ORIGINAL
ARTICLE
E n d o c r i n e
C a r e
The World Health Organization Fracture Risk Assessment Tool (FRAX) Underestimates Incident and Recurrent Fractures in Consecutive Patients With Fragility Fractures Sophie Roux, François Cabana, Nathalie Carrier, Michèle Beaulieu, Pierre-Marc April, Marie-Claude Beaulieu, and Gilles Boire Division of Rheumatology (S.R., N.C., P.-M.A., G.B.), Department of Medicine, Division of Orthopedic Surgery (F.C.), and Department of Family Medicine (M.-C.B.), University of Sherbrooke, Sherbrooke, Québec, Canada J1H 5N4; and Office of Medical Liaison (M.B.), Merck Canada Inc, Montréal, Québec H9H 4M7
Context: The World Health Organization Fracture Risk Assessment tool (FRAX) was developed to identify patients at risk of sustaining a fragility fracture (FF). Objective: The objective of the study was to evaluate estimated FRAX probabilities of FF at the time of a FF and to compare them with the observed incidence of recurrent FF. Methods: A prospective cohort included men and women older than 50 years at the time of a FF. FRAX scores without bone mineral density [FRAX-body mass index (BMI)] were calculated prior to and after the inclusion FF. Recurrent FFs were recorded over a 4-year follow-up. Determinants associated with recurrent FF were determined by univariate and multivariate analyses. Results: FRAX-BMI scores were available in 1399 of the 1409 recruited patients. A high-risk FRAXBMI score was present in only 42.7% patients before and 56.4% after the incident FF. Most FF patients at low or moderate risk before their initial FF were men, younger than 65 years, or without previous FF. Over a median follow-up of 3 years, recurrent FF occurred in 108 patients (2.69 per 100 patient-years). The overall sensitivity of post-FF FRAX to predict a recurrent FF was 71.3% and was specifically lower in patients younger than 65 years (13%) and without previous FF (63%) at inclusion. Conclusions: The FRAX-BMI scores were below the Canadian threshold for treatment in more than half the patients at the time of a FF and in close to a third of patients with recurrent FF. FRAX-BMI severely underestimates the FF risk in patients younger than 65 years old and after a single FF. (J Clin Endocrinol Metab 99: 0000 – 0000, 2014)
ver their lifetime, up to half of women and a quarter of men experience at least one fracture due to bone fragility (1). A fragility fracture (FF) occurs spontaneously or with minimal trauma, usually as a result of a fall from standing height or less (2). According to the World Health Organization classification for bone mineral density (BMD), most FFs occur in individuals who do not have
O
osteoporosis (3, 4). To overcome the limitations of BMD measurements, several multifactorial tools have been developed to accurately identify the patients at high-risk of FF; among these, the World Health Organization Fracture Risk Assessment tool (FRAX) is the most widely used. Other tools with comparable predictive properties include the Canadian Association of Radiologists and Osteopo-
ISSN Print 0021-972X ISSN Online 1945-7197 Printed in U.S.A. Copyright © 2014 by the Endocrine Society Received December 23, 2013. Accepted March 26, 2014.
Abbreviations: BMD, bone mineral density; BMI, body mass index; CAROC, Canadian Association of Radiologists and Osteoporosis Canada; CHUS, Centre Hospitalier Universitaire de Sherbrooke; FF, fragility fracture; FRAX, Fracture Risk Assessment tool; IQR, interquartile range; OPTIMUS, Osteoporosis and Peripheral fractures: Treatment and Investigation in Multidisciplinary Care at the CHUS; PCP, primary care physician; RR, relative risk.
doi: 10.1210/jc.2013-4507
J Clin Endocrinol Metab
jcem.endojournals.org
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
1
2
Roux et al
Accuracy of FRAX in Predicting Fractures
rosis Canada (CAROC) assessment tool, which includes age, femoral neck BMD, corticoid use, and previous FF (5), and the Garvan Fracture Risk Calculator, which also includes fall history (6). FRAX is an algorithm that estimates the 10-year absolute risk of either hip FF or major FF (vertebra, forearm, hip, or proximal humerus) in men and women above the age of 50 years (7). The FRAX score integrates clinical risk factors, alone or in combination with BMD, and has been adapted in several countries, including Canada (8). Evaluating the 10-year fracture risk is currently the recommended approach to assist physicians in treating osteoporosis. Although FRAX is widely promoted for use in clinical practice, it has several limitations (9, 10). In the Global Longitudinal Study of Osteoporosis in Women (GLOW) cohort, neither FRAX nor the Garvan Fracture Risk Calculator was better than the simplest model, which took only age and fracture history into account (11). The cutoffs to identify high-risk patients and the guidelines for therapeutic intervention have been produced by expert committees and vary among countries. Although in some guidelines, patients sustaining a major FF (hip or vertebra) or those with a very low BMD are eligible for pharmacological interventions, regardless of their FRAX score (12, 13), in others the 10-year probabilities of FF alone qualify patients for antiosteoporotic treatment (14). A FF occurring after the age of 50 years is one of the most reliable predictors of subsequent FF. The risk for recurrent FF is maximal in the first few years after the event, gradually diminishing afterward, back to the pre-FF level at 10 years (3, 15). We sought to evaluate the accuracy of Canadian-adjusted FRAX tool in predicting FFs in a Canadian longitudinal prospective cohort of consecutive FF patients, whose 1-year data were recently reported (16). We used the baseline and follow-up data to determine patients’ estimated 10-year fracture risk, either immediately before their initial FF, or considering their recent FF. Because BMD data were not available for a majority of patients, we estimated their 10-year fracture risk using FRAX probabilities without BMD [FRAX-body mass index (BMI)]. We next compared the post-FF FRAXBMI risk with the observed rate of recurrent FF over a 4-year follow-up period. Our findings indicate that the FRAX-BMI scores do not reach the Canadian threshold for pharmacological treatment in more than half of incident FF patients and in close to a third of patients with recurrent FF.
Materials and Methods Study subjects The program Osteoporosis and Peripheral fractures: Treatment and Investigation in Multidisciplinary Care at the CHUS
J Clin Endocrinol Metab
(OPTIMUS) is an ongoing research initiative that combines identification of an initial FF with two incremental and recursive strategies and engages the primary care physicians (PCPs) to manage their own patients. We have recently reported the impact of these strategies on osteoporosis treatment at one year after the FF (16). In this program, patients older than 50 years with X-rayconfirmed fractures (excluding simple avulsion) and treated by orthopedic surgeons at the Centre Hospitalier Universitaire de Sherbrooke (CHUS) were screened by study coordinators. FFs were identified using the Canadian Multicenter Osteoporosis Study questionnaire (17). Patients with multiple FFs during a single event were counted only once. Major fractures (in decreasing order: hip, vertebra, proximal humerus, wrist) prevailed over fractures at other sites (defined as minor fractures). Multiple recurrent FFs occurring during a single or multiple events were defined similarly. Inclusion and exclusion criteria for the OPTIMUS program were described previously (16). Patients were randomly assigned to standard care (no intervention), minimal, or intensive interventions, as described earlier (16). Inpatients with hip FFs were also included but were assessed and treated by rheumatologists and were not assigned to any intervention. Baseline characteristics include demographics (age, gender, weight, height, race, and menopausal status), previous FF or diagnosis of osteoporosis, BMD, comorbidities, currently used medication including antiosteoporosis drugs, estimated calcium and vitamin D intake, tobacco and alcohol consumption, extent of physical activity, and familial history of osteoporosis. Causes of secondary osteoporosis were recorded, including type 1 diabetes, osteogenesis imperfecta, untreated long-standing hyperthyroidism, hypogonadism or premature menopause (⬍45 y), malabsorption and chronic liver disease. All the patients provided consent to contact their PCP and pharmacist. After enrollment, follow-ups were done by phone calls at 4, 8, 12, 24, 36, and 48 months in the intensive intervention and hip FF groups and at 6, 12, 24, 36, and 48 months in the standard care and minimal intervention groups. During phone calls, each patient was asked the following: appointment with their PCP, BMD testing, changes in diet or exercise, additional comorbidities, initiation and adherence with osteoporosis treatment, and details on additional fractures. For the diagnosis of recurrent vertebral fracture, the patient should have had spine radiography and should have been diagnosed by a physician. The Ethics Review Board of the CHUS approved the study.
FRAX calculation The Canadian FRAX tool (http://www.shef.ac.uk/FRAX/ tool.jsp) became available years after the initiation of OPTIMUS and is still infrequently used in our area in which the CAROC tool is preferred. Therefore, FRAX-calculated 10-year probabilities of major FF and hip FF were obtained retrospectively from the baseline data. The FRAX scores were calculated without BMD (FRAX-BMI) for all patients, and with BMD (FRAXBMD) when available (n ⫽ 302). Self-reported ankle fractures prior to the inclusion FF were excluded because qualification for FF was often dubious, and confirmatory X-rays infrequently available, but were considered during the follow-up if they fulfilled the criteria for FF (excluding simple avulsion). The FRAX scores were assigned based on Canadian guidelines (18): high risk (ⱖ20% for major FFs or ⱖ 3% for hip FFs); moderate risk (10%–20% for major FFs and ⬍ 3% for hip FFs); and low risk (⬍10% for major FFs and ⬍ 3% for hip FFs).
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
Statistical analysis Continuous FRAX scores and categorical risks for FF were compared with the baseline demographics and site of FF. Because continuous scores were not normally distributed, the median [and interquartile range (IQR)] was presented, and the scores were compared with Wilcoxon rank-sum test or with KruskalWallis test. Categorical FRAX scores were compared with the 2 test or Fisher exact test if the frequencies were less than 5. The rates of recurrent FF per 100 person-years according to FRAX categories were compared with the 2 test or Fisher exact test. Relative risks (RRs) were estimated with log-binomial univariate regression, and the sensitivity was calculated. A multivariate logbinomial regression was computed to evaluate the variables that predict recurrent FF, including age, gender, osteoporosis treatment, previous FF, minor/major FFs, and FRAX-BMI. Significance was set at P ⬍ .05 and analyses were computed with SAS 9.2 software for Windows (SAS Institute Inc).
jcem.endojournals.org
Table 1.
