Quantification of adipose tissue by MRI: relationship with anthropometric variables.

Quantification of adipose tissue by MRI: relationship with anthropometric variables ROBERT

ROSS, LUC LEGER,

DAVID

MORRIS,

JACQUES

DE GUISE,

AND ROBERT

GUARD0

School of Physical and Health Education, Queens University, Kingston, Ontario K7L 3N6; Departement d’Education Physique, Universite de Montreal, Montreal, Quebec H3C 3J7; Institut de Genie Biomedical, Ecole Polytechnique, Montreal, Quebec H3C 3A7; and Department of Endocrinology, McGill University, Montreal, Quebec H3A 2B4, Canada Ross, ROBERT, Luc LUGER, DAVID MORRIS, JACQUES DE GUISE, AND ROBERT GUARDO. Quantification of adipose tissue by MRI: relationship with anthropometric variables. J. Appl. Physiol. 72(2): 787-795, 1992.-This study had two objectives: 1) to establishmagnetic resonanceimaging (MRI) as a tool for measuringtotal and regional adiposetissue (AT) distribution in humans and 2) to assessthe relationship between selected anthropometric variables and MRI-measured AT. Twenty-seven healthy men varying in age [40.8 t 14.5 (SD) yr], body massindex (28.5 t 4.8), and waist-to-hip ratio (WHR, 0.96 t 0.07) participated in the study. Total AT volume was determined using a linear interpolation of AT areas obtained on consecutive slices (n = 41) taken from head to toe (lo-mm thickness, 50-mm centers). The mean change for repeated measuresof total AT volume was2.9% (range 0.9-4.3%). Large interindividual differences were observed for total AT volume (6.9-59.3 liters), subcutaneousAT (6.3-49.8 liters), and visceral AT (0.5-8.5 liters). Visceral AT represented 18.3%of the total AT. The singlebest predictor of total adiposity was waist circumference (R2 = 0.92). For visceral AT volume, WHR was the strongest anthropometric correlate (r = 0.85, P < 0.01). When controlled for age and adiposity, however, WHR explained only 12% of the variation in absolute visceral AT and 125 cm. In these cases a mouse pointer was used to label the uncounted AT pixels. Visceral and subcutaneous AT regions on each image were labeled by assigning them different color codes. Because of difficulties in distinguishing between bone marrow and AT in the head region (because of its lipid content, bone marrow has pixel values similar to AT), the head was labeled as lean tissue. Similarly, the feet below the ankle were labeled lean tissue.

Reliability

Anthropometric

The reliability of both whole body total AT volume and AT area (cm2) measurements taken at the L,-L, level was assessed from repeated measurements on three male subjects. For each subject a complete data set was acquired (41 images) on two separate occasions during the same day. A single similar AT threshold (110 on a 256 scale) was used to segment all images. Segmentation of AT

The threshold selected for AT was based on the analysis of a sample of typical images and the respective gray level histograms. Results demonstrated that the optimal threshold for AT was 110 (on a scale of 256); above this value pixels were considered as representing AT. The next step involved viewing the pixels that were counted

Calculation of AT Area and Volume

The areas of the respective AT regions in each slice were computed automatically by summing AT pixels and multiplying by the pixel surface area. A previous investigation (19) verified that the error in the spatial dimensions of MR images by use of this system ranged from 0.1 to 1.5%. Thus the AT surface area on the MR images is accurate. The volume (cm3) of the respective AT regions in each slice was calculated by multiplying the AT area (cm2) by the slice thickness (10 mm). AT volume was calculated by adding the volumes of truncated pyramids defined by pairs of consecutive slices as follows V = h X Ni1 [Min

(Si, Si+l) + l/3 Abs (S,+l - Si)]

i=l

where V is the total AT volume, h is the distance between slices, N is the number of slices, Si and Si+l are the areas labeled as AT on the images of two consecutive slices, Min is the minimum value operator, and Abs is the absolute value operator. Total AT volume was calculated using all 41 slices. Visceral AT volume was calculated using the same formula, with the seven slices extending from two below L,--L, to four above. Variables

Body weight was measured on a balance scale calibrated to 0.1 kg. Barefoot standing height was measured using a wall-mounted stadiometer to the nearest 0.1 cm. Skinfold and circumference measurements. were obtained using the procedures described elsewhere (16). Skinfold thickness was obtained using a Harpenden skinfold caliper at the following sites: triceps, biceps, chest, subscapular, iliac crest, rib, thigh, and calf. Circumference measurements obtained with the subjects in a standing position were taken at the following sites: arm, chest, hip, thigh, calf, and waist at the umbilicus level. Waist circumference at the umbilicus level was also obtained with the subject in the supine position. Body fat distribution by anthropometry was estimated using both circumference measures, waist-to-hip ratio (WHR), and skinfold measures, subscapular-to-triceps ratio (STR).

Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (163.015.154.053) on August 17, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

QUANTIFICATION

OF

ADIPOSE

FIG. 1. A: transverse magnetic resonance (MR) image acquired at L,-L, level. A, adipose tissue. Subcutaneous and visceral adipose tissue is clearly distinct from lean tissue (B). B: transverse MR image of upper thigh region. Bone marrow appears with pixel intensity values similar to those of adipose tissue. Because of lack of mobile protons (thus no signal), cortex of bone appears black.

Statistical

Analysis

Data are presented as means + SD. Linear regression analysis was used to assess the simple relationship between variables. Multiple regression analyses were applied to identify the best predictors of total and visceral AT. The regressions were performed in a stepwise manner by use of the variables in Table 6. Partial correlation coefficients were used to examine the independent relationship of WHR to visceral AT after controlling for the effects of age and adiposity. For the cross-validation analysis, we tested the hypothesis that the estimated regression line was not significantly different from the identity line by use of the F test for a general linear hypothesis (8). Statistical procedures were performed using the Crunch Interactive Statistical Package (6). RESULTS

Distinct contrasts between AT and lean tissues were clearly identifiable on the MR images obtained. Figure 1A illustrates an example of an abdominal image obtained at the L,-L, level, and Fig. 1B is an example of a cross-sectional image of the upper thigh. As can be seen in Fig. lA, the subcutaneous and visceral AT compartments are clearly distinct from lean tissue structures and, thus, were segmented in a straightforward manner. Reliability

Reliability data obtained on three subjects are given in Table 1. The segmentation of AT for the three subjects was performed using the same threshold, 110 on a 256

TISSUE

BY

789

MRI

scale, above which pixels were considered AT. For all subjects, no corrections for MRI artifacts were performed. The change between test 1 and 2 for total AT volume ranged from 0.9 to 4.3%, and for AT area measurements obtained from the two transverse images taken at the L,-L, level the difference ranged from 1.4 to 4.2%. These results confirm our previous finding (19) and others (20,25) that MRI measurements show good reproducibility. There is, however, an image acquisition artifact termed “ghosting” that must be corrected interactively. An example of this phenomenon is illustrated in Fig. 2. The arrow in Fig. 2 points to a region in the subcutaneous AT that appears as a shadow. The pixels in this region have an intensity value that is below the other subcutaneous AT pixels. Thus, when the image is segmented, the pixels in the affected area will not be counted. As described above, this problem can be corrected by using straightforward image-processing techniques. To determine the magnitude of the manual corrections, total AT volume values obtained using corrected vs. uncorrected images were compared using 10 of the initial 27 subjects. For total AT volume, the mean difference between the corrected and uncorrected images was 3.3%. However, the error ranged from 0.9 to 9.8%, which supports the observation that segmented MR images require visual verification. The interobserver error for segmenting total AT was tested by comparing the segmentation results obtained on 10 subjects by two individuals. The mean difference for total AT was 1.2%, with a range of l-3%. Taken together, these results indicate that the reliability of the MRI model used is very good. More importantly, however, our results demonstrate that, regardless of the segmentation procedure, until MRI acquisition procedures improve, visual verification of all segmented MR images is required. AT Measured

by MRI

Twenty-seven healthy volunteers recruited from both university personnel and the general public gave their fully informed consent to take part in the study. Their descriptive characteristics are given in Table 2. A wide range in values existed for age-, body mass index-, and MRI-measured total AT volume (6.9-59.2 liters). Although a large range also existed for WHR (0.82-1.17), the mean WHR was 0.96 + 0.07, which is characteristic of an android habitus. The general distribution of AT area measurements per image are illustrated in Fig. 3. Although only one image 1. Reliability data obtained occasions during the same day

TABLE

Total AT Volume, Subj No. 1 2 3 AT,

on two separate

liters

AT Area, cm2 (L,-L,)

Test 1

Test 2

A%

Test 1

Test 2

A%

24.81 22.44 16.52

24.15 22.23 15.80

2.6 0.9 4.3

313.1 283.8 208.4

300.1 279.7 201.6

4.2 1.4 3.3

adipose

tissue;

A%, (test 1 ~ test 2)ltest

1 X 100.


790

QUANTIFICATION

OF ADIPOSE

TISSUE

BY MRI

FIG. 2. Transverse MR image through midabdomen region. Arrow, region within subcutaneous adipose tissue that appear *s as a shadow. This is a randomly occurring MR imaging (MRI) artifact that must be corrected for after segmentation of the image.

was anatomically landmarked (Lb-L,), for most individuals images l-l 7 represent the legs, images 19 (pelvic region)-26 (liver) the abdominal region, and images 2841 the upper torso and arms. ~Inspection of the AT area measurements per image (Fig. 3) shows that the largest mean value for total AT area was obtained at the level of L,-L, (370.8 cm2), which represented 53.6% of the total area. For subcutaneous AT area, the highest mean value (307.01 cm2, 40.4% of total area) was obtained on a transverse image through the buttocks region. For visceral AT, the highest mean value (180.98 cm2, 27.6% of total area) was on the image acquired 10 cm above L,-L,, which is the approximate level of the umbilicus. Mean values for total AT area obtained on images 19-26, the abdominal region, represented 43.0% of the total AT area (41 images). Mean values for total AT area of the legs (images l-l 7) represented 28.9%. TABLE

