DXA Estimates of Vertebral Volumetric Bone Mineral Density in Children: Potential Advantages of Paired Posteroanterior and Lateral Scans

DXA Estimates of Vertebral Volumetric Bone Mineral Density in Children: Potential Advantages of Paired Posteroanterior and Lateral Scans

Journal of Clinical Densitometry, vol. 9, no. 3, 265e273, 2006 Ó Copyright 2006 by The International Society for Clinical Densitometry 1094-6950/06/9:...

388KB Sizes 0 Downloads 4 Views

Journal of Clinical Densitometry, vol. 9, no. 3, 265e273, 2006 Ó Copyright 2006 by The International Society for Clinical Densitometry 1094-6950/06/9:265e273/$32.00 DOI: 10.1016/j.jocd.2006.05.008

Original Article

DXA Estimates of Vertebral Volumetric Bone Mineral Density in Children: Potential Advantages of Paired Posteroanterior and Lateral Scans Mary B. Leonard,*,1,2 Justine Shults,2 and Babette S. Zemel1 1

Department of Pediatrics, The Children’s Hospital of Philadelphia; and 2Department of Biostatistics and Epidemiology, University of Pennsylvania School of Medicine, Philadelphia, PA

Abstract Dual-energy X-ray absorptiometry (DXA) estimates of areal bone mineral density (BMD) are confounded by bone size in children. Two strategies have been proposed to estimate vertebral volumetric BMD: (1) bone mineral apparent density (BMAD) is based on the posteroanterior (PA) spine scan; (2) width-adjusted bone mineral density (WABMD) is based on paired PA lateral scans. The objective of this study was to compare DXA estimates of vertebral bone mineral content (BMC), volume and volumetric BMD obtained from Hologic PA scans (Hologic, Inc., Bedford, MA) alone, and paired PA lateral scans in 124 healthy children, ages 4 to 20 yr. The PA scans were used to estimate bone volume (PA Volume) as (PA Area)1.5 and BMAD as [(PA BMC)/(PA Volume)]. Paired PA lateral scans were used to estimate width-adjusted bone volume (WA Volume) as [(p/4)(PA width)(lateral depth)(vertebral height)] and WABMD as [(lateral BMC)/(WA Volume)]. Generalized estimating equations were used to compare the relationship between scan type (PA vs. paired PA lateral) and bone outcomes, and the effects of height and maturation on this relationship. The estimates of BMC and volume derived from PA scans and paired PA lateral scans were highly correlated (r O 0.97); WABMD and BMAD were less correlated (r 5 0.81). The increases in BMC, volume, and volumetric BMD with greater height and maturation were significantly larger (all p ! 0.001) when estimated from paired PA lateral scans, compared with PA scans alone. The proportion of spine BMC contained within the vertebral body, versus the cortical spinous processes, increased significantly with age ( p ! 0.001) from 28% to 69%. The smaller increases in bone measures on PA scans may have been due to magnification error by the fan beam as posterior tissue thickness increased in taller, more mature subjects, and the distance of the vertebrae from the X-ray source increased. In conclusion, paired Hologic PA lateral scans may increase sensitivity to growth-related increases in trabecular BMC and density in the spine, with less bias due to magnification error. Key Words: Bone mineral density; children; DXA; growth, lumbar spine.

bone dimensions (1e3). Recent years have seen a dramatic increase in studies of the effects of varied childhood diseases on bone health. Dual energy X-ray absorptiometry (DXA) is by far the most common method for the quantitative assessment of bone; however, DXA is significantly limited by the reliance on areal measures of BMD in children of varied size. Areal BMD (g/cm2) is generated by dividing the bone mineral content (BMC) (g) within a defined anatomical region by the projected area of the bone (cm2). This is not a measure of volumetric density (g/cm3), because it provides no information about the depth of bone. Areal BMD

Introduction The increase in bone strength throughout growth and development is the result of gender- and maturation-specific increases in volumetric bone mineral density (BMD) and Received 02/12/06; Revised 05/16/06; Accepted 05/17/06. *Address correspondence to: Mary B. Leonard, MD, MSCE, Division of Nephrology, The Children’s Hospital of Philadelphia, 34th St. and Civic Center Blvd., Philadelphia, PA 19104. E-mail: [email protected]

