Mapping translational research into individualized prognosis of fracture risk


: Dr Tuan V. Nguyen, Bone and Mineral Research Program, Garvan Institute of Medical Research, 384 Victoria Street, Sydney, Australia. Email:


The assessment of fracture risk has until now been based on the measurement of bone mineral density (BMD) and/or a prior fracture. Individuals with BMD T-scores < –2.5 (e.g. osteoporosis) or with prior fractures are indicated for treatment. However, recent data have suggested that 55% of women and 74% of men who sustained a fracture did not have osteoporosis. Therefore, the current strategy reduces a small number of fractures in the general population, and new thinking is required for that majority of individuals whose BMD measurements are at or near (both sides) the current threshold of osteoporosis. An individual's absolute risk of fracture can be estimated from the individual risk profile, which includes age, BMD, weight or body mass index, prior fracture, comorbidities, corticosteroid use, lifestyle factors, and falls. Therefore, risk assessment must simultaneously consider all risk factors to which the individual is exposed. A number of prognostic models and predictive nomograms have been developed to estimate an individual's absolute risk of fracture, but they have not been externally validated. Nevertheless, these prognostic models can be effective tools for individualizing short-term and long-term risks of fracture, which can help patient counseling and selecting appropriate patients for intervention to maximize the benefit of fracture reduction in the general population.


Osteoporosis and its consequence of fragility fracture is increasingly becoming a major public health threat in Asian countries, due mainly to population size and rapid ageing. Approximately three-quarters of the world population is residing in Asia. Moreover, the percentage of the elderly (aged 65 years and above) was about 5.3% in 1995 total population figures, and this percentage is projected to increase to 9.3% in 2025, which represents a 75% increase. The ageing of populations in the South-east Asia region is advancing more rapidly than other regions of the world.1 At present, the number of fractures in Asia is higher than that in European countries combined. Of all the fractures in the world, approximately 17% was found to occur in South-east Asia, 29% in the West Pacific, as compared to 35% occurring in Europe.2 The estimated disability-adjusted life years lost was 8.62 million in South-east Asia and 16.07 million in the West Pacific, which together represents 34% of the world figure.2

Theoretically, any fracture related to low bone density may be considered an osteoporotic fracture. Fractures of the spine (vertebrae), hip and wrist (distal forearm) have long been regarded as typical osteoporotic fractures.3–7 It has been shown that almost all types of fracture occurred more often in patients with low BMD;8,9 therefore the majority of all types of age-related fractures could be osteoporotic in nature. According to this definition, the following fracture types are considered osteoporotic: in women, vertebrae, hip and other femoral, wrist-forearm, humeral, rib, pelvic, clavicle, scapula, sternum, tibia and fibula. The patterns of osteoporotic fractures are similar in men, except tibia and fibular fractures are not considered osteoporotic. In both women and men, fractures occurring at the skull and face; hands and fingers; feet and toes, ankle and patella are classified as not due to osteoporosis.10

Because fragility or osteoporotic fracture is defined as a fracture related with minimal trauma (i.e. a fall from standing height or less),11,12 in the research setting, fractures clearly due to major trauma (such as motor vehicle accidents) or due to underlying diseases (such as cancer or bone-related diseases) were excluded from analysis.7,9,13–18 Nevertheless, results from a recent population-based study19 have shown that the prevalence of osteoporosis in high trauma fracture (i.e. falling from higher than standing height or due to motor vehicle accident) was comparable to that of the low trauma fracture population counterparts. Therefore, the authors have suggested that the exclusion of high trauma fractures from analysis may underestimate the prevalence of osteoporotic fractures in the community.

