SEARCH

SEARCH BY CITATION

Keywords:

  • fracture;
  • osteoporosis;
  • bone loss;
  • BMD;
  • population attributable risk

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Whereas low BMD is known to be a risk factor for fracture, it is not clear whether loss of BMD is also a risk factor. In elderly women, greater loss of BMD at the femoral neck was associated with increased risk of fracture, independent of baseline BMD and age.

Introduction: Baseline measurement of BMD predicts fracture risk. However, it is not clear whether short-term bone loss is an independent risk factor for fractures. This study was designed to investigate the relationship between changes in BMD and fracture risk in elderly women in the general population.

Materials and Methods: A total of 966 women ≥60 years of age (mean, 70 ± 6.7 [SD] years), who had been followed for an average of 10.7 years, were studied. Atraumatic fracture of the proximal femur (hip), symptomatic vertebral fracture, and other major fractures, excluding pathological fractures or those resulting from severe trauma, were recorded and confirmed by radiographs. Femoral neck and lumbar spine BMD was measured by DXA.

Results: During the follow-up period, 224 had sustained a fracture (including 43 hip, 71 symptomatic vertebrae, 37 proximal humerus, 46 forearm and wrist, and 27 rib and pelvis fractures). The annual rate of change in BMD in fracture women (−2.1 ± 4.2%) was significantly higher than that in nonfracture women (−0.8 ± 2.8%; p = 0.005). In the multivariable Cox's proportional hazards analysis, the following factors were significant predictors of fracture risk: femoral neck bone loss (relative hazard [RH], 1.4; 95% CI, 1.1-1.8 per 5% loss), baseline femoral neck BMD (RH, 2.0; 95% CI, 1.7-2.7 per SD), and advancing age (RH, 1.2; 95% CI, 1.1-1.4). The proportion of fractures attributable to the three factors was 45%. For hip fracture, the attributable risk fraction was ∼90%.

Conclusion: Bone loss at the femoral neck is a predictor of fracture risk in elderly women, independent of baseline BMD and age.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

AMONG CLINICAL risk factors, measurement of BMD, particularly at the femoral neck, has the most robust predictive value for risk of various fractures.(1–3) BMD in later life is a dynamic function of peak bone mass achieved during growth and subsequent postmenopausal- and age-related rate of loss. Although a woman may lose one-half of her trabecular bone and one-third of her cortical bone in later life (typically 1% per year), this varies widely between women.(4–7) Some lose >5% per year; others experience no change or even an increase in BMD.(7)

Change in BMD is a result of an imbalance between two continuous remodeling processes of bone formation and bone resorption. A number of prospective studies have suggested that changes in bone turnover markers are associated with changes in BMD,(8–10) but the association is not consistent across studies.(11,12) Furthermore, some prospective studies have also suggested an association between markers of bone turnover and fracture risk, independent of BMD at the menopause.(13–16)

If bone turnover markers, which are suggested to be associated with change in BMD, are predictive of fracture risk, it could be hypothesized that rate of bone loss is an independent risk factor for fracture. This study was designed to assess the contribution of bone loss to the risk of fracture in a group of elderly women by using a longitudinal, population-based study in Australia.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Study design

Data used in this analysis were derived from the Dubbo Osteoporosis Epidemiology Study (DOES), which has been on-going since June 1989. The sampling frame for DOES is the city of Dubbo, New South Wales (Australia), a locality of ∼32,000 people, 98.6% white, of which 1581 men and 2095 women were ⩾60 years of age in 1989. Dubbo's relative isolation in terms of medical care allows virtually complete ascertainment of all fractures.

