- Email: [email protected]
Bone Assessment in Children
Journal of Clinical Densitometry, vol. 8, no. 4, 445–453, 2005 © Copyright 2005 by Humana Press Inc. All rights of any nature whatsoever reserved. 1094-6950/05/8:445–453/$30.00 DOI: 10.1220/1094-6950
Original Article
Bone Assessment in Children Comparison of Fan-Beam DXA Analysis R. J. Shypailo and K. J. Ellis USDA/ARS Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, TX
Abstract The newest version of whole body dual-energy X-ray absorptiometry (DXA) analysis software from Hologic (Discovery 12.1) is designed to enhance bone detection in smaller subjects. We re-analyzed 1127 pediatric scans (ages 1.8–18.5 yr) previously analyzed using software version 11.2. Regression analysis compared new and original results for bone area (BA), bone mineral content (BMC), bone mineral density (BMD), and DXA-derived body weight. Changes in total and regional bone results were compared with age, weight, and height. New results were highly correlated with original analyses (R2 0.9), but there were large differences at the individual subject level. The BA and BMC values increased in subjects less than 40 kg weight, resulting in a lower BMD. Original and new results were equivalent by about age 14 yr in both genders. Regional bone data showed the greatest changes in the legs. The newest software produces significant changes in bone values in subjects weighing less than 40 kg, compared with earlier versions. This effect increases with decreasing body weight. This will impact interpretation of longitudinal pediatric DXA studies, as well as existing pediatric whole body bone reference databases. Investigators must recognize which DXA software version they are using, and which version produced any reference database they may use for comparison. Key Words: DXA; body composition; bone densitometry; clinical/pediatrics; growth and development.
Every stage of DXA technological development, whether in instrumentation or analysis, may produce changes in bone mineral results. The early pencil-beam instruments (Hologic® QDR1000 and QDR-2000; Lunar DPX and DPX-L) have given way to newer fan-beam devices (Hologic QDR4500; Lunar Prodigy) that require more assumptions in the reconstruction of the images and bone data. Investigators comparing the performance of these technologies almost invariably found significant differences in the estimates for bone. Results were usually highly correlated, hence conversion equations were suggested and promulgated (6–9). In conjunction with advancements in scanning techniques came changes in analysis algorithms. Again, investigators comparing analysis versions usually found highly correlated but statistically significant differences, although they were often of smaller magnitude with less clinical importance. Even so, conversion equations were still usually called for (10–13). Maintaining consistency of bone and body composition results is critical because of the importance of establishing reliable normative databases for the interpretation of DXA
Introduction Dual-energy X-ray absorptiometry (DXA) has become a reliable tool for the measurement of bone mineral in vivo (1–4). It has good precision, and operates at a low radiation dose, with ever decreasing scan times resulting from advancements in technology (1,2). A direct result of this success is a growing dependence on DXA measurements for assessment of normal growth and health, evaluation of disease states (e.g., osteoporosis, osteomalacia), and evaluation of therapeutic efficacy (1,2,5). This has taken place concurrently with enhancements in DXA technology that include modifications to hardware, scanning methodology, and analysis algorithms.
Received 4/12/05; Revised 5/19/05; Accepted 5/19/05. *Address correspondence to: Roman J. Shypailo, Children’s Nutrition Research Center, Baylor College of Medicine, 1100 Bates St., Houston, TX, 77030. E-mail: [email protected]
445
446 results. Bone mineral measurements, for example, are used in the assessment of osteopenia and osteoporosis via comparison with a normal reference range of values. For adults, the World Health Organization (WHO) specifies the presence of osteoporosis in cases where bone mineral density (BMD) or bone mineral content (BMC) falls 2.5 standard deviations or more below the mean of a reference young adult population (T-score –2.5) (14,15). The International Commit-tee for Standards in Bone Measurement (ICSBM) has recommended the use of femur measurement data from the latest National Health and Nutrition Examination Survey (NHANES III) as a reference database. These data were exclusively acquired on fan-beam DXA instruments manufactured by Hologic (16,17). Although standardized conversion equations allowed data acquired by other manufacturers’ DXA machines to use the NHANES database as a reference, inconsistencies in standard deviation values caused unreasonable changes in T-scores, so further corrections had to be applied (18). The utility of a consistent reference population database applies equally to studies of childhood growth. Investigators and clinicians have made use of DXA measurements for assessment of various bone diseases (e.g., osteogenesis imperfecta), as well as in the study of soft-tissue composition (e.g., obesity protocols) (19–21). The most recent DXA software modification, targeted at the pediatric population, comes from Hologic. The parameters used to classify pixels in a whole body scan have been adjusted to increase the number of bonecontaining pixels, which affects the resulting estimates of bone area (BA), BMC, and BMD. The dramatic shift in BMD values when changing scan modes from Infant to Adult on a Hologic instrument may have provided the motivation for these changes (22,23). Earlier efforts to solve this problem led to the development of a research analysis software version, recommended for use in children, which increased the number of bone-containing pixels in a whole body scan. The correction tended to produce a systematic shift in BMD from the standard adult whole body analysis. There was a systematic discrepancy when changing from one analysis mode to another (10,24,25). The latest version of Hologic’s analysis software (version 12.1), presumably incorporating enhancements in bone pixel classification, is provided on its newest generation of DXA instruments—the Discovery model, released in 2004. As we have noted in the past, changes in software versions produced noticeable changes in bone analysis results (11–13). We wanted to evaluate the effect of this newest software upgrade on our database of normal children and to explore the possible ramifications of any significant changes that the new analyses may produce.
