HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: 2411051290v1 241/1/190    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Henry, M. J. Articles by Kotowicz, M. A. Search for Related Content PubMed PubMed Citation Articles by Henry, M. J. Articles by Kotowicz, M. A.

Fracture Risk (FRISK) Score: Geelong Osteoporosis Study¹

¹ From the Department of Clinical and Biomedical Sciences, the University of Melbourne, Barwon Health, PO Box 281, Geelong 3220, Victoria, Australia. Received August 2, 2005; revision requested October 6; revision received November 30; accepted January 2, 2006; final version accepted January 13. Supported by the Victorian Health Promotion Foundation, Geelong Region Medical Research Foundation, and NHMRC. Address correspondence to M.J.H. (e-mail: margh{at}BarwonHealth.org.au).

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
Purpose: To develop and evaluate a fracture risk (FRISK) score based on multiple-site bone mineral density (BMD) measurements and other risk factors, to enable prediction of future fracture occurrence.

Materials and Methods: All participants gave written informed consent, and the study was approved by the Barwon Health Research and Ethics Advisory Committee. BMD was measured at the femoral neck and spine in two concurrently recruited groups: women 60 years of age or older who had sustained a low-trauma fracture of the hip, spine, humerus or distal forearm during a 2-year ascertainment period (n = 231; mean age, 74 years ± 7 [standard deviation]) and a population-based random sample of women who had not sustained a fracture during the recruitment period (n = 448; mean age, 72 years ± 8). Falls in the previous year and the number of self-reported fractures in adult life were recorded. Coefficients of a multiple logistic regression model were used as weightings for a combined model. A longitudinal population-based sample was used to assess the fracture risk equation (n = 600; median age, 74 years; interquartile range, 67–82 years).

Results: The FRISK score was obtained from the following equation: 9.304 – 4.735BMD_SP – 4.530BMD_FN + 1.127FS + 0.344NPF + 0.037W, where BMD_SP is spinal BMD (in grams per square centimeter), BMD_FN is femoral neck BMD, FS is falls score, NPF is number of previous fractures, and W is weight (in kilograms). The FRISK score successfully predicted 75% of fractures 2 years after baseline measurements in subjects in the longitudinal study with 68% specificity.

Conclusion: This study resulted in the derivation of a fracture risk score that successfully predicted 75% of fractures 2 years after baseline.

© RSNA, 2006

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
Fragility fractures occur because bones with reduced strength are subjected to critical forces, which most often arise from a fall (1,2). Forces generated from falls will depend on the characteristics of the fall, which are influenced by age and the weight of the faller. Bone mineral density (BMD) is a good surrogate for bone strength. However, a substantial proportion of women with fractures do not have osteoporosis (T score, less than –2.5) according to bone densitometric criteria (3,4) because BMD fails to capture other factors that influence strength, such as bone size, bone geometry, and microarchitectural change (5).

Many studies have focused on single BMD sites to predict fracture (3,6–10). However, it is common in clinical practice for BMD to be measured at both the spine and the hip, but there is little current evidence-based data to assist in making treatment decisions on the basis of multiple-site BMD assessment (11–13). Measurement of BMD at a single anatomic site minimizes the cost of bone mass assessment; however, because BMD measurements can be discordant across anatomic sites (13–15), the site of measurement may be critical for optimal fracture prediction. Discordance may be due to differences in the age at peak BMD (16); relative proportions of cortical and cancellous bone; variable rates of bone loss (17); soft-tissue calcification associated with osteoarthritis, aortic calcification, and osteophytes; or poor precision associated with the proximal femur measurement. When there is discordance in BMD between the sites, how should the different sites be weighted in terms of fracture risk? Is mild reduction in BMD at the hip and spine associated with a similar fracture risk as moderate reduction at one and not at the other? Clearly, there is a need to develop a clinically useful fracture risk assessment tool on the basis of multiple-site BMD measurements and other non–BMD-related risk factors.

The purpose of our study was to develop and evaluate a fracture risk (FRISK) score, on the basis of multiple-site BMD measurements and other risk factors, to enable prediction of future fracture occurrence.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
All participants gave written informed consent, and the study was approved by the Barwon Health Research and Ethics Advisory Committee.

Study Region
Caucasian subjects were recruited from the Barwon Statistical Division (population, 240 334), a region in southeastern Australia that consists of urban, semiurban, and rural communities (16).

