Body composition analysis by dual energy X-ray absorptiometry in female diabetics differ between manufacturers

Transcription

1 European Journal of Clinical Nutrition (1997) 51, 449±454 ß 1997 Stockton Press. All rights reserved 0954±3007/97 $12.00 Body composition analysis by dual energy X-ray absorptiometry in female diabetics differ between manufacturers Center for Clinical and Basic Research, Ballerup, Denmark, and Steno Diabetes Center, Niels Steensens Vej 2, 2820 Gentofte, Denmark Objective: Comparison of body composition results by two dual energy X-ray absorptiometry (DXA) instruments, namely QDR-2000 from Hologic Inc and DPX from Lunar in subpopulations of lean and obese subjects. Design: Cross-sectional study with 85 female diabetics (BMI 18±43 kg/m 2 ) measured with both DXA instruments. Results: The regression lines for fat tissue mass (FTM), FAT% and total body bone mineral content (TBMC), but not lean tissue mass (LTM), were different from the line of identity (P < 0.01). However, the relationships were high (r 2 > 0.95), and the corresponding SEE%'s were low (0.8±4.8%), and were independent of BMI. FTM and FAT% measured by the QDR-2000 were 10% higher, and LTM and TBMC 6% lower, than by DPX (P < 0.001). Conclusions: There were lack of agreements between total body composition results by DPX Lunar, and QDR Hologic Inc. Individual results on the two systems cannot be directly compared. Standardization of body composition measurements by DXA is strongly needed. Descriptors: body composition; dual energy X-ray absorptiometry; diabetes Introduction The advent of dual energy X-ray absorptiometry (DXA), which is increasingly used in clinical and research settings, has improved the precision and accuracy of body composition measurements (Mazess et al, 1990; Svendsen et al, 1993; Haarbo et al, 1991; Lukaski, 1993; Fuller et al, 1992). However, there are still matters concerning the comparability and of DXA that are unsolved. The largest manufacturers of DXA, namely Hologic Inc, Norland and Lunar Rad. Corp, use different hardware and software arrangements for their calibration, data collection and analysis, and DXA instruments from the different manufacturers do not seem to offer directly comparable body composition results (Pritchard et al, 1992; Tothill et al, 1994). The few reports of comparisons of body composition results by DXA instruments from the different manufacturers have primarily focused on healthy normal weight volunteers (Pritchard et al, 1992; Tothill et al, 1994). We have previously found that body composition measurements in adults by a Lunar DPX machine were accurate (Svendsen et al, 1993; Haarbo et al, 1991), whereas the in vivo accuracy of body composition measurements in adults by DXA instruments from Hologic Inc and Norland has not been reported yet. The accuracy of body composition measurements by DXA seems to depend on the composition and mass of soft tissue (Laskey et al, 1992; Jebb et al, 1995). Thus, the comparability of body composition results by DXA instruments from different manufacturers may differ with body size and fatness, and may be different in Correspondence: Dr CN Kistorp, Center of Clinical and Basic Research, Ballerup Byvej 222, DK-2750 Ballerup, Denmark. Received 1 November 1996; revised 4 March 1997; accepted 14 March 1997 normal weight, obese and lean people. In order to investigate this, we compared body composition results by DPX (Lunar) with QDR-2000 (Hologic Inc) in subpopulation of more lean subjects, namely female insulin dependent diabetes mellitus (IDDM) patients, and of more obese subjects, namely female non insulin dependent diabetes mellitus (NIDDM) patients. Methods With the women wearing light indoor clothes and no shoes, weight was measured to nearest 0.1 kg. Total body composition was measured with two total body DXA scanners, namely DPX (Lunar Radiation Corporation, Madison, WI), software version 3.6y, medium scan mode, extended research analysis, and QDR-2000 X-ray bone densitometer, (Hologic Inc, Waltham, MA, fan beam, software version 5.54A). Regional analysis were made by routine according to the respective operators manual. The measurements were completed on the same day for each subject. Total body bone mineral content (TBMC) and density (TBMD), the total body soft tissue mass (STM) and the fat percentage (FAT%) and the fat (FTM) and lean soft tissue mass (LTM) were measured. The FTM is not solely adipose tissue, but the sum of the fatty elements of the soft tissue. Similarly the LTM is not an anatomical entity, but represents the sum of chemical fat free soft tissue elements. Subjects Eighty- ve female diabetes patients from the Steno Diabetes Center outpatient clinic participated in this study. Thirty-two had NIDDM: age: 56.8 (11.8) y (mean(s.d.)); BMI: 28.9 (5.7) kg/m 2. Fifty-three had IDDM: age: 49.9

2 450 Figure 1 Relationships (left) and lack of agreement (right) between body compositions measurements by DPX (Lunar Radiation Corp) and QDR-2000 (Hologic Inc). Broken lines on the left illustrates the lines of identity. Closed circles, NIDDM patients; triangles, IDDM patients.