Shypailo and Ellis The data set consisted of 502 boys and 625 girls. Scale weight ranged from 10.5 to 44.9 kg, with calculated DXA weight ranging from 10.1 to 46.8 kg. The age range was from 1.8 to 18.4 yr. The ethnic distribution was 65% Hispanic, 23% Caucasian, and 10% African American, with the remaining 2% representing other ethnic groups. Correspondence with Hologic (T. Kelly, personal communication) had indicated that subjects weighing more than 40 kg were unaffected by the new analysis, thus only subjects having a body weight of less than 45 kg were included in this analysis. Scale weight was used to exclude the heavier subjects in our database from this study. The maximum DXA weight (the calculated sum of the measured total body BMC, fat, and lean compartments) in our sample was 46.8 kg. The auto-analyze feature was used, allowing us to automatically analyze the scans in batch mode. The original analysis results were first stripped from the scans, although the region of interest placement information was maintained. This ensured that the body image was segmented in exactly the same way for each version of the analysis. The original whole body scan results came from analyses done on the respective DXA machine that had acquired the scans. About 20% of the scans were acquired on the QDR4500A, with the remainder coming from the Delphi-A. Because the QDR-4500A was running software version 8.26, an MS-DOS-based program, all scans coming from that machine were first re-analyzed by software version 11.1; the version running on the Delphi instrument. This produced a baseline dataset having a common analysis software version. Regression analysis was used to compare original and reanalyzed results for BA, BMC, BMD, and DXA weight. Both original and re-analyzed values were compared with body weight to confirm the 40 kg upper threshold. Analysis results for boys and girls were evaluated separately in order to determine any potential gender effects. The effect of the new analysis on regional DXA results was also evaluated. Standard whole body DXA results are comprised of 10 discreet regions that combine to produce the total body values: left arm, right arm, left ribs, right ribs, thoracic spine, lumbar spine, pelvis, left leg, right leg, and head. The extremities and torso regions were summed to produce four regions representing the whole body: arms, legs, trunk, and head. The relative change in each of these regions with the new analysis was obtained. Regression analysis between totalbody BA, BMC, and BMD with body weight and height were also calculated. For all statistical comparisons, a p-value of less than 0.05 was considered significant.
Results Materials and Methods Whole body DXA scans previously acquired on two different DXA fan-beam instruments, a Hologic QDR-4500A and a Hologic Delphi-A, were re-analyzed using the new whole body analysis software from Hologic (version 12.1.3). The analyses were performed on a Discovery QDR Workstation running Hologic QDR Software for Windows® XP. Journal of Clinical Densitometry
The new whole body analysis software did not significantly alter the DXA estimates for body weight. Original and new DXA weight values were highly correlated (R2 = 0.99), and the mean difference (0.937 ± 28.8 g) was not significant (p = 0.277). The results for the individual bone parameters, however, did show significant changes in subjects weighing less than 40 kg. BA, BMC, and BMD values for version 12.1 were highly correlated Volume 8, 2005
DXA Bone Measurements in Children
447
Fig. 1. Effect of Version 12.1 software on estimates of whole-body bone area.
Fig. 2. Effect of version 12.1 software on estimates of bone mineral content. with the corresponding original values from version 11.2. R2 values were 0.92, 0.99, and 0.89, respectively. New BA values were increased compared with the original values in the smaller subjects, and the differences diminished with increasing BA (Fig. 1). BA results for subjects greater than 40 kg, which did not change, Journal of Clinical Densitometry
are also plotted in Fig. 1 and are on the 1:1 line. The range of BA values for these heavier subjects overlapped the range of values obtained for the smaller subjects. Changes in BMC showed a similar pattern to BA, with larger differences at smaller BMC values (Fig. 2). Because the relative BMC increase was smaller Volume 8, 2005
448
Shypailo and Ellis
Fig. 3. Effect of Version 12.1 software on estimates of whole-body BMD.
Fig. 4. Relative change in whole body bone area for version 12.1 analysis compared to subject body weight. Relative change = (new value – original value)/original value × 100. than that for BA, the resultant BMD values were lower than the original values (Fig. 3). These differences diminished with increasing BMD. There was an overlap in BMD values, as observed for BA and BMC, between the two weight groups. Journal of Clinical Densitometry
The relative changes in BA, BMC, and BMD with body weight are presented in Figs. 4–6. The magnitude of the changes in BA and BMC decreased with increasing body weight, approaching no difference (0%) at approx 40 kg. BA Volume 8, 2005
DXA Bone Measurements in Children
449
Fig. 5. Relative change in whole-body bone mineral content resulting from version 12.1 analysis compared to subject body weight.
Fig. 6. Relative change in whole body bone mineral density resulting from version 12.1 analysis compared to subject body weight. was more affected than BMC, having an increase of almost 100% for subjects weighing about 10 kg, vis-à-vis only about 45% for BMC. This resulted in BMD decreases of 25% or more for the smallest subjects. No difference in BMD was in evidence when a subject’s body weight was 40 kg or higher. Journal of Clinical Densitometry
Analysis of variance (ANOVA) indicated no gender difference in the relative changes for BA, BMC, and BMD. The relative changes in BA for the regional DXA results are presented in Table 1. The change is expressed as a percent of the total change. The regional changes were not substantially Volume 8, 2005
450
Shypailo and Ellis Table 1 Relative Change in Regional Bone Area Boys
Fractional changea SD Min Max aFraction
Girls
Arm
Trunk
Leg
Head
Arm
Trunk
Leg
Head
27.6% 7.2% 0% 60%
24.0% 6.0% 0% 48%
47.3% 8.3% 20% 81%
1.1% 1.1% –1% 15%
30.3% 8.1% 0% 62%
21.3% 6.3% –26% 48%
47.6% 8.9% 26% 100%
1.0% 1.0% –1% 12%
of the total change in bone area. Sum of regional changes = 100%.
