Influence of Prevalent Vertebral Fracture on the Correlation between Change in Lumbar Spine Bone Mineral Density and Risk of New Vertebral Fracture: A Meta-Analysis of Randomized Clinical Trials

Background and Objective The correlation between change in bone mineral density (BMD) and the incidence of new vertebral fracture has been drawing attention in regard to evaluation of fracture risk and drug efficacy. We investigated the impact of the prevalence of vertebral fracture on this correlation via a meta-regression analysis with a view to improving evaluation of the correlation. Methods A total of 19 postmenopausal osteoporosis clinical studies involving 62,432 patients in 46 placebo or treatment groups were identified through MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials. We performed a multivariate meta-regression analysis to examine the correlation between the percentage change in lumbar spine BMD from baseline at 3 years and the proportion of subjects experiencing new vertebral fractures, with or without the proportion of subjects with prevalent vertebral fracture as a covariate. We also analyzed the interaction between the subgroups divided by the proportion of subjects with prevalent vertebral fracture and the percentage change in lumbar spine BMD from baseline at 3 years. Results A multivariate meta-regression analysis showed a significant correlation between the change in lumbar spine BMD and the proportion of subjects experiencing new vertebral fracture, and a lower Akaike information criterion was obtained when the proportion of subjects with prevalent vertebral fracture was added as an explanatory variable. Significant interaction between the proportion of subjects with prevalent vertebral fracture and the change in lumbar spine BMD was shown. Conclusions The change in lumbar spine BMD, not BMD T-score at one timepoint, had a significant correlation with the incidence of vertebral fracture. The prediction of the fracture risk by change in lumbar spine BMD was improved by adjusting the proportion of subjects with prevalent vertebral fracture in the study population. The difference of prevalence of vertebral fracture among populations should be considered when the association between change in lumbar spine BMD and incidence of vertebral fracture is examined.