We evaluated different definitions of osteoporosis in a population-based sample of 348 men (age 22–90 years) compared with 351 women (age 21–93 years). Thirty-six men (10%) and 46 women (13%) had a history of osteoporotic fracture (hip, spine, or distal forearm due to moderate trauma at ≥ age 35). In logistic regression analysis, osteoporotic fracture risk was associated with bone mineral density (BMD) at all sites (neck, trochanter, total hip, lumbar spine, and total wrist) in both genders (p < 0.001) except spinal BMD in men. After adjusting for age, total hip BMD was the strongest predictor of fracture risk in women (odds ratio [OR] per 1 SD decline, 2.4; 95% confidence interval [CI], 1.6–3.7), while wrist BMD was best in men (OR, 1.5; 95% CI, 1.1–2.0). Among men but not women, bone mineral apparent density (BMAD) was a better predictor of fracture than BMD (wrist BMAD OR, 1.7; 95% CI, 1.3–2.3). Hip BMD/BMAD decreased linearly from age 20 years onward in both genders, while spinal BMD/BMAD declined after age 40 in women but not in men. In both genders, total wrist BMD/BMAD decreased after age 50. By World Health Organization criteria, the age-adjusted prevalence of osteoporosis at the hip, spine, or wrist was 35% among women ≥50 years of age. A similar approach (BMD > 2.5 SD below the young male mean) produced an osteoporosis prevalence rate in men ≥50 years of age of 19%. Thus, bone density predicts fracture risk in men as it does in women, and the prevalence of osteoporosis in men, using sex-specific normal values, is substantial. These observations indicate a need for better prevention and treatment strategies for men.
In 1994, a study group of the World Health Organization (WHO) proposed that osteoporosis in white women be operationally defined as a bone mineral density (BMD) level 2.5 SD below the mean for normal young women.(1) It was recognized that this cut-off level was an arbitrary choice given the continuous relationship between bone density and fracture risk in women, but the definition produced a number of affected women that was consonant with clinical perceptions of the size of the problem.(2) It was not possible at that time to define osteoporosis in men or nonwhite women due to the dearth of detailed data on the relationship between bone density and fracture risk in these other populations. However, it was suggested that several possibilities for defining osteoporosis in men should be explored: 2.5 SD below the young normal mean for men, 3–4 SD below the male mean, and 2.5 SD below the young normal mean for women.(2) It was thought that the first choice, a gender-specific cut-off level at −2.5 SD might produce a number of affected men much larger than the proportion of the male population that ultimately experiences osteoporotic fractures, estimated at 13% for the lifetime risk of a hip, spine, or forearm fracture.(3) This problem could be dealt with either by use of a more stringent cut-off level to define osteoporosis in men (e.g., BMD 3–4 SD below the mean for young men) or, alternatively, by use of the same absolute value of BMD for men as previously chosen to define osteoporosis for women. At present, however, it is not clear which of these might be the better choice. The objective of the present study was, first, to determine the association between bone density assessed in various ways with fracture risk in white men and then to define normal values and patterns of age-related bone loss in order to estimate the prevalence of osteoporosis in men using these different definitions.
MATERIALS AND METHODS
Following approval by the Mayo Clinic's Institutional Review Board, subjects were recruited from an age-stratified random sample of Rochester, Minnesota residents that was selected using the medical records linkage system of the Rochester Epidemiology Project.(4) Over half of the Rochester population is attended annually at the Mayo Clinic, and the majority are seen in any 3-year period. Thus, the enumerated population (Rochester women seen in 1990 ± 1 year and men seen in 1991 ± 1 year) approximates the underlying population of the community, including both free-living and institutionalized individuals. Altogether, 1138 men were approached for this study, but 239 of them were ineligible (109 were demented; 13 were radiation workers; 91 died before they could be contacted; and 25 others were so debilitated, e.g., terminal cancer, that they were not approached; the final ineligible man was involved in a legal action). Of the 899 eligible men, 348 (39%) participated and provided full study data. There were ∼50 men per decade of age from 20–29 years to age 80 years and over (mean age ± SD, 55.4 ± 19.6 years; range 22–90 years), but the participation rate varied by decade (27, 36, 43, 50, 62, 45, and 29%, respectively, for age groups 20–29 through 80 years and over).
