Comparisons between a pencil beam and two fan beam dual energy X by ghkgkyyt


									The British Journal of Radiology, 74 (2001), 166–176      E   2001 The British Institute of Radiology

Comparisons between a pencil beam and two fan beam
dual energy X-ray absorptiometers used for measuring
total body bone and soft tissue
 Department of Medical Physics, Western General Hospital, Edinburgh EH4 2XU and 2Osteoporosis
Research Unit, Woolmanhill Hospital, Aberdeen AB25 1LD, UK

      Abstract. A pencil beam Hologic QDR 1000W scanner (1000), a fan beam QDR 4500A scanner
      (4500) and a fan beam Lunar Expert scanner (Expert) were compared for bone mineral and body
      composition measurement accuracy. Phantoms were scanned with each instrument to assess
      magnification effects and to compare calibrations for bone mineral and fat proportion. 41
      volunteers were scanned with both the 1000 and the 4500, and 21 patients with both the 4500 and
      the Expert. The height of a bone within the body affected the measured bone mineral content
      (BMC) and, to a lesser extent, the bone mineral density (BMD). There were differences in
      calibration against recognized standards for fat proportion between the three instruments. The
      1000 underestimated low fat proportions and the 4500 underestimated high fat proportions. Fat
      results for the Expert were closer to nominal values. Comparisons on volunteers showed that
      measured mean total body BMD was 4% higher and BMC was 7% higher with the 1000
      compared with the 4500; some regional differences were greater. Mean values of per cent fat were
      equal, but the total and regional regression coefficients were well above unity. Mean BMD was
      3% higher and mean BMC was 10% higher with the Expert compared with the 4500, but most
      regression coefficients for these comparisons were less than unity. Mean values of per cent fat
      were equal, but regression coefficients were above unity. Errors due to magnification are
      acceptable. Differences between the instruments are appreciable, but can be accommodated by

   Dual energy X-ray absorptiometry (DXA) was                   estimate the bone pixel fat proportion from that
introduced originally for the measurement of                    measured in adjacent areas. Manufacturers do not
bone mineral in the spine and hip, but has been                 reveal what assumptions are incorporated into
developed to provide measurements of bone as                    their algorithms, but they probably differ.
well as lean and fat soft tissue in the whole body.             Moreover, the assumptions cannot be universally
The use of bone measurements is well developed,                 valid, given the great variety of composition of
but fat and lean DXA measurements are increas-                  the human body. There are systematic differences
ingly being applied in studies of nutrition [1, 2]              in the results from DXA instruments from
and sports and exercise [3, 4]. The high precision              different manufacturers, both in regional bone
of the method facilitates comparisons between                   measurements [5, 6] and in total body bone and
groups as well as longitudinal studies. The                     soft tissue measurements [7, 8]. There are also
accuracy of the measurements, however, is less                  differences in the results produced by different
well established. This is in part owing to the                  software versions from the same manufacturer
absence of a reliable reference method, but also                [7–10].
stems from the innate limitations of DXA, which                    Recently, a new generation of DXA instru-
are often not appreciated. In X-ray attenuation                 ments has been introduced, using a fan beam
terms there are three main body components:                     geometry to provide faster scanning and improved
bone mineral, and fat and lean soft tissue. With                geometrical resolution. This raises the possibility
only two X-ray energies, only two components                    of errors introduced by magnification effects. We
can be determined in a given pixel. Thus, the fat               therefore investigated the characteristics of the
proportion cannot be directly determined in those               two main brands using phantoms, or models, to
pixels containing bone. It is therefore necessary to            establish their fundamental performance. We also
make assumptions about the distribution of fat to               made direct in vivo comparisons for cross-
                                                                calibration. In addition, comparisons were made
Received 1 March 2000 and in revised form 13 July 2000,         with an established pencil beam instrument. As
accepted 25 September 2000.                                     well as magnification effects, calibration of fat

166                                                                    The British Journal of Radiology, February 2001
Pencil beam and fan beam DXAs for total body bone and soft tissue measurement

