Z-Score Comparability of Bone Mineral Density Reference Databases for Children

Purpose: The diversity of pediatric dual-energy x-ray absorptiometry (DXA) bone mineral density (BMD) reference databases raises questions as to whether they are interchangeable in their application. This study examined the comparability of BMD Z-scores generated from the largest available Hologic DXA databases, applied on BMD results of a large series of unselected pediatric patients. Methods: A total of 2027 BMD scans were extracted from Hologic QDR-4500A machines. Age- and sex-specific BMD Z-scores of children aged 8–17 yr, calculated from six Hologic databases, were compared for lumbar spine (LS) and total body (TB). The final dataset included 708 scans (307 of girls). Results: BMD Z-scores calculated from the six databases were highly correlated but differed significantly (P < 0.001) in both scan regions. Interdatabase Z-score differences (boys/girls, respectively) were up to 0.54/0.55 for LS and 1.0/0.83 for TB. These differences also varied significantly among age groups. In girls, the percentage of LS BMD Z-scores of −2 or below (“low BMD for age”) varied between 15.4 and 27.9% (P < 0.012). The percentage of TB BMD Z-scores of −2 or below varied similarly in boys (P < 0.009). Conclusions: Clinically relevant differences in BMD Z-scores exist between the Hologic databases, revealing a significant potential for misdiagnosis. Ideally, Z-scores should be calculated using model-, brand-, and software-specific reference curves for age, sex, and ethnic group. However, our results can be used to estimate converted values. There are other differences in children’s bone mass, shape, strength, and body size that are not detected by DXA. Direct comparison of the largest bone density reference databases for children revealed clinically relevant differences in calculated bone mineral density Z-scores, with significant potential for misdiagnosis.