Embed Size (px)
Citation preview
Bone 44 (2009) 734–743
Contents lists available at ScienceDirect
Bone
j ourna l homepage: www.e lsev ie r.com/ locate /bone
Review
FRAX® and its applications to clinical practice
John A. Kanis ⁎, Anders Oden, Helena Johansson, Fredrik Borgström, Oskar Ström, Eugene McCloskeyWHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield, UK
⁎ Corresponding author. WHO Collaborating CentreUniversity of Sheffield Medical School, Beech Hill Road, S114 285 1813.
E-mail address: [email protected] (J.A. Kanis
8756-3282/$ – see front matter. Crown Copyright © 20doi:10.1016/j.bone.2009.01.373
a b s t r a c t
a r t i c l e i n f oArticle history:
The introduction of the WH Received 2 December 2008Revised 15 January 2009Accepted 21 January 2009Available online 3 February 2009Edited by: R. Baron
Keywords:FRAX®
Fracture probabilityRisk assessmentCost-effectivenessEfficacy
O FRAX® algorithms has facilitated the assessment of fracture risk on the basis offracture probability. FRAX® integrates the influence of several well validated risk factors for fracture with orwithout the use of BMD. Its use in fracture risk prediction poses challenges for patient assessment, thedevelopment of practice guidelines, the evaluation of drug efficacy and reimbursement, as well as for healtheconomics which are the topics outlined in this review.
Crown Copyright © 2009 Published by Elsevier Inc. All rights reserved.
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734Assessment of patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735
Rationale for use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 736
Clinical guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737Assessment of drug efficacy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739Health economic applications of FRAX®. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 741References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 742
Introduction
FRAX® is a computer based algorithm (http://www.shef.ac.uk/FRAX) that provides models for the assessment of fractureprobability in men and women [1–3]. The approach uses easilyobtained clinical risk factors (CRFs) to estimate 10-year fractureprobability. The estimate can be used alone or with femoral neckbone mineral density (BMD) to enhance fracture risk prediction. Inaddition, FRAX® uses Poisson regression to derive hazard functionsof death as well as fracture. These hazard functions are continuousas a function of time which permit the calculation of the 10-yearprobability of a major osteoporotic fracture (hip, clinical spine,
for Metabolic Bone Diseases,heffield S10 2RX, UK. Fax: +44
).
09 Published by Elsevier Inc. All rig
humerus or wrist fracture) and the 10-year probability of hipfracture. Some of the risk factors affect the risk of death as well asthe fracture risk. Examples include increasing age, low body massindex (BMI), low BMD and smoking. Other risk engines calculate theprobability of a clinical event (e.g. a myocardial infarct) withouttaking into account the possibility of death from other causes. Inaddition, the FRAX® model can be calibrated for different countries[1–3].
Probability of fracture is calculated in men or women from age,body mass index (BMI) computed from height and weight, anddichotomised risk variables that comprise;
a prior fragility fracture,parental history of hip fracture,current tobacco smoking,ever long-term use of oral glucocorticoids,rheumatoid arthritis,
hts reserved.
735J.A. Kanis et al. / Bone 44 (2009) 734–743
other causes of secondary osteoporosis,daily alcohol consumption of 3 or more units daily.
These variables are entered onto the web site. Femoral neck BMDcan additionally be entered as a T-score derived from the NHANES IIIdatabase for female Caucasians aged 20–29 years [4]. When entered,calculations give the 10-year probabilities as defined above with theinclusion of BMD (Fig. 1).
The relationships between risk factors and fracture risk incorpo-rated within FRAX® have been constructed using information derivedfrom the primary data of nine population based cohorts from aroundthe world, including centres from North America, Europe, Asia andAustralia and has been validated in 11 independent cohorts (mainlywomen) with a similar geographic distribution with in excess of 1million patient years [5]. The use of primary data for the modelconstruct permits the determination of the predictive importance in amultivariable context of each of the risk factors, as well as interactionsbetween risk factors, and thereby optimises the accuracy by whichfracture probability can be computed. The large sample permits theexamination of the general relationship of each risk factor by age, sex,duration of follow up and, for continuous variables (BMD and BMI),the relationship of riskwith the variable itself in amanner hitherto notpossible. The use of primary data also eliminates the risk ofpublication bias.
In addition to the clinical risk factors, fracture probability variesmarkedly in different regions of the world [6]. Thus the FRAX® modelsneed to be calibrated to those countries where the epidemiology offracture and death is known. At present FRAX® models are availablefor Austria, China, Germany, France, Italy, Japan, Spain, Sweden,Switzerland, Turkey, and the UK and US. Other models are beingdeveloped, but there are relatively few other countries with sufficientinformation to construct FRAX® models [3], and these are listed belowaccording to categories of hip fracture risk.
(a) Very high risk (e.g. Denmark, Iceland, Norway, Sweden, UnitedStates).
Fig. 1. Chart for input of data and format of results in the UK version of the FRAX® tool. [WithDiseases, University of Sheffield Medical School, UK].
(b) High risk (e.g. Australia, Austria, Canada, Finland, Germany,Greece, Hungary, Italy, Kuwait, Netherlands, Portugal, Singa-pore, Switzerland, Taiwan, UK).
(c) Moderate risk (e.g. Argentina, China, France, Hungary, HongKong, Japan, Spain).
(d) Low risk (e.g. Cameroon, Chile, Korea, Turkey, Venezuela).
Each category of risk has been represented in the FRAX® modelscurrently available (in italics, above). Thus in the absence of a FRAX®
model for a particular country, a surrogate country should be chosen,based on the likelihood that it is representative of the index country.
The obvious application of FRAX® is in the assessment ofindividuals to identify those who would be candidates for pharma-cological intervention, and it has beenwidely used since the launch ofthe web site, currently receiving on average 55,000 hits daily. ButFRAX® should not be used in the clinic without an appreciation of itslimitations as well as its strengths. There are also challenges to befaced in the construct of new clinical guidelines, the assessment ofpharmacologic agents for drug registration (in Europe) and in healtheconomics. These applications are the focus of this review.