Min

Age, yr Ht, cm W kg Waist, cm Hip, cm BMI, kg/m’ WHR STR SOS

40.8k14.5

17

174.7rt5.8 87.1k16.8

101.5L14.9 104.9k9.4

166.0 56.4 75.4 87.6

28.5k4.8

19.8

0.96kO.07 1.64kO.64

0.82 0.88 22.2

60.8rt26.1

AT Area (L4-LJ

The relationships between L,-L, AT area measurements and volume (liters) measures are given in Table 4. Total AT area on the L,-L, image was strongly corre0l

-

0 Visceral l Total

F300

Q a, 200 i?

Range Mean + SD

Relationship Between MRI-Measured and Volume Measurements

2. g 350

2. Descriptive characteristics of subjects

Variable

Large interindividual differences were observed for all MRI-measured variables (Table 3). Total AT volume ranged from 6.9 to 59 liters, subcutaneous volume from 6.3 to 49.8 liters, and visceral AT volume from 0.5 to 8.5 liters. Subcutaneous AT volume [21.2 +- 10.4 (SD) liters] represented 81.2% and visceral AT volume (4.8 + 2.2 liters) 18.3% of the total AT volume.

Max 74 185.0

129.2 135.6 127.0

39.3 1.17

c=” 150

I0 100 .a 2

50 0 0

4

8

12

16

20

24

28

32

36

40

4.32 130.4

Data were obtained from 27 subjs. WHR, waist-toyhip ratio, obtained using umbilicus waist circumference; STR, subscapular-to-triceps skinfold ratio; SOS, sum of skinfolds (biceps, triceps, subscapular, and iliac); BMI, body mass index.

FIG. 3. Distribution of MRI-measured adipose tissue. Image 1 is at the level of the foot; image 41 is at the level of the hands. In general, images l-l 7 represent the legs, 19-26 the abdomen region, and 28-41 the upper torso and arms.


QUANTIFICATION

OF

ADIPOSE

3. MRI AT distribution

TABLE

Mean

BY

791

MRI

5. Correlation matrix between visceral adipose tissue area values obtained on 5 abdominal images and visceral adipose tissue volume TABLE

Range AT Variable

TISSUE

+ SD

Min

Max Variable

L*-L,, cm2 252.8t132.9 117.9t62.1 370.8t165.8 Volume, liters

Subcutaneous Visceral Total Subcutaneous Visceral* Total

47.4 20.0 67.4

21.2tl0.4 4.1t2.2 26.ltl2.1

Variable

609.1 266.6 797.4

6.3 0.5 6.9

1) Vis volume 2) -5 cm 3) L-L, 4) +5 cm 5j +10 cm 6) +15 cm

49.8 8.5 59.2

Data were obtained from 27 subjects. MRI, magnetic resonance imaging. * Derived from 7 abdominal images (see METHODS).

4. Correlation matrix for selected MRI AT variables

TABLE

Variable Variable

1)

2) 3) 4) 5) 6)

LA-L, Sub L,-L, Vis L,-L, Total AT Sub volume Vis volume Total AT

1

2

0.94 0.97 0.48 0.95

0.66 0.95 0.52

3

4

5

6

1

2

3

4

5

0.89 0.95 0.97 0.96 0.96

0.91 0.93 0.85 0.78

0.97 0.91 0.95

0.94 0.88

0.95

6

-

-5 cm, transverse abdominal image acquired 5 cm below L,-L,; +5 cm, +10 cm, and +I5 cm, transverse abdominal images acquired 5,10, and 15 cm above L4-L,, respectively. All coefficients are significant at 0.01 level.

ous AT volume (r = 0.68, P < 0.01) and L,-L, area ratio measurements (r = 0.52, P = O.Ol), confirming the observation that age is a better predictor of relative visceral AT than level of adiposity (3). Prediction of Total and Visceral AT by Anthropometric Variables and Age

