Analysis and Evaluation of DXA in Children and Adolescents

Fig. 6.1

Lumbar spine scans in infants showing (a) anterior-posterior “breathing” movement, (b) lateral movement, and (c) no movement

Contemporary software versions have edge detection algorithms that function well for young children with much lower bone density than adults. The global ROI may be adjusted to the left or right so that it can be centered around the vertebral column; however, the box width is not adjustable. The top line for ROI must be positioned within the intervertebral space between T12 and L1. The bottom line of the ROI must be positioned within the intervertebral space between L4 and L5. These lines may be angled slightly if necessary to account for alterations in spine anatomy such as scoliosis.

Once the global ROI is adjusted, the bone map is either confirmed or edited. Editing should only be done when the bone map is grossly erroneous. The placement of the three horizontal intervertebral lines between L1 and L2, L2 and L3, and L3 and L4 should be confirmed or adjusted as necessary. Interfering factors within the global ROI such as belly button rings can invalidate the results of the spine scan; unlike adults, there are no reference data at this time to determine the Z-scores for individual vertebrae or combinations of vertebrae. Under these circumstances, alternative scans are recommended. Like adults, a deviation in the increasing progression of bone area and bone mineral density from L1 to L4 suggests a possible vertebral deformity or compression fracture, requiring further investigation. There are n o major differences in the final assessment of spine scans for children as compared to adults.

Whole Body Scan Analysis

Specifics of whole body scan acquisition are described in Chap. 5. Prior to analyzing, scan images should be carefully inspected for motion and other artifacts such as snaps, buttons, pocketed cell phones, etc. The quality of the scan is important as poor patient positioning or movement will affect the accuracy of the results. Children should be scanned with their head at the end of the table to achieve uniform positioning for follow-up scans as the child grows. The scan length may be shortened to decrease the scan time. However, care should be taken to make the scan length sufficiently long that the entire child will fit within the scan window. It is helpful to mark the table top at 130, 140, and 150 cm so the operator can determine the appropriate scan length when the child is lying on the table. Figure 6.2 shows whole body scan images with and without motion or other artifacts.

Fig. 6.2

Whole body scans in children illustrating (a) a good scan with no movement, (b) a scan with movement in the head and legs, and (c) a scan of a child with a medical artifact (PICC line)

Manufacturer’s guidelines should be followed for analysis of whole body scans, including defining standard, manufacturer-specified subregions. The technologist must adjust the lines defining the subregions for appropriate scan analysis. There are ten subregions: head, left and right arms, left and right ribs, thoracic and lumbar spine, pelvis, and right and left legs. The bone map is not visible on whole body scans. Therefore, the technologist should examine the BMC and aBMD results within each subregion to confirm that bone is detected. This is particularly important for very young children because of their smaller bone size and lower bone density. Presently, Hologic does not recommend the use of standard whole body scans in children less than 3 years of age (Thomas Kelly, personal communication) because the bone map is likely to be incomplete. The infant whole body scan mode is an alternative, but currently is a tool for research applications only. GE Lunar devices presently do not provide technical support for infant scans.

Of note, Hologic software performs an automatic weight-based adjustment to the edge detection algorithm of whole body scans for childre n under 40 kg (i.e., 88 lb).

Forearm Analysis

Young children may be too small to sit in a chair and achieve the optimal position for a forearm scan. It is often more feasible to position the child on the table. They can be placed in a supine position with their elbow bent at 90 degrees and forearm parallel to the edge of the table, or they can be placed prone on the table with their arm above their head with the forearm parallel to the edge of the table as shown in Fig. 6.3. Regardless of the postural method used, the position of the forearm in the scan field should meet the manufacturer recommendation and the analysis of the forearm scan is the same.

Fig. 6.3

Forearm scan positions in young children showing (a) the supine position, and (b) the prone “superman” position. The superman position should be used only when the supine position is not possible. The arm can be held in position during the scan provided the “helper’s” hands are not in the scan field

In growing children, the growth plate and epiphysis of the radius and ulna will be visible in the image and the ulnar styloid might not be present. In very young children, the ulnar epiphysis might not be visible. Generally, the manufacturer’s guidelines for scan analysis should be followed, but the ROI should be placed so that the ultradistal region excludes the dense bone tissu e at the proximal edge of the growth plate as shown in Fig. 6.4.