The magnitude of osteoporosis-related fracture can be quantified by the lifetime risk of fracture. Due to lack of longitudinal data, there have been few estimates of lifetime risk of fracture in Asian populations. Using population-based data, it has been estimated that the residual lifetime risk of hip fracture for mainland Chinese men and women was ∼2% and 2.4%, respectively.20 However, in the Taiwanese population, the residual lifetime risk of hip fracture was higher: 4.1% for men and 8.8% for women after the age of 50.20 In Caucasian populations, the only longitudinal-based and mortality-adjusted systematic estimate of lifetime risk was from Australia, which suggested that the lifetime risk of hip fracture for men and women was approximately 4% and 9%, respectively.21 Furthermore, approximately 4/10 women and 1/4 men aged 50 years will fracture during their remaining lifetime.21 Given that the current incidence of fracture in Asian countries is half of those in Western countries, it could be inferred that the residual lifetime risk of fracture in Asian women and men is around 25% for women and 15% for men.


During the past three decades, several clinical risk factors have been shown to be associated with fracture risk. These risk factors can be broadly classified into two groups: modifiable and non-modifiable risk factors (Table 1) with modest strengths of association (Table 2). The risk factors not amenable to modification include: the presence of a personal history of fracture occurring after the age of 40; a history of fracture occurring in a close relative; advancing age; being a woman; and genetic factors that are yet to be identified. The risk factors that are potentially modifiable include: current cigarette smoking; low body weight; estrogen deficiency or early menopause (in women) or hypogonadism (in men); lifelong low calcium intake; excessive alcohol intake; inadequate physical activity; poor health or frailty (including rheumatoid arthritis, hyperthyroidism, impaired eyesight, and dementia); long-term exposure to anticovulsant drugs; and a history of falls or recurrent falls. Among these risk factors, four key risk factors are identified: advancing age, personal history of a fragility fracture, family history of fracture, and low bone mineral density (or osteoporosis).

Table 1.  Risk factors for osteoporotic fracture in postmenopausal women
Non-modifiable risk factors
  1. The four risk factors highlighted in bold-faced letters are identified as major risk factors of hip fracture.82

 •1 A history of fracture as an adult
 •2 A family history of fracture (first-degree relative)
 •3 Being Caucasian
 •4 Advanced age
 •5 Being female
 •6 Dementia
Potentially modifiable risk factors
 •1 Cigarette smoking
 •2 Low body weight (< 65 kg)
 •3 Estrogen deficiency (early menopause < age 45, bilateral ovariectomy, prolonged premenopausal amenorrhea > 1 years)
 •4 Low calcium intake
 •5 Excessive alcohol intakes
 •6 Impaired vision
 •7 Multiple falls
 •8 Low levels of physical activity
 •9 Poor health/frailty
Table 2.  Magnitude of association between major risk factors and fracture: summary from meta-analyses
Risk factorRisk ratio and 95% confidence interval associated with
Any fractureOsteoporotic fractureHip fracture
  • These risk ratios were calculated for individuals with body mass index (BMI) = 20 comparing to those with BMI = 25 (as the reference level).

  •  ‡

    These risk ratios were calculated for individuals with BMI = 30 comparing to the reference level. All risk ratios for prior fracture, family history (first-degree relative), corticosteroid use, current smoking, BMI and milk intake were adjusted for BMD.

BMD (per SD)311.45 (1.39–1.51)1.55 (1.47–1.62)2.07 (1.91–2.24)
Prior fracture831.77 (1.64–1.91)1.76 (1.60–1.93)1.62 (1.30–2.01)
Family history of fracture601.19 (1.09–1.30)1.22 (1.10–1.34)1.48 (1.18–1.85)
Corticosteroid use841.57 (1.37–1.80)1.66 (1.42–1.92)2.25 (1.60–3.15)
Current smoking851.13 (1.01–1.25)1.13 (1.00–1.28)1.60 (1.27–2.02)
Body mass index860.98 (0.90–1.08)1.02 (0.92–1.13)1.42 (1.23–1.65)
1.01 (0.91–1.11)0.96 (0.86–1.08)1.00 (0.82–1.21)
Milk intake87NA1.06 (0.95–1.19)1.10 (0.83–1.47)


Advancing age is clearly a major risk factor of fracture, as the incidence of fracture increases exponentially with advancing age in both men and women.15,22–24 Recent data have shown that the 10-year probability of fracture at the forearm, humerus, spine or hip increases as much as 8-fold between ages 45 and 85 for women, and is 5-fold for men.25

