- Email: [email protected]
The Fracture Risk Assessment Tool (FRAX®) predicts fracture risk in patients with chronic kidney disease
clinical investigation
www.kidney-international.org
The Fracture Risk Assessment Tool (FRAX) predicts fracture risk in patients with chronic kidney disease Reid H. Whitlock1,2, William D. Leslie1, James Shaw1, Claudio Rigatto1,2, Laurel Thorlacius1, Paul Komenda1,2, David Collister1, John A. Kanis3,4 and Navdeep Tangri1,2 1
Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, Canada; 2Chronic Disease Innovation Centre, Seven Oaks General Hospital, Winnipeg, Canada; 3Center for Metabolic Bone Diseases, University of Sheffield, Sheffield, UK; and 4Institute for Health and Aging, Catholic University of Australia, Melbourne, Australia
The Fracture Risk Assessment Tool (FRAX) was developed to predict fracture risk in the general population, but its applicability to patients with chronic kidney disease (CKD) is unknown. Using the Manitoba Bone Mineral Density (BMD) Database, we identified adults not receiving dialysis with available serum creatinine measurements and bone densitometry within 1 year. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation. Incident major osteoporotic fractures and hip fractures were ascertained from population-based health care databases. The performance of FRAX, derived without and with BMD, was studied in relation to CKD stage. Among 10,099 subjects (mean age 64 ± 13 years, 13.0% male), 2,154 had eGFR 30-60 mL/min/1.73 m2 (CKD stage 3) and 590 had eGFR <30 mL/min/1.73 m2 (CKD stages 4-5). During a 5-year observation period, 772 individuals experienced a major osteoporotic fracture and 226 had a hip fracture. FRAX predicted risk for major osteoporotic fracture and hip fracture in all eGFR strata. For every standard deviation increase in FRAX score derived with BMD, the hazard ratio (HR) for hip fracture was 4.54 (95% confidence interval [CI] 3.57-5.77) in individuals with eGFR ‡ 60 mL/min/1.73m2, 4.52 (95% CI 3.15-6.49) in individuals with eGFR 30-60 mL/min/1.73m2, and 3.10 (95% CI 1.80-5.33) in individuals with eGFR <30 mL/min/1.73m2. The relationship between FRAX and major osteoporotic fracture was stronger in those with CKD compared to those with preserved eGFR. These findings support the use of FRAX to risk stratify patients with non-dialysis CKD for major osteoporotic fractures and hip fractures. Kidney International (2019) 95, 447–454; https://doi.org/10.1016/ j.kint.2018.09.022 KEYWORDS: bone mineral density; chronic kidney disease; fracture; fracture discrimination; FRAX Copyright ª 2018, International Society of Nephrology. Published by Elsevier Inc. All rights reserved.
O
steoporotic fractures are associated with morbidity, mortality, and high health care costs.1,2 In elderly patients, osteoporotic fractures can result in major functional decline3–7 and lead to a decreased health-related quality of life and early mortality.8–11 At nearly 9 million estimated incident fractures per year, the global burden of osteoporotic fractures is high, and expected to increase in the future with an aging population.12–14 Chronic kidney disease (CKD) is also more common in elderly patients and is associated with increased mortality and morbidity.15,16 For patients over age 65 years, the global prevalence of CKD is approximately 30%, which is 3 times more common than in younger adults.17,18 It follows that CKD and osteoporotic fractures are often co-prevalent in the elderly,18,19 but whether CKD affects fracture risk assessment has not been fully elucidated. In the general population, an individual’s 10-year fracture risk can be estimated using the Fracture Risk Assessment Tool (FRAX). FRAX is widely used in many countries and has been incorporated into numerous guidelines for diagnosis and management of osteoporosis.20,21 Eleven variables plus an optional bone mineral density (BMD) measurement obtained at the femoral neck are used to calculate the probability of fracture, which can then be used to determine the need for preventative therapy. A unique feature of FRAX is that it considers competing mortality in the fracture risk estimation procedure. Notably, the presence of kidney disease is not part of the FRAX risk assessment algorithm. Although BMD is lower in patients with advanced CKD,22 the ability of FRAX to accurately predict fracture risk in patients with CKD is unknown, and existing studies have been limited by their small sample size and number of fracture events, particularly in patients with severely reduced kidney function.23–26 Our primary objective was to determine whether FRAX, derived with or without BMD measurement, predicts incident fracture events in patients with CKD.
Correspondence: Navdeep Tangri, Seven Oaks General Hospital, 2LB10-2300 McPhillips Street, Winnipeg MB R2V 3M3, Canada. E-mail: [email protected]
RESULTS
Received 4 May 2018; revised 13 September 2018; accepted 20 September 2018; published online 20 December 2018
Baseline characteristics and observed events of the cohort are described in Table 1. A total of 10,099 individuals (mean age
Kidney International (2019) 95, 447–454
447
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
64 13 years, 13.0% male) underwent both a dual-energy Xray absorptiometry (DXA) examination and a serum creatinine test within 12 months (Figure 1). The median duration between tests was 6 months, and the interquartile range was 3 to 9 months. Among these, 7355 (72.8%) had an estimated glomerular filtration rate (eGFR) $ 60 ml/min per 1.73 m2, 2154 (21.3%) had an eGFR between 30 and 60 ml/min per 1.73 m2 and 590 (5.8%) had an eGFR < 30 ml/min per 1.73 m2 (380 with eGFR 15–29 ml/min per 1.73 m2, 210 with eGFR < 15 ml/min per 1.73 m2). Compared with individuals with eGFR $ 60 ml/min per 1.73 m2, those subjects with an eGFR < 60 ml/min per 1.73 m2 tended to have a higher risk of fracture and were more likely to be older, male, diabetic, overweight, and have increased albuminuria, lower femoral neck T-scores, and increased mortality. There were 226 (2.2%) incident hip fracture events and 772 individuals (7.6%) with 1 or more major osteoporotic fracture (MOF) events (231 forearm, 206 hip, 180 vertebral, and 158 humerus) over a mean follow-up of 4.6 1.1 years, with higher fracture rates in individuals with lower eGFR. Subjects in the eGFR $ 60 ml/min per 1.73 m2 stratum were significantly less likely to experience a hip fracture or MOF compared with those in the eGFR 30 to 60 ml/min per 1.73 m2 (P < 0.001) or eGFR < 30 ml/min per 1.73 m2 (P < 0.001) groups during the observation period (Figure 2).
