Predictive Value of BMD for Hip and Other Fractures



The relationship between BMD and fracture risk was estimated in a meta-analysis of data from 12 cohort studies of ∼39,000 men and women. Low hip BMD was an important predictor of fracture risk. The prediction of hip fracture with hip BMD also depended on age and z score.

Introduction: The aim of this study was to quantify the relationship between BMD and fracture risk and examine the effect of age, sex, time since measurement, and initial BMD value.

Materials and Methods: We studied 9891 men and 29,082 women from 12 cohorts comprising EVOS/EPOS, EPIDOS, OFELY, CaMos, Rochester, Sheffield, Rotterdam, Kuopio, DOES, Hiroshima, and 2 cohorts from Gothenburg. Cohorts were followed for up to 16.3 years and a total of 168,366 person-years. The effect of BMD on fracture risk was examined using a Poisson model in each cohort and each sex separately. Results of the different studies were then merged using weighted coefficients.

Results: BMD measurement at the femoral neck with DXA was a strong predictor of hip fractures both in men and women with a similar predictive ability. At the age of 65 years, risk ratio increased by 2.94 (95% CI = 2.02-4.27) in men and by 2.88 (95% CI = 2.31-3.59) in women for each SD decrease in BMD. However, the effect was dependent on age, with a significantly higher gradient of risk at age 50 years than at age 80 years. Although the gradient of hip fracture risk decreased with age, the absolute risk still rose markedly with age. For any fracture and for any osteoporotic fracture, the gradient of risk was lower than for hip fractures. At the age of 65 years, the risk of osteoporotic fractures increased in men by 1.41 per SD decrease in BMD (95% CI = 1.33-1.51) and in women by 1.38 per SD (95% CI = 1.28-1.48). In contrast with hip fracture risk, the gradient of risk increased with age. For the prediction of any osteoporotic fracture (and any fracture), there was a higher gradient of risk the lower the BMD. At a z score of -4 SD, the risk gradient was 2.10 per SD (95% CI = 1.63-2.71) and at a z score of -1 SD, the risk was 1.73 per SD (95% CI = 1.59-1.89) in men and women combined. A similar but less pronounced and nonsignificant effect was observed for hip fractures. Data for ultrasound and peripheral measurements were available from three cohorts. The predictive ability of these devices was somewhat less than that of DXA measurements at the femoral neck by age, sex, and BMD value.

Conclusions: We conclude that BMD is a risk factor for fracture of substantial importance and is similar in both sexes. Its validation on an international basis permits its use in case finding strategies. Its use should, however, take account of the variations in predictive value with age and BMD.


OSTEOPOROSIS IS described as a systemic skeletal disease characterized by low bone mass and microarchitectural deterioration of bone tissue, with a consequent increase in bone fragility and susceptibility to fracture.(1) The description focuses on measurements of BMD as well as on bone quality, but at present, only BMD can be readily measured in the clinic. For this reason, the operational criterion for osteoporosis is based solely on BMD measurement,(2) and the majority of intervention studies have used a low value for BMD as the inclusion criterion, as recommended by drug registration agencies in the United States and Europe.(3,4) The positive effect of pharmacological interventions in patients selected on the basis of low BMD values has further consolidated BMD as an important criterion for directing therapy.

As well as providing diagnostic information, low BMD is widely recognized as a major risk factor for fractures.(5) BMD measurements at the hip predict hip fracture with a gradient of risk that is comparable with the gradient of risk for cardiovascular disease provided by measurements of blood pressure.(5) The use of BMD can, however, be enhanced still further by taking account of other independent risk factors.(6,7) The most important of these is age and, for example, at the diagnostic threshold of osteoporosis (a T score of -2.5 SD), the risk of fracture varies markedly with age.(8) Although the general relationship between BMD and fracture risk measured at different sites has been well characterized,(5) there are still gaps in our knowledge even when using the reference site of DXA at the hip. Several cohort studies, for example, suggest that the predictive value of BMD at the hip is similar in men and women,(9,10) although the experience to date is limited and remains controversial.(11) Also, it is uncertain whether the predictive value of BMD attenuates with time after assessment. Theoretical and some empirical evidence suggests that this might be so(12) but others disagree.(13) It is also uncertain whether the gradient of risk provided by BMD is the same in different age groups and whether it also differs according to bone mass.

