You are on page 1of 10

Osteoporos Int (2002) 13:527–536

ß 2002 International Osteoporosis Foundation and National Osteoporosis Foundation
Osteoporosis
International

Position Paper

A New Approach to the Development of Assessment Guidelines for
Osteoporosis
J. A. Kanis1, D. Black2, C. Cooper3, P. Dargent4, B. Dawson-Hughes5, C. De Laet6, P. Delmas7, J.
Eisman8, O. Johnell9, B. Jonsson10, L. Melton11, A. Oden12, S. Papapoulos13, H. Pols14, R. Rizzoli15,
A. Silman16, A. Tenenhouse17, on behalf of the International Osteoporosis Foundation and the National
Osteoporosis Foundation, USA
1
WHO Centre for Metabolic Bone Diseases, University of Sheffield, UK; 2Division of Clinical Epidemiology, University of
California at San Francisco, USA; 3MRC Environmental Epidemiology Unit, University of Southampton, UK; 4INSERM, Paris,
France; 5USDA Human Research Center, Tufts University, Boston, USA (President, NOF); 6Institute for Public Health, Erasmus
MC, Rotterdam, The Netherlands; 7INSERM Research Unit, Hoˆpital Edouard Herriot, Lyon, France (President, IOF); 8Garvan
Institute of Medical Research, St Vincent’s Hospital, Australia; 9Department of Orthopaedics, Malmo¨ General Hospital, Sweden;
10
Dept Economics, Stockholm School of Economics, Sweden; 11Section of Clinical Epidemiology, Mayo Clinic, Rochester, USA;
12
Consulting Statistician, Gothenburg, Sweden; 13Department of Endocrinology, Leiden University Medical Center, The
Netherlands; 14Ziekenhuis Dijkzigt, Rotterdam, The Netherlands; 15Division of Bone Diseases, University Hospital, Geneva,
Switzerland; 16ARC Epidemiology Research Unit, University of Manchester, UK; and 17Division of Bone Metabolism, The
Montreal General Hospital, Canada

Introduction The clinical significance of osteoporosis lies in the
fractures that arise, with their attendant morbidity and
mortality. Low bone mass is an important component of
An increasing awareness of osteoporosis, combined with the risk of fracture, but other skeletal abnormalities
the development of treatments with proven efficacy, will contribute to bone fragility. In addition, a variety of
increase demand for the more effective management of nonskeletal factors contribute to fracture risk, particu-
patients with osteoporosis. This in turn will require larly those related to falls. Thus, ideally the assessment
widespread facilities for the diagnosis and management of fracture risk should encompass all these aspects of
of osteoporosis. Measurements of bone mineral are a risk. For this reason, there is a distinction to be made
central component of any provision that arises from the between the diagnosis of osteoporosis and the assess-
internationally agreed definition of osteoporosis, i.e., a ment of risk. This in turn implies a distinction between
systematic skeletal disease characterized by low bone diagnostic and intervention thresholds. Although diag-
mass and microarchitectural deterioration of bone tissue, nostic thresholds have been defined [3], this paper
with a consequent increase in bone fragility and summarizes the approach by which intervention thresh-
susceptibility to fracture [1,2]. The diagnosis of olds might be defined for clinical practice.
osteoporosis thus centers on the assessment of bone
mass and quality. Since there are no satisfactory clinical
tools widely available to assess bone quality, the Background
diagnosis of osteoporosis depends at present upon the
measurement of skeletal mass [3].
In recent years, a number of approaches to the
assessment of osteoporosis have been formulated based
Correspondence and offprint requests to: Prof. J. A. Kanis, WHO
Collaborating Centre for Metabolic Bone Diseases, University of on a case-finding strategy. In Europe, the approach
Sheffield, Beech Road, Sheffield S10 2RX, UK. Tel: +44 (0)114 271 recommended by the European Foundation for Osteo-
2649. Fax:+44 (0)114 273 9176. porosis (now the International Osteoporosis Foundation,

1. a development committee of the National Osteoporosis Foundation (NOF) began a detailed assessment of the elements required to develop guide- lines [10. The intervention thresholds were risk factors such as prior fragility fractures. Thus.11]. or at the same site individuals identified have a high risk of fracture. Population relative risks for hip fracture over 10 years in Swedish women with osteoporosis according to age. Kanis et al. Thus. if it is cost-effective to treat all individuals with a T-score of Fig. however. the diagnostic threshold – a T-score of 72.5 SD or less. Risk is shown for women at the threshold value for osteoporosis (T = 72. whereas the incidence of fracture rises (Fig. the individuals at high risk go undetected. 1). make a diagnosis and thereafter to treat.5 SD). defined. They are not readily defined by the WHO criteria [3]. For example. An important concept developed was the view that intervention thresholds should be modulated according to risk. An advantage is addition. 2.5 SD. A further relative risk of fracture for a given BMD decreases with drawback is that the iintervention threshold is set at age [15]. the T-score suspicion. It is now evident that the T-score has a different prognostic significance at different ages [8. In and disadvantages with this approach. Fig. . there are many risk factors in addition to age that provide information on fracture risk over and above that provided by BMD alone. corticoster. questionable relevance to other countries [12]. In 1995.5 SD) or those below the threshold (T <72. A. less stringent than 72. there has been an enormous increase in the that guidelines are intuitive to the practice of medicine number of validated assessment techniques. as with the approach of the IOF it has also become by measurement of bone mineral density (BMD) and evident that the use of T-scores or Z-scores is treatment is offered in the presence of osteoporosis as problematic in general practice. Ten-year probability (%) of hip fracture in Swedish women 72. A derived from one site has a different prognostic meaning further advantage is that they are conservative in that all from the same T-score at another site. For example. Moreover. family history and low body mass index [4–7]. 2).5 SD.528 J. by cost-effectiveness analysis of oid use. each with in that the sequence of clinical practice is to alert different performance characteristics. diagnostic thresh- olds are not equivalent to intervention thresholds since the range of risk varies so markedly at any given BMD. but with a different methodology [13.9] (Fig. a prior fragility fracture increases the risk of a further fracture even after adjustment for BMD. More- Individuals with such risk factors are assessed thereafter over.14]. The use of Therein also lies a disadvantage in so far as many relative risks is also problematic. The argument runs that.6-fold per 1 SD change in BMD + the 95% confidence estimate of the gradient [15]. This is confusing for clinicians. There are advantages understood or used by primary care physicians. Hip fracture risk is assumed to increase 2. The NOF provided detailed assessment and intervention strategies based on the modulation of the T-score (or Z-score) in the presence of IOF) was to identify individuals on the basis of strong various risk factors. then it is worthwhile to treat patients with a according to age and T-score for bone mineral density at the femoral prior history of fracture at a T-score that is somewhat neck [9].