FRAX as a predictor of hip and MOF fractures
The effect of a higher FRAX score (with and without BMD measurement) and BMD alone on the risk of hip fracture and MOF in patients with or without reduced kidney function is presented in Table 2. For every SD increase in FRAX computed with BMD, the hazard ratio (HR) for hip fracture was 4.54 in individuals with an eGFR $ 60 ml/min per 1.73 m2 (95% CI: 3.57–5.77, P < 0.001), 4.52 in individuals with an eGFR of 30 to 60 ml/min per 1.73 m2 (95% CI: 3.15–6.49, P < 0.001), and 3.10 in individuals with an eGFR < 30 ml/ min per 1.73 m2 (95% CI: 3.10–5.33, P < 0.001). The interaction between hip fracture probability and eGFR was not significant for FRAX with BMD (P ¼ 0.48) and without BMD (P ¼ 0.23). For MOF probability with BMD, the HR was 1.83 in individuals with an eGFR $ 60 ml/min per 1.73 m2 (95% CI: 1.67–2.01, P < 0.001), 2.33 in individuals with an eGFR of 30 to 60 ml/min per 1.73 m2 (95% CI: 1.96–2.77, P < 0.001), and 2.04 in individuals with an eGFR < 30 ml/min per 1.73 m2 (95% CI: 1.58–2.84, P < 0.001). The interaction of MOF probability and eGFR was significant both with and without BMD, with stronger associations of FRAX in patients with CKD than in those without (P < 0.001). Both FRAX without BMD and BMD alone were also significantly associated with fracture risk irrespective of
Table 1 | Characteristics of study population
Baseline Sex (male) Age Body mass index (kg/m2) Rheumatoid arthritis COPD Alcohol or substance abuse diagnosis Recent glucocorticoid use Prior fracture Parental hip fracture Femoral neck T-score FRAX predicted probability of hip fracture (without BMD) FRAX predicted probability of hip fracture (with BMD) FRAX predicted probability of MOF (without BMD) FRAX predicted probability of MOF (with BMD) eGFR (ml/min per 1.73 m2) Urine ACR (mg/mmol)a Diabetes Observed Events Hip fracture Rate of hip fractures (per 1000 person-yr) MOF Rate of MOF (per 1000 person-yr) Mortality Lost to follow-up
All (n [ 10 099)
eGFR ‡ 60 (n [ 7 355)
eGFR 30–60 (n [ 2 154)
eGFR < 30 (n [ 590)
P value
1308 (13.0%) 64.4 13.3 27.3 5.7 579 (5.7%) 1013 (10.0%) 303 (3.0%) 1193 (11.8%) 1768 (17.5%) 1121 (11.1%) –1.3 1.0 3.9 6.0 2.9 5.0 11.7 9.4 10.8 8.2 73 24 1.5 (0.6–5.3) 2125 (21.0%)
794 (10.8%) 62.2 12.7 27.1 5.6 455 (6.2%) 674 (9.2%) 247 (3.4%) 652 (8.9%) 1258 (17.1%) 837 (11.4%) –1.3 1.0 3.1 5.4 2.4 4.4 10.5 8.6 9.8 7.4 84 15 1.2 (0.6–3.1) 1262 (17.2%)
380 (17.6%) 70.9 12.6 27.7 5.8 97 (4.5%) 253 (11.7%) 40 (1.9%) 384 (17.8%) 399 (18.5%) 231 (10.7%) –1.5 1.1 6.1 7.0 4.2 5.7 15.5 10.4 13.4 9.0 47 8 2.1 (0.6–7.6) 605 (28.1%)
134 (22.7%) 67.0 15.3 28.0 6.0 27 (4.6%) 86 (14.6%) 16 (2.7%) 157 (26.6%) 111 (18.8%) 53 (9.0%) –1.6 1.1 5.4 7.2 4.5 7.1 13.9 10.9 13.4 10.5 18 8 11.5 (1.9–79.5) 258 (43.7%)
<0.001 <0.001 <0.001 0.006 <0.001 0.001 <0.001 0.22 0.168 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
226 (2.2%) 4.57 772 (7.6%) 16.10 1205 (11.9%) 201 (2.0%)
122 (1.7%) 3.38 514 (7.0%) 14.68 592 (8.0%) 158 (2.1%)
81 (3.8%) 7.72 198 (9.2%) 19.47 402 (18.7%) 31 (1.4%)
23 (3.9%) 8.01 60 (10.2%) 21.76 211 (35.8%) 12 (2.0%)
<0.001 <0.001 <0.001 <0.001 <0.001 0.117
ACR, albumin-to-creatinine ratio; BMD, bone mineral density; COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate; FRAX, Fracture Risk Assessment Tool; MOF, major osteoporotic fracture. a Urine ACR obtained for a total of 1210 (11.9%) patients: 727 (9.9%) for eGFR $ 60, 426 (19.5%) for eGFR of 30 to 60, and 147 (23.8%) for eGFR < 30 ml/min per 1.73 m2. Median and interquartile range are presented.
