Appendicular skeletal muscle mass: Measurement by dual-photon absorptiometry

Abstract

Dual-photon absorptiometry (DPA) allows separation of body mass into bone mineral, fat, and fat-free soft tissue. This report evaluates the potential of DPA to isolate appendages of human subjects and to quantify extremity skeletal muscle mass (limb fat-free soft tissue). The method was evaluated in 34 healthy adults who underwent DPA study, anthropometry of the limbs, and estimation of whole-body skeletal muscle by models based on total body potassium (TBK) and nitrogen (TBN) and on fat-free body mass (FFM). DPA appendicular skeletal muscle (22.0 +/- 3.1 kg, mean +/- SD) represented 38.7% of FFM, with similar proportions in males and females. There were strong correlations (all p less than 0.001) between limb muscle mass estimated by DPA and anthropometric limb muscle areas (r = 0.82-0.92), TBK (r = 0.94), and total-body muscle mass based on TBK-FFM (r = 0.82) and TBK-TBN (r = 0.82) models. Appendicular skeletal muscle mass estimated by DPA is thus a potentially practical and accurate method of quantifying human skeletal muscle mass in vivo.

Full text

214 Am J C/in Nuir I990;52:214-8. Printed in USA. © 1990 American Society for Clinical Nutrition
Appendicular skeletal muscle mass: measurement
by dual-photon absorptiometry”2
Steven B Heymsfield, Rebecca Smith, Mary Aulet, Brooke Bensen,
Steven Lichtman, Jack Wang, andRichardNPierson, Jr
ABSTRACT Dual-photon absorptiometry (DPA) allows
separation of body mass into bone mineral, fat, and fat-free
soft tissue. This report evaluates the potential ofDPA to isolate
appendages ofhuman subjects and to quantify extremity skele-
tal muscle mass(limb fat-free soft tissue). The method was eval-
uated in 34 healthy adults who underwent DPA study, anthro-
pomety of the limbs, and estimation of whole-body skeletal
muscle by models based on total body potassium (TBK) and
nitrogen (TBN) and on fat-free body mass (FFM). DPA appen-
dicular skeletal muscle (22.0 ±3. 1 kg, I ± SD) represented
38.7% of FFM, with similar proportions in males and females.
There were strong correlations (all p<0.00 1)between limb
muscle mass estimated by DPA and anthropometric limb mus-
cle areas (r =0.82-0.92), TBK (r =0.94), and total-body mus-
cle mass based on TBK-FFM (r =0.82) and TBK-TBN (r
=0.82) models. Appendicular skeletal muscle mass estimated
by DPA is thus a potentially practical and accurate method of
quantifying human skeletal muscle mass in vivo. Am J
Clin Nutr 1990;52:2 14-8.
KEY WORDS Body composition, dual-photon absorpti-
ometry, skeletal muscle mass, neutron-activation analysis
Introduction
Skeletal muscle represents the largest fraction of fat-free
body mass. Depending on gender, age, and health status, be-
tween one-third and one-half of total body protein is within
skeletal muscle (1).
Despite the obvious significance ofskeletal muscle to physi-
ology and nutrition, methods ofquantification in vivo remain
limited. Although two metabolic end products released from
myocytes, creatinine and 3-methyihistidine, have been used to
estimate whole-body muscle mass, their application is beset
with problems (2, 3). Long urine-collection intervals, the need
for appropriate dietary intake, and concerns related to the met-
abolic origin and distribution of 3-methylhistidine and creati-
nine limit the use ofboth ofthese methods.
At present the most widely accepted methods of evaluating
skeletal muscle mass involve computerized axial tomography,
magnetic-resonance imaging, and ultrasonography performed
in multiple sections ofthe body. Although these methods repre-
sent a technological advance, expense, radiation exposure, urn-
ited instrument access, and concerns for accuracy are often
cited as limitations ofone or the other techniques (4).
The recent development of dual-photon absorptiometry
(DPA) presents a new opportunity to quantify skeletal muscle
mass in vivo. Long recognized for its effectiveness at measuring
bone density, the newly appreciated ability ofDPA to measure
fat and lean components presents an equally significant prom-
ise to the field of body composition research. Because of the
growing number of available whole-body instruments, ex-
tremely low radiation exposure, and ability to define total ap-
pendicular skeletal muscle and bone mass with high precision
(5, 6), these measurements will be widely applicable. Accord-
ingly, in this report we describe the theory behind the use of
DPA in estimating appendicular skeletal muscle mass, the cali-
bration data, and the results ofinitial patient studies.
