We are now fortunate to have a diverse menu of therapeutic options for osteoporosis; however, as Kanis et al.(1) has pointed out, we cannot afford to treat everyone. Available treatments need to be targeted at those individuals with a high risk of fracture, avoiding unnecessary treatment for individuals at low risk of fracture.

The cornerstone of case finding of patients at risk for fracture has been use of BMD measurement using DXA using T-scores as a diagnostic threshold.(2) DXA measurements predict fracture with an increase in fracture risk of ∼1.5- to 2-fold with each SD of change; the increase in risk is called the gradient of risk. However, BMD measurement is not sufficient to identify all patients at risk. BMD measurement is specific but not sensitive. In NORA, a large community study using peripheral densitometry(3) and in the Study of Osteoporotic Fractures(4) and Rotterdam(5) using central measurement techniques, about one half the patients who fractured were not osteoporotic. Thus, if treatment decisions are based on BMD alone, about one half of postmenopausal women who fracture would not be considered for treatment before fracture occurs. Furthermore, in many parts of the world, access to BMD measurement is limited. Even when available, only about one quarter of all women >65 yr of age in the United States have had a BMD measurement.(6) Clinicians have also used patient history to help them with case finding by use of clinical risk factors such as prior fracture. However, there has been a need to develop means to integrate multiple risk factors with or without BMD to increase the gradient of risk. As a result, risk factor assessment scores have been developed that can help healthcare providers identify those who would benefit from BMD testing and those who are at high risk and would benefit from therapy.

Several risk assessment scoring systems have been developed based on both BMD and risk factors for fracture. Among these are risk algorithms developed by the Study of Osteoporotic Fracture,(7) Osteoporosis Canada,(8) and Nguyen et al. in Australia.(9) Recently, a WHO scientific group developed an internet based global fracture risk platform, FRAX, based on 12 prospectively studied population-based cohorts in multiple geographic areas globally using data from the primary databases and that has been validated in other cohorts.(10) There was follow-up for ∼250,000 patient-years of 60,000 men and women, during which >5000 fractures were recorded. FRAX uses risk factors that have been validated in multiple populations, are readily accessible by primary care physicians, are amenable to therapeutic manipulation, and are adjustable for age, sex, and type of fracture. These risk factors are age, sex, BMD of the femoral neck or BMI, prior history of fracture, current smoking, alcohol use, ever glucocorticoid use, secondary osteoporosis, rheumatoid arthritis, and parental history of hip fracture. These clinical risk factors are added together but have different weights. These risk factors improve the gradient of risk of BMD alone. The addition of risk factors to BMD is most valuable for younger women and least valuable for the elderly. If BMD information is not available, these risk factors can be used alone. The output of FRAX is the 10-yr absolute risk of major osteoporotic fracture (hip, clinical spine, forearm, and shoulder) and the 10-yr risk of hip fracture. The risk of different clinical fractures is combined in this model. Rather than simply adding the risk of several osteoporotic fractures, the morbidity and mortality of each fracture is weighed using a utility measurement compared with the utility of a hip fracture. FRAX also incorporates risk of mortality and fracture. Mortality risk is based on age and BMI. Patients with both very low and high BMI have increased risk of mortality.

FRAX represents a significant advance in global case finding of patients at risk. However, this superset of WHO clinical risk factors has significant limitations and may be perceived as complex by some physicians.

FRAX is limited by under-reporting of some risk factors. Radiographic vertebral fractures are not included. There is no dose/response correction for multiple fractures, excessive alcohol use, or high cumulative corticosteroid dose. FRAX is limited by over-reporting of some risk factors such as patient report of diagnosis of rheumatoid arthritis (RA), because many patients have difficulty distinguishing RA from osteoarthritis. In the United States, FRAX data on ethnic groups was modeled from white incident fracture rates and NHANES III BMD data rather than specific fracture incidence rates.