Cross-sectional Cohort of Cases of Fracture
Cases of fracture were ascertained by four research scientists (including K.M.S.) from radiology reports from the two radiology practices that service the region. Vertebral fracture diagnosis was based entirely on the clinical radiologists' interpretation of spine radiographs. Women who sustained a fracture within a 2-year period (February 17, 1994, to February 16, 1996) were invited to participate in the fracture cohort (participation rate, 77% [832 of 1082]) (18). This method of fracture case ascertainment has been validated (2). Only women who had sustained a minimal traumatic fracture of the hip, spine, or humerus or a Colles fracture (fractures included in the World Health Organization definition of osteoporosis [19]) and who were 60 years of age or older were included in this study (n = 231; mean age, 74 years ± 7 [standard deviation]). The median time of assessment was 62 days (interquartile range, 29–99 days) after fracture. Minimal trauma was defined as a fall from less than standing height, a spontaneous fracture, or striking against or being struck accidentally by persons or objects. These patients with fracture included those who had a fracture at more than one anatomic site. The total number of fractures included 54 hip fractures (29 in the femoral neck, 25 in the trochanter), 90 spine fractures, 25 humerus fractures (23 in the proximal humerus, two in the distal humerus) and 65 Colles fractures.

Cross-sectional Control Group
A random age-stratified population-based sample of women was concurrently drawn from the Commonwealth electoral rolls of the same region (participation rate, 77% [1494 of 1938]) (16). The electoral roll provides a complete listing of the adult population because enrollment is compulsory in Australia. Recruitment occurred between January, 1994 and December, 1997. A random selection of women stratified to match the age profile of the region who were 60 years of age or older and who had not sustained a fracture within the 2-year period were used as control subjects for this study (n = 448; mean age, 72 years ± 8).

Longitudinal Population Sample
An age-stratified population-based sample of women aged 60 years and older (n = 600; median age, 74 years; interquartile range, 67–82 years) with a median follow-up time of 5.8 years (interquartile range, 5.2–6.8 years) was recruited from January, 1994 to December, 1997 (16). Fractures that occurred from 2 to 8 years after baseline (24 hip fractures, 41 spine fractures, eight humerus fractures [six in the proximal humerus, two in the distal humerus], and nine Colles fractures) that were reported at each biennial follow-up examination were confirmed by using radiologic reports. These prospectively ascertained fractures were used to assess the fracture risk equations that were derived from the cross-sectional data. Cases were censored from the data set after the first fracture.

Measurements
BMD was measured at the posteroanterior spine (L2 through L4) and the left femoral neck with a densitometer (DPX-L, software version 1.31; Lunar, Madison, Wis) by research scientists who were licensed to operate the bone densitometer (J.A.P., K.M.S., and two other scientists, all with 3 years of experience in bone densitometry). None of the cases or control subjects was excluded on the basis of their lumbar spine measurement. For proximal femur fracture cases, BMD was measured on the side contralateral to the fracture. Short-term precision in vivo BMD measurements yielded mean coefficient of variation values of 0.6% at the posteroanterior spine and 1.6% at the femoral neck. Measurement of a phantom was performed 3 times each week to maintain quality assurance. T scores for the hip and spine were calculated by using the Australian BMD reference range (20). Age, an estimate of the number of falls in the previous year, and the number of fractures resulting from a minimal trauma event since age 20 years were determined by using a questionnaire that was administered by trained interviewers (J.A.P., K.M.S., and two other scientists). A falls score was determined from the number of falls in the previous year, where a score of 1 indicated that the subject never or rarely fell; a score of 2, that the subject fell a few times; a score of 3, that the subject fell several times; and a score of 4, that the subject fell regularly. Weight and height were measured with subjects wearing a hospital gown and with bare or stockinged feet.

Statistical Analysis
Unadjusted comparisons for the cross-sectional cases and control subjects were performed by using the two-sample t test, the Mann-Whitney test, and the ² statistic. The statistic was used to analyze agreement of categories of BMD (osteoporotic, osteopenic, and normal) across anatomic sites.

