Research Department of Human Nutrition, The Royal Veterinary and Agricultural University, Rolighedsvej 30, DK-1958 Frederiksberg C, Denmark. E-mail: email@example.com
Objective: To study the influence of scan velocities of DXA on the measured size of fat mass, lean body mass, bone mineral content and density, and total body weight.
Research Methods and Procedures: The subjects were 71 healthy white adults, 38 women and 33 men. The mean age was 41.7 ± 13.5 years and body mass index was 28.6 ± 5.6 kg/m2. The subjects were scanned consecutively in slow, medium, and fast scan mode by a Lunar DPX-IQ DXA scanner.
Results: Throughout the body mass index and sagittal height ranges, scanned lean body mass significantly decreased with higher scan velocity and lean body mass was 2.7% lower in fast than in medium mode (p < 0.0001). In contrast, fat mass, percentage of body fat, and bone mineral contents were higher with increasing scan velocity. Areas not analyzed by the scanner, so called blue spots, increased with scan velocity and sagittal height, and their presence significantly enhanced the error. Body weight estimated by DXA in slow mode was −0.8% lower than scale weight in the women (p < 0.001) and −0.2% in men (not significant), and the difference was greater with increasing scan velocity.
Discussion: Scan velocity significantly influences the measured fat mass size, lean body mass, bone mineral content, and body weight. To obtain the most accurate results, slow mode is preferable and fast scans should be avoided. Future studies should report and take scan velocity into consideration.
Body composition measured by DXA scanning has become increasingly popular because it is a noninvasive, safe, fast, and easy method. Many studies have assessed the accuracy of DXA measurement in comparison to other gold standard methods. In pigs there is good agreement between total body compositions determined by DXA and post mortem analysis (1). Other studies found differences among different brands of DXA scanners, especially with regard to bone mineral density (BMD) and bone mineral content (BMC) (2, 3, 4, 5, 6), but also in lean body mass and fat mass, when using different scanners from the same manufacturer (7). Due to these differences, the international DXA standardization committee has now derived a set of equations to convert BMD results among different scanners (8, 9). Significant differences in total body composition when using different versions of the analysis software (10) and differences among scanning techniques have also been reported (11). Another study, comparing software developed by the Lunar Radiation Corp. for scanning adults and children, reported systematic differences in estimated BMD and BMC (12).
Numerous studies have been published with data on body composition generated by the Lunar DPX-IQ DXA scanner (Lunar Radiation Corp., Madison, WI), but only a few have provided information about the used scan velocity. The scan velocity could potentially be important for obtaining accurate results especially when scanning obese subjects, because the scan speed determines how long each pixel is analyzed. The Lunar DPX-IQ DXA scanner has three different scan velocities for whole body scans: fast, medium, and slow. There is a difference of 30 minutes between scans performed in fast and slow mode, a difference that makes it tempting to choose the fast mode.
The purpose of this study was to investigate for possible differences in body composition obtained by using three scan velocities in relationship to gender, body size, and sagittal height as a measure of abdominal obesity.
Research Methods and Procedures
Thirty-eight healthy women and 33 healthy men 21 to 68 years old with body mass indexes (BMIs) from 18 to 42 kg/m2 were invited to participate by advertisements in the local newspapers and posters in local supermarkets. None of the subjects had metal implants or were amputated (Table 1). Each subject arrived at the institute after an overnight fast. The subjects were placed on the scan table in the supine position with their arms placed along their thighs. All subjects were mummy wrapped in a broad belt, made of coarse tightly woven cloth that was closed using Velcro straps, to maintain their position within the 62-cm transverse scan area during and between the scans. They were requested to remain on the examination table between each scanning. To avoid attenuation of the X-ray beams, subjects were requested to wear light clothing and remove all removable metal items during the scans.
Table 1. Subject characteristics
Women (n = 38)
Men (n = 33)
Values are means ± SD, and the range is shown in the parentheses.
The study was in accordance with the guidelines of the second Helsinki Declaration and approved by The Scientific Ethical committee for Copenhagen and Frederiksberg. All subjects gave an informed consent.
