Dr Cauley receives research grants from Eli Lilly and Company, Merck & Co. Inc., Novarits, and Pfizer Inc. She also is on the Speaker's Bureau for Eli Lilly and Company. All other authors have no conflict of interest.
BMD at Multiple Sites and Risk of Fracture of Multiple Types: Long-Term Results From the Study of Osteoporotic Fractures
Article first published online: 1 NOV 2003
Copyright © 2003 ASBMR
Journal of Bone and Mineral Research
Volume 18, Issue 11, pages 1947–1954, November 2003
How to Cite
Stone, K. L., Seeley, D. G., Lui, L.-Y., Cauley, J. A., Ensrud, K., Browner, W. S., Nevitt, M. C. and Cummings, S. R. (2003), BMD at Multiple Sites and Risk of Fracture of Multiple Types: Long-Term Results From the Study of Osteoporotic Fractures. J Bone Miner Res, 18: 1947–1954. doi: 10.1359/jbmr.2003.18.11.1947
Dr Cauley receives research grants from Eli Lilly and Company, Merck & Co. Inc., Novarits, and Pfizer Inc. She also is on the Speaker's Bureau for Eli Lilly and Company. All other authors have no conflict of interest.
- Issue published online: 2 DEC 2009
- Article first published online: 1 NOV 2003
- Manuscript Accepted: 10 JUN 2003
- Manuscript Revised: 11 APR 2003
- Manuscript Received: 30 SEP 2002
- bone mineral density;
- population attributable risk
In a large cohort of U.S. women aged 65 and older, we report the relationships of BMD measured at several sites, and subsequent fracture risk at multiple sites over >8 years of follow-up. Although we found almost all fracture types to be related to low BMD, the overall proportion of fractures attributable to low BMD is modest.
Introduction: Although several studies have reported the relationship between bone mineral density (BMD) and subsequent fracture risk, most have been limited by short follow-up time, BMD measures at only one or two sites, or availability of data for only select fracture types.
Materials and Methods: In the multicenter Study of Osteoporotic Fractures (SOF), we studied the relationship of several different BMD measures to fracture risk of multiple types in 9704 non-black women aged 65 and older. We previously reported on the relationship of peripheral BMD measures to risk of several types of fracture during an average 2.2-year follow-up period. In this expanded analysis, we present results of the relationship of both peripheral and central BMD measures and fractures of multiple types during 10.4 and 8.5 years of follow-up, respectively. We also report population attributable risk (PAR) estimates for osteoporosis and risk of several types of fracture.
Results: Our results show that almost all types of fractures have an increased incidence in women with low BMD. However, hip BMD is somewhat more strongly related to most of the fracture types studied than spine or peripheral BMD measures. Nonetheless, the proportion of fractures attributable to osteoporosis (based on a standard definition of osteoporosis) is modest, ranging from <10% to 44% based on the most commonly used definition of osteoporosis (BMD T-score < −2.5).
Conclusion: Finding effective prevention strategies for fractures in older women will require additional interventions beside preventions for bone loss, such as prevention of falls and other fracture risk factors.
Postmenopausal osteoporosis, characterized by low bone mineral density (BMD), is a common condition among elderly women, with major public health consequences including disability and death caused by increased risk of hip and other fractures. Several studies have evaluated the relationship between various measures of BMD and subsequent fracture risk, but most have had relatively short follow-up time, have been limited to BMD measures atone or two sites, or have examined only select fracture sites.(1–13) Using data from the multicenter Study of Osteoporotic Fractures, we were able to prospectively assess the association of several measures of BMD to the long-term fracture risk for all common types of fractures among older women. We also report, for the first time, population attributable risks based on empirical data for a large number of fracture sites.
We previously demonstrated that, of 16 different types of fractures studied, 11 were related to low peripheral bone mass.(14) These findings were based on the first 2.2 years of follow-up. Expanding on these earlier findings, the aims of the present study are as follows: (1) to evaluate the long-term relationship between central BMD (hip and spine), peripheral BMD (forearm and heel), and various fracture types over much longer follow-up time and (2) to estimate the proportion of various types of fractures attributable to low bone mass.