Personal history of fracture

A prior fragility fracture substantially elevates an individual's risk of future fracture.18,26–28 The elevated risk is 1.5- to 9.5-fold depending on age at assessment, number of prior fractures and the site of the incident fracture. Pongchaiyakul and colleagues26 have shown that a pre-existing asymptomatic vertebral fracture increases the risk of a second vertebral fracture and non-vertebral fracture by at least 4-fold. A study of the placebo group in a recent major clinical trial showed that 20% of those who experienced a vertebral fracture during the period of observation had a second vertebral fracture within 1 year. Patients with a hip fracture are at increased risk of a second hip fracture. Pooling the results from all studies (women and men) and for all fracture sites, the risk of subsequent fracture among those with a prior fracture at any site is 2.2 times that of people without a prior fragility fracture (95% confidence interval [CI] 1.9–2.6).28

Family history of fracture

Accumulative data in the past 20 years or so have strongly suggested that a family history of osteoporotic fracture is a major risk factor of fracture. The Study of Osteoporotic Fractures,23 for example, identified a maternal history of hip fracture as a key risk factor for hip fracture in a population of elderly women. A history of hip fracture in a maternal grandmother also carries an increased risk of hip fracture.29

Although most studies have focused on the index person's mother or other female family members, genetic influence on risk of osteoporosis is multifactorial, and one should not ignore a history of osteoporotic fracture in first- or second-degree male relatives. The emphasis on the presence of osteoporotic fractures in patients’ female relatives in epidemiology studies probably reflects the belief that osteoporosis is mostly a disease of women. It is now clear that osteoporosis is common in men; therefore, although the recommendations focus on hip fractures in a patient's mother or grandmother, other family members should be included during assessment of genetic contribution to osteoporosis risk.

Bone mineral density

Bone strength is an important component for the assessment of fracture risk. At present, there are no direct methods for reliably measuring bone strength. However, BMD measured by dual-energy X-ray absorptiometry (DXA) provides a benchmark assessing bone strength. Indeed, in vivo studies have suggested that BMD measurement could ‘explain’ up to 70–75% of the variance in bone strength.30 Due to differences in the measurements of BMD among DXA manufacturers, BMD is often standardized by expressing it as a T-score. The T-score is the number of standard deviations from the young normal mean, taken as aged between 20 and 30 years.

There is a strong, continuous, and consistent relationship between BMD and fracture risk, such that each standard deviation (SD) lowering in BMD is associated with a 1.6-fold increase in fracture risk in both men and women.31 In a recent prospective study on postmenopausal Chinese women, the relative risk of fracture for each SD lowering in femoral neck BMD was 2-fold (95% CI: 1.6–2.5).32 There was evidence suggesting that the magnitude of association between BMD and hip fracture risk (with relative risk being 2.233 to 3.634) is equivalent to or even stronger than the association between serum cholesterol and cardiovascular disease. Thus, measurement of BMD is considered the gold standard for the diagnosis of osteoporosis in elderly men and postmenopausal women.

Given the strong association between BMD and fracture risk, in 1994 the World Health Organization (WHO) expert panel proposed an operational definition of osteoporosis, by which a postmenopausal woman is considered to have osteoporosis if the woman's femoral neck BMD is decreased by at least 2.5 SDs as compared to mean value in young adults.35 Femoral neck BMD is used in the diagnosis of osteoporosis because it is less likely to be affected by osteophytosis which can falsely elevate lumbar spine BMD measurements.36 The WHO classification also includes definition for osteopenia and normal BMD (Table 3). The operational criteria of osteoporosis for women were subsequently adopted for men.37 Although the T-score criteria were criticized as a flawed approach,38 they have been widely used in clinical practice.

