Journal of Clinical Densitometry: Assessment of Skeletal Health, vol. 16, no. 2, 231e237, 2013 Ó Copyright 2013 by The International Society for Clinical Densitometry 1094-6950/16:231e237/$36.00 http://dx.doi.org/10.1016/j.jocd.2012.05.007
Original Article
Validation of the FRAX Predictive Model for Major Osteoporotic Fracture in a Historical Cohort of Spanish Women Cristian Tebe Cordomı,*,1,2 Luıs Miguel del Rıo,3,4 Silvana Di Gregorio,2,4 Lidia Casas,2,5,6 Maria-Dolors Estrada,1,2 Anna Kotzeva,1,2 and Mireia Espallargues1,2 1
Agencia d’Informaci o, Avaluaci o i Qualitat en Salut, Barcelona, Spain; 2CIBER de Epidemiologıa y Salud P ublica, 3 Barcelona, Spain; CETIR Grup Medic, Barcelona, Spain; 4Red de Envejecimiento del Carlos III, Barcelona, Spain; 5 Center for Research in Environmental Epidemiology, Barcelona, Spain; and 6Institut de Recerca Hospital del Mar, Barcelona, Spain
Abstract FRAX is a fracture risk assessment tool to estimate the 10-yr probability of a major osteoporotic fracture or a hip fracture. The aim of the study was to assess the predictive ability of FRAX for major osteoporotic fracture in a cohort of Spanish women. The study was based on a retrospective cohort of women aged 40e90 yr. Patients were followed from their first bone densitometry to the first major osteoporotic fracture event (forearm, proximal humerus, clinical spine, or hip fracture) or for 10 yr whichever comes first. A total of 1231 women were included. Bone mineral density data and self-reported data on risk factors for fracture were obtained. The predictive ability of FRAX was assessed by analyzing calibration and discrimination, with the calculation of observed-to-expected (O/E) fracture ratios and the receiver operating characteristic (ROC) curve, respectively. A total of 222 women (18.1%) reported at least 1 fracture after the first assessment. The incidence of fracture was 14 (95% confidence interval [CI]: 10e17), 19 (95% CI: 15e23), 28 (95% CI: 21e36), and 67 (95% CI: 8e125) cases per 1000 woman-years in women aged !55, 55e64, 65e74, and 75 yr, respectively. The O/E ratio was 3.9 (95% CI: 3.4e4.5; p ! 0.0001). The area under the ROC curve was 61% (95% CI: 57e65%). FRAX underestimated the risk of major osteoporotic fracture in this cohort of Spanish women, particularly in those with a low risk of fracture according to the clinical factors used in the FRAX tool. Our findings highlight the need for validation studies of FRAX in Spain. Key Words: Bone mineral density; cohort studies; FRAX; major osteoporotic fracture; osteoporosis.
at http://www.shef.ac.uk/FRAX/, to compute the 10-yr probability of hip fracture and major osteoporotic fracture (forearm, proximal humerus, clinical spine, or hip fracture). The risk factors used by FRAX to compute the 10-yr probability of fracture are age, sex, body mass index, history of previous osteoporotic fracture, parental history of hip fracture, current tobacco smoking, long-term oral glucocorticoid use, rheumatoid arthritis, other causes of secondary osteoporosis, alcohol consumption measured in units of alcohol (8e10 g of alcohol) per day, and femoral neck bone mineral density (BMD). This last factor is optional but it enhances the performance of the tool.