Fig. 6.4

Placement of regions of interest in forearm scans of children. The distal border of the ultradistal region of interest should be proximal to the growth plate

Proximal Femur Analysis

Acquisition of proximal femur scans is described in Chap. 5. The scans should be inspected for proper positioning. The femur should be rotated 10°–15° and the lesser trochanter is just barely visible. If the lesser trochanter is too prominent, then the leg is not internally rotated sufficiently. In young children, it may be difficult to visualize the lesser trochanter, but in older children, especially adolescent boys, the lesser trochanter is quite prominent. The shaft of the femur should be straight and parallel to the edge of the scan field. The lower edge of the ROI should encompass the lesser trochanter. As with all scans, inspection for motion or other artifacts is important, and the bone map should be inspected to be sure that it is correct.

Once the bone map is complete, the next step is to let the software determine the midline, Ward’s triangle,¹ the base of the greater trochanter, and the neck box placement. The midline should be perpendicular to the narrowest part of the neck region. A misplaced midline can be caused by very low density bone or by unusual proportions in a growing child.

In Hologic systems, increasing the border of the global ROI to include additional soft tissue near the head of the femur may allow the software to find the midline more accurately. If the midline is still incorrectly positioned, it should be manually adjusted. However, manual adjustments should be made with caution because it can be difficult to reproduce the analysis; it is best to use auto-analysis for the midline when possible.

The next step is to inspect the position of the neck box. The default width should not be changed unless the top of the box is in the head of the femur or the bottom is overlapping the ischium . If either of these is the case, the box should be moved or, possibly, narrowed the smallest amount possible. Any changes in the size or location of the neck box should be noted on the report. Additionally, if the neck box width is changed, BMC Z-scores and bone mineral apparent density (BMAD) results cannot be used because the calculations assume this region is at the default box size.

The femoral neck ROI should be positioned so that the lower outer corner is just touching the bone in the neck region at the point where the curve of the greater trochanter meets the curve of the neck region. The other three corners must be in soft tissue. If the patient is very small or has a very short hip axis length, it may not be possible to place the neck box without overlapping the ischium region. In this case, go back to the mapping step and manually delete bone away from the neck region. Once this region is deleted, place the neck box appropriately. Figure 6.5 illustrates good quality proximal femur scans for an older and younger child.

Fig. 6.5

Placement of regions of interest in proximal femur scans in children

GE Lunar models have a slightly different ROI, which uses the midline placement and triangulation to place the lower edge of the global ROI. The neck region for GE Lunar models is also perpendicular to the midline but is located in the center of the whole neck region.

Lateral Distal Femur

The lateral distal femur scan is frequently used to assess bone health in children with chronic diseases who are immobilized as described in Chap. 9. Details of scan procedures are also published [2, 3]. The scan should be acquired with the femur parallel to the long edge of the table using the forearm scan mode. Subregional forearm analysis is used to analyze these scans rather than the auto-analysis function, which requires bone length. The scan is analyzed by placing contiguous ROI boxes around three regions: (1) the anterior distal region of the femur placed proximally to the growth plate, a region that is primarily trabecular bone; (2) the metadiaphysis, which includes both trabecular and cortical bone; and (3) the diaphysis, which is entirely cortical bone (Fig. 6.6). The height of the ROI is determined by the width of the diaphysis, and calculated as twice the diaphyseal width times the aspect ratio. For the first region, only the anterior portion of the bone is included in the ROI. The posterior edge of the ROI is positioned at the midpoint of the bone width at the growth plate. The anterior portion of the first ROI should extend beyond the soft tissue. For regions 2 and 3, the width of the box should be large enough to extend over soft tissue and centered across the bone.

Fig. 6.6

Analysis of lateral distal femur scans in children

The scan should be inspected for movement artifact. With modest movement in one region, it is possible to use the other regions for interpretation. However, with excessive movement, it may be impossible to properly size and align the ROI boxes beca use of the interdependency of their positions.

Manufacturer, Model, and Software Differences

Bone mineral density (BMD) results for the same person measured on different instruments may differ, especially if the instruments are from different manufacturers [4]. These differences may be due machine calibration differences, or unique manufacturer software and acquisition methods. Due to these discrepancies, it is recommended that, whenever possible, the same instrument, model, and software version be used to assess an individual patient over time.