After voiding, body weight was measured on an electronic scale (Lindell 8000; Lindell, Kristianstad, Sweden); subjects wore light clothing and no shoes. Weight was measured to the nearest 0.05 kg. Precision was ±0.05 kg up to 100 kg and ±0.1 kg between 100 and 200 kg. Height was measured to the nearest 0.5 cm with the subject standing against a wall-mounted stadiometer (Hultafors AB, Hultafors, Sweden).
Sagittal height was obtained in the last 54 of the 71 subjects. The participant was placed in a supine position on an examination table. Sagittal height was measured in centimeters as the maximal distance between the top of the examination table and a spirit level placed horizontally above the abdomen at the level of the iliac crest (Table 2).
Table 2. Subject characteristics according to sagittal height
Values are means ± SD and the range is shown in the parentheses.
Sagittal height < 22 cm
(n = 17)
(n = 8)
41.0 ± 14.8
(21.3 to 67.5)
40.4 ± 11.3
(27.2 to 57.7)
Body mass index (BMI; kg/m2)
24.8 ± 3.5
(19.1 to 30.7)
25.4 ± 2.7
(22.7 to 29.0)
Sagittal height (cm)
18.2 ± 2.5
(13.6 to 21.5)
19.0 ± 1.8
(16.8 to 21.7)
Sagittal height 22 to 28 cm
(n = 11)
(n = 15)
51.4 ± 10.1
(31.8 to 67.0)
45.9 ± 10.7
(27.5 to 62.0)
33.6 ± 3.9
(27.7 to 39.1)
31.9 ± 3.8
(23.7 to 40.4)
Sagittal height (cm)
23.9 ± 1.3
(22 to 25.6)
24.0 ± 1.6
(22.1 to 27.3)
Sagittal height > 28 cm (data not shown)
(n = 1)
(n = 2)
The Lunar DPX-IQ DXA (Lunar Radiation Corp.) uses a constant X-ray source of 80 Kvp and a k-edge filter with two different energy levels of 38 and 70 keV. For total body scans the scanner consecutively samples pixels measured 0.48 cm in width and 0.96 cm in height in a series of transverse scans from head to foot. The total scan area measures 61 cm × 196 cm, equivalent to 204 scan lines, each line consisting of maximum of 127 pixels. The total scan typically consists of ∼11,000 pixels, ∼6000 pixels with only soft tissue, and 5000 pixels with both bone and soft tissue (13). For total body scans, this scanner has three different scan modes with the following velocities and durations: fast, 153 mm/s lasting 10 minutes; medium, 77 mm/s lasting 21 minutes; and slow, 38 mm/s lasting 42 minutes.
Each pixel is analyzed as consisting of two compartments. In pixels with soft tissue only, the analysis distinguishes between fat and nonfat tissue. In the presence of bone, it distinguishes between bone and soft tissue. The composition of soft tissue in each pixel containing bone tissue is, thus, indirectly calculated from data on the surrounding pixels and previous knowledge and assumptions concerning soft tissue composition in the two types of pixels in a given location on the body. The R value is the ratio of soft tissue attenuation at 70 keV to that at 38 keV. The measurement of fat is derived from the R ratio. Data can be analyzed as extended analysis, where a special filtration technique is used, or as standard analysis. Extended analysis uses regional R values for the calculations, whereas standard analysis uses the overallR value for calculation.
During the scan process some areas are excluded from analysis in the calculations. This typically happens in extra thick areas or above metal implants. These unanalyzed pixels are called blue spots, being visualized in the scan picture on the screen of the analyzing computer as distinctive blue spots (Figures 1 and 2). In all 213 scan pictures, we recorded whether blue spots were present. Blue spots were considered present if there were at least 10 visible blue spots in the scan picture.
