This article requires a subscription to view the full text. If you have a subscription you may use the login form below to view the article. Access to this article can also be purchased.
Abstract
Background and objectives The Fracture Risk Assessment Tool (FRAX) is widely used to predict the 10-year probability of fracture; however, the clinical utility of FRAX in CKD is unknown. This study assessed the predictive ability of FRAX in individuals with reduced kidney function compared with individuals with normal kidney function.
Design, setting, participants, & measurements The discrimination and calibration (defined as the agreement between observed and predicted values) of FRAX were examined using data from the Canadian Multicentre Osteoporosis Study (CaMos). This study included individuals aged ≥40 years with an eGFR value at year 10 of CaMos (defined as baseline). The cohort was stratified by kidney function at baseline (eGFR<60 ml/min per 1.73 m2 [72.2% stage 3a, 23.8% stage 3b, and 4.0% stage 4/5] versus ≥60 ml/min per 1.73 m2) and followed individuals for a mean of 4.8 years for an incident major osteoporotic fracture (clinical spine, hip, forearm/wrist, or humerus).
Results There were 320 individuals with an eGFR<60 ml/min per 1.73 m2 and 1787 with an eGFR≥60 ml/min per 1.73 m2. The mean age was 67±10 years and 71% were women. The 5-year observed major osteoporotic fracture risk was 5.3% (95% confidence interval [95% CI], 3.3% to 8.6%) in individuals with an eGFR<60 ml/min per 1.73 m2, which was comparable to the FRAX-predicted fracture risk (6.4% with bone mineral density; 8.2% without bone mineral density). A statistically significant difference was not observed in the area under the curve values for FRAX in individuals with an eGFR<60 ml/min per 1.73 m2 versus ≥60 ml/min per 1.73 m2 (0.69 [95% CI, 0.54 to 0.83] versus 0.76 [95% CI, 0.70 to 0.82]; P=0.38).
Conclusions This study showed that FRAX was able to predict major osteoporotic fractures in individuals with reduced kidney function; further study is needed before FRAX should be routinely used in individuals with reduced kidney function.
Introduction
The World Health Organization’s Fracture Risk Assessment Tool (FRAX) is used commonly in the general population to predict the 10-year probability of a major osteoporotic fracture (defined as hip, forearm, clinical vertebral, and humerus fractures) using an algorithm that includes age, sex, and several clinical risk factors for fracture (bone mineral density [BMD] optional) (1,2). The clinical risk factors for fracture incorporated in the FRAX algorithm include the following: parental hip fracture, previous fragility fracture, rheumatoid arthritis, current smoking, secondary osteoporosis (which does not include CKD), low body mass index (BMI) (<19 kg/m2), prolonged glucocorticoid use, and excessive alcohol intake (3–7).
Men and women with CKD have a high fracture risk (8–11). For example, women with moderate declines in kidney function (eGFR 45–59 ml/min per 1.73 m2) are at almost a 4-fold increased risk of fracture compared with women with normal kidney function (11). The clinical utility of FRAX in predicting fracture risk in patients with reduced kidney function is uncertain. CKD is associated with disturbances in mineral metabolism including changes in calcium, phosphate, and parathyroid hormone, which likely alter bone volume, turnover, and mineralization increasing fracture risk (12). Therefore, factors in the FRAX algorithm that are associated with fracture risk in the general population may not accurately predict fracture in individuals with reduced kidney function. One prior study reported on the prognostic value of FRAX in individuals with reduced kidney function; however, this study was cross-sectional and did not include a comparison group of individuals with normal kidney function (13). Our study addresses these limitations. We utilized data from a multicentre cohort study (Canadian Multicentre Osteoporosis Study [CaMos]) to characterize the predictive ability of FRAX in patients with reduced kidney function, and to determine whether the predictive ability differs from individuals with normal kidney function. As a secondary analysis, we examined the ability of FRAX to predict fracture when adding CKD as a secondary cause of osteoporosis in individuals with reduced kidney function. We also assessed the ability of age, T score, and T score with a history of fall to predict fractures in both groups.
