CLINICIANS HAVE ALWAYS BEEN MOVED by the suffering of their patients to develop better approaches to disease management. This is certainly true in the case of osteoporosis where, for example, it may prove difficult to ameliorate the chronic symptoms of a woman suffering from multiple vertebral compression fractures. With a view to preventing such fractures, there has been an explosion of diagnostic and therapeutic technologies in recent years, although many important questions (and exciting opportunities) remain to be explored. However, funding for the necessary research, and for the clinical application of these advances, depends upon societal recognition of the importance of the problem. Authoritative opinions on the subject from leaders in the field are not sufficient to assure this. Instead, data are needed concerning the frequency, cost, and societal impact of osteoporosis and its attendant fractures.
With this in mind, a working group of the World Health Organization (WHO) developed the now-familiar operational definition of osteoporosis, i.e., a bone mineral density (BMD) level more than 2.5 standard deviations (SD) below the young normal mean.1 Some have criticized this approach because it does not lend itself to treatment decisions for individual patients: clinicians may feel that prophylaxis is indicated for patients who do not yet have osteoporosis or even low bone mass, while others may be disinclined to aggressively treat some elderly patients even though they have severe osteoporosis. The need has been for treatment thresholds that, while based on bone density measurements, recognize that other factors such as life expectancy and various risk factors also contribute to fracture prediction. This gap is being filled in part by analyses from the National Osteoporosis Foundation that define subsets of patients in whom various treatments are likely to be cost effective.2 Because specific treatments have different long-term risks and benefits, however, the treatment threshold is unique for each one. This makes perfect sense clinically but it is not possible on this basis to define the prevalence of osteoporosis, which would then rise or fall depending upon the therapy that was envisioned. Thus, to serve the social need, if not to treat patients, it is important to have a more or less consistent definition of osteoporosis in order to document the frequency of the condition for policymakers.
Even if one accepts the WHO definition of osteoporosis, there have been serious limitations with respect to available data. Most important is uncertainty about the prevalence of osteoporosis among men and nonwhite women. Recent estimates indicate that 18% of the cost of osteoporotic fractures in the United States in 1995, or $2.5 billion, was spent treating white men and another 7%, or $900 million, was spent on nonwhite women and men.3 Even among white women, who accounted for 75% of total expenditures, or $10.3 billion, only rough estimates of osteoporosis prevalence have been available, as projected from a small population-based sample of Rochester, Minnesota women.4 The availability of high quality data from a probability sample of the United States population has changed that. Based on data from the first phase of the third National Health and Nutrition Examination Survey (NHANES III), Looker and colleagues determined that the prevalence of osteoporosis of the total hip was 21% among white women compared with 16% among Hispanic women and only 10% among African-American women.5
In this issue of the Journal, these analyses have been enlarged and extended to men.6 Using additional data from the second phase of NHANES III, prevalence rates for osteopenia and osteoporosis of the proximal femur were recalculated for the three groups of American women. For the first time, prevalence rates were also estimated for men. Because it is not clear what standard should be used, two approaches were taken. Based on a definition of osteoporosis using the same absolute bone density cutoff level for men as for women (e.g., 0.56 g/cm2 for femoral neck BMD), the prevalence rates for white, Hispanic, and African-American men age 50 years and over were 4, 2, and 3%, respectively. The lower prevalence in men is consistent with their lower risk of hip fracture7 and reflects the fact that total hip bone density is 12–13% greater in men.8 Because of their larger skeletons, however, areal bone density may be overestimated in men.9 Consequently, osteoporosis prevalence rates were also calculated on the basis of bone density levels more than 2.5 SD below the mean for young men (e.g., a cut-off level of 0.59 g/cm2 for femoral neck BMD). The higher mean value caused the prevalence rates for white, Hispanic, and African-American men to increase to 7, 3, and 5%, respectively. The modest size of this change reflects the fact that bone loss in older men is not as rapid as in older women8 so that the osteoporotic patients are derived disproportionately from the relatively small population of very elderly men.