Results Patient demographics From January 2007 to May 2012, 1409 consecutive consenting patients with incident FFs were enrolled in the prospective cohort. These FFs represented over 80% of all FFs occurring in apt-to-consent patients evaluated by the CHUS orthopedic surgeons (16). Baseline characteristics are shown in Table 1. At inclusion, 104 patients (7.38%) had a secondary cause of osteoporosis (excluding alcohol, tobacco, and corticosteroids), with a significantly higher prevalence in older women (10.28% in women aged 65 y and older) (P ⬍ .001). At baseline, 387 patients (27.5%) were under effective pharmacological treatment for osteoporosis, mainly bisphosphonates; baseline adherence to previously prescribed treatments was not assessed. After 1 year, 775 patients were under effective osteoporosis medications (60.1%), with a mean drug possession of greater than 80% of the doses confirmed by the patients’ pharmacists. In addition, 278 (19.8%) had already suffered a previous FF at baseline, among which 128 (46%) were treated with an effective antiosteoporosis drug. Predicted 10-year absolute risk of fracture before and after the initial fracture We used the Canadian FRAX-BMI tool to determine the 10-year fracture risk. FRAX-BMI scores could not be calculated in 10 patients because of missing weight and/or height; thus, a total of 1399 patients was included in the analysis. The FRAX-BMI scores for major FFs and for hip FFs before and after the initial FF are shown in Table 2. The day before the incident FF, only 42.7% of the patients were at high risk. When the incident FF was imputed into the FRAX-BMI estimation, 56.4% were at high risk (P ⬍ .001, FRAX-BMI before vs after the incident FF). Lower proportions of high-risk scores (36.1% before and 49.5%
3
Baseline Characteristics
Characteristics
Value
Median age, y (IQR) Women, n, % Mean BMI ⫾ SD Tobacco, n, % Nonsmoker Current Smoker Previous smoker Alcohol, n, % Never/occasionally One drink or more by day Personal history of osteoporosis, n, % Family history of osteoporosis, n, % Secondary osteoporosis, n, % Previous FF, n, %a Hip Vertebra Proximal humerus Wrist Other Fragility fracture site at inclusion, n, % Hip Vertebra Proximal humerus Wrist Ankle Other Treatment for osteoporosis at inclusion, n, % Aminobisphosphonate Other effective treatmentb Corticosteroids
67 (59 –79) 1159 (82.3) 26.1 ⫾ 5.7 683 (48.5) 207 (14.7) 519 (36.8) 1283 (91.0) 126 (9.0) 386 (27.5) 304 (22.4) 104 (7.38) 278 (19.8) 43 (16.3) 29 (11.0) 29 (11.0) 100 (37.9) 63 (23.9) 262 (18.6) 21 (1.5) 251 (17.8) 475 (33.7) 265 (18.8) 135 (9.6) 387 (27.5) 330 (23.4) 52 (4.3) 29 (2.6)
a
Self-reported prior ankle fractures excluded. Previous fracture status was unknown for four patients. b Other effective treatments included hormonal replacement therapy, selective estrogen receptor modulators, denosumab, and teriparatide.
after the FF) were observed in initially untreated patients. After the FF, 336 patients (24%) remained in the low-risk category and 274 (19.6%) in the moderate-risk category, indicating that only 192 patients (23.9%) moved from a lower strata to the high-risk category. In the subgroup of patients with BMD data (n ⫽ 302), absolute risk predictions for major FF using FRAX-BMI and FRAX-BMD were strongly correlated (r ⫽ 0.81 before and r ⫽ 0.71 after the baseline FF), as for hip FF (r ⫽ 0.72 before and r ⫽ 0.63 after the baseline FF) (all significant at P ⬍ .0001). Among the 260 hip FF patients, 75.8% were classified as high risk (hip score ⱖ 3%) by FRAX-BMI. Among the patients with major FFs (n ⫽ 999), only 50% were estimated at high risk of major FFs, and 43.7% were at low risk. Most patients with nonmajor FFs (n ⫽ 400) were not considered to be at high risk for major or hip FF. Imputing the initial FF in the FRAX-BMI estimates increased the proportion of patients in the high-risk category for FFs; however, more than one third of patients with major FFs and 60% with minor FFs were still evaluated at low or moderate risk (Table 2).
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
4
Roux et al
Table 2.
Accuracy of FRAX in Predicting Fractures
J Clin Endocrinol Metab
Estimated 10-Year Fracture Risk of Patients With Fragility Fractures Before and After the Inclusion FF Before the Incident Fracturea
Minor and major FF Any FF (n ⫽ 1399) Site of initial FF Hip (n ⫽ 260) Vertebra (n ⫽ 21) Proximal humerus (n ⫽ 247) Wrist (n ⫽ 471) Ankle (n ⫽ 265) Other (n ⫽ 135) Type of initial FF Minor (n ⫽ 400) Major (n ⫽ 999) Age, y ⬍65 (n ⫽ 602) ⱖ65 (n ⫽ 797) Gender Women (n ⫽ 1149) Men (n ⫽ 250) Previous FF No (n ⫽ 1120) Yes (n ⫽ 275) Treatment at inclusion No (n ⫽ 1016) Yes (n ⫽ 383) a
After the Incident Fracturea
FRAX-major med (IQR)
FRAX-hip med (IQR)
High-risk, n, %b
FRAX-major med (IQR)
FRAX-hip med (IQR)
High-risk, n, %b
9.4 (5.3–19)
2 (0.5–7.3)
597 (42.7)
16.0 (10.0 –28.0)
3.7 (1.5–10.0)
789 (56.4)
21.0 (10.0 –27.0)c 9.5 (5.0 –18.0) 11.0 (6.1–21.0) 8.3 (5.4 –17.0) 5.7 (4.0 –10.0) 7.1 (4.3–15.0)
8.3 (3.2–14.0)c 1.8 (0.6 –7.5) 2.6 (0.7– 8.2) 1.5 (0.5–5.6) 0.6 (0.3–2.2) 1.3 (0.3– 4.4)
197 (75.8)c 10 (47.6) 117 (47.4) 171 (36.3) 56 (21.1) 46 (34.1)
29.0 (17.0 –38.0)c 11.0 (9.5–29.0) 11.0 (8.1–18.0) 17.0 (11.0 –29.0) 15.0 (11.0 –25.0) 13.0 (8.5–21.0)
11.5 (5.2–18.5)c 3.6 (1.5–10.0) 1.7 (0.9 – 4.2) 4.0 (1.5–11.0) 3.0 (1.6 – 8.0) 2.9 (1.0 –7.1)
225 (86.5)c 13 (61.9) 146 (59.1) 246 (52.2) 92 (34.7) 67 (49.6)
6.6 (4.1–12)c 11 (6.2–22)
0.9 (0.3–3.1)c 2.9 (0.7–9.1)
103 (25.8)c 11.0 (8.2–18.0)c 497 (49.7) 18.0 (11.0 –32.0)
2.0 (1.0 –5.2)c 4.9 (1.8 –13.0)
159 (39.8)c 630 (63.1)
5.1 (3.8 –7.1)c 17.0 (10.0 –26.0)
0.5 (0.3– 0.8)c 21 (3.5)c 5.9 (2.8 –11.0) 579 (72.6)
1.3 (0.8 –2.0)c 9.0 (4.7–16.0)
83 (13.8)c 706 (88.6)
12.0 (6.1–22.0)c 5.3 (3.8 – 8.5)
2.4 (0.6 – 8.5)c 1.1 (0.4 –3.4)
7.6 (4.7–14.5)c 24.0 (14.0 –36.0)
1.3 (0.4 –5.1)c 386 (34.5)c 14.0 (9.6 –25.0)c 7.5 (3.0 –16.0) 211 (76.7) 24.0 (14.0 –36.0)
3 (1.3–9.0)c 7.5 (3.0 –16.0)
575 (51.3)c 211 (76.7)
7.8 (4.7–16.0)c 15.0 (8.6 –25.0)
1.3 (0.4 –5.6)c 367 (36.1)c 14.0 (9.3–25.0)c 4.4 (1.4 –10.0) 233 (60.8) 22.0 (14.0 –33.0)
2.8 (1.3– 8.6)c 6.8 (2.7–13.0)
503 (49.5)c 286 (74.7)
9.8 (7.7–13.0)c 25.0 (17.0 –35.0)
532 (46.3)c 18.0 (11.0 –31.0)c 4.2 (1.6 –12.0)c 68 (27.2) 9.9 (7.4 –14.0) 2.4 (1.0 –5.3)
679 (59.1)c 110 (44.0)
The 10-year risk of hip fracture or major fracture was calculated both before and after the FF. We used the Canadian BMI-based FRAX score.
b
High-risk score includes a major FRAX score of 20% or greater or a hip FRAX score of 3% or greater.
c
P ⬍ .001 comparison between parameters in each clinical factor category.
While analyzing the FRAX-BMI scores according to baseline characteristics, most patients estimated at low or moderate risk before their FF belonged to defined subcategories: many were men, and most were under 65 years of age (Figure 1), without history of FFs or not receiving antiosteoporotic treatment at baseline (Table 2). When the same analysis was restricted to hip or other major FFs, a similar pattern was observed (Table 3). These findings highlight that the Canadian FRAX-BMI tool does not accurately evaluate the 10-year fracture risk in specific subgroups, particularly among those patients younger than 65 years and in men. Recurrent fractures: sites and rates After 4013 patient-years of follow-up (median 3 y), 122 recurrent FFs occurred in 108 patients (Table 4); 12 patients had two recurrent FFs and one patient had three (only one recurrent FF was considered). Rates of recurrence were 2.69 per 100 patient-years, and the rates were higher after hip (3.96) and wrist FFs (2.96). In the analysis according to the FF site at inclusion, 21.3% of the patients who developed a recurrent FF at a major site had a minor FF at inclusion. Globally, the risk for a recurrent major FF was 57 per 1009 (5.84%) after an
incident major FF and 16 per 400 (4.0%) after an incident minor FF (Table 4). Recurrent fractures according to FRAX-BMI, FRAXBMD, and CAROC scores The FRAX-BMI high-risk category calculated after the incident FF was moderately useful to predict patient’s subsequent FF because only 70% of all recurrent FF patients were estimated at high risk. The hip FRAX-BMI score was more accurate in patients with a recurrent hip FF because 91.7% of them had a score of 3% or greater. However, only 50% of patients with a recurrent major FF had a FRAX-BMI score for major FF of 20% or greater (Figure 2). We could not evaluate the FRAX-BMI specificity in our follow-up study, as FRAX estimates FF risk over a 10-year period. However, because 30% of patients who developed a recurrent FF were not estimated at high risk of FFs, we were confident to analyze the sensitivity of the FRAX-BMI tool (Table 5). Although being at high risk significantly increased the RR of FFs, the sensitivity of the FRAX-BMI tool to detect recurrent FFs was about 70% in the whole cohort. The FRAX-BMD was even less sensitive (40%) to detect patients who would sustain a recurrent FF. In the
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
jcem.endojournals.org
5
high-risk CAROC category to detect recurrent FF was 55% (Table 5). In univariate analyses, a significant increase in RR for recurrent FFs was observed in older patients (ⱖ65 y). Similarly, patients with previous FFs at inclusion (at least two FFs before the recurrent FF) and those already receiving osteoporosis treatment at the time of their initial FF (more likely with a previous FF or a more severe disease) had a significantly higher risk of recurrent FFs. The risk of recurrent FFs was not affected by the site of initial FFs (major/minor). In multivariate analyses, a previous FF at baseline was the only significant factor affecting the risk of recurrent FF (Table 5). The FRAX-BMI scores were also compared within each clinical subgroup of patients, with low statistical power due to small numbers in each subcategory. Most of the patients with low or moderate risk when they developed a recurrent FF (n ⫽ 31) were younger than 65 years (87%), did not have a previous FF (84%), and were not treated for osteoporosis at baseline (68%); the sensitivity of estimated FRAX-BMI high risk to predict recurrent FF in these subgroups was low (13% in patients ⬍ 65 y old, and 63% after a single FF). Most of these 31 patients with an estimated low or moderate risk after inclusion had initially sustained major FF (77%) (Supplemental Table 1). Taken together, our findings suggest that the FRAX-BMI tool underestimates the probabilities of recurrent FF, particularly in younger patients (⬍65 y) and after a single major FF.
Figure 1. Continuous FRAX-BMI probabilities (calculated before and after the incident FF) for major FFs (A) and hip FFs (B) according to age (stratified by 5 y groups). Results are presented as box plots, and the line denotes the median value. Dotted lines represent the cutoffs for the high-risk score: 20% for major FFs and 3% for hip FFs.