Assessment of patients
Rationale for use
Until recently, the majority of clinical guidelines for the manage-ment of osteoporosis have made recommendations for interventionbased predominantly on the basis of the T-score for BMD [3]. In theUK, for example, guidance for the identification of individuals at highfracture risk was provided until recently, by the Royal College ofPhysicians (RCP) [7–9]. The guidance was based on an opportunisticcase finding strategy where physicians are alerted to the possibility ofosteoporosis and high fracture risk by the presence of clinical riskfactors (CRFs) associated with fracture. This provides a trigger for themeasurement of BMD, and treatment was considered in those with aBMD value that lies in the range of osteoporosis as defined by the
permission of the World Health Organization Collaborating Centre for Metabolic Bone
Fig. 2. Ten year probability of hip facture by age and femoral neck BMD in women fromSweden [17 with permission from Springer].
736 J.A. Kanis et al. / Bone 44 (2009) 734–743
World Health Organization (WHO). Treatment was, however, alsorecommended for women with a prior fragility fracture withoutnecessarily measuring BMD. A similar approach has been used inmany European countries [10–12]. The National Osteoporosis Foun-dation (US) adopted a similar approach, though used a less stringentT-score [13].
Since the development of such guidelines, it has become apparentthat the presence of several of the risk factors used to trigger a BMDtest is associated with a fracture risk greater than can be accounted forby BMD alone [14]. Thus, the assessment of fracture risk should takeaccount of specific risk factors that contribute to fracture risk inaddition to BMD, since this would increase the sensitivity for fractureprediction (i.e. detection rate of individuals who would fracture) [15].Themost obvious example is agewhich confers a risk greater than thataccounted for by BMD. For example, from the known relationship ofBMD and fracture risk, it might be expected that hip fracture riskwould increase approximately 4-fold between the age of 55 and85 years in women because of the age-related decrease in bone mass.
Fig. 3. Ten year probability of a major osteoporotic fracture in Caucasianwomen aged 65 ears24 kg/m2.
In practice, hip fracture risk in many countries increases 40-fold [16].Thus, over this age range, the impact of increasing age is 11-foldgreater than the impact of decreasing BMD. The interaction of age andBMD has been formalized in several studies [16,17]. In the case of DXAat the femoral neck, the risk of fracture by age varies markedly at thethreshold of osteoporosis, i.e. with a T-score of exactly −2.5 SD. Thus,at the age of 50 years the 10-year hip fracture probability isapproximately 2% in women but at the age of 80 years it is 12% forthe same T-score (Fig. 2). For any major osteoporotic fracture (hip,forearm, shoulder or clinical spine fracture), the 10-year probability inwomen with a T-score of −2.5 SD ranges from 11% at the age of50 years to 26% at the age of 80 years [16].
This simple example illustrates that fracture risk can be moreaccurately assessed from age and BMD than by BMD alone. Similarconsiderations pertain to the other clinical risk factors used in FRAX®,which each make an independent contribution to fracture risk asillustrated in Fig. 3.
Limitations
Experts in the care of patients with osteoporosis are used tointegrating information derived from multiple risk factors. If aphysician is prepared to treat a patient with a T-score of say,−2.5 SD, then intuitively the fact that a patient is additionally takingoral glucocorticoids would lead him to intervene at a higher T-score.By contrast, primary care physicians in most countries have littleexpert knowledge and it is this constituency for which FRAX® isprimarily designed.
The use of FRAX® with the generation of a number does not,however, replace clinical judgment. For example, several of the clinicalrisk factors identified take no account of dose–response, but give riskratios for an average dose or exposure. By contrast, there is goodevidence that the risk associated with excess alcohol consumption,cigarette smoking and the use of glucocorticoids is dose-responsive[18–20]. In addition, the risk of fracture increases progressively withthe number of prior fractures [21,22]. These limitations should berecognised when interpreting the FRAX® result in the clinic.
It should also be acknowledged that there are many other riskfactors for fracture that are not incorporated into assessment algo-rithms. Examples include the biochemical markers of bone turnover,
from the US according to the presence of the clinical risk factors shown. The BMI is set at
Fig. 4. Management algorithm for the assessment of individuals at risk of fracture[3, with permission of the World Health Organization Collaborating Centre forMetabolic Bone Diseases, University of Sheffield Medical School, UK].
737J.A. Kanis et al. / Bone 44 (2009) 734–743
risk factors for falls and previous exposure to pharmacologicintervention. The reason for not including such risk factors in FRAX®
relates to the paucity of large international data bases, but the clinicianmay wish to take such information into account.
At present the FRAX® tool limits BMD to that measured at thefemoral neck. This is because of the wealth of data available for thissite. It has the advantage that for any given age and BMD, the fracturerisk is approximately the same in men and women [23]. Because ofthis, the T-score is derived from a single reference standard (theNHANES III database for female Caucasians aged 20–29 years) aswidely recommended [3,24]. There are, however, other bone mea-surements that provide information on fracture risk [3]. These includeBMD at other skeletal sites [25], quantitative ultrasound [26] orcomputed tomography [27] and the biochemical indices of boneturnover [28]. The available information was too sparse to provide ameta-analytic framework for the present version of FRAX®, but otherassessment tools should be incorporated into risk assessmentalgorithms when they are more adequately characterised.
Provision is made for the inclusion of many secondary causes ofosteoporosis. A distinction is made between rheumatoid arthritis andother secondary causes. There is good evidence that rheumatoidarthritis carries a fracture risk over and above that provided by BMD[29]. Whereas this may hold true for other secondary causes ofosteoporosis, the evidence base is weak. For this reason, the othersecondary causes of osteoporosis are conservatively assumed tomediate fracture risk as a result of low BMD. It is assumed that theyincrease fracture risk in a manner similar to patients with rheumatoidarthritis. However, when BMD is entered into the FRAX® equations, noweight is accorded by these other secondary causes [1].
For these reasons, the FRAX® tool should not be considered byphysicians as a gold standard, but rather as a platform technology onwhich to build as new validated risk indicators become available.Notwithstanding, the present model provides an aid to enhancepatient assessment by the integration of clinical risk factors aloneand/or in combination with BMD.
Clinical guidelines
The application of this methodology to clinical practice demandsa consideration of the fracture probability at which to intervene,both for treatment (an intervention threshold) and for BMD testing(assessment thresholds). These have been developed for Europe,Canada, Germany, Japan, the UK and US [11,30–34]. There have beentwo approaches to the development of guidelines based on fractureprobability. The first is to ‘translate’ current practice in the light ofFRAX®, and the second has been to determine the threshold fractureprobability at which intervention becomes cost-effective. The latterapproach is discussed in the section on health economic use ofFRAX®.