For these reasons, the aim of this study was to quantify the relationship between BMD and fracture risk as assessed by the reference standard of DXA at the hip.(14,15) In particular, we wished to characterize the predictive ability of BMD measurements by age, gender, time since assessment, and absolute value of BMD for different fracture outcomes on an international basis.


We studied 38,973 men and women from 12 prospectively studied cohorts. Brief details of the cohorts studied are given below and summarized in Table 1.

Table Table 1.. Details of the Cohorts Studied
original image


The Canadian Multicentre Osteoporosis Study (CaMos) is an on-going prospective age-stratified cohort. The study is documenting the incidence of fractures and risk factors in a random sample of 9424 men and women ⩾25 years of age selected by telephone listings. The sampling frame is from nine study centers in seven provinces.(16) Characterization of individuals was by interview. BMD was measured by DXA at the femoral neck (Hologic QDR 1000).


The Dubbo Osteoporosis Epidemiology Study (DOES) is a population-based study with multiple assessments of skeletal status in men and women ⩾60 years of age from Dubbo, Australia.(17,18) Participation in the study was 56% of the population. Baseline measurements included BMD at the femoral neck assessed using DXA (GE-Lunar, DPX, and Prodigy). Fractures are identified through radiologists' reports from the two centers servicing the region.


The EPIDOS cohort is a prospective multicenter study of risk factors for hip fractures that included 7575 French women ⩾75 years of age.(19) Participants were recruited through mailings using large population-based listings such as voter-registration rolls. Baseline characteristics were obtained through a structured questionnaire as well as through clinical and functional examinations and BMD measurement (Lunar DPX). For this analysis, validated fracture follow-up was available for 1183 participants in whom biochemical markers were measured in each hip fracture case and 3 controls with an average follow-up time of 3 years.(20) Femoral neck BMD was measured in 1180 individuals, and an ultrasound scan (Lunar Achilles) was done in 938 women.(21)


The European Vertebral Osteoporosis Study (EVOS) is comprised of age- and sex-stratified random samples from 36 centers in 19 European countries.(22) Equal numbers of men and women were drawn in each center within six 5-year age bands (50-54 up to 75-79 years). A base-line radiograph for vertebral fracture prevalence was undertaken in 15,570 men and women. BMD was measured in 3461 men and women from 13 centers by DXA at the femoral neck using pencil beam machines that were cross-calibrated using the European spine phantom. The sample provided the frame-work for the European Prospective Osteoporosis study (EPOS) where repeated assessment was undertaken in 29 of the centers.(23,24)

Gothenburg I

This study is comprised of four birth cohorts of 2375 randomly sampled men and women ⩾70 years of age followed for up to 20 years(25,26) after a baseline BMD measurement. The participants were drawn randomly from the population register in Gothenburg by the date of birth to provide cohorts of 70, 76, 79, and 85 years of age at the time of study. BMD was measured at the right heel using dual photon absorptiometry.

Gothenburg II

The Gothenburg study was comprised of a randomly drawn population cohort of ∼7000 women 21-89 years of age followed for 3 years.(27) Seventy percent of those invited participated in the study that examined risk factors for osteoporosis by use of a standardized questionnaire. BMD was assessed at baseline at the distal forearm using the Osteometer DTX 200.