Methods
Model
Whole-body DPA partitions body weight into two fractions,
bone ash (calcium hydroxyapatite) and soft tissue, by measur-
ing differential attenuation of photons at two energy levels.
These photons may be produced by ‘53Gd or by an x-ray
source, and the underlying principle is identical in both cases.
The DPA algorithm includes a measure of soft-tissue attenua-
tion at the two energy levels referred to as the R. The R.,.
correlates linearly with the proportion of soft tissue as fat (or
lean). The fat content ofscanned soft tissue in vivo can be esti-
mated by means of a calibration equation (6). This is accom-
plished by first scanning phantoms of known fat content and
establishing the prediction equation for percent fat based on
R5T .Chemically analyzed beefphantoms are used for this pur-
pose. Typical regression lines during calibration are r=0.96-
0.98 (6, 7). In an earlier study we described this calibration pro-
cedure and demonstrated excellent agreement between fat esti-
mated by DPA in healthy nonobese subjects and fat deter-
mined by such conventional methods as hydrodensitometry
CFrom the Department of Medicine, Obesity Research Center, St
Luke’s-Roosevelt Hospital Center, Columbia University College of
Physicians and Surgeons, New York.
2Address reprint requests to SB Heymsfield, Weight Control Unit,
4 1 1 West 1 14th Street, New York, NY 10025.
ReceivedJune 16, 1989.
Accepted for publication October 1 1, 1989.
by guest on July 10, 2011www.ajcn.orgDownloaded from

SKELETAL MUSCLE MEASUREMENT 215
(r =0.94, SEE =1 .82 kg) and neutron-activation analysis
(r =0.95, SEE =1 .68 kg) (6). Hence a given DPA scan, or any
subregion ofa scan, can be analyzed for bone ash, nonosseous
lean tissue, and fat.
The extremities consist primarily of three components-
skeleton, skeletal muscle, and fat. The skeleton or bone shaft
contains a small amount of marrow, which for simplicity may
be ignored in the evaluation of limb composition. We also as-
sume that skin and associated subcutaneous tissue are negligi-
ble in mass relative to the skeletal-muscle component. The de-
fatted and marrow-free bone is a mixture ofwater, protein, and
minerals (ash). Under usual circumstances ash represents 55%
ofwet skeletal weight, with mineral content decreasing slightly
(50-52%) in osteoporotic patients (8, 9). A reasonable assump-
tion, therefore, is that wet bone weight equals bone ash divided
by 0.55 or multiplied by 1.82.
Limb fat is estimated through use ofthe R5T and beef phan-
toms as described above. The actual fat mass in the limb is
determined as percent fat multiplied by soft-tissue mass. Skele-
tal muscle mass is then equal to total limb mass minus the sum
oflimb fat and bone mass.
Upon completion ofthe scan, the DPA software generates an
image ofthe subject’s skeleton (Fig I). Using specific anatomic
landmarks and a cursor, the DPA operator isolates the legs and
arms as shown in Figure 1.Once isolated, the system software
provides the total mass, RST, and bone ash for the identified
region. Wet bone mass (bone ash X 1.82) is next subtracted
from total limb mass, followed by subtraction offat mass calcu-
lated from the RST calibration line. This result then represents
skeletal muscle mass either separately for each limb or for the
summed upper- and lower-limb muscle masses.
Protocol
The DPA muscle-mass method was evaluated in 34 healthy
subjects who underwent DPA, anthropometry, whole-body
counting for total body potassium (TBK), and prompt ‘y-neu-
tron-activation analysis for total body nitrogen (TBN). Four
of the subjects underwent the DPA study on 4 (n =2) or 5
(n =2) consecutive days to establish the between-day coeffi-
cient of variation (CV) for estimates of limb composition. A
single trained observer (RS) read all 34 initial DPA scans and
the serial CV studies were interpreted by investigator MA. A
portion of this database is presented in two earlier unrelated
protocols (6, 10). Each volunteer signed an informed consent
before the study, which was approved by the institutional re-
view boards at St Luke’s-Roosevelt Hospital and at Brookha-
yen National Laboratory.
