Use of Fracture Risk Assessment Tool in clinical practice and Fracture Risk Assessment Tool future directions

Abstract

Fracture Risk Assessment Tool is a free, online fracture risk calculator which can be used to predict 10-year fracture risk for women and men over age 50 years. It incorporates seven clinical risk factors and bone density to give a 10-year risk of major osteoporotic fracture and hip fracture. This dynamic tool can be used with patients at the bedside to help guide treatment decisions. There are some limitations to Fracture Risk Assessment Tool, with the most central limitation being the fact that inputs are binary. Much research has been done to try to refine Fracture Risk Assessment Tool to allow for more accurate risk prediction, and this article describes the data for adjusting Fracture Risk Assessment Tool depending on the clinical scenario such as the dose of glucocorticoid use, presence of diabetes and others. Recently, the new FRAXplus tool has been developed to address many of these concerns and will likely replace the old Fracture Risk Assessment Tool in the future. At the current time, it is available in beta form.

Plain Language Summary

Methods for Refining the FRAX® Tool in Patients with Low Bone Density to Help Improve the Accuracy of Osteoporotic Fracture Risk Prediction

Many patients who have low bone density develop fragility fractures, even those whose bone density is not yet within the osteoporosis range. Thus, in patients with low bone density, the health care team should estimate the risk of fracture to decide which patients should take medications to prevent fractures. Factors such as age, body mass index, steroid use, family history and other clinical factors can influence the fracture risk, in addition to bone density. There is an online calculator called the Fracture Risk Assessment Tool (FRAX®) which allows patients and doctors to integrate these risk factors with bone density in order to estimate the 10 year risk of osteoporotic fractures. FRAX® asks a series of yes/no questions about the patient’s risks for fracture, and also takes into account the patient’s country of residence, age, gender, race and bone density at the femur neck. However, there are some important limitations of this calculator. For example, we think that steroid medications increase the risk of fractures, and the higher the dose, the higher the risk of fractures. However, FRAX® only allows a “yes” or “no” input to the steroid use question. This paper aims to descibe methods for refining the FRAX® calculation to make the fracture risk prediction more accurate. For example, it describes a mathematical adjustment to FRAX® to account for the dose of steroids used. It also reviews methods for FRAX® adjustment for diabetes type 1 and 2, and severity of rheumatoid arthritis, among other considerations. Importantly, there is a new FRAX® tool that is currently in beta testing which will also further refine the accuracy of fracture risk prediction.

Introduction

Osteoporosis is a prevalent condition in the United States. Recent estimates suggest that approximately 16% of women over age 50 years have osteoporosis by bone density criteria, and 52% have low bone mass or ‘osteopenia’. ¹ Approximately one in two postmenopausal women will have an osteoporotic fracture in her lifetime. ² Osteoporotic fractures are an important cause of morbidity and mortality as well health care cost worldwide. The aim of this article is to review the use of the Fracture Risk Assessment Tool (FRAX) calculator in predicting osteoporotic fractures and discuss methods of refining FRAX for various clinical conditions in specific patients.

Bone densitometry is the primary screening tool for the diagnosis of osteoporosis. As bone density declines, the risk of osteoporotic fracture increases. Pharmacologic therapy for osteoporosis is recommended in patients who have the diagnosis of osteoporosis by bone density assessment, those who have sustained low trauma hip or vertebral fractures and those with osteopenia who are at high risk of fracture.

Although those with osteoporosis have a higher risk of fracture than those with osteopenia, osteoporotic fractures are also common in patients whose bone density falls within the osteopenia/low bone mass range. In fact, a greater absolute number of osteoporotic fractures occur in patients whose bone density falls within the osteopenic/low bone mass range. Therefore, risk stratification of patients within the osteopenia/low bone mass range is critically important for reducing the overall burden of osteoporotic fractures.

While bone density is important in the assessment of osteoporotic fracture risk, many other factors contribute to fracture risk in an individual patient. These include, but are not limited to, advancing age, use of glucocorticoids, parental hip fracture, sarcopenia, poor balance/falls, rheumatoid arthritis, cigarette smoking, heavy alcohol use and many others. Fracture rates vary significantly between geographic regions in the world. In addition to other clinical risk factors, for a given bone density, the microarchitectural quality of the bone may vary significantly. Thus, risk stratification of patients within the low bone mass/osteopenia category should incorporate bone density, but ideally would also integrate clinical risk factors which are independent from bone density and a measure of bone quality.

