AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 89:347-357 (1992)

Measurement of Subcutaneous Adipose Tissue Using Ultrasound Images MARIA E. M I R E 2 Cardiovascular Genetics, University of Utah, Salt Lake City, Utah 84108

Subcutaneous fat, Anthropometry, Body composiKEY WORDS tion, Fat patterns ABSTEACT The objectives of this p a p 1 art: to explore the potential cf the ultrasound technique to quantify subcutaneous adipose tissue, and to explain the differences between skinfolds and ultrasound measurements across a large range of ages and levels of adiposity. The sample consisted of 115 men and 117 women aged 35 to 51 years, 132 girls and 145 boys aged 12 to 20 years. Subcutaneous fat thickness was measured at four sites using skinfolds calipers, and at seven sites using a real-time B-mode ultrasound scanner. Anthropometric measurements were obtained, and percent body fat was estimated using electric impedance. The agreement between skinfolds and ultrasound measurements was calculated for each age and sex group. The agreement between techniques, and the levels of correlation between body composition and fatness measurements were high in the sample of young men. However, the results were less consistent in the other groups. Site specific differences were also noted. 0 1992 Wiley-Liss, Inc.

In recent years interest has increased in the distribution of body fat as an indicator of disease risk and as an important determinant of body composition. The measurement of skinfold thickness with a constant-pressure caliper is the most widely used technique t o assess fatness, and it is probably the best way to measure subcutaneous fat in most field situations in terms of cost, speed, and convenience. The shortcomings of this technique are a function of the compressibility of the adipose tissue and the size of the skinfold. This technique is less useful when more precise measurements are needed: e.g., in the obese, the elderly, athletes in training, and those experiencing rapid weight gain or weight loss. To overcome those limitations, some of the techniques used in clinical diagnosis have been applied to the determination of body composition. These techniques contribute information that cannot be obtained with anthropometric measurements alone. Ultrs0 1992 WILEY-LISS, INC

sound imaging is one of the techniques that has been used in body composition research with success, as well as with some contradictory results (Chumlea and Roche, 1986; Jones et al., 1986). Some of the contradictions stem from differences in the equipment, techniques, and goals of each particular study. Some of the studies were aimed at determining the usefulness of ultrasoundmeasured subcutaneous fat to predict body density (Borkan et al., 19821, and other studies have looked at the reliability of this technique in specific groups (Volz and Ostrove, 1984;Chumlea and Roche, 1986;Kuczmarski et al., 1987; Fanelli et al., 1988). THE ULTRASOUND TECHNIQUE

This paper will describe how ultrasound imaging can be used to measure and describe subcutaneous adipose tissue and how _-

_-_-_-

Received January 15,1991; accepted April 12, 1992

348

M.E. RAMIREZ

the ultrasound data differ from and enhance the information obtained with other techniques. The ultrasound measurements are also compared to skinfold thickness and to the body fat contents calculated using anthropometric measurements and electric impedance. My goal has been to explore the potential of the ultrasound technique to increase the understanding of the subcutaneous fat tissue, and to explain the differences between measurements obtained using varying techniques across a large range of ages and levels o f adiposity; it has not been to predict overall fatness. Description of the technique Physical principles of ultrasound. U1trasonography is based on the pulse-echo principle. This term describes the information display system used almost exclusively in ultrasound scanning of any part of the body except the heart. All ultrasound imaging in brightness mode (B-mode) uses a short pulse of ultrasound which is emitted into the body. Although the principle of ultrasonic pulseecho diagnosis is the same as that of an echo traveling in a canyon, several practical differences relate t o the specific “sound” being used. These are a function of the wave length and cycle. The period is the time it takes a sound wave to complete a single cycle, one cycle per second is known as a Hertz; one million cycles per second is called a megahertz IMHzl. In most clinical B-scanning, the sound frequencies may range from 1to 5 MHz. The wavelength is the distance the wave travels during a single cycle, those used in B-mode ultrasound range from 0.3 to 1.5 mm. The amplitude or peak pressure or height of the wave is measured in decibel (dB)notation based on the comparison of the amplitudes of two different sound waves. In abdominal or pelvic ultrasound scanning, the echoes received from the soft tissues range in strength from 0 to 60 decibels. The actual amplitude range of these echoes is therefore 1to 10,000. Interaction of ultrasound and tissue. When an ultrasonic pulse is sent into the soft tissues of the body, it undergoes continuous modification. The most significant

