The predisposition of patients with osteoporosis to fragility fractures has substantial public health implications. Now, two publications in the Journal of Bone & Mineral Research have the potential to improve fracture risk prediction and provision of antiosteoporotic therapy.

The first relates to an established fracture risk assessment tool, FRAX®. As stated by its lead author, William Leslie, of the University of Manitoba, Canada, “10-year fracture risk assessment with FRAX® is increasingly used to guide treatment decisions, and has been incorporated into many national clinical practice guidelines.” However, “a common question from clinicians is whether FRAX® has a role in patients currently or recently receiving treatment for osteoporosis.” To address this issue, the researchers analyzed the performance of FRAX® according to medication use, using data collected between 1996 and 2007 for 35,764 Canadian women ≥ 50 years old who underwent bone mineral density (BMD) testing. Patients were assigned to four treatment categories: untreated; current high adherence users; current low adherence users; and past users.

“FRAX® appropriately stratified major osteoporotic and hip fracture risk within untreated and treated subgroups,” explains Leslie. “Concordance (calibration) plots for major osteoporotic fractures and hip fractures showed good agreement between the predicted and observed 10-year fracture incidence in untreated women and each treated subgroup,” suggesting that medications do not affect FRAX® performance; however, hip fracture incidence in 3,407 patients who received bisphosphonates for ≥ 5 years among the 9,712 current high adherence users was lower than expected. Furthermore, nonprescription and nonpharmacological therapies could not be assessed.

Cross-sectional HRpQCT image of a radius in a 77-year-old female. Image courtesy of Elizabeth Atkinson.

“Since many individuals were initiated on treatment prior to the availability of FRAX®, this potentially expands the use of FRAX® for advising patients on their need for continued treatment,” explains Leslie. “It will be important to confirm our findings in other cohorts, and to see whether there are differences among individual medications,” he concludes.

On this basis, “FRAX® can presumably be used to predict fracture probability in women currently or previously treated for osteoporosis,” says Piet Geusens, of the Maastricht University Medical Center, The Netherlands. However, Geusens also emphasized the need to study the value of FRAX® in fracture risk reduction, as “no studies are available on fracture reduction based on FRAX® alone.”

The second publication, from Elizabeth Atkinson and colleagues from the Mayo Clinic, USA, highlights a novel approach to fracture risk assessment based on combined analysis of a range of bone imaging data. “In addition to the main variables, we routinely obtain extra measured and/or derived quantities that are poorly understood and so generally have been ignored,” states Atkinson. In isolation, these measurements might not have much value in fracture prediction, but together could be highly informative.

Atkinson et al. examined the predictive ability of a gradient boosting machine (GBM) algorithm, which incorporated 267 different bone density, structure and strength measurements derived from dual-energy X-ray absorptiometry (DXA), high-resolution peripheral quantitative CT (HRpQCT) and spiral QCT data. In 322 postmenopausal women, 139 with fractures, 99 distal forearm and 40 vertebral, and 183 without fracture, GBM modeling identified fracture and nonfracture cases with considerably higher accuracy than femoral neck BMD alone. GBM performance was increased if all the variables from DXA, HRpQCT and spiral QCT scans were used, rather than information from each technique alone. “Information is not captured by one or two variables, but involves the interrelation of several aspects, suggesting that there may be additional underutilized information produced by the scans,” says Atkinson.

“The investigators were also able to apply their technique developed for distal radius fractures successfully to the fracture cohort consisting of patients with and without spine fractures, and vice versa,” adds Thomas Link at the University of California, San Francisco, USA. “The results of this study also provide new insights in the important role of bone structure in predicting fracture status,” Link continues; however, “while this is a superb tool for research studies, in a clinical setting it may be more difficult to apply as it is very complex.”

“Work will need to better understand what 'features' the model is capturing with the hope of developing simpler summary measures,” acknowledges Atkinson. “Another important step will be to apply the resulting model to other cohorts; however, this is an exciting approach that ultimately may provide deeper insights into predictor variables that might improve fracture risk assessment.”