Materials and Methods
CaMos
CaMos is a prospective observational study that began in January 1996 (14). Detailed methods concerning CaMos have been published elsewhere (14,15). Briefly, noninstitutionalized individuals were eligible to participate in CaMos if they were aged ≥25 years at the start of the study, lived within a 50-km radius of one of nine major Canadian cities (St. John’s, Halifax, Quebec City, Toronto, Hamilton, Kingston, Calgary, Vancouver, and Saskatoon), and could speak English, French, or Chinese (14). Residential phone numbers were used to randomly select households. Within households, one member who met eligibility criteria was randomly selected; at the baseline interview, 42% of participants contacted agreed to participate (14). In January 1996, participants completed a standardized interviewer-administered questionnaire; the questionnaire was subsequently administered every 5 years. The questionnaire assessed demographics, medication use, nutrition, general health, medical history, fracture risk factors, and fracture events (14). BMD, weight, and height were also assessed at baseline and every 5 years (14). In year 10, blood samples were obtained and serum stored from participants in eight of the nine study centers. Serum creatinine was analyzed by CDL Laboratories Inc (Montreal, QC, Canada). In agreement with the Declaration of Helsinki, written informed consent was provided by study participants. Ethics approval was obtained from McGill University and from each study center’s applicable ethic review board.
Cohort
The beginning date of this study (cohort entry) was the CaMos study year 10—the first time eight of the nine centers assessed blood work. For this analysis, we included individuals who met the following criteria at cohort entry: (1) men and women who were aged ≥40 years, (2) those who had a creatinine value, (3) those who had femoral neck BMD measurement, and (4) those with no prior organ transplant. Creatinine values were missing in those who did not sign the consent form for blood and in those who were from Hamilton (the center that did not collect blood work). We estimated the eGFR using the CKD Epidemiology Collaboration equation (16). We defined kidney function at cohort entry using thresholds defined in the 2012 Kidney Disease Improving Global Outcomes (KDIGO) guidelines (17); an eGFR<60 ml/min per 1.73 m2 was defined as reduced kidney function and an eGFR≥60 ml/min per 1.73 m2 was defined as normal kidney function. We used this classification for our primary analysis. To characterize the degree of renal impairment, we further stratified kidney function in individuals with an eGFR<60 ml/min per 1.73 m2 according to the 2012 KDIGO guidelines as follows: 45–59 ml/min per 1.73 m2 (stage 3a), 30–44 ml/min per 1.73 m2 (stage 3b), 15–29 ml/min per 1.73 m2 (stage 4), and <15 ml/min per 1.73 m2 (stage 5) (17).
BMD
BMD was measured at the femoral neck using the Hologic QDR dual energy x-ray absorptiometry scanner (Hologic, Marlborough, MA) at four centers and the Lunar scanner (GE Healthcare Bio-Sciences Corp, Piscataway, NJ) at five centers. Each center used a spine phantom to monitor longitudinal stability. Standard methods were used to convert Lunar data to corresponding Hologic values (18–21). The Bio-Imaging Bona Fide Phantom (Bio-Imaging Technologies, Newtown, PA) was used to calibrate densitometers at all centers and the coordinating center reanalyzed measurements from each center. Details on the BMD quality assurance/quality control program and cross-calibration were published elsewhere (22). As recommended by the World Health Organization. we calculated femoral neck T scores for both sexes by comparing each individual’s BMD to the Third National Health and Nutrition Examination Survey reference range for white women aged 20–29 years (23).
Fracture Ascertainment
Data on incident clinical fractures were collected over 5 years after cohort entry by self-report from a yearly postal questionnaire or in-person assessment (year 15 of the CaMos study) (15). Fractures were confirmed by the following: structured interview to determine further information (date, fracture location, medical treatment, and cause of fracture [i.e., fall]) and/or verification from the treating physician or hospital (15). We defined fracture as a composite of incident clinical spine, hip, forearm/wrist, and humerus fractures (major osteoporotic fractures) that resulted from low trauma.
Fracture Risk Assessment Using FRAX
We used the Canadian FRAX tool (version 3.7; FRAX Desktop Multi-Patient Entry) to calculate the 10-year probability of a major osteoporotic fracture (with and without BMD) (3). The United States and Canadian versions of FRAX are derived using identical methodology and give similar results with regard to fracture prediction (24,25). A complete list of the variables we used to calculate the FRAX score is provided in Supplemental Table 1. BMI was calculated at cohort entry by dividing weight (in kilograms) by height (in square meters). When BMI (in kilograms per square meter) was missing at year 10, we carried forward values from year 5 of the CaMos study (<0.5% missing). We defined rheumatoid arthritis as a self-report of a diagnosis of rheumatoid arthritis combined with evidence of treatment (prednisone, betamethasone, methotrexate, hydroxychloroquine, leflunomide, etanercept, infliximab, sulfasalazine, adalimumab). Prior corticosteroid use was defined as use of intravenous or oral glucocorticoids for ≥3 months from baseline to cohort entry. Previous fracture was defined as any low-trauma fracture (excluding hands, feet, head, and ankle) occurring before cohort entry. History of parental hip fracture was defined using self-report at year 5 of CaMos. All other clinical risk factors were based on self-report at cohort entry or before.
