Journal of Clinical Densitometry, vol. 3, no. 1, 9 14, Spring 2000 Copyright 2000 by Humana Press Inc. All rights of any nature whatsoever reserved. 0169-4194/00/3:9 14/$11.50 Original Article pqct Measurement of Bone Parameters in Young Children Validation of Technique Teresa L. Binkley, MS 1 and Bonny L. Specker, PHD 1,2 1 Ethel Austin Martin Program in Human Nutrition, South Dakota State University, Brookings, SD; and 2 Department of Pediatrics, University of South Dakota School of Medicine, Brookings, SD Abstract Dual-energy X-ray absorptiometry (DXA) measures areal bone mineral density (BMD) and is affected by bone size. Peripheral quantitative computed tomography (pqct) measures volumetric density and should not be affected by bone size. We hypothesized that pqct could be used to measure geometric parameters in the tibia and that bone size, not density, would correlate with weight and height in 3 and 4 yr olds. Phantom measurements indicate that accurate results for cortical volumetric bone mineral density can be obtained at cortical thickness > 2 mm with voxel/speed sizes of 0.40/20 mm/s. Correlations between pqct measured geometric bone parameters and phantom calculations were observed (all r > 0.96). Baseline data from an ongoing trial of 101 preschool children (53 male) were used to correlate bone parameters and anthropometrics. Total cross-sectional area, cortical area, and cortical thickness correlated with weight (r = 0.54, p < 0.001; r = 0.52, p < 0.001; r = 0.30, p = 0.002) and height (r = 0.41, p < 0.001; r = 0.55, p < 0.001; r = 0.41, p < 0.001). Because of the small cortical thickness at this age (mean = 1.2 mm), the cortical density was not analyzed. In a regression model including height and weight, weight was the only predictor of total cross-sectional area (p < 0.001); cortical thickness was predicted by height (p = 0.006). Both height (p = 0.005) and weight (p = 0.05) predicted the cortical area. In summary, pqct accurately measures the total cross-sectional area, cortical area, and cortical thickness in children age 3 4 yr. Key Words: Peripheral quantitative computed tomography (pqct); cortical density; total cross-sectional area; cortical area; cortical thickness. Introduction Optimal formation and mineralization of the skeletal system during childhood is thought to be important in preventing osteoporosis during adult life (1). Dual-energy X-ray absorptiometry (DXA) is the most widely used technique for measuring bone density in clinical trials; however, methods to Received 07/07/99; Revised 08/20/99; Accepted 09/25/99. Address correspondence to Teresa L. Binkley, Box 2204, South Dakota State University, Brookings, SD 57007. E-mail: Teresa_Binkley@sdstate.edu 9 calculate bone mineral density (BMD) using DXA are dependent on bone area (cm 2 ), not volume (cm 3 ). Peripheral quantitative computed tomography (pqct) measures BMD using a volume (cm 3 ) and has the ability to differentiate cortical bone from trabecular bone. Parameters such as cortical density, total cross-sectional bone area, cortical bone area, and cortical thickness can be assessed using pqct. Although these measurements are expected to be beneficial in the study of bone growth in young children, the question that arises is the extent to which this technique can measure smaller bones.
10 Binkley and Specker The purpose of this study was to use a phantom of known size and density to validate the performance of a pqct densitometer, as well as to test the precision of the instrument in children. Also, baseline data taken from a longitudinal study now underway at South Dakota State University (Brookings, SD) were used to correlate various pqct bone parameter measurements with anthropometric measurements in 3- and 4-yr-old children. Materials and Methods The pqct densitometer used in all measurements was an XCT 2000 (Norland Medical Systems, Inc., Fort Atkinson, WI). This unit has six detectors, with one CT motion or slice equal to 90 projection angles, considered to be 1 block when scanning (range = 1 6 blocks). The size of the field of view in the slice plane can be set by the operator. The sampling resolution, also selected by the user, is referred to as a voxel size (range = 0.2 0.8 mm 2 ). The CT scan speed, in millimeters per second (mm/s), can also be predetermined by the user (range = 10 50 mm/s). The number of blocks, voxel size, and scan speed are all parameters fixed at the time the scan is taken. Once the scan is completed, computer software offers numerous settings for analysis. A scan can be analyzed any number of times with these various settings. Analysis of all scans was done using research software version 5.40 Rev. B, which allows modes (options) for the following: detecting the outer bone edge (Contour Mode), defining the way subcortical and trabecular bone are separated (Peel Mode), and separating cortical bone from trabecular bone (CortMode). The CALCBD function of the software is used to obtain results for total cross-sectional area in millimeters squared and is a function of Contour Mode and Peel Mode selection. All of the mode settings used in our study were selected based on recommendations by the manufacturer and consensus of users working with this particular densitometer (Norland pqct Users Meeting, Sun Valley, ID; August 1998) To detect the outer bone edge, we chose Contour Mode 2. This mode utilizes an