A total of 938 women aged 20 years and over were approached for this study, but 126 were ineligible (89 were demented and could not give informed consent; 11 were pregnant; 9 were radiation workers; 8 were participants in an ongoing clinical trial of osteoporosis prophylaxis; and 9 died before they could be contacted). Of the 812 eligible women, 351 (43%) participated and provided full study data. There were ∼50 women per decade of age from 20–29 years to age 80 years and over. The participation rates by age group were 50, 48, 56, 65, 57, 39, and 22%, respectively. The subjects included 138 premenopausal women (mean age, 35.0 ± 8.6 years; range, 21–54 years) and 213 postmenopausal women (mean age, 67.8 ± 13.2 years; range, 34–93 years).
All subjects were interviewed in accordance with a standard protocol in order to collect clinical data, including a history of fractures that was verified by review of each subject's complete (inpatient and outpatient) medical records in the community. Mayo Clinic records, for example, contain the details of every inpatient hospitalization at its two large affiliated hospitals in Rochester (Saint Marys and Rochester Methodist), every outpatient or office visit at the Clinic, emergency rooms or nursing homes, as well as all radiographic reports and pathology reports, including autopsies.(4) The records contained the clinical history and the radiologist's report of each fracture, but the original roentgenograms were not available for review. Consequently, the diagnosis of vertebral fracture was accepted on the basis of a radiologist's report of compression, wedging, or collapse of one or more thoracic or lumbar vertebrae. The interview and record review were completed independently of any knowledge of each subject's BMD values. There was generally good agreement between interview and medical record data,(5) but where disagreements occurred priority was given to documented medical history; in the absence of documentation to the contrary, the subject's account was accepted. The duration of contemporary documentation in hand averaged 30.8 years (median, 29 years; range, 1–81 years), and ascertainment of clinically evident fractures is believed to be complete. Osteoporotic fractures were defined as clinically recognized fractures of the hip, spine, or distal forearm that resulted from minimal or moderate trauma (e.g., a fall from standing height or less) among persons 35 years of age or older.
BMD in grams per square centimeter was determined for the lumbar spine (L2–L4 in anteroposterior projection), proximal femur (total, femoral neck, and intertrochanteric regions), and wrist (total) using dual-energy X-ray absorptiometry with the QDR2000 instrument (Hologic, Waltham, MA, U.S.A.) with software version 5.40. The coefficients of variation for the spine, hip, and forearm BMD measurements were 0.6, 1.8, and 0.8%, respectively. We also estimated volumetric bone mineral apparent density (BMAD, g/cm3) from these data as previously described,(6) using the following formulas:
where BMC is the bone mineral content and A is the projected bone area. Bone density at each site was categorized into three groups according to WHO criteria as normal, low, or osteoporotic(2) relative to young normal means and SDs. The means (and SDs) for BMD/BMAD of the spine, hip, and forearm depended on the reference population and four different ones were compared: the 50 Rochester women 20–29 years old, all 138 premenopausal women combined, the 48 Rochester men 20–29 years old, and the 148 men 20–49 years old (Table 1).
Table Table 1.. Young Normal Means and SD for Bone Mineral Density (BMD, g/cm2) and Bone Mineral Apparent Density (BMAD, g/cm3) Assessed at Different Skeletal Sites Using Different Reference Populations of Rochester, Minnesota, Men and Women
The relative risk of an osteoporotic fracture was estimated with odds ratios (OR) obtained from multiple logistic regression models, where osteoporotic fracture was the endpoint, while age, gender, BMD (spine, hip, wrist), and BMAD (spine, hip, wrist) were the potential predictors. Variables were selected in a stepwise fashion, entering only variables that were significant at the 0.05 level after adjusting for the other variables in the model. Interactions and higher ordered terms were investigated. Kendall's tau-a was used to compare logistic models when various bone density variables were substituted one for another. Kendall's tau-a is a rank correlation that compares the predicted probabilities from a model with the observed responses (whether or not an osteoporotic fracture was detected). It is a useful statistic for comparing models—the higher the tau-a, the better the model does at predicting the event.
The prevalence of osteoporosis was determined by gender and decade of age using different sets of normal values and summarized by direct age adjustment to the structure of the white population of the United States in 1990.
