The fracture risk index and bone mineral density as predictors of vertebral structural failure.

Structural failure becomes increasingly likely as the load on bone approximates or exceeds the bone's ability to withstand it. The vertebral fracture risk index (FRI) expresses the risk for structural failure as a ratio of compressive stress (load per unit area) to estimated failure stress, and so should be a more sensitive and specific predictor of vertebral fracture than spine areal BMD (aBMD) or volumetric BMD (vBMD), surrogates of bone strength alone. To address this issue, we analyzed the results of a case-control study of 89 postmenopausal women with vertebral fractures and 306 controls in Melbourne, Australia, and a 10-year community-based prospective study in which 30 postmenopausal women who had incident vertebral fractures were compared with 150 controls in Lyon, France. The FRI and vBMD of the third lumbar vertebral body and spine aBMD were derived using dual X-ray absorptiometry. In the cross-sectional analysis, each SD increase in FRI was associated with 2.1-fold (95% confidence interval [CI], 1.55-2.73) increased vertebral fracture risk, while each SD decrease in aBMD or vBMD was associated with 4.0-fold (95% CI, 2.69-6.18 and 2.65-6.94, respectively) increase in risk. Using receiver operating characteristic (ROC) analysis, the FRI was less sensitive and specific than aBMD in discriminating cases and controls (area under ROC, 0.76 vs 0.84, p<0.01). The area under ROC curve did not differ between FRI and vBMD (0.76 vs 0.79, NS). In the prospective data set, the FRI was not predictive [hazard ratio, HR, 1.20 (95% CI, 0.9-1.7)] and was in contrast to aBMD [HR, 2.4 (95% CI, 1.5-3.8)] and vBMD [HR, 2.1 (95% CI, 1.39-3.17)]. There was also lower sensitivity using a cutoff value of FRI>or=1 compared with aBMD T-score of -2.5 SD in both studies. There was poor agreement (kappa=0.13-0.18) between FRI and aBMD T -scores in detecting fractures; each method only identified around 50% of fractured cases. Within the constraints of the sample size, we concluded that applying a biomechanical index such as FRI at the spine is no better in discriminating fracture cases and controls than conventional aBMD or vBMD. The FRI may not predict incident vertebral fractures.