MATERIALS AND METHODS
Subjects were participants in the Study of Osteoporotic Fractures, a prospective study of risk factors for hip and other osteoporotic fractures. During the baseline examination in 1986–1988, 9704 community-dwelling women aged 65 years and older were recruited from population-based listing in four areas of the United States: Baltimore, Maryland; Minneapolis, Minnesota; the Monongahela Valley near Pittsburgh, Pennsylvania; and Portland, Oregon.(15) Black women were excluded at the start of the study because of their low incidence of hip fracture. Women with bilateral hip replacements and those unable to walk without assistance were also excluded.
At the baseline examination (1986–1988), measures of peripheral BMD (distal and proximal radius and calcaneus) were made using single photon absorptiometry (Osteoanalyzer; Dove Medical Systems) Mean CVs between centers were 0.4% for the distal radius, 0.5% for the mid-radius, and 1.2% for the calcaneus.(15)
Between 1988 and 1990, 8134 (87%) of the 9483 surviving members of the cohort returned for a second clinic visit. During this visit, we measured BMD of the proximal femur and subregions (introchanter, trochanter, femoral neck, and Ward's triangle), and of the lumbar spine in the anteroposterior (AP) projection using DXA (QDR 1000; Hologic, Waltham, MA, USA). Mean CVs between centers were 1.2% for the proximal femur and 1.5% for the lumbar spine.(16)
Participants were contacted every 4 months through phone or mail to determine whether any fractures had occurred in the preceding 4-month period. In addition, participants were asked to notify the clinical center as soon as possible after any fracture. All fractures were radiographically confirmed and coded according to degree of trauma. After an average follow-up of more than 10 years, fracture follow-up remains over 98% complete. We exclude fractures resulting from severe trauma, which included fractures resulting from motor vehicle accidents, being struck by a car or other rapidly moving projectile, or assault. In general, falls from greater than standing height were not considered as severe trauma, except in a few special circumstances at the discretion of the physician adjudicator.
Incident spine fractures were assessed by comparing baseline spine radiographs to repeat radiographs taken an average of 3.7 years later. A fracture was defined as a reduction in any vertebral height measure (anterior, medial, or posterior) of any vertebral body (T4-L4) of at least 20% and 4 mm between baseline and follow-up. This analysis was restricted to the 7238 women who had adequate pairs of baseline and repeat spine radiographs and BMD measurements taken at baseline. Analyses of central BMD and incident spine fracture were further restricted to the 6892 participants who returned for the year 2 clinic visit and had BMD measures taken at that visit. More complete details of these measures have been previously published.(17,18)
Peripheral and central BMD and fracture risk
Fractures caused by major trauma, such as motor vehicle accidents (n = 108 since the baseline visit), were excluded from these analyses, as were types of fracture that occurred less than 20 times during the follow-up period (sternum, skull, scapula, coccyx). Using Cox proportional hazards models, we estimated the age-adjusted risk of fracture per SD decrease in BMD. Results are presented as relative hazards with 95% CIs. We analyzed the relation between peripheral BMD and fracture using all fractures that had occurred since the baseline visit when peripheral BMD measures were obtained (mean follow-up, 10.4 years). Similarly, women with a given type of nontrauma fracture during 8.5 years following the second exam (when measures of central BMD were obtained) were compared with women without fractures of any type since the second visit. Because results for subregions of the hip were, in general, similar to those for the total hip and femoral neck, we present results only for BMD of the total hip and the femoral neck subregion. All analyses were conducted using the SAS software (SAS Institute, Cary, NC).
BMD and fracture risk, stratified by follow-up time
To determine if BMD predicts fractures more strongly when the BMD measurements are obtained more proximal to the occurrence of fracture, we repeated the analyses described above stratified on follow-up time (≤5 versus >5 years).