change is attenuation, or the progressive weakening of the sound beam as it travels through tissue. The attenuation of ultrasound is dependent on distance and wavelength; on the density and degree of heterogeneity of the area scanned, and on the number and type of echo interfaces within the tissue. Other processes that influence that attenuation of an ultrasound beam are: absorption, reflection, and scattering. It is impossible to accurately predict the degree of attenuation for something as complex as humar? tissue. Neverthcless, attenuatiuri measurements have been made for many different human tissues with many different frequencies of ultrasound (Parker et al., 1984; Sanders et al., 1980). Types of devices Types of scanners. Numerous types and makes of ultrasound scanners are available for clinical and other uses. Most hospitals have several scanners of different age, size, make, and capabilities. Some of the most sophisticated scanners are large, are expensive and require extensive training to use. With the exception of those used for echocardiography, most B-mode real-time scanners will produce good images of subcutaneous fat tissue, provided that an appropriate transducer is used and the observer is experienced. The term “real-time”refers to the capability for processing information instantly to allow a dynamic presentation of sequential ultrasound images. As different images are produced with shifts in transducer position or with patient motion, these images are presented at a rapid rate in a movie-like fashion. Thus, it is possible to view the effects of motion and compression on the structures being measured. The rate a t which images are generated is called the frame rate. In different scanners the frame rate varies from 15 to 60 frames per second. Most scanners have a freeze-frame capability. This means that an image of each frame can be stored and recorded using X-ray film, instant print cameras or thermal paper printers. To document the motion observed on a series of images, a videotape machine can be used.

ULTRASOUND MEASUREMEPiT OF SUBCUTANEOUS FAT

Types oftransducers. The transducer is a means for converting electrical energy into mechanical sound waves. Transducers vary in the arrangement of the crystals and in wave frequency; they also vary in focal zone, face diameter, and quality. The frequency is determined by the number of times the crystals expand and contract per second. The focal zone is the distance range at which the lateral resolution is best, and the area may vary between 2.4 cm2 to 24.2 cm2. The correct transducer can be selected by matching the depth of the structure to be scanned t o the focal-zone distance, and by matching the size of the transducer’s face to that of the area to be scanned. The highest possible frequency should be used because this will result in superior resolution. To measure subcutaneous fat thickness a 5-MHz linear array transducer provided the best resolution in most cases. Risks. The Environmental Program of the World Health Organization (1982) reported no evidence of adverse health effects in human beings exposed to diagnostic ultrasound. Some biological effects of ultrasound have been reported in laboratory animals and in specific situations such as dental procedures that involve high doses of ultrasound. The level of exposure for measurement of subcutaneous fat is lower than the lowest level reported for diagnostic purposes. Two reasons are the short time the transducer remains over the area to be scanned [less than twenty seconds) and the fact that the areas scanned are large and distributed throughout the body. General scanning technique

Subcutaneous fat scanning has not been used for clinical diagnosis; therefore, few references can be found in the literature to the variations and problems encountered when scanning this tissue. When references exist, they are specific for the site and equipment used. The following recommendations are intended as general guidelines based on the experience of using ultrasound techniques with the sole purpose of measuring subcutaneous adipose tissue. An ultrasound examination involves considerable interaction between the subject and the observer.

349

The subject should have some idea about the procedure and be dressed in a hospital examination gown. The examination room should be private and the light dim in order to clearly visualize the screen. An orderly and consistent procedure should be used to identify the anatomical position of the scan site. The American Institute of U1trasound in Medicine (1980) recommends conventions which enhance descriptions of internal anatomical structures for diagnostic purposes. However, the use of ultrasonogr-aphy for anthropomctry has not beer? widespread, and there are no established protocols. The best technique for the measurements of fat thickness is to have the subject sitting or standing and t o take an approach similar to that used in anthropometry. Anatomical landmarks

An important difference between techniques is that for skinfolds measurements the landmarks are defined in reference to skeletal structures. The site to be measured is marked over the skin; then the caliper is applied directly over the markings. For ultrasound scanning sites defined and marked in this manner may not be helpful. Dim room light and the size of the transducer make any markings difficult to see. More importantly, the image at the site marked on the skin may not correspond to the internal structure it is intended to reflect. In those cases, the examiner will have to decide the exact site at which to measure the fat thickness; this measurement should represent the maximum accumulation of adipose tissue at that anatomical site. For this study, the sites used for ultrasound scanning were located as closely as possible to those used to measure skinfold thickness.’ Compression: The transducer with the contact gel should be held in one hand and applied to the skin with little or no pressure. Good contact can be maintained with all surfaces without compressing the underlying tissues. The real-time images make it possi-

‘Details of the technique and site definition can be obtained from the author.