Introduction The FRAX tool was developed by the World Health Organization (WHO) Collaboration Center for Metabolic Bone Diseases to assess the risk of fracture in men and women aged 40e90 yr (1). The algorithm of the tool has not been published. However, interested parties can use a web application, available Received 03/23/12; Revised 05/15/12; Accepted 05/17/12. *Address correspondence to: Cristian Tebe Cordomı, Msc, Agencia d’Informacio, Avaluacio i Qualitat en Salut, Carrer Roc Boronat 81-95, 2a planta, 08005 Barcelona, Spain. E-mail: ctebe@ aatrm.catsalut.cat
231
232 The FRAX model was developed using 9 population-based cohorts from North America, Asia, Australia, and Europe (including Spain). Till date, adjusted models are available for the following countries: Argentina, Australia, Austria, Belgium, Canada, China, Colombia, Czech Republic, Denmark, Finland, France, Germany, Hong Kong, Hungary, Italy, Japan, Jordan, Lebanon, Malta, Mexico, the Netherlands, New Zealand, Philippines, Poland, Romania, Singapore, South Korea, Spain, Sweden, Switzerland, Taiwan, Tunisia, Turkey, the United Kingdom, and the United States. The US and Singapore models contain ethnic-specific data. The Spanish cohort, which included just more than 200 patients, was from Cantabria and formed part of the European Prospective Osteoporosis Study (2), a prospective follow-up study of the European Vertebral Osteoporosis Study involving patients from Madrid, Oviedo, Las Palmas, and Barcelona (3). FRAX showed high predictive power when validated in independent cohorts from many countries (4), but none of these included Spanish patients. The Spanish FRAX model was calibrated with mortality data from Spain and hip fracture incidence studies conducted in Barcelona, Seville, Madrid, the Canary Islands, Cantabria, and Zamora. The representativeness of the Spanish cohorts has been questioned, mainly owing to the small number of individuals included (5) and the great variability in fracture incidence among the different autonomous regions in Spain (6). The Spanish FRAX model is yet to be validated in an independent cohort. It is known that between-country differences in the incidence of major osteoporotic fractures and mortality contribute to heterogeneity in fracture probability (7,8). Thereafter, FRAX should be validated in local cohorts before clinical use in all countries. Despite FRAX has shown a high predictive power when assessed in independent cohorts (4), local validations may contribute to adjust the model. In USA, for example, validations led to a recalculation of the weight factors in the hip and the major osteoporotic fracture model (9). At present, there are 4 ongoing population-based studies ECOSAP (Ecografıa Osea en Atenci on PrimariadBone Ecography in Primary Care) (10), FRIDEX (Factores de RIesgo de osteoporosis y DEnsitometrıa Osea por absorciometrıa de rayos XdOsteoporosis Risk Factors and Bone Densitometry by X-ray Absorptiometry) (5), FRODOS (Fractures Osteopor otiques de les Dones d’OsonadOsteoporotic Fractures from Osona Women) (11), and ESOSVAL-R (Estudio Osteoporosis Valenciad Valencia Osteoporosis Study) (12), which are expected to contribute to the validation of the FRAX in Spanish population. A recent publication (13) from ECOSAP cohort using a 3-yr prospective has shown that FRAX underestimates osteoporotic fractures. However, data from vertebral fractures were not included and information on BMD was not available. To our knowledge, no other studies have been published validating the Spanish FRAX. The aim of our study was to assess the predictive ability of the Spanish FRAX for major osteoporotic fracture (forearm, proximal humerus, clinical spine, or hip fracture) in women with basal BMD measurements and 10-yr follow-up.
Journal of Clinical Densitometry: Assessment of Skeletal Health
Tebe Cordomı et al.
Methods Study Design and Participants A random sample from the CETIR database (CDB) (14) was included. CETIR is a retrospective cohort of 49,735 women aged 40e90 yr with a first visit for a bone densitometry (BD) between January 1992 and February 2008. BD was performed at CETIR Medical Center in Barcelona by request of a general practitioner or specialist. An interview-led questionnaire was administered by trained technicians at first visits and subsequent follow-ups. Women in the sample who did not have at least 1 follow-up survey in 10 yr or an earlier report of a major osteoporotic fracture were contacted by telephone. Trained personnel performed telephone interviews with those women who agreed to participate.
