Predicting Fracture Risk in Patients with Osteoporosis

In 2014, the fracture risk algorithm FRAX celebrated its 20-year anniversary. Developed by a World Health Organization (WHO) study group in 1994, FRAX is a freely available diagnostic tool used to evaluate the 10-year probability of bone fracture risk, mostly for patients with osteoporosis. It is a part of many national health guidelines worldwide.

Because fracture probability and mortality show marked variation between countries, country-specific models have been developed where sufficient information is available. Currently, FRAX models are available for approximately 56 countries. It has also, like any clinical tool worth its salt in 2015, evolved from a paper-based system to a desktop and smart-phone app.

A recent mini-review in Maturitis looked at the current state of clinical use of FRAX, including its strengths and limitations. Traditional bone mineral density (BMD) measurements used in clinical practice show high specificity for prediction of fractures, but their sensitivity is quite low. Fracture risk algorithms such as FRAX improve prediction by including clinical risk factors that act independently of BMD. Further, a number of studies have demonstrated that the majority of fragility fractures occur in postmenopausal women who do not have osteoporosis.

“One reason for this is that some clinical risk factors other than low BMD contribute independently to fracture risk; inclusion of these risk factors significantly improves prediction of fracture and forms the basis of fracture risk algorithms such as FRAX,” noted review author Juliet Compston, MD, Chairman of the UK National Osteoporosis Guideline Group.

Among the limitations Compston cited:

• FRAX does not take into account important dose-response effects. For example, multiple previous fractures are given the same weight as a single previous fracture and the extent of tobacco use and alcohol abuse, both of which contribute significantly to fracture risk, is not quantified.
• Patients taking glucocorticoids, which are associated with increased risk of fracture, require greater upward adjustments in fracture probability.
• The date used to develop the FRAX algorithm were obtained from treatment naïve individuals; thus, the predictive accuracy of FRAX in individuals currently or previously treated with bone protective therapy is uncertain. Dr. Compston writes, however, that the results of a large study from Canada suggest that the use of FRAX may be valid in the prediction of fracture in postmenopausal women currently or previously treated with bone protective therapy. “In general, there was good agreement between predicted and observed fracture rates in both untreated and treated women, although in high risk women who had been adherent to long term bisphosphonate therapy the observed hip fracture risk was significantly lower than that predicted by FRAX,” Dr. Compston notes.
• The output of FRAX is restricted to hip or a composite of hip, wrist, humerus and clinical spine fracture probability. The majority of fractures in postmenopausal women are non-hip non-vertebral fractures, and many of these occur at sites not included in the FRAX output.
• Falls—a major risk factor for non-vertebral fractures but are not included in FRAX.

Despite these limitations, Compston suggested that FRAX is still comparable to other algorithms being developed, is freely available, and its questionnaire is easy to use and understand.

“Whether individuals identified at high risk by FRAX have a similar response to bone protective therapy as those selected on the basis of low BMD ±â€…fracture is an important issue for future research,” she concluded.