448
Kidney International (2019) 95, 447–454
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
Paents with at least 1 DXA during the study period (n = 42,551)
Exclusions: • No SCr measurement (n = 32,188) • < 20 years of age (n = 150) • On RRT at baseline (n = 114)
Paents included in the cohort for analysis (n = 10,099)
eGFR ≥ 60 ml/min per 1.73 m2 (n = 7,355)
eGFR 30 — 60 ml/min per 1.73 m2
(n = 2,154)
eGFR < 30 ml/min per 1.73 m2 (n = 590)
Figure 1 | Cohort selection. DXA, dual-energy X-ray absorptiometry; eGFR, estimated glomerular filtration rate; RRT, renal replacement therapy (dialysis or transplant); SCr, serum creatinine.
kidney function. However, the magnitude of the hazard ratios as well as model fit and calibration statistics (Brier scores and Akaike Information Criterion) for FRAX with BMD were superior in every eGFR group for both hip fractures and MOFs (Supplementary Table S1). In our cohort, the observed 5-year cumulative probability of incident hip fracture and MOF risk was slightly but consistently higher than the predicted fracture risk across all eGFR strata (Figures 3 and 4). For MOF risk, the observed 5year cumulative fracture percentage was on average approximately 2.5% higher than the predicted fracture risk. For hip fractures, the observed 5-year cumulative fracture percentage was on average approximately 1% higher than the predicted fracture risk (Figure 5). Supplementary analyses
The association between FRAX and MOF or hip fracture, or the interaction between FRAX, eGFR, and fractures was unchanged in the subgroup of participants older than 40 years of age and also when adjusting for bisphosphonate use. FRAX with BMD showed very good discrimination for predicting hip fractures (c-statistic ¼ 0.82) and moderate discrimination for predicting MOF (c-statistic ¼ 0.68) (Supplementary Table S2). After adjusting for clinical risk factors, BMD, and diabetes, eGFR was not independently associated with MOF (HR per 10 ml/min per 1.73 m2 decrease: 1.01, 95% CI: 0.98–1.05, P ¼ 0.39) (Supplementary Table S3). Regarding the
Kidney International (2019) 95, 447–454
association with hip fractures, once adjusted for clinical risk factors, BMD, and diabetes, we found that eGFR was weakly associated with hip fracture (HR per 10 ml/min per 1.73 m2 decrease: 1.07, 95% CI: 1.00–1.14, P ¼ 0.035). Sex was not associated with hip or MOF in these models. Urine albumin-to-creatinine ratio (ACR) measurements were available in 1206 individuals. Of these, 85 MOF and 19 hip fractures were observed during the follow-up period. Urine ACR was not associated with either MOF (P ¼ 0.95) or hip fracture (P ¼ 0.53). DISCUSSION
In this population-based study of more than 10,000 individuals, we found that the FRAX score with or without BMD measurement significantly discriminated fracture risk in patients with nondialysis CKD. Risk for hip fracture did not appear to be influenced by CKD stage, and although FRAX was more strongly predictive of MOF in patients with reduced eGFR, the differences in the hazard ratios between eGFR groups were small, and no linear trend was evident. These findings support the use of FRAX to identify patients with moderate to severe CKD who are at high risk for fracture events, similar to its use in the general population. To our knowledge, this is the largest study to examine fracture risk in a population with advanced CKD using DXA assessment and the FRAX tool. Our assertion that FRAX is valid in patients with CKD is consistent with other smaller studies in the literature. In 2014 Jamal et al.24 used a
449
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
1
FRACTION WITHOUT MOF
0.97
0.94
0.91
eGFR ≥ 60 eGFR 30–60
0.88
Log-rank test P value Comparison eGFR 30–60 vs. eGFR ≥ 60 < 0.001 eGFR < 30 vs. eGFR ≥ 60 < 0.001 eGFR < 30 vs. eGFR 30—60 0.53
eGFR < 30
0.85 0
1
3
2
4
5
YEARS
Figure 2 | Kaplan-Meier curve: time to major osteoporotic fracture. eGFR, estimated glomerular filtration rate; MOF, major osteoporotic fracture.