The FRAX is an online fracture risk calculator which integrates seven clinical risk factors to estimate a 10-year risk of osteoporotic fracture. It was released in 2008 by the World Health Organization Collaborating Centre for Metabolic Bone Diseases in Sheffield, UK. ³ The calculation tool is available at https://frax.shef.ac.uk/FRAX. FRAX integrates clinical risk factors and uses country-specific data to estimate 10-year risk of osteoporotic fractures. It can be calculated using clinical risk factors and bone mineral density at the femur neck (FN), or using clinical risk factors alone if bone density testing is not easily accessible. Both methods have been validated as an accurate fracture prediction tool, although risk prediction is superior using the combination of clinical risk factors and bone density. ⁴

FRAX is intended to be a simple tool for use in the clinical setting. However, it has some important limitations. Most of the clinical risk factor inputs are binary. One must answer ‘yes’ or ‘no’ to glucocorticoid use, for example, without regard to dose or duration of treatment. Similarly, the number of previous fractures, amount and duration of cigarette smoking and other inputs cannot be entered in a quantitative way. Since FRAX was released in 2008, many studies have examined the inputs to FRAX in an attempt to further refine its ability to predict fractures. Various approaches can be used at the bedside to more finely adjust FRAX based on the clinical scenario.

Methods

PubMed searches were conducted between May 2022 and October 2023 using the search terms ‘FRAX’, ‘FRAX and diabetes’, ‘FRAX and glucocorticoids’ and ‘FRAXplus’. An Internet search was also conducted using the search phrase ‘FRAXplus’.

Parental hip fracture

Parental hip fracture is a binary input to FRAX. However, studies have examined whether the age of the parent at the time of hip fracture, gender of the parent or offspring affects this risk. Yang et al. performed a linkage analysis using data from a cohort in Manitoba. They found that the association between parental hip fracture and offspring fracture was the strongest when the parental hip fracture occurred before age 70 years and was not significant if the parental hip fracture occurred after age 80 years. Gender of the parent or offspring did not matter. ⁵ Parental hip fracture younger than age 80 years increased the risk for both hip fracture and major osteoporotic fracture in the offspring. This has led some to propose that when using FRAX, the ‘parental hip fracture’ box should be checked ‘yes’ only if the parent sustained the hip fracture at an age younger than 80 years.

Adjustment for glucocorticoid dose

Glucocorticoids are known to have a detrimental effect on bone. The FRAX tool currently recommends using ‘yes’ for glucocorticoid use ‘if the patient is currently exposed to oral glucocorticoids or has been exposed to oral glucocorticoids for more than 3 months at a dose of prednisolone of 5 mg daily or more’. ⁶ A UK study in 2000 showed that the increased fracture risk seen with systemic steroid use is dose dependent, with doses <2.5 mg prednisone equivalent per day conferring a very small increased risk, 2.5–5 mg per day conferring a moderately increased risk and >7.5 mg per day a much higher risk of fracture. The increased fracture risk seen with glucocorticoids is likely site-specific, with the highest increase in risk seen in vertebral fracture. In fact, the risk of vertebral fracture in the high-dose group was increased fivefold. ⁷ Importantly, the increased fracture risk seems to be at least partly reversible when glucocorticoids are stopped. ⁷

Kanis et al. examined the possibility of adjusting FRAX for glucocorticoid use. The calculation is complex and imperfect because FRAX variables are likely interdependent and data sets are imperfect. They found that the adjustment in fracture risk due to glucocorticoids was age-dependent. In general, glucocorticoids affect fracture risk more in younger patients. One simplified approach to adjusting FRAX for glucocorticoids would be: for a patient on steroids, check ‘yes’ for glucocorticoid use. Then, if the patient is on low-dose glucocorticoids (<2.5 mg prednisone per day), hip fracture risk could be reduced by 30%–40%, whereas major osteoporotic fracture risk could be decreased by 15%–20%. With high-dose steroids (>7.5 mg prednisone per day), the risk of hip fracture should be increased by 10%–25% and the risk of major osteoporotic fracture should be increased by 10%–20%. ⁸ Other age specific algorithms have also been suggested whereby the fracture risk is adjusted based on both glucocorticoid dose and age. ⁸