It is not always feasible to conduct serial measurements on identical instruments, as would be the case for a patient who receives care at multiple clinics or for clinics that have upgraded their DXA technology over time. In an attempt to allow comparisons among manufacturers, the three most common instrument manufacturers (Hologic, GE Lunar, and Norland) established cross-calibration factors [5], which permitted calculation of the standardized BMD (sBMD). The sBMD is expressed in mg/cm² to avoid confusion with manufacturer-specific areal BMDs, which are expressed in g/cm² [5–7]. The equations to convert measurements of spine and hip aBMD were established using scans obtained on adult women and phantoms [5], with further adjustments published for the spine [8] and femur [9]. A comparison of sBMD values for 30 postmenopausal women measured on Hologic Delphi and GE-Lunar Prodigy devices showed differences of 0.5 % for the right total hip and 4.1 % for lumbar spine L1 to L4 [10]. The equations for the spine do not apply well to men [11]. Most conversion equations are based on adults, were developed on older generation DXA scanners, and have not been validated for children. Only one study has established a standardization formula based on a sample that included children, but their results are restricted to whole body scans [12].

If equipment is upgraded or replaced, the differences between devices must be considered. If only software analysis versions are upgraded, it is possible to reanalyze prior scans using the new software so that follow-up values will be less divergent. If the instrument is replaced, it is recommended that cross-calibration be performed by scanning a set of 30 individuals with a range in age, size, and aBMD that reflects the clinical population on both the old and new instruments. Guidelines for performing the cross-calibration can be found at http://www.iscd.org/resources/calculators/. If this is not possible, the change in instrument manufacturer, model, or software should be noted on the clinical report.

Follow-Up Scan Analysis

Analysis of longitudinal changes in growing children can prove to be difficult. Although no one approach has been shown to be best, some practices can minimize error. Prior scan images should be carefully examined during positioning and scan acquisition to optimize comparability of repeated measurements. Good practice should be used in positioning and analysis at all time points. If there is deviation from good practice in the prior or current scan, this should be noted on the report. Documentation of manufacturer, model, and software should be noted on all clinical reports, and specific comments should be included on serial reports when there are changes in the machine model or software used.

Manufacturers recommend that all follow-up scans be analyzed using the compare mode. In rapidly growing children, it is often necessary to make adjustments to the global ROI and subregions, especially if follow-up scans occur over a large period of time. The compare mode can and should be used to ensure similar positioning of the ROI between time points. Regardless of whether or not the compare function is used, it is important that technologists in the same clinic follow a consistent protocol for all children and that researchers report details of scan analyses in manuscripts.

Fundamentals of Evaluation

Age-related increases in BMC and aBMD are complex. BMC and aBMD undergo nonlinear increases relative to age, and the variability in these measures also increases as shown in Fig. 6.7. Like growth in height and weight, DXA results need to be evaluated relative to reference ranges (similar to growth charts) that account for the expected age- and sex-specific changes in growing children. The selection of reference ranges is central to the evaluation process because DXA results for children are expressed as Z-scores, the number of standard deviations above or below the median for age and sex. In older adults, the T-score is used. The T-score compares a DXA result to the mean and standard deviation for young adults when bone mass is at its peak. A T-score should never be used in children as they have not yet reached peak bone mass. For children, Z-scores are used as an indicator of “bone mineral status,” i.e., how an individual’s measurement compares to those of peers with the same age and sex. Use of inappropriate reference ranges can yield Z-scores that misrepresent “bone mineral status” (e.g., low bone density for age). The points below highlight the important considerations in selecting reference data.

Fig. 6.7

Age related increases in BMC and aBMD during growth

Children whose BMC or BMD is close to the median for their age and sex will have a Z-score of approximately zero. A Z-score of 2 corresponds to the 97.7th percentile, and a Z-score of −2 corresponds to the 2.3rd percentile. An advantage of Z-scores over percentile values is that very low or very high values that are outside the reference population distribution (i.e., greater than the 100th percentile or less than the 0th percentile) can be quantified. This is especially important for longitudinal follow-up of children with low BMD as it allows for quantifying the changes in BMD relative to the expected values for age and sex.