FRAX does not include many risk factors that intuitively should be in the model such as low vitamin D, bone loss documented by DXA,(11) bone marker data, use of other skeletal sites such as the total hip or spine or peripheral measurements, or frequent falls. These and other factors may enter the model in the future if global cohort data on these risk factors become available. It is also important that these future risk factors be amenable to therapeutic intervention. For example, a recent Cochrane review found little effect of a comprehensive intervention to reduce falling on fracture risk.(12)

Unfortunately it is not known how FRAX was developed and which regression parameters are used in the model. We do not know all the extrapolations used. We do know that, for example, the gradient of risk for RA was used for secondary osteoporosis. Furthermore, the performance of FRAX in subpopulations has not been fully tested. For example, Hillier et al.,(13) using the SOF cohort, found that FRAX predicted hip fracture better in women with normal and low bone mass than in those with osteoporosis. FRAX may not work as well for men.(14)

The acceptance of FRAX has been slow in primary care practice because of multiple factors; FRAX measurement at this time requires access to the internet and may be perceived as complex. Few primary care providers use other risk scores such as a Gail score for breast cancer or a Framingham score for cardiac events. Many primary care physicians in the United States do not yet have internet access in the examination room. In the near future, FRAX will be added to upgrades of BMD software for compatible machines allowing BMD technologists to enter FRAX information, permitting manufacturers to add a FRAX risk score to the BMD printout. However, the data input by the technologist will need to be reviewed by a physician.

Other simpler risk assessments have also been proposed. Leslie et al.(8) reported recently in JBMR on the Osteoporosis Canada risk factor assessment system that includes age, BMD, previous fracture, and major corticosteroid use (>7.5 mg daily for 3 mo).This system was predictive of 10-yr fracture risk. However, it was not compared with FRAX, nor was it evaluated in another cohort outside Canada. Nguyen et al.(9) developed a simplified risk nomogram using data from the Dubbo Epidemiology Study including age, BMD, prior fracture, and falls. This model has not been validated in other cohorts. Ensrud et al.(7) compared 10-yr risk of hip, osteoporotic, and clinical fractures in women >65 yr of age in the Study of Osteoporotic Fracture using age and femoral neck BMD alone and found similar risk assessment than that calculated using FRAX with or without BMD.(13) In a recent issue of the JBMR, a simplified risk factor assessment in elderly Dutch women was discussed by Plujim et al.(15)

Could patients complete a simplified fracture risk assessment by questionnaire? As shown in the article of Plujim et al.,(15) self-report of several of the FRAX variables such as presence of RA may be overestimated by patient self-report and will require physician confirmation. RA prevalence was 4% in Rotterdam, which had a baseline medical exam, and was 14% in the Longitudinal Aging Study Amsterdam (LASA), where information was obtained using trained interviewers.

Plujim et al.(15) calculated both 5- and 10-yr absolute risk of hip fracture in Dutch elderly women, whereas FRAX only calculates 10-yr fracture risk. In FRAX, the 10-yr time frame was chosen as opposed to lifetime fracture risk for several reasons: assumptions regarding future mortality introduce increasing uncertainties, treatments are not given over a lifetime; the long-term prognostic value of some risk factors decrease over time, and, although 5 yr accommodates clinical trial experience of 3–5 yr, a 10-yr time frame includes time off drug when therapy is stopped.

The fracture risk algorithm reported by Plujim et al.(15) uses five simple questions: age, any prior fracture after age 50, body weight <64 kg, use of a walking aid as a proxy measure of serious immobility, and current smoking, and does not use BMD. As the authors pointed out, a general practitioner or nurse practitioner could easily use the fracture risk score. Women with high-risk scores could be referred for BMD testing or recommended directly for osteoporosis therapies. The risk score needs to be confirmed in other cohorts.