265

266 underestimates the bone density of shorter individuals. Therefore, the uncritical use of areal BMD may lead to spurious associations with other body size-dependent measures, such as muscle strength or dietary intake (4). Varied strategies have been proposed to express DXA spine BMD in a form that is less sensitive to differences in skeletal size in children (5,6). These strategies use the dimensions of the projected posteroanterior (PA) vertebral bone area to estimate vertebral depth and volume. The vertebral PA BMC is then divided by this estimate of vertebral volume to generate bone mineral apparent density (BMAD). These strategies have been used to distinguish between increases in volumetric BMD and bone size in children and adolescents (7,8). An additional limitation of DXA estimates of BMC and bone area is that the fan-beam densitometers produce a magnification error as a function of the distance of the vertebrae from the X-ray source (9). For example, a smaller (thinner) child’s vertebrae will lie closer to the source of the beam, with a consequent overestimation of bone area and BMC (10). This is of particular importance when comparing children of widely variable body size and thickness (11). An alternative approach is to use paired PA lateral vertebral scans to measure vertebral width, height, and depth to estimate bone volume and volumetric BMD. Although this approach requires an additional scan, the paired PA lateral scans offer 3 advantages for the assessment of vertebral volumetric BMD: (1) the addition of the lateral spine permits direct measurement of bone depth, rather than estimating depth from the PA dimensions; (2) the lateral image is edited to isolate the vertebral body, excluding the BMC within the cortical spinous process; and (3) when measured in the supine position with a C-arm, the vertebrae lie a constant distance from the X-ray source and lateral measures are not subject to magnification effects as a function of subject size. Studies in adults have demonstrated that paired PA lateral scans provide better capability for discriminating between vertebral fracture and nonfracture patients compared with BMAD or PA areal BMD (12). Paired PA lateral vertebral scans have been described in healthy children (13,14) and in children enrolled in an intervention study (15); however, no studies have compared PA and paired PA lateral scans to address the potential advantages of paired scans in children. The objective of this study was to compare estimates of BMC, bone volume and volumetric BMD obtained from conventional PA scans and from paired PA lateral scans in healthy children, and to identify sources of variability between these two approaches.

Patients and Methods Study Subjects Lumbar spine scans were obtained in children and adolescents, ages 4 to 20 yr, enrolled as healthy controls for ongoing bone studies in the Nutrition and Growth Laboratory at the Children’s Hospital of Philadelphia (CHOP). The protocol was approved by the institutional review board; all subjects Journal of Clinical Densitometry

Leonard et al. and parents provided written informed consent. Subjects were recruited from the CHOP general pediatric clinics, as well as from the surrounding community using newspaper advertisements and flyers. Subjects with chronic medical conditions or medications that potentially affect growth, pubertal development, nutritional status or dietary intake were excluded.

Anthropometry and Tanner Staging All measurements were performed in the Nutrition and Growth Laboratory at the time of the DXA scans. The measurements consisted of weight, using a digital electronic stand-on scale and height using a wall-mounted Holtain stadiometer (Holtain, Ltd., Crosswell, UK). Pubertal status was determined by physical examination and was classified according to the method of Tanner (16). Age- and gender-specific Z-scores (standard deviation scores) for height, weight, and body mass index (BMI) were calculated using the National Center for Health Statistics 2000 Center for Disease Control growth charts (17). Obesity was defined as BMI greater than the 95th percentile for age and gender (18).

DXA Lumbar Spine Scans The DXA scans of the PA spine (L1eL4) were performed using a Hologic QDR 2000 bone densitometer (software version 4.74; Hologic Inc., Bedford, MA) with a fan beam in the array mode for all subjects. The measurements were performed using standard supine positioning techniques. After completion of the PA scan, the X-ray tube and detector C-arm assembly were rotated 90  , and the lateral scan (L2eL4) was performed without the subject moving from the supine position. All Hologic scanners limit lateral scan acquisition to L2eL4 due to rib interference with L1; therefore, the PA and lateral DXA data in this study are limited to L2eL4. The horizontal lines designating L2eL4 on the lateral scan are generated automatically based on the line placement on the PA scan (Fig. 1). Quality-control PA scans were performed daily using a simulated lumbar spine made of hydroxyapatite encased in epoxy resin. The age-specific coefficients of variation percentage for spine PA BMD were 0.64%, 0.99%, and 1.03% in children ages 6 to 9, 10 to 13, and 14 to 16 yr, respectively (19). A recent study reported the effective radiation dose equivalent from Hologic PA DXA scans of the lumbar spine in children ages 1, 5, 10, and 15 yr, and for adults (20). The effective dose decreased as age increased; the doses were 4.7 microSv and 2.2 microSV in a 1-year-old and an adult, respectively. To our knowledge, the effective dose of lateral spine DXA has not been determined in children; however, the effective dose for lateral spine DXA in adults is less than half the effective dose for PA spine DXA (21). The DXA spine scans were analyzed to generate PA and lateral bone area and BMC. As shown in Fig. 1, the PA measures included the vertebral body and superimposed cortical spinous processes, whereas the region of interest in the lateral scan is defined by the posterior and anterior margins of the vertebral body, excluding the posterior spinous processes. Volume 9, 2006

