Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print January 23, 2007, 10.1148/radiol.2423060111
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2423060111v1
242/3/846    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Siegel, M. J.
Right arrow Articles by White, N. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Siegel, M. J.
Right arrow Articles by White, N. H.
(Radiology 2007;242:846-856.)
© RSNA, 2007

Pediatric Imaging

Total and Intraabdominal Fat Distribution in Preadolescents and Adolescents: Measurement with MR Imaging1

Marilyn J. Siegel, MD, Charles F. Hildebolt, DDS, PhD, Kyongtae T. Bae, MD, Cheng Hong, PhD, MD and Neil H. White, MD, CDE

1 From the Mallinckrodt Institute of Radiology, Washington University School of Medicine, 510 S Kingshighway Blvd, St Louis, MO 63110 (M.J.S., C.F.H., K.T.B., C.H.); and Department of Pediatrics, Washington University School of Medicine, St Louis, Mo (N.H.W.). Received January 18, 2006; revision requested March 21; revision received April 18; accepted May 17; final version accepted July 13. Supported in part by Washington University General Clinical Research Committee (United States Public Health Service grant MO1 RR00036) and a grant from the Society for Pediatric Radiology. Address correspondence to M.J.S. (e-mail: siegelm{at}mir.wustl.edu).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 References
 
Purpose: To prospectively correlate single- and multisection magnetic resonance (MR) imaging measurements with clinical measurements for assessment of abdominal adipose tissue volumes in healthy (control subjects), overweight, and diabetic overweight preadolescents and adolescents.

Materials and Methods: The study was approved by the institutional internal review board and was HIPAA compliant. Informed consent was obtained from parents, and assent was obtained from control subjects and patients. Thirty total study subjects (20 male, 10 female; age range, 10–18 years; mean, 14.5 years) underwent MR imaging, anthropometric measurement, and dual x-ray absorptiometry (DXA). A computer-assisted software program was used to quantify subcutaneous, visceral, and total abdominal adipose tissue volumes. Single-section measurements at disk space L4-L5 and whole-abdominal multisection measurements were compared, and each method was tested for correlations with anthropometric and DXA measurements with Spearman {rho} and Pearson correlation (r) coefficients. Single- and multisection image analyses required 5 and 25 minutes per subject, respectively.

Results: There was a high degree of correlation between single- and multisection MR imaging methods for measurement of subcutaneous (r = 0.97), visceral (r = 0.96), and total abdominal fat (r = 0.97). MR imaging fat measurements strongly correlated with anthropometric measurements ({rho} correlation range, 0.81–0.96; P ≤ .02), with overlapping 95% confidence intervals (CIs) for single- and multisection MR imaging correlations. MR imaging percentage of intraabdominal fat measurements (mean, 23%; 95% CI: 17%, 29%) highly correlated with DXA abdominal fat measurements (mean, 26%; 95% CI: 21%, 31%). Significant differences were found among healthy subjects, overweight patients, and diabetic overweight patients for total fat volumes (P < .001) but not for fat distribution patterns.

Conclusion: Single- and multisection MR imaging measurements for the quantitative assessment of abdominal adipose tissue strongly correlate with clinical and DXA fat measurements.

© RSNA, 2007


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 References
 
The prevalence of overweight children, defined by age and sex and a body mass index (BMI) of more than the 95th percentile on the Centers for Disease Control growth charts, has tripled in the United States in the past 3 decades (1–3). Results of the most recent National Health and Nutrition Examination Survey estimate that 20.6% of children 2–5 years of age, 30.3% of children 6–11 years of age, and 30.4% of adolescents and young adults 12–19 years of age are overweight or at risk for becoming overweight (2). There also is growing evidence that obese children are at greater risk for several metabolic disturbances, including glucose intolerance, insulin resistance, hyperlipidemia, the metabolic syndrome, and diabetes mellitus, as well as for cardiovascular disease and nonalcoholic fatty liver disease (4–8). These metabolic risks appear to be more closely associated with intraabdominal adipose tissue accumulation than with subcutaneous adipose tissue accumulation. Thus, it may be of clinical importance to be able to reliably measure abdominal adipose tissue.

