PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Int J Body Compos Res. Author manuscript; available in PMC 2010 May 12.
Published in final edited form as:
Int J Body Compos Res. 2009; 7(2): 67–72.
PMCID: PMC2868277
NIHMSID: NIHMS133793

Validation of quantitative magnetic resonance for the determination of body composition of mice

Abstract

Objective

The aim of this study was to assess the precision and accuracy of a quantitative magnetic resonance (QMR) instrument for measuring body composition in live, non-anesthetized mice.

Methods

Forty-eight mice of varying strains, ages and body weights (15.3 to 50.2g) were scanned three times each in the QMR instrument. Animals were killed and chemical carcass analysis performed for comparison. Precision was assessed as the coefficient of variation (CV) for the triplicate scans and accuracy was determined by comparing the first QMR data with the chemical analysis. Prediction equations were generated by linear regression analysis and used in a cross-validation study in which 26 mice were scanned once each, killed, and chemical carcass analysis performed.

Results

The mean CV was 1.58% for fat mass (FM) and 0.78% for lean-tissue mass (LTM). QMR significantly (P<0.01) overestimated FM (7.76±5.93 vs. 6.03±5.17g) and underestimated LTM (20.73±6.19 vs. 22.48±6.75g) when compared with chemical carcass analysis. A strong relationship between QMR and chemical data (r2=0.99 and r2=0.97 for fat and LTM respectively; P<0.0001) allowed for the generation of correction equations that were applied to QMR data in the cross-validation study. There was no significant difference between data predicted from QMR and chemical carcass data for FM and LTM (P=0.15 and 0.10 respectively).

Conclusion

The QMR instrument showed excellent precision and data was highly correlated with chemical carcass analysis. This combined with QMR's speed for whole animal analysis (95 seconds) make it a highly feasible and useful method for the determination of body composition in live, non-anesthetized mice.

Introduction

Animal models, such as mice, have offered valuable insight into the etiology and possible treatment of obesity in humans. However, the challenge has been to find reliable and valid methods of evaluating the body composition of these small animals. Chemical carcass analysis serves as the “gold standard,” however this procedure involves killing the animal and thus rules out repeated measurements. In addition, the process of chemical carcass analysis is lengthy and labor-intensive. Methods that provide quick, in-vivo measurements are imperative to tracking changes in body composition that occur over time. Several techniques are currently available such as dual-energy X-ray absorptiometry (DXA), computed tomography (CT), magnetic resonance imaging (MRI), isotope dilution, and total body electrical conductivity (TOBEC) (1).

DXA, CT, and MRI are non-invasive, non-destructive and relatively quick methods for determining body composition. DXA uses two X-ray beams to acquire measures of fat mass (FM), LTM (lean-tissue mass), and also measurements of bone mineral density (BMD) and bone mineral content (BMC), which few other techniques can provide (2). Scan times are approximately six minutes. DXA has gained widespread use and has been used to study body composition in small animals, however the precision and accuracy vary depending upon the instrument and software utilized (3-7). CT and MRI produce estimates of body composition at the tissue and organ level. Images of specific fat pads and organs are produced allowing for detailed measurements of the fat depots and the tissues that make up lean mass. CT has shown documented use in rats (8) and mice (9). MRI use for body composition has increased recently (10-12). Although CT and MRI have been successful in measuring body composition and body fat distribution, measurement and analysis times are lengthy and require highly skilled operators. One limiting factor of the above mentioned DXA, CT, and MRI methods is the requirement of anesthesia or sedation as animals must remain completely motionless during the scans. The deleterious side effects of anesthesia or sedation, which include decreased food intake, reduction in body temperature, and risk of mortality are well-known (12,13).

Two in-vivo methods that do not require anesthesia or sedation are isotope dilution and TOBEC. Isotope dilution provides an indirect measure of fat mass (FM) by determining fat-free mass (FFM) from the calculation of total body water (TBW). The key to this method is the assumption that approximately seventy-three percent of FFM is water; FFM is calculated as TBW/0.73 and FM is then estimated as the difference between body mass and FFM. The technique of total body electrical conductivity (TOBEC) measures FM indirectly by assessing lean mass based upon its electrical conductivity, and then obtaining FM via subtraction. Advantages of using TOBEC include being relatively inexpensive and also portable, however the method lacks the accuracy required to detect and track changes in an animal's body composition (14,15). It has also been indicated that this method may not be good for animals with small fat mass as small errors in lean mass can result in large errors in fat mass (1).