Such estimates of lifetime probability of fracture are of value in considering the future burden of disease in the community and the likely For assessment of intervention strategies.2 1.5 1. consistent with clinical are less relevant for assessing the risk in individuals in practice. Absolute Fracture Risk Absolute risk.3 1. time frame accommodates a treatment for 5 years with Table 1.2 12. evidence for persistence of effect for several years after 10-year probability) increases with age [25]. all ages. which has been improving in all vertebral. lifetime probabilities are in.3 21.Clinically diagnosed fractures.6 15.7 5. 3) but that this may be true also for fracture risk [26]. a shorter effects of intervention strategies on the population. Ten-year probabilities (%) of a first fracture at the sites shown in men and women from Malmo¨ by age [25] Site of fracture Age (years) 50 60 70 80 90 M F M F M F M F M F Distal forearm 1.2 0. significantly poorer than the assessment of BMD alone to predict hip fracture [3. number of treatments. Average long-term fracture probabilities in men and women predictions. gender. whereas the use of serum cholesterol alone has a low gradient of risk.9 10. A 10-year strategies [6].7 3. This probability should be derived not only from age and sex.1 1.. all races and all countries even though the incidence of osteoporotic fractures varies widely by age.9 5. regions of the world [24].1 4. effects on BMD appear to persist appropriate since the remaining lifetime risk for many when treatment is stopped and there is some evidence fractures decreases progressively with age (Fig. i. has generally been calculated in the context of lifetime fracture risks. including but not limited to bone mass measurements.8 Any of these 3.e.. diabetes and serum cholesterol permits the identification of patients at high risk (>20% 5-year risk).8 0.3 0. probability of fracture [14.3 7.e.9 1.4 5.2 3. Moreover.6 0.4 6. For long-term Fig.6 1. but also of the likelihood that individuals with given characteristics at the time of assessment will survive.3 3. For models of the cost-effectiveness of treatment. The optimal duration of treatments is not well whom treatment might be envisioned.7 2.. or more exactly ‘long-term probability’.16].4 a .2 1. but interventions of 3–5 years or so treatments are not presently given for a lifetime. clinical in mortality [23].6 2. account needs to be taken of secular trends from Malmo¨. parathyroid hormone and the bispho- tested using either high-risk or global public health sphonates than for calcium or vitamin D [26]. 3. The the risk during treatment. but also from validated risk assessment tools. They time frame is appropriate.22]. These probabilities take account not only of the incidence of fracture at different ages. stopping treatment is greater for hormone replacement the feasibility of lifelong interventions has never been therapy (HRT).g.9 13.5 6.4 Spinea 1.9 1.0 4.4 7. the simultaneous consideration of smoking.1 5.9 4. The use of absolute fracture risk has the potential to be applicable to both sexes.5 10.0 Proximal humerus 0.1 5. . i.Assessment Guidelines for Osteoporosis 529 These considerations have led to the view that intervention thresholds should be based on absolute risk.5 4.9 7. ethnicity and geography. This is because evaluated.6 7.2 2. proximal humeral and distal forearm fractures combined [25]. the short-term risk (e. due correspond with information available from trials and variably to side effects. low compliance and cost.3 Hip 0. blood pressure.4 1.4 7. In cardiovascular disease.6 18.7 2. Sweden for hip fracture alone and for hip.1 26. Similar ap- proaches are now used in the management of cardiovascular diseases [17–21].3 6. For a shorter-term treatments.