PA and Lateral Spine Scans in Children

267

PA

Lateral

Fig. 1. Paired posterioranterior and lateral lumbar spine dual-energy X-ray absorptiometry (DXA) scans. Using the rotating C-arm, there is an exact line-by-line match between the paired scans. Vertebral body area and bone mineral content (BMC) are measured on the lateral scan anterior to the vertical line intersecting the neural arches. The arrow on the lateral scan indicates the distance from the DXA table to the mid-point of L3; this provides an estimate of posterior soft tissue thickness (lift). The vertical lines defining the region of interest on the lateral scan are placed by the operator, as described in the Hologic Manual. Vertebral volume and volumetric BMD were estimated from the PA scan alone and from the paired PA lateral scans as summarized in Table 1. The estimates based on the PA scan assumed that the average depth of the vertebrae scaled proportionate to the square root of the projected PA area (5); therefore, the volume was estimated as (PA area)1.5 , termed ‘‘PA volume.’’ The BMC measured on the PA scan was then divided by this estimate of volume to generate the BMAD. The estimates based on the paired PA lateral scans were calculated automatically with the manufacturer software that assumed that the vertebral body was an elliptical cylinder (22). The width of the vertebrae on the PA scan was used as an estimate of the major axis of the ellipse and the depth of the vertebrae on the lateral scan was used as the estimate of the minor axis of the ellipse. Volume was calculated as [(p/ 4)(PA vertebral width)(lateral depth)(vertebral height)], termed ‘‘width-adjusted volume’’ (WA volume). The BMC

measured on the lateral scan was then divided by this estimate of volume to generate ‘‘width-adjusted volumetric BMD’’ (WABMD). The use of a fan beam results in magnification error in the PA scan (10,11,23); therefore, an additional measure was obtained to estimate the magnitude of the magnification error. The fan beam in the Hologic QDR 2000 projects from under the examination table and the detectors are above the table. As the distance between the vertebral body and the source of fan beam increases (e.g., overweight subjects), the bone area and BMC are underestimated. In contrast, in small or thin subjects the bone area and BMC are overestimated. To assess the impact of magnification error on the PA scan results, the distance of the vertebral body above the fan beam source was estimated by measuring the distance from the posterior edge of the lateral spine image to the midpoint of the mid-vertebrae (L3) using digital calipers (Fig. 1). Although this was not a true measure of the tissue thickness, the distance on the image scaled proportionately to the true tissue thickness (personal communication, Tom Kelly, Hologic,

Table 1 Estimates of Vertebral Volume (cm3) and Volumetric BMD (g/cm3)

Vertebral volume Volumetric BMD

PA scan alone

Paired PA lateral scans

PA volume 5 (PA area)1.5 BMAD 5 (PA BMC)/(PA volume)

WA volume 5 [(p/4)(PA width)(lateral depth)(height)] WABMD [ (lateral BMC)/(WA volume)

Abbr: BMC, bone mineral content; BMD, bone mineral density; PA, posteroanterior; WA, width-adjusted; WABMD, width-adjusted bone mineral density. Journal of Clinical Densitometry

Volume 9, 2006

268

Leonard et al.

Inc). A radiologic ruler was included in the scan field with the phantom and was used to convert the distance on the image to centimeters of soft tissue thickness (i.e., each 1 cm on the printed DXA image represented 2.08 cm elevation of the vertebrae above the table). Therefore, the distance on each lateral image (see arrow in Fig. 1) was multiplied by 2.08 to provide an estimate of the true distance (cm) from the DXA table to the midpoint of L3; this distance was named ‘‘lift.’’ During the lateral scan, the subject is positioned in the center of the table, a constant distance from the rotated C-arm Xray source; therefore, magnification error on the lateral scan did not vary as a function of subject size. The subjects were positioned by a skilled research technician with the midline of the subject positioned over the midline demarcation on the DXA table.

necessary to improve linearity and eliminate heteroscedascity. The assumptions of the regression models were assessed by the Shapiro-Wilk test of normality of residuals and the Cook-Weisburg test for heteroscedascity. The correlations between bone and anthropometric measures were assessed using Pearson product-moment estimates. The correlation measures the strength of a relation between two variables, not the agreement between them. For example, in a graph of BMAD versus WABMD, if the points all lie along the line of equality, then there is perfect agreement. Conversely, if there is perfect correlation, then the points will lie along any straight line, that is not necessarily BMAD equals WABMD. The traditional Bland-Altman method for assessing agreement between two methods is not appropriate here, because the mean values and variance of the PA and paired PA lateral results differ (24), as shown in Fig. 2. Therefore, the bone estimates obtained from the PA scan alone and from the paired PA lateral scans were compared using generalized estimating equations (GEE), as detailed below (25). The volumetric approaches evaluated here were first developed as a means to adjust for differences in skeletal size; therefore, the BMC results (PA BMC and lateral BMC) and estimates of vertebral volume (PA volume and WA volume)

Statistical Analysis Analyses were conducted using STATA 8.0 (Stata Corp., College Station, TX). Two-sided tests of hypotheses were used and a p value !0.05 was considered statistically significant. The relations between bone measures, age, height, and Tanner stage were explored graphically (Figs. 2e6). Natural log transformations of bone area, BMC, and volume were A

B

500

80 70

400

PA Volume

PA BMC

60 50 40 30 20

300

200 100

10 0

0 0

10

20

30

40

50

60

70

80

0

100

C

200

300

400

500

WA Volume

Lateral BMC 0.30

BMAD

0.25 0.20 0.15 0.10 0.05 0.05

0.1

0.15

0.2

0.25

0.3

WABMD

Fig. 2. Relations between posteroanterior (PA) and paired PA lateral results. The results from the PA scan are shown on the Y-axis and the results from the paired PA lateral scans are shown on the X-axis. The solid line represents the line of identity. (A) Bone mineral content (BMC) (g). (B) Vertebral volume (cm3). (C) Vertebral volumetric bone mineral density (g/cm3). Journal of Clinical Densitometry

