Short-interval reproducibility of quantitative vertebral exclusion criteria for lumbar spine DXA

Lumbar spine DXA is susceptible to focal structural artifacts that usually elevate measured bone mineral density and can obscure osteoporosis. Quantitative vertebral exclusion criteria may help identify scans requiring review, but their short-interval reproducibility is uncertain. We analyzed 886 paired lumbar spine DXA assessments from the Manitoba Bone Density Program performed within 30 days. Quantitative criteria included maximum absolute adjacent-vertebral T-score difference, maximum absolute individual vertebral difference from L1-L4, and sequential bone mineral content/area criteria. Binary flags were generated using thresholds of >0.5, >1.0, >1.5, and >2.0 T-score units. Reproducibility was assessed using intraclass correlation coefficients for continuous measures, weighted kappa statistics for binary criteria, and ROC analyses using first-assessment continuous measures to discriminate second-assessment binary classifications. Mean age was 58.5 ± 12.8 years, mean interval between assessments was 6.7 ± 7.3 days, and 96.6% were women. Continuous T-score discordance measures showed generally good reproducibility, with ICCs ranging from 0.756 to 0.892. Agreement for binary threshold-based criteria ranged from fair to good, with weighted kappa values ranging from 0.379 to 0.682. The highest categorical agreement was observed for maximum adjacent-vertebra discordance at the >2.0 threshold (κ=0.682, 95% CI 0.556-0.807), while individual vertebra versus L1-L4 discordance also showed relatively strong agreement at the >1.0 threshold (κ=0.662, 95% CI 0.603-0.721). ROC analyses showed strong discrimination, with AUCs ranging from 0.835 to 0.952 for maximum adjacent-vertebra discordance and from 0.850 to 0.988 for maximum individual vertebra versus L1-L4 discordance. Quantitative lumbar spine DXA vertebral exclusion criteria were reproducible over short intervals, but reproducibility varied by criterion and threshold. These findings support their use as standardized prompts for review rather than obligatory vertebral exclusion rules.