From the cross-sectional data, optimal thresholds, sensitivity, specificity, positive predictive value, negative predictive value, and the probability of distinguishing fracture cases from control subjects were derived by using receiver operating characteristic (ROC) curves. Combined models of weighted fracture risk scores were obtained by using coefficients of logistic regression equations, and these risk scores were assessed by using ROC curves. Interaction terms were considered, but none were significant. The optimal combined model for fracture risk was derived by applying a backward elimination of regressor variables to logistic regression, and a FRISK score was derived and simplified for use by converting the fracture risk score derived by using logistic regression to a 0–10 scale. Significance was set at a level of = .10. The larger the score, the greater the likelihood of fracture. The optimal cut point in the FRISK score was derived by using ROC curves. Logistic regression was used to assess the increase in fracture risk per unit increase in FRISK score. The FRISK score derived from the cross-sectional data was tested in the longitudinal population sample, and sensitivity and specificity were assessed. Software packages (Minitab, version 13, Minitab, State College, Pa; SPSS, version 11.5, SPSS, Chicago, Ill; and SAS, version 8.2, SAS Institute, Cary, NC) were used for all statistical analyses (performed by M.J.H.).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
BMD across Anatomic Sites
Comparisons between the fracture and control groups are shown in Table 1. Imperfect agreement was found for the comparison of BMD categories at the posteroanterior spine with BMD categories at the femoral neck, with discordance in 48% of the fracture cases ( = 0.21 ± 0.05 [standard error], P < .001) and 44% of the control subjects ( = 0.32 ± 0.03, P < .001) (Table 2).

View larger version (16K): [in this window] [in a new window] [Download PPT slide]   Scores for predicting fracture risk. BMDFN = femoral neck BMD, BMDSP = spine BMD, TSCOREFN = femoral neck T score, TSCORESP = spine T score.

According to the equation, a decrease of 1.0 standard deviation in BMD at the posteroanterior spine increases the FRISK score by 0.6. This increase in fracture risk is the equivalent of decreasing the femoral neck BMD by 1.1 standard deviations, increasing the weight by 17 kg, increasing the number of previous fractures by 1.9, or increasing the falls score by 0.6.

The application of FRISK scores that incorporated combinations of T scores for the hip and spine (0, –1.0, and –2.5) and variable falls scores and numbers of previous fractures (Table 5) indicated that for a "nonfaller" (falls score = 1) of average weight with no previous fracture who had a T score of 0 at the posteroanterior spine and femoral neck, the FRISK score would be 2.4. By contrast, a woman who had fallen frequently (falls score = 4), had sustained three previous fractures, and had a posteroanterior spine and femoral neck T score of –2.5 would have a FRISK score of 9.8. Increasing weight increases the FRISK score.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

Measurement of bone mass at single sites has been used for fracture risk prediction (6,22), and some studies have involved the use of combined models to predict fracture (10,23–25). However, few studies have examined BMD measured at multiple sites (11–13). Wasnich et al (11) observed that prediction of spine fracture was improved by including a combination of bone mass at the distal radius and that at the calcaneus. Furthermore, Hans et al (12) reported that discrimination of patients with hip fracture from those without fracture with quantitative ultrasonography was improved by using multiple sites. Lu et al (13) combined the osteoporotic status of multiple sites within the hip or forearm to improve fracture prediction. Similarly, our analysis predicts fracture risk by combining BMD, falls score, number of previous fractures, and weight. At univariate analysis, BMD was the most sensitive to fracture, but the falls score was the most specific. In the combined model, age did not remain significant, but many of the variables are age related and may act as surrogates for age.

Reduced body weight is associated with an increased risk of hip fracture; however, after adjustment for BMD, reduced body weight was no longer a significant predictor (23), suggesting that the relationship was due to women with low body weights having lower BMDs at the hip. In this study of any low-trauma fracture, increasing weight produced an increase in fracture risk after adjustment for BMD, supporting the notion that a greater body weight increases the force applied to the skeleton during a fall. Approximately 80% of nonspinal fractures result from a fall, and the mechanics of the fall partly determine the site of the fracture (26).

A limitation of this study was that fracture cases were determined by a review of radiographic reports. The radiologists' assessment was subjective, vertebral heights were not measured, and a spinal deformity index was not used. Because spine radiography was not routinely performed in the control group and because approximately two-thirds of vertebral fractures are not associated with fracture-related back pain (27,28), it is likely that spine fractures were underrepresented in this cohort. We also acknowledge that the accuracy of self-reports of previous fractures may have been limited by recall bias. Another limitation of this study was that the only site of the proximal femur that was represented was the femoral neck. This subregion was chosen because it is the site at which measurements are most reproducible (16). More recently it has become common practice for the total hip to be measured. A FRISK score based on hip fracture risk would be of considerable interest, but this study lacked the power to investigate this site alone.