Volume 9, 2006

PA and Lateral Spine Scans in Children

269

80

6.5 PA Volume

6.0

Log Vertebral Volume

70

Percent

60 50 40

WA Volume

5.5 5.0 4.5 4.0 3.5 3.0

30

2.5 100

110

120

140

150

160

170

180

190

Height (cm)

20 4

6

8

10

12

14

16

18

20

22

Age (yr)

Fig. 3. The proportion of the total vertebral bone mineral content (BMC) contained with the vertebral body increases with age. The percentage of the total vertebral BMC contained within the vertebral body was calculated as (lateral BMC/posteroanterior [PA] BMC)  100. were assessed relative to subject height. Because vertebral volumetric BMD increases with puberty (2,26,27), the estimates of volumetric density (BMAD and WABMD) were assessed relative to the Tanner stage. The GEE analyses were used to determine if the results from the PA alone and paired PA lateral scans differed relative to height or Tanner stage. The GEE approach is similar to conducting linear regression, but also adjusts for the fact that two sources of bone measurements (PA alone and paired PA lateral) were collected on each subject. For example, to determine if measures from the PA alone and paired PA lateral scans differed relative to height, the GEE analysis involved first specifying a 2  2 correlation structure to describe the pattern of association between the PA and paired PA lateral scans within each subject. A regression model was then proposed for the expected value of the bone

Fig. 5. Comparison of estimates of vertebral volume relative to body height. measurements that included height, an indicator variable (i[PA]) that took value one for the PA measurement and took value zero for the paired PA lateral scan, and an interaction term height  i(PA) that was constructed as the multiple of height and the indicator variable. An interaction term that differs significantly from zero would indicate that there is a significant difference in the slope of the lines describing the relationship between PA and height versus the paired PA lateral scan versus height. The regression models were also modified to include lift and lift-(i[PA]) interactions to determine if the difference in slopes remained after adjustment for magnification error. The analysis of PA volume and WA volume relative to height was equivalent to the analysis just described. The analysis of BMAD and WABMD relative to the Tanner stage was similar, but with a different regression model for volumetric BMD that included indicator variables for each Tanner stage, with Tanner 1 as the referent stage. Each analysis was also tested for interactions between the scan type (i[PA]) and gender and race to determine if any 0.30

PA BMC

Volumetric BMD

4.0

BMAD WABMD

0.25

4.5 Lateral BMC

3.5

Log BMC

130

3.0 2.5

0.20 0.15 0.10

2.0 0.05

1.5 1.0 100

0.00 110

120

130

140

150

160

170

180

190

Height (cm)

Fig. 4. Comparison of posteroanterior (PA) and lateral bone mineral content (BMC) relative to body height. Journal of Clinical Densitometry

0

1

2

3

4

5

Tanner Stage

Fig. 6. Comparison of estimates of vertebral volumetric bone mineral density (BMD) relative to Tanner stage. Volume 9, 2006

270

Leonard et al.

observed differences between the PA and paired PA lateral results differed according to race or gender.

Results Subject Characteristics Paired PA lateral lumbar spine scans were obtained in 124 children and adolescents (i.e., 46 males and 78 females), ages 4 to 20 yr. The mean (6SD) weight Z-scores were 0.38 6 1.00; height Z-scores 0.32 6 1.04, and BMI Z-scores 0.27 6 1.06, illustrating that these subjects have growth parameters comparable with the United States pediatric population (17). Sixteen percent of the subjects were obese; the prevalence of obesity in children and adolescents nationwide is 15.5% (17). Fifty-three percent of subjects were Caucasian, 35% were African American, and 2% were Hispanic; the remainder was of varied racial background. Fifty-six percent of subjects were pre-pubertal (Tanner stage I). Lift (i.e., the estimate of posterior soft tissue thickness), varied from 3.8 to 10.5 cm (mean 6 SD: 5.6 6 1.4). In univariate analyses, lift was significantly correlated ( p ! 0.001) with height (R2 5 0.50) and with BMI (R2 5 0.77). The BMI and lift were both significantly greater in females compared with males ( p ! 0.001). In a multivariate analysis, lift was independently correlated with BMI and height (R2 5 0.84; both p ! 0.001). That is, in two subjects of equal height, the lumbar vertebrae were elevated a significantly greater distance off the DXA table in the subject with greater BMI.