All scans were performed consecutively on the fasting subjects in all three modes and in the same order: slow, medium, and fast. The same operator (M.K.) performed all scans and analyses. Results were analyzed using Lunar Smart Scan, Version 4.6c as extended analysis. All relative cut locations were positioned as recommended by the manufacturer (14). The DXA whole-body scan provides data for lean body mass, fat mass, BMC, BMD, and total body weight. The percentage of Body fat was computed as fat mass/total body weight × 100%. For quality assurance and equilibration, a calibration block was scanned each morning before scanning the participants. A spine phantom was scanned on a weekly basis; the coefficient of variation was 0.01.
Data Analysis and Statistics
Data from the scans were stored in the Smart Scan database as three individual files (pbio.dbf, biographical data; scan.dbf, scanner related data; and region.dbf, scan data). These three files were updated after every scan, and subsequently all three files were merged using a MS Access database file (Microsoft Corp., Seattle, WA) and imported to a SPSS data file (Chicago, IL) for final calculations.
Descriptive statistics were performed using SPSS for Windows, Version 9.0 and SAS for Windows, Version 6.12 (Cary, NC). Slow was used as reference mode in comparison to medium or fast mode. Measurements were compared using a paired two-sided t test. The level of significance was chosen at p < 0.05. Men and women were analyzed separately. The paired two-sided t test was also used to analyze differences when the participants were grouped according to sagittal height. An additional 2 gender × 2 sagittal height group ANOVA was conducted on the differences among scans with fast, medium, and slow mode to test for gender effect, the sagittal height effect, and the interaction between gender and sagittal height.
To take into account that the measurements at the three scan velocities were correlated within each person, the SAS procedure PROC MIXED was used to make a linear regression model. Adjustment for presence of blue spots was made by including a dichotomous variable indicating presence or absence of blue spots. Dependence of BMI was modeled linearly or quadratically, and it was tested whether sex or BMI modified the effect of modes, i.e., testing for interaction between mode and sex or BMI. The adequacy of each model was assessed by test of normality of the residuals.
Fast mode measurements resulted in lower lean body mass, higher BMC, higher percentage of body fat, and higher fat mass than slow mode (p < 0.001), and this difference was found in both genders (Table 3). Compared with slow mode, medium mode also resulted in lower lean body mass and higher BMC in both genders (p < 0.001). In the men only, fast mode resulted in lower BMD (p < 0.05) compared with medium and slow modes.
Table 3. Body composition at different scan velocities
Values are means ± SD.
The difference in percentage is calculated as slow vs. medium or slow vs. fast.
The manufacturer recommends that scan velocity should be chosen according to trunk height. Fast mode is recommended for scanning subjects who are 15 to 22 cm thick in the trunk region, medium mode for subjects who are 22 to 28 cm thick, and slow mode for subjects who are >28 cm thick (15). We used abdominal sagittal height as an index of trunk height (Table 4) and stratified the subjects into three groups according to sagittal height. We compared all three scan velocities for subjects with sagittal heights below 22 cm and between 22 and 28 cm. The three subjects with sagittal heights above 28 cm were not compared. For sagittal heights below 22 cm, fast mode produced higher fat mass (men, p < 0.05 and women, p < 0.001), lower lean body mass (p < 0.001), higher BMC (p < 0.05), and higher percentage of body fat (p < 0.001) in both genders. Medium mode gave rise to higher fat mass (women, p < 0.05), lower lean body mass (women, p < 0.05), and higher BMC (men, p < 0.05 and women, p < 0.001). For sagittal heights between 22 and 28 cm, fast compared with slow mode, resulted in higher fat mass (men, p < 0.001 and women, p < 0.05), lower lean body mass (p < 0.001), higher BMC (men, p < 0.05 and women, p < 0.001), and higher percentage of body fat (p < 0.05) in both genders. Compared with slow mode, medium mode resulted in lower lean body mass (p < 0.05) and higher BMC (p < 0.05). From ANOVA models of differences, we found no interaction between gender and sagittal height group, and no significant difference between men and women. However, differences between the fast and slow modes were greater for those with sagittal heights above 22 cm compared with those with sagittal heights below 22 cm with respect to lean mass and BMD, and between medium and slow modes regarding BMC.