Dual-photon absorptiometrv. Awhole-body DPA scanner
(DP4, Lunar Radiation, Madison, WI) was used to evaluate
each patient’s total and regional bone ash, RST, and soft-tissue
mass. Each head-to-toe scan required -‘-55 mm. For calibra-
tion, seven frozen beef phantoms of known fat content were
scanned and the results were used to relate R5T to percent fat
(6). The aforementioned procedures were then used to derive
whole-body and appendicular bone ash, fat, and skeletal mus-
cle. The between-day CV for bone ash and percent fat are 1.0%
and 1 .7%, respectively (6). The radiation exposure is 0.02 mGy
per scan.
Total bodi’ potassium. Whole-body #{176}Kcounting was used
to derive TBK ( 1 1). The Brookhaven system consists of 54 so-
dium iodide detectors placed above and below the patient. The
FIG I.Reconstruction of DPA scan demonstrating landmarks that
subdivide body into six regions. The neck cut is made just below the
chin. The rib cuts are made as close to, but not touching, the spine. The
arms are isolated by running a line through the humeral head. The
pelvis cut is placedjust above the pelvic brim and the system computer
automatically draws the lower pelvic lines. The spine cut is placed just
below the last pair ofribs coming out ofT 12.
CV for TBK whole-body counting is 2.4%. The system opera-
tion was described in detail previously (1 1).
Total body nitrogen. Prompt -y-neutron-activation analysis
was used to estimate TBN. This system, described by Vartsky
et al (I 2), uses a plutonium-beryllium source of neutrons that
activates ‘4N nuclei. The 1 735 0prompt y decay of activated
nitrogen is then detected by two sodium iodide crystals that are
mounted above the patient. The present system has a CV of
2.4% in nitrogen-containing phantoms. Additional details of
prompt gamma TBN analysis are reviewed in references 12
and 13.
The absolute TBK and TBN were correlated directly against
by guest on July 10, 2011www.ajcn.orgDownloaded from

216 HEYMSFIELD ET AL
5i±SD.
TABLE I
Subject characteristics and baseline body composition results*
Age Weight Body mass
index Fat
3. kg kg/rn? %
Females(n =16) 56.3 ±22.6 59.7 ± 10.2 22.4 ± 2.9 31.7 ±6.8
Males(n= 18) 48.7± 15.9 72.7± 10.3 23.7±2.9 21.4±6.5
Total(n =34) 52.3 ±19.7 66.6 ±12.1 23.1 ±2.8 26.3 ±8.4
S  SD.
muscle mass estimated by DPA. In addition, total-body skele-
tal muscle mass was calculated by two equations proposed by
Burkinshaw (14, 15). In the first model, skeletal muscle mass
(5MM, kg) is determined by simultaneous measurement of
TBK (mmol) and TBN (g) by use ofthe following relation:
5MM =(TBK -1.33 TBN)/51.2
This model assumes that K/N in skeletal muscle and nonskele-
tal muscle lean tissue is 9 1 and 47 mmol/kg, respectively. The
second model replaces TBN with fat-free body mass(FFM, kg).
Our approach in this study was to use the equation
SMM - (TBK - 48 FFM)/43
in which FFM was calculated as body weight minus DPA total
body fat.
Anthropometric measurements. The anthropometnc mus-
cle-plus-bone area was calculated for upper midarm and mid-
thigh from respective circumference and skinfold measure-
ments. A single trained observer made all ofthe measurements
on the right side of standing patients (14). Midarm and mid-
thigh were identified as being halfway between the acromial
and olecranon processes ofthe scapula and the inferior margin
ofthe ulna and the inguinal crease and proximal border of the
patella, respectively. A calibrated tape measure was used to es-
tablish limb circumferences at each location. The triceps skin-
fold was then measured at the posterior aspect of the upper
midarm, and the results ofthree trials were averaged. The thigh
skinfold was measured at the circumference site in the midsag-
gital plane on the anterior aspect of the thigh. Limb muscle-
plus-bone area was then calculated as
[(circumference) - (ir X skinfoid)]2/4ir
where all units are in centimeters (16).
(2)
Statistical methods
Correlations between muscle mass and other body composi-
tion estimates were examined by using simple linear-regression
analysis (Statst, Statsoft, Tulsa, OK). All group results are ex-
pressed as mean ± SD.