and lean tissue and bone mineral measurements            alone. The effective composition of the material
were investigated. Factors other than accuracy,          of the larger rectangular tank was estimated by
such as precision and radiation dose, have not           scanning the partially water-filled tank with and
been included as these have been considered              without its lid. As this was composed of the same
elsewhere [11, 12].                                      plastic, the effective composition was calculated
   A condensed account of some of this work has          from the measured fat values and the weights of
been published in conference proceedings [13].           the tank, lid and water.
                                                            For the 4500 only, a collection of lard in plastic
                                                         bags, approximately 34 cm wide by 15 cm high by
Methods                                                  60 cm long, was scanned with and without a
  The pencil beam scanner was a Hologic QDR              ‘‘bone’’ of aluminium strips, 4.5 cm wide by
1000W (Hologic Inc, Bedford, MA) (abbreviated            50 cm long, at its centre.
to 1000), used in Edinburgh. Whole body scans               Comparisons between the three scanners for
were obtained and analysed using software V5.55          soft tissue composition were also made using a
in the enhanced mode. The corresponding                  variable composition phantom (VCP) similar to
Hologic fan beam instrument, also used in                that described by Formica et al [15]. This
Edinburgh, was a QDR 4500A (abbreviated to               contained five blocks of acrylic plastic (Perspex),
4500), using software V8.24a:3. The second fan           each 28 cm by 20 cm by 2 cm, with different
beam scanner was a Lunar Expert (Lunar                   degrees of leanness provided by interleaved thin
Corporation, Madison, WI) (abbreviated to                sheets of aluminium instead of vinyl plastic. The
Expert), used in Aberdeen and operated in total          total body analysis protocol for the Expert
body mode, 1.5 mA, medium speed, software v1.9.          instrument requires there to be a ‘‘bone head’’
                                                         in the appropriate position. As its characteristics
                                                         are unimportant, the Lunar spine phantom was
In vitro measurements
                                                         used for this purpose, as suggested by Formica
   The effects of magnification were investigated         et al [15].
by raising phantoms, or their components, to                The accuracy of each scanner in measuring
different heights above the couch, with the              changes in bone density was investigated using the
scanning parameters unchanged. The main phan-            whole body phantom described previously [7].
tom used was a thin-walled plastic tank, with            This consists of sheets of tissue-equivalent hard-
slightly tapered sides, 35 cm by 29 cm at its            board cut and stacked as cylinders to simulate a
centre, simulating a human trunk. It was filled           human body, with a simulated skeleton of
with water to various depths, from 6–18 cm, the          aluminium strips, the number of which could be
added quantities being measured by weighing.             changed in the arms, legs and spine to alter the
Levitation was simulated by raising the tank on          effective bone mineral densities (BMDs).
blocks of plastic foam. The effect of the height of         Measurements were repeated five times. t-tests
bone within the body was examined using an               were used to assess the significance of differences
aluminium rectangle, 29 cm by 4.5 cm, simulating         between means, using a cut-off value of p50.05.
a spine, which was suspended in water of 15 cm
depth. Similar experiments were performed using
a paraffin wax block, 29 cm by 16 cm by 4.5 cm
                                                         In vivo measurements
thick, to simulate fat at various heights.
   Absolute calibrations were checked against               Direct comparisons between the absorpti-
standards recommended by Nord and Payne                  ometers were made by scanning a number of
[14]; that is stearic acid in a block 20 cm by           subjects on each of two instruments within a short
14 cm by 10 cm thick representing 100% fat, and          period of time. There was no special selection of
a solution of 0.6% sodium chloride in water              the patients or volunteers. A fairly wide range of
representing 100% lean. The solution was con-            age, size, adiposity and bone characteristics was
tained in a polyethylene box 21 cm by 15 cm by           achieved. 41 subjects (17 male, 24 female), mean
8 cm high. The molecular composition of the              age 32.9 years (SD 14.8 years), mean total body
plastic suggests that in X-ray attenuation terms it      per cent fat 25.3% (SD 11.9%), BMD
should be equivalent to greater than 100% fat.           1.083 g cm22 (SD 0.093 g cm22), were scanned
This was tested by scanning a slightly larger            with the 1000 and the 4500; 21 subjects (6 male,
water-filled box 20 times with and without the            15 female), mean age 42.0 years (SD 13.7 years),
polyethylene box immersed in it. The differences         mean total body per cent fat 28.8% (SD 9.0%),
in observed fat proportions, coupled with the            BMD 1.132 g cm22 (SD 0.102 g cm22), were
measured weights, permitted calculation of the           scanned with the 4500 and the Expert. The
effective fat proportion of the material of the box.     studies had been approved by the Lothian
The box was also scanned when filled with water           Research Ethics Committee.

The British Journal of Radiology, February 2001                                                           167
                                                                            P Tothill, W J Hannan and S Wilkinson

Results                                                     masses and added weights were highly linear
                                                            (r.0.9999). The regression equations were:
In vitro measurements
                                                                 1000, M~0:985W {0:14, SEE~0:043              ð1Þ
   With both of the Hologic scanners it was found
that the measurements of fat and lean proportions
                                                                 4500, M~0:984W z0:26, SEE~0:047              ð2Þ
depended on the positions of the regions of
interest (ROIs) over the phantoms in the analyses.
                                                                Expert, M~1:068W {1:16, SEE~0:114             ð3Þ
When the phantom was totally included in the
trunk, the bone mineral content (BMC) and BMD               where M is the mass by DXA and W the weight
of any bone included depended on the positioning            by scales, both in kg; and SEE is the standard
of the division between thoracic and lumbar                 error of the estimate. Weights of water ranged
spine. The fat proportion also varied with this             from 7 kg to 20 kg (1 kg corresponds to 9 mm
placing, even when no bone was present. The                 depth of water). Slopes for the 1000 and the 4500
Hologic analyses were therefore carried out with a          were not significantly different from each other,
ROI in the legs, resulting in much more stable              but were significantly lower than unity ( p,0.05).
results. Results from the Expert were independent           Their intercepts were different from each other,
of ROI placing, so the trunk was used, in keeping           and from zero. Although there were statistically
with the ‘‘head’’ position.                                 significant departures of the regression lines from
   Some of the variations of mass with height               unity, these were not of practical importance, with
above the couch were studied at several heights             measured masses being within 2% of the scale
and were found to be linear. For simplicity, the            weights above a water depth of 6 cm. The
results are summarized in Table 1 for two heights,          negative intercept with the Expert was highly
0 cm and 15 cm, which cover most of the likely              significant and the slope was greater than unity,
range of heights of body components.                        so that masses were underestimated at low water
Measurements were repeated five times. The                   depths, for example by 14% at 6 cm depth, which
significance of height dependence was derived                is a typical thickness for much of the arms.
from the differences of mean results at the two                The effective composition of the small poly-
heights ( p,0.05). Both the 4500 and the Expert             ethylene box was determined to be 105% fat and
showed a dependence on height of the phantom                25% lean. These nominal proportions relied on
above the couch at approximately 1.3% per cm;               DXA scanning. As the mass of the box was small
this was negative for the 4500 and positive for the         compared with that of its contents, any inaccu-
Expert, owing to the different positions of the             racy in the fat measurement would have only a
X-ray tubes relative to the couch. The 1000                 small effect on the estimate of the composition of
showed a statistically significant but almost                the combination. When filled with salt solution,
negligible height dependence. The height of the             the effective fat content was 5% compared with
aluminium bone in the tank affected the measured            0% for the solution alone. Results of the
BMC to about the same extent as levitation. With            calibration experiments are presented in Table 2.
both fan beam instruments, the BMD was                      Standard deviations on repeated measurements
affected to a lesser extent, and in the opposite            were all less than 1% fat. The effective fat
direction, as bone area (BA) changed to a greater           proportion of water alone was taken as 9% [14]
extent than BMC. The height of the wax block                and the combination with the box material was
changed the fat content measured with the 4500,             calculated as above. Fat equivalence of the acrylic
but did not significantly ( p50.06) change it with           plastic was calculated by RH Nord (Personal
the Expert.                                                 communication, 1999). The 1000 underesti-
   When different quantities of water were added            mates and the 4500 overestimates fat content of
to the phantom, correlations between measured               the salt solution box. The 4500 appreciably