This study presents a technique that seeks to address the above mentioned issues by measuring and analyzing the body composition of live, non-anesthetized mice in a short amount of time using quantitative magnetic resonance (QMR). The new QMR instrument, the EchoMRI™ 3-in-1 composition analyzer (Echo Medical Systems, Houston, TX) was evaluated for both precision and accuracy. QMR is a technique that relies on the principles of nuclear magnetic resonance (NMR). NMR uses radio waves to manipulate the spins of the nuclei of atoms to determine information about the molecules being tested (16). Any atom with an odd number mass can be used, such as Hydrogen, Carbon-13, Oxygen-17, and Phosphorus-31 as they produce a magnetic moment when their atoms spin. The QMR instrument evaluated in this study uses Hydrogen (Proton)-NMR principles. Protons within different tissues respond differently to disturbances. A magnet within the analyzer is used to align the protons within a sample's tissues which are then hit with radio waves. This causes a temporary disturbance in the magnetic moments (spin states) of the protons. As the protons return to their original positions (relax), they produce peaks on a spectrum whose positions and intensities (amplitudes) are measured and thus reflect their chemical environment (fat or lean). According to Taicher et al. (16), fat produces greater peak amplitude and has a greater rate of relaxation than lean because its hydrogen (proton) density is greater (~40% more hydrogen per unit) (16). Recently Tinsley et al. tested a similar, but different QMR instrument (17), however the validation focused on fat mass and did not give the details for the validation of lean mass. Therefore, the aim of this study was to determination the precision and accuracy of the Echo Medical QMR system for both fat and lean-tissue mass.

Methods

In vivo precision and accuracy

To evaluate the precision and accuracy of QMR (EchoMRI™ 3-in-1; Echo Medical Systems, Houston, TX) for determining fat and lean-tissue mass in vivo, forty-eight mice (24 male and 24 female) were tested. To ensure a range of fatness and leanness, mice of varying strains ranging in weight from 15.3 to 50.2 grams and ranging in age from three to eighteen months were selected. Animals were fasted overnight to ensure that the gastrointestinal tract of the mice were relatively empty. All mice were weighed before scanning and then placed into the holding tube and restricted to the bottom 75mm as stated in the instrument manual. All mice were scanned in triplicate using the “normal” mode of precision; the tube containing the mouse was removed from the instrument and then reinserted for each measure, the mouse was not removed from the container each time. After scanning, mice were immediately killed using carbon dioxide asphyxiation followed by cervical dislocation. After opening the abdominal and thoracic cavities, mice were placed in the oven (70°C) and dried to constant mass to determine total body water. The dried carcass was homogenized using a mortar and pestle and fat mass was determined by extraction with petroleum ether in a Soxhlet apparatus. The remaining fat-free dry mass (after Soxhlet extraction) was weighed, placed in crucibles and combusted in a muffle furnace at 600°C overnight (minimum 8 hours) to obtain ash content. Lean-tissue mass (LTM) was calculated as carcass water plus carcass fat-free dry mass minus carcass ash.

All statistical analyses were conducted using SAS (version 9.1, SAS Institute, Inc., Cary, NC). Precision was taken as the mean intra-individual coefficients of variation (CV%) for the three QMR scans of each animal. Paired t-tests were used to compare QMR data and chemically derived data. Data from the first scan was used in determining accuracy. Only 47 of the 48 animals were used in the accuracy analyses as some of the ash content of one of the carcasses was lost in removal from the furnace. Prediction equations were calculated using linear regression analysis. Data were considered significant when P<0.05.

Cross-validation study

To test the validity of the prediction equations a cross-validation was performed using an independent set of mice. Twenty-six mice (13 males and 13 females) of varying strains weighing 16.6 to 51.5g were obtained that ranged in age from three to fifteen months. Animals were tested the same as for the validation study with the exception of only taking one QMR measurement. These animals did not differ statistically from the original validation group with respect to weight or age (P>0.05).