The UK guidance for the identification of individuals at highfracture risk developed by the National Osteoporosis Guideline Group(NOGG) is an example of the translation of existing guidance providedby the Royal College of Physicians (RCP) [7–9] into probability basedassessment [35]. As with the RCP guidance, the strategy is based onopportunistic case finding where physicians are alerted to thepossibility of increased fracture risk by the presence of clinical riskfactors (CRFs). The clinical risk factors used differ somewhat fromthose of the RCP, and comprised those used in the FRAX® algorithmstogether with low BMI (b19 kg/m2).
The RCP guidance indicates that women with a prior fragilityfracture may be considered for intervention without the necessity fora BMD test, and the management of women over the age of 50 yearson this basis has been shown to be cost-effective [36]. For this reason,the intervention threshold set by NOGGwas at the fracture probabilityequivalent towomenwith a prior fragility fracturewithout knowledgeof BMD [35]. The same intervention threshold was applied to men,
since the effectiveness of intervention in men is broadly similar to thatin women for equivalent risk [37].
In addition to an intervention threshold, assessment thresholdsfor the use of BMD testing were devised. The concept of assessmentthresholds is illustrated in the management algorithm given in Fig. 4[3]. The management process begins with the assessment of fractureprobability and the categorization of fracture risk on the basis ofage, sex, BMI and the clinical risk factors. On this information alone,some patients at high risk may be offered treatment withoutrecourse to BMD testing. As noted, many guidelines [7–13]recommend treatment in the absence of information on BMD inwomen with a previous fragility fracture. Many physicians wouldalso perform a BMD test, but frequently this is for reasons otherthan to decide on intervention for example, as a baseline to monitortreatment. There will be other instances where the probability willbe so low that a decision not to treat can be made without BMD. Anexample might be the well woman at menopause with no clinicalrisk factors. Thus not all individuals require a BMD test. The size ofthe intermediate category in Fig. 4 will vary in different countries,but a pragmatic strategy was used by NOGG because of the limitedfacilities for BMD testing in the UK [38].
The NOGG management strategy requires consideration of twoadditional thresholds.
• a threshold probability below which neither treatment nor aBMD test should be considered (lower assessment threshold)
• a threshold probability above which treatment may be recom-mended irrespective of BMD (upper assessment threshold)
The lower assessment threshold was set to exclude a requirementfor BMD testing in women of average BMI (24 kg/m2) with weak orno clinical risk factors, as given in the RCP and European guidelines[7–12]. The upper threshold was chosen to minimise the probabilitythat a patient characterised to be at high risk on the basis of clinicalrisk factors alone would be reclassified to be at low risk with addi-tional information on BMD [39]. The upper assessment threshold wasset at 1.2 times the intervention threshold. The value 1.2 is arbitraryand determines the number of men and women who would beeligible for a BMD test. With this value, BMD tests would be under-taken in 15–30% of individuals depending on age (excluding thosewith a prior fracture) [35]. Scanning this proportion of the populationmaximises the positive predictive value of the assessment, at leastwith respect to hip fracture [40].
Fig. 5. Management chart for osteoporosis. The darker shaded area in the left hand panel shows the limits of fracture probabilities for the assessment of BMD. The right hand panelgives the intervention threshold [35, with permission from Springer].
Table 1The proportion of women in the UK potentially eligible for treatment. The totalscomprise women with a prior fracture and those with another clinical risk factor (CRF)in whom fracture probability exceeds the intervention threshold [35, with permissionfrom Springer]
Age(years)
Interventionthreshold (%)
Prior fracture%
N1 CRFa
%Total%
50–54 7.5 17.1 6.5 23.655–59 10 21.8 7.0 28.860–64 12.5 26.0 9.3 35.365–69 15 29.0 9.2 38.270–74 20 30.8 7.0 37.875–79 25 35.9 5.5 41.180–84 30 41.3 5.2 46.5
a Excluding prior fracture and probability above the intervention threshold aftertesting with BMD.
738 J.A. Kanis et al. / Bone 44 (2009) 734–743
The management algorithm is shown in Fig. 5 and summarisedbelow [35];
1. Postmenopausal women with a prior fragility fractureshould be considered for treatment; BMD measurementmay sometimes be appropriate, particularly in youngerpostmenopausal women. Men with a prior fragility fractureshould be referred for BMD.
2. Men aged 50 years or more and all postmenopausal womenwith a WHO risk factor or a BMI b19 kg/m2 should havefracture probability assessed using the FRAX® tool withoutmeasurement of BMD.
3. Individuals with probabilities of a major osteoporotic fracturebelow the lower assessment threshold given in Fig. 5 can bereassured. A further assessment is recommended in 5 years orless depending on the clinical context.
4. Individuals with probabilities of a major osteoporotic fractureabove the upper assessment threshold given in Fig. 5 or withprobabilities of a hip fracture above the upper limit (not shown)can be treated without recourse to BMD testing.
5. Individuals with probabilities of a major osteoporotic fracturewithin the limits of the assessment thresholds given in Fig. 5 andwith probabilities of a hip fracture below the limit (not shown)should have a BMD test and probabilities recomputed. If recom-puted probabilities exceed the treatment threshold, interven-tion should be considered. Where probabilities fall below thetreatment threshold, a further assessment is recommended in5 years or less depending on the clinical context.
The transformation of these recommendations into a format that isreadily useable by primary care physicians is not straightforward,given that intervention and assessment thresholds vary by age, andare ideally based on the probability of a hip fracture and a majorfracture. The most accurate approach is a computer program and isavailable through http://www.shef.ac.uk/NOGG/index.htm ordirectly from the FRAX® site (http://www.shef.ac.uk/FRAX/index.htm). Simplified paper charts are also available [41].
Women potentially eligible for treatment using the NOGGalgorithm include those with a prior fracture (irrespective of fractureprobability) and those with a CRF in whom fracture probability layabove the intervention threshold for age. The numbers eligible fortreatment are shown in Table 1. The proportion of the female popu-lation potentially treated varied from 24% to 47%, depending on age.
The proportion of women potentially eligible for treatment increasedwith age, despite the apparently more stringent intervention thresh-old. More women with a prior fracture would be treated than womenwith any other CRF [35].
A similar approach has been developed in Europe by theEuropean Society for Clinical and Economic Aspects of Osteoporosisand Osteoarthritis (ESCEO), and indeed the same interventionthreshold is used. The difference lies in the upper assessmentthreshold in that it is recommended that all patients with proba-bilities above the lower assessment threshold should receive a BMDtest [11].