Table 1. Effect of height above couch

                                            Ratio of value at 15 cm to that at 0 cm
                                            1000                          4500                     Expert

Mass of levitated water-bath                1.01                          0.80                     1.21
Aluminium ‘‘bone’’ BMC                      0.95                          0.79                     1.21
Aluminium ‘‘bone’’ BMD                      ns                            1.03                     0.98
Wax block, mass of fat                      ns                            0.91                     ns

ns, not significantly different from unity (p.0.05).
1000, Hologic QDR 1000W; 4500, Hologic QDR 4500A; Expert, Lunar Expert.
BMC, bone mineral content; BMD, bone mineral density.

168                                                               The British Journal of Radiology, February 2001
Pencil beam and fan beam DXAs for total body bone and soft tissue measurement

Table 2. Calibration of fat proportion against nominal standards

                                                  Percentage fat
                                                  Nominal                    1000                   4500        Expert

0.6% NaC1 in polythene box                    5                              21                      9           4
8 cm water in polythene box                  14                               7                     15          10
Acrylic plastic in VCP                       69                              64                     58          64
Stearic acid block                          100                              96                     82          92
6 cm water in larger tank                    17.5                             8                     17
15 cm water in larger tank                   12.5                             5                     12
Regressions against nominal for first four standards:
  1000, % fat51.03bN26.7a
  4500, % fat50.77bN+4.7a
  Expert, % fat50.94bN21.7
where N5nominal per cent fat

1000, Hologic QDR 1000W; 4500, Hologic QDR 4500A; Expert, Lunar Expert.
VCP, variable composition phantom.
  Significant difference of intercept from zero (p,0.05); bsignificant difference of slope from unity (p,0.05).

underestimates the fat content of the stearic acid                   The lower slope for the 4500 compared with the
standard, whereas Expert results are intermediate                    1000 and the intermediate value for the Expert are
between those from the 1000 and 4500 in all                          in accord with the calibration differences in
respects and are not far from the nominal values.                    Table 2.
These relationships were illustrated by calculating                     The effective BMD of the aluminium strips
regression equations against nominal values. They                    used in the hardboard phantom had been
were highly linear (r.0.999). The equations are                      previously derived by calibration against calcium
included in Table 2. There are appreciable inter-                    hydroxyapatite [7]. When the measured values for
cepts with both Hologic instruments, but the most                    each of the variable components of the skeleton
notable disparity is the reduced slope for the 4500.                 were plotted against these nominal values it was
   The effective composition of the larger empty                     found that there was a BMD threshold for the
plastic tank was 85% fat and 15% lean. This,                         legs of 0.3 g cm22 with the 4500. A somewhat
together with the relative weights of the compo-                     similar anomaly occurred with each of the bones
nents, allowed calculation of the composition of                     for the Expert, as both BMC and BA were grossly
the tank plus water combinations, which are                          underestimated at BMDs below 0.4 g cm22. The
included in Table 2 as the nominal values. The                       upper parts of the plots, above approximately
depth-dependent results for the 4500 agree closely                   0.5 g cm22, were close to linear. Linear regres-
with the nominal values. Values for the 1000 are                     sions were derived; the parameters are presented
lower by 7–9%.                                                       in Table 3. Also included are the measured values
   The larger phantom of lard measured by the                        calculated from the regression equations at a
4500 gave results of 78% fat and 79% fat with and                    nominal BMD of 1.0 g cm22. Deduced BMDs are
without the simulated bone, respectively. A                          all significantly different from unity, and the
smaller sample of lard had been compared with                        differences between bones and between instru-
the stearic acid block and the fat proportions were                  ments are all significant. BMDs and BMCs from
not significantly different, so the underestimate of                  the Expert were all higher than those from the
the high fat proportion in Table 2 is confirmed in                    4500. The equations were used to estimate the
a phantom closer in size to a human trunk.                           measured change in BMD or BMC for a 10%
   The true effective composition of the VCP is                      increase at an initial BMD of 1.0 g cm22; these
not certain. Therefore the observed fat propor-                      results are included in Table 3. Changes were
tions were plotted against the number of alumi-                      nearly all underestimated by the 4500. Disparities
nium sheets. Regressions over the range 5–65% fat                    were less for the Expert.
were highly linear (r.0.997) and the equations                          No changes were made to the ‘‘soft tissue’’ of
were:                                                                the phantom, but the measured body composition
                                                                     demonstrated apparent variations of fat propor-
         1000, % fat~63{10:1N, SEE~1:0                    ð4Þ
                                                                     tion as the BMDs of the limbs and spine were
                                                                     altered. These variations are illustrated for the
         4500, % fat~58{8:3N, SEE~0:9                     ð5Þ
                                                                     limbs in Figure 1. It is evident that the recorded
                                                                     fat proportion varies with the instrument used,
        Expert, % fat~63{8:9N, SEE~0:8                    ð6Þ
                                                                     the part of the body considered and the BMD of
where N is the number of aluminium sheets. The                       the bone inserted into the phantom. Apart from
slopes are significantly different from each other.                   legs measured with the Expert, the recorded fat