At the baseline assessment, 242 (70%) of the men and 186 (53%) of the women had experienced one or more fractures as enumerated in Table 2. The fractures in women were about equally divided between those due to minimal or moderate trauma (e.g., a fall from standing height or less) and those due to severe trauma (e.g., motor vehicle and recreational accidents and falls from heights). Among men, three-fourths of the fractures were due to severe trauma. There were three fractures due to a specific local pathological process (e.g., bone cyst, metastatic malignancy) among the women, including one vertebral fracture and one hip fracture, and one pathological fracture among the men. Altogether, there were 89 fractures of the hip, spine, or distal forearm among the women, 63 of which were due to minimal or moderate trauma, and 113 such fractures among the men, 56 of which were due to minimal or moderate trauma; 46 women (13%) and 36 men (10%) had one or more minimal or moderate trauma fractures at these sites that occurred at age 35 years or after, thus meeting our definition of an “osteoporotic” fracture.
Table Table 2.. Distribution of Fractures Among Age-Stratified Samples of Rochester, Minnesota, Men and Women by Fracture Site and Cause
BMD was associated with the risk of an osteoporotic fracture at all of the sites assessed (total hip, femoral neck, trochanteric region, spine, and total wrist) among the women (all p < 0.001) and at all sites among the men (all p < 0.001) except for the lumbar spine (not significant). After adjustment for age, the ORs varied by gender and, to a lesser degree, with the reference group used to establish normal means and SDs (i.e., men aged 20–29 years, men aged 20–49 years, women aged 20–29 years, or all premenopausal women combined) as shown in Table 3. Generally, there was a stronger relationship between fracture history and BMD among women than men. Within genders, the estimated relative risk of fracture per 1 SD decrease in BMD at each skeletal site was similar regardless of the reference group selected. Likewise, within genders, the ORs for fracture were of similar magnitude for BMD as assessed across the various skeletal sites (Table 3). After adjusting for age, however, total hip BMD was the best predictor of osteoporotic fractures in women (OR per 1 SD decline, 2.4; 95% confidence interval [CI], 1.6–3.7). After hip BMD entered the model, none of the other BMD measurements independently predicted fracture risk in women. By contrast, the OR per 1 SD decline in total hip BMD in men was only 1.3 (95% CI, 0.97–1.7). Total wrist BMD was the stronger predictor of fractures in men (OR, 1.5; 95% CI, 1.1–2.0). In these analyses, and those that follow, ORs for men are based on the 20- to 29-year-old male reference population, while those for women are based on the 20- to 29-year-old female reference population.
Table Table 3.. Age-Adjusted Odds Ratios (and 95% Confidence Intervals) for Osteoporotic Fracture Per 1 SD Decrease in Bone Mineral Density (BMD) or Bone Mineral Apparent Density (BMAD), by Skeletal Site, for Rochester, Minnesota, Men and Women Depending on the Criterion Used for Establishing Normal Means and Standard Deviations
Substituting BMAD for BMD among the women did not improve fracture prediction. The estimated OR actually declined for 14 of the 20 combinations of skeletal site and reference population shown in Table 3. Using the young female reference data, however, the ORs determined for BMAD in the women were somewhat greater than those for BMD when bone density was measured at the total hip and total wrist. Nonetheless, after adjusting for age (and age2) hip BMD was still the strongest predictor of fracture risk in women and none of the other assessments entered the model. For men, the findings were very different. The ORs determined by BMAD were greater than those for BMD in every instance among the men, and in some cases substantially so (Table 3). After adjusting for age, wrist BMAD was the strongest predictor of osteoporotic fracture in the men (OR, 1.7; 95% CI, 1.3–2.3), and none of the other assessments contributed significantly. However, the model incorporating age and wrist BMAD (model tau-a = 0.129) was not substantially better than the previous one with age and wrist BMD (model tau-a = 0.122).