Statistical Analyses
We described continuous variables as means±SDs or medians (interquartile ranges) and categorical variables as proportions. To compare baseline characteristics between adults with an eGFR<60 versus ≥60 ml/min per 1.73 m2, we used the t test or Mann–Whitney U test for continuous variables and the chi-squared test or the Fisher’s exact tests where appropriate for categorical variables. We used area under the receiver operator characteristic curves to determine how well FRAX could discriminate between individuals with a fracture and without a fracture (null value was defined as an area under the curve [AUC] value of 0.5, which indicates that the ability of FRAX to discriminate fracture is no better than chance) (26). To assess differences in fracture discrimination between individuals with an eGFR<60 and ≥60 ml/min per 1.73 m2, we calculated mean differences (95% confidence intervals [95% CIs]) using the two-tailed z test. In an additional analysis, we assessed the predictive discrimination of FRAX (without BMD) including CKD as a cause of secondary osteoporosis in all individuals with an eGFR<60 ml/min per 1.73 m2. The rationale for this was that we wanted to capture some of the unique risk factors for fracture in patients with CKD that are currently not included in the FRAX algorithm (12). It is important to note that only FRAX without BMD can be assessed when including CKD as a secondary cause of osteoporosis because FRAX assumes that secondary causes of osteoporosis effect fracture risk through lowering BMD. We had a maximum of 5 years of follow-up. As a result, to calculate the estimated fracture risk in the cohort using FRAX, we divided the FRAX 10-year risk by 2. The 5-year observed fracture probabilities and 95% CIs were calculated using a survival analysis method that adjusts for the competing risk of death (27). To assess calibration (defined as the agreement between observed and predicted values), we compared the 5-year FRAX estimated fracture risk with the 5-year observed fracture risk. We performed all statistical analysis using the Statistical Analysis System (SAS version 9.3; SAS Institute, Cary, NC). We considered two-sided P values <0.05 as statistically significant.
Results
Baseline Characteristics
We included 320 adults with an eGFR<60 ml/min per 1.73 m2 and 1787 adults with an eGFR≥60 ml/min per 1.73 m2 (Figure 1). During follow-up, 3.3% (n=69) died (5.9% [n=19] with an eGFR<60 and 2.8% [n=50] with an eGFR≥60 ml/min per 1.73 m2) and 3.8% (n=81) were lost to follow-up (8.4% [n=27] with an eGFR<60 and 3.0% [n=54] with an eGFR≥60 ml/min per 1.73 m2). Of the adults with an eGFR<60 ml/min per 1.73 m2, 72.2% (n=231) had stage 3a CKD, 23.8% had stage 3b (n=76), and 4.0% (n=13) had stage 4 or stage 5. When comparing individuals with an eGFR<60 ml/min per 1.73 m2 with individuals with an eGFR≥60 ml/min per 1.73 m2, individuals with reduced kidney function were older (75.9 years versus 65.6 years; P<0.001), more likely to have type 2 diabetes (13.1% versus 6.6%; P<0.001), were more likely to have sustained a previous fracture (25.3% versus 17.1%; P<0.001), and were less likely to report good, very good, or excellent health (87.5% versus 93.6%; P<0.001) (Table 1). Self-reported bisphosphonate use was similar between the two groups (26.9% versus 23.5%; P=0.19).
Study cohort. †Individuals who died or were not reachable at year 11 were excluded because we would not able to obtain fracture data from these individuals. CaMos, Canadian Multicentre Osteoporosis Study.