Population attributable risk
Population attributable risk (PAR) formulas were used to estimate the proportion of fracture cases that can be attributable to low hip or spine BMD.(19) The PAR is a function of the strength of the association between the risk factor (BMD) and the disease (fracture), and the prevalence of the risk factor. For osteoporosis, prevalence varies depending on BMD site and definition of osteoporosis. We used the definition of osteoporosis established by the World Health Organization,(20) as BMD more than 2.5 SDs below that of a young normal population. In addition, to illustrate the impact of change in definition for osteoporosis, we calculated PAR estimates using a cutpoint of 1.5 SDs below young normal means. For hip BMD, young normal values were derived from the National Health and Nutrition Examination Survey.(21) For spine BMD, we used young normal values provided by Hologic.(22) We estimated PAR for these two definitions of osteoporosis and each type of fracture. Analyses were also performed stratified by age.
The average age of participants was 71.7 years (range, 65–99 years) at the baseline examination (Table 1). Mean BMD values are also displayed in Table 1. After an average of 10.4 years of follow-up since baseline (visit 1), 4172 nonspine fractures occurred in 2901 women, including 721 hip fractures, 744 wrist fractures, and 439 humerus fractures. For all analyses, the comparison group included 6137 women who experienced no incident clinical fractures.
Similarly, with an average of 8.5 years of follow-up since the second visit, 3278 fractures occurred in 2372 women, including 622 hip, 561 wrist, and 362 humerus fractures. Between visits 1 and 3 (mean follow-up = 3.7 years), 389 women suffered incident spine fractures.
BMD and fracture risk
Peripheral BMD was significantly associated with all types of fractures studied except facial fractures (Table 2). Significant relative hazard (RH) estimates for fracture and peripheral BMD ranged from a low of 1.15 (95% CI, 1.02–1.30) per SD decrease in BMD (calcaneal BMD and ankle fractures) to a high of 2.44 (1.94–3.08) per SD decrease in BMD (distal radius BMD and lower leg fracture).
Associations between total spine, total hip, and femoral neck BMD and specific types of fractures are displayed in Table 3. Associations with the other BMD subregions of the hip (trochanteric, intertrochanteric, Ward's triangle) were similar and are not presented. Fifteen of the 18 types of fracture analyzed were significantly (p ≤ 0.05) associated with all measures of central BMD. Face and ankle fractures were not significantly associated with any of the central BMD measures. Although low BMD at both the hip and spine predicted risk of incident heel fracture, this association reached statistical significance only in the case of hip BMD. Overall, significant RH estimates ranged from a low of 1.20 (1.04–1.39) per SD decrease in BMD (for spine BMD predicting foot fractures) to a high of 2.50 (1.82–3.44) per SD decrease in BMD (femoral neck BMD and risk of patella fractures).
BMD and fracture risk stratified by follow-up time
Table 4 shows results of analyses of total hip and spine BMD and selected common fracture types stratified by follow-up time. For certain fracture types (e.g., hip, humerus, pelvis, rib), BMD predicts fracture more strongly over the shorter term (≤5 years after BMD assessment) but remains a significant predictor of long-term fractures (>5 years after BMD assessment). Overall, when all nonspine nontrauma fractures were combined, the short-term and long-term results are similar. For example, the RH estimates for all nonspine nontrauma fractures combined per SD decrease in total hip BMD were 1.44 (1.36–1.53) and 1.48 (1.37–1.60), respectively, for follow-up times of ≤5 and >5 years. Similar results were obtained for peripheral BMD sites: RHs for all nonspine nontrauma fractures per SD decrease in calcaneal BMD were 1.44 (1.36–1.52) and 1.33 (1.25–1.41), respectively, for follow-up of ≤5 and >5 years.