A translational approach to guideline development has also beenexamined in Japan [31]. In Japan, the criteria for the diagnosis ofosteoporosis prepared by the Japanese Society for Bone and MineralResearch [42] are based on BMD measurements expressed as apercentages of the young adult mean (YAM) for women. In patientswith no prior fragility fracture a diagnosis of osteoporosis is madeand treatment recommended where the BMD is less than 70% ofYAM. In patients with a previous fracture, osteoporosis is diagnosedand treatment recommended where the BMD is less than 80% ofYAM. In order to compare intervention thresholds using YAM withprobabilities derived from the FRAX® algorithm, T-score equivalentswere used. The T-score equivalent to 70% and 80% of YAM forJapanese people is −2.7 SD and −1.8 SD respectively, using theNHANES III reference for BMD at the femoral neck in Caucasianwomen aged 20–29 years [4]. The probabilities equivalent to the
Fig. 6. Ten year probability of a major osteoporotic fracture inwomen from Japan by ageat the threshold used to start treatment using the Japanese guidelines. The shaded areagives the limits of the intervention threshold in the UK.
739J.A. Kanis et al. / Bone 44 (2009) 734–743
current intervention thresholds vary with age and are shown in Fig. 6against the backdrop of the European and UK intervention thresholdsand show a high degree of concordance with intervention thresholdsdeveloped for the UK and Europe. Other countries that have deve-loped or are developing translational approaches include Canada [33],Poland [43], Hong Kong [44], Switzerland [45] and Germany [30].
These translational approaches from existing treatment guidelinesare characterised by an intervention threshold that increasesprogressively with age. The major reason for this is that the sourceguidelines took little or no account of age. In the UK, for example,intervention is recommended inwomenwith a prior fragility fracture,irrespective of age. Since age is an important independent determi-nant of fracture risk, the fracture probability of an individual with aprior fracture is higher at the age of 70 years than at the age of50 years. This age-dependent increase in the intervention threshold isnot found when intervention thresholds are derived from healtheconomic analyses and are discussed in the section on the healtheconomic use of FRAX®.
Fig. 7. Relationship between 10-year probabilities of a major osteoporotic fracture and the ef[47with permission from Springer]. The horizontal line represents the hazard ratio of 1. Probarisk factors combined with femoral neck BMD (right panel).
Assessment of drug efficacy
Revised guidelines on the evaluation of medicinal products in thetreatment of primary osteoporosis have been developed by theCommittee for Medicinal Products for Human Use (CHMP) andcame into effect at the end of May 2007 [46]. A major departure fromprevious guidance is that there is no longer any distinction betweenprevention and treatment, but an emphasis on the study of patients atrisk from fracture. The preferred metric for expressing risk is the ten-year probability of fracture, in line with the recommendations of theWHO [2,3]. Suggested probabilities as inclusion criteria into phase IIItrials are given as 15–20% for spine fracture, 5–7.5% for hip fractureand 10–15% for major non-vertebral fractures. These are intended toapproximate the fracture risks in previous phase III studies.
As a consequence FRAX® has been applied to several phase IIIstudies in order to determine the enrolment characteristics ofpatients. A development has been to examine whether patients cha-racterised on the basis of fracture probability respond to treatment.An example is provided in a 3-year prospective, randomized, placebo-controlled trial of oral clodronate [47]. Women aged 75 years or moreliving in the general community, identified from general practiceregisters, were given 800 mg oral clodronate or matching placebodaily over three years. Baseline risk factors, including age, BMI, priorfracture, glucocorticoid use, rheumatoid arthritis or other secondarycauses of osteoporosis, smoking and maternal history of hip fracture,were used to compute the 10-year probability of a major osteoporoticfracture. Femoral neck BMD was also measured at entry. The mainoutcome of this analysis was the interaction between fractureprobability and treatment efficacy examined by Poisson regression.Greater fracture reduction was seen at higher fracture probabilities,with or without the use of BMD (Fig. 7) [47]. Efficacy was evident atfracture probabilities that exceeded 20%. The effect was seenirrespective of whether BMD was used in the calculation of fractureprobability.
Similar findings have been found for bazedoxifene [48] and otherclasses of agent [49]. In the phase III study of bazedoxifene, the agentsignificantly decreased the risk of vertebral fractures in postmeno-pausal women [50]. No significant effect was noted on the risk ofclinical fractures, but fracture risk reductionwas reported in a post hocsubgroup analysis in a high risk group categorised on the basis of BMDand prior fracture. The characterisation of the enrolled populationwith FRAX® permitted the re-evaluation of the efficacy of baze-
ficacy of clodronate to reduce fracture risk (hazard ratio with 95% confidence intervals)bilities shown are calculatedwith clinical risk factors alone (left panel) andwith clinical
Fig. 8. Hazard ratio between treatments (bazedoxifene versus placebo) with 95% confidence intervals according to values of 10-year probability of a major osteoporotic fracturecalculated with BMD [48, with permission from Elsevier].
740 J.A. Kanis et al. / Bone 44 (2009) 734–743
doxifene on fracture outcomes avoiding subgroup analysis byexamining the efficacy of intervention as a function of fracture risk.As in the case of clodronate, hazard ratios for the effect of baze-doxifene on all clinical fractures decreased with increasing fractureprobability (Fig. 8). In patients with 10-year fracture probabilities at orabove 16%, bazedoxifene was associated with a significant decrease inthe risk of all clinical fractures. Hazard ratios for the effect ofbazedoxifene on morphometric vertebral fractures also decreasedwith increasing fracture probability (see Fig. 8). In patients with 10-year fracture probabilities above 6.9%, bazedoxifene was associatedwith a significant decrease in the risk of morphometric vertebralfractures. At equivalent fracture probability percentiles, the treatmenteffect of bazedoxifene was greater on vertebral fracture risk than onthe risk of all clinical fractures. For example, at the 90th percentile ofFRAX® probability, bazedoxifene was associated with a relative risk
Table 2Cost-effectiveness analysis by age, sex and race for men and women at average risk and relatreference 60, with permission from Springer]
Age (years) General population Threshold for cost-effectiveness
Fracture probabilitya ICER ($000/QALY)b Relative riskc Fracture probabi
White female50 0.72 380,.7 3.5 2.555 1.18 222.8 2.3 2.860 1.79 139.6 1.6 3.065 2.24 88.4 1.3 2.870 4.66 44.2 0.9 4.075 9.80 cs 0.4 4.480 13.52 cs 0.3 4.085 12.96 cs 0.3 3.3
Average=3.4Black female
50 0.31 932.3 8.1 2.555 0.5 576.2 5.4 2.760 0.75 392.8 3.8 2.965 0.92 268.1 2.9 2.770 1.90 166.7 2.0 3.875 4.08 63.1 1.0 4.280 5.85 28.6 0.7 4.085 5.81 22.2 0.6 3.4
Average=3.3
cs, cost saving; ICER, incremental cost effectiveness ratio.a Average10-year probability of hip fracture (%) in the general population.b Cost-effectiveness of treating individuals at average risk.c The relative risk for hip fracture at which intervention becomes cost-effective.d 10-year probability of hip fracture (%) at which intervention becomes cost-effective.
reduction of 33% (95% CI=7–51%) for all clinical fractures and 51%reduction (95% CI=21–69%) for morphometric vertebral fractures[48].