Measurements

After obtaining written informed consent, subjects were interviewed by a nurse coordinator who administered a structured questionnaire to collect data including age, anthropometric variables, lifestyle, and clinical data.(3,17,18) BMD (g/cm2) at baseline and follow-up (average interval, 2.7 years) were measured in the lumbar spine and femoral neck by DXA using a LUNAR DPX-L densitometer (GE Lunar Corp., Madison, WI, USA). Baseline BMD was measured between 1989 and 1996, with ∼90% of the measurements before 1992. Follow-up BMD was measured between 1990 and 2002, with ∼90% of the measurements before mid-1995. The same instrument and software were used throughout the study. The radiation dose with this method is <0.1 μGy. The coefficient of reliability of BMD in our institution in normal subjects is 0.96, and the CV was 1.5% at the proximal femur.(19) Quality control were performed regularly by using phantom to ensure the reliability of the densitometer. For this study, subjects had had at least two BMD measurements, and for the fracture cases, at least two BMD measurements before the fracture event.

The primary bone mass measurement used in this study was femoral neck BMD. This choice was based on the fact that in elderly women femoral neck BMD measurement is less likely to be affected by degenerative changes than lumbar spine BMD measurement. Thus, the use of femoral neck BMD as an analysis variable permitted a more reliable and valid way to examine the true relationship between fracture and BMD.

Fracture ascertainment

The primary outcome of the study were incidence of low-trauma symptomatic fracture. Low-trauma symptomatic fractures occurring during the study period were identified for residents of the Dubbo local government area through radiologists' reports from the only two centers providing X-ray services as previously described.(3,17) Fractures were only included if the report of fracture was definite and, on interview, had occurred with minimal or no trauma, including a fall from standing height or less. Fractures clearly caused by major trauma such as motor vehicle accidents were excluded from the analysis. Fracture was analyzed as any osteoporotic fracture or as subgroups of hip fractures, vertebral fractures (symptomatic) and other clinical fractures, including proximal humerus, forearm and wrist, rib, and pelvis.

Data analyses

Change of BMD from baseline was calculated for each subject and expressed as percent per year. Cox's proportional hazards regression model was used to estimate relative hazard (RH) for ∼1 SD change in a risk factor. The outcomes examined were fracture incidence and the time to fracture from baseline BMD measurement. The time to fracture was calculated as difference between the date of first BMD measurement and date of fracture. The risk factors considered in this analysis were baseline BMD, bone loss (as a continuous variable), age, and anthropometric variables. The significance of parameter estimates derived from the Cox's proportional hazard model(20) was tested with the likelihood ratio statistic.(21) The assumption of proportional hazards for the levels of each risk factor was tested by evaluating the linearity of plots of log{−log(S[ti]j)}, where S(ti)j describes the jth survival time for the ith level (i = 1, 2) for each risk factor. All analyses were performed with the SAS Statistical Analysis System.(22)

To estimate the relative contribution of various risk factors to the risk of fracture, the population attributable risk fraction (PARF) was calculated. PARF is the proportion of fractures in the population that would be avoided if a particular risk factor or combination of risk factors was not present. To estimate PARF, it was necessary to distinguish a “low-risk” group (assumed to be free of the relevant risk exposures) from a remaining “high-risk” group in which the exposures were present. In this study, the classification of low-risk and high-risk individuals was based on the risk factors that were assessed to be statistically significant in the proportional hazards analysis. An individual was classified as “osteoporosis” if her femoral neck BMD was 2.5 SD or more below the young normal level, equivalent to <0.7 g/cm2, or normal. The threshold of −2.5 SD was chosen so that it was consistent with the WHO definition of osteoporosis. The classification of low versus high bone loss was based on the CV of femoral neck BMD, which was estimated to be 1.5%.(19) In this scenario, if two measurements are performed in the same subject, conservatively, a minimal change that can be considered statistically meaningful is equation image. Therefore, the change was classified as “high risk” if femoral BMD had decreased by at least 5% per year; otherwise a “low risk” was classified.