Discussion
same subset of patients, we also calculated the CAROC score (5). All patients were at least at moderate risk for FFs by definition (at least one FF), and the sensitivity of the
To accurately predict each patient’s risk for a negative outcome is every clinician’s dream. A correct identifica-
Table 3. Estimated 10-Year Fracture Risk of Patients With Fragility Fractures Before and After the Inclusion FF. Analysis Restricted to Major Fractures at Inclusion Before the Incident Fracturea Major FFs only Any major FF (n ⫽ 999) Age, y ⬍65 (n ⫽ 366) ⱖ65 (n ⫽ 633) Gender Women (n ⫽ 864) Men (n ⫽ 135) Previous fracture No (n ⫽ 777) Yes (n ⫽ 219) Treatment at inclusion No (n ⫽ 711) Yes (n ⫽ 288) a
After the Incident Fracturea
FRAX-major med (IQR)
FRAX-hip med (IQR)
High risk, n (%)b
FRAX-major med (IQR)
FRAX-hip med (IQR)
High risk, n (%)b
11 (6.2–22)
2.9 (0.7–9.1)
497 (49.7)
18.0 (11.0 –32.0)
4.9 (1.8 –13.0)
630 (63.1)
5.6 (4.2–7.5)c 18.0 (11.0 –26.0)
0.5 (0.3–1.0)c 7.0 (3.2–12.0)
15 (4.1)c 482 (76.1)
11.0 (8.5–13.0)c 27.0 (18.0 –37.0)
1.5 (1.0 –2.2)c 10.0 (5.3–17.0)
59 (16.1)c 571 (90.2)
13.0 (6.8 –24.0)c 6.8 (4.5–10.0)
3.3 (0.8 –10.0)c 2.0 (0.6 – 4.4)
447 (51.7)c 50 (37.0)
20.0 (13.0 –33.5)c 11.0 (8.6 –15.0)
5.5 (2.0 –14.0)c 3.5 (1.5– 6.4)
553 (64.0)c 77 (57.0)
8.6 (5.4 –18.0)c 26.0 (15.0 –37.0)
1.8 (0.6 – 6.9)c 9.1 (3.4 –18.0)
319 (41.1)c 175 (79.9)
16.0 (11.0 –29.0)c 26.0 (15.0 –37.0)
4.1 (1.7–11.0)c 9.1 (3.4 –18.0)
452 (58.2)c 175 (79.9)
9.2 (5.6 –20.0)c 17.0 (9.3–26.0)
2.1 (0.6 –7.9)c 5.4 (1.6 –11.0)
312 (43.9)c 185 (64.2)
16.0 (11.0 –29.0)c 24.0 (15.0 –35.0)
4.0 (1.6 –11.0)c 7.8 (3.2–14.0)
409 (57.5)c 221 (76.7)
The 10-year risk of hip fracture or of major fracture was calculated both before and after the FF. We used the Canadian BMI-based FRAX score.
b
High-risk score includes a major FRAX score of 20% or greater or a hip FRAX score of 3% or greater.
c
P ⬍ .001 comparison between parameters in each clinical factor category.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
6
Roux et al
Table 4.
Accuracy of FRAX in Predicting Fractures
J Clin Endocrinol Metab
Rates of Recurrent FF According to Sites of Initial FF Site of Reccurent FF
Site of FF at Inclusion
Recurrent FF Cumulative FF Per 100 Proximal Inclusion Re-FF Re-FF, % Time, y Patient-Years Hip Vertebra Humerus Wrist Ankle Othera
Total Hip Vertebra Proximal humerus Wrist Ankle Other sites
1409 262 21 251 475 265 135
a
108b 28 1 14 39 11 15
7.67 10.69 4.76 5.58 8.21 4.15 11.11
4012.68 707.95 58.21 722.25 1317.95 776.5 429.83
2.69 3.96 1.72 1.94 2.96 1.42 3.49
27 11 0 4 7 1 4
14 6 1 2 3 1 1
14 1 0 2 7 2 2
20 2 0 4 9 2 3
7 1 0 1 3 1 1
26 7 0 1 10 4 4
Other sites include the following: upper limb (n ⫽ 2), lower limb (n ⫽ 5), clavicle (n ⫽ 1), pelvis (n ⫽ 2), ribs (n ⫽ 15), and coccyx (n ⫽ 1).
b
A total of 122 FFs occurred in 108 patients. Only the most significant of multiple recurrent FFs are reported here: hip greater than vertebra greater than proximal humerus greater than wrist greater than minor site.
tion of patients at high risk of FFs would allow early treatment and decrease the numbers needed to treat to prevent future injuries. BMD fails to deliver such precise estimates of risk because more than half of FFs occur in nonosteoporotic patients (3). The FRAX tool was developed to overcome the limitations of BMD by integrating several clinical risk factors, including age and past fracture, either alone or in combination with BMD (7). In population studies, fracture rates closely parallel FRAX predicted rates (8). Most guidelines thus build on FRAX to identify men and women in need for preventive treatment. Nonetheless, in the current study, using the Canadian adjusted FRAX-BMI score in consecutive patients included at the time of a FF, we found that just before their
Figure 2. Continuous FRAX-BMI probabilities (accounting for the incident FF) for major FFs (A) and hip FFs (B) according to the type of recurrent FF. Results are presented as dot plots, and the line denotes the median value. Dotted lines represent the cutoffs for the high-risk score: 20% for major FFs and 3% for hip FFs.
FF, more than half the patients were estimated at low or moderate risk for FF. The low sensitivity of FRAX assessment was not driven by inclusion of incident minor FF in our cohort because the same results were obtained when considering baseline major FF only. When the incident FF event was added to the FRAX-BMI calculation, a large proportion, even after major (and even hip) FFs, were still not estimated at high risk for FFs. The Canadian-adjusted FRAX-BMI high-risk category was therefore inadequate to identify most the patients before they actually sustained an incident major FF, and it remained insensitive (71%), even after the FF event. Moreover, over a 4-year follow-up period, one third of the patients who developed a recurrent FF were not estimated at high risk. The Canadian-adjusted FRAX-BMI high-risk category was similarly insensitive to identify patients with any recurrent FF, although FRAX particularly underestimates the impact of a recent major FF on subsequent fracture risk. Limiting analyses to initially untreated patients did not improve FRAX performance. Subgroup analyses further highlighted that the FRAX tool specifically underestimates the probabilities of FFs in patients below the age of 65 years, with a sensitivity of 71% for incident FFs and only 13% for recurrent FFs. The FRAX model does not account for a number of variables that impact the actual 10-year fracture risk: dose and duration of corticosteroids, levels of alcohol or tobacco consumption, activity or duration of predisposing diseases such as rheumatoid arthritis or endocrinopathies (diabetes, hypopituitarism, male hypogonadism), multiple drugs potentially inducing bone loss or increasing fracture risk, and risks associated with falls (10, 19). Some adjustments have been proposed since the FRAX release in 2008 (10). Several studies have suggested that FRAX may not be sufficiently sensitive toward some subgroups of patients. In early postmenopausal women, the FRAX predictive value was low and showed no improvement over
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
Table 5.
jcem.endojournals.org
7
Relative Risk of Recurrent FF According to FRAX and Clinical Variables Total, n
Re-FF n (%)
Re-FF Per 100 Patient-Years
Univariate RR (CI 95%)
Multivariate RR (CI 95%)
610
31 (5.1)
1.74
1
1
789
77 (9.8)
3.49
1.9 (1.3–2.9)b
1.3 (0.7–2.3)
621 778
32 (5.2) 76 (9.8)
1.76 3.50
1 1.9 (1.3–2.8)b
205
12 (5.9)
2.17
1
97
8 (8.2)
3.11
1.4 (0.6 –3.3)
40.0
207 95
12 (5.8) 8 (8.4)
2.17 3.14
1 1.5 (0.6 –3.4)
40.0
152 148
9 (5.9) 11 (7.4)
2.39 2.64
1 1.3 (0.5–2.9)
55.0
603 806
31 (5.1) 77 (9.6)
1.76 3.42
1 1.9 (1.2–2.8)b
1 1.4 (0.8 –2.5)
71.3
250 1159
12 (4.8) 96 (8.3)
1.73 2.89
1 1.7 (1–3.1)
1 1.4 (0.8 –2.5)
11.1
1127 278
70 (6.2) 37 (13.3)
2.21 4.42
1 2.1 (1.5–3.1)c
1 1.7 (1.1–2.5)b
34.6
1022 387
64 (6.3) 44 (11.4)
2.22 3.90
1 1.8 (1.3–2.6)b
1 1.4 (0.9 –2.1)
40.7
400 1009
26 (6.5) 82 (8.1)
2.16 2.92
1 1.3 (0.8 –1.9)
1 1 (0.7–1.6)
75.9
Sensitivity
a
FRAX-BMI (major and hip) Low-moderate (major ⬍ 20% and hip ⬍ 3%) High (major ⱖ 20% and/or hip ⱖ 3%) FRAX-BMI (hip)a Low (⬍3%) High (ⱖ3%) FRAX-BMD (major and hip)a Low-moderate (major ⬍ 20% and hip ⬍ 3%) High (major ⱖ 20% and/or hip ⱖ 3%) FRAX-BMD (hip)a Low (⬍3%) High (ⱖ3%) CAROCa Moderate High Age, y ⬍65 ⱖ65 Gender Men Women Previous FF (at inclusion) No Yes Osteoporosis treatment at inclusion No Yes Site of FF Minor Major a
71.3 70.4
FRAX and CAROC scores include the initial FF
b
P ⬍ .01 compared between each category.
c
P ⬍ .001 compared between each category.