BD and Risk Factors for Major Osteoporotic Fracture BD was performed by trained technicians according to the protocols. Several devices were used throughout the study, all from Lunar Corp., GE Healthcare Madison, WI (models: Advance, Expert and Prodigy). BD results were categorized using the T-score for femoral neck BMD into 3 categories according to the WHO recommendations: normal (T-score O 1), osteopenia (T-score 1 O 2.5), and osteoporosis (T-score 2.5). Questionnaires at each visit collected information on demographics, personal and family history of major osteoporotic fracture and/or osteoporosis, history of other comorbidities, gynecologic and obstetric history, and lifestyle, as well as a validated questionnaire on calcium intake (15). Only comorbidities related to BMD were considered in our study, namely, rheumatoid arthritis, hyperparathyroidism, diabetes mellitus, anorexia nervosa, hyperthyroidism, and secondary osteoporosis. Information on the use of drugs with potential effects on BMD, such as glucocorticoids, anticonvulsants, and diuretics was also collected. Gynecologic and obstetric data included menarche and menopause ages, number of previous gestations (only those of O6 mo), and breastfeeding for longer than 3 mo. Lifestyle questions were focused on weekly physical activity (sedentary !5 h/wk and active 5 h/wk), smoking status (current, former, or never smoker), and alcohol intake measured in units of alcohol per day (!3 and 3 units/d). Finally, height and weight measurements were measured for all participants.
Major Osteoporotic Fractures Location and cause of fractures were reported by participants and not confirmed by imaging studies in all cases. Only major fractures resulting from low-intensity trauma in the forearm, proximal humerus, clinical spine, or hip were included (16).
Statistical Analyses Descriptive statistics included means and standard deviations (SDs) for continuous variables, and frequencies and percentages
Volume 16, 2013
Predictive Ability of FRAX for Osteoporotic Fracture for categorical variables. Using FRAX, we calculated the 10-yr probability of major osteoporotic fracture for each woman. Predictive performance was assessed by examining measures of calibration and discrimination. Calibration measures how the expected and the observed number of fractures differ from each other. Expected fractures result from the sum of the 10-yr probability of fracture for each woman, whereas observed fractures are the total number of fractures observed in women. The sample was split into risk deciles to calculate the number of observed and expected fractures in each group and to test the equal distribution hypothesis using the HosmerLemeshow goodness of fit test. The null hypothesis was rejected when the p value was below 0.05. In addition, the ratio of observed-to-expected (O/E) fractures and their 95% confidence intervals (CIs) was calculated. Discrimination refers to the ability of the risk-prediction model to differentiate between patients who experience a fracture and those who do not. The area under the receiver operating characteristic curve (AUC) quantifies this measure. Values close to 0.5 suggest null capacity to discriminate, whereas those close to 1 suggest maximum capacity. We performed additional sensitivity analyses to replicate the calibration and discrimination analyses, stratifying by age (!55, 55 to !65, 65 to !75, and 75 yr), BMD results (normal, osteopenia, and osteoporosis), and previous fracture. All database management and statistical analyses were performed using the STATA 11.0 statistical software (Stata Corporation, College Station, TX).