Table 4. Body composition at different velocities and sagittal height ranges
Using a linear regression model to test whether the differences among the three scan velocities were modified by BMI, we found that the differences in the estimated fat mass and the percentage of body fat increased with increasing BMI in both medium (p < 0.05) and fast modes (p < 0.0001) with no gender differences (Figures 3 and 4). The difference in measured lean mass increased with increased BMI in both the medium (p < 0.05) and fast modes (p < 0.0001) with no gender differences. Estimated BMC was higher using the fast and medium modes than slow mode (p < 0.0001) independent of BMI, but the estimates were higher in women than in men. For BMD there was no significant difference between results from different scan velocities and no effect of BMI or gender.
When we compared a high accuracy scale weight with body weight estimated by DXA in slow mode, we found a significant systematically lower DXA body weight in women (−0.8%), but not in men (Table 5). The DXA body weight in slow mode was 0.61 kg lower than scale weight in women (p < 0.0001) and 0.25 kg higher in men. The greatest difference between scale weight and DXA body weight (Figure 5 and 6) was between scale weight and DXA weight in fast mode (p < 0.0001).
Table 5. Weight-scale weight and DXA body weight at different velocities and according to sagittal height
The presence of blue spots was more frequent in fast than in medium scan velocities. Blue spots were present in 4 of 71 subjects in medium mode and in 30 of 71 subjects in fast mode. After adjustments were made for blue spots fat mass, percentage of body fat, and BMD were significantly higher (results not shown).
This study demonstrates that scan velocity significantly influences the estimates of whole-body composition. In both genders, fat mass and BMC were measured to be higher, whereas lean body mass and total body weight were measured to be lower by fast mode than by slow mode.
A high-precision electronic scale was used as a reference method. We found the best agreement between scale weight and DXA body weight using slow mode. The large number of blue spots occurring in fast mode, and to some extent in medium mode, could imply that the slow scan mode has a higher accuracy than the fast and medium modes. When blue spots were present, they contributed significantly to the measurement error as indicated by the improvement in fast scans by adjustment for blue spots. This suggests that the loss of accuracy by increasing scan velocity can be explained by the blue spot phenomenon.
There are a number of possible explanations for the discrepancies found between the different scan velocities. With increasing sagittal height, there is an increasing attenuation of radiation through the subject, particularly for the low-energy radiation. The manufacturer recommends that scan velocity should be chosen according to trunk height to compensate for this. Even after stratifying the subjects according to sagittal height, we found significant differences among measurements made by the various scan velocities. We also found differences in the same direction among scan velocities at different values of BMI.
We recommend slow mode as the preferable choice of scan mode, but there are several issues to be addressed before choosing scan velocity. When measuring BMD only in connection with screening large populations for osteoporosis, fast mode may be sufficient. Scanning before and after an intervention in a study should be performed using the same scan mode to obtain comparable results. There is a problem for those who choose to follow the guidelines provided by the manufacturer, especially in studies involving weight loss. Subjects losing weight will normally also reduce their trunk height in connection with the weight loss. The reduced trunk height means that these subjects could move from one scan velocity to another if the manufacturer's recommendations are followed. This could introduce additional differences, caused by changing the scan mode and not the weight loss alone.
Reasons for choosing to scan in fast or medium modes could be that they are less time consuming, thus making it possible to scan far more subjects than in slow mode with the drawback of less accurate results. Subjects who are unable to lie still for the duration of a slow scan, especially obese subjects or those with back pain, may be scanned in medium or fast mode.
In conclusion, this study illustrates the influence of scan velocity for obtaining the most accurate results. The same scan velocity should be chosen, preferably slow, especially when the same subject is scanned repeatedly. It is important to note that it is very difficult to compare scan results among different studies, especially due to the fact that many do not report which scan velocity has been used. We recommend that all scans, if possible, should be performed in slow mode, preferably be analyzed as extended analysis, and scan mode should be reported in publications.
This study was supported by the VELUX Foundation, the Danish National Research Foundation, and the Danish Medical Research Council. We thank senior statistician Claus Holst for advice on the statistical analyses.