cross-sectional convenience sample of 353 CKD patients with mean eGFR of 28 ml/min per 1.73 m2 to demonstrate that FRAX was able to discriminate prevalent fractures in the CKD population, but offered no additional discrimination compared with BMD measurement. Because of the crosssectional nature of the study, the ability of FRAX to predict incident fracture risk was not assessed. More recently, a study by Naylor et al.26 used a larger Canadian prospective cohort stratified by eGFR to show that FRAX could accurately predict MOF in patients with reduced kidney function. However, of the 2107 patients in the study, only 320 had an eGFR < 60 ml/min per 1.73 m2, and only 13 had eGFR < 30 ml/min per 1.73 m2. This small sample size limited their ability to assess FRAX prediction for fracture in those with CKD. Notably, they showed that FRAX calibration was not improved when CKD was added as a secondary cause of osteoporosis in those patients with eGFR < 60 ml/min per 1.73 m2. Our data, based on a much larger sample size with adequate representation of patients with eGFR of 30 to 60 and < 30 ml/min per 1.73 m2, clearly demonstrated that FRAX can predict fracture risk in patients with CKD. Previous studies have examined whether CKD, diminished eGFR, or albuminuria are independent risk factors for
fracture, and found conflicting results. Investigators from the Cardiovascular Health Study measured cystatin C in more than 4000 older adults, and found that eGFR based on cystatin C was associated with a higher risk of incident fractures. Although the investigators adjusted for several possible confounders, they did not have access to BMD measurements and did not adjust for all possible FRAX risk factors.27 In contrast, investigators from the Alberta Kidney Disease Network studied nearly 1.8 million individuals and found that age- and sex-adjusted rates of osteoporotic and hip fracture did not differ in individuals with CKD versus those without.23 While the Alberta study included nearly 130,000 individuals with CKD, no data on BMD or other FRAX risk factors was available. A study of 2982 individuals from the Osteoporotic Fractures in Men cohort found no association between urine albumin levels and risk of incident clinical fracture.28 The investigators did adjust for confounders and had access to BMD measurements, but no women were included in the study. Our findings suggest that a lower eGFR or presence of albuminuria is not associated with MOF risk after adjustment for BMD and FRAX clinical risk factors. There was a weak association between eGFR and hip fracture risk, but this was attenuated after adjustment for diabetes, a leading cause of
Table 2 | Gradient of risk (per SD increase) for incident fractures in groups stratified by kidney function eGFR ‡ 60 (n [ 7 355)
HR (95% CI) Probability of Probability of BMD alonea Probability of Probability of BMD alonea
hip fracture from FRAX with BMD hip fracture from FRAX without BMD MOF from FRAX with BMD MOF from FRAX without BMD
4.54 3.55 3.41 1.83 1.68 1.75
(3.57–5.77) (2.83–4.45) (2.79–4.18) (1.67–2.01) (1.52–1.84) (1.60–1.92)
eGFR 30–60 (n [ 2 154) 4.52 4.34 2.85 2.33 2.31 2.00
(3.15–6.49) (2.95–6.38) (2.20–3.69) (1.96–2.77) (1.91–2.80) (1.71–2.34)
eGFR < 30 (n [ 590) 3.10 1.98 2.48 2.04 1.87 1.87
(1.80–5.33) (1.25–3.15) (1.62–3.78) (1.58–2.84) (1.42–2.46) (1.44–2.42)
Interaction eGFR*FRAX P ¼ 0.47 P ¼ 0.23 P ¼ 0.109 P < 0.001 P < 0.001 P ¼ 0.048
BMD, bone mineral density; CI, confidence interval; eGFR, estimated glomerular filtration rate; FRAX, Fracture Risk Assessment Tool; HR, hazard ratio; MOF, major osteoporotic fracture. a BMD is measured by femoral neck T-score, and its HR is reported per SD decrease.
450
Kidney International (2019) 95, 447–454
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
Figure 3 | Kaplan-Meier curve: time to hip fracture. eGFR, estimated glomerular filtration rate.
CKD. Whether CKD is truly an independent risk factor for fracture remains to be fully understood, and our study calls for better recognition and classification of fracture risk in patients with CKD.29 There are important clinical and research implications of our work. We believe that physicians can consider FRAX scores valid in patients with nondialysis CKD and use them to inform treatment decisions. Bisphosphonates are a frequently used therapy for the general population with osteoporosis and have been shown to be effective in fracture risk reduction.30 This class of medication is not well characterized in CKD and generally not recommended once eGFR is <30 ml/min per 1.73 m2; however, early research into the use of these and other medications for fracture prevention in patients with CKD has shown potential benefits.31–33 Given the morbidity from fractures, and the older CKD population, wider use of
risk assessment tools like FRAX and therapies including bisphosphonates may have a beneficial clinical impact. Although the c-statistics for FRAX with or without BMD were similar to BMD alone in our study population, it is important to note that c-statistics are relatively insensitive to the inclusion of additional risk factors and should not discourage efforts to identify newer biological and/or radiological markers of fracture risk that can markedly improve discrimination and calibration. The strengths of our study include the use of large, robust regional databases that capture all bone densitometry and nearly all fracture events, and include a large number of patients with reduced kidney function. Only 200 patients (2.0%) migrated from Manitoba before the end of the study period, resulting in minimal loss to follow-up. The available BMD and FRAX scores from nearly 600 patients with advanced
15.0%
Fracture percentage
12.0%
9.0%
6.0% 10.3%
9.3% 7.1%
3.0%
6. 7% 4.9%
7.7%
6.7%
6.9%
5.2%
0.0% eGFR ≥ 60 Observed 5-year MOF
eGFR 30 –60
eGFR < 30
Predicted 5-year hip fracture MOF (with BMD)
Predicted 5-year hip fracture MOF (without BMD)
Figure 4 | Observed versus predicted 5-year major osteoporotic fracture probability. BMD, bone mineral density; eGFR, estimated glomerular filtration rate; MOF, major osteoporotic fracture. Note: Error bars represent 95% confidence intervals. Kidney International (2019) 95, 447–454
451
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
10.0%
Fracture percentage
8.0%
6.0%
4.0%
2.0%
3.9%
3.8% 1.7%
1.2%
1.6%
3.0%
2.7%
2.1%
2.2%
eGFR 30 – 60
eGFR < 30
0.0% eGFR ≥ 60 Observed 5-year hip fractures
Predicted 5-year hip fracture FRAX (with BMD)
Predicted 5-year hip fracture FRAX (without BMD)
Figure 5 | Observed versus predicted 5-year hip fracture probability. BMD, bone mineral density; eGFR, estimated glomerular filtration rate. Note: Error bars represent 95% confidence intervals.