The statistical estimation of PARF was based on the “sequential attributable fractions.”(23,24) Briefly, for each threshold criterion used to define high-risk individuals, the expected probability of fracture was calculated from the estimated coefficients of the proportional hazards model. The expected probability was compared with the observed probability, and components of attributable fraction was subsequently estimated for each possible combination of risk factors.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Characteristics of subjects

Data were analyzed from 966 women, 69.9 ± 6.7 (SD) years of age, who had at least two BMD measurements, separated by an average interval of 2.7 years (range, 0.8-4.2 years). The average duration of follow-up for these women was 10.7 years (range, 2.7-13 years), which yielded 10,329 person-years. During the follow-up period, 43 women sustained a hip fracture, 71 symptomatic vertebral fractures, 37 proximal humerus fractures, 46 forearm and wrist fractures, and 27 rib and pelvis fractures. As expected, fracture subjects were older, lighter, and shorter than nonfracture subjects (Table 1). Approximately 62% of fracture subjects were ≥70 years of age. Femoral neck BMD in fracture subjects was ∼1 SD lower than in nonfracture subjects; however, the difference was more pronounced in those with hip fractures. The annual rate of change in femoral neck BMD in fracture subjects was −1.4 ± 4.1%, significantly higher than in nonfracture subjects (p < 0.01). The difference remained statistically significant even after adjusting for age and baseline BMD. Among the fracture group, subjects with hip fracture had the highest rate of bone loss (−2.1 ± 4.2% per year) before the fracture event. However, there was no significant difference in the rate of change in lumbar spine BMD between those with a fracture and those without a fracture (Table 1).

Table Table 1.. Demographic Characteristics, BMD, and Rate of Change in BMD by Fracture Site
Thumbnail image of

Univariate analysis

In univariate analysis, each SD (0.12 g/cm2) lower femoral neck BMD was associated with an RH of 2.1 (95% CI: 1.8-2.5), which was identical to the association between lumbar spine BMD and fracture risk (RH, 2.1; 95% CI: 1.8-2.5). However, the effect of femoral neck BMD on hip fracture risk was more pronounced (RH, 4.3; 95% CI: 3.1-5.9) than on other fractures. The strength of association between fracture risk and second BMD measurement was almost identical to that of the first measurement. Femoral neck bone loss, either expressed as absolute value or percentage change, was significantly associated with an increased RH of hip fracture (RH, 1.8; 95% CI:1.2-2.8) and proximal humerus fracture (RH, 1.7; 95% CI, 1.0-2.7). When all fractures were combined, the association between femoral neck bone loss and fracture risk was also statistically significant (RH, 1.4; 95% CI: 1.1-1.7). Bone loss at the lumbar spine was significantly associated with increased risk of hip and rib and pelvis fractures (Table 2).

Table Table 2.. Predictors of Risk of Fracture: Univariate Analysis
Thumbnail image of

Multivariate analysis

Because BMD measured at different skeletal sites, age, and weight were correlated, it was necessary to assess their independent contributions to fracture risk in a multivariate model. When baseline BMD, rate of loss, and body weight were considered simultaneously in a multivariate Cox's proportional model, the following factors were found to be independent predictors of fracture risk: age (RH, 1.2; 95% CI: 1.1-1.4 per 5 years of age), baseline femoral neck BMD (RH, 2.0; 95% CI, 1.7-2.7 per SD), and rate of loss (RH, 1.4; 95% CI, 1.1-1.8 per 5% decrease). However, analyses of individual fracture sites indicated that age was not significantly associated with risk of fracture at the proximal humerus, forearm and wrist, and rib and pelvis when BMD was considered. Moreover, rate of femoral neck bone loss was significantly associated with hip, symptomatic vertebral, and proximal humerus fractures, but not with forearm, wrist, rib, and pelvis fractures, independent of femoral neck BMD and age (Table 3).

Table Table 3.. Independent Predictors of Fracture Risk
Thumbnail image of

Estimates of attributable fraction

Using the results from the multivariate analysis, the PARF was calculated for each combination of risk factors (Table 4). Among the 966 women in the sample, 26% was osteoporotic (femoral neck BMD T scores ≤−2.5), 10% had experienced femoral neck bone loss of at least 5% per year, and 45% were ≥70 years of age. In total, 56% of women had at least one risk factor (either 70+ years of age, or osteoporosis, or experienced a femoral neck bone loss of at least 5% per year), of which, 19% had two risk factors, and 3% had three risk factors.

