Who Has Osteoporosis? A Conflict Between Clinical and Public Health Perspectives


  • L. Joseph Melton III

    Corresponding author
    1. Department of Health Sciences Research, Mayo Clinic and Mayo Foundation, Rochester, Minnesota, U.S.A.
    • Address reprint requests to: Dr. L. Joseph Melton III, Section of Clinical Epidemiology, Department of Health Sciences Research, Mayo Clinic, 200 First Street S.W., Rochester, MN 55905, U.S.A.
    Search for more papers by this author


NEARLY A decade ago, our group published a Perspective in this journal entitled, “How Many Women Have Osteoporosis?”(1) The article pointed out that the size of the osteoporosis problem could be addressed either with respect to the number of people with low bone mineral density (BMD), a useful in vivo surrogate for “abnormalities in the amount and architectural arrangement of bone tissue leading to impaired skeletal strength,”(2) or in terms of the number of people experiencing characteristic osteoporosis-related fractures, especially those of the hip, spine, or distal forearm. In either instance, a precise delineation of the affected population was limited by the data available at that time, which were restricted almost entirely to white women. However, since that report, an enormous amount of new information has been published about this important problem. In particular, there are now refined estimates of the lifetime risk of osteoporotic fractures that take account of the rising incidence rates observed in some populations as well as improvements in life expectancy. Even more important has been development by the World Health Organization (WHO) of an operational definition of osteoporosis based on bone density(3) that has fostered better estimates of the prevalence of osteoporosis, including a wealth of new data on the prevalence of osteoporosis in men and nonwhite women. On the other hand, the advent of novel densitometric techniques has revealed significant discrepancies in the prevalence estimates depending on which technology is used, and this, in turn, has raised questions about how osteoporosis should be defined.(4) The resulting controversy has exposed a conflict between the need to document the magnitude of the public health problem represented by osteoporosis as opposed to the clinician's need to assess fracture risk in individual patients. Vitriolic attacks on the WHO criteria are a reflection of this growing conflict in perspectives.(5)


The public health impact of osteoporosis relates almost entirely to the associated fractures that are the clinical manifestation of the problem. Osteoporotic fractures initially were identified on epidemiological grounds as those in which the incidence increased dramatically with age, affected women more than men, and disproportionately involved diaphyseal skeletal sites.(6) On that basis, fractures of the proximal femur, vertebrae, and distal radius were linked strongly to osteoporosis,(1) broadening a traditional clinical focus on osteoporosis-related vertebral fractures.(7) However, as population-based densitometric studies subsequently made clear, almost all fractures among elderly women are associated with low bone density(8) and bone density also is related to fracture risk in men.(9,10) Although any fracture can have a devastating impact on the affected individual, hip fractures are by far the most important from the public health perspective. They are the predominant cause of deaths from osteoporosis-related fractures,(11,12) and they are the main source of osteoporosis-related morbidity.(13) An estimated 10% of women who sustain a hip fracture become functionally dependent in the activities of daily living compared with only 4% of those with a vertebral fracture and 1% of women with a distal forearm fracture.(14) Moreover, hip fractures account for the lion's share of medical costs.(15) For example, direct expenditures for osteoporotic fracture care in the United States in 1995 were estimated at $13.8 billion, and hip fractures alone were responsible for 63% of the total. (16)