These results, if confirmed in other clinical settings, have anumber of important implications. First, they dispel a minority viewthat patients identified on the basis of clinical risk factors withFRAX® would not respond to pharmacologic interventions. Indeedboth studies showed that high FRAX® probabilities were associatedwith efficacy, even when BMD was not used to characterise risk.Second, they support the views of the European regulatory agencythat treatments should be developed preferentially to men andwomen at high fracture risk. Third, the finding of greater efficacy athigher fracture probabilities has important implications for healthtechnology assessments and challenges the current meta-analyticapproach. Finally, since treatments directed to high risk patients
ive risk and absolute risk thresholds in the United States [Data extracted from Table 2 of
General population Threshold for cost-effectiveness
lityd Fracture probabilitya ICER $000/QALYb Relative riskc Fracture probabilityd
White male0.34 839.1 6.9 2.40.52 689.4 8.1 4.21.21 367.7 3.4 4.11.53 215.2 2.3 3.52.61 148,.3 1.9 4.83.97 58.9 1.0 3.95.44 36.1 0.7 4.06.9 13.6 0.4 3.1
Average=3.8Black male0.18 1552.3 13.1 2.40.26 1282.8 11.5 3.00.59 700.9 6.5 3.80.75 425.7 4.3 3.31.25 306.9 3.5 4.41.96 138.5 1.9 3.62.72 92.1 1.4 3.73.68 44.6 0.8 3.0Average=3.4
Fig. 9. Correlation between the probability of fracture and cost-effectiveness at the ageof 50 years in women (BMI set to 26 kg/m2). The upper panel shows the ten-yearprobability of hip fracture and the lower panel, the probability of a major osteoporoticfracture. Each point represents a particular combination of BMD and clinical risk factors[35, with permission from Springer]. The horizontal line denotes the threshold for cost-effectiveness (a willingness to pay of £20,000/QALY gained).
Table 3Ten-year probability of a major osteoporotic fracture with 95 percent confidenceintervals (CI) at which it becomes cost-effective to intervene with generic alendronateat a willingness to pay of £20,000/QALY [35, with permission from Springer]
Age (years) 10-year probability of osteoporotic fracture with BMD (%)
Mean Lower CI Upper CI
50 5.6 4.8 6.855 7.6 6.3 8.960 8.4 7.1 9.665 6.1 4.9 7.270 5.0 4.2 5.975 7.3 6.6 8.180 8.3 7.8 8.8
741J.A. Kanis et al. / Bone 44 (2009) 734–743
improve the budget impact, greater efficacy in the higher risk groupswill improve still further the budget impact and the cost-effective-ness of intervention.
Health economic applications of FRAX®
In order to justify resource allocation, it is becoming increasinglyimportant to determine the cost-effectiveness of intervention. Thepreferred approach is cost utility analysis. This integrates the deathsand disability associated with the multiple outcomes by measuringquality of life adjusted years (QALYs). In order to estimate QALYs, eachyear of life is valued according to its utility to the patient that rangesfrom 0, the least desirable health state, to 1, perfect health. In theanalyses reviewed here, costs and effects are measured as cost perQALY gained. The effects of no treatment are compared with an indexintervention.
In the context of osteoporosis and fracture risk, the interventionthreshold that is relevant for clinicians can be defined as theprobability of fracture at which intervention becomes cost-effective.The threshold of cost-effectiveness at which intervention isconsidered appropriate (willingness to pay; WTP) is somewhatarbitrary. One suggested benchmark for establishing WTP thresholdsis to bench mark this to the Gross Domestic Product (GDP)/capita.By using the GDP/capita, differences in ability to pay betweencountries would be acknowledged. The WTP threshold has beensuggested to be somewhere between 1 and 3 times GDP/capita [51].An appropriate threshold would be approximately twice the GDP/capita when considering direct costs to the health service [3,52].
This is equivalent to about £30,000/QALY and close to that used byNICE and others in most appraisals in the UK [53–55]. In the contextof prevention of osteoporosis, NICE used a £20,000/QALY threshold[55].
Several studies have examined intervention thresholds in terms offracture probability [35,56–60]. The majority have expressed resultsas the 10-year probability of hip fracture at which treatment is cost-effective, i.e. using hip fracture as a metric to measure the impact ofall fracture outcomes. The most recent assessment using thisapproach has come from the National Osteoporosis Foundation(NOF) who have updated pre-existing clinical practice guidelineswith a further health economic analysis [60] to reflect the growinginternational consensus that intervention thresholds for osteoporosistreatment should be determined on the basis of fracture probability[2,3,24].
A Markov-cohort model of annual United States age-specificincidence of hip, wrist, clinical spine and other fractures, costs(2005 US dollars), and quality-adjusted life years (QALYs) was used toassess the cost-effectiveness of osteoporosis treatment ($600/yr drugcost for 5 years with 35% fracture reduction). A five-year course oftreatment with a bisphosphonate-like therapy was modelled. Rando-mized clinical trial evidence for the effectiveness of treatment inreducing fracture risk varies by fracture site and the populationstudied [61]. For comparability with other studies, a fracture reductionof 35% (relative risk [RR]=0.65) was assumed [57–59]. Treatmenteffectiveness was assumed towanewhen treatment was stoppedwithan offset time of 5 years as recommended elsewhere [62]. Thethreshold for cost-effectiveness was set at $60,000 per QALY gained.