but this rests on as many assumptions. but no approach is perfect [35. according to the disutility (utility lost) from each type of Against this background. As expected. cause SD) would fall within the threshold risk from the age of substantial impairment in the activities of daily living 65 years onwards.31]. A further consideration is that not all fractures are due to osteoporosis. hip fracture. represent less than 50% of all fractures [30. at the hip in women from Sweden. an offset of effect over the subsequent 5 years. consideration of other fractures requires an probability of hip fracture according to T-score at the evaluation not only of the site-specific pattern of fracture hip and age in Swedish women [9]. hip fracture has a marked effect on intervention thresholds (Fig. A more recent example is the lack of effect of risedronate in elderly women selected by their risk of falling compared with those selected on the basis of low BMD [34].530 J. The same threshold risk is would have a different significance at the age of 50 years attained in women from the age of 57 years. . than at the age of 70 years.42]. type of approach reduces fractures and their conse- when they comprise a much higher proportion of quences to a common currency and permits the fractures. an intervention that prevented 5 fractures per probabilities computed taking other osteoporotic frac- 100 treated patients (number needed to treat. 4). The many fractures that occur in osteoporosis pose Thus. and this is etc. then women with osteoporosis (T-score <72. The left-hand panel shows the probabilities for hip fracture.41. In determining risk thresholds for pharmacologic interven- tions. The right-hand panel includes the [32]. low BMD. Colles’ fracture. The according to the site of fracture. 4. approach used recently is to characterize fractures due to The horizontal dotted line denotes a hypothetical intervention osteoporosis as those (a) associated with low BMD and threshold. this when hip fractures are rare. Thus. The right-hand panel shows probabil- tic fractures. 4 shows the 10-year Moreover. NNT = 20) tures into account [40]. the disutility from osteoporosis as judged by the mortality and morbidity different fractures varies. A. for example. The disutility The Choice of Fracture End-points associated with each fracture is the cumulative loss of utility over time. Ten-year probabilities of the common accepted parameter in the health economic assessment of osteoporotic fractures for Swedish men and women of interventions [43] and can be used to calculate different ages are given in Table 1 [25]. age groups where these other types of fracture Adding all disutility-weighted osteoporotic fractures predominate. Even over the age of 80 years. In order to estimate QALYs.5 fractures and vertebral fractures. Thus. for example. the greatest loss being from experienced by patients and costs to health providers. Although intervention threshold were set at a 10-year probability less devastating than hip fractures. The treatment has been stopped. Ten-year probability of fracture according to age and T-score incidence with age or associations with other osteoporo. in the FIT trial. Time (b) where the incidence increases with age [40]. but also of their morbidity. which ranges from 0 (equal to death) to 1 (perfect health). 10-year probabilities seem fracture. treatment with alendronate had less efficacy on appendicular fractures in women without osteoporosis [33].) and the consequences of these fractures vary substantiated by empirical observation [27. disutilities. therefore. vertebral. each year of life is valued according to its utility. intervention thresholds determined on the probability of hip fracture risk alone would neglect the fracture for 4 vertebral fractures or 20 for less severe many other fractures that occur. account needs to be taken. The ‘exchange rate’ is approximately 1 hip However.28]. for example fractures of the face and skull. distal forerm of 15%. An ities adjusted to take account of fractures in addition to hip fracture. due to variations in rates of bone loss.36–39]. Kanis et al. relative risk) of risk factors that may decrease over time. of those fractures that are due to osteoporosis. Long time frames also pose responsivity of fracture risk to intervention may be a problems in assessing the predictive value (odds ratio or further useful approach when more data are available. If. An approach to same may be true for other risk indicators. Osteoporotic fractures have been variously character- ized on the basis of low-energy trauma. An frames in excess of 10 years may be misleading for alternative is for expert panels to assign the proportion of patients considering treatment when the period of fractures at any one site that are due to osteoporosis greatest fracture risk will occur far in the future when [38. particularly in younger fractures such as Colles’ fracture [40]. Treatments may not affect the risk of this type of fracture. Thus. an incidence with age. Disutility decreases with age due to the Hip fracture is the most important consequence of higher mortality. Quality-adjusted life years (QALYs) are the appropriate. rising Fig. theoretical calculations indicate that the long-term problems for simplifying the assessment of risk because predictive value of BMD for fractures wanes with time of the multiple outcomes (hip. for example resolving the problem is to weight different fractures the ratio of carboxylated to total serum osteocalcin [29]. The left-hand panel in Fig.

biochemical Some examples are given in Table 2. but also on mortality. the available data suggest women.1 1.6 1. for any according to geographic location.8 best documented in the case of hip fracture.40. Nonetheless.47–49].8 13..7 14. i.6 0. which are Australia (Geelong) 0. in those standardized to a reference country. [31]. 5.3 4. but there are a not only upon fracture hazards.5 1.1 6. Pattern of common osteoporotic fractures expressed as a proportion (%) of the total in the USA. The China (Hong Kong) 0.5 5.2 the large variation in fracture risks worldwide.0 0. where the hip fracture rate is known. given age the ratio of hip to other fractures is relatively constant between countries.9 1. [47]. permitting hip fracture probabilities to be that the pattern of fracture is similar.4 4. at least in the Western world (Fig. but prospective studies also show a similar variation in risk [46]. [40] and the UK from Singer et al. Many of these studies are from Turkey (Istanbul) 0.0 in different countries.9 0.7 1.1 0. One challenge is to take account of France (Picardy) 0.e.g. geographic location they include prior fractures.9 Norway (Oslo) 0.6 13.4 0.2 2.3 1. probabilities can be adjusted osteoporotic fractures [31.3 0.3 0. In general.7 upon a commonality in the pattern of fracture incidence UK 0. Sweden [50].9 than 10-fold [44.3 5.5 2. Sweden from Kanis et al. associated with fracture risk. .0 2.2 2.3 1.2 0.6 incidence of hip fracture worldwide varies by more China (Beijing) 0.4 0. BMD measurements alone are not adequate to provide a Over the long term.2 0. the probability is high in associated with osteoporosis. corticosteroids.5 1. and.8 4.2 4.8 2. the probability of fracture depends sole instrument for population screening.1 0. e.7 0.6 0. the Risk Factors risks and consequences for other osteoporotic fractures can be estimated.1 4. use of 10-year probabilities of hip fracture can be computed.7 2. Ten-year probability of hip fracture (%) in men and women of the multiple outcomes of osteoporosis.3 12. Sweden and the UK.7 1.1 0.2 0. Other putative risk factors Fig.2 2.9 0.36..6 2. In general.0 registers.Assessment Guidelines for Osteoporosis 531 development of intervention thresholds that take account Table 2.45]. countries where the risk of hip fracture is high then so On the assumption that other fractures can be computed too is the risk of forearm. Thus. vertebral and other from hip fracture rates.0 4.2 0.4 1. though those have been less well studied. where markers of bone resorption and certain diseases the probability is high in men.2 0. according to age and country [50] Country Men aged (years) Women aged (years) Ethnic and Geographic Applicability 50 60 70 80 50 60 70 80 The validity of this approach to other countries depends USA (Rochester) 0. large number of other risk factors that are consistently and this too varies from country to country.4 7.7 10.9 4. Apart from age. 5). There are also variations in the risk of other osteoporotic fractures. Data from the USA are from Melton et al.4 0.6 1. sex and where mortality risks and hip fracture risks are known.4 1.6 9. low body mass index.

