To FRAX or not to FRAX

The WHO Fracture Risk Assessment Tool (FRAX) calculator is the most thoroughly studied and widely used tool of fracture risk assessment.1 FRAX is now available in 39 countries and is increasingly being used as a guide for clinical decision-making. The specific objective of developing FRAX was to aid clinicians in identifying the most appropriate patients to receive pharmacologic therapy to reduce fracture risk. Prior to FRAX, treatment decisions were based primarily on bone mineral density (BMD) values with the threshold at which treatment was recommended being modified crudely by the simple presence of absence of other clinical risk factors. For example, the 1998 National Osteoporosis Foundation (NOF) guidelines suggested treating all postmenopausal women with a BMD T-score of −2 or lower and all women with a T-score of −1.5 or lower who had one of several risk factors, including being thin, having a fragility fracture as an adult or a family history of fragility fracture.2 These guidelines identified for therapy almost all patients at high risk for fracture (sensitivity of the approach was good) but enfranchised therapy for many younger postmenopausal women with low bone density but who were at modest or even low fracture risk (poor specificity). Understanding this, clinicians asked themselves "Which patients who do not have osteoporosis should be treated?" FRAX was specifically developed to answer that question. After the 1998 NOF guidelines were released, we recognized that individuals could be stratified into gradations of fracture risk more effectively if BMD was combined with other independent clinical risk factors for fracture than could be accomplished by using individual risk factors for such as age, BMD, or fracture history alone.3 The FRAX tool was derived by evaluating the relationships between BMD, clinical risk factors, and fracture risk using individual patient data comprising almost 250,000 patient-years of observation in nine large observational studies performed in different parts of the world.1 Subjects in those studies were generally healthy older adults who were not on osteoporosis treatment. In the United States and Europe, FRAX accurately predicts the 10-year probability of major osteoporotic fracture and hip fracture in patients with characteristics like the study subjects.4, 5 Importantly, because FRAX was designed to identify patients for whom treatment would be beneficial, analyses of clinical trials with different classes of osteoporosis treatments have documented that this is true.6-9 Since its availability in 2008, FRAX estimates of fracture probability have been incorporated, albeit in various ways, into clinical guidelines of several national societies including the NOF.10-12 The fracture risk thresholds at which treatment is recommended in the NOF guidelines were based on clinical and health economic considerations.13 Despite the scientific basis of FRAX and its strengths, several limitations are recognized. FRAX treats all risk factors as categorical, not taking into account the dose or duration of glucocorticoid therapy or the number, type, severity, or recency of a fragility fracture, which are clinical variables known to affect fracture risk.14, 15 In patients with risk factors for fracture that are not accounted for in the FRAX model, such as diabetes, FRAX estimates of fracture probability are less accurate.16 Clinicians often use additional information such as history of fall risk to make treatment decisions. Although these are well-recognized risk factors for fracture, they were intentionally not included in the FRAX calculator. Remember that FRAX was developed to identify patients who would benefit from pharmacologic therapy. While FRAX was being developed, evidence existed that elderly patients with fall-related risk factors but who did not have osteoporosis did not experience fracture risk reduction with bisphosphonate therapy.17 Other risk calculators, such as one based on an American cohort in the Study of Osteoporotic Fractures and the Garvan Fracture Risk Calculator, do take measures of frailty and fall risk into account.18, 19 While these other risk calculators are effective tools to assess fracture risk, they have not been evaluated as a means to identify patients who are responsive to bone-strengthening drugs. When FRAX became available, clinicians were advised not to use the tool in patients receiving osteoporosis treatment. To avoid the use of FRAX in patients on treatment in the United States, the NOF and the International Society for Clinical Densitometry (ISCD) developed the "FRAX filter," which inactivated the calculation of FRAX in DXA software if the patient was receiving osteoporosis therapy.20 The basis of this recommendation was twofold: (1) FRAX was meant to be used to decide who should be treated, a decision already made in patients receiving therapy; and (2) FRAX does not take into account the anticipated effect of the osteoporosis treatment on reducing fracture risk. Whether the latter reason was a valid concern was not known because FRAX was never calibrated in patients who were receiving therapy. This question of whether FRAX can be used in patients receiving osteoporosis therapy was addressed by Dr. William Leslie and his colleagues in this issue of the journal.21 Taking advantage of the Manitoba Bone Density Program database, Leslie's group retroactively calculated FRAX values in more than 37,000 women age 50 years or older who had BMD testing performed between 1997 and 2004. The women were followed for an average of 5.3 years; both fracture incidence and osteoporosis drug therapy were assessed. The FRAX-predicted and observed fracture rates were then compared in the groups of patients who either did or did not receive osteoporosis treatment. Reassuringly, the predicted and observed fracture rates were concordant in patients not receiving osteoporosis treatment. The predicted incidence of major osteoporosis fractures was 10.6% whereas the observed