The cost-effectiveness analysis of treatment relative to no inter-vention among the population at average risk populations rangedfrom over $380,000 per QALY gained in 50-year-old white women tobeing cost-saving in 75-year-old white women (Table 2). Cost-effectiveness was poorer in white men, and poorer still in blackwomen andmen due to the lower risk of fracture in these populations.Although relative risks differed between race groups for each gender,the intervention thresholds by race and sex were relatively constant(see Table 2). In white men and women, treatment became cost-effective at a hip fracture probability of 3.4% and 3.8%, respectively.The corresponding probabilities in blacks were 3.3% and 3.4%. On thisbasis, the NOF chose a 10-year hip fracture probability of 3% as anintervention threshold.
A somewhat different approach has been used in the UK. Fractureprobabilities were computed using the FRAX® tool calibrated to theepidemiology of fracture and death in the UK [1]. The costeffectiveness of generic alendronate in the UK was taken from anindependent paper [36]. The cost of alendronate was set at £95 peryear, which is now a conservative figure since the current price hassubsequently fallen to £24.
In order to determine whether to express intervention thresholdsas the 10-year probability of hip fracture or of a major osteoporoticfracture, the relationship between fracture probabilities expressedusing either outcome and the cost-effectiveness were examined for all
Table 4Ten year probabilities (mean and 95% confidence intervals; CI) of a major osteoporoticfracture (%) by age at or abovewhich treatmentwith risedronate becomes cost-effective[63]
Age 10-year probability of osteoporotic fracture (%) with BMD at a WTP of
(years) £20 000/QALY £30 000/QALY
Probability 95% CI Probability 95% CI
50 17.1 12.1–29.4 10.0 8.6–14.955 19.8 16.1–29.3 14.3 10.6–18.060 23.0 17.5–33.2 16.5 12.6–20.865 18.0 14.4–23.9 11.9 9.4–15.470 16.1 12.9–19.2 9.9 8.9–12.975 17.9 13.8–23.3 13.3 9.7–17.480 18.3 14.7–24.3 15.2 11.6–18.9
742 J.A. Kanis et al. / Bone 44 (2009) 734–743
possible combinations of CRFs at BMD T-scores between 0 and−3.5 SD in 0.5 SD steps (512 combinations) with a BMI set to 26 kg/m2 [35]. Thus, this was not a population simulation, but a display of allpossible combinations.
The relationship of the two probability outputs to cost-effective-ness are shown in Fig. 9. It is evident that a threshold value for hipfracture probability (say 2%) would encompass a very wide range ofcost effectiveness. The relationship of cost-effectiveness with majorosteoporotic fracture probability is, however, more favourable since, atprobability values that might be appropriate as a threshold, the rangeof cost-effectiveness was much less than in the case of hip fractureprobability. For this reason, the major index used to assess interven-tion thresholds in the UK has been the 10-year probability of a majorosteoporotic fracture.
The point estimates for the correlations shown in Fig. 9 permitan estimate of the mean fracture probability for any willingness topay as shown in Table 3 for a WTP of £20,000 [35]. There wasrather little difference in the threshold at different ages with amean value of 7.0%. At a WTP of £10,000 the mean probabilitythreshold was 11.7% and at a WTP of £30,000 was 3.6%. Thus, witha WTP of £20,000, any recommendations for intervention shouldensure that individuals have a fracture probability that exceeds 7%.As can be seen from Fig. 5, all treatment scenarios developed forpostmenopausal women or men aged 50 years or more in the UKwere cost-effective.
The low cost of generic alendronate, and the improvement incost-effectiveness derived thereby, allowed an assessment of thecost-effectiveness of clinical guidelines, whereas intervention thres-holds in the US were limited by cost-effectiveness. The same is nottrue of more expensive interventions. An example is provided inTable 4 for risedronate [63]. At a WTP of £20,000 per QALY, treat-ment with risedronate was cost-effective at a probability thresholdof 19% compared with a threshold of 7% with generic alendronate.
Comparison of intervention thresholds based on cost-effectivenessand those derived from the translational approach from existingguidelines show important differences. Whereas the interventionthreshold rises steeply with age in the latter approach, there is nomarked effect of age in thresholds derived from health economicanalysis (see Tables 3 and 4). The difference arises because healtheconomic analysis takes account of the age-dependent fracture risk,whereas many established guidelines based on clinical imperatives donot. There is no right or wrong methodological approach and theapproach used is dependent on the socioeconomic setting. In the USfor example, screening of men and women (at the ages of 70 and65 years, respectively) with BMD is recommended. In the UK, BMDtests are reserved for individuals with clinical risk factors for fracture.In addition, intervention costs are much lower than in the US. Anintervention threshold based on cost-utility alone in the UK wouldpermit intervention in very many individuals in whom this wasconsidered clinically inappropriate, and many more than would bedeemed eligible in the US.
Because intervention thresholds are in part based on cost-effectiveness, and in part on clinical considerations, the examplesabove are not necessarily applicable to other countries, where theWTP and/or the costs of osteoporosis or intervention may differ.Notwithstanding, a health economic analysis in a ‘European setting’[11] suggests that similar thresholds could apply to several Europeancountries. In the event, intervention thresholds need to be deter-mined on a country by country basis.
References
[1] Kanis JA, Johnell O, Oden A, Johansson H, McCloskey EV. FRAX™ and theassessment of fracture probability in men and women from the UK. Osteoporos Int2008;19:385–97.
[2] World Health Organization. Assessment of osteoporosis at the primary health carelevel. Geneva: WHO; 2007 (www.who.int/chp/topics/rheumatic/en/index.html).
[3] Kanis JA, on behalf of the World Health Organization Scientific Group. Assessmentof osteoporosis at the primary health-care level. Technical report. University ofSheffield, UK: WHO Collaborating Centre; 2008.
[4] Looker AC, Wahner HW, DunnWL, Calvo MS, Harris TB, Heyse SP. Updated data onproximal femur bone mineral levels of US adults. Osteoporos Int 1998;8:468–86.
[5] Kanis JA, Oden A, Johnell O, Johansson H, De Laet C, Brown J, et al. The use of clinicalrisk factors enhances the performance of BMD in the prediction of hip andosteoporotic fractures in men and women. Osteoporos Int 2007;18:1033–46.
[6] Kanis JA, Johnell O, De Laet C, Jonsson B, Oden A, Oglesby A. Internationalvariations in hip fracture probabilities: implications for risk assessment. J BoneMiner Res 2002;17:1237–44.
[7] Royal College of Physicians. Osteoporosis: clinical guidelines for the preventionand treatment. London: Royal College of Physicians; 1999.
[8] Royal College of Physicians and Bone and Tooth Society of Great Britain. Update onpharmacological interventions and an algorithm for management. London UK:Royal College of Physicians; 2000.