350

M.E. RAMIREZ

SUSl = tricipital + suprailiac + subscapular + calf; SUS2 = tricipital + suprailiac + subscapular + calf + epigastric + hypogastric. The agreement between skinfold and ultrasound measurements was tested for each site using correlation coefficients, and by the method described by Bland and Altman SUBJECTS AND METHODS (1986). The effects of age, weight, and adiBody composition and subcutaneous fat posity on the agreement between skinfolds thickness were measured on 509 individu- and ultrasound were also evaluated. The adals: 115 men and 117 women aged 35 to 51 ipose tissue measurements were introduced years, 132 girls aged 11 to 21, and 145 boys in regression analyses aimed to predict body aged 12 to 20 years. All participants are composition. The calculations were made for white, healthy, residents of Salt Lake and the total sample, and by groups defined in terms of sex and age: men, women, boys and Davis Counties, Utah. At the time of the examination, the follow- girls. In addition to the actual measurements of ing data were obtained: date of birth, sex, level of physical activity, and a brief family adipose tissue thickness, the ultrasound imand medical history. The anthropometric ages showed that intermediate layers of inmeasurements obtained were weight in terfaces may be present through the tissue pounds (Detect0 medical scale); stature in at all sites. The compressibility of the adimm (Harpenden fixed stadiometer); bicon- pose tissue and the repeatability of the meadylar diameters at the elbow and knee (in surements are related to the degree of musmm, using a GMP sliding caliper); four cir- cle involvement. The ultrasound images cumferences (in mm, using a metal tape): provide more information about the adipose upper arm, calf, waist, and hip; and four tissue than is possible to obtain with skinskinfolds (Harpenden caliper) the average of fold calipers. This qualitative information three readings at the tricipital, subscapular, was not included in the analysis. suprailiac, and mid-calf. RESULTS Ultrasound measurements of subcutaneAll the variables were normally distribous fat thickness were taken at seven sites: tricipital, subscapular, epigastric, suprail- uted within age and sex groups, after skiniac, hypogastric, anterior thigh, and medial fold and ultrasound measurements were calf. Photographic records were obtained for transformed to decimal logarithms. The all measurements. Body composition was means and standard deviations for the anmeasured using electric impedance (Spec- thropometric measurements are presented trum Bodycomp I1 program, RJL Systems). on Tables 1and 2. All anthropometric, ultrasound and body The correlations between ultrasound and composition measurements were taken on skinfolds measurements for the total sample the right side of the body by the same ob- were high (r = .69 to r = 237). The levels of correlation by site were highest for the calf server. In addition to the measurements obtained (r = .76 to r = .88), and the differences bedirectly, age to the date of the exam was tween techniques were largest for the subcalculated, Body Mass Index [BMI = weight scapular (r = .43 to r = 32). There were (kg)/height2(cm)], Waist-to-Hip Ratio [WHR significant differences by group. The corre= waist circumference (cm)/hip circumferlations between techniques were highest ence (cm)], and Relative Fat Pattern Index among the boys (r = .82 to r = 3 7 ) and low[RFPI = Subscapular Skinfolds/subscapu- est among men (r = .43 to r = .76). lar + suprailiac skinfolds]. We also calcuPartial correlation coefficients between lated one sum of skinfolds (SSKF = tricipi- techniques were calculated for the total tal + suprailiac + subscapular + calf), and sample and by group after adjustment for two sums of ultrasound measurements: BMI and body fat contents. Using the total

ble to measure compression effects and adjust the transducer accordingly. It is important to bear in mind that obtaining reliable and reproducible ultrasound measurements requires technical expertise; until this technique is mastered, scanning may be frustrating and unreliable.