0.91 0.74 0.96

Total AT. The simple correlation coefficients obtained among MRI-measured AT, age, and selected anthropometric variables are given in Table 6. For both total AT volume and L,-L, area measurements, the strongest Sub volume, subcutaneous AT volume; Vis volume, visceral AT volume obtained from 7 abdominal images (see METHODS); Total AT, total anthropometric correlate (including age) was either AT volume. Coefficients shown are significant at 0.01 level; -, not standing or supine waist circumference (r = 0.94, P< significant (P > 0.05). 0.01). When offered in a stepwise regression model with variables given in Table 6, the lated (r = 0.95, P < 0.01) with total AT volume, as was the other anthropometric combination of standing waist circumference and WHR L,--L, subcutaneous AT area (r = 0.95, P < 0.01). Visceral explained 91% of the variation in total AT volume (Table AT area at the L,-L, level was positively correlated with 7). For the prediction of L,-L, total AT, waist circumfervisceral AT volume (r = 0.95, P = 0.01). ence alone explained 91% of the variation in total AT. A correlation matrix for visceral AT area measureThe addition of triceps skinfold and weight increased the ments obtained on five abdominal images and visceral R2 to 93%. Clearly, waist circumference is an excellent AT volume is given in Table 5. For the five abdominal images acquired between 5 cm below L,-L, and 15 cm predictor of abdominal adiposity. Visceral AT. Given the strong relationship observed above (1 image every 5 cm), the correlations with total between L,-L, visceral AT area and visceral AT volume, AT volume ranged from 0.89 to 0.97 (P < 0.01). The preof L,--L, visceral AT is reported. dictive value of the image obtained 5 cm below L,-L, (R2 only the prediction = 0.79) is relatively low and is not as good a predictor as When the variables listed in Table 6 were offered in a of age and the slices obtained from L,-L, to 15 cm above. Althou -gh stepwise regression model, the combination WHR explained 81% of the variation in visceral AT (Tathe correlations observed are biased by the in .clusion of the abdominal image of interest, the visceral AT areas of ble 8). Although subcutaneous skinfold, supine waist circumference, and body mass increased the R2 to 86%, besingle abdominal scans acquired in the umbilicus region among the are, nevertheless, strongly predictive of the volume of cause of a high degree of multicollinearity variables they are not reported. visceral AT obtained by multiple images. Cross validation. Because of the limited number of subjects in this study, rather than cross validate the derived Relative and Absolute Visceral AT in Relation to Total prediction equation, we opted to cross validate the visAT: Effects of Age ceral AT prediction equation of Seidell et al. (22). As Visceral AT volume was positively related to total AT illustrated in Fig. 4, the R2 value obtained when the prevolume (r = 0.66, P < O.Ol), as was visceral AT area with dicted L,--L, visceral AT values were regressed against total AT area measured at the L,-L, level (r = 0.66, P < the MRI-measured values was low (R2 = 0.58, SEE = 0.01). Age was not significantly related to total AT vol- 34%). Analyses indicated that the plotted regression line ume (r = 0.29, P > 0.05). (slope and intercept) was significantly different from the Visceral-to-subcutaneous AT ratio was not signifiline of identity (F = 14.18, P < 0.05). The mean visceral cantly related to total AT for either volume (r = 0.04) or AT value we obtained for L,-L, was 117.9 t 62.1 cm2, L,--L, total AT area (r = 0.21) measurements. Age was, whereas the predicted mean value using the equation of however, positively rela .ted to both visceral-to-subcutaneSeidell et al. was 80.9 t 66.2 cm2 (P < 0.001, paired t test). 0.50 0.98

0.66


792

QUANTIFICATION

OF

ADIPOSE

TISSUE

Anthropometric Waist*, Age,

Wt, kg

Ht, cm

Variables

Yr

Subcutaneous Visceral V/S ratio Total AT

-

0.88

-

0.85 0.52 0.46

-0.50 0.83

-

0.90

BMI, kg/m2

Subcutaneous Visceral V/S ratio Total AT

-

0.74 0.68 -

WHRSU, WHR sured at umbilicus

Skinfold, Hip, cm

Standing

Supine

0.89

0.92

0.93

0.94

0.41

-0.44 0.87

0.58 0.96

0.54 0.95

-0.58 0.85

Volume,

liters

-

0.90

0.90

0.49 -

-

0.89

-

0.57 0.90

0.70 0.94

0.91 0.67

Chest, cm

Subscapular

mm Triceps

Predictors

R2

circumference,

Total AT volume (liters) dard error of estimate.

cm

0.880

0.910 =

1.003X, - 56.475X,

SEE,

4.3 3.7 - 21.364.

0.95 0.42 0.92

0.85 0.43 0.84

0.66 0.82 0.83

0.65 0.78 0.82

0.84 0.56 0.85

0.83 0.64 0.87

0.89 0.56 0.90

0.61 0.85 0.47 0.72

0.59 0.82 0.49 0.70

-

* Circumference

mea-

AT

ratio.

8. Multiple regression equation for predicting visceral AT area (L4-L5) measured by MRI from anthropometric measures TABLE

0.05).