Univariate Correlations Between Bone and Body Size Measures Table 2 summarizes the univariate relations between PA and lateral spine bone measures and subject height. All of the measures were positively correlated (all p ! 0.001). Although the vertebral volumes (PA Volume and WA Volume) were highly correlated (r 5 0.98) and the PA and lateral BMC were highly correlated (r 5 0.97), the estimates of vertebral volumetric BMD (WABMD and BMAD) were less highly correlated (r 5 0.81). The relations between PA and

paired PA lateral estimates of BMC, vertebral volume and volumetric BMD, along with the lines of identity are shown in Fig. 2. Although the PA and paired PA lateral measures were highly correlated, the values differed systematically. The deviation from the line of identity in Fig. 2 illustrates the poor agreement between measures. The PA measures of BMC were consistently greater than the lateral measures due to the exclusion of the posterior spinous processes from the lateral scan. On average, the BMC within the vertebral body, as measured on the lateral scan, contributed 44% of the total vertebral BMC, as measured on the PA scan. The percentage of the spine BMC within the vertebral body was calculated as (lateral BMC/PA BMC)  100. The percentage increased significantly with age ( p ! 0.0001), ranging from 28% in the younger subjects to 69% in the mature subjects (Fig. 3). The volume estimates were also markedly different. The PA volume was approximately 3-fold greater than the WA volume.

PA and Lateral BMC The PA and lateral BMC were assessed relative to height. Fig. 4 demonstrates that PA and lateral BMC increased with height, and that lateral BMC increased more rapidly with increasing height. Use of a GEE model confirmed that the slope was significantly greater for lateral BMC compared with PA BMC ( p ! 0.001): each 1 cm increase in height was associated with a 0.7% greater PA BMC, but a 1.3% greater lateral BMC. There was no evidence of an interaction between scan type (PA alone vs. paired PA lateral) and gender or race. The lesser slope for PA BMC relative to height may be due to magnification error underestimating PA BMC in the taller, and likely ‘‘thicker’’ subjects. Therefore, the measure of posterior soft tissue thickness (i.e., lift) obtained on the lateral scan was added to the model. Lift was significantly and positively associated with PA BMC and with lateral BMC ( p ! 0.02). The differences in slope between lateral BMC and PA BMC relative to height remained significant ( p ! 0.001), adjusted for lift. However, there was

Table 2 Correlations (r) Between Height, PA, and Paired PA Lateral Spine Bone Measures PA alone Height Height Paired PA lateral Lateral area Lateral BMC WA volume WABMD

0.93 0.88 0.93 0.66

PA area

PA BMC

PA volume

BMAD

0.95

0.87

0.93

0.48

0.95 0.97 0.98 0.81

Note: All p ! 0.0001. Abbr: BMAD, bone mineral apparent density; BMC, bone mineral content; PA, posteroanterior; WA, width-adjusted; WABMD, width-adjusted bone mineral density. Journal of Clinical Densitometry

Volume 9, 2006

PA and Lateral Spine Scans in Children a significant interaction ( p 5 0.001) between lift and the source of the BMC (PA vs. lateral). That is, as hypothesized, greater posterior tissue thickness in the taller subjects contributed to the differences in slope between PA and lateral BMC. The adjusted models were as follows: Ln (PA BMC) 5 0.123 þ (0.0067) height þ (0.0012) lift. The Ln (lateral BMC) 5 1.007 þ (0.0117) height þ (0.0031) lift. These models suggest that (for subjects of comparable height) greater lift (suggesting greater BMI) is associated with greater BMC. However, the increase in PA BMC for a given increase in lift is significantly less than the increase in lateral BMC with the same increase in lift. This significant difference in the effect of lift on BMC between scan types may be due to underestimation of PA BMC secondary to magnification error.

PA Volume and WA Volume The PA volume and WA volume were assessed relative to height. Fig. 5 demonstrates that both measures of vertebral volume increased with height. Use of a GEE model confirmed that the slope relative to height was significantly greater for WA volume compared with PA volume ( p ! 0.001), that is, each 1 cm increase in height was associated with a 2.3 % greater WA volume, but a 1.9 % greater PA volume. There was no evidence of an interaction between scan type and gender or race. The lesser slope for PA volume may be related to magnification error. Therefore, the measure of lift obtained on the lateral scan was added to the model. Lift was not significantly associated with WA volume or PA volume. The differences in slope between WA volume and PA volume relative to height remained significant ( p ! 0.001), adjusted for lift. However, there was a significant interaction ( p ! 0.001) between lift and the source of the volume estimate (PA volume vs. WA volume). Again, greater posterior tissue thickness in the taller subjects contributed to the differences in slope. The adjusted models were as follows: the Ln (PA volume) 5 2.63 þ (0.020) height  (0.002) lift. The Ln (WA volume) 5 0.781 þ (0.023) height þ (0.002) lift.

BMAD and WABMD The BMAD and WABMD were assessed relative to the Tanner stage. Fig. 6 demonstrates that both measures of volumetric BMD increased with maturation. The WABMD was significantly greater in the Tanner stage 2 subjects compared with the Tanner stage 1 subjects. The BMAD did not differ significantly between Tanner stages 1 and 2. Both WABMD and BMAD were significantly greater for subjects at each Tanner stage (i.e., stages 3 through 5 compared with Tanner stage 1 [all p ! 0.001]). Use of a GEE model confirmed that the magnitude of the increase in volumetric BMD with the Tanner stage was significantly greater for WABMD than for BMAD ( p ! 0.001). There was no evidence of an interaction between scan type and gender or race. The measure of lift was included in the model. The differences in the increases in the two measures of volumetric BMD with maturation remained significant ( p ! 0.001), Journal of Clinical Densitometry