Baseline characteristics by eGFR
Fracture Risk Prediction and Discrimination
Over an average of 4.8 years of follow-up, there were a total of 64 (3.0%) major osteoporotic fracture events (16 [5.0%] with an eGFR<60 ml/min per 1.73 m2 [2.5% stage 3a, 2.2% stage 3b, and 0.3% stage 4/5] and 48 [2.7%] with an eGFR≥60 ml/min per 1.73 m2). The AUC values for the FRAX models, femoral neck T score alone, age alone, and T score with a previous fall are presented in Table 2. We found that all AUC values were statistically significant (>0.5) regardless of renal function. The major osteoporotic fracture FRAX AUC values were higher in individuals with an eGFR≥60 ml/min per 1.73 m2 compared with individuals with an eGFR<60 ml/min per 1.73 m2. However, these differences did not reach statistical significance (Table 2). Moreover, there were no statistically significant differences in the predictive discrimination of T scores alone, age alone, and T scores with previous falls between individuals with an eGFR<60 versus ≥60 ml/min per 1.73 m2 for major osteoporotic fractures (Table 2).
Area under the curve for incident fracture prediction according to eGFR
Fracture Events and Fracture Risk Calibration
In individuals with an eGFR<60 ml/min per 1.73 m2, the observed major osteoporotic fracture risk (5.3%; 95% CI, 3.3% to 8.6%), calculated using a survival analysis method that adjusts for the competing risk of death (27), was slightly lower than the FRAX-predicted major osteoporotic fracture risk with BMD (6.4%; 95% CI, 6.0% to 6.9%) and was also slightly lower than the FRAX-predicted major osteoporotic fracture risk without BMD (8.2%; 95% CI, 7.6% to 8.7%) (Figure 2). However, the observed and FRAX-predicted fracture risks were concordant with the FRAX-predicted fracture risk within the observed fracture risk 95% CIs. In individuals with an eGFR≥60 ml/min per 1.73 m2, the observed major osteoporotic fracture risk (2.7%; 95% CI, 2.1% to 3.6%) was lower than the FRAX-predicted major osteoporotic fracture risk with BMD (4.6%; 95% CI, 4.5% to 4.8%) and lower than the FRAX-predicted major osteoporotic fracture risk without BMD (5.3%; 95% CI, 5.0% to 5.4%). When including CKD as a cause of secondary osteoporosis in individuals with an eGFR<60 ml/min per 1.73 m2, the calibration of FRAX without BMD did not improve; the FRAX-predicted risk in our cohort was 11.0% (95% CI, 10.3% to 11.7%) compared with an observed major osteoporotic fracture risk of 5.3% (95% CI, 3.3% to 8.6%).
Mean predicted 5-year fracture risk from the Canadian FRAX tool (with and without BMD) and observed 5-year major osteoporotic fracture risk (Kaplan–Meier) according to eGFR. Error bars are 95% confidence intervals. BMD, bone mineral density; FRAX, Fracture Risk Assessment Tool.
Discussion
We found that the discriminative ability of FRAX to predict major osteoporotic fractures was similar and independent of renal function. Furthermore, in individuals with an eGFR<60 ml/min per 1.73 m2, the FRAX-predicted probabilities were comparable to the observed major osteoporotic fracture probabilities. Our finding suggests that FRAX may be a valuable tool for clinicians to accurately assess fracture risk in individuals with reduced kidney function.
The AUC values in individuals with an eGFR<60 ml/min per 1.73 m2 that we found were similar, although slightly lower, to the values found in a cross-sectional study assessing the ability of FRAX to discriminate fracture status in individuals with reduced kidney function (13). Jamal et al. included individuals with an eGFR<90 ml/min per 1.73 m2 and found an AUC of 0.72 (95% CI, 0.65 to 0.78) for FRAX with BMD, whereas we found an AUC of 0.69 (95% CI, 0.54 to 0.83) (13). The AUC values in our study were also similar to average AUC values found in 11 international FRAX validation cohorts (n=230,486) performed in the general population for both FRAX with BMD (AUC 0.62) and FRAX without BMD (AUC 0.60) (2).
In individuals with an eGFR<60 ml/min per 1.73 m2, the AUC values for FRAX with BMD (0.69) and without BMD (0.65) were lower than the AUC value for age alone (0.70), which might suggest that FRAX performs no better than age alone. However, similar results have been found in studies conducted in the general population (28–30) and comparison of AUC values has been criticized as insensitive (31–33). Moreover, because of the small number of fractures in our study, we were not able to test whether these results reached statistical significance because thousands of individuals are required to test whether a statistically significant difference occurs in correlated receiver operator characteristic curves (2,34,35).