Altogether, 104 women (30%) and 91 men (26%) had experienced some other moderate trauma fracture (not hip, spine, or wrist) at age 35 or thereafter. After adjusting for age, femoral neck BMD was the best predictor of these other moderate trauma fractures in women (OR, 1.5; 95% CI, 1.1–2.1), while wrist BMAD was best in men (OR, 1.5; 95% CI, 1.2–1.9). None of the other bone density assessments contributed significantly. Again, however, a model for men that included age and wrist BMD (tau-a = 0.091) was about as good as the one with age and wrist BMAD (tau-a = 0.106). One hundred and ten of the women (31%) and 206 of the men (59%) had one or more fractures due to severe trauma. Neither age nor the bone density assessments predicted severe trauma fractures in women, and wrist BMAD was the only significant predictor in men (OR, 1.2; 95% CI 1.04–1.4).
Based on the cross-sectional data from this study, there was a steady loss of hip BMD with age in both sexes (Fig. 1A). Bone loss in the lumbar spine, however, did not begin before age 40 years in women and was not seen at all in the men (Fig. 1B). There was little loss of wrist BMD before age 50 years in either gender (Fig. 1C). These patterns were identical when BMAD was assessed except that bone density levels were no longer higher in men (data not shown).
The age-adjusted (to 1990 U.S. whites) prevalence of osteoporosis at various skeletal sites using the WHO definition (BMD > 2.5 SD below the young normal mean) but different reference groups is shown in Table 4. The prevalence of osteoporosis at the hip, spine, or wrist among Rochester women was 35% using the mean values from 20- to 29 year-old women; however, use of this reference population produced a prevalence of only 3% among Rochester men. For the hip sites, the SDs among the 20- to 29-year-old men were quite small and produced a larger-than-expected prevalence of osteoporosis at the femoral neck, trochanter, and total hip and an overall prevalence among men of 19%. The prevalence was much lower at the hip sites when the 20- to 49-year-old male reference range was used, and the overall prevalence in men was then only 13%. If a more rigorous standard were employed (BMD > 3.0 SD below the mean for 20- to 29-year-old men), then the prevalence of osteoporosis of the hip, spine, or wrist would be only 9% in men. The age-adjusted prevalence was only 2% when the criterion was a BMD level > 4.0 SD below the young male mean (data not shown). Substitution of BMAD for BMD reduced some of the gender-specific differences (Table 4). Thus, use of the 20- to 29-year-old male reference data produced an overall prevalence rate of 17% for Rochester women and 15% for Rochester men. Conversely, when the 20- to 29-year-old female reference data were used, the age-adjusted prevalence of osteoporosis at the hip, spine, or wrist was 29% in women and 38% in men.
Table Table 4.. Age-Adjusted* Prevalence (%) of Osteoporosis (2.5 SD Below Young Normal Mean) Among Rochester, Minnesota, Men and Women ≥50 Years of Age Depending on the Criterion Used for Establishing Means and Standard Deviations
A definition of osteoporosis in men based on BMD is predicated on showing a strong relationship between bone density and fracture risk in men like that seen for women.(7) Our data confirm earlier reports that bone density is associated with fracture risk in men. Among 1355 Japanese–American men, the incidence of vertebral fractures was increased significantly by 1.8-fold and 1.5-fold for every 1 SD decrease in BMD at the proximal and distal radius, respectively; the age-adjusted OR for other nonviolent fractures was 1.3 at both sites and was not statistically significant.(8) The age-adjusted OR of a low trauma fracture among 498 Nebraska men was 1.7 and 1.8 per 1 SD decrease in BMD of the distal radius and ulna, respectively,(9) and there was a similar relationship between distal radius BMD and fragility fractures in 654 Swedish men.(10) Our results at the total wrist site are entirely consistent: the age-adjusted OR of a moderate trauma fracture of the hip, spine, or wrist was 1.5 and of any other moderate trauma fracture it was 1.4. Comparable results have been reported for bone density measured at the calcaneus,(8,11) a site that was not assessed here. As did we, Nguyen and colleagues found no association between lumbar spine BMD and fracture risk in 709 Australian men.(12) They did find a 2-fold increase in the risk of an atraumatic fracture with each 1 SD decrease in femoral neck BMD in men, a result that was considerably stronger than our age-adjusted OR of only 1.1. The comparability of these two figures is uncertain since the Australian fracture model also included quadriceps strength and body sway as independent predictors of fracture risk. A subsequent univariate analysis of their data showed an OR of about 1.4,(13) which was similar to our unadjusted estimate of 1.5. The age-related decline of femoral neck BMD also appeared to account for a doubling of hip fracture risk among 2446 men in Rotterdam, Netherlands.(14)