In practice, treatment decisions are based not on any single risk factor alone but on a combination of factors (29). The results of our study suggest that an overall assessment of fracture risk with a model that includes a combination of BMDs at the hip and spine, together with the number of previous fractures, a falls score, and weight can predict future fracture with good sensitivity and specificity. We believe the FRISK score presented in this study can be of assistance in making treatment decisions.

	ADVANCE IN KNOWLEDGE

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 

• The fracture risk score enabled prediction of 75% of fractures 2 years after baseline.
  

	ACKNOWLEDGMENTS

 
We thank Biljana Skoric, BSc, and Soheila Panahi, BSc, for their assistance with scanning subjects with the bone densitometer.

	FOOTNOTES

 

Abbreviations: BMD = bone mineral density • FRISK = fracture risk • ROC = receiver operating characteristic

Authors stated no financial relationship to disclose.

Author contributions: Guarantors of integrity of entire study, G.C.N., M.A.K.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, M.J.H., J.A.P., M.A.K.; statistical analysis, M.J.H., J.A.P.; and manuscript editing, all authors

	References

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 

1. Nevitt M, Cummings SR. Falls and fractures in older women. In: Vallas B, Toupet M, Rubenstein L, Albarede J, Christen Y, eds. Falls, balance and gait disorders in the elderly. Paris, France: Elsevier, 1992; 69–80.
2. Pasco JA, Henry MJ, Gaudry TM, Nicholson GC, Kotowicz MA. Identification of incident fractures: the Geelong Osteoporosis Study. Aust N Z J Med 1999;29:203–206.[Medline]
3. Henry MJ, Pasco JA, Seeman E, Nicholson GC, Sanders KM, Kotowicz MA. Fracture thresholds revisited: Geelong Osteoporosis Study. J Clin Epidemiol 2002;55:642–646.[CrossRef][Medline]
4. Siris ES, Chen YT, Abbott TA, et al. Bone mineral density thresholds for pharmacological intervention to prevent fractures. Arch Intern Med 2004;164:1108–1112.[Abstract/Free Full Text]
5. Seeman E. From density to structure: growing up and growing old on the surfaces of bone. J Bone Miner Res 1997;12:509–521.[CrossRef][Medline]
6. Wasnich RD, Ross PD, Heilbrun LK, Vogel JM. Selection of the optimal skeletal site for fracture risk prediction. Clin Orthop Relat Res 1987;216:262–269.
7. Hui SL, Slemenda CW, Johnston CC Jr. Baseline measurement of bone mass predicts fracture in white women. Ann Intern Med 1989;111:355–361.[Medline]
8. Gardsell P, Johnell O, Nilsson BE. Predicting fractures in women by using forearm bone densitometry. Calcif Tissue Int 1989;44:235–242.[Medline]
9. Cleghorn DB, Polley KJ, Bellon MJ, Chatterton J, Baghurst PA, Nordin BE. Fracture rates as a function of forearm mineral density in normal postmenopausal women: retrospective and prospective data. Calcif Tissue Int 1991;49:161–163.[Medline]
10. Nguyen T, Sambrook P, Kelly P, et al. Prediction of osteoporotic fractures by postural instability and bone density. BMJ 1993;307:1111–1115.[Medline]
11. Wasnich RD, Ross PD, Davis JW, Vogel JM. A comparison of single and multi-site BMC measurements for assessment of spine fracture probability. J Nucl Med 1989;30:1166–1171.[Abstract/Free Full Text]
12. Hans D, Srivastav SK, Singal C, et al. Does combining the results from multiple bone sites measured by a new quantitative ultrasound device improve discrimination of hip fracture? J Bone Miner Res 1999;14:644–651.
13. Lu Y, Genant HK, Shepherd J, et al. Classification of osteoporosis based on bone mineral densities. J Bone Miner Res 2001;16:901–910.[CrossRef][Medline]