[9] Royal College of Physicians. Glucocorticoid-induced osteoporosis. Guidelines onprevention and treatment. Bone and Tooth Society of Great Britain, NationalOsteoporosis Society and Royal College of Physicians. London UK: Royal College ofPhysicians; 2002.
[10] Kanis JA, Delmas P, Burckhardt P, Cooper C, Torgerson D, on behalf of the EuropeanFoundation for Osteoporosis and Bone Disease. Guidelines for diagnosis andmanagement of osteoporosis. Osteoporos Int 1997;7:390–406.
[11] Kanis JA, Burlet N, Cooper C, Delmas PD, Reginster JY, Borgstrom F, et al, on behalfof the European Society for Clinical and Economic Aspects of Osteoporosis andOsteoarthritis (ESCEO). European guidance for the diagnosis and management ofosteoporosis in postmenopausal women. Osteoporos Int 2008;19:399–428.
[12] European Community. Report on osteoporosis in the European Community.Strasbourg: EC; 1998.
[13] National Osteoporosis Foundation (NOF). Physician's guide to prevention andtreatment of osteoporosis. Washington, DC: National Osteoporosis Foundation;2003.
[14] Kanis JA. Diagnosis of osteoporosis and assessment of fracture risk. Lancet 2002;359:1929–36.
[15] De Laet C, Oden A, Johansson H, Johnell O, Jonsson B, Kanis JA. The impact of theuse of multiple risk indicators for fracture on case-finding strategies: amathematical approach. Osteoporos Int 2005;16:313–8.
[16] Kanis JA Johnell O, Oden A, Dawson A, De Laet C, Jonsson B. Ten year probabilities ofosteoporotic fractures according to BMD and diagnostic thresholds. Osteoporos Int2001;12:989–95.
[17] Hui SL, Slemenda CW, Johnston CC. Age and bonemass as predictors of fracture in aprospective study. J Clin Invest 1998;81:1804–9.
[18] Van Staa TP, Leufkens HGM, Abenhaim L, Zhang B, Cooper C. Fracture and oralcorticosteroids: relationship to daily and cumulative dose. Rheumatol 2000;39:1383–9.
[19] Kanis JA, Johansson H, Johnell O, Oden A, De Laet C, Eisman JA, et al. Alcohol intakeas a risk factor for fracture. Osteoporos Int 2005;16:737–42.
[20] Kanis JA, Johnell O, Oden A, Johansson H, De Laet C, Eisman JA, et al. Smoking andfracture risk: a meta-analysis. Osteoporos Int 2005;16:155–62.
[21] Delmas PD, Genant HK, Crans JL, Stock GG, Wong M, Siris E, et al. Severity ofprevalent vertebral fractures and the risk of subsequent vertebral and non-vertebral fractures: results from the MORE trial. Bone 2003;33:522–32.
[22] Lunt M, O'Neill TW, Felsenberg D, Reeve J, Kanis JA, Cooper C, et al. Characteristicsof a prevalent vertebral deformity predict subsequent vertebral fracture: resultsfrom the European Prospective Osteoporosis Study (EPOS). Bone 2003;33:505–13.
[23] Kanis JA, McCloskey EV, Johansson H, Oden A, Melton LJ, Khaltaev N. A referencestandard for the description of osteoporosis. Bone 2008;42:467–75.
[24] Kanis JA, Black D, Cooper C, Dargent P, Dawson-Hughes B, De Laet C, et al. A newapproach to the development of assessment guidelines for osteoporosis.Osteoporos Int 2002;13:527–36.
[25] Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bonemineral density predict occurrence of osteoporotic fractures. Br Med J 1996;312:1254–9.
[26] Glüer CC, for the International Quantitative Ultrasound Consensus Group.Quantitative ultrasound techniques for the assessment of osteoporosis: expertagreement on current status. J Bone Miner Res 1997;12:1280–8.
743J.A. Kanis et al. / Bone 44 (2009) 734–743
[27] Genant HK, Engelke K, Prevrhal S. Advanced CT bone imaging in osteoporosis.Rheumatol (Oxf) 2008;47(Suppl 4):iv9–v16.
[28] Delmas PD, Eastell R, Garnero P, Seibel MJ, Stepan J, Committee of ScientificAdvisors of the International Osteoporosis Foundation. The use of biochemicalmarkers of bone turnover in osteoporosis. Osteoporos Int 2000;11(Suppl 6):S2–S17.
[29] Kanis JA, Johansson H, Oden O, Johnell A, de Laet C, Melton III LJ, et al. A meta-analysis of prior corticosteroid use and fracture risk. J Bone Miner Res 2004;19:893–9.
[30] Kurth AA, Pfeilschifter J. Diagnosis and treatment of postmenopausal osteoporosisand osteoporosis in men. German Guidelines Update 2006. Orthopade 2007;36:683–90. German.
[31] Fujiwara S, Nakamura T, Orimo H, Hosoi T, Gorai I, Oden A, et al. Development andapplication of a Japanese model of the WHO fracture risk assessment tool(FRAX™). Osteoporos Int 2008;19:429–48.
[32] Siris E, Delmas PD. Assessment of 10-year absolute risk: a new paradigm withworldwide application. Osteoporos Int 2008;19:383–4.
[33] Siminoski K, Leslie WD, Frame H, Hodsman A, Josse RG, Khan A, et al.Recommendations for bone mineral density reporting in Canada: a shift toabsolute fracture risk assessment. J Clin Densitom 2007;10:120–3.
[34] Dawson-Hughes B, Tosteson AN, Melton III LJ, Baim S, Favus MJ, Khosla S, et al.Implications of absolute fracture risk assessment for osteoporosis practiceguidelines in the USA. Osteoporos Int 2008;19:449–58.
[35] Kanis JA, McCloskey EV, Johansson H, Strom O, Borgstrom F, Oden A, andthe National Osteoporosis Guideline Group. Case finding for the manage-ment of osteoporosis with FRAX® — assessment and intervention thresh-olds for the UK. Osteoporos Int 2008; 19: 1395–1408. Erratum in pressOsteoporos Int. 2008 Nov 20. [Electronic publication ahead of print] PMID:19020922.
[36] Kanis JA, Adams J, Borgström F, Cooper C, Jönsson B, Preedy D, et al. The cost-effectiveness of alendronate in the management of osteoporosis. Bone 2008;42:4–15.