The endpoint of major osteoporosis fracture in the paper of Plujim et al. differs from FRAX in that it includes hip, wrist, pelvis, and humerus fractures but not clinical vertebral fracture. Some of the risk factors used in FRAX such as positive family history, alcohol use, and RA were not included in the Dutch study because the authors felt they were poorly associated or difficult to assess. Plujim et al. pointed out that alcohol self-reporting may be difficult to measure because of social correctness; that family history may be unreliable and biased by longevity of parents; that RA is under-reported in general practice and over-reported by patients who cannot separate RA from osteoarthritis. The patients in the study of Plujim et al.(15) had a minimum age of 60 in Rotterdam and 65 in LASA. FRAX allows us to measure fracture risk between ages 40 and 90 yr in both men and women. The Dutch study is also limited by self-report of fractures and retrospective reporting.

FRAX creates a measure of absolute risk based on risk factors from 12 global cohorts and country-specific incidence rates and thus can be used globally. The model of Plujim et al., however, needs to be validated in another sample of community-dwelling elderly and in younger patients. The respondents were a selective group of relatively healthy older Dutch women.

How will the information from risk assessment measures be used to decide whom to treat? FRAX does not define the intervention threshold or the probability of fracture at which treatment becomes cost-effective in a given country. The intervention threshold is the level of costs and effects associated with a probability of fracture at which an intervention is acceptable to the country and depends on willingness to pay. Here in the United States, Tosteson et al.(16) defined cost-effective thresholds as $60,000/quality-adjusted life-year. In the United States, a 10-yr absolute risk of hip fracture of 3% and a 10-yr absolute risk of major osteoporotic fracture of 20% using FRAX is cost effective.(16) These thresholds, which are used in the new NOF guidelines(17) are, however, difficult for some patients (and some physicians) to understand. Leslie et al.(8) for Osteoporosis Canada used a three-level system: high risk (>20% 10-yr absolute risk of fracture), medium risk (10-20%), and low risk (<10%), similar to cardiovascular risk. This categorization may be easier to understand for patients and may be made easier to understand if the level of risk is color coded such as red = high risk, yellow = moderate risk, and green = low risk.

In conclusion, risk assessment scores are here to stay in osteoporosis. We need to simplify them and make them easier for the general practitioner to use. Risk scores will assist good history taking and help clinicians decide which patients get BMD testing and which patients will receive treatment. We will need to develop simple risk factor assessments that can be completed by patients who could measure their own fracture risk and proactively discuss their risk with their healthcare providers. Healthcare practitioners could confirm or initiate and start therapy if the risk is high, suggest lifestyle change if risk is low with calcium and vitamin D, or, if risk is intermediate, suggest BMD testing. It is also likely risk scores will become more important in design of osteoporosis clinical trials. As the concept of fracture risk becomes more established, healthcare providers will talk more about patients' risk for fracture than their risk for osteoporosis.

How will risk factor assessment impact clinical practice? Healthcare providers will continue to treat older patients with fracture or osteoporosis without FRAX or other risk assessments, but the younger woman with low bone mass and at low risk for osteoporotic fracture may not be targeted for therapy. We will most likely decrease our use of the term osteopenia to avoid confusion by patients that osteopenia is a disease. Instead we will refer to osteopenia as low bone mass at low or high risk of fracture.

In the future, we will be able to use good history taking to identify clinical risk factors, and with new risk factor assessment measures, we will be able to integrate these risk factors with BMD to help us better identify patients at risk. The use of this approach may be more effective in identifying patients at high risk who need treatment and avoid unnecessary treatment for those at low risk, reducing the burden of osteoporotic fracture.