Table 2 Relative Change in Regional Bone Mineral Content Boys
Fractional changea SD Min Max aFraction
Girls
Arm
Trunk
Leg
Head
Arm
Trunk
Leg
Head
17.5% 5.0% 0% 42%
29.2% 6.8% –11% 55%
52.4% 8.3% 13% 77%
1.0% 2.9% 0% 57%
19.1% 5.7% –2% 49%
27.6% 7.1% –19% 56%
52.6% 8.6% 20% 100%
1.0% 1.0% –1% 9%
of the total change in BMC. Sum of regional changes = 100%.
different between boys and girls. Most of the increase in BA occurred in the legs (>47%), while the arms showed less change (28–30%). Smaller changes were seen in the trunk (21–24%). The results for the head region were virtually unaffected by the new analysis, with an average of only about 1% increase in BA with the new software, although there were individual cases where the change was greater than 10%. The patterns of regional relative change for BMC are given in Table 2. The greatest fractional changes (>52%) were present in the legs. The trunk region had higher relative changes in BMC than it had in BA, with increases of 28–29%. The arms had lower fractional changes, about 18–19%. As observed for BA, the head region was relatively unaffected by the new analysis for BMC. None of the changes in regional DXA results were significantly correlated with body weight (R2 < 0.03). A comparison of the results for new and original whole body BMD with age is displayed in Fig. 7. Original BMD values are greater than re-analyzed BMD values for both genders at a young age. This separation decreased with increasing age. On a population basis, BMD values for boys and girls returned to their original values at approximately age 14 yr; however, the ages of subjects affected by the new whole body analysis—that is, subjects weighing less than 40 kg—ranged up to 18.4 yr. The correlations between body weight, height, and the three bone parameters are presented in Table 3. Correlations for BA and BMC with body weight declined with the new software, whereas those for BMD increased from 0.41 to 0.74. The R2 values for height showed a similar increase (0.57 vs 0.78) with BMD from original to new software, whereas correlations with BA and BMC remained relatively constant. In all cases, Journal of Clinical Densitometry
height had higher correlation coefficients than body weight with BA, BMC, and BMD.
Discussion Changes in DXA technology are nothing new. The instruments have evolved since their inception, and the process can be expected to continue. This latest change in software will have a direct affect on a subset of previously acquired scans for smaller children, which may impact the interpretation and use of the pediatric DXA dataset previously accumulated. Because this is a software modification, the data acquisition methodologies, and therefore raw data, are unaltered. This will allow investigators to re-analyze previously acquired DXA scans using the new software. However, the software environment incorporating this algorithm resides on only the newer Microsoft Windows®-based systems. Scans previously acquired on a Hologic Delphi scanner were easily re-analyzed using the new program. The processing of the scans acquired by a Hologic 4500A, using an MS-DOS-based QDR system, was not as straightforward, and required the resetting of several key parameters not normally accessible to the average user. Older scans acquired by a pencil-beam QDR-2000 instrument could not be reanalyzed. Therefore, the newest modifications will probably be directly utilized only by those investigators with the newest Hologic systems. Any new analysis results will subsequently have a bearing on previously acquired data sets for subjects weighing less than 40 kg. For example, if a reference database is established using software version 12.1, this will compel users of older Hologic Volume 8, 2005
DXA Bone Measurements in Children
451
Fig. 7. Whole body bone mineral density change vs age. Comparison of original and version 12.1 results. Table 3 Correlations Between Bone Parameters and Weight and Heighta Original analysis
Wt (kg) Ht (cm)
New analysis
BA
BMC
BMD
BA
BMC
BMD
0.811 0.918
0.810 0.900
0.412 0.572
0.643 0.904
0.734 0.901
0.738 0.782
aR2
values for power function models. All correlations are significant at the 0.01 level.
DXA machines to convert their results to the new standard if their archived scans cannot be directly re-analyzed by the new software. This would be analogous to the use of equations promulgated for converting manufacturer specific adult femur measurements to standardized units based on the NHANES III reference database. Furthermore, because version 12.1 works only with scans acquired utilizing DXA fan-beam technology, it may be more difficult to derive a direct conversion between the newer scanners and older pencil-beam instruments. Clinically, the measurement result most often used for bone is BMD. As we have demonstrated, Hologic Software Version 12.1 will produce lower BMD values than previous scanners for children weighing less than 40 kg. The decrease in BMD, however, is not a systematic shift in the data. Our results show an altered, nonlinear slope for BMD, whether comparing against age or weight. Thus, future longitudinal studies of young children will show a different rate of change in BMD than previously reported, with a child’s BMD increasing more Journal of Clinical Densitometry
rapidly than before. Conversely, BMC and BA have opposite trends, suggesting smaller relative gains over time. Investigators examining regional changes using whole body DXA scans should note that the regional results were not altered uniformly by the new software version. Almost 50% of the changes occurred in the leg region, which may serve to dampen the effects of changes in the remaining regions. It is interesting to note that the results for the head region were not significantly different, most likely indicating that this region was excluded in the new analysis routine. This may give more credence to the approach of using DXA results sans head, as proposed by others (26). Because of the changes resulting from using version 12.1, it may be prudent to evaluate DXA results via comparisons with a reference population. The difficulty here lies with the fact that most normative samples have been based on age, whereas this new correction is based on body weight. Although our re-analyzed results show similar trends vs age and weight, the actual values for BA, BMC, and BMD are different. Children at risk for various health problems, or affected by certain ailments, may fall well below the physical norms for age when compared with a healthy age-matched population. Hence, an underweight child’s DXA results would be altered by version 12.1 to a greater degree in comparison with an age-matched healthy child. The new software algorithm implies that this comparison will be more accurate, but this needs to be verified. The version 12.1 software automatically applies this weight-based adjustment without input from the user. It is based on the DXA estimate of body weight. This makes the adjustment process transparent to the user. However, other Volume 8, 2005
452 anthropometric parameters, such as height, are strongly correlated with whole body bone values, and may deserve consideration as criteria indices for application of adjustment factors. We have shown that height is more strongly correlated with the three bone parameters than body weight (Table 3). The new software version actually reduces the correlation between weight, BA and BMC. Height, however, maintains correlation coefficients of greater than 0.90 for both analysis versions. It is unclear, but it appears that this latest software-driven adjustment appears focused on BMD, possibly at the expense of the accuracy of its components (i.e., BMC and BA). The newest whole body DXA analysis version from Hologic (version 12.1) produces significant changes in BA, BMC, and BMD in subjects weighing less than 40 kg when compared with the previous version. Children in the 10–15 kg weight range, for example, would have decreased total body BMD values of 25% or more. These types of changes will have a significant impact on longitudinal studies of children, as the yearly rate of change for the various bone parameters will be altered. In addition, application of the new software will alter bone mineral results when compared with existing normal reference databases based on earlier software versions or DXA instruments, even from the same manufacturer. Investigators must be aware of what software version was used by their DXA instrument as well as what software version and scanner model was used to derive any comparison normative data set. It is our recommendation that, whenever possible, investigators only compare scans that have a common analysis software version, either through scan reanalysis or adjustment of previous results. Conversion options between instruments and software versions will be made available on the CNRC web site (www.bcm.edu/bodycomplab) for bone assessment in children. The development of a revised Z-score calculator for BMC and BMD, adjusted for this new software version, will also be added to the web site.
