The relationship between baseline bone mineral density and fracture incidence in the placebo groups of randomized controlled trials using individual patient data from the FNIH-ASBMR-SABRE project

We have proposed to the Food and Drug Administration (FDA) that treatment-related increases in total hip BMD (TH BMD) at 2 yr could be a surrogate endpoint for fracture risk reduction in clinical trials. The qualification of a surrogate includes a strong association of the surrogate with the clinical outcome. We compiled a large database of individual patient data (IPD) through the Foundation for the National Institutes of Health-American Society for Bone and Mineral Research- A Study to Advance BMD as a Regulatory Endpoint (FNIH-ASBMR-SABRE) project, and this analysis aimed to assess the relationship between baseline BMD and fracture risk in the placebo groups. We estimated the association of baseline TH, femoral neck (FN), and lumbar spine (LS) BMD with fracture risk using IPD from the combined placebo groups, which included data from 46 666 placebo participants in 25 RCTs. We estimated the relative risk (RR) of fracture per SD decrease in baseline BMD using logistic regression models for radiographic vertebral fractures and proportional hazards models for hip, non-vertebral, "all," and "all clinical" fractures. Total person-years in the combined placebo groups was 250 662 (mean baseline age 70.2 ± 7.2 yr, mean TH BMD T-score -1.97 ± 0.90). We observed significant relationships between baseline TH BMD and vertebral (RR = 1.55/SD), hip (RR = 2.27), non-vertebral (RR = 1.31), all (RR = 1.43), and all clinical (RR = 1.35) fracture risk. Fracture risk estimates were similar for FN BMD and after adjustment for age, race, and study. Fracture incidence increased with decreasing TH BMD quintile, confirming the strong graded association between TH BMD and fracture risk. There was a strong relationship between LS BMD and vertebral fracture risk (RR = 1.56/SD), but only a weak association with non-vertebral (RR = 1.07) and no association with hip (RR = 1.01) fracture risk. These data support the very strong relationship between hip BMD and fracture risk and provide supporting rationale for change in TH BMD as a surrogate for fracture risk reduction in future RCTs.