271 adjusted for lift. In addition, there was a significant interaction ( p ! 0.001) between lift and the source of the volumetric BMD estimate (BMAD vs. WABMD). That is, greater lift in the more mature subjects contributed to the differences in slope between BMAD and WABMD relative to maturation stage. The models for the volumetric BMD estimates are as follows: the WABMD 5 0.1176 þ (0.0011) lift þ x; x 5 0 for Tanner stage 1, and 0.004, 0.0130, 0.0246, and 0.0273 for Tanner stages 2, 3, 4, and 5, respectively. The BMAD 5 0.0847 þ (0.0005) lift þ x; x 5 0 for Tanner stage 1, and 0.0020, 0.0059, 0.0145, and 0.0222 for Tanner stages 2, 3, 4, and 5, respectively. Greater lift was associated with significantly greater WABMD and BMAD (both p ! 0.001). However, within each Tanner stage, a given increase in lift resulted in a lesser increase in BMAD compared with WABMD, suggesting magnification error resulted in an underestimation of volumetric BMD.

Discussion The data presented here in 124 children demonstrated systematic differences in estimates of vertebral BMC, volume, and volumetric BMD obtained from PA scans alone compared with paired PA lateral scans. The systematic differences in PA BMC and lateral BMC were due, in part, to the fact that the lateral spine scans selectively measured the trabecular-rich vertebral bodies without the contributions of the superimposed cortical posterior elements of the spine included in the PA scan. Prior studies have demonstrated that vertebral trabecular volumetric BMD increases significantly during growth and maturation, whereas cortical volumetric BMD remains relatively constant (2,3). Therefore, isolation of the predominantly trabecular vertebral body on the lateral spine highlights growth related increases in trabecular BMC. This is demonstrated by the relatively greater increase in lateral BMC for height compared with PA BMC for height (Fig. 4), and by the observation that an increasing proportion of the L2eL4 BMC is contained within the vertebral body (Fig. 3). The BMAD and WABMD are calculated using the PA BMC and lateral BMC, respectively. Therefore, WABMD may be more sensitive than PA BMC to growth related increases in volumetric BMD in the predominantly trabecular vertebral body. This is supported by the observation of a greater rate of increase in WABMD relative to increasing Tanner stage. These findings are analogous to reports in the elderly that lateral DXA was significantly more sensitive than PA DXA to aging related bone loss in males and females (28e30). In addition, paired PA lateral scans in postmenopausal women were more strongly associated with vertebral fracture than estimates of volumetric BMD based on PA DXA scans (12). The differences in PA volume and WA volume are due to the different assumptions regarding bone geometry. The WA volume is based on the assumption that the vertebral body is approximated by an ellipse, as described in the Methods. In contrast, PA volume is based on the assumption that the Volume 9, 2006

272 average vertebral depth is proportional to the square root of the projected PA bone area (5). That is, PA volume is proportional to, not equivalent to the true bone volume. This is illustrated by our observation that PA volume is approximately 3-fold greater than WA volume. The small, but statistically significant differences in the slopes of WA volume and PA volume with increasing height (Fig. 4) may be due to changes in vertebral width and depth with growth. To our knowledge, our study is the first to address the contribution of magnification error to the differences in PA and lateral measures. Lunar-GE Healthcare (Waukesha, WI) and Hologic QDR (Hologic, Inc., Bedford, MA) scanners have largely converted to fan-beam techniques that introduce magnification errors in measures of bone area and BMC (10). The earlier DXA scanners used a pencil beam design that used a well-collimated pencil beam X-ray coupled to a single detector that moved in a raster pattern across the subject, thus obtaining a true magnification of the region scanned without significant magnification of the scanned bones (10,23). Increasingly, pencil beam scanners have been replaced by a multiple detector array design coupled to a fan-shaped X-ray beam. These fan beam densitometers acquire scans by performing a single sweep of the subject with substantially shorter scanning times and improved geometric resolution. However, the fan beam technique causes an inherent magnification of scanned structures as the distance decreases from the X-ray source, and hence from the apex of the fan beam (10,23). The magnification affects the medial-lateral dimension of the scanned structure, but not the cranial-caudal dimension. In the case of Hologic scanners, the apex of the fan beam is under the subject. Therefore, as the bone is elevated a greater distance off the table, bone geometry and BMC are underestimated. Pocock, et al. (10) have shown that variations in soft tissue thickness between subjects, by altering the distance of the skeleton from the fan beam X-ray source, are sufficiently large to cause clinically significant errors in measures of bone area and BMC (10,23). Of note, the Lunar-GE Healthcare currently manufactures a narrow-angle fan beam densitometer (Lunar Prodigy) that scans in a rectilinear fashion, much like the original pencil beam machines, but with rapid scan time. Two studies demonstrate that this technique is subject to negligible magnification effects (31,32); this latest advance may be particularly important for pediatric applications that rely on PA scans. All measures obtained from the PA scan alone were subject to magnification error that varied as a function of posterior soft tissue thickness. Lateral BMC, area and depth were not subject to magnification error that varied as a function of patient size, because the spine was positioned at a constant distance from the X-ray beam. The WA volume was based on bone width from the PA scan that was subject to magnification error, and bone depth from the lateral scan that was not subject to magnification error. While WA volume is subject to some magnification error, PA volume is subject to greater distortion by magnification error, because the error is amplified when bone area is raised to the 1.5 power in the calculation of PA volume. Of note, only Hologic scanners use C-arm Journal of Clinical Densitometry