14. Abrahamsen B, Hansen TB, Jensen LB, Hermann AP, Eiken P. Site of osteodensitometry in perimenopausal women: correlation and limits of agreement between anatomic regions. J Bone Miner Res 1997;12:1471–1479.[CrossRef][Medline]
15. Larcos G. Predicting clinical discordance of bone mineral density. Mayo Clin Proc 1998;73:824–828.[Medline]
16. Henry MJ, Pasco JA, Nicholson GC, Seeman E, Kotowicz MA. Prevalence of osteoporosis in Australian women: Geelong Osteoporosis Study. J Clin Densitom 2000;3:261–268.[CrossRef][Medline]
17. Dennison E, Eastell R, Fall CH, Kellingray S, Wood PJ, Cooper C. Determinants of bone loss in elderly men and women: a prospective population-based study. Osteoporos Int 1999;10:384–391.[CrossRef][Medline]
18. Sanders KM, Pasco JA, Ugoni AM, et al. The exclusion of high trauma fractures may underestimate the prevalence of bone fragility fractures in the community: the Geelong Osteoporosis Study. J Bone Miner Res 1998;13:1337–1342.[CrossRef][Medline]
19. World Health Organization Study Group. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis: report of a WHO Study Group. World Health Organ Tech Rep Ser 843:1-129. Geneva, Switzerland: World Health Organization, 1994.
20. Henry MJ, Pasco JA, Pocock NA, Nicholson GC, Kotowicz MA. Reference ranges for bone densitometers adopted Australia-wide: Geelong osteoporosis study. Australas Radiol 2004;48:473–475.[CrossRef][Medline]
21. Ross PD, Genant HK, Davis JW, Miller PD, Wasnich RD. Predicting vertebral fracture incidence from prevalent fractures and bone density among non-black, osteoporotic women. Osteoporos Int 1993;3(3):120–126.[CrossRef][Medline]
22. Cummings SR, Black DM, Nevitt MC, et al. Bone density at various sites for prediction of hip fractures. Lancet 1993;341:72–75.[CrossRef][Medline]
23. Ensrud KE, Lipschutz RC, Cauley JA, et al. Body size and hip fracture risk in older women: a prospective study. Study of Osteoporotic Fractures Research Group. Am J Med 1997;103:274–280.
24. Cummings SR, Nevitt MC, Browner WS, et al. Risk factors for hip fracture in white women. Study of Osteoporotic Fractures Research Group. N Engl J Med 1995;332:767–773.
25. Burger H, De Laet CE, Weel AE, Hofman A, Pols HA. Added value of bone mineral density in hip fracture risk scores. Bone 1999;25:369–374.[Medline]
26. Cummings SR, Nevitt MC. Non-skeletal determinants of fractures: the potential importance of the mechanics of falls. Study of Osteoporotic Fractures Research Group. Osteoporos Int 1994;4(suppl 1):67–70.[CrossRef][Medline]
27. Cooper C, Atkinson EJ, O'Fallon WM, Melton LJ 3rd. Incidence of clinically diagnosed vertebral fractures: a population-based study in Rochester, Minnesota, 1985–1989. J Bone Miner Res 1992;7:221–227.[Medline]
28. Kotowicz MA, Melton LJ 3rd, Cooper C, Atkinson EJ, O'Fallon WM, Riggs BL. Risk of hip fracture in women with vertebral fractures. J Bone Miner Res 1994;9:599–605.[Medline]
29. Siminoski K, Leslie WD, Frame H, et al. Recommendations for bone mineral density reporting in Canada. Can Assoc Radiol J 2005;56:178–188.[Medline]

This article has been cited by other articles:

				J. Gorricho, J. Erviti, M. J. Henry, J. A. Pasco, K. M. Sanders, G. C. Nicholson, and M. A. Kotowicz Fracture Risk Score Radiology, October 1, 2007; 245(1): 298 - 298. [Full Text] [PDF]

				R. Lehman Evidently. . . Evid. Based Med., February 1, 2007; 12(1): 8 - 8. [Full Text] [PDF]

				Other articles noted Evid. Based Med., February 1, 2007; 12(1): 31 - 32. [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: 2411051290v1 241/1/190    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Henry, M. J. Articles by Kotowicz, M. A. Search for Related Content PubMed PubMed Citation Articles by Henry, M. J. Articles by Kotowicz, M. A.