In individuals with an eGFR<60 ml/min per 1.73 m2, the observed major osteoporotic fracture risk (5.3%) and FRAX-predicted probability of major osteoporotic fracture risk were similar (6.4% with BMD and 8.2% without BMD). We found that the calibration of FRAX without BMD did not improve when adding CKD as a cause of secondary osteoporosis in individuals with an eGFR<60 ml/min per 1.73 m2; we calculated the FRAX-predicted fracture risk to be 11.0% and the observed major osteoporotic fracture risk was 5.3%. It may be that adding CKD as a cause of secondary osteoporosis does not accurately capture all of the complexities of CKD-mineral and bone disorder (12). In the future, large prospective studies that incorporate CKD specific fracture risk factors (e.g., fibroblast growth factor 23) and include more individuals with advanced CKD are needed.
Our study had some limitations. The small number of fractures limited our statistical power. Thus, we were unable to assess the prognostic value of FRAX for hip fracture alone, we were unable to compare different FRAX models (i.e., assess the performance of FRAX versus age alone), and we were unable to further stratify kidney function into additional eGFR categories. This last point is of particular clinical relevance because as eGFR decreases, the fracture rate increases, which may be largely attributable to changes in bone and mineral metabolism (8,12). Therefore, it may be valuable to assess the performance of FRAX at each stage of CKD. However, even given the small number of fracture events, all of the AUC values for major osteoporotic fracture prediction were statistically significant. The generalizability of our findings may be limited; the majority of our sample was white (≥99%) and individuals with reduced kidney function were largely community-dwelling adults who were unaware they had decreased kidney function. Therefore, these results may not be generalizable to individuals with more severe stages of CKD and diagnosed CKD-mineral and bone disorder. Moreover, we were only able to include Canadians, which may limit the generalizability of the results to different countries; because of the wide variability of fracture rates across countries, FRAX needs to be calibrated separately for each country (36). In addition, a high proportion of individuals with normal kidney function had hyperparathyroidism (>30%), which may limit generalizability to other populations. Previous research found that individuals with moderate declines in kidney function (i.e., eGFR 60–69 ml/min per 1.73 m2) are more likely to have hyperparathyroidism (>20%), which is one potential explanation (37). Moreover, many individuals in our study had low vitamin D levels (approximately 60%); as vitamin D levels decrease, parathyroid hormone levels increase (38).
In summary, FRAX was able to accurately predict fracture risk in this cohort of individuals with an eGFR<60 ml/min per 1.73 m2, which was demonstrated by the similar observed and FRAX-predicted fracture rates. Moreover, FRAX demonstrated major osteoporotic fracture predictive discrimination in individuals with an eGFR<60 ml/min per 1.73 m2, which was similar to individuals with an eGFR≥60 ml/min per 1.73 m2. Therefore, FRAX may be a useful tool for clinicians to use to assess fracture risk in patients with reduced kidney function. Prospective studies with larger sample sizes are needed before FRAX can be recommended to be used routinely for fracture risk assessment in individuals with reduced kidney function.
Disclosures
A.X.G. received an investigator-initiated grant from Astellas and Roche to support a Canadian Institutes of Health Research (CIHR) study in living kidney donors, and his institution received unrestricted research funding from Pfizer. W.D.L. has been on the speaker bureau for Amgen, Eli Lilly, and Novartis, and has had research grants for Novartis, Amgen, and Genzyme. L.-A.F. has been on the speaker bureau for Amgen. CaMos was funded by the CIHR, Merck Frosst Canada Ltd, Eli Lilly Canada Inc, Novartis Pharmaceuticals Inc, The Alliance: sanofi-aventis and Procter and Gamble Pharmaceuticals Canada Inc, Servier Canada Inc, Amgen Canada Inc, the Dairy Farmers of Canada, and the Arthritis Society.
Acknowledgments
We thank all of the Canadian Multicentre Osteoporosis Study (CaMos) participants who made this study possible, as well as members of the CaMos research group, who were instrumental in the ongoing success of the CaMos cohort.
K.L.N. is supported by a Osteoporosis Canada–CaMos Fellowship Award.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
This article contains supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.06040614/-/DCSupplemental.
- Received June 17, 2014.
- Accepted January 5, 2015.
- Copyright © 2015 by the American Society of Nephrology
References