Using total hip BMD to define osteoporosis, the PAR for osteoporosis (defined as total hip BMD T-score ≤ −2.5) and hip fracture is 28% (95% CI 22–33%; Table 5). In other words, we estimate that 28% of hip fracture cases are attributable to osteoporosis as defined using total hip BMD. If spine BMD was used to define osteoporosis, the PAR for osteoporosis and hip fracture was 21% (14–28%). Overall, the PARs for hip and spine BMD and fracture varied from a low of 0–8% for ankle fractures to a high of 39–44% for clavicle fractures.
We also calculated PAR estimates based on an alternative, more conservative definition of osteoporosis (T-score ≤ −1.5). Based on this definition, PAR estimates are somewhat higher than those observed using the more conventional WHO definition of osteoporosis (Table 5), ranging from 12% to 53% when hip BMD is used to define osteoporosis and from 6% to 51% based on spine BMD.
For selected fracture types, we calculated PAR estimates based on the standard WHO definition of osteoporosis (T-score ≤ −2.5) stratified by age: 65–74 years and 75 years and older. For hip, wrist, spine, and rib fractures, PARs were similar for younger and older women. By contrast, for humerus and pelvis fractures, the PARs were significantly higher in older women. For example, PAR estimates for pelvis fractures were 19% (6–37%) for women aged 65–74, and 40% (24–56%) for women aged 75+. This may indicate that with advancing age a larger proportion of certain types of fractures are attributable to low BMD.
Our findings show, using a standard definition of osteoporosis (T-score ≤ −2.5), that <10–44% of most types of fractures can be attributed to low bone mass. We also confirm and expand on previous results from SOF: in our new analyses, which incorporate longer follow-up time, 15 of 18 types of fractures considered were related to all measures of BMD, while one additional fracture type (heel) was related to all but one BMD site. Ankle fractures were related to peripheral, but not central, BMD measures. Face fractures were the only fracture type not associated with any BMD measures. Overall, the relationship between BMD and fracture risk is similar over the short-term (<5 years) and long-term (≥5 years).
Based on the standard WHO definition of osteoporosis, among older women, population attributable risk for osteoporosis and fracture is modest, ranging mostly from <10–44% for specific types of fractures to around 15% for all types of fractures combined. This is considerably lower than PAR for smoking and lung cancer, which has been estimated to be in excess of 80%.(23–25) However, these PAR estimates are comparable with or higher than PAR estimates in the 10–35% range reported for cardiovascular disease based on commonly recognized risk factors such as hypertension,(26,27) high cholesterol levels,(28) and lipid profiles.(29) We recognize that the definition of osteoporosis based on T-scores is arbitrary and controversial,(30) but nonetheless, such definitions are widely used. In addition, even using a more conservative cutpoint to define osteoporosis (T-score ≤ −1.5), we find that at most about 50% of fractures can be attributed to osteoporosis. These results imply that measures to sustain or improve BMD would have some potential to decrease fracture rates and subsequent morbidity and disability as well as to reduce the economic burden of osteoporosis, which has been estimated at as much as $13.8 billion in 1995.(31) However, while the importance of low BMD on fracture risk should not be underestimated, there may be other equally important factors contributing to risk of fractures, including risk factors for falls.(32) Furthermore, results from some clinical trials have indicated that observed reductions in fracture risk exceed those predicted based on change in BMD alone,(33) possibly because of additional effects of treatment on bone turnover and/or geometric properties of bone.
Melton et al.(34) used the Delphi method(35) to calculate osteoporosis attribution probabilities (PAR). The Delphi method involved assembly of an expert panel to make judgments about probability of fractures of different types based on patient age, gender, and race. For white women, 65–84 years of age, the expert panel concluded that osteoporosis is responsible for approximately 90% of hip fractures, 70% of distal radius forearm fractures, 90% of all spine fractures, and 50% of all other fractures (BMD site not specified). In contrast, we found much lower PARs for this age and race group: 28% for hip fracture, 25% for spine fracture, 16% for wrist fracture, and 13% for all fracture types (using hip BMD, the least conservative estimate). An earlier expert panel reported attribution probabilities of 51%, 71%, and 91% for hip fractures in white women aged 45–59, 60–74, and ≥75, respectively.(36) Another study also reported much higher PARs for fractures of the pelvis, spine, wrist, humerus, and leg.