Acknowledgments We would like to thank Tom Kelly of Hologic for his technical assistance. This work is a publication of the USDA/ARS Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, and Texas Children’s Hospital, Houston, TX. Funding has been provided from the USDA/ ARS under Cooperative Agreement No. 58-6250-6-001. The contents of this publication do not necessarily reflect the views or policies of the USDA, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government.
References 1. Albanese CV, Diessel E, Genent HK. 2003 Clinical applications of body composition measurements using DXA. J Clin Densitom 6:75–85. 2. Kolta S, Ravaud P, Fechtenbaum J, Dougados M, Roux C. 2000 Follow-up of individual patients on two DXA scanners of the same manufacturer. Osteoporosis Int 11:709–713. 3. Blake GM, Harrison EJ, Adams JE. 2004 Dual X-ray absorptiometry: cross-calibration of a new fan-beam system. Calcif Tissue Int 75:7–14.
Journal of Clinical Densitometry
Shypailo and Ellis 4. Litaker MS, Barbeau P, Humphries MC, Gutin B. 2003 Comparison of Hologic QDR-1000/W and 4500W DXA scanners in 13- to 18-year olds. Obes Res 11:1545–1552. 5. Tylavsky A, Lohman TG, Dockrell M, et al. 2003 Comparison of the effectiveness of 2 dual-energy X-ray absorptiometers with that of total body water and computed tomography in assessing changes in body composition during weight change. Am J Clin Nutr 77:356–363. 6. Ellis KJ, Shypailo RJ. 1998 Bone mineral and body composition measurements: cross-calibration of pencil-beam and fan-beam dual-energy X-ray absorptiometers. J Bone Miner Res 13: 1613–1618. 7. Tothill P, Hannan WJ, Wilkinson S. 2001 Comparisons between a pencil beam and two fan beam dual energy X-ray absorptiometers used for measuring total body bone and soft tissue. Br J Radiol 74:166–176. 8. Oldroyd B, Smith AH, Truscott JG. 2003 Cross-calibration of GE/Lunar pencil and fan-beam dual energy densitometers— bone mineral density and body composition studies. Eur J Clin Nutr 57:977–987. 9. Soriano JM, Ioannidou E, Wang J, et al. 2004 Pencil-beam vs fan-beam dual-energy X-ray absorptiometry comparison across four systems: body composition and bone mineral. J Clin Densitom 7:281–289. 10. Laskey MA, Prentice A. 1999 Comparison of adult and paediatric spine and whole body software for the Lunar dual energy X-ray absorptiometer. Br J Radiol 72:967–976. 11. Vozarova B, Wang J, Weyer C, Tataranni PA. 2001 Comparison of two software versions for assessment of body-composition analysis by DXA. Obes Res 9:229–232. 12. Cordero-MacIntyre ZR, Peters W, Libanati CR, et al. 2002 Reproducibility of DXA in obese women. J Clin Densitom 5:35–44. 13. 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 4:587–591. 14. 1994 Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. World Health Organ Tech Rep Ser 843:1–129. 15. Kanis JA, Melton LJ, Christiansen C, Johnston CC, Khaltaev N. 1994 The diagnosis of osteoporosis. J Bone Miner Res 11: 1199–1201. 16. Hanson J. 1997 Standardization of proximal femur BMD measurements. International Committee for Standards in Bone Measurement. Osteoporos Int 7:500–501. 17. Looker AC, Wahner HW, Dunn WL, et al. 1998 Updated data on proximal femur bone mineral levels in US adults. Osteoporosis Int 8:468–489. 18. Binkley N, Kiebzak GM, Lewiecki EM, et al. 2005 Recalculation of the NHANES database SD improves T-score agreement and reduces osteoporosis prevalence. J Bone Miner Res 20:195–201. 19. Crabtree NJ, Kibirige MS, Fordham JN, et al. 2004 The relationship between lean body mass and bone mineral content in paediatric health and disease. Bone 35:965–972. 20. Specker B, Johannsen N, Binkley T, Finn K. 2001 Total body bone mineral content and tibial cortical bone measures in preschool children. J Bone Miner Res 16:2298–2305. 21. Himes JH, Obarzanek E, Baranowski T, Wilson DM, Rochon J, McClanahan BS. 2004 Early sexual maturation, body composition, and obesity in African-American girls. Obes Res 12: 64S–72S. 22. Shypailo RJ, Ellis KJ. 2000 Accuracy of dual-energy X-ray absorptiometry in infant scans. Ann NY Acad Sci 904:98–100.
Volume 8, 2005
DXA Bone Measurements in Children 23. Picaud JC, Duboeuf F, Vey-Marty V, et al. 2003 First all-solid pediatric phantom for dual X-ray absorptiometry measurements in infants. J Clin Densitom 6:17–23. 24. Wang J, Thornton JC, Horlick M, Formica C, Wang W, Pierson RN. 1999 Dual X-ray absorptiometry in pediatric studies: changing scan modes alters bone and body composition measurements. J Clin Densitom 2:135–141.
Journal of Clinical Densitometry
453 25. Hammami M, Koo WW, Hockman EM. 2004 Technical considerations for fan-beam dual-energy x-ray absorptiometry body composition measurements in pediatric studies. J Parenter Enteral Nutr 28:328–333. 26. Roubenoff R, Kehayias JJ, Dawson-Hughes B, Heymsfield SB. 1993 Use of dual-energy X-ray absorptiometry in body-composition studies: not yet a “gold standard.” Am J Clin Nutr 58:589–591.