Results A random sample of 2086 women was drawn from the CDB. Of these, 355 had at least 1 follow-up survey 10 yr after inclusion in the CDB or an earlier report of a major osteoporotic fracture and were included in the study cohort. The rest (1731 women) were contacted by telephone. The final cohort consisted of 1231 women, after the exclusion of 533 women who did not answer the telephone, 189 women with a wrong telephone number, 28 women who did not want to participate, and 105 women for other reasons (Fig. 1). Characteristics of women included were not different from those who were not included (data not shown), except from the fact that the included women were significantly younger ( p ! 0.05). The median follow-up time was 10.95 yr (interquartile range: 0.52), which is the equivalent of 12,015.71 woman-years of follow-up. Mean age was 56.8 yr (SD: 7.8) and 82% of women were younger than 65 yr. A family history of osteoporosis and/or major osteoporotic fracture was reported by 23.5% of the women, and 15% reported at least 1 previous osteoporotic fracture. The mean T-score in this population was 1.4 (SD: 1.1), and 16% of the women were classified in the osteoporosis group (T-score ! 2.5). A description of the major osteoporotic fracture risk factors by age group is shown in Table 1. A total of 222 women (18.1%) reported at least 1 major osteoporotic fracture during follow-up, with an incidence of 19 fractures per 1000 women-year. The most frequently reported fractures were wrist and spine (106 and 78 women, Journal of Clinical Densitometry: Assessment of Skeletal Health
233
Fig. 1. Women included in the study. respectively); humerus fracture was reported in 25 women and hip fracture in 13 women. The incidence of fracture by age was 14 cases (95% CI: 10e17) per 1000 women-year in women younger than 55 yr, 19 (95% CI: 15e23) per 1000 women-year in women aged 55e64 yr, 28 (95% CI: 21e36) per 1000 women-year in women aged 65e74 yr, and 67 (95% CI: 28e125) per 1000 women-year in women aged 75 yr or older. Table 2 shows the calibration and discrimination assessment statistics for total major osteoporotic fractures. For calibration, the number of observed fractures was 3.9 times higher than the expected number (95% CI: 3.4e4.5). Sensitivity analyses stratified by age, BMD, and previous fracture showed O/E ratios of more than 2 times, statistically significant ( p ! 0.05) in all groups. Additional analyses per deciles (Table 3) showed lower O/E ratios in the high-risk deciles, and the calibration chi-square test was statistically significant ( p ! 0.0001). For discrimination, the AUC for the FRAX major osteoporotic fracture was 61% (95% CI: 57e65%; Fig. 2). The AUC ranged from 47% in women aged 75 yr or older to 63% in the group with the lowest BMD (osteoporosis group) in stratified analyses by age, BMD, and previous fracture. FRAX was developed to estimate the risk of fracture in people not taking prescription drugs for osteoporosis. In women interviewed by telephone, we recorded information on osteoporosis treatment during the last 10 yr; 436 (35.4%) reported having received osteoporosis treatment (78% bisphosphonates). The rate of fracture was significantly higher in women without treatment than in women with treatment (23.4% vs 8.3%, p ! 0.001). Moreover, the O/E ratio was also higher for those who had not received any drugs Volume 16, 2013
234
Tebe Cordomı et al. Table 1 Basal Characteristics of Women Included in the Study !55 yr (n 5 542)
Basal characteristics Body mass index (kg/m2) Menarche (yr) Menopause (yr) Number of gestations of O6 mo Calcium intake (mg/d) T-score
Mean
SD
55 to !65 yr (n 5 463) Mean
SD
65 to !75 yr (n 5 215) Mean
SD
75 yr (n 5 11) Mean
SD
26.2 12.8 45.3 2.2
4.0 1.5 5.1 1.4
26.8 13.2 47.2 2.4
3.7 1.7 5.8 1.5
27.1 13.3 47.6 2.1
3.7 1.7 5.9 1.6
27.0 13.6 47.2 3.0
4.1 1.9 7.4 1.4
825.4 1.0
309.1 1.0
804.4 1.6
276.2 1.0
765.0 2.2
239.0 0.8
757.3 2.6
334.2 0.6
n
%
n
%
n
%
n
%
Previous fracture Family history of osteoporosis/fracturea Body mass index !20 kg/m2 Rheumatoid arthritis Secondary osteoporosis Nulliparous Breastfeeding 3 mo Glucocorticoids Calcium intake !500 mg/d Sedentary Current smoker
49 134
9.0 24.7
77 114
16.6 24.6
56 41
26.1 19.1
3 0
27.3 0.0
16 3 8 47 301 14 35 374 58
3.0 0.6 1.5 8.7 55.5 2.6 6.6 69.0 10.7
9 4 14 51 292 18 48 341 19
1.9 0.9 3.0 11.2 63.1 3.9 10.5 73.7 4.1
4 2 7 34 138 9 19 167 4
1.9 0.9 3.3 15.6 64.2 4.2 9.0 77.7 1.9
1 0 0 1 10 0 3 9 0
9.1 0.0 0.0 9.1 90.9 0.0 27.3 81.8 0.0
Femoral neck BMD result Osteopenia Osteoporosis
254 29
46.9 5.4
271 73
58.5 15.8
105 90
48.8 41.9
4 7
36.4 63.6
Abbr: BMD, bone mineral density; SD, standard deviation. a Family history of osteoporosis and/or major osteoporotic fracture.
for osteoporosis in the last 10 yr (O/E: 5.1; 95% CI: 4.4e5.8) than for those who had received drugs (O/E: 1.8; 95% CI: 1.3e2.5).