Several clinical methods, including anthropometry and dual x-ray absorptiometry (DXA), have been used as surrogates for estimating body fat, but these can be imprecise and do not allow the evaluation of visceral fat content (9,10). Computed tomography (CT) more accurately depicts the amount of visceral fat, but it requires radiation exposure, which makes it difficult to justify for pediatric studies, especially when repeated longitudinal determinations are necessary. Only a few reports with MR to measure body fat in children have been published, and most have included healthy or obese subjects alone, used time-intensive T1-weighted magnetic resonance (MR) sequences, or have measured abdominal adipose tissue with a single-image acquisition (8–11). The purpose of our study was to prospectively correlate single- and multisection MR imaging measurements with clinical measurements for the assessment of abdominal adipose tissue volumes in healthy (control subjects), overweight, and diabetic overweight preadolescents and adolescents.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 References
 
Participants
This study was approved by the institutional internal review board and was Health Insurance Portability and Accountability Act compliant. Informed consent was obtained from parents, and assent was obtained from control subjects and patients. We included 30 patients and control subjects (20 male, 10 female; age range, 10–18 years; mean age, 14.5 years) who were recruited through the in-house diabetes clinic or through in-hospital advertising. Nine nondiabetic overweight patients and 10 type 2 diabetic overweight patients, as defined with a BMI of more than the 95th percentile, and 11 nondiabetic patients of a normal weight (control subjects) were enrolled. We included both obese and normal-weight participants to examine a wide spectrum of body fat levels and distributions. Exclusion criteria for MR imaging included implanted metallic devices, such as intracerebral aneurysm clips, implanted cardiac pacemakers, prosthetic ear devices, or epicardial wires.

Clinical Measurements
Anthropometric parameters, including weight, height, waist circumference (measured in the orthostatic position at the midpoint between the lateral iliac crest and the lowest rib), abdominal height (measured with the control subject or patient supine on an examination table by using a commercial abdominal caliper to determine the sagittal abdominal diameter from the back to a point on the abdomen midway between the left and right iliac crests), and BMI (weight in kilograms divided by height in square meters), were obtained. Anthropometric parameters were obtained by a research nurse who had previously been trained and certified to perform these procedures as part of involvement in a large multicenter clinical trial. Anthropometric measurements were obtained on the same day DXA scanning was performed, except in seven study subjects who did not undergo DXA. In these cases, anthropometric measurements were obtained on the same day MR imaging was performed.

DXA (QDR 2000/W, Hologic, Waltham, Mass; or Lunar DPX, United Medical Technologies, Fort Myers, Fla) with dual radiation and two peak kilovoltages of 70 and 140 keV was performed within a 2-week time period. Fifteen study subjects underwent both DXA and MR imaging within a 3-day period. (Eleven underwent both on the same day.) All studies performed on the same day or within 3 days were in the order of MR imaging first, then DXA. Because of scheduling and equipment maintenance problems, eight study subjects underwent MR imaging 3 weeks to 3 months before DXA and one underwent DXA 2 months before MR imaging. Truncal fat mass and soft-tissue fat free mass representing lean mass were calculated by the research nurse with standard software that was part of the DXA system. From these data, the percentage of fat in the trunk was calculated. Our precision for DXA fat measurement was 4.5%.

MR Imaging
MR imaging of the abdomen was performed with a 1.5-T magnet (Siemens Medical Systems, Iselin, NJ) by using a phased-array body coil. Study subjects were imaged in a supine position with both arms parallel to their body. Whole-volume coverage of the abdomen was obtained by using a true fast imaging with steady-state precession (FISP) sequence during breath holding (repetition time msec/echo time msec, 4.3/2.1; flip angle, 58°; matrix, 256 x 150; acquisition time, 14 seconds; section thickness, 10 mm; no gap). The field of view, which varied with the size of the patient or control subject, was typically 250 x 360 mm. The number of images in each breath-hold set was 15, and two neighboring sets were acquired to cover the abdomen above the diaphragm to below the iliac crest. In addition, a spoiled gradient-echo T1-weighted sequence (391/3.5; flip angle, 40°; matrix, 256 x 160; section thickness, 10 mm; no gap; acquisition time, 20 seconds) was acquired for the same anatomic coverage. This image set was used to complement the true FISP image set to resolve any anatomic ambiguity and assist the image analysis. Total duration for the five breath-hold scans, including the scout image scan, was about 5 minutes.

Image analysis was performed with software developed at our institution for volumetric measurement of muscle and fat in the abdomen and thigh. Images were transferred electronically to a workstation for assessment of subcutaneous and intraabdominal fat and total abdominal tissue volume. By using semiautomated image segmentation software implemented in the Analyze software system (Mayo Clinic Foundation, Biomedical Imaging Resource, Rochester, Minn), the fat region was segmented and its area was calculated for each image section. We used a region-based thresholding method (12,13) to segment the fat regions. This method is based on the property of our acquired MR images in which the signal intensities of fat pixels are higher than those of muscle and visceral organ pixels (Fig 1).