Table 3.  WHO diagnostic categories for BMD in postmenopausal women
NormalBMD not more than 1 standard deviation (SD) below the peak BMD in young adult mean (T-score > –1)
OsteopeniaBMD between 1 and 2.5 SD below the young adult mean (T-score between –1 and –2.5)
OsteoporosisBMD 2.5 SD or above the young adult mean (T-score at or below –2.5)
Established osteoporosisBMD 2.5 SD or more below the young adult mean and the presence of one or more fragility fracture

Who should have a BMD measurement?

At the population level, an important issue is to decide who should undergo BMD measurement and further evaluation based on the results of BMD scan. DXA technology is relatively expensive and is not widely available in most developing Asian countries. Therefore, mass screening using DXA scanning is not recommended or not feasible without some selection of the target population. Guidelines from the International Osteoporosis Foundation,39 National Osteoporosis Foundation,40 and National Institutes of Health41 suggest that a case-finding strategy be adopted to identify individuals ideally eligible for a BMD scan. According to this strategy, the following individuals are recommended to have a BMD measurement: (a) women over 65 years or men over 70 years; and (b) women under 65 years or men under 70 years with risk factors of fracture as shown in Table 1.

A number of clinical prediction rules, including the Osteoporosis Self-assessment Tool for Asians (OSTA), have been developed42–44 to identify ‘candidates’ for a BMD scan. Most of these scores, being based only on age and body weight, are a simple prediction rule that can potentially be useful in identifying women at high risk of osteoporosis. However, these scoring systems have good sensitivity but poor specificity;45,46 therefore, they can be useful in ruling-out osteoporosis.

Quantitative ultrasonography (QUS) is a portable, less expensive, less time-consuming without radiation technology that is commonly used in Asian countries. However, in view of the limited evidence of its clinical usefulness, it is recommended that this technology is not used for the diagnosis of osteoporosis. However, recent prospective epidemiological studies have shown that QUS was an independent predictor of fracture risk in the elderly population. Pongchaiyakul and colleagues42 have recently developed a nomogram (Fig. 1), which combines information from QUS, age and weight to predict the probability of osteoporosis for an Asian woman. This multivariable-based nomogram has been shown to significantly increase the sensitivity and specificity over and above that of the Osteoporosis Self-assessment Tool for Asians (OSTA) and the Khon Khean Osteoporosis Study (KKOS) scores, with the area under the receiver operating characteristics being 0.86.42 This nomogram approach can potentially offer a practical and reliable means to assess the likelihood of osteoporosis for a woman.

Figure 1.

 Nomogram for predicting the risk of osteporosis. Instruction for usage: Mark an indvidual's age on the ‘Age’ axis, and draw a vertical line to the ‘Point’ axis to determine the number of points the individual receives for her age. Repeat this process for the weight and quantitative ultrasonography (QUS) T-score. Add the number of points from each predictor. Mark this sum on the ‘Total Points’ axis, and draw a vertical line down to meet the ‘Risk of Osteoporosis’ axis, to find the woman's probability of having osteoporosis. Example: Mrs X, 70 years old, weighs 50 kg and has a QUS T-score of −2; her points for age is approximately 50, her weight points is 67; and QUS points is 66. Her total points is therefore 50 + 67 + 66 = 183, and her probability of having osteoporosis is around 0.36. In other words, in 100 women like her, one would expect 36 of them to have osteoporosis.42


Bone is a net result of two counteracting processes of bone resorption and bone formation, often referred to as bone remodeling. Bone remodeling is a normal, natural process that maintains skeletal strength, enables repair of microfractures and is essential for calcium homeostasis. During the remodeling process, osteoblasts produce a number of cytokines, peptides and growth factors that are released into the circulation. Their concentration thus reflects the rate of bone formation. Bone formation markers include serum osteocalcin, bone-specific alkaline phosphatase and procollagen I carboxyterminal propeptide (PICP).

Osteoclasts produce bone degradation products that are also released into the circulation and are eventually cleared via the kidney. These include both enzymes and nonenzymatic peptides derived from cellular and non-cellular compartments of bone. Most biochemical indices of bone resorption are related to collagen breakdown products such as hydroxyproline or the various collagen cross-links and telopeptides. Bone resorption markers include urinary hydroxyproline, urinary pyridinoline (PYR), urinary deoxypyridinoline (D-PYR) as well as collagen Type I cross-linked N telopeptide (NTX) and collagen Type I cross-linked C telopeptide (CTX).