The prediction equations generated from the validation study were applied to the cross-validation QMR data to produce predicted values of fat and lean-tissue masses. Validity was evaluated by comparing the QMR predicted variables with chemical carcass analysis variables by paired t-tests, and linear regression analysis to determine if the predicted results passed through the origin and had a slope that did not differ from unity. Additionally, the pure error and root mean square error (RMSE) for each equation were calculated to assess performance. Pure error was calculated as the square root of the sum of squared differences between the predicted and carcass values divided by the number of animals in this study (pure error=√(Σ(predicted-carcass)2/26)). Root Mean Square Error (RMSE) was determined using SAS.

All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Alabama at Birmingham.

Results

Validation Study

The CV of fat ranged from 0.08 to 4.38% with a mean and standard deviation of 1.58 ± 0.98%. There was a significant relationship between the weight of the animal and the fat CV% (P=0.02; Fig. 1A). Precision for LTM ranged from 0.07 to 2.08% with a mean and standard deviation of 0.78 ± 0.46%. There was no significant relationship between the weight of the animal and the lean CV% (P=0.95; Fig. 2B).

Figure 1
Precision for QMR over three repeated measurements of mice (n=47): (A) There was a significant relationship between mouse weight and FM CV%.; (B) There was no significant relationship between mouse weight and LTM CV%.
Figure 2
(A) Relationship between fat mass determined by chemical carcass analysis and by QMR (dashed line represent the line of identity, solid line represents line of best fit). (B) Residual plot of the differences between QMR fat and carcass fat indicating ...

QMR significantly overestimated FM relative to chemical carcass analysis (7.76±5.93g vs. 6.03±5.17g; P<0.01). However, there was a highly significant relationship between QMR- and chemically-derived FM (r2=0.99; P<0.0001; Fig. 2A). A residual plot showed a significant bias for QMR to overestimate fat more as fat mass increased (P<0.05; Fig. 2B). In contrast to fat mass, LTM was significantly underestimated by QMR (20.73±6.19g vs. 22.48±6.75g; P<0.01), but a strong relationship between variables was observed; r2=0.97, P<0.0001 (Fig. 3A). A residual plot revealed a significant bias for QMR to underestimate lean more as lean mass increased (P<0.05; Fig. 3B). The strong relationships between chemically-derived variables and QMR variables allowed for the generation of correction equations for both FM and LTM by linear regression analysis (Table 1).

Figure 3
(A) Relationship between lean mass determined by chemical carcass analysis and by QMR (dashed line represent the line of identity, solid line represents line of best fit). (B) Residual plot of the differences between QMR lean and carcass lean indicating ...

Cross-validation study

There was no significant difference between QMR predicted fat values and chemically-derived fat values (6.37±6.07g vs. 6.24±5.93g, P=0.15). The intercept of the relationship was not significantly different than zero (P=0.80) and the slope was not significantly different than 1.0 (P=0.08; Fig. 4A). The residual plot of FM showed no significant bias (P=0.17) (Fig. 4B). Pure error for the prediction equation for FM was calculated as 0.45g and the RMSE was 0.43g.

Figure 4
(A) Relationship between fat mass determined by chemical carcass analysis and by prediction equations from QMR values (n=26) (dashed line represents the line of identity; solid line represents line of best fit). (B) Residual plot of the differences between ...

QMR predicted LTM was not significantly different than chemical carcass lean values (21.80±4.31g vs. 22.18±5.16g, P=0.10). However, the intercept of the line for QMR predicted LTM variables was significantly different than zero (P=0.007) and the slope of the line was significantly different than one (P=0.002; Fig. 5A). A residual plot for LTM also showed a significant bias for QMR to underestimate LTM as the lean mass increased (P<0.01; Fig. 5B). Pure error was 1.21g and RMSE was 0.89g.

Figure 5
(A) Relationship between lean mass determined by chemical carcass analysis and by prediction equations from QMR values (n=26) (dashed line represents the line of identity; solid line represents line of best fit). (B) Residual plot of the differences between ...