7. Multiple regression equation for predicting total adipose tissue volume measured by MRI from anthropometric measures

WHR

cm2

0.94 -

TABLE

Waist WHR

Variable

cm

L,--L5,

Xl

MRI

6. Correlation coefficients between MRI measurements and anthropometric variables and age

TABLE

x2

BY

stan-

Relationship Between WHR and Both Absolute and Relative Visceral AT WHR vs. visceral AT volume. As reported above, without controlling for adiposity, WHR was significantly correlated with absolute visceral AT volume and L,-L, area measurements. Because it is possible that these correlations reflect the influence of age or total AT volume, we entered these variables into a stepwise regression and partial correlations were calculated. For visceral AT volume, the strongest correlate was WHR (r = 0.85, P < O.Ol), and the highest partial correlation coefficient obtained was for age (r = 0.39, P < 0.05). After controlling for WHR and age, the partial coefficient for total AT volume was not statistically significant (r = 0.36, P > 0.05). When controlled for age, the correlation between visceral AT volume and WHR was r = 0.69 (P < 0.01). When controlled for both age and total AT volume, however, the partial correlation coefficient for WHR did not attain statistical significance (r = 0.35, P > 0.05). WHR vs. L,-L, visceral AT area. For visceral AT area measured at the L,-L, level, the strongest correlate was age (r = 0.85, P < O.Ol), and the highest partial correlation obtained was for WHR (r = 0.60, P < 0.01). When controlled for both age and adiposity, the partial correlation for WHR (r = 0.26) was not significant. When the relationship between age and L,-L, visceral AT area was corrected for WHR, the correlation coefficient was 0.68 (P < 0.01). WHR vs. visceral-to-subcutaneous ratio. We next evalu-

ated the relationship among WHR, age, total AT, and the visceral-to-subcutaneous ratio. Age remained the strongest correlate for both visceral-to-subcutaneous AT vol-

Predictors

R2

SEE

P

Xl

Age

x2

WHR

0.713 0.807

33.3 27.3

0.10). DISCUSSION

The results of this study demonstrate that MRI is a feasible and reliable method for assessing both total and regional AT distribution in humans. This finding supports a previous study that used an animal model to validate the accuracy of MRI to measure AT (19). To our knowledge, no other study has used MRI to measure whole body AT volumes, although others have used MRI to measure AT area from single slices (20, 25). We are aware of two studies that have used CT to measure whole body AT! volumes in men (12, 26). The values we obtained for total and visceral AT volumes are slightly lower than previously reported (12), although the percent contribution of visceral to total AT volume was similar (18 vs. 20%). The variance between the studies may be partially explained by differences in adiposity, because the mean body mass index (weight/height’) for our subjects (mean 28.5) was lower than theirs (mean 31.5). For AT area measured at the L,-L, level, the values obtained are of the same order of magnitude as those previously reported using CT (23). Kvist et al. (13) observed that a reduction in the number of images used to calculate total AT volume from the


QUANTIFICATION

OF ADIPOSE

r2 = 0.54, SEE = 40.1 cm2 (34%)

Line

I

0

0

50

I

I

I

100

150

200

Predicted

Visceral

AT L4-L5

I

250

I

300

(cm2)

FIG. 4. 4.

Comparison of predicted and MRI-measured visceral adipose tissue values. Predicted values were obtained using equation of Seidell et al. (22): visceral AT (cm2)Os5= 0.350 (BMI) + 0.405 (skinfolds)0*5 + 33.118 (WHR)‘s5 + 0.068 (age) - 37.322, where AT is adipose tissue, BMI is body mass index, and WHR is waist-to-hip ratio.

original 23 to 9 resulted in a small error of -1%. We performed a similar analysis and determined that, for total AT volume, a selective reduction of scans from 41 to 20 resulted in a high correlation (r = 0.99) and a standard error of -1.5%. Further reduction in the number of images resulted in an error of >2% (data not shown). For CT, a reduction in the number of images acquired to calculate AT volume has the practical advantage of reducing the subjects’ exposure to ionizing radiation. Because MRI is not subject to such restrictions, a reduction in data acquisition has no practical advantage. Furthermore a reduction in the number of images acquired would not substantially decrease the time required to gather the data, because most MRI systems can acquire data for several body slices in the same time it takes for a single slice. Thus, for example, were we to reduce the number of images in each set from 7 to 4, the total time required to obtain the MRI data would remain the same. Therefore, if MRI is to be used as a criterion method of body composition assessment, we recommend that -40 images be acquired. In fact, because the abdomen is a region of physiological interest and -45% of the total AT is located in this region, future studies would no doubt benefit from acquiring additional MRI data in this area. The observation that AT area measurements obtained at the L,--L, level were highly predictive of the corresponding volume measurements confirms the observation of Kvist et al. (12), who made similar observations with both male and female subjects. Borkan et al. (4) reported that the AT area measurements from multiple slices taken in the umbilicus region were highly correlated with one another. These observations suggest that, in cross-sectional studies, total and visceral AT determined from single scans at the level of the umbilicus are highly predictive of visceral volume measurements acquired by multiple scans. What has not been demonstrated, however, is whether diet or exercise intervention would result in selective effects on subcutaneous or visceral AT depots in the abdomen region that would be identified only through multiple scan acquisitions. MRI is particularly suited to intervention studies, because nu-