Discussion FRAX underestimates the risk of major osteoporotic fracture in our cohort. This underestimation was smaller in women aged older than 65 yr and in women with osteoporosis confirmed by BD or with a previous fracture. To our knowledge, no validation studies on the predictive capacity of FRAX have been published for the Spanish population. Final results from the 4 ongoing studies aiming at validating this model are not expected for another 2 yr. These studies ECOSAP (10), FRIDEX (5), ESOSVAL-R (12), and FRODOS (11) analyze population in hospital and primary care settings. A recent publication with 3-yr follow-up results from the ECOSAP study (13) showed an underestimation of the risk of main osteoporotic fractures (forearm, proximal humerus, hip, and not clinical spine fracture), highlighting the fact that the linearity of the tool is limited. Preliminary Journal of Clinical Densitometry: Assessment of Skeletal Health
findings from the FRIDEX study, in turn, indicated a potential overestimation of predicted major fractures (17). Validation studies in other countries have shown that the local versions of the model have limitations with regard to its predictive performance. One of these studies, conducted by Ensrud et al (18), was based on a prospective cohort of 6252 women aged older than 65 yr from 4 regions in the USA and a mean follow-up of almost 7 yr. The study compared the predictive capacity of FRAX with models based on femoral neck BMD and age only or on femoral neck BMD and fracture history only. Results suggested that FRAX did not offer a superior predictive performance compared with other models. These results are consistent with reports from the development and validation study of FRAX (1), which showed that the predictive performance of FRAX for hip fracture was slightly better than that of a model based on BMD alone; and that the discrimination capacity was very similar when both models were compared. Another prospective study, performed by Sandhu et al (19) with a sample of 144 women and 56 men from Australia and a mean follow-up of almost 4 yr, compared the predictive Volume 16, 2013
Predictive Ability of FRAX for Osteoporotic Fracture
235
Table 2 Calibration and Discrimination Statistics for the FRAX Major Osteoporotic Fracture Model Observed Characteristics
Expected
Calibration
Discrimination
N
Fx
No Fx
Fx
No Fx
O/E
1231
222
1009
56.84
1174.16
3.99
3.42
4.46
0.61
0.57
0.65
Age (yr) 40 to !55 55 to !65 65 to !75 75
543 463 215 11
74 87 56 6
469 376 159 5
13.48 21.31 20.48 1.57
529.52 441.69 194.52 9.43
5.49 4.08 2.73 3.18
4.37 3.31 2.10 1.32
6.89 5.04 3.55 7.63
59.69 58.19 50.79 46.67
52.57 51.12 41.41 6.82
66.82 65.25 60.17 86.51
BMDa Normal Osteopenia Osteoporosis
398 634 199
47 125 51
351 509 148
8.37 26.81 21.66
389.63 607.19 177.34
5.61 4.63 2.35
4.22 3.88 1.79
7.47 5.51 3.10
53.66 57.12 62.87
44.81 51.68 53.84
62.51 62.56 71.90
Basal fracture No Yes
1046 185
174 49
872 136
38.77 18.07
1007.23 166.93
4.46 2.71
3.84 2.05
5.18 3.59
59.31 61.59
54.71 52.27
63.91 70.91
All patients
95% CI
AUC
95% CI
Abbr: AUC, area under the receiver operating characteristic curve; BMD, bone mineral density; CI, confidence interval; Fx, fracture; O/E, observed/expected. a Femoral neck bone mineral density.
