Background: There is significant inter and intraobserver variability in diagnosing vertebral fractures in children.
Purpose: We aimed to evaluate the diagnostic accuracy of morphometric vertebral fracture analysis (MXA) using a 33-point software program designed for adults, on dual-energy x-ray absorptiometry (DXA) images of children.
Materials and methods: Lateral spine DXA images of 420 children aged between 5 and 18 years were retrospectively reviewed. Vertebral fracture assessment (VFA) by an expert pediatric radiologist using Genant's semiquantitative scoring system served as the gold standard. All 420 DXA scans were analyzed by a trained radiographer, using semi-automated software (33-point morphometry). VFA of a random sample of 100 DXA was performed by an experienced pediatric clinical scientist. MXA of a random sample of 30 DXA images were analyzed by three pediatric radiologists and the pediatric clinical scientist. Diagnostic accuracy and inter and intraobserver agreement (kappa statistics) were calculated.
Results: Overall sensitivity, specificity, false positive (FP) and false negative (FN) rates for the radiographer using the MXA software were 80%, 90%, 10%, and 20% respectively and for mild fractures alone were 46%, 92%, 8%, and 54% respectively. Overall sensitivity, specificity, FP, and FN rates for the four additional observers using MXA were 89%, 79%, 21%, and 11% respectively and for mild fractures alone were 36%, 86%, 14%, and 64% respectively. Agreement between two expert observers was fair to good for VFA and MXA [kappa = 0·29 to 0·76 (95% CI: 0·17–0·88) and 0·29 to 0·69 (95% CI: 0·17–0·83)] respectively.
Conclusion: MXA using a 33-point technique developed for adults is not a reliable method for the identification of mild vertebral fractures in children. A pediatric standard is required which not only incorporates specific vertebral body height ratios but also the age-related physiological changes in vertebral shape that occur throughout childhood.