Figure 1
View larger version (38K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1: MR images in 10-year-old overweight girl. Upper left: Transverse true FISP abdominal MR image (4.3/2.1; flip angle, 58°; matrix, 256 x 150; section thickness, 10 mm) acquired as part of standard MR examination. From this image, a pixel was selected within subcutaneous or visceral fat region. Lower left: Subcutaneous fat region. Lower right: Visceral fat region. Fat regions were segmented separately with region-based thresholding segmentation method, and areas were measured. Upper right: Total abdominal fat was calculated (not directly measured) as the sum of measured subcutaneous fat and measured intraabdominal fat. We measured the entire abdomen to have a denominator for the computation of percentage of abdominal fat (total abdominal fat/entire abdomen), which was compared with DXA measurement.

 
A seed pixel was selected within the subcutaneous or visceral fat region by a single research radiologist (C.H.) with 5 years of experience analyzing MR images for volume measurement. The computer program then automatically connected and grouped the pixels of similar signal intensities. The degree of region connectivity and growth was adjusted by controlling the level of thresholds displayed in the program. Any erroneous extension or connection to the fat regions was manually separated. The visceral (mesenteric and omental) and subcutaneous fat regions were segmented and computed separately. Total abdominal fat was calculated as the sum of the measured subcutaneous and visceral fat volumes. In addition, the cross-section of the entire abdomen was segmented from the surrounding background air by using the same region-based thresholding method but with different threshold values. Measurement of the entire abdomen served as the denominator for computation of percentage of abdominal fat (ie, total abdominal fat/entire abdomen).

Both multi- and single-section methods were used. For the multisection method, the volume (cubic centimeters) of intraabdominal fat was computed by summing the area measurements from contiguous 1-cm sections. The ratio of the volume of abdominal fat to the total tissue volume of the abdominal region was used to determine the percentage of abdominal fat. For the single-section method, the above procedures were performed only at the level of L4-L5. MR imaging measurements were performed by a single experienced research radiologist (C.H.) who is responsible for all MR segmentation analyses at our institution. This radiologist was blinded to the classification of the participant (healthy control subject, overweight patient, diabetic overweight patient). The single-level image analysis was performed in 5 minutes, whereas the multilevel image analysis took approximately 25 minutes per patient or control subject. To evaluate measurement variability, the total tissue volume of the abdomen was measured twice for all participants (C.H.).

Statistical Analysis
We compared fat measurements from L4-L5 (single-section method) with the more time-intensive fat measurements from the entire subcutaneous, visceral, and abdominal areas (multisection method) to determine the extent to which MR measurements of fat correlated with anthropometric and DXA measurements. As part of these determinations, normal curves were fitted to the data distributions and normality was tested with the Shapiro-Wilk W test. For the anthropometric measurements, the only data set that was normally distributed was abdominal height (P > .12). Because of this, we calculated Spearman rank correlation {rho} coefficients in addition to Pearson correlation coefficients (r). For {rho} correlations, we calculated 95% confidence intervals (CIs). Regression plots of selected variables were used to demonstrate association. We also calculated {rho} correlations among patient groups (control subjects, overweight patients, and overweight diabetic patients) and among anthropometric measurements (waist circumference, abdominal height, and BMI).

We additionally determined the extent to which there were significant differences in MR measurements among patient groups (overweight patients, diabetic overweight patients, control subjects). Group determination was on the basis of BMI scores. For comparison of groups, the equality of variances was tested with the O'Brien, Brown-Forsythe, Levene, and Bartlett test. Groups were tested for statistically significant differences with the Welch analysis of variance and Kruskal-Wallis tests. Post hoc testing was performed with the Tukey-Kramer honestly significant difference test if data were normally distributed and the Wilcoxon test if data were not normally distributed. For normally distributed L4-L5 data, power analyses were performed to determine the smallest difference between means that could be detected. To illustrate the determination of group membership on the basis of MR measurement of fat, a probability plot was created by means of nominal logistic regression analysis. Our final assessment was to compare percentage of fat determined with DXA measurements of the trunk with percentage of fat determined with MR imaging measurements of the abdominal area. Both of these percentages were normally distributed (P ≥ .26). Ninety-five percent confidence intervals were calculated for these percentages, and percentages were compared by means of regression analysis and a paired t test. A power analysis was also performed with these data.

To determine the precision of MR measurements, repeat measurements were performed for the total volume of the abdomen. As recommended for determination of measurement precision (14), we calculated root-mean-square standard deviation, root-mean-square coefficient of variation, and root-mean-square percentage coefficient of variation. For these calculations, a standard deviation, coefficient of variation, and percentage coefficient of variation were determined for each patient. From these values, the root-mean-square standard deviation, root-mean-square coefficient of variation, and root-mean-square percentage coefficient of variation were calculated. Arithmetic means for standard deviation, coefficient of variation, and percentage coefficient of variation tend to cause underestimation of Gaussian error; therefore, the root-mean-square values are preferred in precision studies (15). Calculations of arithmetic means for standard deviation, coefficient of variation, and percentage coefficient of variation also result in smaller values than do calculations of root mean squares, and this would result in the precision appearing better than it actually is.