Table Table 4.. Population Attributable Risk Fraction Estimates for Age, Osteoporosis, and Femoral Neck Bone Loss
Thumbnail image of

Using the sequential attributable fraction analysis, it was estimated that 45% of fractures were attributable to the three risk factors, of which, 13% was attributable to high rate of bone loss, and ∼32% was attributable to osteoporosis. However, the majority of these attributable proportions were associated with advancing age. Indeed, osteoporosis and advancing age (15% in the population) were the two most important risk factors of fracture, because the combination of the two factors accounted for the majority of PARF. The largest PARF was observed in hip fractures, where almost 90% was attributable to advancing age, osteoporosis, and high rate of femoral neck bone loss. In contrast, the lowest attributable fraction was found in forearm and wrist fracture (PARF, 19%).

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

To our knowledge, this is the first prospective study examining the association between bone loss and fracture in the general untreated population. This study's results suggest that femoral neck bone loss was an independent predictor of future fracture risk, independent of baseline BMD levels and advancing age. It was further shown in this study that 45% of fracture cases could be attributable to osteoporosis (based on the most common definition of BMD T scores ≤−2.5), high rate of bone loss, and advancing age. This result, if confirmed in independent studies, suggests that effective primary prevention of bone loss (or preservation of BMD) through pharmacological intervention or primary prevention, initiated before the menopause, may be helpful in the reduction of fracture incidence in the general population.

This study's result may also help explain the antifracture efficacy observed in clinical trials. Recent clinical trials with antiresorptive therapies (e.g., potent bisphosphonates) have shown a modest increase in BMD (typically 4-8%) among low-BMD women. This modest increase has been associated with an approximate halving of spine and nonspine, including hip fracture. Based on the BMD-fracture risk relationship observed in short-term observational epidemiological studies, such a fracture reduction has been suggested to be higher than expected,(25) and led to a postulation that this excess of benefit is in some way caused by improvement in bone quality or other unmeasured architectural characteristics.(26) Whereas the postulation is eminently possible, the rate of bone loss per se was not considered in the assessment of the BMD-fracture risk relationship. This study suggests that the excess of benefit could be explained by the rate of bone loss. Indeed, in normal ambulatory elderly women, a 5% loss of femoral neck BMD could independently (of baseline BMD) translate into a 40% increase in the risk of fracture; therefore, a therapy with modest effect on BMD could bring about a significant reduction of fracture risk that is not captured by differences in baseline BMD.

This study's results confirm that BMD is one of the primary indicators of hip fracture risk.(1–3,27) BMD measured at the femoral neck was consistently a significant predictor of hip fracture, and the discriminant power was better than that measured at the lumbar spine. Furthermore, the effect of lumbar spine BMD on vertebral fracture was more pronounced than its effect on nonvertebral fractures. Whereas the lesser predictive power of lumbar spine BMD may also be caused, in part, by confounding by osteoarthritic degenerative change, these observations are consistent with the sensitivity of discrimination of fracture risk being site-specific.

Although the association of lower BMD with fracture is well recognized, it remains largely unknown how much longitudinal bone loss might contribute to fracture risk. In this study, each 0.12-g/cm2 loss in femoral neck BMD was associated with a 3.1-fold increase in hip fracture risk. This association is somewhat lower than each 0.12-g/cm2 difference in cross-sectional BMD, but it is consistent with earlier data from forearm BMD, which suggested that “fast bone losers” (defined as loss of at least 3% per year) had higher risk of vertebral fractures.(28,29)

The mechanism of the association between bone loss and fracture risk could not be determined in this epidemiological study. However, greater bone loss could be a marker of underlying poor health. In this study, women who did not survive the follow-up period had significantly higher rate of bone loss than those who survived. Greater bone loss would be expected to be associated with cumulative macro- and microarchitectural damage,(30) resulting in weaker bone than suggested by overall BMD alone.