3 (14.8) y; BMI: 23.2 (2.9) kg/m 2. All of the subjects were relatively healthy, that is none of the female diabetics suffered from heart failure, renal disease or were using diuretics. The study was carried out in accordance with the Declaration of Helsinki II, and with approval of the ethical committee of Copenhagen County. Statistical analysis Linear regression analysis were used to determine the relationship between the two DXA machines. The lack of agreements between QDR and DPX were plotted as a function of the means, according to the statistical procedure of Bland & Altman (1986). Analysis were performed with the SAS computer software package (version 6.0: SAS Institute Inc. Cary NC, 1989). Results The r 2 of the linear regressions between weight measured by a scale weight and the weight calculated by QDR-2000 or DPX were both 0.99, and the corresponding SEE's were 0.7 kg and 0.9 kg, respectively. The mean weight was 0.03 kg higher by QDR-2000 and 0.2 kg lower by DPX than by the scale weight (not signi cantly different from zero). Figure 1 illustrates the linear regressions of FTM, LTM and FAT% and Table 1 gives the relationships between all of the total body components measured by QDR-2000 and DPX. The determination coef cients r 2 were above 0.95 for all comparisons, ranging from 0.95± 0.99, and the corresponding SEE% ranging from 0.8±4.8%. The SEE's of the regression lines were small and were similar along the regression line, that is in lean and obese subjects (Figure 1). Furthermore, there were no signi cant association between the residuals and the size of any of the body components (P > 0.05). The slopes of all the regression lines were signi cantly different from one (P < 0.01), and the regression lines for weight, FTM, FAT%, STM and TBMC were signi cantly different from the line of identity (P < 0.05). There were lack of agreement for all the body components between the DPX and QDR-2000 (Table 1 and Figure 1). Thus the FAT% and FTM measured by the QDR-2000 were about 10% higher (3.7% and 2.5 kg, respectively) and the LTM 5% (2.1 kg) lower than by DPX (P < 0.05). The TBMC and TBMD were 6% and 3.4%, respectively, lower by QDR-2000 than by DPX (P < 0.05). The difference between the two instruments in weight, FTM and STM, but not LTM or FAT%, were signi cantly correlated to the average values of the respective parameters (Table 1 and Figure 1), and for weight, FTM and LTM also with body mass index (BMI) (P < 0.01). The waist to hip ratio (WHR) and the FAT% was signi cantly higher in the NIDDM patients compared to the IDDM (0.91 and 42.5% vs 0.82 and 31.4%, respectively, P < 0.05). Linear regressions revealed that the relationships between DPX and QDR-2000 were independent of weight, BMI and WHR, but were signi cantly (P < 0.05) dependent of the type of diabetes (dummy code: NIDDM:0, IDDM:1) and the FAT%. When both the type of diabetes and the FAT% were entered in the regression models, only the FAT%, but not the type of diabetes, remained signi cant (P < 0.05). Separate linear regression analyses of DPX versus QDR-2000 were performed for all the body components for the IDDM and the NIDDM subpopulation, respectively (Table 2). The regression equations for the IDDM were statistically signi cantly different from those for NIDDM for all the body components (P < 0.01), except for TBMC. However, the determination coef cients remained rather similar, ranging from 0.93±0.99, and the SEE's for the pooled population, and the NIDDM and the IDDM subpopulation, respectively were also quite similar. There was also signi cantly lack of agreement (P < 0.01) for all the body components, (except for weight in IDDM) in both subpopulation between the two DXA instruments (Table 2). The mean differences for all the body components between the two instruments were all signi cantly different in the NIDDM and in the IDDM subpopulation (P < 0.01). Table 3 gives regional comparisons between DPX and Hologic Discussion The present study examined comparability of body composition measurements by two DXA instruments from two 451 Table 1 Linear regressions and lack of agreement between DPX (Lunar Rad. Corp) and QDR-2000 (Hologic Inc) Linear regression Agreement Equation P a SEE r 2 SEE% DPX-QDR (mean 2s.d.) r b Weight (kg) QDR ˆ 1.03DPX P < c d DPX ˆ 0.98QDR STM (kg) QDR ˆ 1.02DPX P < c d DPX ˆ 0.98QDR FTM (kg) QDR ˆ 1.06DPX 1.2 P < c d DPX ˆ 0.93QDR LTM (kg) QDR ˆ 0.95DPX 0.1 NS c 0.16 NS DPX ˆ 1.01QDR Fat% QDR ˆ 0.99DPX 3.9 P < c NS DPX ˆ 0.97QDR TMBC (g) QDR ˆ 1.06DPX P < c d DPX ˆ 0.92QDR TBMD (g/cm 2 ) QDR ˆ 0.98DPX NS c NS DPX ˆ 0.97QDR a Different from the line of identity. b Correlation between the difference of DPX minus QDR and the average value of DPX and QDR. c Signi cantly different from zero (P < 0.05). d P < NS P > 0.05.