TISSUE

BY MRI

793

merous slices (20-30) of the abdomen could be acquired sequentially without hazard to the subject. Several studies have shown that abdominal distribution of AT, as indicated by the WHR, is an independent predictor of metabolic aberrations including insulin resistance (ll), hyperlipidemia (1, 7), hypertension (27), and atherosclerosis (14). It is generally believed that the correlative power of WHR is its ability to predict absolute or relative amounts of visceral AT. Because the liberated free fatty acids from much of the visceral AT have direct access to the liver via the portal vein, it has been proposed that visceral AT is a likely mediator for some of the apparent effects of abdominal adiposity on glucose and lipid metabolism (15). What has yet to be firmly established however, is the ability of WHR to predict absolute or relative visceral AT. Three previous studies reported significant relationships between WHR and CT-measured visceral-to-subcutaneous AT ratio (2, 9, 18). Baumgartner et al. (3), however, reported that WHR was significantly associated with absolute and relative visceral AT in women only. Seidell et al. (22) reported that WHR did not correlate significantly with visceral-tosubcutaneous ratio measured at the L, level after adjusting for age and body mass index. Kvist et al. (13) reported significant relationships between WHR and visceral AT; the R2 values were 69 and 35% for absolute and relative visceral AT, respectively, but they did not investigate the effects of age or adiposity. The results of this study show that for absolute visceral AT volume, after controlling for both age and adiposity, WHR explained only 12% of the variation in visceral AT. When abdominal obesity was considered from visceral AT area measurements obtained at the L,-L, level, WHR explained 7% of the variation in visceral AT. Furthermore, after controlling for age and adiposity, the observed relationship between WHR and relative visceral AT was nonexistent. These results suggest that, although WHR is an independent predictor of numerous metabolic aberrations, its correlative power may not be explained by the ability to predict either absolute or relative visceral AT. Indeed, our findings suggest that age is a better predictor of relative visceral AT mass than either total adiposity or WHR. One factor that may partially explain our finding of a reduced correlation between visceral AT and WHR is the level of waist circumference we chose to calculate WHR. Because it is well known that varying the level of waist circumference can significantly affect the calculated WHR (21), WHR values obtained using a waist circumference other than the umbilicus may improve the relationship between WHR and visceral AT. For example, Peiris et al. (18) observed that a WHR calculated using a waist circumference obtained at the level of minimum girth remained significantly correlated to visceral AT after adjusting for age and body mass index. Thus future studies that would assess the ability of WHR to predict visceral AT might benefit from calculating WHR values by use of more than one waist circumference; however, waist circumference should be obtained using accepted standards (i.e., level of minimum girth and umbilicus). Furthermore it would be ideal if these studies would simultaneously assess metabolic parameters to evaluate


794

QUANTIFICATION

OF ADIPOSE

whether the WHR that is best related to visceral AT is also associated with clinical abnormalities or metabolic risk factors. Baumgartner et al. (3) noted that WHR circumference measures were normally obtained with the subject in a standing position, whereas CT (or MRI) transverse scans were obtained with the subject supine. Because the effects of gravity may change the position of intra-abdominal mass in a standing vs. a supine position, it was hypothesized that the correlations between WHR and visceral AT measures might improve if the circumference measures were taken in a supine position. We calculated WHR using a supine waist circumference taken at the umbilicus level but observed no improvement in the relationship between WHR and absolute or relative visceral AT. Thus, movements in abdominal mass may not be a factor that explains the reduced correlation between WHR and visceral AT. One of the potential benefits of assessing abdominal adiposity by MRI or CT is the development of mathematical equations from external anthropometry that can predict MRI-measured visceral AT. Two previous studies have published prediction equations for visceral AT (12, 22). We chose to cross validate the equation of Seidell et al. (22) for predicting visceral AT in men, inasmuch as it was developed with a larger number of subjects (n = 71). Although the descriptive characteristics of the subjects in the two studies were similar and the dependent variable, visceral AT area, was obtained at a similar anatomic level, the predictive equation did not tally strongly with direct measurements. The variance between the measured and predicted values may be partially explained by the fact that the independent variables were obtained at slightly different anatomic locations. Thus, although our observations do not necessarily negate the validity of the equation of Seidell et al. (22), the results do emphasize the need to develop prediction equations based on large numbers of subjects. Inasmuch as MRI is not subject to the restrictions associated with CT, larger numbers of subjects could be assessed and more robust prediction equations developed. In conclusion, MRI offers a reliable and accurate measure of total and regional AT distribution in humans. MRI will be of particular value in assessing the impact of intervention studies such as the effects of diet and exercise on regional AT distribution, inasmuch as multiple scans can be acquired without threat to the subject. Further study is required using MRI or CT to clarify whether WHR is an independent predictor of visceral AT. The valuable technical assistance of Gilles Leroux, Andre Cormier, and Yves Martel is gratefully acknowledged. We also thank Dr. Jean Lambert for reviewing the statistical procedures. This study was financially supported by Canadian Fitness and Lifestyle Research Institute Grant 0052 1128 1014. Address for reprint requests: R. Ross, School of Physical and Health Education, Queen’s University, Union St., Kingston, Ontario K7L 3N6, Canada. Received 12 April 1991; accepted in final form 6 September 1991. REFERENCES