There is some agreement that the proximal femur is an optimal site to assess osteoporosis in women(15) because hip BMD predicts hip fractures better than measurements at other skeletal sites and predicts fractures in general just as well as the other measurements.(7) By contrast, bone density assessed at the wrist was the best predictor of osteoporotic fractures in men in this study, although the advantage over other sites was not great. It is not clear why this should be, particularly since men do not have a high likelihood of experiencing a distal forearm fracture. Nonetheless, unlike women, fracture prediction in men was not improved by moving to a hip measurement site. Indeed, at every skeletal site assessed, there was a weaker relationship between BMD and fracture risk in men compared with women, as seen also in some other studies.(8,12) In part, this is due to the fact that men's bones are bigger and larger bones are more resistant to fracture at any given level of bone density.(16) Areal bone density does not completely correct for the men's larger bone size,(17) but a further correction for bone volume (e.g., BMAD) did improve fracture prediction in men, bringing the estimated OR per 1 SD change more in line with those of Rochester women. Fracture risk in 258 elderly Finnish men and women was almost identical at comparable levels of calcaneal bone density assessed as gram per cubic centimeter.(18) Although BMAD predicted fractures in men better than BMD, the improvement was modest, and BMD was the better predictor of osteoporotic fractures in women. Others have also found that the theoretical advantages of BMAD over BMD for fracture prediction are slight in practice.(19,20)
The prevalence of osteoporosis depends on the reference population that is used to determine normal values, the cut-off level selected to define the condition and the pattern of bone loss in the community. The choice of a reference population can have a substantial impact on the estimated prevalence of osteoporosis.(21,22) We used reference data from the Rochester population and, within genders, it mattered little whether “young normal” was defined as age 20–29 years or age 20–49 years (or all premenopausal women combined). For women, the overall prevalence of osteoporosis at the hip, spine or wrist changed only from 35% to 41% using the different gender-specific standards, while the men changed from 19% to 13%. This modest difference is due to the fact that bone loss was not seen at the spine or wrist in these cross-sectional data before about age 50 years, despite the linear decrease in hip BMD from age 20 years onward. Use of male normal values for women, however, resulted in female prevalence estimates that were extremely high (66–72%), while female normal values produced prevalence estimates in men that were very low (3–4%). All of these figures are based on use of the WHO cut-off level to define osteoporosis at 2.5 SD below the young normal mean.(1) Originally there was some concern that using a comparable definition of osteoporosis for men (2.5 SD below the young male mean) as for women (2.5 SD below the young female mean) would produce similar prevalence figures in the two genders,(23) whereas the lifetime risk of a hip, spine, or forearm fracture was 40% in white women and only 13% in white men.(3) However, BMD values declined less rapidly over life in men than women in Rochester as seen also by others(24) so that the overall age-adjusted prevalence (35% in Rochester women and 19% in Rochester men), appears to be of reasonable magnitude relative to the lifetime fracture risks previously reported for each gender. Prevalence estimates were similar when BMAD was used with gender-specific reference values (29% in women and 15% in men). Use of more stringent criteria for men (e.g., BMD 3.0 or 4.0 SD below the young male mean) produced overall prevalence estimates in men that seem too low (9% and 2%, respectively). It must be re-emphasized, however, that this empiric approach is entirely arbitrary and that better ways might be developed to select an appropriate cut-off level to define the condition.
Using the WHO definition, the age-adjusted prevalence of osteoporosis at the total hip alone, based on our gender-specific reference data for 20- to 29-year-old women, was 14% for Rochester women compared with a similar 15% figure for white women from the Third National Health and Nutrition Examination Survey (NHANES III), using means and SDs from 409 white women aged 20–29 years.(25) However, the prevalence was estimated at only 4% for white men in the NHANES III data, using means and SDs from 382 white men 20–29 years old to establish the osteoporosis cut-off level. This contrasts sharply with our estimate of 16% using reference data from 20- to 29-year-old Rochester men. The discrepancy is explained by the fact that the mean total hip BMD level in the young Rochester men was higher (1.08 vs. 1.04 g/cm2), and the SD at this site was much smaller (0.113 vs. 0.144 g/cm2) compared with NHANES data. The reference data for Rochester men 20–49 years of age at the total hip site (mean, 1.06 g/cm2; SD, 0.140 g/cm2) corresponded more closely to the NHANES III figures and produced a comparable prevalence of osteoporosis at the total hip in men of 3%.