4 452 Table 2 Separate linear regressions and lack agreement between DPX (Lunar Rad. Corp) and QDR-2000 (Hologic Inc). For IDDM and NIDDM subpopulations Linear regression Agreement Equation P a SEE r 2 SEE% DPX-QDR (mean 2s.d.) r b Weight (kg) IDDM QDR ˆ 0.99DPX 0.1 NS NS 0.04 NS NIDDM QDR ˆ 1.02DPX NS c d STM (kg) IDDM QDR ˆ 0.99DPX 0.4 NS c 0.09 NS NIDDM QDR ˆ 1.02DPX NS c d FTM (kg) IDDM QDR ˆ 1.02DPX 1.8 P < c NS NIDDM QDR ˆ 1.06DPX 1.4 P < c d LTM (kg) IDDM QDR ˆ 0.99DPX NS c NS NIDDM QDR ˆ 0.91DPX 1.2 NS c 0.46 d Fat% IDDM QDR ˆ 0.95DPX 4.7 P < c 0.08 NS NIDDM QDR ˆ 0.95DPX 6.2 P < c 0.18 NS TMBC (g) IDDM QDR ˆ 1.04DPX P < c d NIDDM QDR ˆ 1.10DPX P < c d TBMD (g/cm 2 ) IDDM QDR ˆ 0.98DPX 7 0 NS c 0.04 NS NIDDM QDR ˆ 1.03DPX 7 0 NS c NS a Different from the line of identity. b Correlation between the difference of DPX minus QDR and the average value of DPX and QDR. c Signi cantly different from zero (P < 0.05). d P < NS P > different manufacturers in populations of rather lean and obese females, namely in NIDDM and IDDM patients, and whether body size or fatness affect this comparability. The correlations between the body composition measurements by QDR-2000 and DPX were high (r 2 > 0.95). However, there were lack of agreements between the two instruments in all the body components measured. Thus, the TBMC and TBMD were signi cantly higher by DPX than by the QDR-2000 (3% and 6%, respectively), which is in accordance with a previous study by Laskey et al (1991). On the other hand, we found that the FAT% measured by DPX was 3.7% lower than by QDR At rst sight these results are not in agreement with previous results. Both Tothill et al (1994) and Pritchard et al (1992) found Table 3 Linear regressions and lack of agreement between DPX Lunar and QDR-2000 Hologic. Regional body composition Linear regression Agreement Equation P a SEE r 2 SEE% DPX-QDR (mean 2s.d.) r b STM (kg) Limb pooled QDR ˆ 0.83DPX 3.7 P < c 0.62 d IDDM QDR ˆ 0.91DPX 1.9 P < c 0.33 d NIDDM QDR ˆ 0.82DPX 3.7 P < c 0.69 d Trunk pooled QDR ˆ 1.18DPX P < c d IDDM QDR ˆ 1.06DPX NS c NIDDM QDR ˆ 1.21DPX P < c FTM (kg) Limb pooled QDR ˆ 0.86DPX 3.3 P < c 0.45 d IDDM QDR ˆ 0.98DPX 2.4 P < c NS NIDDM QDR ˆ 0.83DPX 3.3 P < c 0.54 d Trunk pooled QDR ˆ 1.30DPX P < c d IDDM QDR ˆ 1.10DPX P < NS d NIDDM QDR ˆ 1.38DPX P < c d LTM (kg) Limb pooled QDR ˆ 0.83DPX 0.2 NS c 0.41 d IDDM QDR ˆ 0.85DPX NS c 0.29 NS NIDDM QDR ˆ 0.82DPX 0.2 NS c 0.54 d Trunk pooled QDR ˆ 0.97DPX 1.3 NS c NS IDDM QDR ˆ 1.06DPX70.4 NS c d NIDDM QDR ˆ 0.91DPX 2.3 P < c 0.22 NS Fat% Limb pooled QDR ˆ 0.87DPX 12.9 P < c 0.10 NS IDDM QDR ˆ 0.88DPX 12.1 P < c 0.03 NS NIDDM QDR ˆ 0.74DPX 18.6 P < c 0.26 NS Trunk pooled QDR ˆ 1.21DPX P < NS d IDDM QDR ˆ 1.07DPX P < c d NIDDM QDR ˆ 1.29DPX P < c d a Different from the line of identity. b Correlation between the difference of DPX minus QDR and the average value of DPX and QDR. c Signi cantly different from zero (P < 0.05). d P < NS P > 0.05.