J. A., K. A. SOBOCHINSKY, D. S. FRIEDMAN, J. J. BARA. RIMM, AND H. W. GRUCHOW. Body fat distribution, plasma lipids and lipoprotiens. Arteriosclerosis 8: 88-94, 1988.

1. ANDERSON, BORIAK, A.

TISSUE

BY MRI

M., T. J. COLE, AND A. K. DIXON. Obesity: new insight into the anthropometric classification of fat distribution shown by computed tomography. Br. Med. J. 290: 1692-1694,1985. 3. BAUMGARTNER, R. N., S. B. HEYMSFIELD, A. F. ROCHE, AND M. BERNARDINO. Abdominal composition by computed tomography. 2. ASHWELL,

Am. J. Clin. Nutr. 48: 936-945, 1988. 4. BORKAN, G. A., S. G. GERZOF, A. H. ROBBINS, D. E. HULTS, C. K. SILBERT, AND J. E. SILBERT. Assessment of abdominal fat content by computed tomography. Am. J. C&n. Nutr. 36: 172-177,1982. 5. BORKAN, G. A., AND D. E. HULTS. Relationships between com-

puted tomography tissue areas, thicknesses and total body composition. Ann. Hum. BioZ. 10: 537-546, 1983. 6. Crunch Interactive Statistical Package. San Francisco, CA: Crunch Software, 1985. 7. DESPRI?S, J. P., C. ALLARD, A. TREMBLAY, J. TALBOT, AND C. BouCHARD. Evidence for a regional component of body fatness in the association with serum lipids in men and women. MetaboZism 34: 967-973,1985. 8. DRAPER, N.

R., AND H. SMITH. Applied Regression Analysis (2nd ed.). New York: McGraw-Hill, 1981. 9. FUJIOKA, S., Y. MATSUZAWA, K. TOKUNAGA, AND S. TARUI. Contribution of intra-abdominal fat accumulation to the impairment of glucose and lipid metabolism in human obesity. MetuboZism 36: 54-59,1987. 10. HAYES, P.

A., P. S. SOWOOD, A. BELYANIN, AND F. W. SMITH. Subcutaneous fat thickness measured by magnetic resonance imaging, ultrasound and calipers. Med. Sci. Sports Exercise 20: 303-

309,1988. 11. KISSEBAH, HARTZ, R.

A. H., N. VYDELINGUM, R. MURRAY, D. J. EVANS, A. J. K. KALKHOFF, AND P. W. ADAMS. Relation of body fat distribution to metabolic complications of obesity. J. CZin. Endo-

crinol. Metab. 54: 254-260, 1982. 12. KVIST, H., B. CHOWDHURY, U. GRANGARD, SJ~STR~~M. Total and visceral adipose tissue

U. TYLI%N,

AND L.

volumes derived from tomography in adult men and

measurements with computed women: predictive equations. Am. J. CZin. Nutr.

48: 1351-1361,

1988. 13. KVIST,

H., L. SJ~STR~M, AND U. TYLBN. Adipose tissue volume determinations in women by computed tomography: technical considerations. Int. J. Obes. 10: 53-67, 1986. 14. LARSSON, B., K. SVARDSUDD, L. WELIN, L. WILHELMSEN, P. BJ~RNTORP, AND G. TIBBLIN. Abdominal adipose tissue distribution, obesity and risk of cardiovascular disease and death: 13 year follow up of participants in the study of men born in 1913. Br. Med. J. 288: 1401-1404,1984. 15. LEIBEL, R. L., N. K. EDENS,

AND

the control of body fat distribution 417-443,1989. 16. LOHMAN, Anthropometric

T. G., A. F. ROCHE, Standardization

S. K. FRIED. Physiologic basis for in humans. Annu. Reu. Nutr. 9: AND R. MARTELLO Reference Manual.

(Editors). Champaign,

IL: Human Kinetics, 1988. P., AND P. G. MORRIS. NMR Imaging in Biomedicine. Toronto: Academic, 1982. 18. PEIRIS, A. N., M. I. HENNES, D. J. EVANS, C. R. WILSON, M. B. LEE, AND A. H. KISSEBAH. Relationship of anthropometric measurements of body fat distribution to metabolic profile in premenopausal woman. In: Health ImpZications of Regional Obesity, edited by P. Bjorntorp, U. Smith, and P. Lonnroth. Stockholm: Almqvist & Wiksell, 1987, p. 179-188. 19. Ross, R., L. LINGER, R. GUARDO, J. DE GUISE, AND B. G. PIKE. Adipose tissue volume measured by magnetic resonance imaging and computerized tomography in rats. J. AppZ. Physiol. 70: 216417. MANSFIELD,

2172,199l. 20. SEIDELL,

J. C., C. J. C. BAKKER, AND K. VAN DER KOOY. Imaging techniques for measuring adipose-tissue distribution-a comparison between computed tomography and 1.5-T magnetic resonance.