algorithm that sets a density threshold automatically and searches to find voxels that meet specified conditions to be marked as the outer bone edge. For separating subcortical from trabecular bone, Peel Mode 2, which implements an operatorselected threshold density of 400 mg/cm 3,was used. Any voxel with a density higher than this threshold is considered subcortical or cortical, whereas voxels with lower values are designated as trabecular. Another function, CORTBD, gives results for cortical density, cortical area, and cortical thickness. CortMode 1 was the setting for separating cortical from trabecular bone. In this mode, a density threshold is entered by the operator and voxels with a lower density are removed, leaving the cortical shell. The default threshold of 710 mg/cm 3 was used. Cortical thickness can be assessed using two different software procedures, either an iterative contour detection, which is a direct measurement of cortical thickness based on the true shape of the bone, or a threshold algorithm, often referred to as the circular ring model (2). In our study, the circular ring model for thickness with a threshold of 710 mg/cm 3 was used. Validation of the XCT 2000 was performed using a phantom supplied by Dr. Thomas Hangartner, Director, BioMedical Imaging Laboratory (Dayton, OH). This apparatus consisted of a Lucite cylinder that held an insert with a graduated thickness ranging from 0.3 mm to 4 mm. The inner diameter of the insert was 20 mm throughout, with the wall thickness equal to 4, 3, 2, 1, and 0.3 mm (see Fig. 1). To vary the density, two inserts were used, one of aluminum, the other of a poly(vinyl chloride) (PVC) material. Both the Lucite tube and insert were filled with distilled water at room temperature. For each Fig. 1. Phantom sketch showing a constant inner opening diameter and graduated wall thickness.
pqct Measurement of Bone Parameters in Children 11 thickness, on each insert, three scans were done using both one and two blocks with the following voxel/speed combinations: voxel = 0.4/speed = 20; voxel=0.3/speed = 15; voxel = 0.2/speed = 10. This amounted to a total of 18 scans at each thickness. A reproducibility study of the XCT 2000 involved duplicate scans on 11 children ages 3 5 yr. The radiation dose for each scan was 8 mrem (central dose supplied by the manufacturer) and written parental consent was obtained to do two scans at the 20% distal tibia site of the left leg. This site is located by calculating 20% of the total distance between the tibial tuberosity and the medial malleolus and marking that calculated distance from the distal end. The patient s leg was repositioned between the scans without the use of a scout view. Settings for the scan were voxel=0.4, speed = 20, and blocks = 1. Calculations for total cross-sectional area, cortical area, and cortical thickness were obtained using the software settings stated previously. Coefficients of variation (CV) were calculated on duplicate scans (3) as where CV = Sd of difference between two scans Overall mean for that measurement SD = sq rt ( sum (difference between two ) scans)2 2N either phantom at any voxel/speed combination. In both the PVC and aluminum phantom, cortical density measurements decreased significantly between the 1.0- and 0.3-mm wall thickness (p < 0.001). In the aluminum phantom, there was no significant difference between mean density readings for any speed when the wall thickness was greater than 1 mm (mean density = 1877.0 mg/cm 3 ). For the PVC insert, the smaller wall thickness had more of an effect on the density reading than the larger wall thickness, tested by the interaction of wall thickness and speed (p < 0.001) (see Fig. 2). For other significant differences in the mean density readings of the PVC insert, see Table 1. The total cross-sectional area and ring area were calculated for each thickness of the phantom and then compared to total cross-sectional area and cortical area readings. Calculated total cross-sectional area versus measured total cross-sectional area showed a strong correlation for both PVC (r = 0.962) and aluminum (r = 0.978), as did calculated ring area versus measured cortical area (r = 0.999 PVC; r = 0.997 Al). Phantom thickness versus measured cortical thickness also showed a strong correlation (r = 0.998 PVC; r = 0.990 Al) (see Fig. 3). Neither speed nor number of blocks had an effect on the ability of the machine to measure total cross-sectional area, cortical area, or cortical thickness. Baseline data from an ongoing exercise and calcium trial of 101 preschool children (53 male) were collected by pqct scan at the 20% distal tibia site of the left leg. Written parental consent was obtained before scanning. Settings for the scan were voxel = 0.4, speed = 20, and blocks = 1. Calculations for cortical density, total cross-sectional area, cortical area, and cortical thickness were obtained with the same software settings as stated previously. Height (cm) and weight (kg) were measured at the time of the scan. For tests involving children, protocols were approved by the Human Subjects Committee at South Dakota State University. Results Validation Results The number of blocks used (one vs two) to obtain the scan had no effect on cortical density readings for Fig. 2. PVC phantom density reading versus phantom wall thickness for scan speeds of ( ) 10 mm/s; ( ) 15 mm/s; ( ) 20 mm/s.