Results
Subjects
There were 18 male and 16 female subjects (Table I) with
average age for the pooled group of52.3 ±19.7 (i± SD). Over-
all the group was relatively lean, with a body mass index of 23
±2 kg/rn2 and a percentage fat by DPA for men and women
of2l.4 ± 6.5% and 31.7 ± 6.8%, respectively.
DPA limb composition
(1) The repeated studies on four ofthe subjects resulted in CVs
of7.0 ± 2.4%, 2.4 ± 0.5%, and 3.0 ± 1.5% (1± SD) for upper-
extremity, lower-extremity, and combined-limb appendicular
skeletal muscle masses, respectively.
The bone, skeletal muscle, and fat content of the limbs as
estimated by DPA is presented in Table 2. Males had more
bone and skeletal muscle and less fat than did females for both
lower and upper extremities. Males also had more upper-ex-
tremity than lower-extremity skeletal muscle (upper/lower
=0.57) than did females (0.44). The ratio of bone to skeletal
muscle tended to be higher for both upper and lower extremi-
ties in males (0.089 and 0.142) than in females (0.085 and
0. 1 18) and more bone was present relative to skeletal muscle
in lower(pooled value =0. 135) than upper(0.094) extremities.
Skeletal muscle mass
No definitive methods are available for quantifying whole-
body skeletal muscle mass in vivo. DPA was therefore evalu-
ated in relation to available markers of skeletal muscle by
simple linear-regression analysis. The results ofTBK and TBN
estimates are presented in Table 3 along with calculated total-
body skeletal muscle mass (TBK, TBN, and FFM), anthropo-
metric limb muscle areas, and combined (upper and lower)
DPA limb muscle mass.
TBK was highly correlated with DPA extremity muscle mass
(3) fl pooled data for the 34 subjects (r =0.94, p<0.001 ; Table 4
and Fig 2). The correlation between DPA muscle and TBN
was also significant (r =0.78, p<0.001) and ofsimilar magni-
TABLE 2
Limb composition analyses from dual-photon absorptiometry*
Lower extremities Upper extremities
Bone Skeletal muscle Fat Total Bone Skeletal muscle Fat Total
kg kg
Females
Males
Total
1.3±0.3
2.0±0.4
1.7±0.5
11.0±2.2
14.1±1.7
12.6±2.5
5.9±2.0
4.1±1.6
5.0±2.0
18.2±3.5
20.2±2.9
19.3±3.4
0.4±0.1
0.7±0.1
0.6±0.2
4.7±0.9
7.9±1.6
6.4±2.1
3.7± 1.6
3.1±1.5
3.4±1.6
8.8±2.2
11.7±2.7
10.4±2.9
by guest on July 10, 2011www.ajcn.orgDownloaded from

0Males
.Females
0
2000 3000 4000 5000
TBK
(mmol)
SKELETAL MUSCLE MEASUREMENT 217
*Calculated from TBK and TBN.
tCalculated from TBK and FFM.
TABLE 3
Results ofbody composition studies*
TBK TBN 5MM 1SMM2 AMA TMA AMA +TMA SMM3
mmo/ kg kg kg cm cm2 cm2 kg
Females 2152±394 1.35±0.27 7.0± 5.9 5.0±4.6 31.9± 5.8 136.6±28.6 168.3±31.4 15.7±2.9
Males 3459±578 1.78±0.25 21.4± 8.0 17.1 ±7.1 57.6± 9.5 169.5±28.0 227.1 ±36.3 22.0±3.2
Total 2844±823 1.58±0.34 14.6± 10.1 11.4±8.6 45.5± 15.1 154.0±32.7 199.3±44.9 19.0±4.3
*SMM I, SMM2, and SMM3 are skeletal muscle mass calculated from, respectively, TBK and TBN (Eq 1), TBK and FFM (Eq 2), and the sum
ofupper and lower extremity 5MM estimated by DPA (Eq 3). AMA is arm muscle-plus-bone area; TMA is thigh muscle-plus-bone area.
tude to the correlation between DPA muscle and body weight
(r =0.80, p<0.001).
Total body estimates of skeletal muscle mass (Eqs 1 and 2)
were on average smaller than limb muscle mass estimated by
DPA (I 4.6 and 1 1 .4 kg, respectively, vs 19.0 kg). Although sev-
eral negative values were observed in the former, both calcu-
lated muscle estimates were significantly correlated with DPA
skeletal muscle (both r=0.82, p<0.00 1; Table 4).