hip BMD (20). In the French cohort Os des Femmes de Lyon (OFELY), greater than 50% of women with an incident major FF had a FRAX score less than 9%, and 90% had less than 24%. FRAX score severely underestimated the risk of major FF in women older than 65 years with low BMD (21). In a small study in Switzerland, half the patients showed low FRAX risk on the day before FF (22). Our observations in a larger cohort confirm the low sensitivity of the FRAX tool to predict initial FF and demonstrate for the first time that post-FF FRAX is also rather insensitive to predict recurrent FF. In the FRAX algorithm adapted for the Canadian population (8, 23), a high 10-year fracture risk is defined as 20% or more for major FFs and 3% or more for hip FFs. Treatment is recommended after hip or spine FFs, after two FF events at any sites, as well as with a high-risk FRAX score (18). Our re-FF study results are consistent with the recent report that initial minor FFs increase the risk of recurrent FFs at both major and minor sites (24). A minor
FF event provides the opportunity for early pharmacological intervention, at least in the first 3–5 years after the initial FF, during which risk of recurrent FF is maximal (15, 25). Because the Canadian FRAX seems less sensitive among younger groups, adjustment of the threshold should be considered to trigger treatment in younger FF patients (below 65 y), similar to what was proposed in other countries (14, 22, 26). However, the degree of adjustment required is not clear, and further studies are needed. Our study has a number of strengths. First, it represents a relatively large longitudinal cohort of consecutive FF patients with minimal exclusion bias. As such, our cohort differs from administrative cohorts based on BMD testing. Prevalence of BMD testing varies significantly across geographical areas, socioeconomic status, and age (27–30). BMD-based cohorts often consist of younger and healthier individuals, thus underestimating FF rates in the general population. Second, all baseline and most recurrent
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
8
Roux et al
Accuracy of FRAX in Predicting Fractures
FFs were documented by radiographs, obviating falsepositive fracture reports. Third, we lost to follow-up less than 10% of the patients at 12 months and less than 20% at 4 years. Fourth, it is known that FFs at different sites do not have the same strength to predict recurrent FFs (15, 25), although the FF site is not included in the FRAX analysis. We included the sites of the FFs (major vs minor), in addition to treatment, gender, and age, to provide a more detailed analysis. Our study also presents some limitations. First, we used FRAX without BMD (FRAX-BMI) rather than FRAXBMD, and FRAX scores were obtained after the FF from baseline data. Even though FRAX-BMI has been shown to correlate well with FRAX-BMD in Canadians (8, 31), femoral neck BMD is more related to hip FFs and spine BMD to vertebral FFs (32); BMD has been shown to improve the fracture risk estimation at these sites when integrated within the FRAX calculation (33). In our subset of 302 patients with available BMD, the ability of FRAXBMD to predict recurrent FFs was lower relative to FRAXBMI, despite a good correlation between both scores; analyses using FRAX-BMD and CAROC were less robust due to small numbers of patients with BMD, however. Second, our study included only a small number of vertebral FFs, as expected for patients recruited from orthopedic fracture clinics. Systematic detection of vertebral fractures at baseline by radiography would have increased the FRAX scores because vertebral fractures are often asymptomatic (34), and incorporating vertebral FF might have increased the observed rates of recurrent FF because vertebral FFs are strong predictors of subsequent FFs. Finally, only patients with FFs were studied. This may have amplified the limitations of FRAX-BMI at inclusion, although this study design was adapted to the evaluation of recurrent FF. The risk of recurrent fracture is highest in the first years following a FF (15) and may then be transiently increased relative to what can be expected from estimated 10-year FRAX probability, in which risk is considered constant over time (35). In conclusion, data from our prospective cohort of patients with FF suggest that the Canadian-adjusted FRAX tool largely underestimates the risk of FFs prior to an incident FF, particularly in younger patients (⬍65 y old), in men, and in the absence of previous FFs. In addition, FRAX also underestimates the risk of recurrent FFs after an incident FF, particularly in individuals younger than 65 years. As a consequence, the current Osteoporosis Canada recommendations, to restrict pharmacological treatment to FF patients with hip or vertebra or repeat FF and to those with estimated high FRAX (or CAROC) risk, neglect a significant proportion of patients who will present recurrent fractures. This restrictive approach may signif-
J Clin Endocrinol Metab
icantly hamper the potential decrease in recurrent FFs that post-FF intervention programs (such as Fracture Liaison Services) want to attain (36).
Acknowledgments We are especially indebted to our study coordinators, Noémie Poirier and Line Larrivée, for their continuing involvement and dedication. This study was registered at clinicaltrials.gov with an identification of NCT00512499. None of the funding sources had any role in the design of the study, collection, analysis or interpretation of the data, or in the decision to publish this article. Address all correspondence and requests for reprints to: Gilles Boire, MD, MSc, Centre Hospitalier Universitaire de Sherbrooke, 3001, 12th Avenue North, Division of Rheumatology, Room 3853, Sherbrooke, Québec, Canada J1H 5N4. E-mail: [email protected]. This work was supported by unrestricted research grants from Merck Canada, The Alliance for Better Bone Health (Procter & Gamble; Sanofi-Aventis), Amgen Canada, Novartis Pharmaceuticals Canada Inc, Warner Chilcott Canada, Eli Lilly Canada, and Servier Canada and by the Centre de Recherche Clinique Étienne-LeBel from the Centre Hospitalier Universitaire de Sherbrooke, which received a team grant from the Fonds de Recherche du Québec-Santé. Disclosure Summary: M.-C.B., P.-M.A., and F.C. have nothing to declare. S.R. and G.B. have received lecture fees or advisory board fees from Merck Canada, Procter & Gamble Canada, Sanofi-Aventis Canada, Amgen Canada, and Novartis Canada Inc and S.R. from Eli Lilly Canada and Servier Canada Inc. M.B. is an employee of Merck Canada.
References 1. Dennison E, Mohamed MA, Cooper C. Epidemiology of osteoporosis. Rheum Dis Clin North Am. 2006;32:617– 629. 2. Bessette L, Ste-Marie LG, Jean S, et al. The care gap in diagnosis and treatment of women with a fragility fracture. Osteoporos Int. 2008; 19:79 – 86. 3. Kanis JA, Johnell O, De Laet C, et al. A meta-analysis of previous fracture and subsequent fracture risk. Bone. 2004;35:375–382. 4. Pasco JA, Seeman E, Henry MJ, Merriman EN, Nicholson GC, Kotowicz MA. The population burden of fractures originates in women with osteopenia, not osteoporosis. Osteoporos Int. 2006;17:1404 – 1409. 5. Leslie WD, Berger C, Langsetmo L, et al. Construction and validation of a simplified fracture risk assessment tool for Canadian women and men: results from the CaMos and Manitoba cohorts. Osteoporos Int. 2011;22:1873–1883. 6. Nguyen ND, Frost SA, Center JR, Eisman JA, Nguyen TV. Development of prognostic nomograms for individualizing 5-year and 10-year fracture risks. Osteoporos Int. 2008;19:1431–1444. 7. Kanis JA, Johnell O, Oden A, Johansson H, McCloskey E. FRAX and the assessment of fracture probability in men and women from the UK. Osteoporos Int. 2008;19:385–397. 8. Leslie WD, Lix LM, Johansson H, Oden A, McCloskey E, Kanis JA. Independent clinical validation of a Canadian FRAX tool: fracture
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
9. 10. 11.
12.
13.
14.
15. 16.
17.
18.
19.
20.
21.
22.
prediction and model calibration. J Bone Miner Res. 2010;25:2350 – 2358. Chapurlat R. Contribution and limitations of the FRAX tool. Joint Bone Spine. 2013;80:355–357. McCloskey E, Kanis JA. FRAX updates 2012. Curr Opin Rheumatol. 2012;24:554 –560. Sambrook PN, Flahive J, Hooven FH, et al. Predicting fractures in an international cohort using risk factor algorithms without BMD. J Bone Miner Res. 2011;26:2770 –2777. Briot K, Cortet B, Thomas T, et al. 2012 update of French guidelines for the pharmacological treatment of postmenopausal osteoporosis. Joint Bone Spine. 2012;79:304 –313. Leslie WD, Schousboe JT. A review of osteoporosis diagnosis and treatment options in new and recently updated guidelines on case finding around the world. Curr Osteoporos Rep. 2011;9:129 –140. Compston J, Bowring C, Cooper A, et al. Diagnosis and management of osteoporosis in postmenopausal women and older men in the UK: National Osteoporosis Guideline Group (NOGG) update 2013. Maturitas. 2013;75:392–396. Johnell O, Kanis JA, Oden A, et al. Fracture risk following an osteoporotic fracture. Osteoporos Int. 2004;15:175–179. Roux S, Beaulieu M, Beaulieu MC, Cabana F, Boire G. Priming primary care physicians to treat osteoporosis after a fragility fracture: an integrated multidisciplinary approach. J Rheumatol. 2013; 40:703–711. Kreiger N, Tenenhouse A, Joseph L, et al. The Canadian multicentre osteoroposis study (CaMos): background, rationale, methods. Can J Aging. 1999;18:376 –387. Papaioannou A, Morin S, Cheung AM, et al. 2010 clinical practice guidelines for the diagnosis and management of osteoporosis in Canada: summary. CMAJ. 2010;182:1864 –1873. Giangregorio LM, Leslie WD, Lix LM, et al. FRAX underestimates fracture risk in patients with diabetes. J Bone Miner Res. 2012;27: 301–308. Tremollieres FA, Pouilles JM, Drewniak N, Laparra J, Ribot CA, Dargent-Molina P. Fracture risk prediction using BMD and clinical risk factors in early postmenopausal women: sensitivity of the WHO FRAX tool. J Bone Miner Res. 2010;25:1002–1009. Sornay-Rendu E, Munoz F, Delmas PD, Chapurlat RD. The FRAX tool in French women: how well does it describe the real incidence of fracture in the OFELY cohort? J Bone Miner Res. 2010;25:2101– 2107. Aubry-Rozier B, Stoll D, Krieg MA, Lamy O, Hans D. What was
jcem.endojournals.org
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34. 35. 36.
9
your fracture risk evaluated by FRAX(R) the day before your osteoporotic fracture? Clin Rheumatol. 2013;32:219 –223. Fraser LA, Langsetmo L, Berger C, et al. Fracture prediction and calibration of a Canadian FRAX(R) tool: a population-based report from CaMos. Osteoporos Int. 2011;22:829 – 837. Ahmed LA, Center JR, Bjornerem A, et al. Progressively increasing fracture risk with advancing age after initial incident fragility fracture: The Tromso Study. J Bone Miner Res. 2013;28:2214 –2221. Gehlbach S, Saag KG, Adachi JD, et al. Previous fractures at multiple sites increase the risk for subsequent fractures: the Global Longitudinal Study of Osteoporosis in Women. J Bone Miner Res. 2012; 27:645– 653. Kanis JA, McCloskey EV, Johansson H, Strom O, Borgstrom F, Oden A. Case finding for the management of osteoporosis with FRAX—assessment and intervention thresholds for the UK. Osteoporos Int. 2008;19:1395–1408. Rubin KH, Abrahamsen B, Hermann AP, Bech M, Gram J, Brixen K. Prevalence of risk factors for fractures and use of DXA scanning in Danish women. A regional population-based study. Osteoporos Int. 2011;22:1401–1409. Morris CA, Cabral D, Cheng H, et al. Patterns of bone mineral density testing: current guidelines, testing rates, and interventions. J Gen Intern Med. 2004;19:783–790. Davisson L, Warden M, Manivannan S, et al. Osteoporosis screening: factors associated with bone mineral density testing of older women. J Womens Health (Larchmt). 2009;18:989 –994. Rubin KH, Friis-Holmberg T, Hermann AP, Abrahamsen B, Brixen K. Risk assessment tools to identify women with increased risk of osteoporotic fracture: complexity or simplicity? A systematic review. J Bone Miner Res. 2013;28:1701–1717. Leslie WD, Morin S, Johansson H, et al. Fracture risk assessment without bone density measurement in routine clinical practice. Osteoporos Int. 2012;23:75– 85. Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ. 1996;312:1254 –1259. Leslie WD, Lix LM, Johansson H, Oden A, McCloskey E, Kanis JA. A comparative study of using non-hip bone density inputs with FRAX(R). Osteoporos Int. 2012;23:853– 860. Xu W, Perera S, Medich D, et al. Height loss, vertebral fractures, and the misclassification of osteoporosis. Bone. 2011;48:307–311. Silverman SL, Calderon AD. The utility and limitations of FRAX: a US perspective. Curr Osteoporos Rep. 2010;8:192–197. Akesson K, Marsh D, Mitchell PJ, et al. Capture the fracture: a best practice framework and global campaign to break the fragility fracture cycle. Osteoporos Int. 2013;24:2135–2152.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
ARTICLE
E n d o c r i n e
C a r e
The World Health Organization Fracture Risk Assessment Tool (FRAX) Underestimates Incident and Recurrent Fractures in Consecutive Patients With Fragility Fractures Sophie Roux, François Cabana, Nathalie Carrier, Michèle Beaulieu, Pierre-Marc April, Marie-Claude Beaulieu, and Gilles Boire Division of Rheumatology (S.R., N.C., P.-M.A., G.B.), Department of Medicine, Division of Orthopedic Surgery (F.C.), and Department of Family Medicine (M.-C.B.), University of Sherbrooke, Sherbrooke, Québec, Canada J1H 5N4; and Office of Medical Liaison (M.B.), Merck Canada Inc, Montréal, Québec H9H 4M7
Context: The World Health Organization Fracture Risk Assessment tool (FRAX) was developed to identify patients at risk of sustaining a fragility fracture (FF). Objective: The objective of the study was to evaluate estimated FRAX probabilities of FF at the time of a FF and to compare them with the observed incidence of recurrent FF. Methods: A prospective cohort included men and women older than 50 years at the time of a FF. FRAX scores without bone mineral density [FRAX-body mass index (BMI)] were calculated prior to and after the inclusion FF. Recurrent FFs were recorded over a 4-year follow-up. Determinants associated with recurrent FF were determined by univariate and multivariate analyses. Results: FRAX-BMI scores were available in 1399 of the 1409 recruited patients. A high-risk FRAXBMI score was present in only 42.7% patients before and 56.4% after the incident FF. Most FF patients at low or moderate risk before their initial FF were men, younger than 65 years, or without previous FF. Over a median follow-up of 3 years, recurrent FF occurred in 108 patients (2.69 per 100 patient-years). The overall sensitivity of post-FF FRAX to predict a recurrent FF was 71.3% and was specifically lower in patients younger than 65 years (13%) and without previous FF (63%) at inclusion. Conclusions: The FRAX-BMI scores were below the Canadian threshold for treatment in more than half the patients at the time of a FF and in close to a third of patients with recurrent FF. FRAX-BMI severely underestimates the FF risk in patients younger than 65 years old and after a single FF. (J Clin Endocrinol Metab 99: 0000 – 0000, 2014)
ver their lifetime, up to half of women and a quarter of men experience at least one fracture due to bone fragility (1). A fragility fracture (FF) occurs spontaneously or with minimal trauma, usually as a result of a fall from standing height or less (2). According to the World Health Organization classification for bone mineral density (BMD), most FFs occur in individuals who do not have
O
osteoporosis (3, 4). To overcome the limitations of BMD measurements, several multifactorial tools have been developed to accurately identify the patients at high-risk of FF; among these, the World Health Organization Fracture Risk Assessment tool (FRAX) is the most widely used. Other tools with comparable predictive properties include the Canadian Association of Radiologists and Osteopo-
ISSN Print 0021-972X ISSN Online 1945-7197 Printed in U.S.A. Copyright © 2014 by the Endocrine Society Received December 23, 2013. Accepted March 26, 2014.