5 Table 4 Difference in Fat% between DXA from Lunar and Hologic Inc. with advancing software and instrument development Lunar Hologic Inc Fat% 453 n DPX QDR-1000 W QDR-2000 W Pencil beam Pencil beam Pencil beam software version software version software version Fan beam software version Diff. a left-right Accumulated diff. b Lunar-Hologic Tothill et al (3.6) 5.35 (5.51p) 3.7% Pritchard et al % 3.6% Mazess et al R±3.6R 0.2% Van Loan et al R±3.6R 72.0% 70.9% Abrahamsen et al % 71.2% A 76.4% Spector et al A 74.3% 75.4% 73.9% The present study y 5.54A 73.7% a Fat% by software version and DXA instrument on the left minus that on the right. b Accumulated difference in Fat% between DPX, software version 3.4 and QDR-2000 W, software version 5.54A, calculated from the reports in the Table. that FAT% measurements by DPX was about 3.7% higher than by the QDR However, when interpreting body composition results by DXA, attention must be given not only to which manufacturer the DXA instrument comes from, but also to the software version, to the generation of the DXA instrument, and to whether pencil or fan beam have been used. New software versions may cause changes in body composition results (Van Loan et al, 1995). For Hologic Inc the introduction of fan beam in QDR-2000 has increased the FAT% signi cantly (Abrahamsen et al, 1995). Table 4 shows the impact of the introduction and development of new software and DXA machines on the difference in measured FAT% between DPX from Lunar and QDR from Hologic. The accumulated effect of the introduction of new software and instruments generations is a shift in the difference in FAT% between DPX and QDR at 3.7% to 73.9%, which is more in accordance with our nding at 73.7%. The fact that the DXA results vary with manufacturer, software version, generation of DXA scanner, scan and beam mode and calibration procedure means that individual results on the two systems cannot be directly compared. The regression equations found were independent of body size (weight, BMI) but were statistically signi cantly dependent on the type of diabetes. A reasonable explanation could be the difference in fat distribution, namely the waist to hip ratio between the IDDM and the NIDDM patients. However, our results suggest that it is the difference in fatness between IDDM and NIDDM that causes this. However, speci c regression equations for the IDDM or NIDDM subpopulation did not improve the r 2 or the SEE's signi cantly as compared to the regression equations for the pooled population of IDDM and NIDDM patients. Thus, to compare body composition results by DPX, software version 3.6y, and by QDR-2000, fan beam and software version 5.54A, the regression equations given in the pooled population can be used (Table 1). These regression equations seem valid in females with a BMI ranging from 18±43 kg/m 2. However, they do introduce an extra nonsystematic error of 0.8±4.8% (SEE%), which must be considered and accounted for when comparing cross-sectional data from different instruments or when using different instruments in longitudinal clinical trials. We nd the results of the differences in the regional body composition results by DPX and Hologic-2000 dif cult to interpret. We do not know the accuracy error with neither DPX or Hologic-2000 of the regional body composition results, as we do it for total body. The optional standard de nitions of the limb and trunk regions seems to be the same by DPX and the Hologic DXA do only measure the soft tissue composition directly in pixels without bone. Thus the trunk with a high bone to non bone pixel ratio (pelvis, spine, ribs and sternum) may have a higher accuracy error than the limbs. On the other hand, the soft tissue composition is not measured if the height of the absorber is too low, as in hands and feet. Discrepancies in the approach to deal