The British Journal of Radiology, February 2001                                                                      169
                                                                           P Tothill, W J Hannan and S Wilkinson

Table 3. Regression equations for correlations between measured and nominal bone mineral density (BMD) and
bone mineral content (BMC) in hardboard plus aluminium whole body phantom, with measured change for a
nominal 10% change at a BMD of 1.0 g cm22. Y5A+BX; C5measured BMD or BMC at a nominal BMD of
1.0 g cm22; D5measured change for a nominal change of 10%

                                       A              B               r               C             D

BMD 4500                 Legs           0.22             0.77          0.990           0.99             7.7
                         Arms           0.11             0.88          0.998           0.99             8.9
                         Spine          0.08             0.82          0.986           0.90             9.1
BMD Expert               Legs           0.09             0.96          0.999           1.05             9.2
                         Arms           0.05             0.99          0.999           1.04             9.5
                         Spine          0.10             0.85          0.990           0.96             9.0
All intercepts (A) are significantly higher than zero and all slopes (B) are significantly lower than 1.0 (p,0.05)
BMC 4500                 Legs            (14)            0.90          0.999           663             (9.8)
                         Arms            (20)            0.88          0.996           283              9.3
                         Spine           (15)            0.64          0.994           144              8.7
BMC Expert               Legs             (1)            1.06          0.999           761           (10.0)
                         Arms           211             (1.00)         0.999           290            10.4
                         Spine          236              1.13          0.997           189            11.9
Nominal                  Legs                                                          720
                         Arms                                                          300
                         Spine                                                         200
Figures in parentheses are not significantly different from zero (A), unity (B) or 10.0 (D) (p.0.05)

4500, Hologic QDR 4500A; Expert, Lunar Expert.

proportion at the lowest BMD was lower than at               Table 4 as differences between the means and
any higher BMD. Above a BMD of 0.5 g cm22,                   correlations from linear regression equations. As
the fat proportion recorded by the 1000 was                  well as illustrating the differences, the latter serve
higher than that recorded by the 4500, which was             as cross-calibration formulae, essential in follow-
in turn greater than that measured by the Expert,            ing-up subjects initially measured on the 1000 and
the biggest disparity being 10% fat. With the 4500           subsequently on the 4500. The results were
and the Expert, there was a small but significant             analysed separately for women and men. No
( p,0.05) reduction of fat proportion with                   significant differences of the regression equations
increasing BMD. Measurement for the arms                     were found, so the combined results for all 41
differed from the legs for each instrument. The              subjects are presented. The measured mean BMD
whole phantom had been constructed from the                  and BMC are higher with the 1000 compared with
same batch of hardboard sheets, so no differences            the 4500 by up to 11%, and the slopes of the
of composition between the sections were likely.             regression equations are greater than unity, with
                                                             the notable exception of the head. While the
In vivo measurements                                         means and regression equations should provide
  Direct comparisons between the 1000 and the                sufficient information regarding the differences
4500 patients’ measurements are presented in                 between the instruments, these are further illu-
                                                             strated by plotting the differences between pairs
                                                             of results against the mean values for total body
                                                             BMD, BMC and per cent fat [16] (Figure 2).
                                                             There are significant biases for BMD and BMC
                                                             and significant slopes of regression lines for BMC
                                                             and fat proportion.
                                                                It is common in clinical practice to compare
                                                             individual results with normal reference ranges,
                                                             defining a T-score as the difference, expressed in
                                                             terms of the standard deviation, between the
                                                             measured BMD and the mean BMD of a young
                                                             normal subject in the reference range, and a
                                                             Z-score as the corresponding comparison with an
                                                             age-matched normal. The manufacturer uses the
                                                             same reference ranges for the 4500 and the 1000,
Figure 1. Variation of measured fat proportion in the        with standard deviations of 0.087 g cm22 for
limbs of the hardboard phantom with bone mineral
density (BMD) of the aluminium skeleton, measured            women and 0.095 g cm22 for men. The depar-
using the Lunar Expert scanner, the Hologic QDR              tures of our regression line from unity range from
4500A scanner or the Hologic QDR 1000W scanner.              0.036 g cm22 at 0.8 g cm22 to 0.056 g cm22 at

170                                                                The British Journal of Radiology, February 2001
Pencil beam and fan beam DXAs for total body bone and soft tissue measurement