Normative Data

Ideal Characteristics of Reference Data

The first step in the interpretation of DXA measurements in pediatric patients is appropriate selection of reference data. Reference data should have several characteristics. The data should be derived from a sample of healthy children who are representative of the overall population. Healthy children can be defined as those who are free of chronic diseases, medication use, and physical limitations that might affect bone mineral accrual. They should also be of normal nutritional, growth, and developmental status because these are known to affect bone health. Because of the possibility of regional differences in lifestyle, ethnic composition, sunlight exposure, and so forth, a multiregional sample is optimal to capture normal variability. The sample needs to be of sufficient size to adequately characterize the variability in bone measures for both boys and girls. Because the variability in bone measures increases with age, it is important to assure sufficient sample sizes at all ages so that the age-dependent differences in variability can be well characterized.

Calculation of the Z-Score

The purpose of reference data is to gather the necessary information for calculating Z-scores (standard deviation scores). Expression of DXA outcomes in children as Z-scores requires knowing age- and sex-specific characteristics of the BMC or aBMD distributions so that a child’s measurement is compared to those of the same age and sex in the reference population.

The simplest approach to calculate Z-scores involves use of age and sex-specific means and standard deviations from reference data as shown in Eq. (1).

$Z-\mathrm{score}=\left(\mathrm{observed}-\mathrm{mean}\right)/\mathrm{standard}\ \mathrm{deviation}$

(1)

This approach assumes that the age- and sex-specific values are normally distributed, and that the sample sizes for each age and sex group are large enough to characterize the distribution. It also assumes that annual age increments are an appropriate grouping; during periods of rapid bone accrual, smaller age groupings may be needed. A limitation of this approach is that random sampling fluctuations can result in erratic differences between adjacent age groups.

The distribution of bone measures is sometimes skewed, so more sophisticated biostatistical techniques have been used, including parametric regression modeling [13, 14], multivariate semi-metric smoothing [15] and the LMS method [16]. These techniques provide “smoothed” values over the age range to avoid random sampling fluctuations. Bone outcomes become increasingly variable with advancing age and are often skewed owing to variation in timing of puberty (i.e., early and later maturing children), and the increasing influence of environmental, genetic, and behavioral factors on bone outcomes. The LMS method of characterizing reference values addresses these concerns and provides a method for smoothing across age ranges, accounting for changes in variation with age, and skewness in the distribution [16]. The LMS method uses a power transformation to normalize data. The optimal power to obtain normality is calculated for a series of age groups, and the trend is summarized by a smooth (L) curve. Smoothed curves for the median (M) and coefficient of variation (S) are also calculated, and these three measures, L, M, and S, are used to describe the data distribution. When the LMS method is applied, the distribution and exact percentiles can be described using Eq. (2) as follows:

$\mathrm{percentile}\ \mathrm{curve}=M\times \left[{\left(1+\left(L\times S\times Z\right)\right)}^{\left(1/L\right)}\right]$

(2)

using the L (skewness), M (median), and S (coefficient of variation) values, and Z corresponding to the Z-score equivalent of standard percentiles (e.g., the 50th percentile is a Z-score of zero; the 2.3rd percentile is equivalent to a Z-score of −2.0).

For an individual, the exa ct Z-score can be calculated as:

$Z-\mathrm{score}=\left(\left[\left(\mathrm{Observed}\ \mathrm{value}/M\right)L\right]-1\right)/\left(L\times S\right)$

(3)

Selection of Reference Data

Most DXA software includes pediatric reference data and provides the calculated Z-score in the DXA report, but good documentation regarding the source of the reference data is sometimes lacking. Clinicians and researchers should be aware of the reference data used to generate Z-scores and the impact that this may have on interpretation of results. Options for selecting pediatric reference data include: (1) using the database incorporated into the manufacturer’s software; (2) comparing your data to published data; or (3) using locally collected normative values. Each approach has its benefits and limitations . The database incorporated into the manufacturer’s software may derive from a single published study, a blending of several published studies, or other adjustments that are not explicitly described. For example, the U.S. reference data from the Bone Mineral Density in Childhood Study (see below) were acquired on Hologic DXA devices. These data have been blended into Hologic pediatric databases. Likewise, GE Lunar incorporated these reference data into the enCore 2011 software, presumably with adjustments because the reference data were collected using Hologic DXA devices. As with all decisions related to acquiring, analyzing, and interpreting DXA scans, the most important point is to be aware of the sources used to calculate Z-scores , the limitations involved and providing documentation to ease interpretation of changes at follow-up.

Only gold members can continue reading. Log In or Register to continue