Markers of bone formation and resorption have been shown to be related with bone loss, with higher rates of bone resorption being associated with rapid bone loss.66 Some cross-sectional studies47,48 have shown that bone turnover rates as evaluated by markers increase at menopause and remain elevated. Bone turnover rate in postmenopausal women correlates negatively with BMD.49

A number of population-based epidemiological studies have shown that markers of bone resorption were associated with fracture risk. In the Rotterdam study which involved approximately 8000 individuals (60% women aged 55 years and over) van Daele and colleagues showed that the relative risk per SD increase in urinary deoxypyridioline (DPD) was 3.0 (95% CI: 1.3–8.6).50 There was evidence suggesting that the use of bone turnover markers and BMD could increase the identification of high risk individuals. Results from the EPIDOS study49 have shown that the combination of femoral neck BMD and of bone resorption markers increased the predictive power for hip fractures (RR 5–6). A nested case control in elderly men (aged 60+ years) from the Dubbo Osteoporosis Epidemiology Study demonstrated that accelerated bone resorption was associated with increased risk of osteoporotic fracture, independent of BMD; thus, combining measurements of BMD and bone turnover improved fracture prediction in elderly men.51 Furthermore, recent data from a prospective study of 1112 frail elderly men and women indicated that high bone turnover was an independent predictor of ‘all cause’ mortality.52

Bone turnover biomarkers as a group can potentially be a valuable tool for monitoring response to antiresorptive therapy in clinical trials. Normalization of bone formation and resorption markers following antiresorptive therapy has been prospectively observed.53,54 Reduction in biochemical markers appears to be correlated with a decrease in vertebral fracture incidence55 in some studies, but is not necessarily always predictive of response to therapies. Nevertheless, the predictive value of biomarkers in assessing an individual patient56 can be limited by their high variability within individuals.51,57


A major priority in osteoporosis research at present is the translation of risk factors into simple and accurate prognostic models to identify individuals who are at high risk of fractures in the future and to treat them appropriately so that their fracture risk will be reduced.58 The assessment of fracture risk has until now been largely based on the measurement of BMD and a history of prior fracture. This is logical, since there is a strong association between BMD and the risk of fracture.31,33,59 Furthermore, a history of postmenopausal fracture is also a strong risk factor of subsequent fracture.60 More importantly, the effect of prior fracture on subsequent fracture is independent of BMD. The National Osteoporosis Foundation guidelines recommend treatment in the following clinical situations: women with T-scores below –2 with no risk factors; women with T-scores below –1.5 with one or more risk factors for fracture (including a prior fracture); and women with a prior vertebral or hip fracture. Australian experts suggested that treatment should be initiated for any postmenopausal woman with osteoporosis or a prior fracture.61 This strategy is logical and evidence-based because results from randomized clinical trials show that treating these patients (e.g. with osteoporosis and/or a prior fracture) did reduce their fracture risk.33,62

However, there is a problem of treatment initiation based on a BMD cut-off value. Although the risk of fracture is directly related to BMD at all levels, and there is no threshold value for BMD that accurately separates those who will from those who will not sustain a fracture. Indeed, even at the lowest BMD range, only some individuals will sustain a fracture; on the other hand, a high BMD does not confer total protection against a fracture. Therefore, the dichotomization of BMD into osteoporosis versus non-osteoporosis by a threshold can be ineffective at the population level. Among individuals aged 60+ years, 55% and 74% of fracture cases occurred in non-osteoporotic women and men, respectively.63 As a result, treatment of individuals with osteoporosis can reduce only a modest number of fractures in the general population.

