Obstetrics & Gynecology:
Age as a Predictor of Osteoporotic Fracture Compared With Current Risk-Prediction Models
Jiang, Xuezhi MD; Westermann, Lauren B. DO; Galleo, Gabriella V. BS; Demko, John BS; Marakovits, Kimberly A. BA; Schnatz, Peter F. DO
Departments of Obstetrics–Gynecology and Internal Medicine, the Reading Hospital, Reading, Pennsylvania; Philadelphia College of Osteopathic Medicine, Philadelphia, Pennsylvania; the Departments of Obstetrics–Gynecology and Internal Medicine, Jefferson Medical College of Thomas Jefferson University, Philadelphia, Pennsylvania.
Corresponding author: Xuezhi Jiang, MD, Assistant Professor of Obstetrics–Gynecology, Jefferson Medical College of Thomas Jefferson University, Clerkship Director, The Reading Hospital, Department of Obstetrics–Gynecology–R1, PO Box 16052, Reading, PA 19612-6052; e-mail: firstname.lastname@example.org.
The authors thank Taghogho Agarin, MD, Marianne Muchura, MD, Charnetta Smith, MD, Jessica Abrantes, Alison Romegialli, David Cunningham, Sarah Dobrowolski, and Barbara Levarge, MD, for their help with patient recruitment and data collection.
Presented as a poster at the American College of Obstetricians and Gynecologists' 59th Annual Clinical Meeting, April 30–May 4, 2011, Washington, DC.
Financial Disclosure The authors did not report any potential conflicts of interest.
OBJECTIVE: To compare several fracture risk-prediction models and their predictive values.
METHODS: Women older than age 49 years were sent for dual-energy X-ray absorptiometry screening between January 2007 and March 2009. Data collection included multiple osteoporosis risk factors. The ability to identify fractures was analyzed and compared using the North American Menopause Society 2006 and 2010 Position Statements, The Fracture Risk Assessment Tool, along with age alone. The area under the curve (AUC) comparison with chance (AUC 0.50) and paired AUC comparisons between models were used to investigate the efficacy of each model in predicting osteoporotic fractures.
RESULTS: Among the 615 women studied, with mean (standard deviation) age of 61.4 (8.3) years and 94.5% being white, 15 have experienced a fracture. All screening approaches were significantly better than chance at predicting fractures. Paired comparisons of the detection ability of fracture prediction models showed no significant differences. Age alone was a significant predictor for fracture (AUC 0.79, 95% confidence interval [CI] 0.67–0.91, P<.001) with the optimal cutoff age of 65 years, which was associated with a sensitivity (95% CI) of 80% (77–83%) and specificity (95% CI) of 73% (70–77%). Compared with young postmenopausal women (younger than 65 years), the odds ratio (95% CI) of fractures in older women (65 years or older) is 10.2 (2.32–44.97). In addition, when age was added, it significantly increased the AUC of each model.
CONCLUSION: These data suggest that all current screening modalities are effective in predicting fracture but not significantly better than age alone. Age should be considered carefully while evaluating patients for osteoporosis screening and treatment.
LEVEL OF EVIDENCE: II
Osteoporosis has become a major health concern throughout the world. It is often asymptomatic and unrecognized in the early stages, causing the disease to be frequently underdiagnosed and undertreated.