As a result of this disproportionate impact, it is appropriate for health authorities to focus on hip fracture as the primary measure of osteoporosis. Indeed, one of the objectives of Healthy People 2000 was to reduce annual hip fracture incidence in elderly women and men from 714 per 100,000 in 1988 to 607 per 100,000 by 2000.(17) Unfortunately, the rate had risen to 934 per 100,000 in 1996,(18) when 390,000 hip fractures were observed in the United States,(19) and the great concern is that growing numbers of elderly people will drive the number of hip fractures, and their associated costs, even higher in the future.(20) On a global basis, the 323 million individuals 65 years of age and over in 1990 will grow to an estimated 1555 million by 2050, and this demographic trend alone could cause the number of hip fractures worldwide to increase from an estimated 1.7 million in 1990 to a projected 6.3 million in 2050.(21) Although hip fracture incidence has stabilized in some countries,(22,23) increases are still occurring in many areas of the world,(24) and any rise in the incidence rates will increase future fractures still further. Assuming a 1% annual increase in age-adjusted incidence, the projected number of hip fractures worldwide in 2050 could be 8.2 million; if incidence rates stabilized in Europe and North America but increased by 3% annually in the rest of the world, the total could exceed 21 million.(24) Even this alarming figure could be conservative because life expectancy also is rising worldwide. Our estimated lifetime risk of hip fracture (17.5% in white women and 6.0% in white men) was based on a life expectancy in women and men of 78.9 years and 72.3 years, respectively.(1) In a similar analysis, the lifetime risk of hip fracture in Swedish women and men was 13.9% and 4.6%, respectively, but this rose to 22.7% and 11.1%, respectively, when projected improvements in mortality were accounted for.(25) If, in addition, age-adjusted hip fracture incidence rates were to rise by just 1% annually in Sweden, the lifetime risk could increase to an incredible 34.9% in women and 17.0% in men.

If hip fractures are most important from a public health perspective and if BMD assessed at the proximal femur is the best predictor of hip fracture risk,(26) then it also is appropriate for health authorities to focus on osteoporosis of the hip to the exclusion of other sites. Based on an older definition of osteopenia as a BMD value of more than 2 SD below the young normal mean, we estimated that 29% of postmenopausal white women might be affected at the proximal femur.(1) However, the WHO definition of osteoporosis requires a BMD value more than 2.5 SD below the young normal mean for white women.(3) Using this definition, about 17% of postmenopausal white women in the United States have osteoporosis at the total hip site based on data from the Third National Health and Nutrition Examination Survey (NHANES III), a large probability sample of the United States population.(27) The proportion varies somewhat when bone density is assessed at the different subregions, ranging from 13% with osteoporosis at the trochanteric region to 20% at the femoral neck. This translates to about 6 million white women in this country with osteoporosis at the femoral neck and another 15 million with low bone mass, that is, BMD more than 1.0 SD but less than 2.5 SD below the young normal mean.(27) An even greater number of women are affected when additional skeletal sites are considered. Based on data from Rochester, MN, the prevalence of osteoporosis at the proximal femur, lumbar spine, or total wrist sites among postmenopausal white women was 35%.(9)

Because of insufficient data about the relationship between BMD and fracture risk in men or nonwhite women, the WHO did not offer a definition of osteoporosis for these other groups,(3) even though together they account for a substantial proportion of the cost of osteoporotic fractures.(16) However, women of all races lose bone from the proximal femur in a similar fashion.(28,29) If normative data for white women are used to define osteoporosis for nonwhite women, then the prevalence of the condition appears to be lower in some of the other groups. For example, only 12% of Hispanic women and just 8% of postmenopausal black women have osteoporosis as assessed at the total hip site by NHANES.(27) The prevalence of osteoporosis is not well established among women of Asian heritage, but their BMD levels appear to be lower than those in white women.(30,31) However, it has become clear that greater bone density in black women compared with white women and white women compared with Asian women is partly artifactual. This is because areal BMD (g/cm2) corrects for the area scanned but does not completely account for the fact that wider bones also are thicker.(32) Thus, areal BMD values are confounded by skeletal size and, when body size is adjusted for, race-specific differences in bone density are reduced or eliminated.(31, 33–36) This might be expected from the fact that the prevalence of vertebral fractures seems similar in women of different races.(37,38) The risk of hip fractures is greater among white women, but lower BMD levels among Asian women would have predicted higher hip fracture incidence rates in that group, not the lower rates actually observed.(39,40) This suggests that lower hip fracture rates in Asian women are caused by some other factor besides bone density, such as biomechanical differences(30) or a reduced risk of falling.(41)