CKD (eGFR < 30 ml/min per 1.73 m2) is a unique feature of this study, and extends our confidence in FRAX risk assessment to all patients with more advanced CKD. Our study also has limitations. Only adult Canadians who had both serum creatinine and BMD measured within 1 year were included, and as a consequence our cohort was predominantly female. As such, the applicability of our findings to male patients with CKD and to those living in other regions is unknown and needs further validation. Our classification of subjects into eGFR categories was based upon a single serum creatinine, which may not represent the subject’s true baseline renal function. However, measurement error should bias results toward the null; therefore our results are likely to be conservative. In addition, although the discrimination of the FRAX score in our population was similar (cstatistics within 0.01) to what was reported for hip and MOF events in the general population,34 calibration may require adjustment because the observed incidence of hip and MOF events was slightly higher than predicted across all eGFR strata. However, this may not be due to CKD, and may reflect characteristics intrinsic to our cohort, who may have had a higher fracture risk compared with the general population based on our selection criteria, as it was also seen in those with normal eGFR. In other populations, such as patients with type 2 diabetes where FRAX has underestimated risk, inclusion of rheumatoid arthritis as a proxy for type 2 diabetes has been proposed to improve calibration.35,36 Whether a similar substitution would be beneficial for patients with CKD is unknown. We also cannot rule out that confounding by unmeasured variables may have affected fracture risk: for example, we did not account for race, hyperparathyroidism, vitamin D deficiency, or other changes of renal osteodystrophy, as these variables were not routinely available in our data sets. Finally, although the FRAX score was initially designed to evaluate 10-year fracture risk, we have used half 452
the 10-year risk to estimate 5-year risk, which is a valid approach based on previous literature.26,37 In summary, our findings support the use of BMD testing and use of the FRAX tool to determine fracture risk in patients with nondialysis CKD. Additional studies synthesizing the evidence and comparing the effectiveness of newer agents for fracture risk reduction in patients with CKD are needed to inform treatment decisions. METHODS Study design and population We conducted a retrospective cohort study that used anonymously linked administrative health databases from Manitoba, a Canadian province with a population of w1.3 million, where all residents are covered through a single-payer universal health care system. All databases used in the analysis were from the Data Repository housed at the Manitoba Centre for Health Policy at the University of Manitoba. The study was approved by the Health Research Ethics Board for the University of Manitoba, and data access was granted by the Health Information Privacy Committee, Diagnostic Services of Manitoba, and the Winnipeg Regional Health Authority. Our cohort consisted of all individuals age 20 years and older who underwent both a DXA examination and a serum creatinine test within 12 months between the years 2005 and 2010. The DXA appointment date was used as the index date. Individuals in the cohort were stratified into 3 groups according to their eGFR as calculated by the Chronic Kidney Disease Epidemiology Collaboration equation38: eGFR $ 60 ml/min per 1.73 m2 (normal kidney function or CKD stage 1 and 2), eGFR of 30 to 60 ml/min per 1.73 m2 (CKD stage 3), and eGFR < 30 ml/min per 1.73 m2 (CKD stages 4 and 5). The lowest eGFR value within 12 months of the index date was used. Serum creatinine was obtained through the Diagnostic Services of Manitoba laboratory database, which serves >70% of the province. Patients on dialysis at the time of the index date were excluded from the analysis because the sample size was too low to yield enough incident fracture events. Patients known to have a renal transplant were also excluded.
Kidney International (2019) 95, 447–454
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
Clinical risk factor collection Results from the DXA examination were captured in the Manitoba BMD database. In addition to the BMD measurements obtained by DXA, the Manitoba BMD database also captured clinical risk factors for fracture. Height and weight were measured with a wall-mounted stadiometer and scale at the time of DXA appointment, and were used to calculate body mass index. Parental hip fracture was selfreported. Age and sex were captured from the Manitoba Health Insurance Registry. Recent glucocorticoid use (>90 days dispensed in the year prior to DXA testing) was obtained through the provincewide Drug Prescription Information Network. A diagnosis of rheumatoid arthritis, prior fracture, alcohol and/or substance abuse, or chronic obstructive pulmonary disease (the latter two being proxies for high alcohol intake and smoking, respectively) were obtained through hospital abstracts and physician claims databases using methods that have been previously described.34 Probability of fracture Ten-year probability of an MOF and hip fracture were calculated for each subject by the Canadian FRAX tool (version 3.11, FRAX Desktop Multi-Patient Entry) using the clinical risk factors described above and femoral neck BMD expressed as a T-score (the number of SDs above or below young adult mean BMD based on white female reference data from the National Health and Nutrition Examination Survey III survey).39 The Canadian FRAX tool has been independently validated for fracture risk prediction in our population.34,40 Outcomes of interest The presence of hip, clinical vertebral, forearm, or humerus fracture codes (collectively designated as MOFs) was determined through a combination of hospital discharge abstracts and physician billing claims using validated algorithms.41,42 A hip fracture was identified if it was the primary diagnosis code for a hospitalization. A vertebral fracture was identified if it was the primary diagnosis code for a hospitalization or there was a single physician visit. For forearm and humerus fractures, at least 1 hospitalization (with a primary diagnosis code) or 2 separate physician visits (within 90 days) were required. Incident fractures were ascertained for up to 5 years after the index date. In order to exclude possible recurrent fractures, we required that a fracture of the same type did not occur in the previous 6 months. To increase accuracy for an acute event, the study required hip and forearm fractures to be accompanied by a sitespecific procedural code. Fractures associated with high trauma codes were excluded.