These differences in PARs are remarkable. The Delphi method has been used in a variety of healthcare applications over the past three decades and is recognized as a systematic, literature-based, scientific method that uses expert group judgment in the absence of adequate data.(34) While our study is biased by our limited and somewhat select population, it is based on real data and therefore likely represents a better estimate of the true impact of osteoporosis and fracture. Marshall et al.(37) also estimated PAR for hip fracture to be 21–36% depending on the lifetime incidence of hip fracture used. While we cannot directly compare results because of slightly different cutpoints and normal definitions used, these levels compare favorably to the 28% PAR we have reported for hip BMD and hip fracture. Additional data from other populations are needed to resolve this question adequately.
For humerus and pelvis fractures, the PARs were somewhat higher for older women, while results for other fracture sites are similar in older and younger women. This differs somewhat from both expert panel reports that suggest that PARs for hip, wrist, and spine fractures are higher for older women.(34,36)
The choice of a reference population affects the prevalence of osteoporosis, which in turn affects the PAR estimate. The debate over accurate reference data has been ongoing,(38,39) with the NHANES being the accepted reference population for hip BMD (based on non-Hispanic white women aged 20–29 years). Our attributable risks using spine BMD are generally consistent with those obtained using hip BMD, suggesting our reference population for spine BMD is appropriate.
Our statistical estimates of PAR suggest that eliminating osteoporosis would modestly reduce risk of fractures. However, prevention of more fractures in populations will require interventions besides treatments for osteoporosis. For example, it may be necessary to focus additional interventions on prevention of falls and other fracture risk factors, in addition to treatment of osteoporosis, to prevent a majority of fractures that occur in older women.
BMD and fracture risk
Our findings extend our earlier published results from SOF(14) that were based on the relationship of peripheral BMD sites with risk of multiple types of fractures over a 2.2-year follow-up period. Specifically, compared with the earlier publication, our new results extend BMD measures to include both peripheral and central sites and encompass a much longer follow-up time (10.4 and 8.5 years on average, respectively, for analyses of peripheral and central BMD and fracture risk) allowing for increased power for analysis of specific fracture types. Our long-term results show that all 10 of the fracture types that were previously found to be associated with peripheral BMD over the shorter term remain significantly related to BMD over the long-term and are related to both peripheral and central BMD. Additionally, we found five new fracture types (heel, femur, elbow, finger, lower leg, and patella) to be related to both peripheral and central BMD, whereas ankle fractures were found to be related to peripheral but not central BMD. Facial fractures are the only fracture type not related to either peripheral or to central BMD. For a few fracture types, BMD appears to be a stronger predictor over the shorter term than over the longer term (Table 4), although when all fracture types are combined there is very little difference between the short and long-term relationship of BMD and risk of fracture.
Our risk estimates for BMD and fracture are similar to those reported by others.(3,40) It seems that hip BMD is more strongly related to most fracture types than spine or peripheral BMD measures. In a few cases, we found that site-specific measures of BMD tended to have the strongest association with fracture at that site, similar to findings of others.(40) For example, our findings confirm that hip BMD is a better predictor of hip fracture than other measurements, a finding already published by our group.(13) Results for BMD of the total hip and subregions of the hip yielded similar results. However, by contrast, hip and spine BMD yielded similar estimates for prediction of spine fractures.