Leonard et al. technology to obtain the lateral scan with the subject in a supine position. The Lunar-GE and Norland Medical Systems, Inc. (Fort Atkinson, WI) machines require that that the patient rolls on their side. The Lunar-GE lateral scans are subject to variable magnification as a function of patient size when assessed with fan beam technology. The Norland DXA machines only use pencil beam technology; therefore, they are not subject to magnification error. To assess the impact of magnification error on the difference between PA and paired PA lateral scans, the distance of the vertebrae relative to the table was estimated from the lateral scan. A direct measure of soft tissue thickness was necessary. Adjustment for BMI or weight and height would not accurately characterize lift because variability in fat distribution and body composition results in varying posterior tissue thickness. Furthermore, greater BMI likely results in true increases in BMC and volumetric BMD (33,34). Therefore, adjustment for BMI would not isolate the effect of magnification error from the true effects of BMI on bone mass. The data presented here confirm that magnification error contributes to the observed difference in BMC, bone volume, and volumetric BMD relative to height and Tanner stage. We hypothesized that taller, more mature subjects would have greater posterior tissue thickness that would result in a progressive underestimation of PA BMC, volume, and volumetric BMD with increasing height and maturation, compared with the paired PA lateral scans. These data demonstrate that Hologic paired PA lateral scans have multiple advantages that will likely increase their utility in pediatric studies. First, the BMC is limited to the vertebral body (excluding spinous processes), increasing sensitivity to growth-related increases in trabecular BMC. Second, the vertebral volume is based on measured PA and lateral dimensions, likely providing a more accurate estimate of volume as dimensions change with growth. Third, the lateral scans are not subject to magnification error. The absence of magnification error for lateral BMC may be especially important in longitudinal studies in children and adolescents where growth-related increases in body ‘‘thickness’’ may mask significant gains in BMC and bone size. Future studies comparing these DXA techniques to 3-dimensional imaging techniques are necessary. The pattern of increases in vertebral spine volumetric BMD with Tanner stage observed here is consistent with quantitative computed tomographic (QCT) studies of vertebral volumetric BMD in children and adolescents (2,26,27). However, a recent study comparing QCT volumetric BMD with areal BMD and BMAD based on Hologic PA DXA scans demonstrated that BMAD only slightly improved the density correlations (i.e., the correlation [r] between PA areal BMD and QCT volumetric BMD was 0.62, and the correlation between BMAD and QCT volumetric BMD was 0.70). Subsequent studies are needed to determine if width-adjusted BMD provides more accurate estimates of volumetric BMD. Most importantly, studies are needed to determine the test characteristics (i.e., sensitivity and specificity) of the varied BMD imaging strategies in the identification of patients with fractures. Volume 9, 2006

PA and Lateral Spine Scans in Children

Acknowledgment This protocol was supported by the National Institutes of Health grants (no. K08-DK02523 [MBL] and 1-R03DK058200 [MBL]), and the General Clinical Research Center (grant no. M01RR00240), The Children’s Hospital of Philadelphia, University of Pennsylvania School of Medicine.

References 1. Seeman E. 1997 From density to structure: growing up and growing old on the surfaces of bone. J Bone Miner Res 12: 509e521. 2. Gilsanz V, Kovanlikaya A, Costin G, Roe TF, Sayre J, Kaufman F. 1997 Differential effect of gender on the sizes of the bones in the axial and appendicular skeletons. J Clin Endocrinol Metab 82:1603e1607. 3. Gilsanz V, Skaggs DL, Kovanlikaya A, et al. 1998 Differential effect of race on the axial and appendicular skeletons of children. J Clin Endocrinol Metab 83:1420e1427. 4. Prentice A, Parsons TJ, Cole TJ. 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 60:837e842. 5. Carter DR, Bouxsein ML, Marcus R. 1992 New approaches for interpreting projected bone densitometry data. J Bone Miner Res 7:137e145. 6. Kroger H, Vainio P, Nieminen J, Kotaniemi A. 1995 Comparison of different models for interpreting bone mineral density measurements using DXA and MRI technology. Bone 17:157e159. 7. Katzman DK, Bachrach LK, Carter DR, Marcus R. 1991 Clinical and anthropometric correlates of bone mineral acquisition in healthy adolescent girls. J Clin Endocrinol Metab 73:1332e1339. 8. Lu PW, Cowell CT, SA LL-J, Briody JN, Howman-Giles R. 1996 Volumetric bone mineral density in normal subjects, aged 5-27 years. J Clin Endocrinol Metab 81:1586e1590. 9. Griffiths MR, Noakes KA, Pocock NA. 1997 Correcting the magnification error of fan beam densitometers. J Bone Miner Res 12:119e123. 10. Pocock NA, Noakes KA, Majerovic Y, Griffiths MR. 1997 Magnification error of femoral geometry using fan beam densitometers. Calcif Tissue Int 60:8e10. 11. Cole JH, Scerpella TA, van der Meulen MC. 2005 Fan-beam densitometry of the growing skeleton: are we measuring what we think we are? J Clin Densitom 8:57e64. 12. Jergas M, Breitenseher M, Gluer CC, Yu W, Genant HK. 1995 Estimates of volumetric bone density from projectional measurements improve the discriminatory capability of dual X-ray absorptiometry. J Bone Miner Res 10:1101e1110. 13. Henry YM, Fatayerji D, Eastell R. 2004 Attainment of peak bone mass at the lumbar spine, femoral neck and radius in men and women: relative contributions of bone size and volumetric bone mineral density. Osteoporos Int 15:263e273. 14. Wu XP, Yang YH, Zhang H, et al. 2005 Gender differences in bone density at different skeletal sites of acquisition with age in Chinese children and adolescents. J Bone Miner Metab 23:253e260. 15. Antoniazzi F, Zamboni G, Bertoldo F, et al. 2003 Bone mass at final height in precocious puberty after gonadotropin-releasing hormone agonist with and without calcium supplementation. J Clin Endocrinol Metab 88:1096e1101.