ULTRASOUND MEASUREMENT OF SUBCUTANEOUS FAT

351

TABLE 1. Means and standard deviations for body size measurements by sex and age groups Boys ( n = 145) . Mean SD

Men Women (n = 117) (n = 115) -. - __ Mean SD Mean SD ..____ ___ Weight (kg) Height (mm)

BMI' WTIRZ % Fat3 Waist4 Hip4 Upper arm4 Calfl Elbow5 IchPS

11 60 3 0.8 5 93 64 26 23 4 4

85.9 1,800 26 0.93 19 965 1,028 319 389 74 100

66.4 1657 24 0.77 29 784 1009 285 363 64 92

~

13 60 4 01 5 104 96 41 34 4 6

.-

61.7 1735 20 0.82 13 742 904 264 346 70 97

__

14 104 3 0.1 5 88 76 33 32 4 4

Girls (n =: 132) SD - _Mean _ _ ___ .. .

.. .

53.1 1645 20 0.74 23 666 893 240 333 61 88

10 70 3 0.04

5 59 111 28 32 3 5

Body Mass Index. Waist-to-Hip Ratio. Electric Impedance, RJL Systems *Circumference in millimeters. Diameter in millimeters.

TABLE 2. Means and standard deviations for subcutaneous fat measurements by sex and age groups Men Mean

SD

Women ( n = 117) Mean SD

3.9 12.8

2.0 4.3

9.8 21.8

3.6 19.9

1.6 7.0

7.2 17.7

(n = 115) _________ Tricipital Ultrasound Skinfold Subscapular Ultrasound Skinfold Suprailiac Ultrasound Skinfold Calf Ultrasound Skinfold Epigastric Ultrasound Hypogastric Ultrasound Thigh Ultrasound

___I___

Boys

Girls

___ ( n = 132)

(n = 145) -_____ Mean

SD

Mean

SD

4.9 7.0

3.0 9.9

2.4 5.3

5.3 14.5

3.5 5.3

3.9 18.3

3.1 8.2

1.9 8.9

1.3 5.0

2.5 11.3

1.9 5.5

3.5 3.6

9.3 21.1

5.6 9.8

3.5 9.0

4.5 7.6

4.9 12.5

3.5 6.2

3.3 11.3

1.7 5.0

7.3 21.7

3.5 8.3

2.5 9.9

1.8 5.1

4.5 4.8

2.8 6.5

19.7

8.8

17.9

11.3

5.2

8.9

7.8

9.0

19.9

10.7

19.4

11.9

6.7

10.8

9.4

9.3

3.2

1.3

5.1

2.7

1.9

4.3

3.i

8.96

'All measurements in millimeters; SD = standard deviation

sample, adjustment for BMI resulted in a significant decrease in the levels of correlation at the suprailiac site. Adjustment for total body fat decreased the correlations between techniques at the subscapular and tricipital sites. When the data were analyzed by group, the correlations among men were high for the triceps and calf sites after adjustments for BMI. In contrast, the adjustments decreased the correlation at the suprailiac site. Among boys, adjustment for BMI decreased the levels of correlation at the subscapular; and adjustment for fat contents reduced the correlation at the tricipital site. Among girls

and women, adjustments for BMI and body fat contents resulted in lower levels of correlation at all sites (Table 3). The agreement between techniques was also measured using the approach described by Bland and Altman (1986). The mean difference and the confidence limits of the agreement at each site were similar for the total sample and within each group. However, the results were very different when the comparisons were made with the total skinfold (which is the true value), and with one half of the skinfold. The confidence limits between techniques were significantly smaller when the half skinfold was used,

352

M.E. W I R E Z

TABLE 3. Total an.d partial correlation coefficients between skinfblds and ultrasomd ineasurenzents of subcutaneous fat bv p r o m "

Tricipital Total Partial

__ BMI 9% Fat

Subscapular Total Partial

__ BMI % Fat

Suprailiac Total Partial

-

BMI cz F2t

Calf Total Partial I/

BMI % Fat

All correlations are statistically significant a t P

Girls

Total sample

0.86 0.80 0.65

0.75 0.47 0.49

0.86 0.80 0.61

0.66 0.38 0.44

0.82 0.60 0.73

0.71 0.59 0.54

0.69 0.39 0.56

0.59 0.45 5. $4

0.84 0.54 0.60

0.85 0.69 n.59

0.84 0.44 9.60

0.84 0.65 9.69

0.76 0.71 0.72

0.77 0.59 0.56

0.87 0.83 0.73

0.88 0.72 0.72

0.87 0.85 0.71

Men

Women ___

Boys

0.71 0.67 0.69

0.78 0.45 0.51

0.43 0.17 0.35

=

0.001.