. which carries a individuals with an average BMD for age but a relative very high risk of hip fracture in men and women. Thus. and the relative risk falls only to 1. sex and type of fracture objective of risk assessment is to identify individuals at . It is thus evident that Gradients of risk exceeding 3.532 J. menopause are important factors. Although BMD at the hip predicts hip fracture as well as There are.53].. The use of the . therefore not much greater than the risk associated with either factor alone. hip fracture ensure that the risk identified can be modified by the risk was increased 1. but only by 1. Thus.5 per 1 SD change. an early menopause is a 15% of the female population at highest risk at the significant risk factor for any osteoporotic fracture in menopause (Table 3). combination of a continuous variable (e. selection of risk factors for use in fracture prediction. Fortunately. Thus. to the population risk according to its prevalence at each . The combined risk is the assessment of ‘preventable risk’.. different risk factors have different relevance at For example. A final con- dependent. Validated in multiple populations population risk as a reference is appropriate since the . 73% of fractures are may in part be due to the different pattern of fracture missed for comparable specificity. age and fracture pattern). The importance of these risk should have a utility even where clinical risk factors factors differs. For example. BMD) and For example. These sensitivity falls to 27%. if it were wished to select for treatment the different ages. the weight of each if not. For forearm fractures low BMI and early marked impact on the detection rate (see Table 3). Different risk factors may also have gradient of risk. prior fragility fracture) ankle fractures but these factors do not contribute yields a continuous variable with a higher gradient of significantly to the risk of forearm fractures at the risk than the continuous variable alone. The .3.. significance for fracture in the elderly. example body mass index (BMI) and BMD in the elderly Similarly. but their use may be mutually skeletal measurments where these are known to exclusive. they should incorporate other laboratory- [52. . however. The need to be chosen with care. Kanis et al. such as family history of fracture and smoking require to These considerations are important because guidelines be validated more widely.g. In other settings. since they Other risk factors are more or less dependent. being categorized by age (i. be reluctant to target Similarly.9-fold by smoking. but guidelines should adjusted for BMD. Readily assessable by primary care practitioners high risk compared with the general population. An example is cigarette smoking and BMD. this risk factor can be utilized accommodate measurements by peripheral devices to enhance risk assessment in conjunction with BMD. a maternal family tribution to fracture risk prediction. they predict fractures significantly when the one is bone mineral measurements make an important con- adjusted for the other. but is of uncertain specificity of 85% but a sensitivity of only 45% [15].87 compared with the general population [15]. for example 1. a The low sensitivity indicates that the majority of family history of hip fracture appears to be a risk factor fractures (55%) would occur in those women categorized in the elderly but is not a consistent risk factor at as being at low risk. account of age and the distribution of fracture types that These considerations indicate that risk assessment varies with age.g.e.6 compared with An example of the last point is dementia. the use of hip BMD would have a perimenopausal women [54]. the different relevance for different fracture sites. so that risk needs to be categorized by age since. Conversely. In addition. Intuitive rather than counterintuitive to medical care. Thus. A third category of risk factors are partially contribute significantly to fracture risk. intended intervention. neither technique is adequate for population screening. Clinical risk factors to be used for fracture prediction strength of predictive value and independence. including validated ultrasound technologies. They should be: weighting can be based on the combined relative risks adjusted to the population by age.g. risk factors other than age or BMD can be used to identify patients in whom Risk Assessment densitometry is indicated or to enhance the value of BMD. the with age (e.55]. hip fractures are common in the elderly). Some are more or less independent in that alone are available for assessment. age dominates all assessments. so that the focus needs to be on fold when adjusted for BMD [10]. For example. for are reimbursed in many countries such as the USA. Physicians might. age itself is a dominant risk should be based on an estimate of fracture probability factor. First. either BMI or BMD can be utilized to derived risk factors such as biochemical tests and other characterize risk. high BMI and smoking are risk factors for a dichotomous variable (e. With a technique with a poorer menopause. however. Adjusted for age. For example. derived from a panel of risk factors. Emphasis to date has history of hip fracture increases the risk of hip fracture been on measurements by dual-energy X-ray absorptio- two-fold [51]. a dichotomous risk factor should be adjusted osteoporosis treatments on the diagnosis of dementia.9 when metry (DXA) at the hip. sideration in the context of intervention thresholds is to Among elderly women in the SOF study.0 per SD have sensitivities guidelines based on risk factor assessment need to take of >50% in the example used above. A. several considerations in the blood pressure predicts cardiovascular disease [22. risk of 1. Contribute to a risk that is amenable to the therapeutic adjustment of relative risk in this way applies to manipulation intended continuous variables as well as dichotomous variables. a Z-score of 71 for BMD at the hip has a relative risk for hip fracture of 2. This has a same age [54].