Additionally, FRAX was modeled for the entire cohort alongside eGFR as a continuous variable. In these models we tested for an eGFR*FRAX interaction to examine whether eGFR modified the effect of FRAX on the risk of fracture. Hip fractures and MOF outcomes were modeled separately. Patients were censored in the analysis at the time of death or if they emigrated from Manitoba as determined by the Manitoba Health Insurance Registry. Loss to follow-up due to migration was less than 2% and did not differ between groups. There were no missing data. The observed 5-year cumulative fracture probability was derived using a modified Kaplan-Meier method with competing mortality framework whereby individuals who died before experiencing a fracture event were not censored at the time of death and instead were assigned a follow-up time of 5 years.43 This approach is equivalent to other competing risk approaches (e.g., that of Satagopan et al.).44 The observed values among kidney disease groups were compared with the log-rank test. The 5-year predicted fracture probability (one-half of the 10-year probability) was compared with observed values to assess FRAX calibration.37 Supplementary analyses We also conducted several subgroup and sensitivity analyses. First, we excluded individuals younger than 40 years of age, and examined hazard ratios for differences and P values for interaction between the FRAX score and the observed outcomes of hip and MOF. Second, we adjusted each model for past bisphosphonate use (at least 90 days in the year before the index date) and examined for the same differences. Third, we also calculated concordance statistics (c- statistics) among kidney disease groups as an alternative measure of the ability of FRAX and BMD alone to discriminate fracture risk. Furthermore, we examined the association of measures of CKD and fracture risk. Cox proportional hazards models of eGFR as a predictor of time to fracture were modeled univariately and in the presence of the FRAX clinical risk factors, femoral neck BMD, and diabetes. Diabetes was included as it is the primary cause of CKD in the majority of individuals, and has been shown to be independently associated with fracture risk.45–47 A diagnosis of diabetes was captured through hospital discharge abstracts, physician claims, or a prescription of insulin or anti-diabetes medication. Cox proportional hazards models of urine ACR as a predictor of fracture were modeled univariately. Urine ACR was captured through the Diagnostic Services of Manitoba database and was tested by using the highest value within 12 months of the index date. Both eGFR and urine ACR were modeled as continuous variables. DISCLOSURE
Data analysis All analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, NC). Descriptive statistics consisted of baseline demographic information, clinical risk factors, FRAX scores, and outcomes of interest. Continuous variables were presented with mean and SD and were compared using analysis of variance or Mann-Whitney U Test. Categorical variables were presented as frequency and percentage and were compared with the chi-square test. Cox proportional hazards models were used to test for associations between FRAX score and the outcomes of interest (incident fracture within 5 years). FRAX score (with and without BMD) was log-transformed and modeled in all 3 stratified groups with the hazard ratios presented per SD increase. Model fit statistics and calibration were assessed using the Akaike Information Criterion and the Brier Score. Kidney International (2019) 95, 447–454
Our study was funded by Diagnostic Services Manitoba (DSM). DSM had no role in the study design, collection, analysis and interpretation of data, writing the report, or decision to submit the report for publication. NT has received honoraria or research support from AstraZeneca, Otsuka Inc., and Tricida Inc. All the other authors declared no competing interests. ACKNOWLEDGMENTS
This work was supported through funding provided by the Department of Health of the Province of Manitoba to the University of Manitoba (HIPC no. 2015/2016-12). The results and conclusions are those of the authors, and no official endorsement by Manitoba Health was intended or should be inferred. Data used in this study are from the Population Health Research Data Repository housed at the Manitoba Centre for Health Policy, University of Manitoba, and were derived from data provided by Manitoba Health, Winnipeg Regional 453
clinical investigation
Health Authority, and Diagnostic Services Manitoba. This article has been reviewed and approved by the members of the Manitoba Bone Density Program Committee. SUPPLEMENTARY MATERIAL Table S1. Akaike Information Criterion and Brier scores for Fracture Risk Assessment Tool (FRAX) with bone mineral density (BMD) measurement and BMD alone as a predictor of fracture. Table S2. C-statistics for Fracture Risk Assessment Tool (FRAX) as a predictor of fracture within 5 years. Table S3. Hazard ratios and 95% confidence intervals for estimated glomerular filtration rate as a risk factor for fracture within 5 years. Supplementary material is linked to the online version of the paper at www.kidney-international.org. REFERENCES 1. Johnell O, Kanis JA. An estimate of the worldwide prevalence and disability associated with osteoporotic fractures. Osteoporosis Int. 2006;17:1726–1733. 2. Hopkins RB, Burke N, Von Keyserlingk C, et al. The current economic burden of illness of osteoporosis in Canada. Osteoporosis Int. 2016;27: 3023–3032. 3. Center JR, Nguyen TV, Schneider D, et al. Mortality after all major types of osteoporotic fracture in men and women: an observational study. Lancet. 1999;353:878–882. 4. Ioannidis G, Papaioannou A, Hopman WM, et al. Relation between fractures and mortality: results from the Canadian Multicentre Osteoporosis Study. Can Med Assoc J. 2009;181:265–271. 5. Morin S, Lix LM, Azimaee M, et al. Mortality rates after incident nontraumatic fractures in older men and women. Osteoporos Int. 2011;22: 2439–2448. 6. Tosteson AN, Gottlieb DJ, Radley DC, et al. Excess mortality following hip fracture: the role of underlying health status. Osteoporos Int. 2007;18: 1463–1472. 7. Johnell O, Kanis JA, Oden A, et al. Mortality after osteoporotic fractures. Osteoporos Int. 2004;15:38–42. 8. Hallberg I, Rosenqvist AM, Kartous L, et al. Health-related quality of life after osteoporotic fractures. Osteoporosis Int. 2004;15:834–841. 9. Tarride J, Burke N, Leslie WD, et al. Loss of health related quality of life following low-trauma fractures in the elderly. BMC Geriatrics. 2016;16:84. 10. Al-Sari UA, Tobias J, Clark E. Health-related quality of life in older people with osteoporotic vertebral fractures: a systematic review and metaanalysis. Osteoporos Int. 2016;27:2891–2900. 11. Borgstrom F, Lekander I, Ivergard M, et al. The International Costs and Utilities Related to Osteoporotic Fractures Study (ICUROS)–quality of life during the first 4 months after fracture. Osteoporos Int. 2013;24:811–823. 12. Oden A, McCloskey EV, Kanis JA, et al. Burden of high fracture probability worldwide: secular increases 2010-2040. Osteoporos Int. 2015;26:2243– 2248. 13. Gullberg B, Johnell O, Kanis JA. World-wide projections for hip fracture. Osteoporosis Int. 1997;7:407–413. 14. Schneider EL, Guralnik JM. The aging of America. Impact on health care costs. JAMA. 1990;263:2335–2340. 15. Go AS, Chertow GM, Fan D, et al. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med. 2004;351: 1296–1305. 16. Tonelli M, Wiebe N, Culleton B, et al. Chronic kidney disease and mortality risk: a systematic review. J Am Soc Nephrol. 2006;17:2034–2047. 17. Arora P, Vasa P, Brenner D, et al. Prevalence estimates of chronic kidney disease in Canada: results of a nationally representative survey. Can Med Assoc J. 2013;185:E417–E423. 18. Hill NR, Fatoba ST, Oke JL, et al. Global prevalence of chronic kidney disease - a systematic review and meta-analysis. PLoS One. 2016;11, e0158765. 19. Klawansky S, Komaroff E, Cavanaugh PF, et al. Relationship between age, renal function and bone mineral density in the US population. Osteoporosis Int. 2003;14:570–576. 20. Kanis JA, Harvey NC, Cooper C, et al. A systematic review of intervention thresholds based on FRAX: a report prepared for the National Osteoporosis Guideline Group and the International Osteoporosis Foundation. Arch Osteoporos. 2016;11:25.