TABLE 4. Limits o f agreement between skinfold and half skinfold thickness with ultrasound measurements of subcutaneous fat

Tricipital Ultrasound vs. skinfold Ultrasound vs. 112 skinfold Subscapular Ultrasound vs. skinfold Ultrasound vs. 112 skinfold Suprailiac Ultrasound vs. skinfold Ultrasound vs. 112 skinfold Calf Ultrasound vs. skinfold Ultrasound vs. 112 skinfold

' SD d

=

-2SD1

d2

t 2SD

1.0 -2.6

9.1 4.1

17.2 9.8

-1.7 -1.6

11.1 4.1

23.9 9.8

--3.0 -4.6

8.5 1.2

20.1 6.7

-0.5 -1.0

9.9 2.8

20.4 6.6

Standard deviation. mean difference between techniques

=

TABLE 5. Ci,rrelutiori coefficients uf the rneun dif%wice beiweeri techniques with fatness neasurenzrnis Tricipital Subscapular ~.MD' BMI Percent fat

-0.34* 0.00 --0.15"

Suprailiac .

Calf

-0.17" -0.03 -0.06

0.43* 0.21' 0.34"

0.64" 0.63" 0.43"

*Correlation significant a t P >.001. 'MMU = mean difference skinfold ultrasound

-

___~_.______.~

~

because both the mean difference and the standard error were significantly smaller. The agreement was significantly influenced by the subject's degree of fatness (Tables

4.5). Age effects on subcutaneous fat thickness were different in each group: in men, age was significantly correlated with fat thickness at the suprailiac and epigastric sites.

However, in the adult female sample no significant age effects on fat thickness, fat distribution, or total adiposity were observed. Among boys age was significantly correlated with all subcutaneous fat measurements. Among girls age was significantly correlated with total fatness and with subcutaneous fat at the tricipital and subscapular sites. To determine how the different measurements of subcutaneous fat thickness correlate with indices of adiposity and fat distribution; the correlation coefficients between weight, body mass index, percent fat, and waist-to-hip ratio with subcutaneous fat measurements, and with the sums of those measurements were calculated. The results show that in men total body fat and WHR were highly correlated with the sum of all ultrasound measurements (SUSB). Among women, the sums of ultrasound measurements were more highly correlated with weight, BMI, and percent fat, than with the sum of skinfolds. In the girls' sample, the correlation between the sum of skinfolds and percent fat was higher. Among boys, BMI was highly correlated with subscapular skinfolds, and percent fat had the highest correlation with the suprailiac skinfolds. The waist-to-hip ratio was not significantly correlated with subcutaneous fat measurements (Tables 6, 7). In separate regression analyses (stepwise) BMI and percent fat were introduced as dependent variables, with the single adiposity measurements and the sums of those

ULTRASOUND MEASUREMENT OF SUBCUTANEOUS FAT

353

TABLE 6. Correlation coefficients between skinfold and ultrasound measurements of subculaneous fat thickness by age and sex groups * Men ........ . BMI %Fat __-.- A52

WT

-A!? Trieipital Ultrasound Skinfold Subscapular Ultrasound Skinfold Suprailiac Ultrasound Skinfcld

-0.02 0.27 0.33 -0 09 0.33 0.43

0.31

-0 10 0.43 0.49 -0.05 0.45 0.62 -

0.28

-0.25 0.31 0.44 0.19 0 5 4 n m

Calf Ultrasound -0.05 Skinfold 0.09. .._.. Epigastric Ultrasound 021 Hypogastric Ultrasound -008 __ Thigh Ultrasound 003 ~~