Abbreviations: BMD, bone mineral density; BMI, body mass index; CAROC, Canadian Association of Radiologists and Osteoporosis Canada; CHUS, Centre Hospitalier Universitaire de Sherbrooke; FF, fragility fracture; FRAX, Fracture Risk Assessment tool; IQR, interquartile range; OPTIMUS, Osteoporosis and Peripheral fractures: Treatment and Investigation in Multidisciplinary Care at the CHUS; PCP, primary care physician; RR, relative risk.
doi: 10.1210/jc.2013-4507
J Clin Endocrinol Metab
jcem.endojournals.org
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
1
2
Roux et al
Accuracy of FRAX in Predicting Fractures
rosis Canada (CAROC) assessment tool, which includes age, femoral neck BMD, corticoid use, and previous FF (5), and the Garvan Fracture Risk Calculator, which also includes fall history (6). FRAX is an algorithm that estimates the 10-year absolute risk of either hip FF or major FF (vertebra, forearm, hip, or proximal humerus) in men and women above the age of 50 years (7). The FRAX score integrates clinical risk factors, alone or in combination with BMD, and has been adapted in several countries, including Canada (8). Evaluating the 10-year fracture risk is currently the recommended approach to assist physicians in treating osteoporosis. Although FRAX is widely promoted for use in clinical practice, it has several limitations (9, 10). In the Global Longitudinal Study of Osteoporosis in Women (GLOW) cohort, neither FRAX nor the Garvan Fracture Risk Calculator was better than the simplest model, which took only age and fracture history into account (11). The cutoffs to identify high-risk patients and the guidelines for therapeutic intervention have been produced by expert committees and vary among countries. Although in some guidelines, patients sustaining a major FF (hip or vertebra) or those with a very low BMD are eligible for pharmacological interventions, regardless of their FRAX score (12, 13), in others the 10-year probabilities of FF alone qualify patients for antiosteoporotic treatment (14). A FF occurring after the age of 50 years is one of the most reliable predictors of subsequent FF. The risk for recurrent FF is maximal in the first few years after the event, gradually diminishing afterward, back to the pre-FF level at 10 years (3, 15). We sought to evaluate the accuracy of Canadian-adjusted FRAX tool in predicting FFs in a Canadian longitudinal prospective cohort of consecutive FF patients, whose 1-year data were recently reported (16). We used the baseline and follow-up data to determine patients’ estimated 10-year fracture risk, either immediately before their initial FF, or considering their recent FF. Because BMD data were not available for a majority of patients, we estimated their 10-year fracture risk using FRAX probabilities without BMD [FRAX-body mass index (BMI)]. We next compared the post-FF FRAXBMI risk with the observed rate of recurrent FF over a 4-year follow-up period. Our findings indicate that the FRAX-BMI scores do not reach the Canadian threshold for pharmacological treatment in more than half of incident FF patients and in close to a third of patients with recurrent FF.
Materials and Methods Study subjects The program Osteoporosis and Peripheral fractures: Treatment and Investigation in Multidisciplinary Care at the CHUS
J Clin Endocrinol Metab
(OPTIMUS) is an ongoing research initiative that combines identification of an initial FF with two incremental and recursive strategies and engages the primary care physicians (PCPs) to manage their own patients. We have recently reported the impact of these strategies on osteoporosis treatment at one year after the FF (16). In this program, patients older than 50 years with X-rayconfirmed fractures (excluding simple avulsion) and treated by orthopedic surgeons at the Centre Hospitalier Universitaire de Sherbrooke (CHUS) were screened by study coordinators. FFs were identified using the Canadian Multicenter Osteoporosis Study questionnaire (17). Patients with multiple FFs during a single event were counted only once. Major fractures (in decreasing order: hip, vertebra, proximal humerus, wrist) prevailed over fractures at other sites (defined as minor fractures). Multiple recurrent FFs occurring during a single or multiple events were defined similarly. Inclusion and exclusion criteria for the OPTIMUS program were described previously (16). Patients were randomly assigned to standard care (no intervention), minimal, or intensive interventions, as described earlier (16). Inpatients with hip FFs were also included but were assessed and treated by rheumatologists and were not assigned to any intervention. Baseline characteristics include demographics (age, gender, weight, height, race, and menopausal status), previous FF or diagnosis of osteoporosis, BMD, comorbidities, currently used medication including antiosteoporosis drugs, estimated calcium and vitamin D intake, tobacco and alcohol consumption, extent of physical activity, and familial history of osteoporosis. Causes of secondary osteoporosis were recorded, including type 1 diabetes, osteogenesis imperfecta, untreated long-standing hyperthyroidism, hypogonadism or premature menopause (⬍45 y), malabsorption and chronic liver disease. All the patients provided consent to contact their PCP and pharmacist. After enrollment, follow-ups were done by phone calls at 4, 8, 12, 24, 36, and 48 months in the intensive intervention and hip FF groups and at 6, 12, 24, 36, and 48 months in the standard care and minimal intervention groups. During phone calls, each patient was asked the following: appointment with their PCP, BMD testing, changes in diet or exercise, additional comorbidities, initiation and adherence with osteoporosis treatment, and details on additional fractures. For the diagnosis of recurrent vertebral fracture, the patient should have had spine radiography and should have been diagnosed by a physician. The Ethics Review Board of the CHUS approved the study.
FRAX calculation The Canadian FRAX tool (http://www.shef.ac.uk/FRAX/ tool.jsp) became available years after the initiation of OPTIMUS and is still infrequently used in our area in which the CAROC tool is preferred. Therefore, FRAX-calculated 10-year probabilities of major FF and hip FF were obtained retrospectively from the baseline data. The FRAX scores were calculated without BMD (FRAX-BMI) for all patients, and with BMD (FRAXBMD) when available (n ⫽ 302). Self-reported ankle fractures prior to the inclusion FF were excluded because qualification for FF was often dubious, and confirmatory X-rays infrequently available, but were considered during the follow-up if they fulfilled the criteria for FF (excluding simple avulsion). The FRAX scores were assigned based on Canadian guidelines (18): high risk (ⱖ20% for major FFs or ⱖ 3% for hip FFs); moderate risk (10%–20% for major FFs and ⬍ 3% for hip FFs); and low risk (⬍10% for major FFs and ⬍ 3% for hip FFs).
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
Statistical analysis Continuous FRAX scores and categorical risks for FF were compared with the baseline demographics and site of FF. Because continuous scores were not normally distributed, the median [and interquartile range (IQR)] was presented, and the scores were compared with Wilcoxon rank-sum test or with KruskalWallis test. Categorical FRAX scores were compared with the 2 test or Fisher exact test if the frequencies were less than 5. The rates of recurrent FF per 100 person-years according to FRAX categories were compared with the 2 test or Fisher exact test. Relative risks (RRs) were estimated with log-binomial univariate regression, and the sensitivity was calculated. A multivariate logbinomial regression was computed to evaluate the variables that predict recurrent FF, including age, gender, osteoporosis treatment, previous FF, minor/major FFs, and FRAX-BMI. Significance was set at P ⬍ .05 and analyses were computed with SAS 9.2 software for Windows (SAS Institute Inc).
jcem.endojournals.org
Table 1.