454
RH Whitlock et al.: Utility of FRAX in CKD
21. Kanis JA, Johansson H, Oden A, et al. Worldwide uptake of FRAX. Arch Osteoporos. 2014;9:166. 22. Bucur RC, Panjwani DD, Turner L, et al. Low bone mineral density and fractures in stages 3-5 CKDCKD: an updated systematic review and metaanalysis. Osteoporosis Int. 2015;26:449–458. 23. Elliott MJ, James MT, Quinn RR, et al. Estimated GFR and fracture risk: a population-based study. Clin J Am Soc Nephrol. 2013;8:1367–1376. 24. Jamal SA, West SL, Nickolas TL. The clinical utility of FRAX to discriminate fracture status in men and women with chronic kidney disease. Osteoporos Int. 2014;25:71–76. 25. Naylor KL, McArthur E, Leslie WD, et al. The three-year incidence of fracture in chronic kidney disease. Kidney Int. 2014;86:810–818. 26. Naylor KL, Garg AX, Zou G, et al. Comparison of fracture risk prediction among individuals with reduced and normal kidney function. Clin J Am Soc Nephrol. 2015;10:646–653. 27. Fried LF, Biggs ML, Shlipak MG, et al. Association of kidney function with incident hip fracture in older adults. J Am Soc Nephrol. 2007;18: 282–286. 28. Fink HA, Vo TN, Langsetmo L, et al. Association of increased urinary albumin with risk of incident clinical fracture and rate of hip bone loss: the Osteoporotic Fractures in Men Study. J Bone Miner Res. 2017;32: 1090–1099. 29. Moe SM. Renal osteodystrophy or kidney-induced osteoporosis? Curr Osteoporos Rep. 2017;15:194–197. 30. Crandall CJ, Newberry SJ, Diamant A, et al. Comparative effectiveness of pharmacologic treatments to prevent fractures: an updated systematic review. Ann Intern Med. 2014;161:711. 31. Miller PD, Roux C, Boonen S, et al. Safety and efficacy of risedronate in patients with age-related reduced renal function as estimated by the Cockcroft and Gault method: a pooled analysis of nine clinical trials. J Bone Miner Res. 2005;20:2105–2115. 32. Jamal SA, Bauer DC, Ensrud KE, et al. Alendronate treatment in women with normal to severely impaired renal function: an analysis of the Fracture Intervention Trial. J Bone Miner Res. 2007;22:503–508. 33. Wilson LM, Rebholz CM, Jirru E, et al. Benefits and harms of osteoporosis medications in patients with chronic kidney disease: a systematic review and meta-analysis. Ann Intern Med. 2017;166:649–658. 34. Leslie WD, Lix LM, Johansson H, et al. Independent clinical validation of a Canadian FRAX tool: Fracture prediction and model calibration. J Bone Miner Res. 2010;25:2350–2358. 35. Leslie WD, Rubin MR, Schwartz AV, Kanis JA. Type 2 diabetes and bone. J Bone Miner Res. 2012;27:2231–2237. 36. Leslie WD, Hough S. Fracture risk assessment in diabetes. Diabetic Bone Disease. 2016:45–69. 37. Leslie WD, Majumdar SR, Morin SN, et al. FRAX for fracture prediction shorter and longer than 10 years: the Manitoba BMD registry. Osteoporos Int. 2017;28:2557–2564. 38. Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150:604–612. 39. Looker AC, Wahner HW, Dunn WL, et al. Updated data on proximal femur bone mineral levels of US adults. Osteoporos Int. 1998;8:468–489. 40. 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. 41. Lix LM, Azimaee M, Osman BA, et al. Osteoporosis-related fracture case definitions for population-based administrative data. BMC Public Health. 2012;12:301. 42. O’Donnell S. Canadian Chronic Disease Surveillance System (CCDSS) Osteoporosis Working Group. Use of administrative data for national surveillance of osteoporosis and related fractures in Canada: results from a feasibility study. Arch Osteoporos. 2013;8:143. 43. Leslie WD, Lix LM, Wu X, Manitoba Bone Density Program. Competing mortality and fracture risk assessment. Osteoporos Int. 2013;24:681–688. 44. Satagopan JM, Ben-Porat L, Berwick M, et al. A note on competing risks in survival data analysis. Br J Cancer. 2004;91:1229–1235. 45. Giangregorio LM, Leslie WD, Lix LM, et al. FRAX underestimates fracture risk in patients with diabetes. J Bone Miner Res. 2012;27:301–308. 46. Jha V, Garcia-Garcia G, Iseki K, et al. Chronic kidney disease: global dimension and perspectives. Lancet. 2013;382:260–272. 47. Schwartz AV, Vittinghoff E, Bauer DC, et al. Association of BMD and FRAX score with risk of fracture in older adults with type 2 diabetes. JAMA. 2011;305:2184–2192.