Currently, the most common osteoporosis screening tool is the dual-energy X-ray absorptiometry (DXA) device. Generally, the measurements are taken from the spine, hip, wrist, and femur.1 Dual-energy X-ray absorptiometry scanning, although widely used, is rather limited. This test can be helpful in determining how dense bones are, but not necessarily how strong they are.2 Furthermore, only one factor is considered when bone mineral density (BMD) measurements are the sole marker used to identify patients at risk for fracture or the need for treatment. Therefore, in 2008, the World Health Organization developed a more systematic approach to fracture prediction, the Fracture Risk Assessment Tool.3,4 Based on a series of risk factors, the Fracture Risk Assessment Tool is used to calculate a 10-year probability that an individual will sustain a hip fracture or a major osteoporotic fracture, including hip, proximal humerus, wrist, and clinical vertebral fractures.3 In 2008 and 2010, the National Osteoporosis Foundation and North American Menopause Society, respectively, recommended that osteoporosis should be treated when a patient has a T-score −2.5 or less or when a patient with a T-score between −1.0 and −2.5 has a 10-year hip fracture risk of 3% or greater or a 10-year major fracture risk of 20% or greater according to the Fracture Risk Assessment Tool.5,6 Although the Fracture Risk Assessment Tool is a major step forward in our approach to osteoporosis screening and treatment, it is far from perfect and is still being refined.7 The Fracture Risk Assessment Tool fails to recognize several important risk factors for fracture in its calculation, including bone turnover markers,8 vitamin D deficiency,9 risk of falls,10 physical activity,11 previous osteoporosis treatment, and certain medications12,13 (including aromatase inhibitors14 and androgen deprivation therapy15). Because the Fracture Risk Assessment Tool does not account for the number of previous fractures, a patient with two or more previous osteoporotic fractures is considered to have the same relative risk as a patient with only one previous fracture.16 Also, the Fracture Risk Assessment Tool is only intended to be used for those patients who are not being treated for osteoporosis.17 With these stated limitations, and the fact that the Fracture Risk Assessment Tool is relatively new, data assessing its validity and predictive capabilities will be extremely helpful.
The North American Menopause Society first proposed evidence-based management guidelines for osteoporosis in 200218 with subsequent updates in 2006 and 2010.6,19 These algorithms help to identify patients in need of screening and treatment. However, the predictive power of these tools is not well understood.
Multiple study results indicate that age is a particularly strong risk factor for fracture.6,20–22 The hypothesis of the current study is that age alone may be as predictive as the Fracture Risk Assessment Tool and the North American Menopause Society Position Statement treatment guidelines. We seek to determine the predictive values of the various osteoporosis management models and to compare the predictive value of age alone with these models.
PATIENTS AND METHODS
This research and protocol was reviewed and approved by the Hartford Hospital institutional review board and the current research is being done under the review and approval of The Reading Health System institutional review board. Postmenopausal women aged 49 years and older presenting for a screening DXA scan were given a description of this project and were invited to take part in the study. Menopausal-aged women were recruited between January 1, 2007, and March 1, 2009. After their recruitment and consent, each individual was contacted by telephone by a member of the research team, verbally consented, and interviewed. This phone survey included questions regarding each person's age, weight, height, race, osteoporosis risk factors, and medications along with family and personal medical histories; in addition, women were asked to report any fragility fractures of the spine or hip that occurred after the age of 50.23 Each woman had a DXA scan at one of four Jefferson radiology testing sites in the greater Hartford, Connecticut, area. The results from each scan were obtained and incorporated into the database. Three of the women surveyed failed to provide the researcher with their age and were consequently eliminated from the study. Osteoporosis was defined as a T-score of −2.5 or lower and low bone mass was defined as a T-score between −1 and −2.5.24 The T-score used for the various algorithms in this analysis, including the Fracture Risk Assessment Tool, is the worst overall score among the BMD measurements of L1–L4, the femoral neck, and the total femur.