For all tests, {alpha} was set at .05. A P value of less than .05 was considered to indicate a significant difference. Statistical testing was performed with software (JMP, SAS Institute, Cary, NC; and MedCalc Statistics for Biomedical Research, MedCalc Software, Mariakerke, Belgium). Power analyses were performed with another software (Power and Precision; Biostat, Englewood, NJ).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 ADVANCES IN KNOWLEDGE
 References
 
DXA measurements were missing in seven study subjects, abdominal height measurements were missing in three study subjects, waist circumference measurement was missing in one study subject, and MR imaging measurements were missing in two study subjects because of subject failure to return for clinic visits (Table 1).


View this table:
[in this window]
[in a new window]

  		Table 1. Clinical Measurement and DXA-determined Percentage of Fat in Trunk of 30 Subjects

 
Correlation of MR Measurements with Clinical Measurements
Correlation coefficients between MR measurements and anthropometric measurements were calculated for 18 comparisons (Table 2). For the non–normally distributed data, {rho} was equal to or higher than r for 15 (83%) of 18 comparisons. Overall, there was a high degree of correlation ({rho} range, 0.81–0.96; P ≤ .02) between MR measurements and anthropometric measurements (waist circumference, abdominal height, and BMI). Nine of the comparisons were for multisection MR imaging (subcutaneous, visceral, abdominal), and nine were for single-section MR imaging (subcutaneous fat at L4-L5, visceral fat at L4-L5, abdominal fat at L4-L5). The range of the {rho} correlations for multisection MR imaging was 0.86–0.94 and that for single-section MR imaging was 0.81–0.96, with 95% CIs overlapping for corresponding correlations for the two methods (Table 2). Among anthropometric measurements, the highest correlations were for waist circumference (n = 6; {rho} range, 0.92–0.96), with the correlations for BMI (n = 6; {rho} range, 0.81–0.91) and abdominal height (n = 6; {rho} range, 0.82–0.95) being similar to each other. The correlations were similar for visceral ({rho} range, 0.82–0.92), subcutaneous ({rho} range, 0.81–0.96), and abdominal ({rho} range, 0.89–0.96) measurements.


View this table:
[in this window]
[in a new window]

  		Table 2. Pearson and Spearman Correlations between MR Measurements and Anthropometric Measurements

 
Within the control subjects, the range of {rho} was 0.18–0.90. The ranges within overweight subjects and overweight diabetic subjects were 0.50–1.00 and 0.33–0.88 (Table 3), respectively. Spearman {rho} correlations among anthropometric measurements (waist circumference, abdominal height, and BMI) ranged from 0.81 to 0.95 for control subjects, 0.86 to 0.96 for overweight subjects, and 0.83 to 0.88 for overweight diabetic subjects. When the subjects were combined, the three correlations were all 0.97.


View this table:
[in this window]
[in a new window]

  		Table 3. Spearman Correlation between MR Measurements and Anthropometric Measurements for Participant Groups

 
Results of regressions of BMI versus total abdominal fat and total fat at the level of L4-L5, with markers indicating group membership (control subjects, overweight patients, or overweight diabetic patients), were similar, as were the correlations ({rho} = 0.89 and 0.91, respectively) (Fig 2). Similarly, results of regression of the multisection data versus the single-section data for subcutaneous, visceral, and abdominal measurements resulted in high correlations ({rho} = 0.97, 0.96, and 0.98, respectively) (Fig 3).


Figure 2A
View larger version (13K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2a: Graphs of regressions of (a) BMI versus total abdominal fat (a) and (b) BMI versus total fat at level of L4-L5 (t). Units = cubic centimeters. Regression formula in a is BMI = –18.87 + 0.0018a (r = 0.90, {rho} = 0.89, P < .01). Regression formula in b is BMI = –17.11 + 0.0367t (r = 0.92, {rho} = 0.91). In b, bands around regression line are 95% CIs. Total number of participants is 28 instead of 30 because two subjects did not undergo MR imaging. x = control subjects, {circ} = overweight patients, bullet = overweight diabetic patients.

 

Figure 2B
View larger version (15K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2b: Graphs of regressions of (a) BMI versus total abdominal fat (a) and (b) BMI versus total fat at level of L4-L5 (t). Units = cubic centimeters. Regression formula in a is BMI = –18.87 + 0.0018a (r = 0.90, {rho} = 0.89, P < .01). Regression formula in b is BMI = –17.11 + 0.0367t (r = 0.92, {rho} = 0.91). In b, bands around regression line are 95% CIs. Total number of participants is 28 instead of 30 because two subjects did not undergo MR imaging. x = control subjects, {circ} = overweight patients, bullet = overweight diabetic patients.