with these and related issues may, besides the calibration, cause different accuracy errors and absolute values of the regional body composition results. Interestingly we found the greatest lack of agreement in the FAT% in the limbs, not in the trunk. Conclusions There were lack of agreements ranging from 3±10% between total body composition results by DPX Lunar, and QDR-2000 Hologic Inc. However the determination coef cients were high and SEE's of the linear regressions of the results between the two DXA instruments were small. The manufacturers of the DXA instruments should standardize and intercalibrate their instruments. AcknowledgementsÐWe thank Drs Christian Hassager and Claus Christiansen, Center for Clinical and Basic Research, as well as Dr Hans-Henrik Parving, Steno Diabetes Center, for their valuable help and inspiring discussions. References Abrahamsen B, Gram J, Hansen TB & Beck-Nielsen J (1995): Cross calibration of QDR-2000 and QDR-1000 dual energy X-ray densitometers for bone mineral and soft-tissue measurements. Bone 16, 385± 390. Bland JM & Altman DG (1986): Statistical methods for assessing agreement between two methods of clinical measurement. Lancet II, 307± 310. Fuller NJ, Laskey MA & Elia M (1992): Assessment of the composition of major body regions by dual energy X-ray absorptiometry (DEXA), with special reference to limb muscle mass. Clin. Physiol. 12, 253±266. Haarbo J, Gotfredsen A, Hassager C & Christiansen C (1991): Validation of body composition by dual energy X-ray absorptiometry (DEXA). Clin. Physiol. 11, 331±341. Jebb SA, Goldberg GR, Jennings G & Elia M (1995): Dual energy X-ray absorptiometry measurements of body composition: effects of depth and tissue thickness, including comparisons with direct analysis. Clin. Sci. 88, 319±324. Laskey MA, Flaxman ME, Barber RW, Trafford S, Hayball MP, Lyttle KD, Crisp AJ & Compston JE (1991): Comparative performance in vitro and in vivo of Lunar DPX and Hologic QDR-1000 dual energy X- ray absorptiometers. Br. J. Radiol. 64, 1023±1029.

6 454 Laskey MA, Lyttle KD, Flaxman ME & Barber RW (1992): The in uence of tissue depth and composition on the performance of the Lunar dual energy X-ray absorptiometer whole-body scanning mode. Eur. J. Clin. Nutr. 46, 39±45. Lukaski HC (1993): Soft tissue composition and bone mineral status: evaluation by dual energy X-ray absorptiometry. J. Nutr. 123, 438±443. Mazess RB, Barden HS, Bidek JP & Hanson J (1990): Dual energy X-ray absorptiometry for total body and regional bone mineral and soft tissue composition. Am. J. Clin. Nutr. 51, 1106±1112. Mazess RB, Bisek J, Trempe J & Poutchot S (1991): Effects of tissue thickness on body composition measurements using dual energy X-ray absorptiometry. Madison WI: Lunar Corporation. Pritchard JE, Nowson CA, Strauss BJ, Carlson JS & Kaymakci B (1992): Evaluation of dual energy X-ray absorptiometry as a method of measurement of body fat. Eur. J. Clin. Nutr. 47, 216±228. SAS Institute Inc (1989): Guide for personal computers, version 6 edition. Gary, North Carolina: SAS Institute Inc. Spector E, LeBlanc A & Shackelford L (1995): Hologic QDR 2000 whole body scans: A comparison of three combinations of scan modes and analysis software. Osteoporosis Int. 5, 440±445. Svendsen OL, Haarbo J, Hassager C & Christiansen C (1993): Accuracy of measurement of body composition by dual energy X-ray absorptiometry in vivo. Am. J. Clin. Nutr. 57, 605±608. Tothill P, Avenell A, Love J & Reid DM (1994): Comparison between Hologic, Lunar and Norland dual energy X-ray absorptiometers and other techniques used for whole body soft tissue measurements. Eur. J. Clin. Nutr. 48, 781±794. Van Loan MD, Keim NL, Berg K & Mayclin PL (1995): Evaluation of body composition by dual energy X-ray absorptiometry and two different software packages. Med. Sci. Sports Exerc. 27, 587±591.