In this study, the Fracture Risk Assessment Tool was used to determine the 10-year probability that postmenopausal women will sustain any one of four major osteoporotic fractures (hip, proximal humerus, wrist, and vertebral) and the 10-year probability for a hip fracture.3 The objective of this study was to compare the likelihood of identifying fractures generated by age alone, the Fracture Risk Assessment Tool, and the North American Menopause Society osteoporosis treatment guidelines from 20067 and 2010.6 The data and information from all women, who were originally recruited for a study analyzing the association of previous pregnancies, breastfeeding, or both with osteoporosis,23 were entered into the Fracture Risk Assessment Tool and algorithms from the 2006 and 2010 North American Menopause Society osteoporosis position statements. If the outcome variable, such as fracture, was a part of the predictive factors in the model, the risk factor was not used to predict that same outcome (eg, for the 2006 North American Menopause Society position statement, when assessing the ability to predict a fracture, the knowledge of a fracture was not used in the model, although BMD information was available to be used in the model for the calculation). Fracture information was likewise not used in the Fracture Risk Assessment Tool (Table 1). The women's ages were stratified to assess how much, if at all, age plays a factor in determining whether there is a difference among the three methods. A secondary goal of this study was to find out how accurate the Fracture Risk Assessment Tool is in identifying women with a T-score of −2.5 or less at high risk of fractures. In this case, the BMD was not entered into the Fracture Risk Assessment Tool and patients were considered at high risk of fractures when they had a probability of hip fracture greater than 3% or a probability of a major osteoporotic fracture greater than 20%.
The a priori power analysis determined the need for approximately 600 participants to detect a 10% difference in the prevalence of osteoporosis between a breastfeeding and nonbreastfeeding group (eg, 20% compared with 10%, respectively). For this assessment using a finite sample of approximately 615 cases, a χ2 of proportions would afford at least 80% power to detect a maximum difference of 8% in the probabilities generated by any two of the prediction models (eg, 55% compared with 63%, 60% compared with 68%). Data were evaluated for normality of distribution. Area under the curve (AUC) analyses using logistic regression were performed to compare each model with “chance” (AUC 0.5) and to evaluate usefulness of each model in predicting study outcomes. Paired AUC comparisons were conducted between all models to evaluate predictive superiority of each assessment tool. All data were analyzed using SAS 9.2 at an α level of .05 such that differences and results yielding P<.05 will be deemed statistically significant.
All women older than 49 years who presented to one of the four radiology centers in Hartford, Connecticut, for a DXA scan were offered enrollment into this study. A total of 1,253 women presented; 1,124 of them were eligible for the study, whereas 129 women met one or more exclusion criteria. The exclusion criteria included an unsigned Health Insurance Portability and Accountability Act form, being younger than age 49 years, lack of time to participate, and being unavailable for follow-up. Of all eligible women, 724 were enrolled into the study, and a total of 615 were successfully contacted and agreed to complete the telephone survey with mean (standard deviation) age of 61.4 (8.3) years. Of the 615 women, 13 of them were of “mixed race.” All women who were of mixed race in addition to whites were entered into the Fracture Risk Assessment Tool as white because the white race has the highest fracture risk of all races. The racial composition of the sample of 615 women was 94.5% white, 0.7% Hispanic, 1.8% African American, 1.2% Asian, and 0.2% Native American. The remaining 1.6% could not identify themselves with any of the previously mentioned races and, therefore, considered themselves to be of “other” race (Fig. 1). Fifteen women have experienced a fracture, with mean age (standard deviation) of 70.7 (9.5) years, and were significantly older than those who have not experienced a fracture (61.2 [8.09] years, P<.001). Of the 15 women, nine (60%) were diagnosed with osteoporosis, and this was significantly higher than those who have not experienced a fracture (59 [10%], P<.001). However, there was no significant difference in the percentage of white women with fractures among all who had a fracture and those who did not. (100% compared with 94%, P=.34).
This chart reveals t...Image Tools
After all women were entered into the Fracture Risk Assessment Tool, nine of the 15 women who actually experienced a fracture before undergoing risk assessment were found to meet criteria for treatment. In other words, nine of the 15 women with fractures were identified by the Fracture Risk Assessment Tool as having a high enough risk of fracture to require treatment based on a 10-year probability of hip fracture of 3% or greater or a probability of a major osteoporotic fracture of 20% or greater. Next, when all 15 women were entered into the North American Menopause Society 2006 and 2010 position statement algorithms, nine and 12 women were considered to need treatment, respectively.