 

Figure 3A
View larger version (16K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3a: Graphs of regressions of (a) subcutaneous (s), (b) visceral (v), and (c) abdominal fat (a) measurements from entire abdomen versus corresponding L4-L5 measurements (x). Units = cubic centimeters. Bands around regression lines are 95% CIs. Regression formula in a is s = –589 + 19x (r = 0.97, {rho} = 0.97, P < .01). Regression formula in b is v = –131 + 23x (r = 0.96, {rho} = 0.96, P < .01). Regression formula in c is a = –745 + 19x (r = 0.97; {rho} = 0.98, P < .01). Total number of participants is 28 instead of 30 because two subjects did not undergo MR imaging.

 

Figure 3B
View larger version (13K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3b: Graphs of regressions of (a) subcutaneous (s), (b) visceral (v), and (c) abdominal fat (a) measurements from entire abdomen versus corresponding L4-L5 measurements (x). Units = cubic centimeters. Bands around regression lines are 95% CIs. Regression formula in a is s = –589 + 19x (r = 0.97, {rho} = 0.97, P < .01). Regression formula in b is v = –131 + 23x (r = 0.96, {rho} = 0.96, P < .01). Regression formula in c is a = –745 + 19x (r = 0.97; {rho} = 0.98, P < .01). Total number of participants is 28 instead of 30 because two subjects did not undergo MR imaging.

 

Figure 3C
View larger version (16K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3c: Graphs of regressions of (a) subcutaneous (s), (b) visceral (v), and (c) abdominal fat (a) measurements from entire abdomen versus corresponding L4-L5 measurements (x). Units = cubic centimeters. Bands around regression lines are 95% CIs. Regression formula in a is s = –589 + 19x (r = 0.97, {rho} = 0.97, P < .01). Regression formula in b is v = –131 + 23x (r = 0.96, {rho} = 0.96, P < .01). Regression formula in c is a = –745 + 19x (r = 0.97; {rho} = 0.98, P < .01). Total number of participants is 28 instead of 30 because two subjects did not undergo MR imaging.

 
With regard to measurement precision, the mean for the first set of MR measurements for the total volume of the abdomen was 12 728 cm3 (95% CI: 10 171, 15 285 cm3). The corresponding values for the repeat measurements were 12 993 cm3 (95% CI: 10 238, 15 748 cm3). The root-mean-square standard deviation was 477 cm3, the root-mean-square coefficient of variation was 0.038, and the root-mean-square percentage coefficient of variation was 3.79%.

Differences among Patient Groups
Regarding adipose tissue distribution for the control subjects, overweight subjects, and diabetic overweight subjects (Table 4), for all comparisons, variances were demonstrated to be unequal: One or more of the tests of equality of variance indicated a statistically significant difference (P < .05). Because of this, Welch analysis of variance tests were used. In addition, nine of 18 data sets contained nonnormal data distributions (Table 4). Because of this, nonparametric Kruskal-Wallis tests were performed in addition to Welch analysis of variance tests.


View this table:
[in this window]
[in a new window]

  		Table 4. Differences in Fat Volumes among Patient Groups

 
Significant differences were found in the amount of adipose tissue among overweight diabetic, overweight, and control subjects: Overweight diabetic subjects had the most fat, and control subjects had the least fat (P < .001) (Fig 4). Differences in amount of fat were least pronounced for visceral measurements, for which significant differences existed between control subjects and overweight diabetic patients and between control subjects and overweight patients but not between diabetic overweight and overweight patients (P = .052) (Table 3). The probability that study subjects would have diabetes increased with increasing fat volume on the basis of MR measurements (Fig 5).


Figure 4
View larger version (12K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4: Ninety-five percent confidence interval mean diamond plots for subcutaneous fat for three patient groups. Horizontal line = grand mean. Heights of diamonds represent 95% CIs, and widths of diamonds are proportional to sample sizes. If overlap line of one diamond is closer to another diamond's mean than is that diamond's overlap line, there is no difference between the groups. Point placement is shifted to avoid overlap.

 

Figure 5
View larger version (10K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5: Cumulative probability plot of association between MR total fat measurements (cubic centimeters) for L4-L5 and patient group membership. Lines of fit partition whole probability into response categories. Probability of a participant being a control subject can be read directly from y axis. Probability of patient having diabetes is distance between curves, as read on y axis. Probability of patient being overweight is the distance on y axis from second curve to top of graph (1 minus the axis reading). For example, if MR measurement is 220 cm3, probability of the patient being a control subject would be about 5%, probability of the patient having diabetes would be about 20%, and probability of the patient being overweight (without diabetes) would be about 70%.