2 91.8 75.5 56 2.2 Swedish population [59].1 2.2 4. 2.1 40.0 85.06 RR/SD Population RR compared with the risk for the general population [15].6 85.7 (b) Women aged 65 years 1. The effect of risk factors alone or in combination on the relative risk of hip fracture in women at the age of 80 years.1 60.6 Prior fracture Yes 39 3.9 2.9 0. The effect of gradient of risk for fracture prediction (RR/SD) Where P is the prevalence and OR the odds ratio.3 18.0 4.4 8.3 2. according to different population cut-offs to define a high-risk category [15] Gradient High-risk category (% of population) of risk (RR/SD) 5 10 15 25 PPV Sensitivity Specificity PPV Sensitivity Specificity PPV Sensitivity Specificity PPV Sensitivity Specificity (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (%) (a) Women aged 50 years 1.3 Low BMD + high CTX As above 16 4.4 6.0 24.58] are given in Table 4 together with 10-year hip fracture probabilities applied to the 2.1 95.82 29.5 5.3 90.4 3.6 59.2 2.6 6.87 = 3.4 1.0 95.0 90.0 56.Adjusted to the population.9 96.3 11.5 5.2 17.2 96.5 77.6 30.2 2.0 23.6 96.4 11.2 3.1 2.0 5.6 51.0) of relative risk.1 75.1 11.7 42.3 95.2 22. it were intended to OR/P (OR + (1 7 P)) = 1.8 The selection of patients with risk factors that provide 3.5 13. i.6 25.1 37.0 a high gradient of risk increases sensitivity.1 40.6 3.8 77.6 6.2 28.0 the EPIDOS study [57. the fewer the patients identified above any given population is threshold of risk.4 87.9 5.5 95.8 59.2 63.7 90.6 4.7 85.8 29.4 2.86 Table 5.9 77.6 7.Assessment Guidelines for Osteoporosis 533 Table 3. carboxy-terminal crosslinked telopeptide of type I collagen.40 23.74 40.4 17.1 31.9 4.1 3.6 5.6 90.9 2. found in 7.3 15.8 25.5 3.5 2.5 1.7 49. for example. Further examples of combinations of risk factors from 1.5 85.9 46.4 75.50 47.3 2.5 3.0 6.2 60.3 a .6 77.3 0.4 87.0 22. sensitivity and specificity of measurements to predict hip fracture over 15 years or to death in women aged 50 years (a) or 65 years (b).1 1.7 10.8 44.9 18.9 6.8 52.6 23.3 3.1 86.9 77.3 90..8 Table 4.6 34.5 25.6 25.2 26.0 2.5 37.4 39.7 3.0 36.8 18.5 95.0 75.8 90.7 0.9 38.1 85.6 12.4 92.9 18.5 75.1 2. over a wide range of assumptions.8 2.0 29.4 0.2 1.e.0 2.2 4.0 Low BMD T-score <72.1 56.7 95.8 69.8 90.0 95. the lower the gradient of [56].6 5.7 87.4 4.3 3.1 2. An example is provided by poor visual acuity.5 32.7 31.5 4. If.9 11. age band. The increase in risk compared with the general risk.5 92.8 77. Estimates of positive predictive value (PPV).9 6.39 36.5 48.4 13.8 High CTX Above premenopausal values 23 2.7 2.5 85.6 66.7 91.0 9.9 57.8 1.7 92.0 3.9 6.3 49.0 76.6 71.5 3.4 detection rate of the test.0 7.2 17.0 11.2 3.0 34.6 11.0 46.1 85.8 11.8 75.0 18.77 28.3% improves the detection rate.6 95.3 51.8 30.4 1.4 76.1 Prior fracture + high CTX As above 12 5.0 12.4 17.5 7.4 24.4 24. was associated with a two-fold higher population that can be selected with a given threshold risk of hip fracture than better visual acuity (OR = 2. The right-hand column gives the probability of hip fracture within the next 10 years [59] Risk factors Threshold values Prevalence Risk Relative Ten-year (%) ratio riska probability (%) Average – 100 1.0 4.3 97.0 90.1 41.4 75. This not only In the EPIDOS study poor acuity (<2/10).6 5.2 48.4 26.1 43. Thus.0 75. it also enlarges the of the population. the 4.5 Low BMD + prior fracture As above 23 4.5 2.9 23.5 26.5 29.0 31.86 6 1. on the proportion of the population identified according to the threshold of relative risk [60] in this example the combined risk of low BMD (Z = 71 SD) and poor visual acuity is 1.3 34.2 3.1 95.1 36.4 87.8 87.2 2.9 17.4 85.3 27.0 97.0 15.1 60.7 19.3 92.3 44.0 44.7 3.6 4. without trading off specificity .0 34.6 51.0 41.6 20.3 5. Conversely.5 15. CTX.