Discussion

This study demonstrates the precision and accuracy of a quantitative magnetic resonance instrument (EchoMRI 3-in-1 Composition Analyzer) in measuring FM and LTM in live, non-anesthetized mice. Precision was calculated as the mean intra-individual coefficient of variation (CV) of the triplicate QMR scans for each mouse. The QMR instrument in this study was shown to be highly precise for in-vivo measurements with very small CVs for both fat and lean mass (FM=1.6%, LTM=0.78%) indicating good repeatability. This method can be compared to another well-known method for in vivo body composition analysis, dual-energy X-ray absorptiometry (DXA). Nagy and Clair (6) validated DXA (GE-Lunar PIXImus) for in vivo body composition using twenty-five mice which were measured three times each. Precision for FM and LTM was found to be 2.2% and 0.86%, respectively (6). However, it should be noted that precision for DXA measurements varies depending upon the DXA instrument and software used. Tinsley et al. (17) analyzed precision of a QMR instrument (Echo Medical Systems) in comparison to DXA (Norland P-DXA; Fort Atkinson, WI) and chemical carcass analysis using fifty-one mice scanned alive and post mortem (17). They found QMR's precision to be “superior” to that of DXA for fat mass. CVs ranged from 0.34% to 0.71% for QMR versus 3.06% to 12.60% for DXA for animals post-mortem. For live animals, CVs for fat of 0.9-1.6% were obtained from QMR. However, there was no mention of precision for lean-tissue mass and no mention of precision for DXA live-animal scans in that study (17).

Of great interest in this study was the ability of the QMR instrument to accurately measure or predict FM and LTM in mice in vivo. Overall, FM was significantly overestimated and LTM was significantly underestimated in mice. A portion of the overestimation of FM could be explained by the method of fat extraction used (Soxhlet fat extraction using petroleum ether). Petroleum ether is a solvent that extracts fats mainly in the form of triglycerides but does not extract polar lipids (18). The QMR instrument used in this study measures total body lipids (non-polar as well as polar) [19]. Total lipid is composed of approximately 83% triglycerides and 17% structural lipids (18). Since this study's chemical carcass analysis method leaves behind 17% of the fat that the QMR measures, it follows that there should be an overestimation by QMR of ~17%. However, the data reveal a mean overestimation of 39% of FM by QMR. So, there remains some discrepancy that is not explained by the extraction solvent. Another possible source of error could be intrinsic to the instrument itself due to an inaccurate calibration.

In contrast to fat mass, LTM was significantly underestimated by QMR when compared to chemical carcass analysis. As stated earlier, LTM was calculated as carcass fat-free dry mass minus carcass ash content plus carcass water. Again, the method of fat extraction uses petroleum ether, which only extracts 83% of total lipids thus leaving behind approximately 17%. This lipid remains as part of the fat-free dry mass making it a larger number than it should be. So, if this larger number is used in the above calculation for LTM, it will, of course, give a higher LTM for carcass analysis which presents itself as an underestimation by QMR.

Because values for fat and LTM from QMR differed from that of chemical carcass analysis we calculated prediction equations using the data from the original 47 animals. The relationship between QMR and carcass-derived data was strong (r2=0.99 and 0.97 for fat and LTM respectively; Table 1). These equations were then applied to the 26 animals from the independent cross validation sample. The predicted data were compared directly to the data from chemical carcass analysis. Predicted FM values were not significantly different than chemically-derived values and the regression intercept was not significantly different than zero and the slope was not significantly different than one. Pure error and RMSE were small (0.45g and 0.43g) indicating good strength of the equation used. Data for predicted LTM differed slightly from that for FM. Predicted LTMs were not significantly different than those measured by chemical carcass analysis, suggesting that the equation worked appropriately. However, upon further examination of the relationship between the predicted and actual values, one sees that the regression equation intercept is significantly different than zero and that the slope of the relationship differed from one. Thus, although the pure error and RMSE were small (1.21 and 0.89g), there were potentially problems with the cross-validation. The problem appears to be that as the mass of LTM increases above the group mean (approx 23g) that the predicted values will underestimate the data from chemical carcass analysis.

Techniques that offer in vivo measurements are imperative to studies that need to track changes in body composition over a period of time and those in which very expensive or rare animal models are used. The new technique of quantitative magnetic resonance presented in this study not only measures body composition in vivo, but does not require anesthesia or sedation which avoids several deleterious side effects. It can be concluded from this study that quantitative magnetic resonance is a valid and useful method for determining body composition in live, non-anesthetized, non-sedated mice. Additionally, the QMR instrument's excellent precision, strong correlation with chemical carcass analysis data, and speed of scan and analysis (95 seconds) make it a very practical and reliable tool. However, caution should be taken in applying the prediction equation for LTM derived in this study to animals that have large lean tissue masses as those values may be underestimated.