12 Binkley and Specker Table 1 Mean Density Readings (mg/cm 3 ) for Each PVC Phantom Wall Thickness Mean (SD) density reading per wall thickness Voxel/speed 0.3 mm 1 mm 2 mm 3 mm 4 mm Significant differences a 0.20/10 mm/s 853 (12) 952 (4) 963 (3) 963 (6) 956 (7) 0.3 mm 1, 2, 3, 4 mm 0.30/15 mm/s 778 (6.7) 926 (2) 952 (2) 951 (1) 945 (2) 0.3 mm 1, 2, 3, 4 mm 1 mm 0.3, 2, 3, 4 mm 4 mm 0.3, 1, 2 mm 0.40/10 mm/s 740 (7.0) 903 (3) 935 (1) 941 (1) 940 (2) 0.3 mm 1, 2, 3, 4 mm 1 mm 0.3, 2, 3, 4 mm a Significant difference in means (p < 0.05). Reproducibility Results Coefficient of variation for pqct bone parameters from duplicate scans in children aged 3 5 yr ranged from 3.6 7.8% (see Table 2). The child with the lowest mean cortical thickness (0.924 mm) had the largest discrepancy in cortical thickness measurements. When this child was excluded, the coefficients of variation were 3.1% for total area, 4.5% for cortical area, and 6.8% for cortical thickness. There was no obvious reason for the discrepant results, as both scans appeared acceptable. Preliminary Data Results Total cross-sectional area, cortical area, and cortical thickness were correlated with weight (r = 0.54, p < 0.001; r = 0.52, p < 0.001; r = 0.30, p = 0.002) and height (r = 0.41, p < 0.001; r = 0.55, p < 0.001; r = 0.40, p < 0.001). Because of the small cortical thickness at this age (mean = 1.2 mm), the relationship between cortical density and height/weight was not analyzed. There were no differences between females and males in mean ± SEM in total cross-sectional area (179.1 ± 4.5 vs 175.5 ± 4.4; p = 0.57), cortical area (50.7 ± 1.8 vs 53.2 ± 1.7; p = 0.29), or cortical thickness (1.19 ± 0.05 vs 1.26 ± 0.05; p = 0.29). In a regression model with height and weight as predictors, height was the only significant predictor of cortical thickness (p = 0.006), whereas total crosssectional area was predicted by weight (p < 0.001). Both height (p = 0.005) and weight (p = 0.05) predicted the cortical area (see Fig. 4). Discussion Although dual-energy X-ray absorptiometry (DXA) is considered the method of choice for evaluating bone mineral status, pqct offers two possible advantages for measurement of bone in growing children. One advantage is that the true volumetric bone mineral density is measured. The other improvement is the ability to measure total cross-sectional and cortical area of bone and cortical thickness. In young children, volumetric density may be a better expression for BMD than that given by the areal density measurement from DXA, which is influenced by bone size (4). The results of our study show that pqct, using the XCT 2000 clinical model, can be used to measure cortical density accurately when the cortical thickness of the bone is 2 mm or greater with voxel/speed sizes of 0.40/20 mm/s. A study by Hangartner and Gilsanz found the minimal thickness to be 2 2.5 mm for accurate density readings using the same phantom on two different QCT machines (5). Unfortunately, the mean cortical thickness of the 20% distal tibia in 3- and 4- yr-old children (mean = 1.2 mm) is too small to get an accurate pqct cortical density reading with the above voxel/speed settings. The other benefit of pqct, the ability to differentiate cortical bone, makes geometric assessments possible. Although DXA is beneficial in obtaining total bone mineral context (BMC) changes during growth (6), pqct may be helpful in determining particular changes in bone parameters such as total
pqct Measurement of Bone Parameters in Children 13 Table 2 Coefficient of Variation for Repeated Scans in Children Aged 3 5 yr % CV (N = 11) Mean (N = 11) Total area (cm 2 ) 3.6 191 Cortical area (cm 2 ) 5.4 54 Cortical thickness (mm) 7.8 1.19 Fig. 3. Phantom pqct results for PVC insert ( ) and aluminum insert ( ) compared to calculated or known values ( ) for each wall thickness. Top panel: total cross-sectional area results versus calculated crosssectional area for each wall thickness (r = 0.962 PVC; r = 0.978 Al); middle panel: cortical area results versus calculated ring area for each wall thickness (r = 0.999 PVC; r = 0.997 Al); bottom panel: thickness results versus known wall thickness (r = 