This approach does not accommodate agents that have significant extraskeletal benefits and The objective of assessment is to provide information risks such as HRT or the SERMs. proximal humerus or clinical vertebral fracture) in men and women from Sweden according to age and relative risk [61]. result in thresholds for each intervention.1% of the general population. For .g. However. The general based on effectiveness [6].. The three decisions are: for their provision is that cost-effectiveness is critically dependent on these extraskeletal risks and benefits. based on cost-effective. The argument for not that can be used for decision-making in the management providing different thresholds is that they evolve of patients. because costs of treatment depend upon such thresholds for they will vary according to costs and efficacy [10]. a test with a and only thereafter consider the method of treatment. No further assessment or treatment required example. gradient of 2. with treatment in women with osteoporosis would mean ment that it would be cost-effective to treat women with a . since in The diagnosis of osteoporosis is made from the many instances intervention thresholds will be less measurement of BMD. Further assessment indicated. but practitioners first decide whom to treat risk of 3 times higher or more than average. relationship between relative risk and fracture prob. there are obvious arguments for not providing score of 72. treatment is recommended in those with a BMD below Health economic justifications. a significant reduction of breast cancer risk . Current clinical guidelines follow the The choice of a probability threshold for each decision principle that BMD measurements are indicated in point is complex and depends upon not only absolute individuals with risk factors for fracture and that risk. there is would detect 6. intervention and assessment threshold irrespective of the intervention intended. In this sense BMD is utilized as a risk assessment. but a test with a gradient of risk of 3 per SD ness and side effects of interventions differ. that fracture risk country at each cost. a critical value. Fig. e. diagnostic site. therefore. the threshold of BMD for intervention will vary according to the risk Summary and Conclusions factor profile including the risk contributed by age. 6. treatment without this effect.0 per SD would identify 2. Other inadequate evidence to provide a hierarchy of treatment examples are provided in Table 5 [60].5 SD. In Europe the critical value corresponds to a T- However. 6 [61]. It is evident. BMD. however. whereas in the USA less stringent separate thresholds for each treatment available in each criteria are used. identify individuals from the general population with a unwieldy. There are broadly three decisions that can be guidelines towards product specificity. Although it is likely that effective- population. Kanis et al.534 J. In some countries reimbursement for the ness. Ten-year probability (%) of an osteoporotic fracture (hip. Note that DXA at the hip is considered to be the diagnostic test for osteoporosis. diagnostic assess. Treatment indicated irrespective of any diagnostic lower (less stringent) probability threshold than with a assessment. A. DXA at the hip is the appropriate stringent than the diagnostic threshold. distal forearm. but the efficacy of interventions and their costs. It is considered. Not only is this complexity at any given T-score varies markedly according to age . The argument based on fracture probability. One approach is to determine cost utility with a basket of treatments of average cost and Intervention Thresholds effectiveness.7% of the and not vice versa. that guidelines should focus on developing a standard ability is shown in Fig.

San Diego: Academic Press. 22. Yeo WW. Gilbert C.12:519–28. Thus. Oden A. et al. Musliner TA.348:121–2.348:387–8. Effect and 4. Adams JE. prophylaxis and 24. Corrected. The sex and age distributions of population.27:585–90.8:599–603. Guidleines for diagnosis and management of osteoporo. Kakonen S-M. European Quantitation of Osteoporosis Study. Symptomatic fracture incidence in elderly men and women. than the diagnostic threshold. Long-term risk of osteoporotic fracture in application to screening for postmenopausal osteoporosis. treatment of osteoporosis. Applegate WB. Johnell O. World Health Organization. Crowson CS. J 8. 1998. Disease. Oden A. Bone 2000. Jackson PR. Kanis JA. Eisman 1988. 7. Delmas P. Osteoporos Int 2000. prophylaxis and and projections. Effect of alendronate on and US practice guidelines for osteoporosis. Osteoporos Int 1997. JC. Olsson JO. An update on the instances intervention thresholds will be less stringent diagnosis and assessment of osteoporosis with densitometry. Kanis JA. Oden A. Glu¨er CC for the Committee of Scientific Advisors. Sernbo I. Johnell O. intervention thresh. criteria for osteopenia and osteoporosis. 5. Trends Endocrinol risk of fracture in women with low bone density but without Metab 2000. Q J Med 1994. Zippel H. Mancia G. 1998. Outcome of osteoporotic 11. Calcif Tissue Int 1999. Greendale GA. Jonsson B. The Dubbo Osteoporosis Epidemiology Study (DOES). Wood D. Johnell O.11:28–32. A 18. Kanis JA. since in many International Osteoporosis Foundation. Cooper C. Physicians guide to prevention fractures: In: Marcus R. promising approach is to weight fracture probabilities Pyorella K. United Nations. Reeve J. 16. 