It is well established that bone loss is resulted from an imbalance between two processes of bone resorption and bone formation. The two processes can be measured by bone turnover markers. Indeed, recent data have suggested that the risk of bone loss in postmenopausal women could be predicted by a single baseline biochemical measure of bone turnover(10); however, the relationship was not always consistent.(11,12) Moreover, patients with hip fracture had much higher rates of bone turnover than those without fractures,(13) and the combination of bone turnover markers and BMD could improve the prediction of fracture risk.(13,31) Taken together, this study's results suggest that short-term bone loss could be an indication of fracture risk and that reliable bone turnover markers could be useful in the assessment of fracture risk.

How much fracture incidence can be attributed to baseline BMD and how much to bone loss? In this study, 45 of every 100 fractures were attributable to only three risk factors, namely, osteoporosis, high rate of femoral neck bone loss, and advancing age. However, the attributable fraction for high rate of bone loss was modest, because the combination of osteoporosis and advancing age accounted for most of the attributable fraction. This is because the prevalence of osteoporosis (26%) was higher than the prevalence of high bone loss (10%). Furthermore, 46% of all fractures occurred in women with osteoporosis; on the other hand, 19% of all fractures occurred in women with femoral neck bone loss of at least 5% per year. Therefore, it is not surprising to observe that the attributable risk estimate for high rate bone loss (13%) was lower than for low BMD (32%). However, the attributable proportion was not mutually exclusive, because in either case, there was an interaction between advancing age and osteoporosis, or advancing and high rate of bone loss that significantly contributed to the overall PARF. For example, the combination of age and osteoporosis could explain up to 51% of hip fractures and 24% of clinical vertebral fractures. This clearly suggests that the combination of three risk factors could significantly improve the prediction of fracture risk derived from two or one risk factor.

The relations between these risk factors and fracture was continuous, and it is difficult to define an optimal threshold. Therefore, it must be recognized that any classification of low risk versus high risk based on BMD, bone loss, and age is arbitrary. However, the arbitrary classification in this study was unlikely to overestimate the population attributable risk fraction, because even in the low risk group, the effect of BMD, bone loss, and advancing age on fracture was appreciable. It can thus be argued that the population attributable risk fraction in this study represents a conservative estimate tending to underestimate the potential contributions of these risk factors.

This study also suggests that the existence of other independent risk factors for fracture cannot be ruled out, because 55% of fracture cases were not explained by osteoporosis, bone loss, and advancing age. However, additional risk factors would have to be sufficiently prevalent and strongly related to fracture to make a substantial contribution to the fracture incidence in the general population. In any case, a greater emphasis on the search for additional and non-BMD predictors of fractures is warranted. On the other hand, greater emphasis on the control and improved understanding of the determinants of BMD and bone loss is likely to be critical to the control of fracture incidence at the population level.

The prospective design of this study overcomes many biases inherent in cross-sectional or case-control studies of the association between BMD, bone loss, and fractures. Furthermore, because two measurements of BMD were made before fracture occurrence, it allows associative inference on bone loss and fractures to be drawn with greater confidence. The study was based on a large sample of women with long duration of follow-up. Finally, women who participated in this study were essentially volunteers and were generally healthier than those who did not participate.(32) Therefore, the magnitude of association between bone loss and fracture risk could have been underestimated.

In conclusion, rate of bone loss at the femoral neck is a significant predictor of subsequent osteoporotic fractures, independent of baseline BMD and age. Although the biological mechanism for this predictive relationship is not certain, it seems the combination of low BMD, high rate of bone loss, and advancing age could identify a subgroup of women who were at particularly high risk of fractures, who are clearly an ideal group for intervention to reduce the community burden of fractures.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

The authors thank the expert assistance of Sr Janet Watters and Sr Donna Reeves for the interview, data collection, and measurement of BMD. We also appreciate the invaluable help of the staff of Dubbo Base Hospital. The authors thank Natasa Ivankovic for management of the database. This work was partly supported by the National Health and Medical Research Council of Australia and by untied grants from Lilly US, Merck, Sharp & Dohme Australia, and Aventis Australia.