Rheumatoid arthritis–disease severity

In many cases, secondary causes of osteoporosis such as hyperparathyroidism increase fracture risk by lowering bone density. However, rheumatoid arthritis is a secondary cause of osteoporosis whose effect on FRAX is independent from bone density. For this reason, rheumatoid arthritis is an independent risk on the FRAX calculator. Some have speculated that fracture risk may vary with disease activity, but this has not been substantiated in clinical studies. ⁹

Adjustment for diabetes

The relationship between diabetes and fracture risk is complex. Patients with both type 1 and type 2 diabetes have significantly higher fracture risk, with the risk associated with type 1 diabetes much higher than type 2. The exact cause of these increased risks are unknown, but likely involve neuropathy and resultant difficulty with balance, frailty, decreased mobility, diabetes medications, hypoglycemia and deterioration of bone microarchitecture.

Type 1 diabetes

Type 1 diabetes is one of the ‘secondary causes of osteoporosis’ clinical risk factor inputs to the FRAX calculator. However, this input is removed from the FRAX calculation when bone mineral density (BMD) is included, as it has been assumed in the past that the increased fracture risk seen with type 1 diabetes is attributable to lower bone density. However, this assumption has been called into question in recent years. In 2015, Shah et al. ¹⁰ demonstrated that the risk of hip and spine fractures is approximately three- to four-fold higher in patients with type 1 diabetes than in controls with the most striking difference seen in hip fracture. In addition, hip fractures were observed in patients with type 1 diabetes who were 10–15 years younger than controls. ¹¹ However, when bone density was examined, the Shah’s group found only a modest reduction in BMD in patients with type 1 diabetes, seen most clearly at the FN. ¹² Specifically, they observed that FN bone density in patients with type 1 diabetes was −0.055 g/cm² lower than controls, a small but statistically significant difference. The difference in bone density was not sufficient to explain the very significantly higher risk of osteoporotic fractures. ¹⁰ For this reason, FRAX probably significantly underestimates the fracture risk in patients with type 1 diabetes. The ability of FRAX to predict fractures in patients with type 1 diabetes has not been studied extensively, and there is no proposed modification to FRAX to account for the much higher fracture risk seen in type 1 diabetes.

Type 2 diabetes

The relationship between type 2 diabetes and fracture risk is also not mediated through change in bone density. In fact, type 2 diabetes is associated with higher body mass index (BMI) and higher bone density, but a paradoxical increased risk for major osteoporotic fractures. ¹³ This increased risk is not integrated into FRAX. In order to adjust for the increased risk of fracture in type 2 diabetes, an International Osteoporosis Foundation working group recommends checking ‘yes’ on the rheumatoid arthritis input in patients with type 2 diabetes.^13,14 In addition, the fracture risk in a diabetic patient with a score −2.0 could be considered equivalent to a non-diabetic with a T-score of −2.5. ¹⁴

Recency of previous fracture

Previous fracture is a strong predictor of future fracture. The FRAX tool uses a binary ‘yes’ or ‘no’ for previous fracture. Importantly, the recency of the previous fracture modifies risk such that the highest increased risk is seen within the first year of the initial fracture. The concept of a changing relative risk depending on recency to the first fracture has been termed ‘imminent fracture risk’.^15,16 A method for adjusting fracture risk based on recency of vertebral fracture has been proposed by Kanis et al. ¹⁷ Their calculations suggest that vertebral fracture within 2 years increases risk of future fracture by 1.5- to 2.5-fold in women younger than 70 years compared to women with more remote vertebral fracture. In general, practitioners should keep in mind that osteoporotic fractures tend to cluster in time and that the highest risk of recurrent fracture is seen in the first year after the initial fracture. Prompt treatment to reduce fracture risk is crucial.

Discordance between hip and spine bone density

The FRAX calculator uses FN bone density to calculate 10-year fracture risk. In the clinical setting, it is not uncommon for patients to have significant discordance between the hip and spine. Should the FRAX be adjusted if the spine is much lower than the FN or vice versa? At least two studies have attempted to answer this question. In 2010, Leslie et al. ¹⁸ published an analysis from the Manitoba BMD database and suggested that a simple method for adjusting FRAX for spine–hip discordance was to ‘increase/decrease FRAX estimate for a major fracture by one tenth for each rounded T-score difference between LS and FN’. In 2014, Johansson et al. analysed data from several large cohorts and largely substantiated Leslie’s calculations. They examined the difference in the T-score between the spine and hip for >21,000 patients. They found that it was quite uncommon for patient to have a discordance of 2 SD or more (2.5% of the group). Moderate discordance (1–2 SD) was seen in about 21%. They also found that for each 1 SD difference between the hip and spine, the fracture risk could be increased/decreased by 10%. For example, if the patient’s T-score at the spine was −2.2 and hip −1.2, the FRAX risk could be initially calculated using the FN data, and then increased by 10%. Importantly, in adjusting fracture risk in this manner, it was uncommon to re-classify patients into ‘treatment’ or ‘no treatment’ group and more commonly downgraded patients out of the ‘treatment’ group than the other way around. ¹⁹