                                                         phantom Equations (4) and (5) in the text. All
                                                         regressions had a negative intercept. The slopes
                                                         are very similar, and all significantly greater than
                                                            Results of the comparisons in vivo between the
                                                         Expert and the 4500 are presented in Table 5.
                                                         Again, correlations for males and females were
                                                         combined as there were no differences between
                                                         the regression equations. Most of the mean values
                                                         of BMC and BMD were higher for the Expert
                                                         than for the 4500 by up to 25%; the head was an
                                                         exception. In the regressions of Expert against
                                                         4500 there were significant positive intercepts for
                                                         all BMD values and for most BMC values. The
                                                         slopes were significantly below unity for BMD.
                                                         Bland–Altman plots [16] for the total body results
                                                         are presented in Figure 4. There were significant
                                                         biases for BMD and BMC, and a significant slope
                                                         for fat proportion.
                                                            Total body fat correlations between the Expert
                                                         and the 4500 in vivo are compared with the
                                                         phantom measurements in Figure 5. The two
                                                         comparisons in vitro have similar slopes and
                                                         intercepts, but the regression in vivo has a steeper
                                                         slope and a negative intercept.
                                                            Although the mean value of total body BMD
                                                         was only 3% higher for the Expert than the 4500,
                                                         the characteristics of the regression equation
                                                         meant that there was a 10% difference at the
Figure 2. Differences between total body bone            lowest BMD. The possible impact of this on
mineral density (BMD), bone mineral content (BMC)        clinical assessment of patients was examined by
and per cent fat measurements from the two Hologic       comparing T- and Z-scores using the manufac-
scanners, Hologic QDR 1000W and QDR 4500A,
plotted against mean values. #, Males; , females.
                                                         turers’ reference ranges. There were no mean
                                                         differences between the T- or Z-scores of the
                                                         Expert and the 4500, and the regression equations
1.2 g cm22, so the derivation of T- and Z-scores         were not significantly different from unity. The
without correction would lead to some disparities        calibration differences would not, therefore, lead
in diagnosis.                                            to different clinical assessments.
   The soft tissue results are presented as fat
proportions. Similar conclusions were reached if
the comparisons were based on fat masses, but the
results were then more subject to differences in
ROI selection, especially in the comparisons
between the 4500 and the Expert. For the 1000
and 4500 comparisons, there is no mean differ-
ence of total fat proportion. This finding results
from disparities in opposite directions in different
parts of the body by up to 20%, and a balance
between slopes that are substantially above unity
and negative intercepts in the regression equa-
tions. The fat proportion in the head is particu-
larly disparate, with an average difference of
   The total body comparisons between the 1000
and the 4500 in vivo are compared with the
                                                         Figure 3. Fat percentage measured by the Hologic
phantom measurements in Figure 3. The regres-            QDR 1000W plotted against that measured by
sion lines for the comparisons in vitro are taken        the Hologic QDR 4500A. VCP, variable composi-
from Table 2 and the variable composition                tion phantom.

The British Journal of Radiology, February 2001                                                          171
                                                                                      P Tothill, W J Hannan and S Wilkinson

Table 4. Comparisons between Hologic QDR 1000W (1000) and Hologic QDR 4500A (4500) in vivo: means,
ratios and parameters of linear regressions, 10005a+b64500 (n541)

                Mean for             Ratio                a                  b               r                  SE      SE%
                1000                 1000/4500

BMD (g cm22)
Total        1.083                   1.04                 20.006             1.053           0.975              0.028   2.6
Arms         0.739                   1.03                 20.090a            1.169b          0.974              0.025   3.4
Legs         1.220                   1.09                 20.020             1.112b          0.976              0.043   3.5
Trunk        0.846                   1.05                 0.024              1.022           0.974              0.028   3.3
Head         1.930                   0.96                 0.381a             0.769b          0.975              0.063   3.3
BMC (g)
Total        2166                    1.07                 2171a              1.156b          0.988              84      3.9
Arms         290                     ns                   228                1.104b          0.969              25      8.6
Legs         876                     1.11                 225                1.162b          0.979              53      6.1
Trunk        572                     1.10                 234                1.175b          0.985              26      4.5
Head         429                     0.97                 26                 0.911b          0.966              19      4.4
Per cent fat
Total        24.8                    ns                   26.9a              1.270b          0.996              1.04    4.2
Arms         30.7                    1.20                 0.5                1.228b          0.989              2.08    6.8
Legs         29.2                    ns                   24.7a              1.174b          0.994              1.38    4.7
Trunk        21.7                    0.94                 29.7a              1.361b          0.994              1.46    6.7
Head         14.8                    0.76                 25.6a              1.113           0.985              0.38    2.6

ns, not significantly different from unity (p.0.05).
  Significant difference of intercept from zero (p,0.05); bsignificant difference of slope from unity (p,0.05).
SE, standard error; BMD, bone mineral density; BMC, bone mineral content.

                                                                        The mean values of fat proportion are not
                                                                     significantly different, except for the arms, but the
                                                                     regression slopes are appreciably greater than
                                                                     unity, with some negative intercepts.

                                                                       The results of the experiments with varying
                                                                     depths of water presented in Equations (1–3),
                                                                     coupled with the relationship between weight and
                                                                     thickness, show that the 4500 fulfills reasonably
                                                                     well the claim by Hologic [17] that, provided the
                                                                     subject remains in contact with the couch, the
                                                                     mass is accurately recorded for all thicknesses. In
                                                                     contrast, the accuracy of mass determined by the
                                                                     Expert varies with water depth. With both fan

Figure 4. Differences between total body bone
mineral density (BMD), bone mineral content (BMC)
and per cent fat measurements from the two fan
beam scanners, Hologic QDR 4500A and Lunar                           Figure 5. Fat proportion measured by the Lunar
Expert, plotted against mean values. #, Males; ,
                                                              N      Expert plotted against that measured by the Hologic
                                                                     QDR 4500A. VCP, variable composition phantom.

172                                                                         The British Journal of Radiology, February 2001
Pencil beam and fan beam DXAs for total body bone and soft tissue measurement