The Adult Health Study (AHS) was established to document the late health effects of radiation exposure among atomic bomb survivors in Hiroshima and Nagasaki. The original AHS cohort consisted of about 15,000 atomic bomb survivors and 5000 controls selected from residents in Hiroshima and Nagasaki using the 1950 national census supplementary schedules and the Atomic Bomb Survivors Survey. AHS subjects have been followed through biennial medical examinations since 1958 with a participation rate of ∼80% throughout this period. BMD at the spine and femoral neck was measured at each examination using DXA (Hologic QDR-2000) in 1994. Trained nurses interviewed the subjects about fracture at each examination.(28,29)


The Kuopio osteoporosis risk factor and prevention (OSTPRE) study in Finland was comprised of a postal enquiry sent to all 14,220 women 47-56 years of age who were residents of the Kuopio province in 1989.(30) A total of 13,100 women responded to the enquiry, of whom 938 were excluded for incomplete information. This left a study population of 12,162 women. A random stratified sample of 3222 women underwent bone densitometry at the femoral neck using the Lunar DPX.(30)


The OFELY cohort is comprised of a cohort of 1039 women 31-89 years of age stratified by age randomly selected from the regional section of a large health insurance company (Mutuelle Generale d'Education Nationale, Lyon, France).(31) Eighteen percent of women contacted participated in the study. Baseline characteristics were obtained using a standardized questionnaire, including the documentation of prior wrist, humeral, vertebral, and hip fractures that occurred after 45 years of age. Only low-trauma fractures (falls from a standing height or less) were recorded. BMD was measured at the spine (L1-L4), proximal femur, distal radius, and whole body by DXA using a QDR 2000 (Hologic). Women were reviewed annually and fractures were registered. Peripheral fractures were confirmed by radiography. Vertebral fractures were identified from sequential X-rays of the thoracic and lumbar spine by morphometry in 80% of patients.(32)


The Rochester cohort was recruited from two random population samples stratified by decade of age, one comprising women who were subsequently followed for up to 20 years(13) and another sample of women and men followed for 8 years.(33) BMD of the right femoral neck was measured by dual photon absorptiometry in the first cohort (cross-calibrated to DXA) and by DXA (Hologic QDR 2000) in the second group. Fractures were ascertained by periodic interview combined with review of the inpatient and outpatient medical records of all local care providers.


The Rotterdam study, begun in 1990, is a prospective cohort study that aimed to examine and follow-up all residents ⩾55 years of age living in Ommoord, a district of Rotterdam.(34) By 1993, 7983 residents had been included (response rate 78%). BMD was assessed at the femoral neck by DXA using a Lunar DPX-L.(9) Fracture follow-up was undertaken using an automated link with general practitioner computer systems and hospital admission data. For this analysis, validated fracture follow-up was available for 7774 participants (3065 men) with an average follow-up time of 6 years. Femoral neck BMD was measured in 5776 individuals (2432 men).


The Sheffield cohort was comprised of women ⩾75 years of age selected randomly from the population of Sheffield, UK, and surrounding districts between 1993 and 1999. Approximately 35,000 women, identified from general practitioner listings, were contacted by letter and invited to attend for assessment of their skeletal status. A total of 5873 women were willing to attend for the screening visit. Of these, 281 were excluded, and the remainder were randomly allocated after informed consent to treatment with the bisphosphonate clodronate or to an identical placebo. The material used for this study is comprised of 2148 women allocated to treatment with placebo.(35) All women had baseline assessment of BMD undertaken at the femoral neck using the Hologic 4500, and 1701 women had an ultrasound scan (Cuba Clinical; McCue). Outcomes were assessed by 6-month home visits.

Baseline and outcome variables

BMD was assessed by multiple techniques. For the study we used BMD assessed at the femoral neck by DXA available in all cohorts except for Gothenburg. The z score of BMD for each sex and each cohort was computed from the regression of BMD by age because the variance by age was homogenous. Ultrasound measurements were made in women only from EPIDOS and the Sheffield cohort. Both broadband attenuation (BUA) and speed of sound (SOS-VOS) values were used in the calculations.

