FRAX: A Valuable Tool, But It Has Limitations

By Chad Deal, MD

Case Presentation

B.S. is a 65-year-old white woman who presented for osteoporosis evaluation. She had been treated with a bisphosphonate for three years immediately following menopause (ages 52 to 55). She had a lumbar spine T-score of –2.3 and a femoral neck T-score of –2.2. Her current bone density showed a significant decline of 8.6 percent in the spine and 7.0 percent in the hip when compared with her bone density three years earlier. Laboratory tests did not reveal a secondary cause for low bone mass and bone loss. She was taking adequate calcium and vitamin D and walked for exercise four times a week.

Her 10-year absolute fracture risk based on the FRAX® tool was 1.9 percent for hip fracture and 10.0 percent for major osteoporotic fractures. Current National Osteoporosis Foundation guidelines recommend treatment if the 10-year fracture risk is 3 percent or greater for the hip or 20 percent or greater for a major osteoporotic fracture (hip, spine, wrist or humerus). Even though B.S.’s fracture risk was below treatment thresholds, she was started on a bisphosphonate.

FRAX Caveats: The Devil’s in the Details

The appropriate use of FRAX as a tool for guiding treatment decisions is important in the clinic. With FRAX, as with all tools, understanding its strengths as well as its limitations is essential to making appropriate treatment decisions. The limitations are often called FRAX caveats. In the case of B.S., the most important limitation is that the FRAX model does not adjust for patients with rapid bone loss. Allowing this patient to continue losing bone at a rate of 7 to 8 percent every three years would result in increasing fracture risk over time. This is a case when treatment for prevention is appropriate in spite of her FRAX-generated 10-year fracture risk being below the National Osteoporosis Foundation treatment threshold.

The sidebar lists clinical risk factors considered in FRAX. Many of these risk factors are dichotomous, involving a yes/no response (see screen shot), yet in the real world, these risk factors are more nuanced and complex. As a result, the dichotomous nature of some variables can result in either over- or underestimation of fracture risk. The risks for fracture in the FRAX model are averages in a large population of patients. Consider the following examples.

Fracture history. A patient with one vertebral fracture has a fivefold increase in fracture risk, while a patient with two vertebral fractures has a 12-fold increase in risk. Despite this difference, FRAX allows previous fracture to be reported only as “yes” or “no” with no adjustment for multiple fractures or for fracture site.

Smoking and alcohol use. FRAX assigns the same risk to a patient regardless of whether she smokes one pack a day or two packs a day or whether her alcohol use is three or six units daily. Moreover, “no” is the technically accurate entry for “current smoking” for a patient who quit a 40-year smoking habit six months ago, yet this leaves the skeletal effects of decades of smoking totally uncaptured.

Glucocorticoid use. The same increase in fracture risk is assigned by FRAX for a patient on 60 mg of prednisone for temporal arteritis as for a patient who was on 5 mg of prednisone for three months five years ago.

Rheumatoid arthritis. The severity of rheumatoid arthritis is likely to affect fracture risk (severe disease having a greater effect than mild disease), yet FRAX does not account for disease severity.

Family and personal fracture history. Family history of osteoporosis is represented only by parental hip fracture, not by spine fracture or other fracture types. Although personal history is represented by both fracture types, many spine fractures are asymptomatic and are thus not included in the FRAX calculation unless X-rays are reviewed or ordered.

Other Limitations

Additionally, the FRAX model does not include some factors that affect fracture risk, such as bone turnover and falls. The model also is hip-centric; when lumbar spine density is lower than hip density, the model will underestimate fracture risk. “FRAX-hip” might be a more precise name for the tool. Moreover, the model’s output ‒ 10-year fracture risk ‒ is absolute, with no confidence intervals. Whereas FRAX may assign a patient a risk of 19 percent, fracture risk (and life) is not so exact, and a standard deviation around the estimate would be welcome.

A final limitation, and one with relevance to our case patient B.S., is that FRAX is designed for use in treatment-naïve patients, with no guidance for a patient who took a bisphosphonate five or 10 years ago or who took estrogen but discontinued two years ago.

Still a Valuable Tool ‒ If Used Appropriately

The purpose of reviewing all the limitations of FRAX is to help the clinician wield this tool in an appropriate manner, not to negate its importance. The FRAX tool is a significant advance in global case finding of patients at increased risk for fracture compared with other methods available for predicting fracture risk before FRAX was introduced in 2008.

FRAX is a great starting point for assessing fracture risk, but its usefulness in guiding treatment decisions for individual patients depends, like the usefulness of all tools, on the knowledge and skill of the user

Dr. Deal is Director of the Center for Osteoporosis and Metabolic Bone Disease as well as Vice Chair for Quality and Outcomes in the Department of Rheumatic and Immunologic Diseases.