Of the three methods, the North American Menopause Society 2010 position statement identified patients with fractures most often. All women deemed treatment-worthy according to the Fracture Risk Assessment Tool and the 2006 North American Menopause Society treatment guidelines were also found to need treatment according to the 2010 North American Menopause Society guidelines. In other words, the North American Menopause Society 2010 position statement did not fail to recognize any women who were considered to need treatment according to the two other methods of identifying treatable women.
In AUC analyses, all prediction models studied are significantly better than chance and useful for correctly predicting fractures (Table 2). There were no significant differences noted between any two models from paired AUC comparisons. Age alone is a significant predicting factor for fracture (AUC 0.79, 95% confidence interval [CI] 0.67–0.91, P<.001) with the optimal cutoff at age 65 years, which was associated with a sensitivity (95% CI) of 80% (77–83%) and a specificity (95% CI) of 73% (70–77%) (Fig. 2). Compared with younger postmenopausal women (younger than 65 years), the crude and adjusted odds ratios (95% CI) of fracture in older women (65 years or older) are 10.4 (2.89–37.22) and 10.2 (2.32–44.97), respectively (Table 3). When age was added to the models as another predicting factor, AUCs of each model were all significantly increased (Table 4).
A secondary goal of this study was to determine the accuracy of the Fracture Risk Assessment Tool in identifying those with a T-score of −2.5 or less at high risk for an osteoporotic fracture as well as the North American Menopause Society 2006 and 2010 position statement algorithms in identifying those with a T-score of −2.5 or less as candidates for DXA testing. There were a total of 70 women in the study who had a T-score of −2.5 or less. Of these 70, the Fracture Risk Assessment Tool was able to identify 30 women (43%) with a T-score of −2.5 or below at high risk of an osteoporotic fracture and labeled 111 out of 545 women with a T-score greater than −2.5 (20%) at high risk of an osteoporotic fracture with sensitivity, specificity, and overall classification accuracy of 43% (95% CI 39–47%), 80% (95% CI 77–83%), and 75% (95% CI 72–78%), respectively. The North American Menopause Society 2006 and 2010 position statement algorithms were able to identify 55 (79%) and 57 (81%) of women with a T-score of −2.5 or below as candidates for DXA testing, respectively. According to the current guideline,6 all women aged 65 years and older are recommended to have BMD measured, regardless of clinical risk factors. Using 65 years of age as a cutoff, 32 (46%) women with a T-score of −2.5 or below were identified as candidates for DXA testing. The sensitivity, specificity, and overall classification accuracy of these osteoporosis screening modalities are listed in Table 5. Because early detection of low bone mass may facilitate early intervention to prevent osteoporosis and fractures, the ability of North American Menopause Society guidelines in identifying women with low bone mass as candidates for DXA testing was also assessed. The sensitivity (95% CI) of the 2006 and 2010 North American Menopause Society algorithms in identifying women with low bone mass as candidates for DXA testing are 55% (51–59%) and 57% (53–61%), respectively.
Osteoporosis is a burdensome medical condition that affects millions nationwide and throughout the world.25 Properly following osteoporosis screening and treatment guidelines is without question important. Postmenopausal women are at risk for osteoporosis and subsequent fractures. However, as previously mentioned, there are numerous other risk factors that have been shown to be associated with fractures. In addition to DXA scanning for BMD, there are a number of fracture risk assessment tools currently available.