 
Power analyses were performed for subcutaneous L4-L5 data and for abdominal L4-L5 data (which were normally distributed) (Table 4) to determine the smallest differences between means that could be detected between overweight diabetic subjects and control subjects. For subcutaneous L4-L5 data, the smallest difference that could be detected was 88 cm3. For abdomen L4-L5 data, the smallest difference that could be detected was 96 cm3.

MR and DXA Measured Percentage of Fat
The mean percentage of truncal fat, as determined with MR imaging, was 23% (95% CI: 17%, 29%), and the mean percentage of fat for the entire abdominal area, as determined with DXA, was 26% (95% CI: 21%, 31%). These two sets of measurements were highly correlated (r = 0.96; 95% CI: 0.90, 0.98). The mean difference between the percentages was 2.4% (95% CI: 0.4%, 4.3%), which was statistically significant (P = .02). The statistical power of this test was 81%. The scatterplots of the percentage of fat at DXA versus the percentage of fat at MR imaging were similar (Fig 6).


Figure 6
View larger version (13K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 6: Graph of regression of percentage of fat determined with DXA for trunk (d) versus the percentage of fat determined with MR imaging for entire abdominal area (m). Bands around regression lines are 95% CIs. Regression formula is d = –78.87 + 0.767m (r = 0.96 [95% CI: 0.90, 0.98], P < .01).

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
ADVANCES IN KNOWLEDGE
 References
 
In our study, waist circumference, abdominal height, and BMI significantly correlated with adipose tissue distribution, with the range of {rho} correlations being 0.81–0.96. This high degree of correlation may be partially attributable to the differences among our three groups (control subjects, overweight patients, and overweight diabetic patients), as indicated by the relatively lower within-group correlations for our data. Our results are, nevertheless, similar to those of prior investigators who have shown significant correlation between total body fat and BMI (9) and waist circumference (11). Despite this correlation, determination of body fat content by using these noninvasive methods can be difficult (16). These widely used clinical indexes are subject to inter- and intraobserver error and poor reproducibility and give no information on the regional distribution of adipose tissue. In addition, there can be variations among ethnic groups (17).

More specialized techniques, such as bioimpedance analysis and DXA, have been used to estimate the percentage of lean and fat mass (18). In our study, the difference between percentage of fat determinations with MR imaging and with DXA was only 2.4%. These data are in agreement with those of other investigators and confirm the usefulness of this MR technique (18). Although DXA is accurate, this technique requires exposure to ionizing radiation. Furthermore, DXA gives no breakdown of subcutaneous and intraabdominal adipose tissue volumes.

A variety of cross-sectional imaging techniques have been used to estimate body fat content. Ultrasonography is quick and readily available, but reproducibility is variable and the total volume of visceral fat cannot be measured (19–22). CT also is fast and available and can help give insight about the regional distribution of fat. Results of studies in adults have reported an association between excess intraabdominal fat, as determined with CT, and metabolic abnormalities, including glucose intolerance, insulin resistance, diabetes mellitus, hyperlipidemia, and the metabolic syndrome (20,23,24). However, radiation exposure limits the usefulness of CT in children, especially if serial measurements are to be performed as part of intervention studies and clinical trials. MR imaging is becoming more widely used in the evaluation of abdominal fat composition. Results of previous investigations (25,26) in adults have shown the accuracy of MR imaging for abdominal fat analysis compared with that of clinical measures, CT scanning, and cadaveric studies.

In the pediatric population, there is a paucity of data on the use of MR imaging methods for fat analysis. A review of the literature shows that most of these studies have been performed with time-intensive T1-weighted sequences, single-section or multisection acquisitions alone, or manual measurements (9–11,27). In our study, we used a fast MR sequence and semiautomated measurements to characterize the quantity and distribution of abdominal adipose tissue. Our data clearly show that automated analysis of a single image at the level of L4-L5 can be used to characterize subcutaneous, intraabdominal, and total abdominal fat in subjects and that the results are nearly equivalent to those from multisection MR volumes. Because a single-level technique is rapid, it has the potential to reduce imaging time, making MR imaging a feasible study for younger patients. With a 1.5-T magnet, it is possible to perform single-level breath-hold T1-weighted fast imaging within a few seconds.