Important changes in thinking are needed for the majority of individuals whose BMD measurements are at or near (both sides) the current threshold of osteoporosis. First, it should be recognized that BMD, like other risk factors, provides quantitative and not qualitative information on fracture. Osteoporosis or low BMD is only one of many risk factors for fracture. At any given level of BMD, fracture risk varied widely in relation to the burden of other risk factors, such as advancing age, gender, genetics, family history of fracture, increased bone loss, low body weight, falls, and smoking behaviour. Thus, for any one individual, the likelihood that fracture will occur depends on a combination of those risk factors (Fig. 2).32 This means that two individuals, both with osteoporosis, can have different risks of fracture because they have different non-BMD risk profiles; on the other hand, an osteoporotic individual can have the same risk of fracture as a non-osteoporotic individual due to the difference in constellation of risk factors between the two individuals.

Figure 2.

 Incidence of hip fracture (per 1000 person-years) stratified by femoral neck BMD T-score and number of risk factors. For any given BMD level, the incidence of hip fracture increases exponetially with the number of risk factors.32

Second, the metric of risk should be absolute risk rather than relative risk. In the osteoporosis context, absolute risk is the probability of sustaining a fracture over a period. Relative risk describes the change (either increase or decrease) in the likelihood of fracture in a population in comparison to another (referent) population. Therefore, relative risk, by its ratio nature, is more suitable to a group of individuals or populations than to an individual. Absolute risk is more relevant to an individual, because it provides information about the absolute expected likelihood of fracture for the individual. A relative risk of 0.5 can mean a difference between 0.2% and 0.1% risks, but it can also mean a difference between 20% and 10% risks. Thus, knowing a therapy that reduces the risk from 20% to 10% (absolute risk) is much more useful than a relative risk of 50% (relative risk).

Third, prognosis is about imparting information of risk to an individual, and each individual is a unique case, in the sense that any two individuals are unlikely to have the same risk profile. Therefore, the risk of osteoporosis should ideally be individualized. One approach to increase the uniqueness of prediction is to consider the risk in its continuous scale, rather than in dichotomized scale based on a cut-off value. The uniqueness of prediction, or individualization of risk, can also be increased by considering multiple risk factors in a multivariable model, since the more factors considered, the greater likelihood of uniqueness of an individual is defined. Indeed, there is more than one way that an individual can attain the risk conferred by either low BMD or a prior fracture. For example, we have demonstrated that virtually all women aged 70 years with BMD T-scores less than –1.5 and all 80-year-old men with BMD T-scores less than –1.0 can be considered at ‘high risk’.64 On the other hand, 60-year-old men or women without a prior fracture and a fall are considered high risk, even when their BMD T-scores are below –2.5.64 This demonstrates the informativeness of a multivariable prognostic model, and the limitation of a risk stratification-based approach for risk assessment for an individual. The risk of fractures for an individual is the sum of probabilities that the individual is exposed to multiple risk factors.


Individualized prognosis refers to the use of prognostic methods that incorporate an individual's unique risk profile to achieve accurate and reliable prognosis of fracture for the individual, and to help improve the management of the individual's predisposition to fracture. The approach of individualized prognosis must be distinguished from the approach of risk stratification. In risk stratification, the estimate of risk is applicable to a group of individuals rather than to an individual. For example, the stratification of BMD measurement into osteoporosis versus non-osteoporosis based on the T-score treats two women with T-scores of –2.4 and –2.6 into two distinct groups despite the trivial numerical difference and that the two women may have comparable risks of fracture if other risk factors are considered. Moreover, such a stratification approach classifies an 80-year-old woman with a T-score of –2.5 and a 70-year-old woman with a T-score of –3.0 into a single group, despite the two women having different risk profiles! In contrast to the risk-grouping approach, the individualized prognosis approach recognizes that the four individuals are different, and that they should have different fracture risks as one would logically expect. Although this risk-grouping approach is simple and sometimes useful in clinical practice, its predictive value is poorer than the individualized approach due to the arbitrariness of the cut-off value.65