This study evaluated three fracture risk-prediction models. In our analysis, all three prediction models were effective tools for predicting fractures. However, it appears that all of these models are no better predictors of fracture than age alone. The data indicate that age may be at least as good of a fracture predictor as the North American Menopause Society 2010 guidelines and the Fracture Risk Assessment Tool with BMD. A large prospective cohort study20 showed that simple models based on age and BMD alone, or age and fracture history alone, predicted the 10-year risk of hip, major osteoporotic, and clinical fractures as well as the more complex Fracture Risk Assessment Tool. In 2009, Donaldson et al22 analyzed data from the placebo groups of the Fracture Intervention Trial and reported that once femoral neck BMD and age are known, the eight additional risk factors in the Fracture Risk Assessment Tool do not significantly improve the prediction of vertebral fracture. Another group of investigators21 evaluated an assessment tool called the FRACTURE index using data from the Study of Osteoporotic Fractures, including a total population of 7,782 women aged 65 years and older. The index was comprised of a set of seven variables including age, T-score, fracture after age 50 years, maternal hip fracture after age 50 years, a weight of 125 pounds or less (57 kg), smoking status, and the need for women to use their arms to stand up from a chair. The study found that the relationship between age and the risk of hip fracture was very strong with age being the single most important component of the index. It was also reported by Kanis et al26 that the risk of increasing age is at least 10-fold greater than the risk of a decreasing BMD.
The Fracture Risk Assessment Tool without BMD was only able to identify 43% of women with osteoporosis at high risk of fracture. Logically, failure to identify those with osteoporosis and appropriately intervene in those at high risk of fractures will increase the overall number of osteoporotic fractures. Therefore, there is much room for improving the clinician's ability to predict and prevent osteoporotic fractures.
While assessing the ability to identify women with osteoporosis as appropriate candidates for DXA testing, the sensitivity, specificity, and accuracy of the 2006 and 2010 North American Menopause Society osteoporosis screening guidelines are comparable. A higher sensitivity is required for a good screening test. Sensitivity is the highest for the 2010 North American Menopause Society screening guideline and is the lowest using a cutoff age of 65 years. However, the low sensitivity of the 2006 and 2010 North American Menopause Society guidelines in identifying women with low bone mass as candidates for DXA testing indicates that the current models would leave many of these high-risk individuals with undetected low bone mass untreated and at risk for significant morbidity and mortality. Without a “perfect” fracture prediction model, clinicians must use these tools carefully. Both a fracture prediction model and well-known risk factors should be considered in the physician's and patient's decision to pursue treatment.2
This study was limited by the fact that data were gathered retrospectively and that all participants with a fracture had already experienced it before their entrance into the study. Most of the screening tools used have been developed in the last few years; therefore, long-term prospective data currently do not exist. It is clear that more prospective research is necessary to develop a screening model with improved predictive ability, and further data are needed to determine the validity of the currently available models. Additional limitations include the fact that the power calculation was for a previous research question, study participants are predominantly white who were preselected by their DXA screening requisition, and the fact that there was no prospective validation of these models.
In conclusion, it is not recommended to abandon the osteoporotic fracture prediction models, because there is no better alternative model available currently. Although age alone should not be used as an independent predictor of fractures, data from this study suggest that age should be carefully considered when evaluating patients for osteoporosis screening and treatment. It appears that although useful, the current models need additional refinement to provide the best fracture prediction for our patients. This will, consequently, improve surveillance, diagnosis, and therapeutic approaches to osteoporosis management.
1. Dontas IA, Yiannakopoulos CK. Risk factors and prevention of osteoporosis-related fractures. J Musculoskelet Neuronal Interact 2007;7:268–72.
2. Dawson-Hughes B, Tosteson ANA, Melton LJ III, Baim S, Favus MJ, Khosla S, et al.. Implications of absolute fracture risk assessment for osteoporosis practice guidelines in the USA. Osteoporos Int 2008;19:449–58.
3. Fardellone P. Predicting the fracture risk in 2008. Joint Bone Spine 2008;75:661–4.
5. National Osteoporosis Foundation. Clinician’s guide to prevention and treatment of osteoporosis. Washington, DC: National Osteoporosis Foundation; 2008.
6. Management of osteoporosis in postmenopausal women: 2010 position statement of The North American Menopause Society. Menopause 2010;17:25–54.
7. Ettinger B, Black DM, Dawson-Hughes B, Pressman AR, Melton LJ III. Updated fracture indices rates for the US version of FRAX®
. Osteoporosis Int 2010;21:25–33.