By using MR imaging measurements, we found that normal-weight, overweight, and diabetic overweight subjects had similar fat distribution patterns, with a predominance of subcutaneous adipose tissue. However, we found a significant difference in the total amount of adipose tissue among the overweight patients, diabetic overweight patients, and control subjects. Overweight diabetic subjects had the most fat, and control subjects had the least fat. We also found significant differences in intraabdominal fat volume between control subjects and overweight diabetic subjects and between control subjects and overweight subjects but not between overweight diabetic and overweight subjects. The overall fat volume was highest in diabetic overweight patients, which can be explained by the more severe obesity in diabetic overweight subjects evaluated as part of this study. Increased intraabdominal fat is known to be associated with metabolic disorders such as diabetes, although we did not find this correlation; this suggests that the diabetes in our study population was an independent event and not the result of increased intraabdominal fat. However, further investigation with larger populations will be required to determine the relationship between diabetes and obesity in terms of intraabdominal fat accumulation in adolescent patients.

A limitation of our study was that we had no absolute reference standard for documenting the regional distribution of abdominal fat. We compared our investigational method for fat analysis with clinical parameters, which are imperfect reference standards. CT might have been a better reference standard examination, but this would be difficult to justify given the necessary ionizing radiation exposure. Another limitation was that a single observer performed the semiautomated analysis of the MR images. This observer is, however, experienced, as indicated by the relatively high level of precision for this preliminary study (root-mean-square percentage coefficient of variance, 3.79%). In addition, this observer was blinded to subject classification.

In conclusion, we report a method for imaging abdominal fat composition in subjects by using a fast T1-weighted MR sequence at a single level. This technique for adipose tissue estimation has the potential to be useful in many clinical and research applications, including planning patient treatment, monitoring interventions, and implementing multicenter clinical trials or epidemiologic studies.


   ADVANCES IN KNOWLEDGE
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
References
 


   ACKNOWLEDGMENTS
 
We acknowledge Michelle Sadler, who performed most of the recruiting and clinical measurements for this project, and Laura Gallagher, RT, who monitored the MR imaging studies and quality assessment.


   FOOTNOTES
 

Abbreviations: BMI = body mass index • CI = confidence interval • DXA = dual x-ray absorptiometry • FISP = fast imaging with steady-state precession

Authors stated no financial relationship to disclose.

Author contributions: Guarantor of integrity of entire study, M.J.S.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, M.J.S., K.T.B., C.H.; clinical studies, K.T.B., C.H., N.H.W.; statistical analysis, C.F.H.; and manuscript editing, M.J.S., C.F.H., K.T.B., N.H.W.


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 ADVANCES IN KNOWLEDGE
 References
 

