As with other laboratory tests, the numeric value reported is meaningless without comparison to the appropriate normal controls. After comparison is made, the reported value is given as a percentile or a standard deviation score, the Z-score. A Z-score of zero is equivalent to the mean, and Z-scores of −1 and +1.5 are equivalent to values one standard deviation below and 1.5 standard deviations above the mean, respectively. The T-score (comparison of the current Z-score with peak adult BMD) is used in adult interpretation of DXA but should not be included in the pediatric DXA report [3
]. Because the T-score is a measure of bone density loss
since early adulthood, its use in children whose BMD has yet to peak will always yield a low result. Because the World Health Organization’s DXA-based definitions of osteopenia and osteoporosis are in terms of T-scores, T<−1.0 and T<−2.5, respectively [3
], a different terminology is needed for pediatric patients. It is recommended that the phrase “low bone density” be used in DXA reports [3
]. Some clinicians and researchers use the terms osteopenia and osteoporosis in children when Z-scores are less than −1.0 and −2.5, respectively. It is important to note that the diagnosis of osteoporosis should not be made on DXA results alone but should take into account other patient factors.
Much of the research in pediatric DXA has focused on determining which factors most influence BMD and should be accounted for in the development of normative pediatric datasets. The factors age, gender, ethnicity, and physiologic maturity level have been extensively studied and are included in most current normative datasets provided by the major DXA manufacturers. Some of the earliest reports indicate the influence of age on BMD values, and several authors have presented normal data accounting for age [10
]. The effect that age has on BMD is largely related to the increase in frame size that occurs with increasing chronologic age. Increasing height and weight strongly correlate with increasing BMC and BMD. The changes in height and weight are most pronounced during the pubertal growth spurt. BMD increases rapidly during early puberty [21
], but because the age of pubertal onset is quite variable, physiologic maturity has a stronger influence on BMD than age. Several authors have included Tanner stage or gynecologic age as a primary factor in their normative datasets [21
]. Weight is an important factor influencing BMD for multiple reasons [4
] and has been included in normal datasets [10
]. Ethnicity has been determined to be important in the analysis of BMD results, with black children showing significantly higher BMD values than non-black children [10
]. Multiple factors are thought to account for this, including increased cross-sectional area in the axial skeleton and thicker trabecula in cancellous bone [36
]. Lifestyle and anthropometric factors might also play a role [42
There are numerous published pediatric normative datasets, many of which are summarized in Table . These datasets have been developed using a variety of scanners and processing software and are based on various combinations of demographic and physiologic patient variables. Rather than simplifying pediatric DXA interpretation, the sheer number of available normal databases has made DXA interpretation complex, confusing, and at times erroneous [43
]. To report the numeric result generated from the manufacturer’s automated processing without consideration of factors specific to the patient being studied is unacceptable. This often will lead to misdiagnoses and can result in inappropriate therapy [44
]. In fact, the diagnosis of osteoporosis in a child based on a DXA result is often a misinterpretation of the scan data. Gafni and Baron [45
] found this to be the case in more than half of the pediatric patients referred to them with the diagnosis of osteoporosis. The most common causes for misdiagnosis were the use of T-scores, inappropriate normative datasets, inadequate ROIs, and inattention to short stature.
Normative pediatric DXA databases (C/B/H/A/O Caucasian/black/Hispanic/Asian/other, GA gestational age, (L) longitudinal study, SA surface area)
As with any other radiologic study, a methodical evaluation of the results should be undertaken in order to minimize the risk of misdiagnosis. The radiologist needs to review all input data, including patient age, gender, ethnicity, weight, height, and Tanner stage (if provided). Patient positioning should be evaluated, and the ROIs need to be analyzed for artifact and appropriateness. Comparison should be made with previous studies to ensure consistency of positioning and ROI selection. In addition, changes in patient height, weight, and Tanner stage should be noted. After these steps have been taken, interpretation of the numeric result is performed. An appropriate database for comparison purposes is selected. Ideally, this is based on data generated locally using the same equipment and technologists, but this is rarely possible. Normative data provided by the DXA manufacturers can be used, but historically these datasets do not include the parameters currently thought to be most important for interpretation. At a minimum, patient body size (height and weight) and physiologic maturity (Tanner stage, gynecologic or bone age) should be factors included in the normative dataset. Ethnicity and gender are also frequently included in the generation of normative data and are generally thought to affect BMD significantly. Table summarizes a large number of normal databases that can be used to best match the patient scan to be analyzed.
More complex and scientifically rigorous analyses of the BMD result have been suggested. Molgaard et al. [46
] described a three-step analysis of BMD. Bone length is categorized as short or long by assessing the patient’s height for age. Height is highly correlated with BMC [12
] and thus needs to be accounted for, especially when BMD Z-score is abnormal (Z=±2). Bone width is categorized as thin or thick by assessing BMC for height. Last, bone mineralization is categorized with assessment of BMC for BA. The first step takes into account height because of its profound effect on BMC; taller children will have a higher bone content. However, bone width is also important in determining bone content, thus the second step accounts for this. The last step is often the only step performed by many radiologists with reporting of BMC for BA which, by definition, is BMD. Height for age can be assessed using standard growth charts corrected for age, gender, and ethnicity. Percentile rankings can be easily converted to a Z-score [47
]. Bone area for height tables are available for select groups, but normative data for all pediatric patients need to be developed. Using this three-step analysis, Molgaard and Michaelsen [48
] found that the causes for low BMC might be various combinations of factors such as short stature and thin bones, as in children with cystic fibrosis, or short stature and reduced mineralization, as in children with milk allergy. BMC might be normal despite short stature in the presence of wide bones, as in children with previously treated leukemia.
In summary, an abnormal BMD Z-score should lead to evaluation of confounding patient factors that influence BMD, including height, weight, and physiologic maturity, before a diagnosis of low bone density is made. Until the manufacturers’ databases sufficiently account for the physiologic factors that most impact BMD results, normative data derived locally or from the medical literature should be used in pediatric DXA interpretation. The report should include the DXA equipment and software algorithm used (pediatric or adult, low bone density or standard), the source of the normative reference data, the Z-score (not the T-score), and an impression giving a clinical context for the result. The diagnosis of “low bone density” does not rest solely on the DXA numeric result, and the report should indicate which patient factors were incorporated into the final impression. A specimen DXA report and the examination protocol from Columbus Children’s Hospital are given in Appendix