~

~

0.49

Women ~ WT BMI %Fat

-

~

- - --

Boys Girls .... WT BMI %;Fat--Age WT BMI %Fat -- . . .~

~

~

-0.02 0.68 0.72 _ . 0.07 0.81 0.81

0.73 0.81

-0.46 -0.51

0.26 0.52 0.41 0.67

0.77 0.16 0.61 0.68 0.76 0.25 0.76 0.81

0.66 0.81

0.08 0.65 0.61 0.03 0.73 0.75

0.59 0.69

--0.28 0.60 0.70 -0.26 0.65 0.81

0.57 0.21 0.58 0.62 0.71 0.23 0.72 0.77

0.53 0.73

0.03 0.79 0.83 0.13 0.76 0.78

0.79 0.78

-0.51 0.52 0.69 -0.68 0.47 0.75

0.69 0.16 0.73 0.78 0.81 0.15 0.72 0.78

0.77 0.78

0 0 3 066 067 007 _ _ 069 069

070 071

- 0 5 6 027 054 -053 026 056

077

016 - _ 069 074 014 071 077 -

072

073

0.50

~

0.50

-.-

053

0.31 0.41

0.36

0.38 0.45

0.40

0 3 3 045

042

_010 _ 066 065

070

-043

0 5 3 069

077

012 __ 065 0 7 3

072

062 0 7 1

065

_007 _ 065 072

078

-051

054 073

080

0 _ 1_5 070 077

074

004 024

013

0 1 3 065 068

062

-044

040 064

080

0 0 8 0 6 7 072

067

-

078

~

The levels of correlation are significant a t P

6

0 05. Underlined numbers denote correlation not significant a t P

s :

0.05

TABLE 7. Correlation coefficients between the sum of skinfolds and the sum of ultrasound measurements of fat thickness with measurements of total body fatness and fat placement Men SUS12 SUSP3 SSK' __________ -

~

Weight BMI % Fat WHR RFPI

0.45 0.55 0.49 0.28 -0.32

0.45 0.58 0.55 0.34 -0.33

0.55 0.69 0.64 0.51 -0.43

__ SSK 0.76 0.77 0.77 0.35 -0.23

Women SUSl SUS2_ ~

_

0.84 0.83 0.83 -0.31 -0.21

0.80 0.86 0.86 0.38 -0.24

Boys Weight BMI %) Fat WHR RFPI

0.35 0.64 0.78 0.49 -0.55

'SSK = (Subscapular Sk + Tricipital Sk + Suprailiac Sk + Calf Sk) x .5. SUSl = Subscapular US + Tricipital US + Suprailiac US -t Calf US. Tricipital US iSuprailiac US * Calf US Epigastric ITS +

measurements as the independent variables. The results showed that for the total sample, single skinfolds had greater predictive value for overall fatness than the sums of the measurements. BMI was best predicted by subscapular skinfolds, and body fat content was best predicted by tricipital skinfolds (results not shown). When the same analyses were performed for each group, different variables were retained. The variables listed in table 8 are those where the regression coefficient was significant to P > .001. In men, BMI and percent fat were best predicted using the ultrasound measurement of subcutaneous fat at the lower abdomen (hypogastric). Among women, BMI was best predicted by

Girls

-

SUS2 _SSK SUS2 ~ - _SSK _ - SUSl _ _ _-__ _ _SUSl ~ __ _ .0.50 0.70 0.80 0.58 -0.54

0.55 0.75 0.82 0.61 -0.52

0.74 0.80 0.81 0.00 -0.13

0.75 0.81 0.77 0.04 0.09

0.74 0.81 0.78 0.08 -0.15

+ Hg-posastric US

the ultrasound measurement a t the suprailiac site, and percent fat was best predicted by the sum of the ultrasound measurements. In boys, the subscapular skinfold was highly predictive of BMI, and the sum of ultrasound thickness (SUSS) was the best predictor of body fat contents. In girls, the tricipital skinfold was the best predictor of body fat contents, and both the sum ofultrasound measurements and the tricipital skinfolds were highly predictive of BMI (Table 8). DISCUSSION The results are consistent with those from other studies that have shown that some of the individual skinfolds are more highly cor-

354

M.E. RNWIREZ

TABLE 8. Summary of the stepwise regression analysis for BMI and percent fat by sex and age groups * BODY MASS INDEX ~-

Variable Hypogastric Subscapular US Subscapular Sk Variable Subscapular Sk Suprailiac Sk Tricipital US

Men

Women ~~

~

~

Partial R.

Pi

Variable

Partial R.

P