Peripheral measurements of BMD were performed in three cohorts: Rochester (one-third radius), Gothenburg I (DPA at the heel) for both men and women, and Gothenburg II (Osteometer DTX 200) at the radius only (DPX 200) in women.

Fracture ascertainment was undertaken by self-report (Sheffield, Kuopio, EVOS/EPOS, Hiroshima, EPIDOS, and OFELY) and/or verified from hospital or central databases (Gothenburg, CaMos, DOES, Kuopio, Sheffield, EVOS/EPOS, Rotterdam, Rochester). The EPOS study also included sequential systematic radiography to define incident vertebral deformities. Information on all clinical fractures was used for this report. In addition, fractures considered to be caused by osteoporosis were analyzed, and finally, hip fracture alone was considered separately. An osteoporotic fracture was one considered to be caused by osteoporosis by the investigator. For the EVOS/EPOS study, osteoporotic fractures comprised hip, forearm, humeral, or limb fractures. For the CaMos Study, they comprised fractures of the spine, pelvis, ribs, distal forearm, forearm, and hip. In the other cohorts, fractures at sites characteristic for osteoporosis(36) were extracted.

In the calculations of fracture outcome, vertebral fractures assessed by morphometry were not included, only clinical vertebral fractures. Fractures were categorized as follows: any fracture (all fractures reported in the studies), osteoporotic fractures, and hip fracture alone. In a separate calculation, osteoporotic fractures excluding hip fractures were used.

Statistical methods

The risk of fracture was estimated by the use of Poisson regression applied to each cohort and each sex separately. Covariates included time since start of follow-up and current age. We additionally tested interactions with age and BMD z scores in the model. The β-values for each coefficient are age-dependent, βk + βk + 1.age. The estimated value of βk + βk + 1.age was determined for each age from the age of 50-85 years together with the variance. The results of each analysis and the two sexes were weighted according to the variance and merged to determine the weighted mean and SD. The risk ratios at different BMD z scores were given by e(weighted mean coefficient) and expressed as a risk ratio per SD decrease in BMD, where the SD was sex-specific. We additionally adjusted the data using the SD of the young female reference range (age, 20-29 years) from the NHANES III data,(37) so that the gradient of risk in men and women could be compared using the same SD.

We used a fixed effects model rather than a random effects model because the latter weights the smaller cohorts disproportionately. In addition, the fixed effect model generally gives a more conservative point estimate for the risk ratio, albeit with wider confidence estimates. Heterogeneity in outcome variables between cohorts was tested by means of the I2 statistic,(38) and where more than moderate heterogeneity was found (>50%), risk ratios were recomputed using the random effects model to determine whether the significance of estimates had changed.


Details of the cohorts are provided in Table 1. The total sample studied for DXA at the femoral neck was 38,973 men and women followed for 168,366 person-years. During this time there were 3694 fractures, 2606 of which were considered to be caused by osteoporosis, and 971 hip fractures. In the cohorts where ultrasound was undertaken, a total of 4992 measurements were made with a follow-up of 15,337 person-years. During this time, 425 fractures were documented, of which 288 were hip fractures. For peripheral BMD measurements, a total of 9728 individuals were followed for 48,905 person-years, during which time 1010 fractures were recorded, of which 836 were considered to be osteoporotic, including 288 hip fractures. For hip fracture outcomes there was a low degree of heterogeneity (I2 = 24%; 95% CI = 0-60%), but heterogeneity was marked for other fracture outcomes (I2 = 65%; 95% CI = 36-80%; p < 0.001). When current age was used with BMD z score as an interaction term (Age × BMD), there was very low heterogeneity for osteoporotic fracture or for hip fracture outcomes and BMD (I2 = 11% and 27%, respectively; p > 0.17) and for the interaction term (I2 = 0% and 23% respectively; p > 0.20).