Use of FRAX on therapy

FRAX was originally intended to be used in treatment naïve, osteopenic patients to identify those at high risk of fracture in whom treatment could be considered. Traditionally, the FRAX has not been used in patients who are currently on treatment. This is in part due to the fact that bisphosphonates reduce fracture risk to a degree which is not proportional to change in bone density, that is, fracture risk reduction is greater than would be expected based on improvement in bone mineral density. ²⁰ However, studies have shown that FRAX risk continues to predict risk of fracture even in most patients who are currently on treatment. ²¹ This may not apply to patients who have been on continuous and consistent bisphosphonate therapy for 5 years or more. In clinical practice, calculation of FRAX in some patients on therapy may help with treatment decisions about whether to continue or stop therapy.

Trabecular bone score

Trabecular bone score (TBS) is a textural assessment of bone quality which is calculated from lumbar spine DXA data. More homogeneous bone structure results in a higher TBS and more heterogeneous bone results in a lower TBS. Multiple clinical trials have validated the TBS as a predictor of major osteoporotic fracture in women and older men. ²² It may be particularly effective at predicting fracture in those with type 2 diabetes or taking glucocorticoids. Importantly, TBS is not affected by degenerative changes in the spine which artifactually increase bone mineral density assessment in many cases. Despite its value as a predictive tool, clear TBS thresholds for treatment have not been established.

A method for incorporating TBS into FRAX has been developed. On the FRAX calculation page, an option for ‘adjust with TBS’ is available. This allows the user to input TBS, if available, to further improve the accuracy of risk prediction. This is particularly important in younger women with low TBS for whom low TBS is most predictive for fracture. Note that TBS should not be used to monitor response to treatment with bisphosphonates or other antiresorptives, since TBS changes are generally very small.

FRAXplus, beta version

In early 2023, researchers from the University of Sheffield released a beta version of a new FRAX calculator which incorporates many of the modifications described above. Using the FRAXplus calculator, the clinician has the opportunity to input the recency of the osteoporotic fracture, site of the osteoporotic fracture, glucocorticoid dose, presence and duration of type 2 diabetes, TBS, number of falls within the preceding year, hip axis length and discordance between the bone density at the hip and spine. This new version of FRAX allows for further refinement of osteoporotic risk, and it is available at: https://www.fraxplus.org/frax-plus. To use it, credits must be purchased on the website, and proceeds will support ongoing FRAX development. ²³ At the current time, the website notes that this beta version is undergoing public testing and it is suggested not to use for clinical purposes. ²⁴

Discussion

Most osteoporosis medications are inexpensive, effective and have a favourable side effect profile; however, patients frequently have serious concerns about the side effects of these medications. There is much patient education and counselling required upon initiation of these medications, and unfortunately, they are significantly under-prescribed. ²⁵ The FRAX tool can and should be used at the bedside, as a tool for shared decision-making between patient and health care provider. It is easily accessed using any Internet search engine. It allows patients to visualize and understand risks. However, it is an imperfect tool, and these are several helpful and easy techniques as described above to further refine risk. There are likely other risk factors for fracture that could be used as modifiers for FRAX that have not been included in this review due to lack of data.

Since its release in 2008, various groups have studied FRAX’s ability to predict risk of fracture and have proposed methods for refining FRAX. The new FRAXplus tool was developed to improve fracture risk prediction and will likely replace the original FRAX tool in the future. No doubt, fracture risk assessment will continue to undergo refinements and modifications over time as research in this field progresses.

Conclusion

Although each individual patient should always be considered in their unique clinical context, using FRAX provides substantial assistance to the clinician–patient team by producing quantitative, individualized risk estimates to guide treatment decisions. In future, FRAXplus may provide a more refined tool for this purpose.