Acknowledgments

We thank Amanda Watts for her assistance in this study. This work was supported by the UAB Clinical Nutrition Research Center (P30DK56336), the Alabama Neuroscience Blueprint Core Center (P30ARC46031), and the UAB Diabetes Research and Training Center (P60DK079626).

References

1. Johnson MS, Nagy TR. Animal body composition methods. In: Heymsfield SB, Lohman TG, Wang Z, Going SB, editors. Human Body Composition. Second. Human Kinetics; 2005. pp. 141–50.
2. Nagy TR. The use of dual-energy X-ray absorptiometry for the measurement of body composition. In: Speakman JR, editor. Body Composition Analysis of Animals: A Handbook of Non-Destructive Methods. Cambridge: Cambridge University Press; 2001. pp. 211–29.
3. Brommage R. Validation and calibration of DEXA body composition in mice. Am J Physiol Endocrinol Metab. 2003;285:E454–9. [PubMed]
4. Hunter HL, Nagy TR. Body composition in a seasonal model of obesity: longitudinal measures and validation of DXA. Obesity Research. 2002;10:1180–7. [PubMed]
5. Johnson MS, Landy NM, Potter EP, Nagy TN. Comparison of software versions for body composition analysis using the PIXImus dual-energy X-ray absorptiometer. Int J Body Composition Research. 2005;3:69–72. [PMC free article] [PubMed]
6. Nagy TR, Clair AL. Precision and accuracy of dual-energy X-ray absorptiometry for determining in vivo body composition of mice. Obesity Research. 2000;8:392–8. [PubMed]
7. Speakman JR, Booles D, Butterwick R. Validation of dual energy X-ray absoptiometry (DXA) by comparison with chemical analysis of dogs and cats. Int J Obes. 2001;25:439–47. [PubMed]
8. Ross R, Léger L, Guardo R, et al. Adipose tissue volume measured by magnetic resonance imaging and computerized tomography in rats. Journal of Applied Physiology. 1991;70:2164–72. [PubMed]
9. Nagy TR, Johnson MS. Measurement of relative body and liver fat in small animals using peripheral quantitative computed tomography. Int J Body Comp Res. 2004;4:155–60. [PMC free article] [PubMed]
10. Tang H, Vasselli JR, Wu EX, et al. High-resolution magnetic resonance imaging tracks changes in organ and tissue mass in obese and aging rats. American Journal Physiology. 2002;282:R890–9. [PubMed]
11. Changani KK, Nicholson A, White A, et al. A longitudinal magnetic resonance imaging (MRI) study of differences in abdominal fat distribution between normal mice, and lean overexpressors of mitochondrial uncoupling protein-3 (UCP-3) Diabetes, Obesity and Metabolism. 2003;5:99–105. [PubMed]
12. Kunnecke B, Verry P, Benardeau A, Kienlin MV. Quantitative body composition analysis in awake mice and rats by magnetic resonance relaxometry. Obesity Research. 2004;12:1604–15. [PubMed]
13. Foster MA, Fowler PA. Non-invasive methods for assessment of body composition. Proc Nutr Soc. 1988;47:375–85. [PubMed]
14. Scott I, Colin S, Mitchell PI, Evans P. The use of total body electrical conductivity (TOBEC) to determine body composition in vertebrates. In: Speakman JR, editor. Body Composition Analysis of Animals: A Handbook of Non-destructive methods. Cambridge, UK: Cambridge University Press; 2001. pp. 127–60.
15. Wirsing AJ, Steury TD, Murray DL. Noninvasive estimation of body composition in small mammals: a comparison of conductive and morphometric techniques. Physiol and Biochem Zoology. 2002;75:489–97. [PubMed]
16. Taicher GZ, Tinsley FC, Reiderman A, Heiman ML. Quantitative magnetic resonance (QMR) method for bone and whole-body-composition analysis. Anal Bioanal Chem. 2003;377:990–1002. [PubMed]
17. Tinsley FC, Taicher GZ, Heiman ML. Evaluation of a quantitative magnetic resonance method for mouse whole body composition analysis. Obes Res. 2004;12:150–60. [PubMed]
18. Comizio R, Pietrobelli A, Tan YX, et al. Total body lipid and triglyceride response to energy deficit: relevance to body composition models. Am J Physiol Endocrinol Metab. 1998;274:E860–6. [PubMed]
19. Echo Medical Systems. Frequently asked questions. [Electronic Version] 2006. http://www.echomedical.com.