incidence was 10%. Both the FRAX-predicted and observed incidence of hip fracture were 1.9%. These results confirm again that the FRAX calculator is accurately calibrated to predict fracture probability in a healthy population of postmenopausal white women, this time in Manitoba. In patients who had previously taken an osteoporosis drug (most often estrogen) before their inclusion in the study or in patients given a prescription for an osteoporosis drug but who were not adherent to therapy, predicted and observed fracture rates were also similar. These data are consistent with prior observations that poor adherence to bisphosphonate therapy blunts the effectiveness of treatment.22 Also, the skeletal benefits of estrogen therapy wane quickly upon discontinuation, so no residual effect of prior estrogen therapy would have affected the fracture rates in those who stopped treatment.23 The most interesting results, and the major focus of the study, involved the evaluation of FRAX in patients on osteoporosis therapy who were adherent to their treatment. In these women, FRAX estimates of fracture probability clearly correlated with observed fracture incidence; higher FRAX scores were associated with higher incidence of both major osteoporotic and hip fractures. However, FRAX did not accurately predict the hip fracture rate in these treated patients. The observed incidence of hip fracture was 39% lower (confidence interval, 17%–60%) than was predicted by the FRAX model in these adherent treated patients, consistent with meta-analysis of the effectiveness of alendronate.24 In contrast, when the FRAX-predicted and observed incidence of major osteoporotic fracture rates were compared, no difference was observed. As with all good studies, these results provide us with some answers and but also leave us with some new questions. If osteoporosis treatment reduced the incidence of hip fracture probability, why was an offset in calibration of major osteoporotic fracture probability was not also observed in treated patients? Maybe this result is not as surprising as it might seem at first. Oral bisphosphonates have only a modest effect on overall nonvertebral fracture risk, reducing relative risk by 16% to 20%.24, 25 It is difficult, even in a large database, to observe such a small difference between predicted and observed fracture risk. Additionally, as Dr. Leslie points out, due to poor compliance or persistence of therapy or other reasons, the effectiveness of treatment in daily clinical practice may not be as robust as the effects observed in the pristine environment of a clinical trial. Another possible explanation is that nonvertebral fracture risk reduction has only been observed in patients known to have osteoporosis by BMD testing or history of a hip or spine fracture.26 The average femoral neck T-score in the Manitoba cohort who were adherent to therapy was −1.8. Perhaps osteoporosis therapy was simply not effective in reducing overall clinical fracture risk in this population, and the FRAX-predicted fracture probability—which did not take into account an effect of treatment—accurately predicted fracture incidence because there was no treatment effect. So, can FRAX be used in patients receiving osteoporosis therapy? I think that the answer is a clear, "Yes!" The Manitoba results demonstrate that higher FRAX scores identify patents on therapy more likely to fracture than those with lower scores and that FRAX can be used, without adjustment, to predict the probability of major fractures in patients on current pharmacologic therapy. For hip fracture probability, an adjustment must be made for those patients on treatments known to reduce hip fracture risk. Knowing that FRAX can be used in treated patients, the next question becomes "Why should one want or need to use FRAX in a treated patient?" There are at least two clinical situations in which this might be considered. Many patients currently receiving osteoporosis drugs do not meet current indications for treatment. Retroactively calculating FRAX, based on a patient's age, BMD, and risk factors at the time treatment was begun, could be used to determine whether treatment is appropriate, but that clinical information is often not available. Thus, clinicians are left with the goal of determining fracture probability based on current clinical information—including FRAX scores—to decide whether the osteoporosis treatment was justified. The other situation involves the potential use of FRAX to aid in the decision about whether a bisphosphonate drug holiday would be appropriate. Is it important to know that FRAX can be used in patients on therapy? I personally do not think so. Certainly FRAX cannot be used to monitor response to an osteoporosis treatment. Many of us were making the decisions about stopping treatment in patients who did not need it or recommending a drug holiday before FRAX was available. It is not clear that knowing a patient's FRAX score improves our ability to make these decisions. However, for clinicians who feel that the additional information from FRAX would be useful, the data from Manitoba provide the basis to do so. At least, the question of whether FRAX works in patients on therapy can now be laid aside as answered. An additional comment is appropriate. Dr. Leslie and his colleagues are to be congratulated for taking advantage of their unique and well-characterized set of data to address a question posed by practicing clinicians. It is refreshing to observe clinical investigators be cognizant of questions confronted in daily practice and to use their large database to answer them. Such interplay between those attempting to make clinical decisions with incomplete evidence and investigators who have the capacity to provide the evidence needs to occur more frequently. MRM has received research grants from Amgen, Lilly, Merck, and Procter & Gamble, and consultant fees and/or honoraria from Amgen, Lilly, Merck, Novartis, and Warner-Chilcott.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.