Results Patient demographics From January 2007 to May 2012, 1409 consecutive consenting patients with incident FFs were enrolled in the prospective cohort. These FFs represented over 80% of all FFs occurring in apt-to-consent patients evaluated by the CHUS orthopedic surgeons (16). Baseline characteristics are shown in Table 1. At inclusion, 104 patients (7.38%) had a secondary cause of osteoporosis (excluding alcohol, tobacco, and corticosteroids), with a significantly higher prevalence in older women (10.28% in women aged 65 y and older) (P ⬍ .001). At baseline, 387 patients (27.5%) were under effective pharmacological treatment for osteoporosis, mainly bisphosphonates; baseline adherence to previously prescribed treatments was not assessed. After 1 year, 775 patients were under effective osteoporosis medications (60.1%), with a mean drug possession of greater than 80% of the doses confirmed by the patients’ pharmacists. In addition, 278 (19.8%) had already suffered a previous FF at baseline, among which 128 (46%) were treated with an effective antiosteoporosis drug. Predicted 10-year absolute risk of fracture before and after the initial fracture We used the Canadian FRAX-BMI tool to determine the 10-year fracture risk. FRAX-BMI scores could not be calculated in 10 patients because of missing weight and/or height; thus, a total of 1399 patients was included in the analysis. The FRAX-BMI scores for major FFs and for hip FFs before and after the initial FF are shown in Table 2. The day before the incident FF, only 42.7% of the patients were at high risk. When the incident FF was imputed into the FRAX-BMI estimation, 56.4% were at high risk (P ⬍ .001, FRAX-BMI before vs after the incident FF). Lower proportions of high-risk scores (36.1% before and 49.5%
3
Baseline Characteristics
Characteristics
Value
Median age, y (IQR) Women, n, % Mean BMI ⫾ SD Tobacco, n, % Nonsmoker Current Smoker Previous smoker Alcohol, n, % Never/occasionally One drink or more by day Personal history of osteoporosis, n, % Family history of osteoporosis, n, % Secondary osteoporosis, n, % Previous FF, n, %a Hip Vertebra Proximal humerus Wrist Other Fragility fracture site at inclusion, n, % Hip Vertebra Proximal humerus Wrist Ankle Other Treatment for osteoporosis at inclusion, n, % Aminobisphosphonate Other effective treatmentb Corticosteroids
67 (59 –79) 1159 (82.3) 26.1 ⫾ 5.7 683 (48.5) 207 (14.7) 519 (36.8) 1283 (91.0) 126 (9.0) 386 (27.5) 304 (22.4) 104 (7.38) 278 (19.8) 43 (16.3) 29 (11.0) 29 (11.0) 100 (37.9) 63 (23.9) 262 (18.6) 21 (1.5) 251 (17.8) 475 (33.7) 265 (18.8) 135 (9.6) 387 (27.5) 330 (23.4) 52 (4.3) 29 (2.6)
a
Self-reported prior ankle fractures excluded. Previous fracture status was unknown for four patients. b Other effective treatments included hormonal replacement therapy, selective estrogen receptor modulators, denosumab, and teriparatide.
after the FF) were observed in initially untreated patients. After the FF, 336 patients (24%) remained in the low-risk category and 274 (19.6%) in the moderate-risk category, indicating that only 192 patients (23.9%) moved from a lower strata to the high-risk category. In the subgroup of patients with BMD data (n ⫽ 302), absolute risk predictions for major FF using FRAX-BMI and FRAX-BMD were strongly correlated (r ⫽ 0.81 before and r ⫽ 0.71 after the baseline FF), as for hip FF (r ⫽ 0.72 before and r ⫽ 0.63 after the baseline FF) (all significant at P ⬍ .0001). Among the 260 hip FF patients, 75.8% were classified as high risk (hip score ⱖ 3%) by FRAX-BMI. Among the patients with major FFs (n ⫽ 999), only 50% were estimated at high risk of major FFs, and 43.7% were at low risk. Most patients with nonmajor FFs (n ⫽ 400) were not considered to be at high risk for major or hip FF. Imputing the initial FF in the FRAX-BMI estimates increased the proportion of patients in the high-risk category for FFs; however, more than one third of patients with major FFs and 60% with minor FFs were still evaluated at low or moderate risk (Table 2).
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
4
Roux et al
Table 2.
Accuracy of FRAX in Predicting Fractures
J Clin Endocrinol Metab
Estimated 10-Year Fracture Risk of Patients With Fragility Fractures Before and After the Inclusion FF Before the Incident Fracturea
Minor and major FF Any FF (n ⫽ 1399) Site of initial FF Hip (n ⫽ 260) Vertebra (n ⫽ 21) Proximal humerus (n ⫽ 247) Wrist (n ⫽ 471) Ankle (n ⫽ 265) Other (n ⫽ 135) Type of initial FF Minor (n ⫽ 400) Major (n ⫽ 999) Age, y ⬍65 (n ⫽ 602) ⱖ65 (n ⫽ 797) Gender Women (n ⫽ 1149) Men (n ⫽ 250) Previous FF No (n ⫽ 1120) Yes (n ⫽ 275) Treatment at inclusion No (n ⫽ 1016) Yes (n ⫽ 383) a
After the Incident Fracturea
FRAX-major med (IQR)
FRAX-hip med (IQR)
High-risk, n, %b
FRAX-major med (IQR)
FRAX-hip med (IQR)
High-risk, n, %b
9.4 (5.3–19)
2 (0.5–7.3)
597 (42.7)
16.0 (10.0 –28.0)
3.7 (1.5–10.0)
789 (56.4)
21.0 (10.0 –27.0)c 9.5 (5.0 –18.0) 11.0 (6.1–21.0) 8.3 (5.4 –17.0) 5.7 (4.0 –10.0) 7.1 (4.3–15.0)
8.3 (3.2–14.0)c 1.8 (0.6 –7.5) 2.6 (0.7– 8.2) 1.5 (0.5–5.6) 0.6 (0.3–2.2) 1.3 (0.3– 4.4)
197 (75.8)c 10 (47.6) 117 (47.4) 171 (36.3) 56 (21.1) 46 (34.1)
29.0 (17.0 –38.0)c 11.0 (9.5–29.0) 11.0 (8.1–18.0) 17.0 (11.0 –29.0) 15.0 (11.0 –25.0) 13.0 (8.5–21.0)
11.5 (5.2–18.5)c 3.6 (1.5–10.0) 1.7 (0.9 – 4.2) 4.0 (1.5–11.0) 3.0 (1.6 – 8.0) 2.9 (1.0 –7.1)
225 (86.5)c 13 (61.9) 146 (59.1) 246 (52.2) 92 (34.7) 67 (49.6)
6.6 (4.1–12)c 11 (6.2–22)
0.9 (0.3–3.1)c 2.9 (0.7–9.1)
103 (25.8)c 11.0 (8.2–18.0)c 497 (49.7) 18.0 (11.0 –32.0)
2.0 (1.0 –5.2)c 4.9 (1.8 –13.0)
159 (39.8)c 630 (63.1)
5.1 (3.8 –7.1)c 17.0 (10.0 –26.0)
0.5 (0.3– 0.8)c 21 (3.5)c 5.9 (2.8 –11.0) 579 (72.6)
1.3 (0.8 –2.0)c 9.0 (4.7–16.0)
83 (13.8)c 706 (88.6)
12.0 (6.1–22.0)c 5.3 (3.8 – 8.5)
2.4 (0.6 – 8.5)c 1.1 (0.4 –3.4)
7.6 (4.7–14.5)c 24.0 (14.0 –36.0)
1.3 (0.4 –5.1)c 386 (34.5)c 14.0 (9.6 –25.0)c 7.5 (3.0 –16.0) 211 (76.7) 24.0 (14.0 –36.0)
3 (1.3–9.0)c 7.5 (3.0 –16.0)
575 (51.3)c 211 (76.7)
7.8 (4.7–16.0)c 15.0 (8.6 –25.0)
1.3 (0.4 –5.6)c 367 (36.1)c 14.0 (9.3–25.0)c 4.4 (1.4 –10.0) 233 (60.8) 22.0 (14.0 –33.0)
2.8 (1.3– 8.6)c 6.8 (2.7–13.0)
503 (49.5)c 286 (74.7)
9.8 (7.7–13.0)c 25.0 (17.0 –35.0)
532 (46.3)c 18.0 (11.0 –31.0)c 4.2 (1.6 –12.0)c 68 (27.2) 9.9 (7.4 –14.0) 2.4 (1.0 –5.3)
679 (59.1)c 110 (44.0)
The 10-year risk of hip fracture or major fracture was calculated both before and after the FF. We used the Canadian BMI-based FRAX score.
b
High-risk score includes a major FRAX score of 20% or greater or a hip FRAX score of 3% or greater.
c
P ⬍ .001 comparison between parameters in each clinical factor category.
While analyzing the FRAX-BMI scores according to baseline characteristics, most patients estimated at low or moderate risk before their FF belonged to defined subcategories: many were men, and most were under 65 years of age (Figure 1), without history of FFs or not receiving antiosteoporotic treatment at baseline (Table 2). When the same analysis was restricted to hip or other major FFs, a similar pattern was observed (Table 3). These findings highlight that the Canadian FRAX-BMI tool does not accurately evaluate the 10-year fracture risk in specific subgroups, particularly among those patients younger than 65 years and in men. Recurrent fractures: sites and rates After 4013 patient-years of follow-up (median 3 y), 122 recurrent FFs occurred in 108 patients (Table 4); 12 patients had two recurrent FFs and one patient had three (only one recurrent FF was considered). Rates of recurrence were 2.69 per 100 patient-years, and the rates were higher after hip (3.96) and wrist FFs (2.96). In the analysis according to the FF site at inclusion, 21.3% of the patients who developed a recurrent FF at a major site had a minor FF at inclusion. Globally, the risk for a recurrent major FF was 57 per 1009 (5.84%) after an
incident major FF and 16 per 400 (4.0%) after an incident minor FF (Table 4). Recurrent fractures according to FRAX-BMI, FRAXBMD, and CAROC scores The FRAX-BMI high-risk category calculated after the incident FF was moderately useful to predict patient’s subsequent FF because only 70% of all recurrent FF patients were estimated at high risk. The hip FRAX-BMI score was more accurate in patients with a recurrent hip FF because 91.7% of them had a score of 3% or greater. However, only 50% of patients with a recurrent major FF had a FRAX-BMI score for major FF of 20% or greater (Figure 2). We could not evaluate the FRAX-BMI specificity in our follow-up study, as FRAX estimates FF risk over a 10-year period. However, because 30% of patients who developed a recurrent FF were not estimated at high risk of FFs, we were confident to analyze the sensitivity of the FRAX-BMI tool (Table 5). Although being at high risk significantly increased the RR of FFs, the sensitivity of the FRAX-BMI tool to detect recurrent FFs was about 70% in the whole cohort. The FRAX-BMD was even less sensitive (40%) to detect patients who would sustain a recurrent FF. In the
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
jcem.endojournals.org
5
high-risk CAROC category to detect recurrent FF was 55% (Table 5). In univariate analyses, a significant increase in RR for recurrent FFs was observed in older patients (ⱖ65 y). Similarly, patients with previous FFs at inclusion (at least two FFs before the recurrent FF) and those already receiving osteoporosis treatment at the time of their initial FF (more likely with a previous FF or a more severe disease) had a significantly higher risk of recurrent FFs. The risk of recurrent FFs was not affected by the site of initial FFs (major/minor). In multivariate analyses, a previous FF at baseline was the only significant factor affecting the risk of recurrent FF (Table 5). The FRAX-BMI scores were also compared within each clinical subgroup of patients, with low statistical power due to small numbers in each subcategory. Most of the patients with low or moderate risk when they developed a recurrent FF (n ⫽ 31) were younger than 65 years (87%), did not have a previous FF (84%), and were not treated for osteoporosis at baseline (68%); the sensitivity of estimated FRAX-BMI high risk to predict recurrent FF in these subgroups was low (13% in patients ⬍ 65 y old, and 63% after a single FF). Most of these 31 patients with an estimated low or moderate risk after inclusion had initially sustained major FF (77%) (Supplemental Table 1). Taken together, our findings suggest that the FRAX-BMI tool underestimates the probabilities of recurrent FF, particularly in younger patients (⬍65 y) and after a single major FF.
Figure 1. Continuous FRAX-BMI probabilities (calculated before and after the incident FF) for major FFs (A) and hip FFs (B) according to age (stratified by 5 y groups). Results are presented as box plots, and the line denotes the median value. Dotted lines represent the cutoffs for the high-risk score: 20% for major FFs and 3% for hip FFs.