ability of the UK FRAX, the USA FRAX, and the Garvan model. Results showed that the FRAX model performed well in women but had poor discriminatory performance in men. According to the authors, the differences might have been owing to the small number of men compared with women in the FRAX development studies. Sex-related differences might, therefore, also exist in countries with different regional fracture incidences. Because hip fracture incidence varies throughout Spain (6), hip fracture risk could be overestimated or underestimated in certain regions. Data in our initial cohort did not discriminate between a family history of fracture and a family history of osteoporosis. Therefore, we cannot exclude a recall bias. The current
classification system used in osteoporosis diagnoses was first introduced 20 yr ago, implying that a family history of osteoporosis might have been underreported by the older women in the study. Nonetheless, all the women interviewed later by telephone were asked about a family history of fracture, excluding those caused by high-impact trauma. Twenty-four percent of women who reported a family history of fracture or osteoporosis in the basal visit did not report a family history of fracture in the second interview by telephone. The percentage of women with a family history of fracture thus decreased from 23.5% to 17.8%, suggesting a lack of
Table 3 Distribution of the FRAX Estimated and Observed Risk of Incident Fractures by Deciles of Risk Risk deciles
Observed (%)
FRAX (%)
1 2 3 4 5 6 7 8 9 10
0.94 1.31 1.12 1.46 1.83 1.64 2.87 1.65 2.09 4.00
0.12 0.16 0.18 0.23 0.28 0.35 0.46 0.57 0.80 1.82
Journal of Clinical Densitometry: Assessment of Skeletal Health
Fig. 2. Area under the receiver operating characteristic curve for the FRAX major osteoporotic fracture model. ROC curve, receiver operating characteristic curve. Volume 16, 2013
236 specificity and a probable overestimation of the incidence. In the validation study, we observed an underestimation of this risk, which might also have been underestimated owing to the recall bias. Major osteoporotic fractures were reported by the study participants but were not always confirmed by radiology. Studies assessing the reliability of self-reports of fractures vs that of diagnoses confirmed by imaging studies have concluded that reported fractures provide a good estimate of fractures of the hip and the forearm/wrist, but not of the spine (20e22). Nevertheless, the estimated prevalence of hip and forearm/wrist fractures based on self-reports is approx 10% lower than the prevalence based on image test diagnoses. A potential underestimation of previous fractures would have increased the O/E ratio. These would be consistent with the underestimation we observed in the risk of major osteoporotic fractures. Limitations of the CETIR cohort have been described elsewhere (23). Its main weaknesses are a lack of external validity (the women included were referred for BD by a physician); recall bias and misclassification; and a potential competitive risk owing to the high morbidity and mortality associated with hip fracture (24). Nonresponse errors can increase the sampling error, by decreasing the sample size, and introduce bias in the results. In our case, women included in the analysis were significantly younger from those who were not included. Such nonresponse bias may have decreased the number of fractures in older women. Notwithstanding these limitations, the study has important strengths. First, the population included in the study was clinically relevant and not involved in the development of the FRAX model and second, the follow-up time was 10 yr. FRAX is a simple and useful fracture risk evaluation tool, and despite its lack of validation, it is used in the Spanish population. Examples of this include the assessment of differences in fracture risk between patients with and without osteoporosis treatment (25), the measurement of the absolute risk of fracture in women living in rural areas (26), and the comparison of the application of several clinical guidelines in a sample of patients with osteoporotic fractures according to the FRAX models (27). All these studies not only highlighted the advantages of using the model but also pointed out the lack of validation studies in the Spanish population, intimating both the authors and readers to exert caution in interpreting results (28). In conclusion, the present study shows that the predictive performance of FRAX in the Spanish population varies according to fracture type and risk factors. The model underestimates the risk of major osteoporotic fracture, especially in women with a low risk of fracture according to the clinical factors used in the FRAX tool. Our findings highlight the need for validation studies of FRAX in the Spanish population.
Acknowledgments The authors thank M. Millaret and M. Garcia for their help with documentation, E. Bonel and MJ. Garcia from CETIR Journal of Clinical Densitometry: Assessment of Skeletal Health
Tebe Cordomı et al. Center Medic for their fieldwork, and N. Paladio for an enthusiastic review of the final draft of the manuscript. This study was funded by the Plan de Calidad para el Sistema Nacional de Salud in collaboration with the Instituto Carlos III and the Agencia d’Informacio, Avaluacio i Qualitat en Salut. The first and the last 3 authors belong to the RAR research group (research group in health service and outcomes assessment), recognized by the Catalan Government (2005SGR00171).