beam scanners, raising the bath above the couch                      differences in calibration for soft tissue composi-
leads to substantial errors in mass determination                    tion between the 1000 and the 4500. The
(Table 1), although it is unlikely that this would                   fundamental calibration of DXA instruments
have much influence on normal scanning in vivo.                       stems from the relationship between the measured
More importantly, Table 1 shows that the height                      ‘‘r-values’’, the ratios of the low and high energy
of a bone within the body affects the accuracy of                    attenuations, for known standards such as stearic
BMC determination, although BMD is little                            acid and water. Hologic use such standards [20].
affected because BA changes similarly. These BA                      Other materials, such as acrylic plastic and
changes are, no doubt, associated with procedures                    aluminium, are used to monitor calibration
to correct for magnification, but there may also be                   during scanning. The same step phantom is
less valid links between BMC and BA, which we                        used for the 1000 and the 4500. Our phantom
have demonstrated previously [18, 19].                               measurements provide comparisons of the basic
   In practice, in vivo, the heights of the different                calibrations. The results in Table 2 suggest that in
components of the skeleton above the couch                           both comparative and absolute terms the 1000
range from around 2 cm to 20 cm, and with the                        underestimates low fat proportions and the 4500
fan beam scanners the accuracy of the BMC                            underestimates high ones. The different slopes are
measurements will vary correspondingly. Table 1                      also reflected in the VCP results in Equations (4)
suggests that, because of the different X-ray tube                   and (5), and in the comparisons in vivo in Table 4.
orientations, the Expert and the 4500 might differ                   They also explain the apparent discrepancies
in, for example, BMC measurements in the arms,                       between Table 2, where fat per cent is higher
where the bones are relatively near the couch.                       for the 4500 than the 1000 in the water-bath, and
Indeed, Table 5 shows that the Expert/4500 ratio                     Figure 1, where the converse is true for the
is lower for the arms than for the remainder of the                  hardboard phantom. Indeed, the regression equa-
body, in keeping with the above deductions. It is                    tion for total per cent fat in vivo in Table 4
not likely that magnification effects will affect the                 predicts a 1000 measurement of 6% fat compared
accuracy of BMC measurements in an individual                        with a 4500 value of 10%, and 44% fat 1000
during weight change, although there are other                       compared with 40% fat 4500, values not far from
characteristics of DXA that will do this [18, 19].                   those observed in the phantom measurements.
However, we have shown that there were                                  The final results of scanning in vivo depend on
significant differences of a few per cent between                     additional assumptions, regarding, for example,
some bone and fat measurements when a                                fat distribution. The calibration philosophy
volunteer was scanned prone and supine with                          adopted by Hologic is to adjust QDR 4500A
the Expert and the 4500 [13].                                        body composition results to provide measure-
   All of the experiments demonstrate consistent                     ments of per cent fat that are equivalent to

Table 5. Comparisons between Lunar Expert and Hologic QDR 4500A (4500) in vivo: means, ratios and param-
eters of linear regressions, Expert5a+b64500 (n521)

                 Mean                Ratio                   a                b               r                 SE     SE%
                 (Expert)            Expert/4500

BMD (g cm22)
Total        1.164                   1.04                    0.32a            0.75b           0.92              0.04   3.4
Arms         0.955                   1.28                    0.41             0.73            0.50              0.09   9.4
Legs         1.289                   1.12                    0.45a            0.72b           0.76              0.07   5.4
Trunk        0.961                   ns                      0.25a            0.76b           0.85              0.05   5.2
Head         1.872                   0.82                    0.73a            0.50b           0.89              0.11   5.9
BMC (g)
Total        2528                    1.10                    387a             0.93            0.96              105    4.2
Arms         341                     1.07                    33               0.97            0.92              27     7.9
Legs         994                     1.17                    195a             0.94            0.92              60     6.0
Trunk        775                     1.26                    145a             1.02            0.93              50     6.5
Head         419                     0.82                    No significant    correlation
Per cent fat
Total        29.5                    ns                      26.8a            1.27b           0.99              1.2    4.1
Arms         38.5                    1.27                    0.4              1.25b           0.98              3.4    8.8
Legs         32.9                    ns                      29.8a            1.28b           0.99              2.4    7.3
Trunk        27.6                    1.07                    24.6a            1.25b           0.98              2.6    9.4
Head         18.8                    ns                      No significant    correlation

ns, not significantly different from unity (p.0.05).
  Significant difference of intercept from zero (p,0.05); bsignificant difference of slope from unity (p,0.05).
SE, standard error; BMD, bone mineral density; BMC, bone mineral content.

The British Journal of Radiology, February 2001                                                                          173
                                                                      P Tothill, W J Hannan and S Wilkinson