Anthropometric muscle-plus-bone areas were significantly
(p <0.001) correlated with DPA extremity muscle mass, with
r=0.82 for upper limb r=0.88 for lower limb, and r=0.92
for the sum of upper- plus lower-limb muscle-plus-bone areas.
Thus two ofthe indirect markers ofskeletal muscle mass, TBK
and anthropometric muscle-plus-bone areas, were highly cor-
related with DPA muscle mass. Significant but weaker associa-
tions were observed between DPA muscle and whole-body
skeletal muscle derived by the TBK-TBN and TBK-FFM
models.
Discussion
Despite the obvious physiological relevance of quantifying
skeletal muscle mass, no definitive whole-body in vivo method
is yet available. The DPA technique described herein advances
our measurement capability by providing a practical approach
to estimating appendicular skeletal muscle mass. The extrem-
ity muscle per se is of intense interest and, moreover, the ap-
pendages account for a large portion (73-75%) oftotal skeletal
TABLE 4
Correlations between DPA appendicular skeletal muscle and other
body composition estimates
Equation rSEE p
Body weight 0.29x +0.0 0.80 2.7 <0.001
TBK 0.005x +5.01 0.94 1.6 <0.001
TBN l0.06x+ 3.14 0.78 2.8 <0.001
Skeletal muscle I  0.35x +1 3.83 0.82 2.6 <0.001
Skeletal muscle 2t 0.41x +14.33 0.82 2.6 <0.00 1
Arm muscle-plus-bone
area 0.24x +8.25 0.82 2.5 <0.001
Thigh muscle-plus-bone
area 0.l2x+ 1.04 0.88 2.1 <0.001
Arm +thigh muscle-
plus-bonearea 0.09x+ 1.33 0.92 1.8 <0.001
muscle mass (17). The between-day CV for the method (3%
for total extremity muscle) is within the range of other body
composition techniques, such as whole-body counting for po-
tassium. Hence the DPA approach brings within range the ca-
pability of reproducibly estimating all but one-fourth of skele-
tal muscle mass.
Skeletal muscle mass derived by DPA was highly correlated
with other regional (anthropometry) and total-body (whole-
body counting, neutron activation) estimates of muscle mass.
These associations demonstrate the potential of using DPA to
explore other methods ofquantifying muscle. For example, the
sum ofanthropometric limb muscle-plus-bone areas showed a
strong correlation (r =0.92) with DPA total-extremity muscle
mass, suggesting the potential for developing anthropometric
limb-muscle-mass prediction equations. Another example is
the demonstration that the neutron-activation and the whole-
body counting models (Eqs 1 and 2) for partitioning FFM into
muscle and nonmuscle components provides muscle estimates
that on average are too low (1 1-15 kg vs 19 kg for DPA limb
muscle), with negative values observed in some cases. These
models therefore need to be reconsidered and perhaps revised
in light ofthe present findings.
The DPA skeletal-muscle-mass method has several possible
limitations worthy of discussion. At present our gadolinium
system has a long scan time (55 mm), restricting the study to
patients with sufficient endurance. New x-ray-based dual-pho-
ton systems (DEXA) reduce scan time to  15 mm, thus par-
tially alleviating this problem. The minimal radiation doses are
<0.01 of 1% ofannual background, or ‘--2 h background mdi-
30
25
DPA
Skeletal Muscle
(kg) 20
15
10 -
1000
FIG 2. DPA appendicular skeletal muscle mass vs total body potas-
sium (TBK). Thirty-four observations pooled for males (n =18) and
females (n =I6) (DPA skeletal muscle =0.005 TBK +5.0, SEE =I .6
kg. r=0.94, p <0.001).
by guest on July 10, 2011www.ajcn.orgDownloaded from

218 HEYMSFIELD ET AL
ation. An additional problem is that the method developed in
the present study includes bone marrow and skin in the esti-
mate of limb muscle mass. However, these contributions are
relatively minor because the amount of nonfat extremity mar-
row and skin found in healthy young adults averages 1.2 kg, or
<5% of skeletal muscle as reported in this study (17). Finally,
changes in muscle hydration would alter the ratio between
muscle cell mass and total muscle weight, and this confounding
factor should be considered in the interpretation of results in
patients with edema.