The gradients of risk afforded by BMD at the femoral neck for the prediction of any fracture, any osteoporotic fracture, and a hip fracture are shown in Table 2. Gradients of risk were highest for hip fracture, lowest for any fracture, and intermediate for osteoporotic fracture. There was no difference in the gradient of risk between men and women. For hip fracture risk, the gradient of risk per SD was marginally higher in women than in men, but this was not apparent when gradients of risk were examined by unit of BMD (see Table 2) or by age.

Table Table 2.. Gradient of Risk per SD Decrease in z Score of BMD in Men and Women
original image

Effect of age

Gradients of risk were not the same at all ages (Table 3). For any fracture, the gradient of risk increased significantly with age. For example, at the age of 55 years, the gradient of risk was 1.35 (95% CI = 1.71-1.57) and 1.29 (1.20-1.39) per SD decrease in men and women, respectively. At the age of 85 years, the gradient of risk was 1.60 (95% CI = 1.36-1.88) in men and 1.54 (1.44-1.65) in women. There was no significant difference between men and women. A similar effect was seen for osteoporotic fracture (Table 3).

Table Table 3.. Gradient of Risk (RR/SD) for Osteoporotic Fracture in Men and Women
original image

In the case of femoral neck BMD and the prediction of hip fracture, there was also a significant trend with age, but this was in the opposite direction. Risk ratios decreased with age, with no differences between men and women (Fig. 1). The data for men and women combined are shown in Table 4.

Table Table 4.. Gradient of Risk (RR/SD) for Hip Fracture in Men and Women Combined by Age
original image
Figure FIG. 1..

Hip fracture prediction (risk ratio/SD change) with femoral neck BMD in men and women by age.

In view of the increasing gradient of risk for osteoporotic fracture with age, but the converse effect for hip fractures, gradients of risk by age were recalculated for osteoporotic fractures minus hip fractures. The gradient of risk for osteoporotic fracture, without hip fracture included, increased with age, although the mean gradient was less (rising from 1.19 at the age of 50 years to 1.55 at the age of 85 years). Thus, the rise in gradient of risk with age was not solely caused by the inclusion of an increasing proportion of hip fractures (because DXA predicts hip fracture with higher gradients of risk than other fractures).

Time since measurement

For any fracture, the predictive ability of BMD remained stable as time elapsed after measurement. For osteoporotic fracture there was a small attenuation of gradient of risk with time after assessment and for hip fracture the effect was more marked, but in neither instance was this significant. Data for hip fractures are shown in Table 5.

Table Table 5.. Gradient of Risk for Hip Fracture (RR/SD) With Time Since Assessment in Men and Women Combined
original image

Baseline BMD for all fracture outcomes

The gradient of risk per SD decrease in BMD for any fracture decreased significantly with increasingly worse z scores. For hip fracture, the effect was modest and not significant, whereas for any fracture or any osteoporotic fracture the effect was more pronounced (Table 6).

Table Table 6.. Risk Ratio (RR) for 1 SD Change in z Score in Men and Women Combined
original image


The information from the two cohorts with available data on ultrasound (Sheffield and EPIDOS) is summarized in Table 7. For BUA, the gradient of risk was significant for fracture when not adjusted for BMD. The gradient of risk for hip fracture decreased significantly with age (p < 0.01) in a manner qualitatively similar to that seen for hip fracture risk prediction with DXA. At the age of 60 years, the risk ratio per SD decrease was 3.82 (95% CI = 2.05-7.12), and fell to 2.00 (95% CI = 1.69-2.36) at the age of 80 years.

Table Table 7.. Gradients of Risk (RR/SD) for BUA and SOS Using Quantitative Ultrasound in Women With and Without Adjustment for BMD at the Hip
original image

Risk prediction for hip fracture was attenuated with adjustment for BMD, but BUA was still a significant independent risk indicator (Fig. 2). For SOS-VOS, the gradient of risk was somewhat lower than that provided by BUA and was downward adjusted from 1.50 to 1.23 with hip BMD in the model (see Table 7).