There are few normative data with which to assess the prevalence of osteoporosis in men. Based on the same absolute bone density cut-off level for men as for women (femoral neck BMD below 0.56 g/cm2), the prevalence of osteoporosis among white, Hispanic, and black men age 50 years and over was 4, 2, and 3%, respectively, in the NHANES study.(27) The lower prevalence in men is consistent with their lower risk of hip fracture(6) and reflects that fact that hip bone density is 12-13% greater in men.(29) However, because of their larger skeletons, bone density is overestimated in men relative to women.(32) When osteoporosis prevalence was calculated on the basis of BMD levels more than 2.5 SD below the mean for young men (femoral neck BMD below 0.59/cm2), the higher mean value caused the figures for white, Hispanic, and black men to increase to 7, 3, and 5%, respectively.(27) However, in Rochester, use of the same absolute cut-off values for defining osteoporosis in men as in women produced an estimated prevalence of osteoporosis of the hip, spine, or distal forearm of only 3%,(9) which appears much too low if the lifetime risk of osteoporotic fractures in men of about 13% is taken as a rough benchmark.(1) Sex-specific normal values produced an estimated prevalence of osteoporosis in men from Rochester of 19%. Again, however, adjusting for bone size by calculating bone mineral apparent density (BMAD; g/cm3) greatly reduces the sex-specific differences.(42,43) Comparable lumbar spine BMAD levels are consistent with recent evidence suggesting that the prevalence of vertebral fracture is similar in men and women,(44) but hip fracture incidence rates are still twice as high in women as in men(6) despite comparable femoral neck BMAD levels in the two sexes.(43)


Although public health authorities may be justified in emphasizing hip fractures, clinicians have an equal obligation to deal with the adverse effects on their patients of all fractures. These other osteoporotic fractures, exclusive of the hip, account for over 2 million physician outpatient visits each year in the United States, along with 600,000 emergency room encounters.(16) In aggregate, these other fractures are more common than hip fractures. The lifetime risk of a clinically evident vertebral fracture in white women and men from age 50 years onward is about 15.6% and 5.0%, respectively, whereas that of distal forearm fracture is 16.0% and 2.5%, respectively(1); the lifetime risk of all other fractures in white women has been estimated at 31%.(45) There is some evidence that the incidence of these other fractures also has been increasing over time.(46) Few forearm fracture patients are disabled as a result of the fracture(14) but severe vertebral fractures, which affect up to 10% of postmenopausal white women, (47–50) may cause disfigurement, persistent pain, and a greatly reduced quality of life.(51) Indeed, the adverse influence of vertebral fractures on most activities of daily living is almost as great as that seen for hip fractures, and even wrist fractures interfere significantly with activities such as meal preparation.(52) Finally, there is growing evidence that vertebral fractures, though not distal forearm fractures,(11) are associated with an increased risk of death.(53)

If the focus is on fractures in general, rather than on hip fractures alone, then the various densitometric measurements seem to perform comparably in predicting fracture risk.(26) However, as newer techniques were used more widely, it became obvious that the estimated prevalence of osteoporosis, using the WHO definition, can vary substantially from one skeletal site or technology to another.(54,55) Although the decision to treat a patient involves more considerations than just BMD,(56) and the WHO definition was not meant to be used as an unequivocal treatment threshold,(57) this has become widespread practice.(58) Consequently, discrepancies in osteoporosis prevalence appear to imply important differences in the proportion of patients for whom treatment might be indicated. (59–62) For example, in one population of patients, 17% would have been classified as osteoporotic if BMD was measured at the femoral neck compared with only 6% if an intertrochanteric measurement had been obtained and 34% had BMD been assessed at Ward's triangle.(60) This discrepancy relates partly to the difference in SDs, which can be quite wide for less precise techniques. Thus, only 3% of 60-year-old white women have heel ultrasound measurements more than 2.5 SD below the young normal mean for that technology compared with 38% with lateral spine dual-energy X-ray absorptiometry (DXA) measurements below this level.(63) However, there also are different patterns of age-related bone loss at the various skeletal sites, and this has an influence both on the choice of a “normal” population for setting standards and on the apparent prevalence of osteoporosis later in life.(43) Finally, discrepancies in the normative data used by different densitometer manufacturers create additional problems. (64–66)