Marshall et al.(37) performed a meta-analysis of various BMD measures predicting hip, wrist, spine, or all fractures. Summarizing 11 separate study populations with approximately 90,000 person-years of observation time, they reported similar summary relative risks to what we have reported. Overall, most risk estimates were around 1.5 (1.4, 1.6). However, consistent with our findings, the relationships were markedly stronger for spine BMD predicting risk of spine fracture (2.3; 1.9–2.8 in the Marshall study compared with 2.1; 1.8–2.4 in our study); and for hip BMD predicting risk of hip fracture (2.6; 2.0–3.5 in the Marshall study compared with 2.4; 2.1–2.7 observed in our study using femoral neck BMD). However, our results differed considerably from those reported by Varenna et al.(41) Based on a case-control study design, these authors reported a very strong relationship between lumbar spine BMD and foot fracture (relative risk = 2.4; 1.9–3.1). In contrast, we observed a RH of only 1.2 (1.0, 1.4) for the relationship between total spine BMD and risk of foot fracture.
Although our study is very large, has a prospective design, and has a very long follow-up, there are a few limitations. Our results are limited to ambulatory, white, postmenopausal women. We recently enrolled black women and hope to address these questions in that population in the near future. While we were able to address many different fracture sites, we had insufficient data to look at parts of the long bones separately, such as the distal femur versus the proximal femur and the tibia versus the fibula.
In summary, we found that risk of almost all types of fractures except face are related to low BMD and that all BMD sites predicted almost all types of fractures. Hip BMD had somewhat stronger association to all fractures combined than did BMD of the spine or peripheral sites. Despite the strong associations between BMD and fracture risk, less than one-half of fractures in older women are caused by osteoporosis according to our classical epidemiologic estimates of attributable risk. Therefore, finding effective prevention strategies for fractures in older women will require additional interventions besides treatment for osteoporosis, such as prevention of falls and other fracture risk factors.
This study was supported by Public Health Service Grants AG05407, AR35582, AG05394, AR35584, and AR35583.
- 11985 Prediction of postmenopausal fracture risk with use of bone mineral measurements. Am J Obstet Gynecol 153:745–751., , ,
- 21989 Predicting fractures in women by using forearm bone densitometry. Calcif Tissue Int 44:235–242., ,
- 31989 Baseline measurement of bone mass predicts fracture in white women. Ann Intern Med 111:355–361., ,
- 41983 Relationship between bone mineral content and frequency of postmenopausal fractures. Acta Med Scand 213:61–63., , , ,
- 51993 Prediction of osteoporotic fractures by postural instability and bone density. BMJ 307:1111–1115., , , , , ,
- 61992 Axial and appendicular bone density predict fractures in older women. J Bone Miner Res 7:633–638., , , , ,
- 71998 Hip fracture prediction in elderly men and women: Validation in the Rotterdam study. J Bone Miner Res 13:1587–1593., , , , ,
- 81998 Short-term and long-term fracture prediction by bone mass measurements: A prospective study. J Bone Miner Res 13:107–113., ,
- 91998 How hip and whole-body bone mineral density predict hip fracture in elderly women: The EPIDOS Prospective Study. Osteoporos Int 8:247–254., , , , , , , , , ,
- 101999 Separate and combined value of bone mass and gait speed measurements in screening for hip fracture risk: Results from the EPIDOS study. Osteoporos Int 9:188–192., , , , , , ,
- 111997 Calcaneal bone mineral density predicts fracture occurrence: A five-year follow-up study in elderly people. J Bone Miner Res 12:1075–1082., , , , ,
- 121997 Bone density variation and its effects on risk of vertebral deformity in men and women studied in thirteen European centers: The EVOS Study. J Bone Miner Res 12:1883–1894., , , , , , , , , , , , , , , , ,
- 131993 Bone density at various sites for prediction of hip fractures. The Study of Osteoporotic Fractures Research Group. Lancet 341:72–75., , , , , , , , ,
- 141991 Which fractures are associated with low appendicular bone mass in elderly women? The Study of Osteoporotic Fractures Research Group. Ann Intern Med 115:837–842., , , , ,
- 151990 Appendicular bone density and age predict hip fracture in women. The Study of Osteoporotic Fractures Research Group. JAMA 263:665–668., , , , , , , , , ,
- 161992 Age-related decrements in bone mineral density in women over 65. J Bone Miner Res 7:625–632., , , ,
- 171999 Defining incident vertebral deformity: A prospective comparison of several approaches. The Study of Osteoporotic Fractures Research Group. J Bone Miner Res 14:90–101., , , , ,
- 181998 The association of radiographically detected vertebral fractures with back pain and function: A prospective study. Ann Intern Med 128:793–800., , , , , , , ,
- 191981 Statistical Methods for Rates and Proportions, 2nd ed. Wiley, New York, NY, USA, pp 75–77.
