The relationship between treatment-related changes in total hip BMD measured after 12, 18, and 24 mo and fracture risk reduction in osteoporosis clinical trials: the FNIH-ASBMR-SABRE project

Abstract

There is a strong association between total hip bone mineral density (THBMD) changes after 24 mo of treatment and reduced fracture risk. We examined whether changes in THBMD after 12 and 18 mo of treatment are also associated with fracture risk reduction. We used individual patient data (n = 122 235 participants) from 22 randomized, placebo-controlled, double-blind trials of osteoporosis medications. We calculated the difference in mean percent change in THBMD (active-placebo) at 12, 18, and 24 mo using data available for each trial. We determined the treatment-related fracture reductions for the entire follow-up period, using logistic regression for radiologic vertebral fractures and Cox regression for hip, non-vertebral, "all" (combination of non-vertebral, clinical vertebral, and radiologic vertebral) fractures and all clinical fractures (combination of non-vertebral and clinical vertebral). We performed meta-regression to estimate the study-level association (r2 and 95% confidence interval) between treatment-related differences in THBMD changes for each BMD measurement interval and fracture risk reduction. The meta-regression revealed that for vertebral fractures, the r2 (95% confidence interval) was 0.59 (0.19, 0.75), 0.69 (0.32, 0.82), and 0.73 (0.33, 0.84) for 12, 18, and 24 mo, respectively. Similar patterns were observed for hip: r2 = 0.27 (0.00, 0.54), 0.39 (0.02, 0.63), and 0.41 (0.02, 0.65); non-vertebral: r2 = 0.27 (0.01, 0.52), 0.49 (0.10, 0.69), and 0.53 (0.11, 0.72); all fractures: r2 = 0.44 (0.10, 0.64), 0.63 (0.24, 0.77), and 0.66 (0.25, 0.80); and all clinical fractures: r2 = 0.46 (0.11, 0.65), 0.64 (0.26, 0.78), and 0.71 (0.32, 0.83), for 12-, 18-, and 24-mo changes in THBMD, respectively. These findings demonstrate that treatment-related THBMD changes at 12, 18, and 24 mo are associated with fracture risk reductions across trials. We conclude that BMD measurement intervals as short as 12 mo could be used to assess fracture efficacy, but the association is stronger with longer BMD measurement intervals.

Keywords: Bone mineral density; clinical fracture; fracture risk reduction; hip fracture; meta-regression; non-vertebral fracture; osteoporosis medication; randomised controlled trial; surrogate; vertebral fracture.

Graphical Abstract

Figure 1

Meta-regression plots displaying the association of active-placebo differences in THBMD percent change and vertebral and all clinical fracture risk reduction for 12- (A, D), 18- (B, E), and 24-mo (C,F) BMD measurement intervals, respectively. Dashed lines represent 95% prediction intervals. Individual trials are represented by circles with areas that are approximately proportional to the number of fractures in the trial. Drugs of the same class are represented by symbols of the same color. The red horizontal line is the odds/hazard ratio of 1.0 (no treatment effect on fracture risk), and the STE is the point where the upper 95% prediction interval intersects this line.