1994. probability appears to be an appropriate time frame.8(Suppl 4):1–88. et al. National Osteoporosis Foundation. Black DM. 31. Johnell O. McClung MR.140:199–270. Dawson A. Melton LJ. The treatment of osteoporosis. 15.90:107–110. The Sheffield table for primary prevention of coronary heart disease. 1994 revision of the United Nations global population estimates 2. Kanis JA. Payne validity in different regions of the world and their JN. Thompson DE. Jones G. Account also needs to be taken of sclerosis 1998.94:646–50. Dawson A. Jonsson B. Ruudgren A. Cooper C. Johnell O. Kanis JA.81:1804–9. JA. et al. London: RCP.87:203–9.11:120–7.11:669–74. Am J Med 1991. so that models of risk of coronary artery disease: an updated Sheffield table. Osteoporosis. Cummings SR. Koshi K. Dawson A. Technical report series 843. Clinical guidelines for the Lovgren T. Kanis JA. Palermo L. Minnesota: comparison of urban with rural rates and cost of screening and treatment strategies for osteoporosis: a and changes in urban rates over time. 6. New York: UN. Amsterdam: Excerpta Medica. Ramsay LE. National Osteoporosis Foundation.280:2077–82. Analyses of the effectiveness Olmsted County. Dequeker J. Am J Med 1993. vertebral fractures: results from the fracture intervention trial. Kanis JA.9:29– basis for development of practice guidelines. 1998. Jackson PR. Kanis JA. Osteoporos Int 2000. Cooper C. Steinbuch M. At what coronary risk level is it cost-effective to worldwide. Jonsson B. 28. Burckhardt P.4:277–82. 14. Dargent-Molina P. Anonymous. morbidity. 1995. Kelly PJ. Johnston CC. Statin therapy – what now? Drug Ther Bull that of other fractures which contribute significantly to 2001. Miller P. Faergeman O. Osteoporosis: recent advances in pathogen- esis and treatment. Redlund-Johnell I. Yeo WW. References 23. Lunt M. 13.11:192–202. Laippala P. European Commission. Sambrook PN.10:193–9. et al. predictors of fracture in a prospective study. 3. Barrett-Connor E. Haq IU. Johansson C. A 10-year fracture to the population of Sweden. adults by the ratio of carboxylated to total serum osteocalcin. Haq IU. Slemenda CW.7:390–406. Osteoporos Int 37. Royal College of Physicians. Aihie A. 1. Luukinen H. prevention and treatment of osteoporosis. DeBacker G. A further challenge for the future will be to initiate cholesterol lowering drug treatment in primary preven- identify risk factors that predict fracture with high tion? Eur Heart J 2002. 33. Torgerson D. Department of Health. Bone 2000. Kanis JA. 25. in independent cohorts. Kanis JA. J Clin Invest 30. Benson WG. Roux . Kroger H.11(Suppl 2):S103.39:17–21. Consensus development conference. Risk of hip olds need to differ from diagnostic thresholds and be fracture in Sweden according to relative risk: an analysis applied based on fracture probabilities. Lancet 1996. Health. Dere W. 1999. 9. Malmo¨. This has led to the view that men and women with osteoporotic fractures of spine and hip. London: Department of Osteoporos Int 1999. Fracture incidence in 10. Osteoporos Int 2001.15:2473–8. Osteoporos Int 2000. Consensus development conference. Luxembourg: Office for Official Publications of the ments overestimates risk. Diagnosis. BMD is utlized as a risk assessment. In this sense. sis. Oden A.64:191–9. Bone density reduction in various measurement sites in 34. Diagnosis. 32. Prevention of coronary heart disease in clinical practice: recommendations of the second joint task force of according to the disutility incurred compared with hip European and other societies on coronary prevention. Assessment of fracture risk and its Dawson A. DeLaet C. irrespective of a fixed BMD threshold. Torgerson D. Pickin DM. Johnell O. Oden A. They need to fracture risk in postmenopausal women. Osteoporos Int 1999. Osteo- Risk of hip fracture according to the World Health Organization poros Int 1994. Dawson A. Comparison of the European Barrett Connor E. Nguyen PN. Strong prediction of fractures amongst older prevention and treatment of osteoporosis. Lancet prediction can be constructed and ultimately validated 1996. Targeting lipid-lowering drug therapy for primary prevention independent contributions. on Mellstrom D. O’Fallon WM. particularly in younger individuals. An assessment tool for predicting development of assessment guidelines. Ramsay LE. 1998. Freedman D. Johnell O. Does the predictive value for hip fracture of a behalf of the European Foundation for Osteoporosis and Bone calcaneal bone mineral measurement change with time [abstract]. Osteoporos Int 2000. 26. 12. De Laet C. Lindsay There are a number of problems to be overcome in the R. Hui SL. the large geographic variation in fracture probabilities 19. Quick reference primary guide on the offset of effect of treatments for hip fracture on health outcomes. Lifetime risk of hip fracture is underestimated. 21. 29. Black DM. Osteoporos Int 1999.Assessment Guidelines for Osteoporosis 535 and other risk factors. European Communities. Kelsey J (eds) and treatment of osteoporosis. Age and bone mass as Bone Miner Res 2000. 1998:635–44. 20. 27. Jonsson B. Pettersson K. take account of not only the risk of hip fracture but also 17. Report on osteoporosis in the European Prediction of fracture from low bone mineral density measure- Community. Birkenhager JAMA 1998. Dere W. Dawson A. Geneva: WHO. Hoseyni MS. Johnell O. Johannesson M. The interventions should be targeted to those at high risk. Jonsson B. in press. Graham I. Geusens P. Athero- fracture probability.26:387–91.