References 1. Kanis JA, Johnell O, Oden A, et al. 2008 FRAX and the assessment of fracture probability in men and women from the UK. Osteoporos Int 19(4):385e397. 2. Reeve J, Silman A. EPOS Study Group. 1997 Epidemiology of osteoporotic fractures in Europe: towards biologic mechanisms. The European Prospective Osteoporosis Study Group. Osteoporos Int 7(3):78e83. 3. O’Neill TW, Marsden D, Matthis C, et al. 1995 Survey response rates: national and regional differences in a European multicentre study of vertebral osteoporosis. J Epidemiol Community Health 49(1):87e93. 4. Kanis JA, Oden A, Johnell O, et al. 2007 The use of clinical risk factors enhances the performance of BMD in the prediction of hip and osteoporotic fractures in men and women. Osteoporos Int 18(8):1033e1046. 5. Azagra R, Roca G, Encabo G, et al. 2011 Prediction of absolute risk of fragility fracture at 10 years in a Spanish population: validation of the WHO FRAX tool in Spain. BMC Musculoskelet Disord 12:30. 6. Alvarez-Nebreda ML, Jimenez AB, Rodrıguez P, Serra JA. 2008 Epidemiology of hip fracture in the elderly in Spain. Bone 42(2): 278e285. 7. Elffors I, Allander E, Kanis JA, et al. 1994 The variable incidence of hip fracture in southern Europe: the MEDOS Study. Osteoporos Int 4(5):253e263. 8. Cummings SR, Melton LJ. 2002 Epidemiology and outcomes of osteoporotic fractures. Lancet 359(9319):1761e1767. 9. Ettinger B, Black DM, wson-Hughes B, et al. 2010 Updated fracture incidence rates for the US version of FRAX. Osteoporos Int 21(1):25e33. 10. Gonzalez-Macıas J, Marın F, Vila J, et al. 2004 Prevalencia de factores de riesgo de osteoporosis y fracturas osteopor oticas en una serie de 5.195 mujeres mayores de 65 a~ nos. [Risk factors for osteoporosis and osteoporotic fractures in a series of 5,195 women older than 65 years]. [Article in Spanish]. Med Clin (Barc) 123(3):85e89. 11. Kanterewicz E, Puigoriol E, Peris P, et al. 2009 Relaci on del telopeptido carboxiterminal del colageno tipo I serico con la densidad mineral osea y el consumo de farmacos en mujeres posmenopausicas. Datos preliminares del estudio FRODOS. [Relationship between C-telopeptides of type I collagen serum values with bone mineral density and antiosteoporotic drug intake in postmenopausal women. Preliminary data from the FRODOS study]. [Article in Spanish]. Med Clin (Barc) 133(16):609e614. 12. Sanfelix-Genoves J, Peir o S, Sanfelix-Gimeno G, et al. 2010 Development and validation of a population-based prediction scale for osteoporotic fracture in the region of Valencia, Spain: the ESOSVAL-R study. BMC Public Health 10:153. 13. Gonzalez-Macıas J, Marin F, Vila J, Dıez-Perez A. 2011 Probability of fractures predicted by FRAX and observed incidence in the Spanish ECOSAP study cohort. Bone 50:373e377. 14. Tebe C, Casas L, Del Rio L, et al. 2011 Factores de riesgo de fracturas por fragilidad en una cohorte de mujeres espa~ nolas. Volume 16, 2013
Predictive Ability of FRAX for Osteoporotic Fracture
237