Dr Eisman received funding and serves as a consultant for Aventis, Eli Lilly and Company, MSD, NPS Pharmaceuticals, Novartis, Organon, Roche, and Servier. All other authors have no conflict of interest.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  • 1
    Cummings SR, Black DM, Rubin SM 1989 Lifetime risks of hip, Colles', or vertebral fracture and coronary heart disease among white postmenopausal women. Arch Intern Med 149: 24452448.
  • 2
    Cummings SR, Nevitt MC, Browner WS, Stone K, Fox K, Ensrud K, Cauley J, Black D, Vogt TM 1995 Risk factors for hip fractures in white women. N Engl J Med 332: 767773.
  • 3
    Nguyen T, Sambrook P, Kelly P, Jones G, Lord S, Freund J, Eisman J 1993 Prediction of osteoporotic fractures by postural instability and bone density. BMJ 307: 11111115.
  • 4
    Riggs BL, Wahner HW, Seeman E, Offord KP, Dunn WL, Mazess RB, Johnson KA, Melton LJ 1992 Changes in bone mineral density of the proximal femur and spine with aging: Differences between the postmenopausal and senile osteoporosis syndromes. J Clin Invest 70: 716723.
  • 5
    Jones G, Nguyen TV, Sambrook PN, Kelly PJ, Eisman JA 1994 Progressive femoral neck bone loss in the elderly: Longitudinal findings from the Dubbo Osteoporosis Epidemiology Study. BMJ 309: 691695.
  • 6
    Ensrud KE, Palermo L, Black DM, Cauley JA, Jergas M, Orwoll ES, Nevitt MC, Fox KM, Cummings SR 1995 Hip and calcaneous bone loss increase with advancing age: Longitudinal results from the Study of Osteoporotic Fracture. J Bone Miner Res 10: 17781787.
  • 7
    Nguyen TV, Sambrook PN, Eisman JA 1998 Physical activity, weight change, bone loss and fracture in elderly women: The Dubbo Osteoporosis Epidemiology Study. J Bone Miner Res 13: 14581467.
  • 8
    Garnero P, Sornay-Rendu E, Duboeuf F, Delmas PD 1999 Markers of bone turnover predict postmenopausal forearm bone loss over 4 years: The OFELY study. J Bone Miner Res 14: 16141621.
  • 9
    Bauer DC, Sklarin PM, Stone KL, Black DM, Nevitt MC, Ensrud KE, Arnaud CD, Genant HK, Garnero P, Delmas PD, Lawaetz H, Cummings SR 1999 Biochemical markers of bone turnover and prediction of hip bone loss in older women: The study of osteoporotic fractures. J Bone Miner Res 14: 14041410.
  • 10
    Dresner-Pollak R, Parker RA, Poku M, Thompson J, Seibel MJ, Greenspan SL 1996 Biochemical markers of bone turnover reflect femoral bone loss in elderly women. Calcif Tissue Int 59: 328333.
  • 11
    Keen RW, Nguyen T, Sobnack R, Perry LA, Thompson PW, Spector TD 1996 Can biochemical markers predict bone loss at the hip and spine?: A 4-year prospective study of 141 early postmenopausal women. Osteoporos Int 6: 399406.
  • 12
    Looker AC, Bauer DC, Chesnut CH III, Gundberg CM, Hochberg MC, Klee G, Kleerekoper M, Watts NB, Bell NH 2000 Clinical use of biochemical markers of bone remodeling: Current status and future directions. Osteoporos Int 11: 467480.
  • 13
    Garnero P, Sornay-Rendu E, Claustrat B, Delmas PD 2000 Biochemical markers of bone turnover, endogenous hormones and the risk of fractures in postmenopausal women: The OFELY study. J Bone Miner Res 15: 15261536.
  • 14
    Garnero P, Hausherr E, Chapuy MC, Marcelli C, Grandjean H, Muller C, Cormier C, Breart G, Meunier PJ, Delmas PD 1996 Markers of bone resorption predict hip fracture in elderly women: The EPIDOS Prospective Study. J Bone Miner Res 11: 15311538.