[37] Kanis JA, Stevenson M, McCloskey EV, Davis S, Lloyd-Jones M. Glucocorticoid-induced osteoporosis: a systematic review and cost-utility. analysis. HealthTechnol Assess 2007;11:1–256.
[38] Kanis JA, Johnell O, on behalf of the Committee of Scientific Advisors of theInternational Osteoporosis Foundation. Requirements for DXA for the manage-ment of osteoporosis in Europe. Osteoporos Int 2005;16:229–38.
[39] Johansson H, Oden A, Johnell O, Jonsson B, De Laet C, Oglesby A, et al.Optimisation of BMD measurements to identify high risk groups for treatment —a test analysis. J Bone Miner Res 2004;19:906–13.
[40] Johansson H, Kanis JA, Oden A, Johnell O, McCloskey E. BMD, clinical risk factorsand their combination for hip fracture prevention. Osteoporos Int in press.
[41] National Osteoporosis Guideline Group on behalf of the Bone ResearchSociety, British Geriatrics Society, British Society of Rheumatology, Society ofEndocrinology, British Orthopaedic Association, Primary Care RheumatologySociety, Osteoporosis 2000 and Osteoporosis Dorset. Osteoporosis: clinicalguideline for prevention and treatment. Sheffield, UK: University of SheffieldPress; 2008.
[42] Orimo H, Sugioka Y, Fukunaga M, et al. Diagnostic criteria of primary osteoporosis.J Bone Miner Metab 1998;16:139–50.
[43] Czerwiński E, Badurski JE, Marcinowska-Suchowierska E, Osieleniec J. Currentunderstanding of osteoporosis according to the position of the World HealthOrganization (WHO) and International Osteoporosis Foundation. Ortop TraumatolRehabil 2007;9:337–56.
[44] Tsang SWY, Kung AWC, Kanis JA, Johansson H, Oden A. Ten-year fractureprobability in Hong Kong southern Chinese according to age and BMD femoralneck T-scores. Osteoporos Int. 2009, in press.
[45] Lippuner K, Johansson H, Kanis JA, Rizzoli R. Remaining lifetime and absolute 10-year probabilities of osteoporotic fracture in Swiss men and women. OsteoporosInt 2008 Oct 31 [Electronic publication ahead of print] PMID: 18974918.
[46] Committee for Medicinal Products for Human Use (CHMP). Guideline on theevaluation of medicinal products in the treatment of primary osteoporosis. RefCPMP/EWP/552/95Rev.2. London, CHMP. Nov 2006.
[47] McCloskey E, Johansson H, Oden A, Aropuu S, Jalava T, Kanis JA. Efficacy ofclodronate on fracture risk in women selected by 10-year fracture probability.Osteoporos Int in press.
[48] Kanis JA, Johansson H, Oden A, McCloskey EV. (2008) Bazedoxifene reducesvertebral and clinical fractures in postmenopausal women at high risk assessedwith FRAX®. Bone in press.
[49] Kanis JA. Data held in confidence and presently embargoed. Personalcommunication.
[50] Silverman SL, Christiansen C, Genant HK, Vukicevic S, Zanchetta JR, de Villiers TJ,et al. Efficacy of bazedoxifene in reducing new vertebral fracture risk inpostmenopausal women with osteoporosis: results from a 3-year, randomized,placebo- and active-controlled clinical trial. J Bone Miner Res 2008 Jul 29[Electronic Publication ahead of print]PMID: 18665787.
[51] Eichler HG, Kong SX, Gerth WC, Mavros P, Jonsson B. Use of cost-effectivenessanalysis in health-care resource allocation decision-making: how are cost-effectiveness thresholds expected to emerge? Value Health 2004;7:518–28.
[52] Kanis JA, Jonsson B. Economic evaluation of interventions for osteoporosis.Osteoporos Int 2002;13:765–7.
[53] Kanis JA, Brazier JE, Stevenson M, Calvert NW, Lloyd Jones M. Treatment ofestablished osteoporosis: a systematic review and cost-utility analysis. HealthTechnol Assess 2002;6:1–146.
[54] National Institute for Health and Clinical Excellence. Final appraisal determination.Alendronate, etidronate, risedronate, raloxifene, strontium ranelate and teripara-tide for the secondary prevention of osteoporotic fragility fractures in post-menopausal women. London: NICE; 2008.
[55] National Institute for Health and Clinical Excellence. Final appraisal determination.Alendronate, etidronate, risedronate, raloxifene and strontium ranelate for theprimary prevention of osteoporotic fragility fractures in postmenopausal women.London: NICE ; 2008.
[56] Kanis JA, Oden A, Johnell O, Jonsson B, de Laet C, Dawson A. The burden ofosteoporotic fractures: a method for setting intervention thresholds. OsteoporosInt 2001;12:417–27.
[57] Kanis JA, Borgstrom F, Zethraeus N, Johnell O, Oden A, Jonsson B. Interventionthresholds for osteoporosis in the UK. Bone 2005;36:22–32.
[58] Kanis JA, Johnell O, Oden A, Borgstrom F, Johansson H, De Laet C, et al. Interventionthresholds for osteoporosis in men and women: a study based on data fromSweden. Osteoporos Int 2005;16:6–14.
[59] Borgstrom F, Johnell O, Kanis JA, Jonsson B, Rehnberg D. At what hip fracture risk isit cost-effective to treat? International intervention thresholds for the treatment ofosteoporosis. Osteoporos Int 2006;17:1459–71.
[60] Tosteson AN, Melton III LJ, Dawson-Hughes B, Baim S, Favus MJ, Khosla S, et al.National Osteoporosis Foundation Guide Committee. Cost-effective osteoporosistreatment thresholds: the United States perspective. Osteoporos Int 2008;19:437–47.
[61] Cranney A,Wells G,Willan A, Griffith L, Zytaruk N, Robinson V, et al. Meta-analysesof therapies for postmenopausal osteoporosis. II. Meta-analysis of alendronate forthe treatment of postmenopausal women. Endocr Rev 2002;23:508–16.
[62] Tosteson AN, Jonsson B, Grima DT, O'Brien BJ, Black DM, Adachi JD. Challenges formodel-based economic evaluations of postmenopausal osteoporosis interven-tions. Osteoporos Int 2001;12:849–57.
[63] Borgström F, Strom O, Coelho J, Johansson H, Oden A, McCloskey E, et al. The cost-effectiveness of risedronate in the UK for the management of osteoporosis. 2008Osteoporos Int in press.