Recently, we and others56,64,66,67 have developed a number of prognostic models, in which multiple risk factors are simultaneously considered in a multivariable model. Most prognostic models require tedious computation, which can be impractical in the primary care setting, even with internet implementation. One alternative approach is to visually translate a statistical model into a nomogram so that it can be readily used in clinical practice. The idea of using nomograms in clinical medicine is not new. In 1928 Henderson described complex blood flow in a graphical format that he called a ‘nomogram’.68 Since then, the development and utilization of nomograms have exploded in clinical medicine. The medical literature records more than 1700 nomograms.69 Several nomograms have been developed and used in the field of oncology, and it has been demonstrated that nomograms exhibit a better performance than risk-grouping categorization,70,71 because a nomogram estimates a continuous probability of an event, which yields more accurate predictions than models based on risk grouping. The use of nomogram-based prognostic models obviates the need for grouping individuals by some arbitrary thresholds, and as a result, increases the uniqueness of an individual and allows the risk of fracture to be individualized. Moreover, nomogram-based prognostic models have been shown to out-perform clinical judgement,72 because they can objectively incorporate many risk data. Because of their objectivity, multivariable prognostic models can reduce the variability in risk estimates.

In addition to a website (, we have also developed nomograms for easy assessment of fracture in the primary care setting. Thus, the multivariable model recognizes the fact that there are different ways two individuals can attain the same risk level. For example, a 60-year-old woman with a BMD T-score = –2.5 and a history of fracture is predicted to have the same 5-year risk of fracture as an 80-year-old woman with a T-score = –1 without a previous fracture (Fig. 3).

Figure 3.

 Nomogram for predicting the 5-year and 10-year probability of hip fracture for a woman. Instruction for usage: Mark the age of an individual on the ‘Age’ axis and draw a vertical line to the ‘Point’ axis to determine how many points toward the probability of hip fracture the individual receives for his/her age value. Repeat the process for each additional risk factor. Sum the points of the risk factors. Locate the final sum on the ‘Total points’ axis. Draw a vertical line down to the 5-year or 10-year risk line to find the individual's probability of sustaining a hip fracture within the next 5 or 10 years. Example: Mrs. Smith, 70 years old, has a BMD T-score of −2.5, had a prior fracture and a fall in the past 12 months; her points for age is approximately 10, her BMD points is 65; prior fracture point is 8 and fall point is 3. Her total points is therefore 10 + 65 + 8 + 3 = 86, and her probability of having a hip fracture is around 0.091 in the next 5 years and 0.174 in the next 10 years. In other words, in 100 women like her, one would expect 9 and 17 of them will have a hip fracture in the next 5 years and next 10 years, respectively.56

The predicted risk of fracture is a continuous probabilistic variable ranging from 0 to 1. This raises the issue of selecting an optimal cut-off predicted probability to classify an individual into fracture or non-fracture. This is not an easy task, because the cut-off value – if it exists at all – depends on the complex risk-benefit consideration, and perhaps more importantly, an individual's perception of risk, which is beyond the scope of the present study. The level of predicted risk at which that individual is prepared to take action is dependent on the risk perception of the individual, which is not easily quantified. Nevertheless, the predicted probability of fracture from the present prognostic models can be viewed as a measure of severity of osteoporosis for an individual.

The individualization of fracture risk can help select patients suitable for intervention. The critical question of who should be treated can only be answered by a complete evaluation of an individual's risk profile, and to this end, the nomogram-based estimate can be helpful. This leads to the need to set absolute risk levels that treatment can be cost-effective. In a recent analysis, it was suggested that treatment is cost-effective (based on the criteria of £30,000 per quality-adjusted life-year gained) if an individual's 10-year risk of hip fracture is between 1.2% and 9.0%, dependent on age.73

The individualization of fracture prognosis may also be used to optimize the number needed to treat (NNT). Results from several randomized clinical trials74 indicate that the patient NNT to prevent one vertebral fracture compared to the control, ranged between 8 and 83 (Table 4). For hip fracture, the NNT ranged between 91 and 250.39 The NNT varied inversely with the background risk, such that treatment of high-risk individuals yielded lower NNT (Fig. 4). The large variability in the NNTs among trials was due to the variability in fracture rates among the study samples, despite the fact that patients were selected on the basis of having osteoporosis and/or a prevalent vertebral fracture. However, the variability is expected given the multiple risk factors that affect the incidence of fractures. In the presence of such variability, selecting patients based on their absolute risk of fracture (rather than based on a BMD threshold value) may improve the consistency of therapeutic efficacy and efficiency of trials. Although trials specifically testing the efficacy of multivariable risk-based therapy have not been done, it seems likely that such an approach would prove more cost-effective and would yield a more consistent NNT.