  • 15
    Szulc P, Chapuy MC, Meunier PJ, Delmas PD 1996 Serum undercarboxylated osteocalcin is a marker of the risk of hip fracture: A three year follow-up study. Bone 18: 487488.
  • 16
    Riis BJ, Hansen MA, Jensen AM, Overgaard K, Christiansen C 1996 Low bone mass and fast rate of bone loss at menopause: Equal risk factors for future fracture: A 15-year follow-up study. Bone 19: 912.
  • 17
    Jones G, Nguyen T, Sambrook PN, Kelly PJ, Gilbert C, Eisman JA 1994 Symptomatic fracture incidence in elderly men and women: The Dubbo Osteoporosis Epidemiology Study. Osteoporos Int 4: 277282.
  • 18
    Nguyen TV, Kelly PJ, Sambrook PN, Gilbert C, Pocock NA, Eisman JA 1994 Life-style factors and bone density in the elderly: Implications for osteoporosis prevention. J Bone Miner Res 9: 13391346.
  • 19
    Nguyen TV, Sambrook PN, Eisman JA 1997 Source of variability in bone density: Implication for study design and analysis. J Bone Miner Res 12: 124135.
  • 20
    Cox DR 1972 Regression models and life-tables (with discussion). J Roy Statist Soc Series B 34: 187220.
  • 21
    Lee ET 1980 Statistical Models for Survival Data Analysis. Lifetime Learning Publications, Belmont, CA, USA.
  • 22
    SAS Institute 1992 SAS Technical Report P-229, SAS/STAT Software: Changes and Enhancements, Release 6.07. SAS Institute, Cary, NC, USA.
  • 23
    Eide GE, Gefeller O 1995 Sequential and average attributable fractions as aids in the selection of preventive strategies. J Clin Epidemiol 48: 645655.
  • 24
    Gefeller O, Land M, Eide GE 1998 Averaging attributable fractions in the multifactorial situation: Assumptions and interpretation. J Clin Epidemiol 5: 437441.
  • 25
    Cummings SR, Karpf DB, Harris F, Genant HK, Ensrud K, LaCroix AZ, Black DM 2002 Improvement in spine bone density and reduction in risk of vertebral fractures during treatment with antiresorptive drugs. Am J Med 112: 281289.
  • 26
    Chesnut CH, Rosen CJ for the Bone Quality Discussion Group 2002 Reconsidering the effects of antiresorptive therapies in reducing osteoporotic fracture. J Bone Miner Res 16: 21632172.
  • 27
    Marshall D, Johnell O, Wedel H 1996 Meta analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ 312: 12541259.
  • 28
    Christiansen C, Hansen MA, Overgaard K, Riis BJ 1993 Prediction of future fracture risk. In ChristiansenC, RiisBJ (eds.) Proceedings of the Fourth International Symposium on Osteoporosis and Consensus Development Conference Handelstrykkeriet Aalborg ApS, Aalborg, Denmark, pp. 5254.
  • 29
    Hansen M, Overgaard K, Riis B, Christiansen C 1991 Role of peak bone mass and bone loss in postmenopausal osteoporosis. BMJ 303: 961964.
  • 30
    Heaney RP 1989 Osteoporotic fracture space: A hypothesis. Bone Miner 6: 113.
  • 31
    Akesson K, Ljunghall S, Jonsson B, Sernbo I, Johnell O, Gardsell P, Obrant KJ 1995 Assessment of biochemical markers of bone metabolism in relation to the occurrence of fracture: A retrospective and prospective population-based study of women. J Bone Miner Res 10: 18231829.
  • 32
    Center JR, Nguyen TV, Schneider D, Sambrook PN, Eisman JA 1999 Mortality after all major types of osteoporotic fracture in men and women: An observational study. Lancet 353: 878882.