  1. Fowler-Brown A, Kahwati LC. Prevention and treatment of overweight in children and adolescents. Am Fam Physician 2004;69:2591–2598.[Medline]
  2. Ogden CL, Flegal KM, Carroll MD, Johnson CL. Prevalence and trends in overweight among US: children and adolescents, 1999–2000. JAMA 2002;288:1728–1732.[Abstract/Free Full Text]
  3. Prevalence of overweight among children and adolescents: United States, 1999–2000. Centers for Disease Control and Prevention, National Center of Health Statistics. http:/www.cdc.gov/nchs/products/pubs/pubd/hestats/overwght99.htm. Accessed April 7, 2006.
  4. Eckel RH, Grundy SM, Zimmet PZ. The metabolic syndrome. Lancet 2005;365:1415–1428.[CrossRef][Medline]
  5. Schwimmer JB, Deutsch R, Rauch JB, et al. Obesity, insulin resistance, and other clinicopathological correlates of pediatric nonalcoholic fatty liver disease. J Pediatr 2003;143:500–505.[CrossRef][Medline]
  6. Must A, Jacques PF, Dallal GE, Bajema CJ, Dietz WH. Long-term morbidity and mortality of overweight adolescents: a follow-up of the Harvard Growth Study of 1922 to 1935. N Engl J Med 1992;327:1350–1355.[Abstract]
  7. Goran MI, Gower BA. Relation between visceral fat and disease risk in children and adolescents. Am J Clin Nutr 1999;70:149S–156S.[Medline]
  8. Owens S, Gutin B, Ferguson M, Allison J, Karp W, Le NA. Visceral adipose tissue and cardiovascular risk factors in obese children. J Pediatr 1998;133:41–45.[CrossRef][Medline]
  9. Chan YL, Leung SS, Lam WW, Peng XH, Metreweli C. Body fat estimation in children by magnetic resonance imaging, bioelectrical impedance, skinfold and body mass index: a pilot study. J Paediatr Child Health 1998;34:22–28.[CrossRef][Medline]
  10. de Ridder CM, de Boer RW, Seidell JC, et al. Body fat distribution in pubertal girls quantified by magnetic resonance imaging. Int J Obes 1992;16:443–449.[Medline]
  11. Fox K, Peters D, Armstrong N, Sharpe P, Bell M. Abdominal fat deposition in 11-year-old children. Int J Obes 1993;17:11–16.[Medline]
  12. Chapman AB, Guay-Woodford L, Grantham JJ, et al. Renal structure in early autosomal-dominant polycystic kidney disease (ADPKD): the Consortium for Radiologic Imaging Studies of Polycystic Kidney Disease (CRISP) cohort. Kidney Int 2003;64:1035–1045.[CrossRef][Medline]
  13. Bae KT, Zhu F, Chapman AB, et al. MR imaging evaluation of hepatic cysts in early ADPKD: CRISP cohort. Clin J Am Soc Nephrol 2006;1:64–69.
  14. Gluer CC, Blake G, Lu Y, Blunt BA, Jergas M, Genant HK. Accurate assessment of precision errors: how to measure the reproducibility of bone densitometry techniques. Osteoporos Int 1995;5:262–270.[CrossRef][Medline]
  15. Bonnick SL. Bone densitometry in clinical practice: application and interpretation. Totowa, NJ: Humana, 1995.
  16. Widhalm K, Schonegger K, Huemer C, Auterith A. Does the BMI reflect body fat in obese children and adolescents? a study using the TOBEC method. Int J Obes 2001;25:279–285.[CrossRef][Medline]
  17. Deurenberg P, Deurenberg-Yap M, Guricci S. Asians are different from Caucasians and from each other in their body mass index/body fat per cent relationship. Obes Rev 2002;3:141–146.[CrossRef][Medline]
  18. Bertin E, Marcus C, Ruiz JC, Eschard JP, Leutenegger M. Measurement of visceral adipose by DXA combined with anthropometry in obese humans. Int J Obes Relat Metab Disord 2000;24:263–270.[CrossRef][Medline]
  19. Armellini F, Zamboni M, Rigo L, et al. The contribution of sonography to the measurement of intra-abdominal fat. J Clin Ultrasound 1990;18:563–567.[Medline]
  20. Armellini F, Zamboni M, Robbi R, et al. Total and intraabdominal fat measurements by ultrasound and computerized tomography. Int J Obes 1993;17:209–214.[Medline]
  21. Ribeiro-Filho FF, Faria AN, Azjen S, Zanella MT, Ferreira SR. Methods of estimation of visceral fat: advantages of ultrasonography. Obes Res 2003;11:1488–1494.[Medline]
  22. Tornaghi G, Raiteri R, Pozzato C, et al. Anthropometric or ultrasonic measurements in assessment of visceral fat? a comparative study. Int J Obes 1994;18:771–775.[Medline]
  23. Sjostrom L. A computed tomography based multicompartment body composition technique and anthropometric predictions of lean body mass, total and subcutaneous adipose tissue. Int J Obes 1991;15:19–30.
  24. Fujioka S, Matsuzawa Y, Tokunaga K, Tarui S. Contribution of intra-abdominal fat accumulation to the impairment of glucose and lipid metabolism in human obesity. Metabolism 1987;36:54–59.[CrossRef][Medline]
  25. Abate N, Burns D, Peshock RM, Garg A, Grundy SM. Estimation of adipose tissue mass by magnetic resonance imaging: validation against dissection in human cadavers. J Lipid Res 1994;35:1490–1496.[Abstract]
  26. Brennan DD, Whelan PF, Robinson K, et al. Rapid automated measurement of body fat distribution from whole-body MRI. AJR Am J Roentgenol 2005;185:418–423.[Abstract/Free Full Text]
  27. Brambilia P, Manzoni P, Sironi S, et al. Peripheral and abdominal adiposity in childhood obesity. Int J Obes 1994;18:795–800.[Medline]



This article has been cited by other articles:


Home page
 Am. J. Clin. Nutr.Home page
J. A Morrison, C. J Glueck, P. S Horn, G. B Schreiber, and P. Wang
Pre-teen insulin resistance predicts weight gain, impaired fasting glucose, and type 2 diabetes at age 18-19 y: a 10-y prospective study of black and white girls
Am. J. Clinical Nutrition, September 1, 2008; 88(3): 778 - 788.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
D. D. Brennan, P. F. Whelan, K. Robinson, O. Ghita, J. M. O'Brien, R. Sadleir, and S. J. Eustace
Reply
Am. J. Roentgenol., December 1, 2007; 189(6): W385 - W385.
[Full Text] [PDF]

Home page
 RadiologyHome page
G. Taylor
Good Fat, Bad Fat-Does Location Matter?
Radiology, March 1, 2007; 242(3): 645 - 646.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2423060111v1
242/3/846    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Siegel, M. J.
Right arrow Articles by White, N. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Siegel, M. J.
Right arrow Articles by White, N. H.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE