|Home | About | Journals | Submit | Contact Us | Français|
To compare the accuracy of body composition measurements to small, defined changes in fat mass between dual X-ray absorptiometry (DXA) and air-displacement plethysmography (ADP).
Fifty-six healthy adults, 29 women and 27 men (age, 38 ± 12.4 years; BMI, 27.6 ± 5.8 kg/m2) were included in the study. Exclusion criteria were pregnancy, indwelling metal hardware or pacemakers, or weight exceeding DXA table limit (>350 lbs). All individual testing was completed within a 2-hour period. Fat packets were prepared using lard wrapped in plastic and applied exogenously in defined locations. Each participant completed body composition measurements with ADP and DXA (both testing modalities completed with and without 1 kg of exogenously applied fat mass).
Both DXA and ADP were highly accurate in detecting an overall increase in body mass associated with exogenously applied 1kg of fat mass (0.99 kg vs. 0.97 kg, respectively). DXA more accurately detected exogenous fat increase as fat mass compared to ADP (0.93 kg; 90% CI for the mean of the difference: 0.83 to 1.03 kg vs. 0.45 kg; 90% CI: 0.19 to 0.71 kg, respectively). The accuracy of body mass detection was similar for males and females (0.97 vs. 1.02 for DXA and 0.92 vs. 1.02 for ADP, respectively), though accuracy in detecting added mass as fat was less accurate in males than females (0.84 vs. 1.00 for DXA and 0.39 vs. 0.51 for ADP, respectively)
Both DXA and ADP are accurate in detecting an overall increase in body mass associated with exogenously applied 1kg of fat mass. However, DXA is more accurate than ADP in correctly identifying the increase in body mass as fat mass, as opposed to fat free mass.
The prevalence of obesity in both children and adults is increasing in the US and numerous other countries . Mirroring this trend, studies have shown an increase in obesity related morbidity and mortality . The ability to accurately assess and monitor body composition changes is of paramount importance to both clinicians and researcher tracking response to nutritional, medication, and lifestyle changes [3, 4]. There is no accepted single tool for measurement of obesity and change in fat mass that provides a sufficiently accurate and precise estimate . The reference technique currently is the 4-compartment analysis , which is based on the measurement of body volume, total body water, bone mineral content, and weight. These measurements are often defined from a combination of bioelectrical impedance analysis, dual X-ray absorptiometry (DXA), air-displacement plethysmography (ADP). Among the clinically accepted and relatively inexpensive methods, DXA and ADP are frequently used in clinical research and practice.
ADP is based upon the two-component model, which divides the body into fat mass (FM) and fat-free mass (FFM). While demonstrated to be accurate in numerous studies [3, 5, 6], the model assumes constancy in composition of the FFM, which can be altered by hydration and disease processes . DXA is based upon the three-compartment model, which adds determination of bone mineral content (BMC) with FM and FFM. To assess changes in fat mass using an experimental model, previous studies have demonstrated that DXA can accurately quantify small changes in exogenous fat[4, 7]. In these studies, extra-corporal fat (e.g., lard packets) were placed on specific areas of the test subjects’ body. Though limitations of DXA in these settings were highlighted (namely decreased accuracy with increasing truncal fat mass), several studies have reported it as an accurate means of assessing fat mass in obesity and modest fat changes as would be expected in clinical or research settings[9, 10]. A similar study with ADP also demonstrated accuracy in detecting small changes in fat mass . A number of studies have compared the accuracy of ADP and DXA in tracking weight loss over time, with conflicting results [11, 12]. However, no study to date has directly compared the abilities of DXA and ADP to accurately detect small, defined changes in fat mass.
Thus, the goal of this study was to compare the accuracy of body composition measurements to small, clinically relevant defined changes in fat mass between DXA and ADP. The change in fat mass was simulated by attaching exogenous fat packets to the abdomen and thighs during the measurement.
Fifty-six 19 to 65 year old adults, including 29 females (8 African American, 16 Caucasian, 5 Hispanic) and 27 males (4 African American, 18 Caucasian, 5 Hispanic) participated in the study. Volunteers were recruited using flyers, e-mail distribution lists, and personal contact. All participants signed an informed consent document approved by the university-affiliated institutional review board. The age distribution of these participants is shown in Table 1.