per cent fat results generated by a multicompo-          four subjects had more than 35% fat [22]. From
nent model including body water estimates                the same centre, Evans et al [2] also compared a
by deuterium dilution, bone mass measurements            four-compartment hydrodensitometry technique
by DXA and body density measurements by                  with QDR 1000W DXA in 27 obese women. They
hydrodensitometry [11]. They do not indicate the         were primarily interested in measuring changes
basis of calibration of the QDR 1000W, but say           during dieting and reported only mean results of
that per cent fat results from the 4500 are 2–3%         fat proportions. When all six groups were
lower than earlier QDR scanners at a nominal             combined, the mean per cent fat was 43.0% and
value of 25%. By direct comparison, we found no          there was no significant difference between the
difference at this degree of fatness. Kelly [11] gives   DXA and four-compartment hydrodensitometry
no indication of whether the difference varies with      results. On balance, therefore, there is no evidence
fat proportion.                                          that the 1000 was inaccurate at high fat propor-
   The comparisons between a 4500 and a four-            tions in vivo. Coupled with our phantom results
compartment hydrodensitometry model have not             and comparisons between the 1000 and 4500
yet been reported in full. Visser et al [21] report      in vivo, we conclude that there may be an
only fat-free mass. There was a high correlation         underestimate of high fat proportions by the
between the two results (r50.99), but there was a        4500, but that more evidence is required.
mean fat-free mass 1.8 kg higher by DXA. Total              Few other direct comparisons of total body
body mass by DXA was 1.1% higher than body               measurements in vivo between a QDR 1000W and
weight, so presumably there was good agreement           a QDR 4500 have been reported. Bouyoucef et al
on mean fat mass. No fat proportions are quoted.         [24] included measurements of total body bone
While errors in lean mass and fat mass are likely        mineral on seven subjects, but no regional
to be highly correlated, the data on fat-free mass       measurements or soft tissue results. There are
presented by Visser et al [21] are not incompatible      disparities between their quoted regression equa-
with the discrepancies of fat proportion that we         tion and the plot, but it appears that total body
have shown.                                              BMC and BMD were about 3% higher for the
   The underestimates of low fat proportions by          4500 than the 1000. This is in contrast with our
the 1000 in the phantom experiments are sup-             findings that the reverse was true, and emphasizes
ported by our experience that some anorexic              the fact that such comparisons can only apply to
subjects have recorded apparently negative per           the actual instruments studied. In an abstract
cent fat. Prior et al [22] have compared body            from conference proceedings, Fuerst and Genant
composition measurements by a Hologic                    [25] report comparisons in vivo between a 4500
QDR1000W with a four-compartment hydroden-               and a Hologic QDR 2000, used in pencil beam
sitometry model in 172 young women and men,              mode. High correlations were found. For BMD,
using comparable software to ourselves. They             the slope for 4500/2000 was 0.97 (r50.98). For per
found that there was a very good agreement of            cent fat, the slope for 4500/2000 was 0.95
mean values, but that the 1000 underestimated fat        (r50.99). While the identity between results
in the leanest women and overestimated it in             from the QDR 2000 and QDR 1000W cannot
the fattest, the regression equation being               be guaranteed, these results support our findings
4C50.8561000+3.30, where 4C is the per cent              shown in Table 4 and Figure 4. With an under-
fat from the four-compartment model. This lends          estimate of 5% fat for the 4500 relative to the
support to our conclusions regarding the calibra-        2000, Fuerst and Genant [25] comment that the
tion of the 1000 at low fat proportions. The Prior       manufacturer was investigating this apparent
regression equation for men was less disparate,          incorrect calibration.
4C50.9061000+0.75, the intercept being signifi-              There are more uncertainties regarding soft
cantly different from that for women.                    tissue calibration of the Expert. Measurements
   Arngrimsson et al [23], from the same group,          with the smaller boxes and blocks reported in
included comparisons between a Hologic QDR               Table 2 show results intermediate between those
1000W and a four-compartment model in a study            from the 4500 and the 1000, with values not far
of relatively lean subjects of runners and controls.     from nominal. However, they do not agree with
The 1000 consistently underestimated per cent fat,       the composition in vivo (Figure 5).
on average by 2.8% fat. All the evidence thus               Results of varying the BMD of the limbs and
suggests that the 1000 underestimates low fat            spine (Table 3) may be compared with those
proportions.                                             already published for the Hologic QDR 1000 [7].
   Our conclusion from the phantom measure-              The 4500 results are similar to those from the
ments that the 4500 underestimates high fat              1000. An increase in BMD of 10% is under-
proportions is less well supported. The Prior            estimated by both instruments. There is a similar
regression equation predicts that the 1000 over-         BMD threshold in the legs, which is not so severe
estimates fat proportion in fatter women, but only       with the 4500 as with the 1000. The Lunar Expert

174                                                            The British Journal of Radiology, February 2001
Pencil beam and fan beam DXAs for total body bone and soft tissue measurement

gave results closer to the nominal changes than          shown that corrections by the manufacturer to
either of the Hologic scanners in measuring BMD          deal with magnification effects in fan beam
changes. BMD measurements with the Lunar                 absorptiometers have been reasonably, but
Expert were higher than with either Hologic              not completely, effective. There are calibration
scanner, in keeping with the well known Lunar/           differences between the instruments, but the
Hologic bone calibration differences. The Expert         comparisons in vivo provide a means of cross-
results can also be compared with measurements           calibration. Some anomalies remain, which would
of the same phantom using a pencil beam Lunar            be worthy of more attention by the manufac-
DPX presented in an earlier publication [7]. The         turers. It must be remembered that the results
main difference is that the Expert bone results          apply only to the three particular scanners
were unreliable below a BMD of 0.5 g cm22, a             examined. We have no evidence regarding agree-
higher limit than for the Lunar DPX. As a                ment with other models, nor indeed other
relatively simple bone phantom was used, these           examples of the same models, or the effect of
results may not be fully representative of the           different software.
situation in vivo, but they highlight the limitations
and differences of the scanners.
   A by-product of the variable BMD experiment
was the soft tissue composition results presented
in Figure 1. Percentage fat differences recorded by         We are grateful to Professor DM Reid for
the three instruments are in accordance with the         encouragement and access to the facilities of the
calibration experiments, being up to 10% fat             Osteoporosis Research Unit in Aberdeen, and to
between the 1000 and the Expert for the legs. The        Carol Millar and Charles Sidey for skilled
graphs also demonstrate that the measurement of          technical assistance in Edinburgh.
per cent fat varies with position in the body, the
arms showing a higher value than the legs with
the Hologic instruments and a lower value with           References
the Expert. The apparent dependence of per cent           1. Van     Loan     MD.    Is   dual-energy    X-ray
fat on the BMD, although inaccurate, is under-               absorptiometry ready for prime time in the
standable at the lowest levels of BMD.                       clinical evaluation of body composition? Am
Unrecognized bone will contribute to the lean                J Clin Nutr 1998;68:1155–6.
                                                          2. Evans EM, Saunders MJ, Spano MA, Arngrimsson
component and lower the recorded value. There is
                                                             SA, Lewis RD, Cureton KJ. Body composition
not such an obvious explanation for the smaller              changes with diet and exercise in obese women: a
but significant reduction of per cent fat with                comparison of estimates from clinical methods and
increasing BMD above 0.5 g cm22 with the                     a 4-component model. Am J Clin Nutr 1999;70:
Expert and the 4500.                                         5–12.
                                                          3. Pichard C, Kyle UG, Gremion G, Gerbase M,
   Comparisons of bone and fat measurements
                                                             Slosman DO. Body composition by x-ray
in vivo in Tables 4 and 5 demonstrate the need for           absorptiometry and bioelectrical impedance in
and value of such cross-calibration. For the two             female runners. Med Sci Sports Exerc 1997;
Hologic scanners, the standard errors of 3% for              29:1527–34.
total body BMD and 4% for BMC, while greater              4. Rankinen T, Lyytikainen S, Vanninen E, Penttila I,
                                                             Rauraman R, Uusitupa M. Nutritional status of the
than the precision obtainable with either scanner,
                                                             Finnish elite ski jumpers. Med Sci Sports Exerc
should allow some longitudinal measurements to               1998;30:1592–7.
span a change of instrument using the cross-              5. Genant HK, Grampp S, Gluer CC, Faulkner KG,
calibration equations. Standard errors for the               Jergas M, Engelke K, et al. Universal
Expert/4500 comparisons are rather higher, but it            standardization for dual X-ray absorptiometry:
is less likely that longitudinal studies would               patient and phantom cross-calibration results.
                                                             J Bone Miner Res 1994;9:1503–14.
involve such a change of brands. The comparisons          6. Tothill P, Fenner JAK, Reid DM. Comparisons
reported in Table 4 show that the results for the            between three dual-energy X-ray absorptiometers
head are markedly different from those for the               used for measuring spine and femur. Br J Radiol
rest of the body. It has to be remembered that the           1995;68:621–9.
fat proportion in the brain cannot be measured by         7. Tothill P, Avenell A, Reid DM. Precision and
                                                             accuracy of measurements of whole-body bone
DXA as it is surrounded by the skull; thus it has            mineral: comparisons between Hologic, Lunar and
to be assumed and this assumption may also                   Norland dual-energy X-ray absorptiometers. Br
affect bone measurements. The disparities suggest            J Radiol 1994;67:1210–7.
to us that Hologic changed their assumptions in           8. Tothill P, Avenell A, Love J, Reid DM.
this respect, perhaps as part of a procedure to              Comparisons between Hologic, Lunar and
                                                             Norland dual-energy X-ray absorptiometers
adjust the total body fat calibration to agree with          and other techniques used for whole-body soft
other techniques.                                            tissue measurements. Eur J Clin Nutr 1994;48:
   In conclusion, the phantom measurements have              781–94.