Volume 8, 2005
Original Article
Bone Assessment in Children Comparison of Fan-Beam DXA Analysis R. J. Shypailo and K. J. Ellis USDA/ARS Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, TX
Abstract The newest version of whole body dual-energy X-ray absorptiometry (DXA) analysis software from Hologic (Discovery 12.1) is designed to enhance bone detection in smaller subjects. We re-analyzed 1127 pediatric scans (ages 1.8–18.5 yr) previously analyzed using software version 11.2. Regression analysis compared new and original results for bone area (BA), bone mineral content (BMC), bone mineral density (BMD), and DXA-derived body weight. Changes in total and regional bone results were compared with age, weight, and height. New results were highly correlated with original analyses (R2 0.9), but there were large differences at the individual subject level. The BA and BMC values increased in subjects less than 40 kg weight, resulting in a lower BMD. Original and new results were equivalent by about age 14 yr in both genders. Regional bone data showed the greatest changes in the legs. The newest software produces significant changes in bone values in subjects weighing less than 40 kg, compared with earlier versions. This effect increases with decreasing body weight. This will impact interpretation of longitudinal pediatric DXA studies, as well as existing pediatric whole body bone reference databases. Investigators must recognize which DXA software version they are using, and which version produced any reference database they may use for comparison. Key Words: DXA; body composition; bone densitometry; clinical/pediatrics; growth and development.
Every stage of DXA technological development, whether in instrumentation or analysis, may produce changes in bone mineral results. The early pencil-beam instruments (Hologic® QDR1000 and QDR-2000; Lunar DPX and DPX-L) have given way to newer fan-beam devices (Hologic QDR4500; Lunar Prodigy) that require more assumptions in the reconstruction of the images and bone data. Investigators comparing the performance of these technologies almost invariably found significant differences in the estimates for bone. Results were usually highly correlated, hence conversion equations were suggested and promulgated (6–9). In conjunction with advancements in scanning techniques came changes in analysis algorithms. Again, investigators comparing analysis versions usually found highly correlated but statistically significant differences, although they were often of smaller magnitude with less clinical importance. Even so, conversion equations were still usually called for (10–13). Maintaining consistency of bone and body composition results is critical because of the importance of establishing reliable normative databases for the interpretation of DXA
Introduction Dual-energy X-ray absorptiometry (DXA) has become a reliable tool for the measurement of bone mineral in vivo (1–4). It has good precision, and operates at a low radiation dose, with ever decreasing scan times resulting from advancements in technology (1,2). A direct result of this success is a growing dependence on DXA measurements for assessment of normal growth and health, evaluation of disease states (e.g., osteoporosis, osteomalacia), and evaluation of therapeutic efficacy (1,2,5). This has taken place concurrently with enhancements in DXA technology that include modifications to hardware, scanning methodology, and analysis algorithms.
Received 4/12/05; Revised 5/19/05; Accepted 5/19/05. *Address correspondence to: Roman J. Shypailo, Children’s Nutrition Research Center, Baylor College of Medicine, 1100 Bates St., Houston, TX, 77030. E-mail: [email protected]
445
446 results. Bone mineral measurements, for example, are used in the assessment of osteopenia and osteoporosis via comparison with a normal reference range of values. For adults, the World Health Organization (WHO) specifies the presence of osteoporosis in cases where bone mineral density (BMD) or bone mineral content (BMC) falls 2.5 standard deviations or more below the mean of a reference young adult population (T-score –2.5) (14,15). The International Commit-tee for Standards in Bone Measurement (ICSBM) has recommended the use of femur measurement data from the latest National Health and Nutrition Examination Survey (NHANES III) as a reference database. These data were exclusively acquired on fan-beam DXA instruments manufactured by Hologic (16,17). Although standardized conversion equations allowed data acquired by other manufacturers’ DXA machines to use the NHANES database as a reference, inconsistencies in standard deviation values caused unreasonable changes in T-scores, so further corrections had to be applied (18). The utility of a consistent reference population database applies equally to studies of childhood growth. Investigators and clinicians have made use of DXA measurements for assessment of various bone diseases (e.g., osteogenesis imperfecta), as well as in the study of soft-tissue composition (e.g., obesity protocols) (19–21). The most recent DXA software modification, targeted at the pediatric population, comes from Hologic. The parameters used to classify pixels in a whole body scan have been adjusted to increase the number of bonecontaining pixels, which affects the resulting estimates of bone area (BA), BMC, and BMD. The dramatic shift in BMD values when changing scan modes from Infant to Adult on a Hologic instrument may have provided the motivation for these changes (22,23). Earlier efforts to solve this problem led to the development of a research analysis software version, recommended for use in children, which increased the number of bone-containing pixels in a whole body scan. The correction tended to produce a systematic shift in BMD from the standard adult whole body analysis. There was a systematic discrepancy when changing from one analysis mode to another (10,24,25). The latest version of Hologic’s analysis software (version 12.1), presumably incorporating enhancements in bone pixel classification, is provided on its newest generation of DXA instruments—the Discovery model, released in 2004. As we have noted in the past, changes in software versions produced noticeable changes in bone analysis results (11–13). We wanted to evaluate the effect of this newest software upgrade on our database of normal children and to explore the possible ramifications of any significant changes that the new analyses may produce.