These problems matter little to public health authorities, whose focus is on the population rather than the individual patient. Public health interventions (e.g., adequate nutrition and exercise) are applied broadly, without assessment of or reference to individual risk, and are directed at social and environmental factors most likely responsible for differences in fracture risk within populations. (67–70) By contrast, the clinical approach to osteoporosis management focuses on the identification and treatment of high-risk individuals.(71) In this respect, any specific bone density level, much less the WHO definition, is inherently limited.(57) Fracture risk varies continuously with BMD,(3) and there are numerous predictors of fracture risk that are independent of bone density. In the most comprehensive report, an exhaustive set of potential risk factors for hip fracture was evaluated in a prospective study of 9516 white and Asian-American women.(72) The independent predictors of fracture risk in a multivariate analysis adjusting for age and calcaneus BMD included maternal history of hip fracture, weight gain since the age of 25 years (protective), greater height, poorer self-rated health status, history of hyperthyroidism, use of long-acting benzodiazepines, current caffeine intake, more hours of standing each day, inability to rise from a chair, impaired depth perception or contrast sensitivity, and resting heart rate more than 80 beats per minute. Although other investigators have identified different sets of risk factors,(73) the point to be made is that hip fracture incidence was 17 times greater among 15% of the women who had five or more risk factors, exclusive of bone density, compared with 47% of the women who had two risk factors or less.(72) However, women with five or more risk factors had an even higher risk of hip fracture if their bone density Z score was in the lowest tertile.

How is this information to be used by the practicing clinician? Bone densitometry is essential for guiding therapies directed specifically at skeletal metabolism, but the prediction of fracture risk in the individual patient might still be sharpened by incorporating clinical risk factors and/or by adding a biochemical measurement of bone turnover to complement the bone density data. Moreover, if absolute fracture risk could be estimated accurately, and a clinically significant level of risk could be agreed on,(57) there would be no need for reference to age-, race-, or gender-specific normal databases or any concern about T scores (other than the not inconsiderable fact that reimbursement polices often are based on T scores). Admittedly, the practical difficulties involved are formidable. In particular, it may be difficult to achieve consensus on the level of absolute fracture risk that should be of clinical concern and to adjust payment policies accordingly. Nonetheless, methods for accomplishing this task are now being explored. In one example, a model was constructed to predict absolute fracture risk over the succeeding 5 years for postmenopausal white women depending on their age, hip BMD level, and the presence or absence of several risk factors, including previous fracture history in the patient (after the age of 40 years), low body weight (≤127 lb), cigarette smoker (current), and family history of fractures (mother or father with hip, spine, or forearm fracture ≥ age 50 years); treatment appeared to be more cost-effective at any given BMD level when additional risk factors were present, that is, when the absolute fracture risk was greater.(45) Unique, cost-effective treatment thresholds such as these, if they can be simplified sufficiently for everyday application, can account for the fact that specific therapies have different long-term risks, benefits, and costs, and their use could help ensure that treatment with potent pharmacologic agents is rewarded with commensurate social benefits in terms of reducing the adverse outcomes of osteoporotic fractures.


It is clear from the previous discussion that there is no simple answer to the question, Who has osteoporosis? From the clinical perspective, this probably is not even the right question. The clinician is faced with the decision of whether or not to treat and, if so, with what agent. In addition to balancing the risks and benefits of a specific intervention, in conjunction with the patient's preferences, the clinician needs to know the patient's fracture risk. It does not matter so much in this instance whether or not the patient has osteoporosis by WHO criteria. Indeed, the clinician needs sufficient latitude to initiate treatment to prevent the development of osteoporosis in patients who do not already have it or, in some elderly individuals with osteoporosis, to withhold aggressive treatment that might be counterproductive. Instead, what is required is an estimation of the likelihood of fracture over the coming 5-10 years, that is, absolute fracture risk,(57) and efforts are underway to develop this information. However, individual fracture risk and drug-specific treatment thresholds do not provide a basis for the diagnosis of existing osteoporosis, which is sometimes needed for reimbursement of bone densitometry, and neither speaks to the social burden of osteoporosis. Indeed, the prevalence of osteoporosis would appear to vary widely depending on the type of therapy that was envisioned if the definition was linked to the proportion of the population in whom a specific treatment was indicated. Although the WHO definition of osteoporosis is arbitrary, it is well established. From the public health perspective, then, the answer to the question of who has osteoporosis probably can best be addressed by continuing to assess bone density levels in the proximal femur.(57)


The author thanks Mrs. Mary Roberts for help in preparing the manuscript. This study was supported by grants AR27065 and AG04875 from the National Institutes of Health, U.S. Public Health Service.