8. Unnanuntana A, Gladnick BP, Donnelly E, Lane JM. The assessment of fracture risk. J Bone Joint Surg Am 2010;92:743–53.
9. Gerdhem P, Ringsberg KA, Obrant KJ, Akesson K. Association between 25-hydroxyvitamin D levels, physical activity, muscle strength and fractures in the prospective population-based OPRA Study of Elderly Women. Osteoporosis Int 2005;16:1425–31.
10. Tinetti ME. Preventing falls in elderly persons. N Engl J Med 2003;348:42–9.
11. Gregg EW, Cauley JA, Seeley DG, Ensrud KE, Bauer DC. Physical activity and osteoporotic fracture risk in older women. Ann Intern Med 1998;129:81–8.
12. Leipzig RM, Cumming RG, Tinetti ME. Drugs and falls in older people: a systematic review and meta-analysis. I. Psychotropic drugs. J Am Geriatr Soc 1999;47:30–9.
13. Leipzig RM, Cumming RG, Tinetti ME. Drugs and falls in older people: a systematic review and meta-analysis. II. Cardiac and analgesic drugs. J Am Geriatr Soc 1999;47:40–50.
14. Howell A, Cuzick J, Baum M, Buzdar A, Dowsett M, Forbes JF, et al.. Results of the ATAC (Arimidex, Tamoxifen, Alone or in Combination) trial after completion of 5 years’ adjuvant treatment for breast cancer. Lancet 2005;365:60–2.
15. Shahinian VB, Kuo YF, Freeman JL, Goodwin JS. Risk of fracture after androgen deprivation for prostate cancer. N Engl J Med 2005;352:154–64.
16. van den Bergh JP, van Geel TA, Lems WF, Geusens PP. Assessment of individual fracture risk: FRAX and beyond. Curr Osteoporos Rep 2010;8:131–7.
17. Watts NB, Ettinger B, LeBoff MS. Perspective FRAX facts. J Bone Miner Res 2009;24:975–9.
18. North American Menopause Society. Management of postmenopausal osteoporosis: position statement of the North American Menopause Society. Menopause 2002;9:84–101.
19. North American Menopause Society. Management of osteoporosis in postmenopausal women: 2006 position statement of The North American Menopause Society. Menopause 2006;13:340–67.
20. Ensrud KE, Lui LY, Taylor BC, Schousboe JT, Donaldson MG, Fink HA, et al.. A comparison of prediction models for fractures in older women: is more better? Arch Intern Med 2009;169:2087–94.
21. Black DM, Steinbuch M, Palermo L, Dargent-Molina P, Lindsay R, Hoseyni MS, et al.. An assessment tool for predicting fracture risk in postmenopausal women. Osteoporos Int 2001;12:519–28.
22. Donaldson MG, Palermo L, Schousboe JT, Ensrud KE, Hochberg MC, Cummings SR. FRAX and risk of vertebral fractures: the fracture intervention trial. J Bone Miner Res 2009;24:1793–9.
23. Schnatz PF, Barker KG, Marakovits KA, O’Sullivan DM. Effects of age at first pregnancy and breast-feeding on the development of postmenopausal osteoporosis. Menopause 2010;17:1161–6.
24. World Health Organization. Assessment of fracture risk and application to screening for postmenopausal osteoporosis. Geneva, (Switzerland); World Health Organization;1994.
25. National Osteoporosis Foundation. Clinician’s guide to prevention and treatment of osteoporosis. Washington, DC: National Osteoporosis Foundation; 2010.
26. Kanis JA, Johnell O, Oden A, Dawson A, De Laet C, Jonsson B. Ten year probabilities of osteoporotic fractures according to BMD and diagnostic thresholds. Osteoporos Int 2001;12:989–95.
© 2013 by The American College of Obstetricians and Gynecologists.
ACOG MEMBER SUBSCRIPTION ACCESS
If you are an ACOG Fellow and have not logged in or registered to Obstetrics & Gynecology, please follow these step-by-step instructions to access journal content with your member subscription.