  1. Top of page
  • 1
    Kanis JA, Dawson A . Oden A, Johnell O, de Laet C, Jonsson B 2001 Cost effectiveness of preventing hip fracture in the general female population. Osteoporos Int 12: 356361.
  • 2
    Kanis JA, Melton LJ III , Christiansen C, Johnston CC, Khaltaev N 1994 The diagnosis of osteoporosis. J Bone Miner Res 9: 11371141.
  • 3
    Siris ES, Chen YT, Abbott TA, Barrett-Connor E, Miller PD, Wehren LE, Berger ML 2004 Bone mineral density thresholds for pharmacological intervention to prevent fractures. Arch Intern Med 164: 11081112.
  • 4
    Wainwright SA, Marshall LM, Ensrud KE, Cauley JA, Black DM, Hillier TA, Hochberg MC, Vogt MT, Orwoll ES 2005 Study of Osteoporotic Fractures research group hip fracture in women without osteoporosis. J Clin Endocrinol Metab 90: 27872793.
  • 5
    Van der Klift M, de Laet CE, McCloskey EV, Johnell O, Kanis JA, Hofman A, Pols HA 2004 Risk factors for incident vertebral fractures in men and women: The Rotterdam Study. J Bone Miner Res 19: 11721180.
  • 6
    Neuner JM, Binkley N, Sparapani RA, Laud PW, Nattinger AB 2006 Bone density testing in older women and its association with patient age. J Am Geriatr Soc 54: 465469.
  • 7
    Ensrud KE, Lui L, Taylor B, Schousboe J, Donaldson M, Fink H, Cauley J, Hillier T, Browner W, Black D, Cummings S 2008 A comparison of prediction models for hip, osteoporotic fracture and any clinical fracture in older women: Is more better ? J Bone Miner Res 24: S18.
  • 8
    Leslie WD, Tsang JF, Lix LM 2009 Simplified system for absolute fracture risk assessment: clinical validation in Canadian women. J Bone Miner Res 24: 353360.
  • 9
    Nguyen ND, Frost SA, Center JR, Eisman JA, Nguyen TV 2008 Development of prognostic nomograms for individualizing five year and ten year fracture risks. Osteoporos Int 19: 431444.
  • 10
    Kanis JA, on behalf of the World Health Organization Scientific Group 2007 Assessment of Osteoporosis at the Primary Health Care Level. WHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield, Sheffield, UK.
  • 11
    Berger C, Langsetmo L, Joseph L, Hanley DA, Davison S, Josse RG, Prior JC, Kreiger N, Tenenhouse A, Goltzman D for the CaMos Research Group 2009 Association between change in BMD and fragility fracture in women and men. J Bone Miner Res 24: 361370.
  • 12
    Gillespie LD, Gillespie WJ, Robertson MC, Lamb SE, Cumming RG, Rowe BH 2003 Interventions for preventing falls in elderly people. Cochrane Database Syst Rev 4: CD000340
  • 13
    Hillier TA, Cauley JA, Rizzo JH, Pedula KL, Vesco KK, Taylor BC, Bauer DC, Ensrud KE, Black DM, Cummings SR 2003 WHO absolute fracture risks in older women. How does prediction vary for incident hip and nonspine fracture across T scores. J Bone Miner Res 23: S17.
  • 14
    Sandhu SR, Nguyen ND, Center JR, Pocock NA, Eisman JA, Nguyen TV Validation of the FRAX algorithm and Garvan nomogram for predicting fracture risk. Presented at the Annual Meeting of the Australia and New Zealand Bone Mineral Society, August 29, 2008, Melbourne, Australia. Available at Accessed April 8, 2009.
  • 15
    Pluijm S, Koes B, de Laet C, Van Schoor NM, Kuchuk NO, Rivadeneira F, Mackenbach JP, Lips P, Pols HA, Steyerberg EW 2009 A simple risk score for the assessment of absolute fracture risk in general practice based on two longitudinal studies. J Bone Miner Res 24: 768774.
  • 16
    Tosteson ANA, Melton LJ, Dawson-Hughes B, Baim S, Favus MJ, Khosla S, Lindsay RL 2008 Cost-effective osteoporosis treatment thresholds: The US perspective from the National Osteoporosis Foundation Guide Committee. Osteoporos Int 19: 437447.
  • 17
    National Osteoporosis Foundation 2009 NOF's new clinician guide to prevent and treat osteoporosis. Available at Accessed March 2, 2009.