It is apparent that reference data from NHANES III, a probability sample of the entire United States population, provide superior normative values for hip BMD than does a small sample of men from Rochester. However, there are no comparable national data for BMD of the spine or wrist since those sites were not assessed in NHANES III. This is important because the prevalence of osteoporosis is expected to be higher if a number of skeletal sites are assessed simultaneously as demonstrated in this study. None of the 48 men aged 20–29 years and only one of the 50 women aged 20–29 years had osteoporosis of the hip, spine, or wrist. Among women age 50 years and older, however, the age-adjusted prevalence of osteoporosis at the hip, spine, or distal forearm was 35%. When the age-specific prevalence among Rochester women is extrapolated to the population of the United States in 1990, it suggests that 11.5 million white women might be affected. A similar analysis based on an earlier age-stratified sample of Rochester women assessed with dual-photon absorptiometry produced an estimated 1990 prevalence of 9.4 million white women in the United States with osteoporosis at the hip, spine, or wrist.(3) The estimated number of white men in the United States with osteoporosis in 1990, based on the age-specific prevalence of osteoporosis at the hip, spine, or wrist in Rochester men, is 4.5 million.
The main limitations of this analysis are the relatively small number of subjects, the marginal response rates, and the cross-sectional nature of the analysis. Low participation rates in the oldest age groups, for example, raise the possibility that frail individuals with low bone density and high fracture prevalence were disproportionately excluded, leading to a conservative bias in the fracture ORs. However, analysis of an earlier sample of women from the community demonstrated that the prevalence of osteoporotic fractures was similar whether assessed in study participants alone or in participants and nonparticipants combined.(26) In any event, our cross-sectional estimates of the odds of fracture per 1 SD change in BMD are entirely consistent with prospective data both in white women and white men. Because of the racial composition of the community (96% white in 1990), we are unable to address the relationship between bone density and fracture risk in other ethnic groups. It is important that this issue be resolved since expenditures for treating osteoporotic fractures in nonwhite women and men in the United States approached $1 billion in 1995.(27)
We believe that fracture ascertainment in this study is complete insofar as the fractures were recognized clinically because we reviewed inpatient and outpatient medical records for each subject that spanned 30 years on average. Undiagnosed vertebral fractures are most likely to have been missed since only severe deformities come to clinical attention.(28) Of those with 4 SD deformities by vertebral morphometry in this study (data not shown), 31% of the men and 45% of the women also had a clinical diagnosis of vertebral fracture, while 38% and 59%, respectively, had at least one fracture of the spine, hip, or distal forearm. The significance of any discrepancies is uncertain because it is not clear that all of these deformities represent osteoporotic vertebral fractures. Consequently, we did not consider fracture misclassification to be an important problem here. Likewise, we measured bone density using accurate methods, although we recognize that osteoarthritic changes may cause overestimation of BMD in the lumbar spine and mask any age-related bone loss.(29–33) Consequently, the latter approach may be of limited value in assessing osteoporosis in men and older women. In addition, it remains to be determined whether BMAD would be more useful than BMD in the global assessment of osteoporosis. While BMAD did not appear to be superior to BMD for fracture prediction in this population, gender-specific differences were minimized by substituting volumetric for areal bone density, as reported by others.(34) Use of BMAD instead of BMD has also been shown to reduce the discrepancies in bone density values among men and women of other races.(35–37) When these remaining questions are answered, it should be possible to propose an operational definition for osteoporosis that addresses the entire population.
We would like to thank Mrs. Veronica L. Gathje and Mrs. Margaret F. Holets for help with data collection, Mrs. Cindy Crowson for assistance with data analysis, and Mrs. Mary Roberts for help in preparing the manuscript. This work was supported by research grant AR27065 from the National Institute of Arthritis, Musculoskeletal and Skin Diseases, United States Public Health Service.