Table 4.  Incidence of morphometric vertebral fractures observed in large-scale randomized controlled clinical trials (RCT)
RCT and referenceAgentRisk profilePlaceboActiveNNT
  1. Prev fx = prevalent fracture; T = BMD T-score; NNT = Number needed to treat.

FIT-I74AlendronatePrev fx, T < –2.50.1500.08014
PROOF88CalcitoninPrev fx, T < –2.50.1560.1082 1
MORE-289RaloxifenePrev fx, T < –2.50.2120.14715
VERT-US90RisedronatePrev fx, T < –2.50.1630.11320
VERT-MN91RisedronatePrev fx, T < –2.50.2900.1819
Neer, 200192PTH 20 mgPrev fx, T < –2.50.1400.05011
Neer, 200192PTH 40 mgPrev fx, T < –2.50.1400.04010
FIT-293AlendronateNo prev fx0.0270.01583
FIT-293AlendronateNo prev fx, T < –2.50.0420.02148
MORE89Risedronate 60No prev fx0.0450.02345
Risedronate 120No prev fx0.0450.02859
TROPOS94StrontiumT < –2.50.1290.11259
Strontium95StrontiumPrev fx0.3280.2098
HORIZON96ZoledronatePrev fx0.1090.03313
Figure 4.

 Relationship between background risk (x-axis) and number needed to treat (NNT).

An important question is that whether treatment of individuals based on absolute risk of fracture can reduce fracture risk. McCloskey and colleagues75 have conducted a clinical study in which 5212 women aged 75 years and older were randomized into two groups, with the placebo group receiving calcium and vitamin D and the clodronate group receiving clodronate (800 mg daily orally). The ten-year probability of fracture was computed for each woman using baseline clinical risk factors including body mass index, prior fracture, glucocorticoid use, parental hip fracture, smoking, alcohol and secondary osteoporosis. Results indicated that in women lying at the top 25th percentile of fracture probability (average probability of 24%), treatment reduced the risk of fracture by 23% over 3 years (hazards ratio [HR] 0.77, 95% CI 0.63–0.95). However, among those in the top 10% percentile (average fracture probability of 30%), treatment reduced the fracture risk by 31% (HR 0.69, 0.53–0.90).75 Thus, treatment of women at high risk or moderate risk could reduce fractures.


During the past three decades, several risk factors, including low BMD, advancing age and a history of fracture, have been shown to be associated with fracture risk.23,31,34,76–79 Moreover, it has recently been shown that the risk of fracture increased with the cumulative presence of the number of risk factors.33 The issue at hand is how to translate the knowledge of risk factors into prognostic models for individualizing fracture risk in clinical practice. Some prognostic models have recently been developed.56,64,66 However, because these models have not been externally validated, their validity and accuracy in Asian populations is not clear. Therefore, external validation should be a priority of research in the application of risk assessment models.

As a fictional surgeon Angelfinger says, ‘We all know what a prognosis is! The problem is to know it – in each case’ (in Hilden80 quoted from Windeler81), individualization of risk – or the prediction of risk for an individual given a risk profile – is a fundamental to practising medicine. Since fracture risk is determined by multiple factors, any unidimensional risk assessment is unlikely to be helpful. A multivariable-based nomogram can be an effective tool for individualizing short-term and long-term absolute risks of fracture, which can help patient counseling and selecting appropriate patients for intervention to maximize the benefit of fracture reduction in the general population.


The authors would like to thank Dr Nguyen D. Nguyen for help in the preparation of graphs and data analysis. The work is supported by a grant from the National Health and Medical Research Council.