[Risk factors for fragility fractures in a cohort of Spanish women.] [Article in Spanish]. Gac Sanit 25(6):507e512. Musgrave KO, Giambalvo L, Leclerc HL, et al. 1989 Validation of a quantitative food frequency questionnaire for rapid assessment of dietary calcium intake. J Am Diet Assoc 89(10): 1484e1488. Kanis JA, Oden A, Johnell O, et al. 2001 The burden of osteoporotic fractures: a method for setting intervention thresholds. Osteoporos Int 12(5):417e427. Azagra R, Roca G, Zwart M, Encabo G. 2011. Diferencias en los valores predictivos de la herramienta FRAX(tm) entre poblaci on espa~nola y del Reino Unido y consideraciones sobre el umbral de intervencion. [Differences in the predictive values of the FRAX tool between the Spanish and United Kingdom population and considerations about the intervention threshold]. [Article in Spanish]. Med Clin (Barc) 137(15):713e714. Ensrud KE, Lui LY, Taylor BC, et al. 2009 A comparison of prediction models for fractures in older women: is more better? Arch Intern Med 169(22):2087e2094. Sandhu SK, Nguyen ND, Center JR, et al. 2010 Prognosis of fracture: evaluation of predictive accuracy of the FRAX algorithm and Garvan nomogram. Osteoporos Int 21(5):863e871. Ismail AA, O’Neill TW, Cockerill W, et al. 2000 Validity of selfreport of fractures: results from a prospective study in men and women across Europe. EPOS Study Group. European Prospective Osteoporosis Study Group. Osteoporos Int 11(3):248e254. Honkanen K, Honkanen R, Heikkinen L, et al. 1999 Validity of self-reports of fractures in perimenopausal women. Am J Epidemiol 150(5):511e516. Chen Z, Kooperberg C, Pettinger M, et al. 2004 Validity of selfreport for fractures among a multiethnic cohort of postmenopausal women: results from the women’s health initiative observational study and clinical trials. Menopause 11(3):264e274.
23. Tebe C, Espallargues M, Estrada M, et al. 2010 Analisis de los factores de riesgo y la probabilidad de fractura por fragilidad en una cohorte de mujeres con indicaci on de densitometrıa osea. [Risk factor analysis and probability of fragility fracture in a cohort of women with bone densitometry indication]. Plan de Calidad para el Sistema Nacional de Salud del Ministerio de Sanidad y Polıtica Social. Ministerio de Ciencia e Innovaci on. Agencia d’Informaci o, Avaluaci o i Qualitat en Salut de Catalu~ na. Informes de Evaluaci on de Tecnologıas Sanitarias, AATRM n um. 2009/04. 24. Bentler SE, Liu L, Obrizan M, et al. 2009 The aftermath of hip fracture: discharge placement, functional status change, and mortality. Am J Epidemiol 170(10):1290e1299. 25. Roig Vilaseca D, G omez Vaquero C, Hoces Otero C, Nolla JM. 2010 Riesgo de fractura calculado con el ındice FRAX en pacientes con osteoporosis tratados y no tratados. [Fracture risk calculated with the FRAX index in patients with osteoporosis treated and untreated]. [Article in Spanish]. Med Clin (Barc) 134:189e193. 26. G omez-Navarro R. 2010 Aplicaci on de la herramienta FRAX(R) para la determinaci on del riesgo de fractura en mujeres de un ambito rural. [Use of the FRAX algorithm to calculate the fracture risk in women of a rural area]. [Article in Spanish]. Rev Esp Salud Publica 84(3):321e330. 27. Naranjo A, Ojeda-Bruno S, Francisco-Hernandez F, et al. 2011 Aplicaci on de las guıas clınicas de prevenci on secundaria de fractura osteopor otica y del ındice FRAX en una cohorte de pacientes con fractura por fragilidad. [Application of guidelines for secondary prevention of fracture and the FRAX index in patients with fragility fracture]. [Article in Spanish]. Med Clin (Barc) 136:290e292. 28. Dıez-Perez A. 2010 El debate sobre la escala FRAX. [The FRAX tool debate]. [Article in Spanish]. Rev Osteoporos Metab Miner 1:5e6.
15.
16. 17.
18. 19. 20.
21. 22.
Journal of Clinical Densitometry: Assessment of Skeletal Health
Volume 16, 2013