0.998 PVC; r = 0.990 Al). cross-sectional area, cortical area, and cortical thickness. As growth occurs, changes in these bone parameters may be observed. When testing various scan speeds, we found that scan speed had a small effect on density at the smaller phantom wall thickness. Although there were no significant differences in measured density in either the PVC or aluminum insert at thickness sites 1 mm or greater using the 10 mm/s speed, there is a feasibility issue when measuring young children because of the time it takes to complete a scan at this speed. At a speed setting of 20 mm/s, it takes approximately 104 s to complete a scan at the 20% distal tibia site, whereas at the same site, using a speed of 10 mm/s, it would take about 181 s. In a study such as ours, where the subjects are 3 and 4 yr olds, time becomes a critical factor in obtaining an acceptable scan, and we have chosen the 20-mm/s speed. Preliminary results from the preschool study show that the mean cortical thickness of the 20% distal tibia site is too small to accurately analyze cortical density. Other bone parameters measured at this age can be measured using pqct. We found no difference in bone size because of gender. Inconsistent findings have been reported as to whether bone size differs with gender at this age, which may be caused in part by the use of different absorptiometers, differences in sites measured, or sample sizes insufficient to detect differences (7 10). Findings from this study indicate that pqct can be used to accurately measure total cross-sectional area, cortical area, and cortical thickness in the bones of young children 3 to 4 yr of age at a scan speed of 20 mm/s and voxel size of 0.40 mm. Total cross-sectional area, cortical area, and cortical thickness increase with increasing height and weight, but no gender differences in bone size were noted. Acknowledgments The authors gratefully acknowledge G. Niemeyer and D. Schiferl, Norland/Stratec Medical Systems,
14 Binkley and Specker Fort Atkinson, WI, for their technical expertise. Also, many thanks to T. Hangartner, Biomedical Imaging Laboratory, Dayton, OH, for allowing the use of his phantom. Fig. 4. pqct geometric bone parameter results from 3- to 4-yr-old children scanned at the 20% distal tibia site (N = 101). Top panel: total cross-sectional area versus weight (r = 0.54, p < 0.001); middle panel: cortical area versus height (r = 0.55, p < 0.001); bottom panel: cortical thickness versus height (r = 0.41, p < 0.001). References 1. Schonau E. 1998 The development of the skeletal system in children and the influence of muscular strength. Horm Res 49:27 31. 2. Louis O, Willnecker J, Soykens S, Van den Winkel P, Osteaux M. 1995 Cortical thickness assessed by peripheral quantitative computed tomography: accuracy evaluated on radius specimens. Osteoporos Int 5:446 449. 3. Mueller WH, Martorell R. 1988 Reliability and Accuracy of Measurement. In: Anthropometric Standardization Reference Manual, Lohman T, Roche A, Martorell R, eds. Human Kinetics Books, Champaign, IL, 84. 4. Prentice A, Parsons T, Cole T. 1994 Uncritical use of bone mineral density in absorptiometry may lead to size related artifacts in the identification of bone mineral determinants. Am J Clin Nutr 60:837 842. 5. Hangartner T, Gilsanz V. 1996 Evaluation of cortical bone by computed tomography. J Bone Miner Res 11:1518 1524. 6. Brunton A, Weiler H, Atkinson S. 1997 Improvement in the accuracy of dual energy x-ray absorptiometry for whole body an regional analysis of body composition: validation of using piglets and methodologic considerations in infants. Pediatr Res 41:590 596. 7. Gilsanz V, Boechat MI, Roe TF, Loro ML, Sayre JW, Goodman WG. 1994 Gender differences in vertebral body sizes in children and adolescents. Radiology 190:673 677. 8. Henderson RC. 1991 Assessment of bone mineral content in children. J Pediatr Orthop 11:314 317. 9. Ruprich RC, Specker BL, Lieuw-A-Fa M, Ho M. 1996 Gender and race differences in bone mass during infancy. Calcif Tissue Int 58:395 397. 10. Southard RN, Morris JD, Mahan JD, et al. 1991 Bone mass in healthy children: measurement with quantitative DXA. Radiology 179:735 738.