Journal of Clinical Densitometry

273 16. Tanner JM. 1962 Growth at Adolescence. 2nd ed. Oxford: Blackwell Scientific Publication. 17. Ogden CL, Flegal KM, Carroll MD, Johnson CL. 2002 Prevalence and trends in overweight among US children and adolescents, 1999e2000. JAMA 288:1728e1732. 18. Krebs NF, Jacobson MS. 2003 Prevention of pediatric overweight and obesity. Pediatrics 112:424e430. 19. Shepherd JA, Fan B, Sherman M, et al. 2004 Pediatric DXA precision varies with age (Abstract). J Bone Miner Res 19(Suppl 1): S234. 20. Thomas SR, Kalkwarf HJ, Buckley DD, Heubi JE. 2005 Effective dose of dual-energy X-ray absorptiometry scans in children as a function of age. J Clin Densitom 8:415e422. 21. Lewis MK, Blake GM, Fogelman I. 1994 Patient dose in dual x-ray absorptiometry. Osteoporos Int 4:11e15. 22. Blake GM, Warner HW, Fogelman I. 1999 The evaluation of osteoporosis: dual energy X-ray absorptiometry and ultra sound in clinical practice. 2nd ed. London: Blackwell Science. 23. Blake GM, Parker JC, Buxton FM, Fogelman I. 1993 Dual X-ray absorptiometry: a comparison between fan beam and pencil beam scans. Br J Radiol 66:902e906. 24. Bland JM, Altman DG. 1986 Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1:307e310. 25. Diggle PJ, Liang KY, Zeger SL. 1986 Longitudinal data analysis using generalized linear models. Oxford: Oxford Science Publication. 26. Gilsanz V, Roe TF, Mora S, Costin G, Goodman WG. 1991 Changes in vertebral bone density in black girls and white girls during childhood and puberty. N Engl J Med 325: 1597e1600. 27. Gilsanz V, Gibbens DT, Roe TF, et al. 1988 Vertebral bone density in children: effect of puberty. Radiology 166:847e850. 28. Zmuda JM, Cauley JA, Glynn NW, Finkelstein JS. 2000 Posterior-anterior and lateral dual-energy x-ray absorptiometry for the assessment of vertebral osteoporosis and bone loss among older men. J Bone Miner Res 15:1417e1424. 29. Finkelstein JS, Cleary RL, Butler JP, et al. 1994 A comparison of lateral versus anterior-posterior spine dual energy x-ray absorptiometry for the diagnosis of osteopenia. J Clin Endocrinol Metab 78:724e730. 30. Grampp S, Genant HK, Mathur A, et al. 1997 Comparisons of noninvasive bone mineral measurements in assessing age-related loss, fracture discrimination, and diagnostic classification. J Bone Miner Res 12:697e711. 31. Laskey MA, Murgatroyd PR, Prentice A. 2004 Comparison of narrow-angle fan-beam and pencil-beam densitometers: in vivo and phantom study of the effect of bone density, scan mode, and tissue depth on spine measurements. J Clin Densitom 7: 341e348. 32. Mazess RB, Hanson JA, Payne R, Nord R, Wilson M. 2000 Axial and total-body bone densitometry using a narrow-angle fan-beam. Osteoporos Int 11:158e166. 33. Leonard MB, Shults J, Wilson BA, Tershakovec AM, Zemel BS. 2004 Obesity during childhood and adolescence augments bone mass and bone dimensions. Am J Clin Nutr 80:514e523. 34. Petit MA, Beck TJ, Shults J, Zemel BS, Foster BJ, Leonard MB. 2005 Proximal femur bone geometry is appropriately adapted to lean mass in overweight children and adolescents. Bone 36: 568e576.

Volume 9, 2006