The NHANES data enjoy several notable strengths. First, there are few, if any, other examples of bone densitometry data sampled from ages, genders, ethnic groups, and geographic regions representative of an entire nation. While the response rate of 63% was not as high as obtained in some Scandinavian studies, the authors did not find many important differences between eligible subjects who did or did not volunteer.6 However, frail or demented persons are no doubt underrepresented among the study subjects. Exclusion of such individuals is inevitable in any prospective study involving volunteers10 but suggests that the osteoporosis prevalence rates estimated for the oldest age groups are conservative. A second major strength is the large sample size, with bone density measurements on over 6000 individuals aged 50 years and older. Even that was not sufficient to assure an adequate sample of some groups, e.g., Asian-Americans. However, the pattern of age-related bone loss from the proximal femur in Asian men and women resembles that in whites.11 Finally, the NHANES measurements were made using accurate and precise dual-energy X-ray absorptiometry (DXA) technology, with good quality control, and they were made at the proximal femur. Such measurements have been shown in prospective studies to be the best predictor of hip fracture,12 the osteoporotic fracture with the greatest consequences in terms of morbidity, mortality, and cost.7
As Looker and colleagues point out, however, there are also a number of important issues that remain to be resolved.6 One is the issue of peak bone mass. The WHO approach to assessing osteoporosis is based on the notion that the biomechanical competence of the skeleton is optimal among young adults but is degraded as bone loss ensues with aging. Thus, bone density later in life is evaluated with respect to mean levels among normal young people. There is evidence from prospective studies that bone loss from the proximal femur begins in the early 20s13,14 so Looker and colleagues based their prevalence rates on the mean bone densities (and standard deviations) seen among 20- to 29-year-olds.6 However, cross-sectional data show no loss of bone from the distal forearm until the time of menopause.15,16 Findings from the lumbar spine are less clear but seem to indicate continued bone gain until the 30s.13,17 Thus, there may be no age at peak bone mass that applies consistently to all skeletal sites. Consequently, the estimated prevalence of osteoporosis at any site (hip, spine, or forearm) among the Rochester women was based on mean bone density levels for all adult premenopausal women combined.4 Such a choice results in a somewhat lower mean bone density for the “normal” population and a lower estimate of the prevalence of osteoporosis.
There is also some controversy about the proper skeletal site to measure when assessing osteoporosis risk. This dispute again results from failure to distinguish the clinical from the public health perspectives on osteoporosis. The clinician may want to measure bone density in the spine, especially in perimenopausal women and patients who present with vertebral fractures. This is because of concern that such patients may have accelerated loss of cancellous bone from the spine that is not reflected in bone density measurements made at the hip. Health policymakers, however, may believe that it is important to hold down screening costs for a condition as widespread as osteoporosis, that bone density measurements at the hip predict future fractures as well as other measurements and predict hip fractures better, and that hip fractures represent the manifestation of osteoporosis that is most important to control from a societal perspective. Thus, it is entirely appropriate as Looker and colleagues have done to focus on the prevalence of osteoporosis of the proximal femur while recognizing that there may be clinical indications for measuring bone density at other skeletal sites.
The most important issue, however, is whether the −2.5 SD cutoff level should be used to diagnose osteoporosis in men and women of other races, where the relationship between bone density and fracture risk may differ from that in white women. Thus, the lower risk of hip fractures among Asian women compared with white women has been attributed to a reduced risk of falling,18 particularly since bone density among Asian women is not higher19 and vertebral fractures are as frequent as in white women.20 By contrast, lower hip fracture risk among African-American women has been attributed to greater bone mass and different biomechanical characteristics of the proximal femur.21 Finally, some vertebral fractures in men are blamed on excessive occupational loading of the spine rather than osteoporosis,22 whereas distal forearm fractures in men never increase with age like they do in women7 despite the fact that elderly men lose bone from the forearm as they age23 and have quite a high risk of falling.24 The fact that bone size independently influences fracture risk25 adds an additional complication.
The most pressing need at this point, then, is for direct assessment of the relationship between bone density and the risk of various fractures in men and women of different races in order to validate a diagnostic threshold at −2.5 SD in these other groups. It is already clear from epidemiologic data that osteoporotic fractures represent a threat to each of these groups,7 but such studies might indicate that different definitions of osteoporosis would be more useful in some situations. If so, new prevalence estimates could be generated from the NHANES data already available. However, Looker and colleagues have provided an important service by documenting that osteoporosis is a serious problem in the United States—by any definition.