Discussion
same subset of patients, we also calculated the CAROC score (5). All patients were at least at moderate risk for FFs by definition (at least one FF), and the sensitivity of the
To accurately predict each patient’s risk for a negative outcome is every clinician’s dream. A correct identifica-
Table 3. Estimated 10-Year Fracture Risk of Patients With Fragility Fractures Before and After the Inclusion FF. Analysis Restricted to Major Fractures at Inclusion Before the Incident Fracturea Major FFs only Any major FF (n ⫽ 999) Age, y ⬍65 (n ⫽ 366) ⱖ65 (n ⫽ 633) Gender Women (n ⫽ 864) Men (n ⫽ 135) Previous fracture No (n ⫽ 777) Yes (n ⫽ 219) Treatment at inclusion No (n ⫽ 711) Yes (n ⫽ 288) a
After the Incident Fracturea
FRAX-major med (IQR)
FRAX-hip med (IQR)
High risk, n (%)b
FRAX-major med (IQR)
FRAX-hip med (IQR)
High risk, n (%)b
11 (6.2–22)
2.9 (0.7–9.1)
497 (49.7)
18.0 (11.0 –32.0)
4.9 (1.8 –13.0)
630 (63.1)
5.6 (4.2–7.5)c 18.0 (11.0 –26.0)
0.5 (0.3–1.0)c 7.0 (3.2–12.0)
15 (4.1)c 482 (76.1)
11.0 (8.5–13.0)c 27.0 (18.0 –37.0)
1.5 (1.0 –2.2)c 10.0 (5.3–17.0)
59 (16.1)c 571 (90.2)
13.0 (6.8 –24.0)c 6.8 (4.5–10.0)
3.3 (0.8 –10.0)c 2.0 (0.6 – 4.4)
447 (51.7)c 50 (37.0)
20.0 (13.0 –33.5)c 11.0 (8.6 –15.0)
5.5 (2.0 –14.0)c 3.5 (1.5– 6.4)
553 (64.0)c 77 (57.0)
8.6 (5.4 –18.0)c 26.0 (15.0 –37.0)
1.8 (0.6 – 6.9)c 9.1 (3.4 –18.0)
319 (41.1)c 175 (79.9)
16.0 (11.0 –29.0)c 26.0 (15.0 –37.0)
4.1 (1.7–11.0)c 9.1 (3.4 –18.0)
452 (58.2)c 175 (79.9)
9.2 (5.6 –20.0)c 17.0 (9.3–26.0)
2.1 (0.6 –7.9)c 5.4 (1.6 –11.0)
312 (43.9)c 185 (64.2)
16.0 (11.0 –29.0)c 24.0 (15.0 –35.0)
4.0 (1.6 –11.0)c 7.8 (3.2–14.0)
409 (57.5)c 221 (76.7)
The 10-year risk of hip fracture or of major fracture was calculated both before and after the FF. We used the Canadian BMI-based FRAX score.
b
High-risk score includes a major FRAX score of 20% or greater or a hip FRAX score of 3% or greater.
c
P ⬍ .001 comparison between parameters in each clinical factor category.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
6
Roux et al
Table 4.
Accuracy of FRAX in Predicting Fractures
J Clin Endocrinol Metab
Rates of Recurrent FF According to Sites of Initial FF Site of Reccurent FF
Site of FF at Inclusion
Recurrent FF Cumulative FF Per 100 Proximal Inclusion Re-FF Re-FF, % Time, y Patient-Years Hip Vertebra Humerus Wrist Ankle Othera
Total Hip Vertebra Proximal humerus Wrist Ankle Other sites
1409 262 21 251 475 265 135
a
108b 28 1 14 39 11 15
7.67 10.69 4.76 5.58 8.21 4.15 11.11
4012.68 707.95 58.21 722.25 1317.95 776.5 429.83
2.69 3.96 1.72 1.94 2.96 1.42 3.49
27 11 0 4 7 1 4
14 6 1 2 3 1 1
14 1 0 2 7 2 2
20 2 0 4 9 2 3
7 1 0 1 3 1 1
26 7 0 1 10 4 4
Other sites include the following: upper limb (n ⫽ 2), lower limb (n ⫽ 5), clavicle (n ⫽ 1), pelvis (n ⫽ 2), ribs (n ⫽ 15), and coccyx (n ⫽ 1).
b
A total of 122 FFs occurred in 108 patients. Only the most significant of multiple recurrent FFs are reported here: hip greater than vertebra greater than proximal humerus greater than wrist greater than minor site.
tion of patients at high risk of FFs would allow early treatment and decrease the numbers needed to treat to prevent future injuries. BMD fails to deliver such precise estimates of risk because more than half of FFs occur in nonosteoporotic patients (3). The FRAX tool was developed to overcome the limitations of BMD by integrating several clinical risk factors, including age and past fracture, either alone or in combination with BMD (7). In population studies, fracture rates closely parallel FRAX predicted rates (8). Most guidelines thus build on FRAX to identify men and women in need for preventive treatment. Nonetheless, in the current study, using the Canadian adjusted FRAX-BMI score in consecutive patients included at the time of a FF, we found that just before their
Figure 2. Continuous FRAX-BMI probabilities (accounting for the incident FF) for major FFs (A) and hip FFs (B) according to the type of recurrent FF. Results are presented as dot plots, and the line denotes the median value. Dotted lines represent the cutoffs for the high-risk score: 20% for major FFs and 3% for hip FFs.
FF, more than half the patients were estimated at low or moderate risk for FF. The low sensitivity of FRAX assessment was not driven by inclusion of incident minor FF in our cohort because the same results were obtained when considering baseline major FF only. When the incident FF event was added to the FRAX-BMI calculation, a large proportion, even after major (and even hip) FFs, were still not estimated at high risk for FFs. The Canadian-adjusted FRAX-BMI high-risk category was therefore inadequate to identify most the patients before they actually sustained an incident major FF, and it remained insensitive (71%), even after the FF event. Moreover, over a 4-year follow-up period, one third of the patients who developed a recurrent FF were not estimated at high risk. The Canadian-adjusted FRAX-BMI high-risk category was similarly insensitive to identify patients with any recurrent FF, although FRAX particularly underestimates the impact of a recent major FF on subsequent fracture risk. Limiting analyses to initially untreated patients did not improve FRAX performance. Subgroup analyses further highlighted that the FRAX tool specifically underestimates the probabilities of FFs in patients below the age of 65 years, with a sensitivity of 71% for incident FFs and only 13% for recurrent FFs. The FRAX model does not account for a number of variables that impact the actual 10-year fracture risk: dose and duration of corticosteroids, levels of alcohol or tobacco consumption, activity or duration of predisposing diseases such as rheumatoid arthritis or endocrinopathies (diabetes, hypopituitarism, male hypogonadism), multiple drugs potentially inducing bone loss or increasing fracture risk, and risks associated with falls (10, 19). Some adjustments have been proposed since the FRAX release in 2008 (10). Several studies have suggested that FRAX may not be sufficiently sensitive toward some subgroups of patients. In early postmenopausal women, the FRAX predictive value was low and showed no improvement over
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
Table 5.
jcem.endojournals.org
7
Relative Risk of Recurrent FF According to FRAX and Clinical Variables Total, n
Re-FF n (%)
Re-FF Per 100 Patient-Years
Univariate RR (CI 95%)
Multivariate RR (CI 95%)
610
31 (5.1)
1.74
1
1
789
77 (9.8)
3.49
1.9 (1.3–2.9)b
1.3 (0.7–2.3)
621 778
32 (5.2) 76 (9.8)
1.76 3.50
1 1.9 (1.3–2.8)b
205
12 (5.9)
2.17
1
97
8 (8.2)
3.11
1.4 (0.6 –3.3)
40.0
207 95
12 (5.8) 8 (8.4)
2.17 3.14
1 1.5 (0.6 –3.4)
40.0
152 148
9 (5.9) 11 (7.4)
2.39 2.64
1 1.3 (0.5–2.9)
55.0
603 806
31 (5.1) 77 (9.6)
1.76 3.42
1 1.9 (1.2–2.8)b
1 1.4 (0.8 –2.5)
71.3
250 1159
12 (4.8) 96 (8.3)
1.73 2.89
1 1.7 (1–3.1)
1 1.4 (0.8 –2.5)
11.1
1127 278
70 (6.2) 37 (13.3)
2.21 4.42
1 2.1 (1.5–3.1)c
1 1.7 (1.1–2.5)b
34.6
1022 387
64 (6.3) 44 (11.4)
2.22 3.90
1 1.8 (1.3–2.6)b
1 1.4 (0.9 –2.1)
40.7
400 1009
26 (6.5) 82 (8.1)
2.16 2.92
1 1.3 (0.8 –1.9)
1 1 (0.7–1.6)
75.9
Sensitivity
a
FRAX-BMI (major and hip) Low-moderate (major ⬍ 20% and hip ⬍ 3%) High (major ⱖ 20% and/or hip ⱖ 3%) FRAX-BMI (hip)a Low (⬍3%) High (ⱖ3%) FRAX-BMD (major and hip)a Low-moderate (major ⬍ 20% and hip ⬍ 3%) High (major ⱖ 20% and/or hip ⱖ 3%) FRAX-BMD (hip)a Low (⬍3%) High (ⱖ3%) CAROCa Moderate High Age, y ⬍65 ⱖ65 Gender Men Women Previous FF (at inclusion) No Yes Osteoporosis treatment at inclusion No Yes Site of FF Minor Major a
71.3 70.4
FRAX and CAROC scores include the initial FF
b
P ⬍ .01 compared between each category.
c
P ⬍ .001 compared between each category.