Figure FIG. 2..

Hip fracture prediction (risk ratio RR/SD) with broadband ultrasound attenuation in men and women by age.

Peripheral BMD

The performance characteristics of peripheral BMD measurements are also shown in Table 7. The risk ratios were marginally higher in men compared with women, but differences were not significant. In contrast with DXA at the hip, the gradient of risk was similar for any fracture, any osteoporotic fracture, and for hip fracture alone. As was the case for DXA, the gradient of risk was significantly higher at younger ages for the prediction of hip fractures.


A large number of studies have examined the relationship between BMD measured at various sites and fracture risk, and the available data were summarized by meta-analysis in 1996.(5) With DXA measured at the hip, fracture risk increased 2.6-fold per SD decrease in BMD, and the risk of any fracture increased 1.6-fold per SD. Since then, many additional studies with longer follow-up times have shown broadly similar findings.(9,39–47)

This analysis derives similar values as found in the previous meta-analysis and subsequent studies. However, an advantage of this study is that the meta-analysis is based on individual data from each cohort and not on published values with summary statistics. This eliminates publication bias and has the advantage that important potential interactions between BMD and fracture risk could be examined. This study also covers several areas of the world, including cohorts from North America, Europe, Australia, and Asia. The homogeneity of the data suggest that our findings can be applied generally, at least in the developed world. In contrast, the data we show for ultrasound and peripheral BMD measurements are from a much more restricted sample.

Age at assessment was an important factor that affected the predictive value of femoral neck BMD. For the prediction of osteoporotic fracture, there was a modest and not significant increase in predictive power with age. This might be expected because osteoporotic fractures include hip fractures, which increase with age and are predicted more accurately by BMD at the hip than other fractures. The exclusion, however, of hip fractures in the analysis did not markedly alter the relationship with age. A much more marked and significant relationship was found between the gradient of risk for hip fracture. Gradients of risk (RR/SD) were higher at younger ages than in the elderly. Similar findings have been reported in the EPIDOS study, with a lower predictive ability of DXA for hip fractures in women at the higher ages.(39) The gradient of risk for hip fractures with a T score of -2.5 SD or less was 4.4 (3.6-5.5) at ages <80 years and 2.5 (2.0-3.1) at ages of ⩾80 years. In the Study of Osteoporotic Fractures (SOF), differences with age were less marked. For hip fractures in women assessed by DXA at the hip, the gradient of risk between the ages of 65-79 years was 2.9 (2.2-3.9), and in women ⩾80 years was 2.1 (1.4-3.2).(48)

The reason for the decreasing gradient of risk for hip fracture with age is not known. It is possible that extraskeletal risk factors, such as liability to falls, affect the gradient, but this seems unlikely because most falls do not cause hip fracture. A more plausible hypothesis is that age adversely affects the structural or material properties of the femur.

The decreasing gradient of hip fracture risk with age should not be misinterpreted to infer a decrease in absolute risk with age. Indeed, at any given age, 10-year hip fracture probability increases with decreasing T score (Table 8). The probabilities in Table 8 are also compared with those previously published assuming a fixed gradient of risk (2.6/SD).(8) Compared with a fixed gradient of risk, probabilities are underestimated in the young and overestimated in the elderly.

Table Table 8.. Ten-Year Probability of Hip Fracture (%) at the Hip in Men and Women From Sweden According to Age and T Score for BMD
original image

For example, an individual with a z score of −2 SD at the age of 50 years would have a hip fracture risk that was 13 times that of an individual with an average BMD for age (Table 4), whereas the risk would be increased only 5-fold at the age of 80 years. This may be fortuitously advantageous in the assessment of risk. In younger patients, greater reliance can be placed on BMD, whereas in the elderly, risk can be additionally assessed by clinical risk factors, many of which increase in frequency with age.