- 20World Health Organization 1994 Assessment of fracture risk and its application to screening for postmenopausal osteoporosis: Report of WHO Study Group. World Health Organization, Geneva, Switzerland.
- 211997 Prevalence of low femoral bone density in older U.S. adults from NHANES III. J Bone Miner Res 12:1761–1768., , , , , , , ,
- 221990 Sex differences in peak adult bone mineral density. J Bone Miner Res 5:1169–1175., , ,
- 231986 Smoking and lung cancer with special regard to type of smoking and type of cancer. A case-control study in north Sweden. Br J Cancer 53:673–681.,
- 241988 Cigarette smoking and lung cancer in New Mexico. Am Rev Respir Dis 137:1110–1113., , ,
- 251984 A cohort analysis of lung cancer and smoking in British males. Am J Epidemiol 119:624–641.,
- 262001 Risk factors for congestive heart failure in US men and women: NHANES I epidemiologic follow-up study. Arch Intern Med 161:996–1002., , , , ,
- 272000 Reducing the global burden of blood pressure-related cardiovascular disease. J Hypertens Suppl 18:S3–S6., ,
- 281998 Lipoprotein(a) and cholesterol levels act synergistically and apolipoprotein A-I is protective for the incidence of primary acute myocardial infarction in middle-aged males. An incident case-control study from Sweden. J Intern Med 244:425–430., , , , , ,
- 292001 Low triglycerides-high, high-density lipoprotein cholesterol and risk of ischemic heart disease. Arch Intern Med 161:361–366., , ,
- 302000 Osteoporotic fractures, ageing, and the bone density T-score. Clin Rheumatol 19:171–173.
- 311997 Medical expenditures for the treatment of osteoporotic fractures in the United States in 1995: Report from the National Osteoporosis Foundation. J Bone Miner Res 12:24–35., , ,
- 321995 Risk factors for hip fracture in white women. Study of Osteoporotic Fractures Research Group. N Engl J Med 332:767–773., , , , , , , ,
- 332000 Fracture risk reduction with alendronate in women with osteoporosis: The Fracture Intervention Trial. FIT Research Group. J Clin Endocrinol Metab 85:4118–4124., , , , , , , ,
- 341997 Fractures attributable to osteoporosis: Report from the National Osteoporosis Foundation. J Bone Miner Res 12:16–23., , , , , , , , ,
- 35LinstoneHA, TuroffM (eds.) 1975 The Delphi Method: Techniques and Applications. Addison-Wesley, Reading, MA, USA, p. 620.
- 361988 The direct medical costs of osteoporosis for American women aged 45 and older, 1986. Bone 9:271–279., , ,
- 371996 Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 312:1254–1259., ,
- 381996 Discrepancies in normative data between Lunar and Hologic DXA systems. Osteoporos Int 6:432–436., ,
- 391997 The effects of standardization and reference values on patient classification for spine and femur dual-energy X-ray absorptiometry. Osteoporos Int 7:200–206., , , ,
- 401993 Long-term fracture prediction by bone mineral assessed at different skeletal sites. J Bone Miner Res 8:1227–1233., , , ,
- 411997 Is the metatarsal fracture in postmenopausal women an osteoporotic fracture? A cross-sectional study on 113 cases. Osteoporos Int 7:558–563., , , , ,