Shypailo and Ellis The data set consisted of 502 boys and 625 girls. Scale weight ranged from 10.5 to 44.9 kg, with calculated DXA weight ranging from 10.1 to 46.8 kg. The age range was from 1.8 to 18.4 yr. The ethnic distribution was 65% Hispanic, 23% Caucasian, and 10% African American, with the remaining 2% representing other ethnic groups. Correspondence with Hologic (T. Kelly, personal communication) had indicated that subjects weighing more than 40 kg were unaffected by the new analysis, thus only subjects having a body weight of less than 45 kg were included in this analysis. Scale weight was used to exclude the heavier subjects in our database from this study. The maximum DXA weight (the calculated sum of the measured total body BMC, fat, and lean compartments) in our sample was 46.8 kg. The auto-analyze feature was used, allowing us to automatically analyze the scans in batch mode. The original analysis results were first stripped from the scans, although the region of interest placement information was maintained. This ensured that the body image was segmented in exactly the same way for each version of the analysis. The original whole body scan results came from analyses done on the respective DXA machine that had acquired the scans. About 20% of the scans were acquired on the QDR4500A, with the remainder coming from the Delphi-A. Because the QDR-4500A was running software version 8.26, an MS-DOS-based program, all scans coming from that machine were first re-analyzed by software version 11.1; the version running on the Delphi instrument. This produced a baseline dataset having a common analysis software version. Regression analysis was used to compare original and reanalyzed results for BA, BMC, BMD, and DXA weight. Both original and re-analyzed values were compared with body weight to confirm the 40 kg upper threshold. Analysis results for boys and girls were evaluated separately in order to determine any potential gender effects. The effect of the new analysis on regional DXA results was also evaluated. Standard whole body DXA results are comprised of 10 discreet regions that combine to produce the total body values: left arm, right arm, left ribs, right ribs, thoracic spine, lumbar spine, pelvis, left leg, right leg, and head. The extremities and torso regions were summed to produce four regions representing the whole body: arms, legs, trunk, and head. The relative change in each of these regions with the new analysis was obtained. Regression analysis between totalbody BA, BMC, and BMD with body weight and height were also calculated. For all statistical comparisons, a p-value of less than 0.05 was considered significant.
Results Materials and Methods Whole body DXA scans previously acquired on two different DXA fan-beam instruments, a Hologic QDR-4500A and a Hologic Delphi-A, were re-analyzed using the new whole body analysis software from Hologic (version 12.1.3). The analyses were performed on a Discovery QDR Workstation running Hologic QDR Software for Windows® XP. Journal of Clinical Densitometry
The new whole body analysis software did not significantly alter the DXA estimates for body weight. Original and new DXA weight values were highly correlated (R2 = 0.99), and the mean difference (0.937 ± 28.8 g) was not significant (p = 0.277). The results for the individual bone parameters, however, did show significant changes in subjects weighing less than 40 kg. BA, BMC, and BMD values for version 12.1 were highly correlated Volume 8, 2005
DXA Bone Measurements in Children
447
Fig. 1. Effect of Version 12.1 software on estimates of whole-body bone area.
Fig. 2. Effect of version 12.1 software on estimates of bone mineral content. with the corresponding original values from version 11.2. R2 values were 0.92, 0.99, and 0.89, respectively. New BA values were increased compared with the original values in the smaller subjects, and the differences diminished with increasing BA (Fig. 1). BA results for subjects greater than 40 kg, which did not change, Journal of Clinical Densitometry
are also plotted in Fig. 1 and are on the 1:1 line. The range of BA values for these heavier subjects overlapped the range of values obtained for the smaller subjects. Changes in BMC showed a similar pattern to BA, with larger differences at smaller BMC values (Fig. 2). Because the relative BMC increase was smaller Volume 8, 2005
448
Shypailo and Ellis
Fig. 3. Effect of Version 12.1 software on estimates of whole-body BMD.
Fig. 4. Relative change in whole body bone area for version 12.1 analysis compared to subject body weight. Relative change = (new value – original value)/original value × 100. than that for BA, the resultant BMD values were lower than the original values (Fig. 3). These differences diminished with increasing BMD. There was an overlap in BMD values, as observed for BA and BMC, between the two weight groups. Journal of Clinical Densitometry
The relative changes in BA, BMC, and BMD with body weight are presented in Figs. 4–6. The magnitude of the changes in BA and BMC decreased with increasing body weight, approaching no difference (0%) at approx 40 kg. BA Volume 8, 2005
DXA Bone Measurements in Children
449
Fig. 5. Relative change in whole-body bone mineral content resulting from version 12.1 analysis compared to subject body weight.
Fig. 6. Relative change in whole body bone mineral density resulting from version 12.1 analysis compared to subject body weight. was more affected than BMC, having an increase of almost 100% for subjects weighing about 10 kg, vis-à-vis only about 45% for BMC. This resulted in BMD decreases of 25% or more for the smallest subjects. No difference in BMD was in evidence when a subject’s body weight was 40 kg or higher. Journal of Clinical Densitometry
Analysis of variance (ANOVA) indicated no gender difference in the relative changes for BA, BMC, and BMD. The relative changes in BA for the regional DXA results are presented in Table 1. The change is expressed as a percent of the total change. The regional changes were not substantially Volume 8, 2005
450
Shypailo and Ellis Table 1 Relative Change in Regional Bone Area Boys
Fractional changea SD Min Max aFraction
Girls
Arm
Trunk
Leg
Head
Arm
Trunk
Leg
Head
27.6% 7.2% 0% 60%
24.0% 6.0% 0% 48%
47.3% 8.3% 20% 81%
1.1% 1.1% –1% 15%
30.3% 8.1% 0% 62%
21.3% 6.3% –26% 48%
47.6% 8.9% 26% 100%
1.0% 1.0% –1% 12%
of the total change in bone area. Sum of regional changes = 100%.