The British Journal of Radiology, February 2001                                                            175
                                                                        P Tothill, W J Hannan and S Wilkinson

 9. Van Loan MD, Keim NL, Berg K, Mayclin PL.             18. Tothill P, Hannan WJ, Cowen S, Freeman CP.
    Evaluation of body composition by dual energy x-ray       Anomalies in the measurement of total-body bone
    absorptiometry and two different software packages.       mineral by dual-energy X-ray absorptiometry
    Med Sci Sports Exerc 1995;27:587–91.                      during weight change. J Bone Miner Res 1997;
10. Patel R, Blake GM, Herd RJM, Fogelman I. The              12:1908–21.
    effect of weight change on DXA scans in a 2-year      19. Tothill P, Laskey MA, Orphanidou CI. Anomalies
    trial of etidronate therapy. Calcif Tissue Int            in dual-energy X-ray absorptiometry measurements
    1997;61:693–9.                                            of total-body bone mineral during weight change
11. Kelly TL. Body composition determinations using           using Lunar, Hologic and Norland instruments. Br
    high speed fan-beam technology. Hologic Insights          J Radiol 1999;72:661–9.
    1999;9(2):14–7.                                       20. Kelly TL, Berger N, Richardson TL. DXA body
12. Steel SA, Baker AJ, Saunderson JR. An assessment          composition: theory and practice. Appl Radiat Isot
    of the radiation dose to patients and staff from a        1998;49:511–3.
    Lunar Expert XL fan-beam densitometer. Physiol        21. Visser M, Fuerst T, Lang T, Salamone L, Harris
    Meas 1998;19:17–26.                                       TB. Validity of fan-beam dual-energy X-ray
13. Tothill P, Hannan WJ. Comparisons between                 absorptiometry for measuring fat-free mass and
    Hologic QDR 1000W, QDR 4500A and Lunar                    leg muscle mass. J Appl Physiol 1999;87:
    Expert DXA scanners used for measuring total-             1513–20.
    body bone and soft tissue. Ann N Y Acad Sci           22. Prior BM, Cureton KJ, Modlesky C, Evans EM,
    2000;904:63–71.                                           Sloniger MA, Saunders M, et al. In vivo validation
14. Nord RH, Payne RK. Standards for body                     of whole body composition estimates from dual-
    composition in DEXA. In: Ring EFJ, editor.                energy X-ray absorptiometry. J Appl Physiol 1997;
    Current research in osteoporosis and bone mineral         83:623–30.
    measurement. London: British Institute of             23. Arngrimsson SA, Evans EM, Saunders MJ,
    Radiology, 1990:27–8.                                     Ogburn CL III, Lewis RD, Cureton KJ.
15. Formica C, Nieves J, Dixon J, Munn L, Lindsay R,          Validation of body composition estimates in
    Miller CG. A DXA body composition cross-                  male and female distance runners using estimates
    calibration phantom. Osteoporosis Int 1997;7:             from a four-compartment model. Am J Hum Biol
    278 (abstract).                                           2000;12:301–14.
16. Bland JM, Altman DG. Statistical methods for          24. Bouyoucef SE, Cullum ID, Ell PJ. Cross-calibration
    assessing agreement between two methods of                of a fan-beam X-ray densitometer with a pencil-
    clinical measurement. Lancet 1986;i:307–10.               beam system. Br J Radiol 1996;69:522–31.
17. Kelly TL, Shepherd JA, Steiger P, Stein JA.           25. Fuerst T, Genant HK. Evaluation of body
    Accurate body composition assessment using fan-           composition and total bone mass with the
    beam DXA: technical and practical considerations.         Hologic QDR 4500. Osteoporosis Int 1996;6(S1):
    J Bone Miner Res 1997;12:S269 (abstract).                 203 (abstract).

176                                                             The British Journal of Radiology, February 2001

To top