Am. J. CZin. Nutr. 51: 953-957, 1990. 21. SEIDELL, J. C., M. CIGOLINI, J. CHARZEWSKA, F. CONTALDO, ELLSINGER, AND P. BJORNTORP. Measurement of regional bution of adipose tissue. In: Obesity in Europe. Proceedings First European Congress on Obesity, edited by P. Bjorntorp

B. M. distriof the

and S.

Rossner. London: Libby, 1988, p. 91-99. J. C., A. OOSTERLEE, M. THIJSSEN, J. BUREMA, P. DEURJ. HAUTVAST, AND J. JOSEPHUS. Assessment of intra-abdominal and subcutaneous abdominal fat: relation between anthro-

22. SEIDELL, ENBERG,


QUANTIFICATION

OF

pometry and computed tomography. Am. J. Clin. Nutr. 45: 7-13, 1987. 23. SHUMAN, W. P., L. L. NEWELL MORRIS, D. L. LEONETTI, P. W. WAHL, V. M. MOCERI, A. A. Moss, AND W. FUJIMOTO. Abnormal body fat distribution detected by computed tomography in diebetic men. Invest. Radiol. 21: 483-487, 1986. 24. SPARROW, D., G. A. BORKAN, S. G. GERZOF, C. WISNIEWSKI, AND C. K. SILBERT. Relationship of fat distribution to glucose tolerance: results of computed tomography in male participants of the normative aging study. Diabetes 35: 411-415, 1986.

ADIPOSE

TISSUE

BY

MRI

795

25. STATEN, M. A., W. G. TOTTY, AND W. M. KOHRT. Measurement of fat distribution by magnetic resonance imaging. Invest. Radial. 24: 345-349,1989. 26. TOKUNAGA, K., Y. MATSUZAWA, K. ISHIKAWA, AND S. TARUI. A novel technique for the determination of body fat by computed tomography. Int. J. Obes. 7: 437-445, 1983. 27. WEINSER, R. L., D. J. NORRIS, R. BIRCH, R. S. BERNSTEIN, J. WANG, M.-U. YANG, R. N. PIERSON, JR., AND T. B. VAN ITALLIE. The relative contribution of body fat and fat pattern to blood pressure level. Hypertension 7: 578-585, 1985.


Prediction of whole-body fat percentage and visceral adipose tissue mass from five anthropometric variables.

Segmentation and quantification of adipose tissue by magnetic resonance imaging.

Epicardial Adipose Tissue Is Nonlinearly Related to Anthropometric Measures and Subcutaneous Adipose Tissue.

Brown adipose tissue quantification in human neonates using water-fat separated MRI.

Different imaging modalities in quantification of epicardial adipose tissue thickness.

Anthropometric variables and risk of breast cancer.

[Relationship between anthropometric variables and muscle dysmorphia in gymnasts in the province of Alicante].

Derivation and validation of simple anthropometric equations to predict adipose tissue mass and total fat mass with MRI as the reference method.

Relationship between fatty acids of adipose tissue and plasma cholesterol.

Intramuscular adipose tissue determined by T1-weighted MRI at 3T primarily reflects extramyocellular lipids.

Predicting visceral adipose tissue by MRI using DXA and anthropometry in adolescents and young adults.

Relationship between uptake of norepinephrine by hypothalamic homogenates and the activity of brown adipose tissue.

Epicardial adipose tissue: relationship between measurement location and metabolic syndrome.

Qualitative characterization of adipose tissue by MDCT.

Relationship between grip, pinch strengths and anthropometric variables, types of pitch throwing among Japanese high school baseball pitchers.

Essential fatty acid composition of human colostrum triglycerides: its relationship with adipose tissue composition.

Interrelationships between anthropometric variables and overweight in childhood and adolescence.

C3 and C4 are strongly related to adipose tissue variables and cardiovascular risk factors.

Using fingerprint measures to predict other anthropometric variables.

Correlations between Areas, Volumes or Body Fat and Anthropometric Variables.

Relationship between epicardial adipose tissue thickness and vitamin D in patients with metabolic syndrome.

Best single-slice location to measure visceral adipose tissue on paediatric CT scans and the relationship between anthropometric measurements, gender and VAT volume in children.

Quantification of intrapancreatic fat in type 2 diabetes by MRI.

The association of total sialic acid and malondialdehyde levels with metabolic and anthropometric variables in obesity.