by anthropometry in the thigh but remains better in the calf. It is uncertain as to whether or not the differences in reproducibility between methods vary according to the degree of adiposity in the limbs.
CONCLUSIONS This study has used an established reference method, MRI, to assess the relative value of anthropometry and a fundamental BIA approach for estimating muscle and AT in defined sections of thigh and lower leg (calf). Although both methods estimated the MRI values with substantial individual variability, predictions were better using the fundamental BIA method. The measurement precision for the fundamental BIA method was similar to that for anthropometry in the thigh but was better in the calf.
Assessing muscle mass in the lower limb
ACKNOWLEDGMENTS We are grateful to the subjects for their participation in the study, and to the following for their expert help and advice : Mr G. Johnson of the MRI Unit at Addenbrooke’s Hospital, Cambridge ; Dr T. J. Cole, Dr Odile Dewit, Dr Gail Goldberg, Ms Anne Nugent and Dr J. C. K. Wells of the MRC Dunn Nutrition Centre, Cambridge ; and Professor B. H. Brown of The Royal Hallamshire Hospital, Sheffield.
REFERENCES 1 Lukaski, H. C. (1996) Estimation of muscle mass. In Human Body Composition (Roche, A. F., Heymsfield, S. B. and Lohman, T. G., eds.), pp. 109–128, Human Kinetics, Champaign, IL, U.S.A. 2 Evans, W. J. and Campbell, W. W. (1993) Sarcopenia and age-related changes in body composition and functional capacity. J. Nutr. 123, 465–468 3 Eliakim, A., Burke, G. S. and Cooper, D. M. (1997) Fitness, fatness, and the effect of training assessed by magnetic resonance imaging and skinfold-thickness measurements in healthy adolescent females. Am. J. Clin. Nutr. 66, 223–231 4 Fiatarone, M. A. and Evans, W. J. (1993) The etiology and reversibility of muscle dysfunction in the aged. J. Gerontol. 48, 77–83 5 Ross, R., Pedwell, H. and Rissanen, J. (1995) Effects of energy restriction and exercise on skeletal muscle and adipose tissue in women as measured by magnetic resonance imaging. Am. J. Clin. Nutr. 61, 1179–1185 6 Welsman, J. R., Armstrong, N., Kirby, B. J., Winsley, R. J., Parsons, G. and Sharpe, P. (1997) Exercise performance and magnetic resonance imaging-determined thigh muscle volume in children. Eur. J. Appl. Physiol. 76, 92–97 7 Martin, A. D., Spenst, L. F., Drinkwater, D. T. and Clarys, J. P. (1990) Anthropometric estimation of muscle mass in men. Med. Sci. Sport Exerc. 22, 729–733 8 Lukaski, H. C. (1997) Sarcopenia : assessment of muscle mass. J. Nutr. 127, 994S–997S 9 Heymsfield, S. B., McManus, C., Smith, J., Stevens, V. and Nixon, D. W. (1982) Anthropometric measurement of muscle mass : revised equations for calculating bone-free arm muscle area. Am. J. Clin. Nutr. 36, 680–690
10 Knapik, J. J., Staab, J. S. and Harman, E. A. (1996) Validity of an anthropometric estimate of thigh muscle crosssectional area. Med. Sci. Sport. Exerc. 28, 1523–1530 11 Brown, B. H., Karatzas, T., Nakielny, R. and Clark, R. G. (1988) Determination of upper arm muscle and fat areas using electrical impedance measurements. Clin. Phys. Physiol. Meas. 9, 47–55 12 Pearson, K. and Bell, J. (1919) A study of the long bones of the English skeleton. Part 1 : The femur (chapters 1 to 4). Draper’s Company Research Memoirs, Biometric Series X, Cambridge University Press, Cambridge, U.K. 13 Organ, L. W., Bradham, G. B., Gore, D. T. and Lozier, S. L. (1994) Segmental bioelectrical impedance analysis : theory and application of a new technique. J. Appl. Physiol. 77, 98–112 14 Aaron, R., Huang, M. and Shiffman, C. A. (1997) Anisotropy of human muscle via non-invasive impedance measurements. Phys. Med. Biol. 42, 1245–1262 15 Geddes, L. A. and Baker, L. E. (1967) The specific resistance of biological material – a compendium of data for the biomedical engineer and physiologist. Med. Biol. Eng. 5, 271–293 16 Duck, F. A. (1990) Physical Properties of Tissues, Academic Press, London 17 Snyder, W. S., Cook, M. J., Nasset, E. S., Karhausen, L. R., Parry-Howells, G. and Tipton, I. H. (1975) Report of the Task Group on Reference Man, Pergamon Press, Oxford 18 Bland, J. M. and Altman, D. G. (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1, 307–310 19 Fuller, N. J., Jebb, S. A., Goldberg, G. R. et al. (1991) Inter-observer variability in the measurement of body composition. Eur. J. Clin. Nutr. 45, 43–49 20 Ward, L. C., Byrne, N., Rutter, K. et al. (1997) Reliability of multiple frequency bioelectrical impedance analysis : an inter-machine comparison. Am. J. Hum. Biol. 9, 63–72 21 Rosendahl, T. (1945) Concluding studies on the conducting properties of human skin to alternating current. Acta Physiol. Scand. 9, 39–49 22 Nyboer, J. (1970) Electrical Impedance Plethysmography, 2nd edn, C. C. Thomas Publishers, Springfield, IL, U.S.A. 23 Yamamoto, T. and Yamamoto, Y. (1976) Electrical properties of the epidermal stratum corneum. Med. Biol. Eng. Comp. 14, 151–158 24 Cook, H. F. (1951) The dielectric behaviour of some types of human tissues at microwave frequencies. Br. J. Appl. Physiol. 2, 295–300 25 England, T. S. and Sharples, N. A. (1949) Dielectric properties of the human body in the microwave region of the spectrum. Nature (London) 163, 487–488 26 England, T. S. (1950) Dielectric properties of the human body for wave-lengths in the 1–10 cm range. Nature (London) 166, 480–481
Received 5 November 1998/8 February 1999; accepted 12 March 1999
# 1999 The Biochemical Society and the Medical Research Society
657
