A critical appraisal of vertebral fracture assessment in paediatrics.

Purpose: There is a need to improve our understanding of the clinical utility of vertebral fracture assessment (VFA) in paediatrics and this requires a thorough evaluation of its readability, reproducibility, and accuracy for identifying VF.

Methods: VFA was performed independently by two observers, in 165 children and adolescents with a median age of 13.4 years (range, 3.6, 18). In 20 of these subjects, VFA was compared to lateral vertebral morphometry assessment on lateral spine X-ray (LVM).

Results: 1528 (84%) of the vertebrae were adequately visualised by both observers for VFA. Interobserver agreement in vertebral readability was 94% (kappa, 0.73 [95% CI, 0.68, 0.73]). 93% of the non-readable vertebrae were located between T6 and T9. Interobserver agreement per-vertebra for the presence of VF was 99% (kappa, 0.85 [95% CI, 0.79, 0.91]). Interobserver agreement per-subject was 91% (kappa, 0.78 [95% CI, 0.66, 0.87]). Per-vertebra agreement between LVM and VFA was 95% (kappa 0.79 [95% CI, 0.62, 0.92]) and per-subject agreement was 95% (kappa, 0.88 [95% CI, 0.58, 1.0]). Accepting LVM as the gold standard, VFA had a positive predictive value (PPV) of 90% and a negative predictive value (NPV) of 95% in per-vertebra analysis and a PPV of 100% and NPV of 93% in per-subject analysis.

Conclusion: VFA reaches an excellent level of agreement between observers and a high level of accuracy in identifying VF in a paediatric population. The readability of vertebrae at the mid thoracic region is suboptimal and interpretation at this level should be exercised with caution.