Table 2 Relative Change in Regional Bone Mineral Content Boys
Fractional changea SD Min Max aFraction
Girls
Arm
Trunk
Leg
Head
Arm
Trunk
Leg
Head
17.5% 5.0% 0% 42%
29.2% 6.8% –11% 55%
52.4% 8.3% 13% 77%
1.0% 2.9% 0% 57%
19.1% 5.7% –2% 49%
27.6% 7.1% –19% 56%
52.6% 8.6% 20% 100%
1.0% 1.0% –1% 9%
of the total change in BMC. Sum of regional changes = 100%.
different between boys and girls. Most of the increase in BA occurred in the legs (>47%), while the arms showed less change (28–30%). Smaller changes were seen in the trunk (21–24%). The results for the head region were virtually unaffected by the new analysis, with an average of only about 1% increase in BA with the new software, although there were individual cases where the change was greater than 10%. The patterns of regional relative change for BMC are given in Table 2. The greatest fractional changes (>52%) were present in the legs. The trunk region had higher relative changes in BMC than it had in BA, with increases of 28–29%. The arms had lower fractional changes, about 18–19%. As observed for BA, the head region was relatively unaffected by the new analysis for BMC. None of the changes in regional DXA results were significantly correlated with body weight (R2 < 0.03). A comparison of the results for new and original whole body BMD with age is displayed in Fig. 7. Original BMD values are greater than re-analyzed BMD values for both genders at a young age. This separation decreased with increasing age. On a population basis, BMD values for boys and girls returned to their original values at approximately age 14 yr; however, the ages of subjects affected by the new whole body analysis—that is, subjects weighing less than 40 kg—ranged up to 18.4 yr. The correlations between body weight, height, and the three bone parameters are presented in Table 3. Correlations for BA and BMC with body weight declined with the new software, whereas those for BMD increased from 0.41 to 0.74. The R2 values for height showed a similar increase (0.57 vs 0.78) with BMD from original to new software, whereas correlations with BA and BMC remained relatively constant. In all cases, Journal of Clinical Densitometry
height had higher correlation coefficients than body weight with BA, BMC, and BMD.
Discussion Changes in DXA technology are nothing new. The instruments have evolved since their inception, and the process can be expected to continue. This latest change in software will have a direct affect on a subset of previously acquired scans for smaller children, which may impact the interpretation and use of the pediatric DXA dataset previously accumulated. Because this is a software modification, the data acquisition methodologies, and therefore raw data, are unaltered. This will allow investigators to re-analyze previously acquired DXA scans using the new software. However, the software environment incorporating this algorithm resides on only the newer Microsoft Windows®-based systems. Scans previously acquired on a Hologic Delphi scanner were easily re-analyzed using the new program. The processing of the scans acquired by a Hologic 4500A, using an MS-DOS-based QDR system, was not as straightforward, and required the resetting of several key parameters not normally accessible to the average user. Older scans acquired by a pencil-beam QDR-2000 instrument could not be reanalyzed. Therefore, the newest modifications will probably be directly utilized only by those investigators with the newest Hologic systems. Any new analysis results will subsequently have a bearing on previously acquired data sets for subjects weighing less than 40 kg. For example, if a reference database is established using software version 12.1, this will compel users of older Hologic Volume 8, 2005
DXA Bone Measurements in Children
451
Fig. 7. Whole body bone mineral density change vs age. Comparison of original and version 12.1 results. Table 3 Correlations Between Bone Parameters and Weight and Heighta Original analysis
Wt (kg) Ht (cm)
New analysis
BA
BMC
BMD
BA
BMC
BMD
0.811 0.918
0.810 0.900
0.412 0.572
0.643 0.904
0.734 0.901
0.738 0.782
aR2
values for power function models. All correlations are significant at the 0.01 level.
DXA machines to convert their results to the new standard if their archived scans cannot be directly re-analyzed by the new software. This would be analogous to the use of equations promulgated for converting manufacturer specific adult femur measurements to standardized units based on the NHANES III reference database. Furthermore, because version 12.1 works only with scans acquired utilizing DXA fan-beam technology, it may be more difficult to derive a direct conversion between the newer scanners and older pencil-beam instruments. Clinically, the measurement result most often used for bone is BMD. As we have demonstrated, Hologic Software Version 12.1 will produce lower BMD values than previous scanners for children weighing less than 40 kg. The decrease in BMD, however, is not a systematic shift in the data. Our results show an altered, nonlinear slope for BMD, whether comparing against age or weight. Thus, future longitudinal studies of young children will show a different rate of change in BMD than previously reported, with a child’s BMD increasing more Journal of Clinical Densitometry
rapidly than before. Conversely, BMC and BA have opposite trends, suggesting smaller relative gains over time. Investigators examining regional changes using whole body DXA scans should note that the regional results were not altered uniformly by the new software version. Almost 50% of the changes occurred in the leg region, which may serve to dampen the effects of changes in the remaining regions. It is interesting to note that the results for the head region were not significantly different, most likely indicating that this region was excluded in the new analysis routine. This may give more credence to the approach of using DXA results sans head, as proposed by others (26). Because of the changes resulting from using version 12.1, it may be prudent to evaluate DXA results via comparisons with a reference population. The difficulty here lies with the fact that most normative samples have been based on age, whereas this new correction is based on body weight. Although our re-analyzed results show similar trends vs age and weight, the actual values for BA, BMC, and BMD are different. Children at risk for various health problems, or affected by certain ailments, may fall well below the physical norms for age when compared with a healthy age-matched population. Hence, an underweight child’s DXA results would be altered by version 12.1 to a greater degree in comparison with an age-matched healthy child. The new software algorithm implies that this comparison will be more accurate, but this needs to be verified. The version 12.1 software automatically applies this weight-based adjustment without input from the user. It is based on the DXA estimate of body weight. This makes the adjustment process transparent to the user. However, other Volume 8, 2005
452 anthropometric parameters, such as height, are strongly correlated with whole body bone values, and may deserve consideration as criteria indices for application of adjustment factors. We have shown that height is more strongly correlated with the three bone parameters than body weight (Table 3). The new software version actually reduces the correlation between weight, BA and BMC. Height, however, maintains correlation coefficients of greater than 0.90 for both analysis versions. It is unclear, but it appears that this latest software-driven adjustment appears focused on BMD, possibly at the expense of the accuracy of its components (i.e., BMC and BA). The newest whole body DXA analysis version from Hologic (version 12.1) produces significant changes in BA, BMC, and BMD in subjects weighing less than 40 kg when compared with the previous version. Children in the 10–15 kg weight range, for example, would have decreased total body BMD values of 25% or more. These types of changes will have a significant impact on longitudinal studies of children, as the yearly rate of change for the various bone parameters will be altered. In addition, application of the new software will alter bone mineral results when compared with existing normal reference databases based on earlier software versions or DXA instruments, even from the same manufacturer. Investigators must be aware of what software version was used by their DXA instrument as well as what software version and scanner model was used to derive any comparison normative data set. It is our recommendation that, whenever possible, investigators only compare scans that have a common analysis software version, either through scan reanalysis or adjustment of previous results. Conversion options between instruments and software versions will be made available on the CNRC web site (www.bcm.edu/bodycomplab) for bone assessment in children. The development of a revised Z-score calculator for BMC and BMD, adjusted for this new software version, will also be added to the web site.