All testing was completed during a single visit at the Clinical Research Center at Vanderbilt University. Females of childbearing potential completed beta-HCG pregnancy test due to ionizing radiation exposure during the DXA testing. Each participant had two DXA scans and six ADP measurements with or without exogenous fat packets used to simulate changes in fat mass. Fat packets were constructed using commercially available lard (assumed to be 100% fat) packaged in thin, plastic bags. Packets were of uniform size, thickness, and weight.
DXA scanning was performed using Lunar iDXA scanner (GE Lunar Medical Systems, Madison, WI, USA) with Encore 2007 software, version 11.40.004, and calibrated daily according to manufacturer specifications. DXA estimates body composition from the attenuation of X-rays between 38 and 76 KeV. Complete technical aspects of the DXA technique has been described in detail elsewhere . The iDXA instrument used in this study has an intra-individual coefficient of variation for FM (kg) of 0.96 ± 1.06%. All participants wore lightweight cotton clothing during the whole—body DXA scanning in the supine position. Data collected included fat mass (kg), fat free mass (kg), % fat mass, % fat free mass, total body mass (kg), tissue mass (kg), bone mineral content (grams), arm mass (lean, fat, BMC), leg mass (lean, fat, BMC), trunk mass (lean, fat, BMC), android mass (lean, fat, BMC) and gynoid mass (lean, fat, BMC). Whole body scans were than repeat with the lard packets of specific weight (four 250g packets) attached to their abdomen (2 packets) and thighs (2 packets) to simulate increased adiposity as a result of weight gain. Air was manually expelled from the packets to assure accuracy of measurement. All scans for a particular person were completed during the same study visit and analyzed by the same investigator (LEW).
Before each testing, the BOD POD (Gold Standard Body Composition Tracking System, version 4.2.4, Life Measurement, Concord, CA) was calibrated according to manufacturer specifications. In ADP, the volume of object is measured indirectly by measuring the volume of air it displaces inside an enclosed chamber (plethysmography). Thus, human body volume is measured when a person sits inside the chamber and displaces a volume of air equal to his or her body volume. Body volume is calculated indirectly by subtracting the volume of air remaining inside the chamber when the subject is inside from the volume of air in the empty chamber, as described previously. Participants were instructed to sit in the BOD POD still but relaxed, and breath normally. A minimum of 2 consecutive measurements of body volume (Vb) was complete for each person. A third measurement was obtained of the initial 2 varied by more than 150mL. Predicted respiratory volumes were used in calculation of body volume, as described previously . Body density (Db) was calculated from BOD POD’s integrated computed body volume and measured weight using the formula: Db = weight/Vb. The percentage of body fat was calculated from body density using the Siri formula for Caucasians and the Schutte formula for African Americans. The ADP instrument used in this study has an intra-individual coefficient of variation for FM (kg) of 4.7 ± 6.1%. Resulting data included fat mass (kg), fat free mass (kg), % fat mass, % fat free mass, and total body mass (kg). Subjects were then instructed to hold lard packets as previously described in specified locations (abdomen, thighs) for repeat BOD POD testing.
Body weight was measured by the research staff to the nearest 0.1 kg with a digital scale (Detecto-Medic, Detecto Scales, Inc., Northbrook, IL) while participants wore lightweight clothing and no shoes. Height was measured using a wall-mounted stadiometer (Perspective Enterprises, Portage, MI). Body mass index (BMI) was calculated as a ratio of weight (kg) to height (meters squared).
The primary aim for the analysis was to test whether each instrument (DXA or ADP) could measure fat mass within 20% (0.2 kg) of the added fat mass (1 kg). We used two one-sided tests commonly used for equivalence trials and we calculated the 90% confidence intervals (CI) obtained from a mixed-effects model, which is an equivalent to the two one-sided tests [17-19]. Results for males and females were also examined separately due to known differences in regional distribution of fat mass potentially affecting differences in body composition. Data are presented as means, standard deviations (SD) or ranges. All analyses were performed using the programming language R version 2.12.2 .
The coefficient of variation of the DXA measurements within each individual were ≤ 5%. For the combined study population, the difference in the fat mass before and after adding 1 kg of fat mass was equivalent to the true added fat mass within 20% limit of equivalence (mean of the difference = 0.93 kg; 90% CI for mean of the difference = 0.83 to 1.03 kg) (Table 2). The difference in the fat free mass before and after adding 1 kg of fat mass was not significantly different from the true added fat free mass of 0 kg (mean of the difference = 0.03 kg, p = 0.77). The difference in the body mass before and after adding fat mass of 1 kg was equivalent to the true added fat mass within 10% limit of equivalence (mean of the difference = 0.99 kg; 90% CI for mean of the difference = 0.93 to 1.04 kg).
When stratified by sex, variations from the above pattern were seen in the determination of fat mass. The difference in fat mass before and after adding fat mass of 1 kg was equivalent to the true added fat mass within 20% limit of equivalence in females (mean of the difference = 1.0 kg; 90% CI for mean of the difference = 0.88 to 1.14 kg) (Figure 1), but was not equivalent in males (mean of the difference = 0.84 kg; 90% CI for mean of the difference = 0.71 to 0.97 kg) (Figure 2).
The coefficient of variation of the ADP measurements within each subject varied widely up to 32%. For the combined study population, the difference in the fat mass before and after adding fat mass of 1 kg was not equivalent to the true added fat mass within 20% limit of equivalence (mean of the difference = 0.45 kg; 90% CI for mean of the difference = 0.19 to 0.71 kg) (Table 2). The difference in the fat free mass before and after adding fat mass of 1 kg was significantly different from the true added fat free mass of 0 kg (mean of the difference = 0.47 kg, p = 0.001). The difference in the body mass before and after adding fat mass of 1 kg was equivalent to the true added fat mass within the 20% limit of equivalence (mean of the difference = 0.97 kg; 90% CI for mean of the difference = 0.93 to 1.02 kg). When stratified by sex, no variations from the above pattern were seen.
This study demonstrates that both DXA and ADP are highly accurate in detecting an overall increase in body mass associated with exogenously added 1kg of fat equal to approximately 1 to 2% of total body mass. However, DXA proved more accurate than ADP in correctly identifying the increase as fat mass, as opposed to fat free mass. This was true for the whole group and when stratified by sex, despite the statistically significant difference in fat mass distribution between males and females. These findings are important in regards to the clinical utility of both ADP and DXA in tracking the total body and fat mass changes that would be expected in various nutritional, medication, and lifestyle-modification weight loss interventions [21, 22].
We also found that both DXA and ADP were more accurate in females than males at detecting an increase in fat mass. This disparity potentially could be caused by the differences in the baseline body fat distribution between the two groups. While the BMI for males and females was not statistically different (p=0.31), males had an overall greater body mass (p <0.05) while females had higher truncal fat mass (p=0.03) and leg fat mass (p <0.05) as determined by baseline DXA measurements. The difference between males and females remained significant after adjusting for BMI in the analysis. The confidence intervals were nearly identical and hence the results and conclusions remained the same (data not shown).
Second, the two methods differed significantly in detecting added fat as FM or FFM. DXA primarily detected the exogenous fat as an increase in total FM (mean increase of 0.93 kg), while ADP detected it as a nearly equal increase in both FFM (0.47 kg) and FM (0.45 kg). The detection of exogenous FM by DXA in males was more accurate than by ADP (0.84 kg vs. 0.39 kg, respectively). These results are in contrast to a previous report in which exogenously applied oil packets were accurately detected as FM by ADP . The discrepancy between the two studies could be caused by difference in consistency of added fat (solid and soft lard versus liquid oil).