47. Looker A. Jacobs J. Osteoporos Int 1999. Dequeker 60. Tuppurainen M. Osteoporos Int 1998. Markers of bone turnover predict hip fractures in al. Epidemiology of osteoporosis. Cummings SR. Elffors L. Jonsson B. BMJ 1996.10:240–7. 1992. Melton LJ. Jonsson B. Bacon WE. Johnell O. Kanis JA. Navi CR. fracture: a population-based study of 12. sources. Johnell O. markers of bone resorption add to bone mineral density and 43. Chesnut CH. Fall-related factors 7:414–25. in press.000 adults: the influence of age Ten year risk of osteoporotic fracture and the effect of risk factors and gender. women: validation in the Rotterdam Study. 37. Delmas PD. Case J. DeLaet CEDH. Sanders KM. 1986. DeLaet CED.9:271–9. Wedel H. von Overbeck J.80:234–8. Osteoporos Int 1992.6:69–75. Johnell O. Stone K. Hans D. Ray NF. Which fractures are associated with low appendicular ed. Oden A. Russell LB.13:523–526. Chan JK. Lancet 42. Ugoni AM. N Engl J Med 1995. Pols lHAP. Sanders KM. University Press. et Muller C.13:1587–93. population based study. Kanis JA. Scott JC.536 J. Received for publication 18 February 2002 Accepted in revised form 22 February 2002 . Weinstein MC. AM. Allander E. Singer BR. Johnell O. Kanis JA.4:253–63. Johnell O. Garnero P. Melton LJ. Gullberg B. C.115:837–42. Phillips S. (1997) Incidence and direct medical costs of hospitalizations due 56. Kanis JA. The hospital burden of vertebral risedronate on the risk of hip fracture in elderly women. Baudoin C. Wright WE. Oglesby AK.2:298–302. Do Bone 1998. National Osteoporosis Foundation. 58. elderly women. Kanis et al. Philadelphia: Lippincott-Raven. Johnell O.312:1254–9. Martin TJ.12:16–23. The EPIDOS prospective study. Allander A. Siegel JE. Fractures attributable to osteoporosis: report from the sectional analysis. Bone 2002. Garnero P. Osteoporosis: etiology. Christie J. Marcelli C. Favier F. The burden of osteoporotic fractures: a method for setting 55. Osteoporos Int 1997. In: Riggs BL. et al. Osteoporos Int 2002.8:563–9. De Laet CED. Osteoporos Int 1998. Weel 1997. Melton LJ. J Bone Miner Osteoporos Int 1996. De Laet C. underestimate the prevalence of bone fragility fractures in the 54. Einhorn Bone density and risk of hip fracture in men and women: cross TA.12:417–27. Grandjean H. Nevitt MC. Burger H.348:145–9. finding and opportunistic screening: a mathematical approach. 57. Oden A. Johnell O.16(Suppl 1):S414. Bone Ensrud KE. Harris T. McLauchlan CJ. Osteoporos Int 1994. Chapuy MC.7:138–44. Dargent-Molina P. Kanis JA. The variable incidence of hip fracture in Southern Europe. 46. Seeman E.315:221–5. Alhava E. Robinson CM. Seeley DG. Epidemiology of fracture in 15. Hofman A.332:767–73. Osteoporos Int Med 2001. De Laet C. Perrelet R. Gullberg B. J Bone Miner Res Martin TJ. Gold MR. Schott AM.344:222–34. and risk of hip fracture: the EPIDOS prospective study. Garnero P. Oden A. International comparison of hip fracture rates in 1988–1989. Bossard H. Meta-analysis of how well intervention thresholds. Nevitt MC. Gullberg B. Seeman E. Cummings SR. incidence of hip fracture in Europe: a study of national register Biochemical markers and the assessment of fracture probability. 52. Jonsson B. 2nd Group. 45.192 perimenopausal 40. The exclusion of high trauma fractures may 1998.8:25–31. Hansherr E. Maggi S. Skoric International variations in hip fracture probabilities: implications B. Burger H. J The MEDOS study. Marshall D. Ugoni AM. Honkanen R. Giaconi J.13:S51–9. Hauser E. AEAM.and gender-specific rate of fractures in Australia: a for assessment guidelines. J Bone Miner Res 2002.11:1531–8. Pasco JA. for the EPIDOS group. A. 1997. Kanis JA. Osteoporos Int 2001. et al.30:251–8. 61. Res 1996. Fox N. J Bone Miner Res 53. of osteoporosis from American women aged 45 and older. Grandjean H. van Hout BA. Johnell O. Kanis JA. Hip fracture prediction in elderly men and 39. women. Dargent-Molina P. BMJ 1997. for the Study of Osteoporotic Fractures Research editors. Cost. De Laet C. community: the Geelong osteoporosis study. 1995:225–47. 50. Browner WS. New York: Oxford fracture in elderly women? The EPIDOS prospective study. 44. Oden A. Age. 49. Schott to osteoporotic fractures in Switzerland. Pitt FA. Epidemiology of fractures. et al. Bone Miner Res 2001. 36. Hoffman A. Pols HAP. Pasco JA. measures of bone mineral density predict occurrence of 41. on screening strategies. Genant HK. N Engl J fracture: a study of national register sources. Thamer M. J Bone Miner Res Relationships between risk factors and fractures differ by type of 1998. Fox KM. 1996. Oden A. et al. Effect of 48. Jonsson B. et al. De Laet C. Risk factors for hip fracture in white women. Johnell O. Saarikoski S. diagnosis and management.13:1337–42. Browner WS. The direct medical cost 1996. et al. J Bone Joint Surg Br 1998. The apparent 59. et al. et al. Nicholson GC. Kanis JA. for the Intervention Programme Study Group. Lippuner K. Jaeger P osteoporotic fractures. et al. Dawson A. bone mass in elderly women? Ann Intern Med 1991. Van Hout BA. 35. 51. Kroger H. Kanis JA. 38. ultrasonographic heel measurement for the prediction of hip effectiveness in health and medicine. Dawson A.