A concern with long-term prediction is that predictive value of risk factors may decrease with time because of variable rates of bone loss with time. A previous study by Kanis et al.(12) modeled the decrease in the predictive ability of BMD with time after measurement, a phenomenon that has been supported by some but not all empirical studies.(13,49,50) Huang et al.(49) showed a nonsignificant reduction in predicting vertebral fractures by time after measurement. After 2.7 years of follow-up, the relative risk per SD decrease in BMD was 1.7 at the forearm and 1.8 at lumbar spine. During a follow-up of 8 years, the gradients of risk were 1.5 and 1.7, respectively. A reduction by time after measurement was also observed by Düppe at al.(50) With single photon absorptiometry at the forearm, the gradient of hip fracture risk was 1.86 (1.01-3.44), and after 15-25 years was 1.31 (0.79-2.17) per SD decrease. In this study, we showed that for hip fracture there was a slight attenuation in the predictive ability for hip fractures the longer the follow-up, whereas this was not the case for any fracture. A similar result was found by Melton et al.,(13) where femoral neck BMD predicted the risk of osteoporotic fracture as well in the first 10 years of follow-up (HR = 1.38; 1.10-1.74) as in the subsequent 10 years (HR = 1.39; 95% CI = 1.05-1.84). Because the decrease in predictive ability was rather small, it will not markedly affect the computation of 10-year fracture probability that has been proposed as an index of risk in assessment.(14,51)

A more important interaction was the effect of baseline BMD measurement itself on predictive value. Gradients of risk were lower the higher the z score. The effect was marked for the prediction of osteoporotic fracture or any fracture and was also evident for hip fracture prediction, but was not statistically significant. Reasons for this are conjectural but might be related to the lower body mass index that is associated with lower values of BMD, and also perhaps associated with co-morbidity, muscle weakness, or less padding with fat to protect against injury.(18,19,52) In addition, lower BMD values may be associated with structural changes in bone (e.g., increased bone area but thinner cortices or higher rates of bone remodeling) that reduce resistance to fracture.(53) Irrespective of the mechanism, the effect is large so that account should be taken of this for the optimal assessment of patients.

There has been some debate as to whether the gradient of risk of BMD for fracture is the same or differs between men and women.(54) Differences are described in selected cohorts, or when gender-specific SDs are used.(11,55) In this study, we found no significant differences between men and women as has been found in other population-based samples.(9,10) Indeed, the age-specific risk of hip fracture at a given hip BMD in men seems to be the same in women with the same BMD and age. When incidence and mortality are combined to give long-term probabilities, these are also similar for men and women for a given BMD value, except for the highest ages where the probability of fracture is higher in women because of their lower mortality.(8)

We conclude that assessment of hip BMD provides a strong risk indicator for fracture risk that is largely independent of sex. Its predictive value is not significantly attenuated with time after assessment over a 10-year interval, suggesting that it can be used to compute long-term fracture probabilities. The significance of BMD as a risk factor depends on the level of BMD when used to predict any fracture or any osteoporotic fracture. In addition, the gradient of risk for hip fracture is higher in the younger individuals. These characteristics need to be accounted for to make best use of BMD in a clinical setting.


We are grateful to the Alliance for Better Bone Health, Hologic, IGEA, Lilly, Lunar, Novartis, Pfizer, Roche, Wyeth, and the EU (FP3/5; QLK6-CT-2002-00491) for supporting this study and the International Osteoporosis Foundation, the International Society for Clinical Densitometry, and the National Osteoporosis Foundation for their unrestricted support of this work.

Dr Eisman serves as a consultant and receives support from Aventis, Eli Lilly and Company, MSD, Novartis, NPS Pharmaceuticals, Organon, Roche, and Servier. Dr Meunier serves as a consultant for Aventis, Eli Lilly and Company, Merck, Nycomed, Procter & Gamble, and Servier. Dr Pols serves as a consultant for Amgen, Merck, and Organon. All other authors have no conflict of interest.