hip BMD (20). In the French cohort Os des Femmes de Lyon (OFELY), greater than 50% of women with an incident major FF had a FRAX score less than 9%, and 90% had less than 24%. FRAX score severely underestimated the risk of major FF in women older than 65 years with low BMD (21). In a small study in Switzerland, half the patients showed low FRAX risk on the day before FF (22). Our observations in a larger cohort confirm the low sensitivity of the FRAX tool to predict initial FF and demonstrate for the first time that post-FF FRAX is also rather insensitive to predict recurrent FF. In the FRAX algorithm adapted for the Canadian population (8, 23), a high 10-year fracture risk is defined as 20% or more for major FFs and 3% or more for hip FFs. Treatment is recommended after hip or spine FFs, after two FF events at any sites, as well as with a high-risk FRAX score (18). Our re-FF study results are consistent with the recent report that initial minor FFs increase the risk of recurrent FFs at both major and minor sites (24). A minor
FF event provides the opportunity for early pharmacological intervention, at least in the first 3–5 years after the initial FF, during which risk of recurrent FF is maximal (15, 25). Because the Canadian FRAX seems less sensitive among younger groups, adjustment of the threshold should be considered to trigger treatment in younger FF patients (below 65 y), similar to what was proposed in other countries (14, 22, 26). However, the degree of adjustment required is not clear, and further studies are needed. Our study has a number of strengths. First, it represents a relatively large longitudinal cohort of consecutive FF patients with minimal exclusion bias. As such, our cohort differs from administrative cohorts based on BMD testing. Prevalence of BMD testing varies significantly across geographical areas, socioeconomic status, and age (27–30). BMD-based cohorts often consist of younger and healthier individuals, thus underestimating FF rates in the general population. Second, all baseline and most recurrent
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
8
Roux et al
Accuracy of FRAX in Predicting Fractures
FFs were documented by radiographs, obviating falsepositive fracture reports. Third, we lost to follow-up less than 10% of the patients at 12 months and less than 20% at 4 years. Fourth, it is known that FFs at different sites do not have the same strength to predict recurrent FFs (15, 25), although the FF site is not included in the FRAX analysis. We included the sites of the FFs (major vs minor), in addition to treatment, gender, and age, to provide a more detailed analysis. Our study also presents some limitations. First, we used FRAX without BMD (FRAX-BMI) rather than FRAXBMD, and FRAX scores were obtained after the FF from baseline data. Even though FRAX-BMI has been shown to correlate well with FRAX-BMD in Canadians (8, 31), femoral neck BMD is more related to hip FFs and spine BMD to vertebral FFs (32); BMD has been shown to improve the fracture risk estimation at these sites when integrated within the FRAX calculation (33). In our subset of 302 patients with available BMD, the ability of FRAXBMD to predict recurrent FFs was lower relative to FRAXBMI, despite a good correlation between both scores; analyses using FRAX-BMD and CAROC were less robust due to small numbers of patients with BMD, however. Second, our study included only a small number of vertebral FFs, as expected for patients recruited from orthopedic fracture clinics. Systematic detection of vertebral fractures at baseline by radiography would have increased the FRAX scores because vertebral fractures are often asymptomatic (34), and incorporating vertebral FF might have increased the observed rates of recurrent FF because vertebral FFs are strong predictors of subsequent FFs. Finally, only patients with FFs were studied. This may have amplified the limitations of FRAX-BMI at inclusion, although this study design was adapted to the evaluation of recurrent FF. The risk of recurrent fracture is highest in the first years following a FF (15) and may then be transiently increased relative to what can be expected from estimated 10-year FRAX probability, in which risk is considered constant over time (35). In conclusion, data from our prospective cohort of patients with FF suggest that the Canadian-adjusted FRAX tool largely underestimates the risk of FFs prior to an incident FF, particularly in younger patients (⬍65 y old), in men, and in the absence of previous FFs. In addition, FRAX also underestimates the risk of recurrent FFs after an incident FF, particularly in individuals younger than 65 years. As a consequence, the current Osteoporosis Canada recommendations, to restrict pharmacological treatment to FF patients with hip or vertebra or repeat FF and to those with estimated high FRAX (or CAROC) risk, neglect a significant proportion of patients who will present recurrent fractures. This restrictive approach may signif-
J Clin Endocrinol Metab
icantly hamper the potential decrease in recurrent FFs that post-FF intervention programs (such as Fracture Liaison Services) want to attain (36).
Acknowledgments We are especially indebted to our study coordinators, Noémie Poirier and Line Larrivée, for their continuing involvement and dedication. This study was registered at clinicaltrials.gov with an identification of NCT00512499. None of the funding sources had any role in the design of the study, collection, analysis or interpretation of the data, or in the decision to publish this article. Address all correspondence and requests for reprints to: Gilles Boire, MD, MSc, Centre Hospitalier Universitaire de Sherbrooke, 3001, 12th Avenue North, Division of Rheumatology, Room 3853, Sherbrooke, Québec, Canada J1H 5N4. E-mail: [email protected]. This work was supported by unrestricted research grants from Merck Canada, The Alliance for Better Bone Health (Procter & Gamble; Sanofi-Aventis), Amgen Canada, Novartis Pharmaceuticals Canada Inc, Warner Chilcott Canada, Eli Lilly Canada, and Servier Canada and by the Centre de Recherche Clinique Étienne-LeBel from the Centre Hospitalier Universitaire de Sherbrooke, which received a team grant from the Fonds de Recherche du Québec-Santé. Disclosure Summary: M.-C.B., P.-M.A., and F.C. have nothing to declare. S.R. and G.B. have received lecture fees or advisory board fees from Merck Canada, Procter & Gamble Canada, Sanofi-Aventis Canada, Amgen Canada, and Novartis Canada Inc and S.R. from Eli Lilly Canada and Servier Canada Inc. M.B. is an employee of Merck Canada.
References 1. Dennison E, Mohamed MA, Cooper C. Epidemiology of osteoporosis. Rheum Dis Clin North Am. 2006;32:617– 629. 2. Bessette L, Ste-Marie LG, Jean S, et al. The care gap in diagnosis and treatment of women with a fragility fracture. Osteoporos Int. 2008; 19:79 – 86. 3. Kanis JA, Johnell O, De Laet C, et al. A meta-analysis of previous fracture and subsequent fracture risk. Bone. 2004;35:375–382. 4. Pasco JA, Seeman E, Henry MJ, Merriman EN, Nicholson GC, Kotowicz MA. The population burden of fractures originates in women with osteopenia, not osteoporosis. Osteoporos Int. 2006;17:1404 – 1409. 5. Leslie WD, Berger C, Langsetmo L, et al. Construction and validation of a simplified fracture risk assessment tool for Canadian women and men: results from the CaMos and Manitoba cohorts. Osteoporos Int. 2011;22:1873–1883. 6. Nguyen ND, Frost SA, Center JR, Eisman JA, Nguyen TV. Development of prognostic nomograms for individualizing 5-year and 10-year fracture risks. Osteoporos Int. 2008;19:1431–1444. 7. Kanis JA, Johnell O, Oden A, Johansson H, McCloskey E. FRAX and the assessment of fracture probability in men and women from the UK. Osteoporos Int. 2008;19:385–397. 8. Leslie WD, Lix LM, Johansson H, Oden A, McCloskey E, Kanis JA. Independent clinical validation of a Canadian FRAX tool: fracture
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
doi: 10.1210/jc.2013-4507
9. 10. 11.
12.
13.
14.
15. 16.
17.
18.
19.
20.
21.
22.
prediction and model calibration. J Bone Miner Res. 2010;25:2350 – 2358. Chapurlat R. Contribution and limitations of the FRAX tool. Joint Bone Spine. 2013;80:355–357. McCloskey E, Kanis JA. FRAX updates 2012. Curr Opin Rheumatol. 2012;24:554 –560. Sambrook PN, Flahive J, Hooven FH, et al. Predicting fractures in an international cohort using risk factor algorithms without BMD. J Bone Miner Res. 2011;26:2770 –2777. Briot K, Cortet B, Thomas T, et al. 2012 update of French guidelines for the pharmacological treatment of postmenopausal osteoporosis. Joint Bone Spine. 2012;79:304 –313. Leslie WD, Schousboe JT. A review of osteoporosis diagnosis and treatment options in new and recently updated guidelines on case finding around the world. Curr Osteoporos Rep. 2011;9:129 –140. Compston J, Bowring C, Cooper A, et al. Diagnosis and management of osteoporosis in postmenopausal women and older men in the UK: National Osteoporosis Guideline Group (NOGG) update 2013. Maturitas. 2013;75:392–396. Johnell O, Kanis JA, Oden A, et al. Fracture risk following an osteoporotic fracture. Osteoporos Int. 2004;15:175–179. Roux S, Beaulieu M, Beaulieu MC, Cabana F, Boire G. Priming primary care physicians to treat osteoporosis after a fragility fracture: an integrated multidisciplinary approach. J Rheumatol. 2013; 40:703–711. Kreiger N, Tenenhouse A, Joseph L, et al. The Canadian multicentre osteoroposis study (CaMos): background, rationale, methods. Can J Aging. 1999;18:376 –387. Papaioannou A, Morin S, Cheung AM, et al. 2010 clinical practice guidelines for the diagnosis and management of osteoporosis in Canada: summary. CMAJ. 2010;182:1864 –1873. Giangregorio LM, Leslie WD, Lix LM, et al. FRAX underestimates fracture risk in patients with diabetes. J Bone Miner Res. 2012;27: 301–308. Tremollieres FA, Pouilles JM, Drewniak N, Laparra J, Ribot CA, Dargent-Molina P. Fracture risk prediction using BMD and clinical risk factors in early postmenopausal women: sensitivity of the WHO FRAX tool. J Bone Miner Res. 2010;25:1002–1009. Sornay-Rendu E, Munoz F, Delmas PD, Chapurlat RD. The FRAX tool in French women: how well does it describe the real incidence of fracture in the OFELY cohort? J Bone Miner Res. 2010;25:2101– 2107. Aubry-Rozier B, Stoll D, Krieg MA, Lamy O, Hans D. What was
jcem.endojournals.org
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34. 35. 36.
9
your fracture risk evaluated by FRAX(R) the day before your osteoporotic fracture? Clin Rheumatol. 2013;32:219 –223. Fraser LA, Langsetmo L, Berger C, et al. Fracture prediction and calibration of a Canadian FRAX(R) tool: a population-based report from CaMos. Osteoporos Int. 2011;22:829 – 837. Ahmed LA, Center JR, Bjornerem A, et al. Progressively increasing fracture risk with advancing age after initial incident fragility fracture: The Tromso Study. J Bone Miner Res. 2013;28:2214 –2221. Gehlbach S, Saag KG, Adachi JD, et al. Previous fractures at multiple sites increase the risk for subsequent fractures: the Global Longitudinal Study of Osteoporosis in Women. J Bone Miner Res. 2012; 27:645– 653. Kanis JA, McCloskey EV, Johansson H, Strom O, Borgstrom F, Oden A. Case finding for the management of osteoporosis with FRAX—assessment and intervention thresholds for the UK. Osteoporos Int. 2008;19:1395–1408. Rubin KH, Abrahamsen B, Hermann AP, Bech M, Gram J, Brixen K. Prevalence of risk factors for fractures and use of DXA scanning in Danish women. A regional population-based study. Osteoporos Int. 2011;22:1401–1409. Morris CA, Cabral D, Cheng H, et al. Patterns of bone mineral density testing: current guidelines, testing rates, and interventions. J Gen Intern Med. 2004;19:783–790. Davisson L, Warden M, Manivannan S, et al. Osteoporosis screening: factors associated with bone mineral density testing of older women. J Womens Health (Larchmt). 2009;18:989 –994. Rubin KH, Friis-Holmberg T, Hermann AP, Abrahamsen B, Brixen K. Risk assessment tools to identify women with increased risk of osteoporotic fracture: complexity or simplicity? A systematic review. J Bone Miner Res. 2013;28:1701–1717. Leslie WD, Morin S, Johansson H, et al. Fracture risk assessment without bone density measurement in routine clinical practice. Osteoporos Int. 2012;23:75– 85. Marshall D, Johnell O, Wedel H. Meta-analysis of how well measures of bone mineral density predict occurrence of osteoporotic fractures. BMJ. 1996;312:1254 –1259. Leslie WD, Lix LM, Johansson H, Oden A, McCloskey E, Kanis JA. A comparative study of using non-hip bone density inputs with FRAX(R). Osteoporos Int. 2012;23:853– 860. Xu W, Perera S, Medich D, et al. Height loss, vertebral fractures, and the misclassification of osteoporosis. Bone. 2011;48:307–311. Silverman SL, Calderon AD. The utility and limitations of FRAX: a US perspective. Curr Osteoporos Rep. 2010;8:192–197. Akesson K, Marsh D, Mitchell PJ, et al. Capture the fracture: a best practice framework and global campaign to break the fragility fracture cycle. Osteoporos Int. 2013;24:2135–2152.
The Endocrine Society. Downloaded from press.endocrine.org by [${individualUser.displayName}] on 24 June 2014. at 00:23 For personal use only. No other uses without permission. . All rights reserved.