It is also possible that the testing accuracy of the two instruments tested in this study had an impact on the results. As noted previously, the intra-individual coefficient of variation of ADP for fat mass was larger than that of DXA (4.7 ± 6.1% vs. 0.96 ± 1.06%, respectively). Corresponding with this variation, the range in measurement values for ADP was relatively large, ranging from 1 kg overestimation to 3 kg underestimation of added 1 kg fat packets. Despite this variability, the average overall underestimation of FM by ADP and DXA reflects a true difference in accuracy between the instruments. Recent updates in iDXA software now allow segmentation abdominal fat into subcutaneous fat and visceral fat components. While this further information is important in the clinical application of DXA, it is far beyond the capacities of ADP and consequently not utilized for the purpose of this current comparison study.
It has been suggested that the accuracy of ADP in determining fat mass is dependent upon the hydration of FFM, posing a potential source of error when tracking weight change over time . Le Carvennec and colleagues reported that the presence of small quantities of water (applied in packets exogenously in a similar method to our study) was detected largely as an increase in fat-free mass . Several studies with DXA have shown less of an effect of hydration on accuracy in determining fat mass and fat-free mass [24-26]. This is an important factor to consider, since interventions in clinical and research settings routinely focus on decreasing fat mass while preserving fat-free mass. However, our study design limited potential hydration effect by completing all of the testing and data collection within the same study visit (~2 hours). While this study provided an excellent comparison between accuracy of DXA and ADP in assessing small changes in fat mass, it did not address FFM hydration changes over time that could occur in clinical setting. Thus, in future studies these inherent shortcomings of DXA and ADP or any 2-compartment or 3-compartment model as a stand-alone test must be considered.
In addition to hydration, several studies have investigated the effect of differences in hair and clothing on BOD POD testing accuracy [27, 28]. The BOD POD uses the pressure-volume relationship to determine body volume, but has to make corrections for isothermal air found in the lungs, skin surface, clothing, and hair . In particular, the type of clothing worn caused an underestimation of body fat . Therefore, it is possible that lard packets utilized in our study had the potential to modify the isothermal air layer at the skin surface area, and consequently affect the results. Any changes in body or clothing positioning between repeated measurements could lead to changes in the isothermal layer, resulting in testing inaccuracy and presenting a limitation of the study. While the BOD POD correctly measured an increase in total body mass, a within-person difference in fat mass increase could have resulted from changes in the isothermal layer. However, in the study by Fields et al, the difference in fat mass calculations by BOD POD between subjects wearing either a form-fitting swimsuit or a hospital gown was only ~5%, suggesting that potential inconsistencies in the isothermal layer in the current study would not fully account for our observed between-person differences.
A limitation in both the current study and that reported by Le Carvennec et al, is the assumption that exogenous addition of fat mass correctly reflects the physiological distribution of increased fat mass within the body. This assumption cannot be confirmed using our experimental design, however the result showed the accurate detection of an overall increase in body mass by both methods. The addition of 1 kg of exogenous fat mass was chosen in the study design to simulate a typical, short-term change in body composition. However, this predetermined unit of change also poses a potential limitation in assessing the full sensitivity and accuracy of the two testing instruments. It is possible that testing with incremental increases in exogenous fat mass both less than and greater than 1 kg (eg. 250g, 500g, 1500g, 2000g ect.) might have further defined the range of accuracy for both instruments. Additionally, the accuracy of the results might have been further demonstrated by the inclusion of daily calibration measurements with standardized oil and water packets. However the coefficients of variation reported previously, as well as consistency in daily calibration with manufacturer phantoms (data not shown), suggest acceptable repeatability.
In summary, this study demonstrated that both DXA and ADP are accurate in detecting an overall increase in body mass caused by exogenously added ~1kg of fat mass. However, DXA is more accurate than ADP in correctly identifying the increase in body mass as fat mass.
JW contributed to data analysis and interpretation and writing of the article. MB contributed to the study conception, design, data analysis and interpretation, and revision of the article. JK contributed to study conception, design, and conduct of the experiments. LK contributed to design and conduct of the experiments. LC contributed to design and data analysis and interpretation.
We thank Jeff Kantor for his assistance with recruitment of participants and data collection for this study.
This project was supported in part by R01 DK69465 and the Vanderbilt CTSA grant UL1 RR024975 from NCRR/NIH.
CONFLICTS OF INTEREST The authors declare no conflict of interest.