In summary, we describe a promising new approach for esti-
mating the amount of skeletal muscle in the appendages. The
initial results in healthy adults indicate that limb muscle mass
determined by the DPA approach is highly correlated with
muscle estimates established by other available techniques. Fu-
ture studies are needed to evaluate the method’s applicability
in patients with altered body habitus and disease states. The
relative safety and low radiation exposure of the method and
the growing number and accessibility of available instruments
suggest that DPA may be a practical and widely applica-
ble technique of evaluating extremity skeletal muscle mass
invivo. II
References
1. Heymsfield SB, McManus C, Stevens V, Smith J. Muscle mass:
reliable indicator of protein-energy malnutrition severity and out-
come. Am J Clin Nutr l982;35: 1 192-9.
2. Heymsfield SB, Arteaga C, McManus C, et al. Measurement of
muscle mass in humans: validity ofthe 24-hour urinary creatinine
method. Am J Clin Nutr 1983; 37:478-93.
3. Buskirk ER, Mendez J. Lean body tissue assessment, with empha-
sis on skeletal muscle mass. In: Roche AF, ed. Body composition
assessments in youth and adults. Report ofthe Sixth Ross Confer-
ence on Medical Research. Columbus, OH: Ross Laboratories,
1985:59-65.
4. Heymsfield SB. Human body composition: analysis by computer-
ized axial tomography and nuclear magnetic resonance. In: Nor-
gan GN, ed. Proceedings of Euro-Nut Conference. The Hague:
CIP-Gegevens Konenkeijke Bibliotheek, 198:105-12.
S. Mazess RN, Peppler WW, Gibbons M. Total body composition
by dual-photon (‘53Gd) absorptiometry. Am J Clin Nutr l984;40:
834-9.
6. Heymsfield SB, Wang J, Funfar J, Kehayias JJ, Pierson RN. Dual
photon absorptiometry: accuracy of bone mineral and soft tissue
mass measurement in vivo. Am J Clin Nutr 1989;49:1283-9.
7. WangJ, Heymsfield SB, Aulet M, Thornton JC, Pierson RN. Body
fat from body density: underwaterweighing vs dual photon absorp-
tiometry. Am J Physiol 1989;256:E829-34.
8. Woodard HQ. The composition ofhuman corticol bone. Clin Or-
thop 1964;37: 187-93.
9. Burnell JM, Baylink DJ, Chestnut CH III, Mathews MW, Teubner
El. Bone mineral and matrix abnormalities in post menopausal
osteoporosis. Metabolism 1982; 31:1 1 13-20.
10. Heymsfield SB, Wang J, Kehayias JJ, Heshka S, Lichtman S, Pier-
son RN Jr. Chemical determination of human body density in
vivo: relevance to hydrodensitometry. Am J Clin Nutr 1989; 50:
1282-89.
1 1. Cohn SH, Shukle KK, Dombrowski CS, Fairchild RG. Design and
calibration of “broad-beam” 238Pu, Be neutron source for total-
body neutron activation analysis. J Nucl Med 1972; 13:487-92.
12. Vartsky D, Ellis KJ, Cohn SH. In vivo quantification ofbody nitro-
gen by neutron capture prompt gamma-ray analysis. JNucl Med
l979;20:l 158-65.
13. Cohn SH, Vartsky D, Yasumura S, et al. Compartmental body
composition based on total body nitrogen potassium, and calcium.
Am J Physiol 1980;239:E524-30.
14. Burkinshaw L. Measurement ofhuman body composition in vivo.
In: Orton CG, ed. Progress in medical radiation physics. New
York: Plenum, 1985;2:l 13-37.
15. Burkinshaw L. Models ofthe distribution ofprotein in the human
body. In: Ellis KJ, Yasumura S, Morgan WD, eds. In vivo body
composition studies. London: Institute of Physical Sciences in
Medicine, 1987:15-24.
16. Wright RA, Heymsfield SB. Nutritional assessment. Boston:
Blackwell Scientific Publications, 1984.
17. Snyder WS, Cook Mi, Nasset ES, Karhausen LR, Howells GP,
Tipton IH, eds. Report ofthe task group on reference man. Oxford,
England: International Commission on Radiological Protection,
1984. (ICRP report 23.)
by guest on July 10, 2011www.ajcn.orgDownloaded from