Acknowledgments We would like to thank Tom Kelly of Hologic for his technical assistance. This work is a publication of the USDA/ARS Children’s Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, and Texas Children’s Hospital, Houston, TX. Funding has been provided from the USDA/ ARS under Cooperative Agreement No. 58-6250-6-001. The contents of this publication do not necessarily reflect the views or policies of the USDA, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government.
References 1. Albanese CV, Diessel E, Genent HK. 2003 Clinical applications of body composition measurements using DXA. J Clin Densitom 6:75–85. 2. Kolta S, Ravaud P, Fechtenbaum J, Dougados M, Roux C. 2000 Follow-up of individual patients on two DXA scanners of the same manufacturer. Osteoporosis Int 11:709–713. 3. Blake GM, Harrison EJ, Adams JE. 2004 Dual X-ray absorptiometry: cross-calibration of a new fan-beam system. Calcif Tissue Int 75:7–14.
Journal of Clinical Densitometry
Shypailo and Ellis 4. Litaker MS, Barbeau P, Humphries MC, Gutin B. 2003 Comparison of Hologic QDR-1000/W and 4500W DXA scanners in 13- to 18-year olds. Obes Res 11:1545–1552. 5. Tylavsky A, Lohman TG, Dockrell M, et al. 2003 Comparison of the effectiveness of 2 dual-energy X-ray absorptiometers with that of total body water and computed tomography in assessing changes in body composition during weight change. Am J Clin Nutr 77:356–363. 6. Ellis KJ, Shypailo RJ. 1998 Bone mineral and body composition measurements: cross-calibration of pencil-beam and fan-beam dual-energy X-ray absorptiometers. J Bone Miner Res 13: 1613–1618. 7. Tothill P, Hannan WJ, Wilkinson S. 2001 Comparisons between a pencil beam and two fan beam dual energy X-ray absorptiometers used for measuring total body bone and soft tissue. Br J Radiol 74:166–176. 8. Oldroyd B, Smith AH, Truscott JG. 2003 Cross-calibration of GE/Lunar pencil and fan-beam dual energy densitometers— bone mineral density and body composition studies. Eur J Clin Nutr 57:977–987. 9. Soriano JM, Ioannidou E, Wang J, et al. 2004 Pencil-beam vs fan-beam dual-energy X-ray absorptiometry comparison across four systems: body composition and bone mineral. J Clin Densitom 7:281–289. 10. Laskey MA, Prentice A. 1999 Comparison of adult and paediatric spine and whole body software for the Lunar dual energy X-ray absorptiometer. Br J Radiol 72:967–976. 11. Vozarova B, Wang J, Weyer C, Tataranni PA. 2001 Comparison of two software versions for assessment of body-composition analysis by DXA. Obes Res 9:229–232. 12. Cordero-MacIntyre ZR, Peters W, Libanati CR, et al. 2002 Reproducibility of DXA in obese women. J Clin Densitom 5:35–44. 13. 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 4:587–591. 14. 1994 Assessment of fracture risk and its application to screening for postmenopausal osteoporosis. World Health Organ Tech Rep Ser 843:1–129. 15. Kanis JA, Melton LJ, Christiansen C, Johnston CC, Khaltaev N. 1994 The diagnosis of osteoporosis. J Bone Miner Res 11: 1199–1201. 16. Hanson J. 1997 Standardization of proximal femur BMD measurements. International Committee for Standards in Bone Measurement. Osteoporos Int 7:500–501. 17. Looker AC, Wahner HW, Dunn WL, et al. 1998 Updated data on proximal femur bone mineral levels in US adults. Osteoporosis Int 8:468–489. 18. Binkley N, Kiebzak GM, Lewiecki EM, et al. 2005 Recalculation of the NHANES database SD improves T-score agreement and reduces osteoporosis prevalence. J Bone Miner Res 20:195–201. 19. Crabtree NJ, Kibirige MS, Fordham JN, et al. 2004 The relationship between lean body mass and bone mineral content in paediatric health and disease. Bone 35:965–972. 20. Specker B, Johannsen N, Binkley T, Finn K. 2001 Total body bone mineral content and tibial cortical bone measures in preschool children. J Bone Miner Res 16:2298–2305. 21. Himes JH, Obarzanek E, Baranowski T, Wilson DM, Rochon J, McClanahan BS. 2004 Early sexual maturation, body composition, and obesity in African-American girls. Obes Res 12: 64S–72S. 22. Shypailo RJ, Ellis KJ. 2000 Accuracy of dual-energy X-ray absorptiometry in infant scans. Ann NY Acad Sci 904:98–100.
Volume 8, 2005
DXA Bone Measurements in Children 23. Picaud JC, Duboeuf F, Vey-Marty V, et al. 2003 First all-solid pediatric phantom for dual X-ray absorptiometry measurements in infants. J Clin Densitom 6:17–23. 24. Wang J, Thornton JC, Horlick M, Formica C, Wang W, Pierson RN. 1999 Dual X-ray absorptiometry in pediatric studies: changing scan modes alters bone and body composition measurements. J Clin Densitom 2:135–141.
Journal of Clinical Densitometry
453 25. Hammami M, Koo WW, Hockman EM. 2004 Technical considerations for fan-beam dual-energy x-ray absorptiometry body composition measurements in pediatric studies. J Parenter Enteral Nutr 28:328–333. 26. Roubenoff R, Kehayias JJ, Dawson-Hughes B, Heymsfield SB. 1993 Use of dual-energy X-ray absorptiometry in body-composition studies: not yet a “gold standard.” Am J Clin Nutr 58:589–591.
Volume 8, 2005