Kidney International (2019) 95, 447–454
www.kidney-international.org
The Fracture Risk Assessment Tool (FRAX) predicts fracture risk in patients with chronic kidney disease Reid H. Whitlock1,2, William D. Leslie1, James Shaw1, Claudio Rigatto1,2, Laurel Thorlacius1, Paul Komenda1,2, David Collister1, John A. Kanis3,4 and Navdeep Tangri1,2 1
Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, Canada; 2Chronic Disease Innovation Centre, Seven Oaks General Hospital, Winnipeg, Canada; 3Center for Metabolic Bone Diseases, University of Sheffield, Sheffield, UK; and 4Institute for Health and Aging, Catholic University of Australia, Melbourne, Australia
The Fracture Risk Assessment Tool (FRAX) was developed to predict fracture risk in the general population, but its applicability to patients with chronic kidney disease (CKD) is unknown. Using the Manitoba Bone Mineral Density (BMD) Database, we identified adults not receiving dialysis with available serum creatinine measurements and bone densitometry within 1 year. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation. Incident major osteoporotic fractures and hip fractures were ascertained from population-based health care databases. The performance of FRAX, derived without and with BMD, was studied in relation to CKD stage. Among 10,099 subjects (mean age 64 ± 13 years, 13.0% male), 2,154 had eGFR 30-60 mL/min/1.73 m2 (CKD stage 3) and 590 had eGFR <30 mL/min/1.73 m2 (CKD stages 4-5). During a 5-year observation period, 772 individuals experienced a major osteoporotic fracture and 226 had a hip fracture. FRAX predicted risk for major osteoporotic fracture and hip fracture in all eGFR strata. For every standard deviation increase in FRAX score derived with BMD, the hazard ratio (HR) for hip fracture was 4.54 (95% confidence interval [CI] 3.57-5.77) in individuals with eGFR ‡ 60 mL/min/1.73m2, 4.52 (95% CI 3.15-6.49) in individuals with eGFR 30-60 mL/min/1.73m2, and 3.10 (95% CI 1.80-5.33) in individuals with eGFR <30 mL/min/1.73m2. The relationship between FRAX and major osteoporotic fracture was stronger in those with CKD compared to those with preserved eGFR. These findings support the use of FRAX to risk stratify patients with non-dialysis CKD for major osteoporotic fractures and hip fractures. Kidney International (2019) 95, 447–454; https://doi.org/10.1016/ j.kint.2018.09.022 KEYWORDS: bone mineral density; chronic kidney disease; fracture; fracture discrimination; FRAX Copyright ª 2018, International Society of Nephrology. Published by Elsevier Inc. All rights reserved.
O
steoporotic fractures are associated with morbidity, mortality, and high health care costs.1,2 In elderly patients, osteoporotic fractures can result in major functional decline3–7 and lead to a decreased health-related quality of life and early mortality.8–11 At nearly 9 million estimated incident fractures per year, the global burden of osteoporotic fractures is high, and expected to increase in the future with an aging population.12–14 Chronic kidney disease (CKD) is also more common in elderly patients and is associated with increased mortality and morbidity.15,16 For patients over age 65 years, the global prevalence of CKD is approximately 30%, which is 3 times more common than in younger adults.17,18 It follows that CKD and osteoporotic fractures are often co-prevalent in the elderly,18,19 but whether CKD affects fracture risk assessment has not been fully elucidated. In the general population, an individual’s 10-year fracture risk can be estimated using the Fracture Risk Assessment Tool (FRAX). FRAX is widely used in many countries and has been incorporated into numerous guidelines for diagnosis and management of osteoporosis.20,21 Eleven variables plus an optional bone mineral density (BMD) measurement obtained at the femoral neck are used to calculate the probability of fracture, which can then be used to determine the need for preventative therapy. A unique feature of FRAX is that it considers competing mortality in the fracture risk estimation procedure. Notably, the presence of kidney disease is not part of the FRAX risk assessment algorithm. Although BMD is lower in patients with advanced CKD,22 the ability of FRAX to accurately predict fracture risk in patients with CKD is unknown, and existing studies have been limited by their small sample size and number of fracture events, particularly in patients with severely reduced kidney function.23–26 Our primary objective was to determine whether FRAX, derived with or without BMD measurement, predicts incident fracture events in patients with CKD.
Correspondence: Navdeep Tangri, Seven Oaks General Hospital, 2LB10-2300 McPhillips Street, Winnipeg MB R2V 3M3, Canada. E-mail: [email protected]
RESULTS
Received 4 May 2018; revised 13 September 2018; accepted 20 September 2018; published online 20 December 2018
Baseline characteristics and observed events of the cohort are described in Table 1. A total of 10,099 individuals (mean age
Kidney International (2019) 95, 447–454
447
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
64 13 years, 13.0% male) underwent both a dual-energy Xray absorptiometry (DXA) examination and a serum creatinine test within 12 months (Figure 1). The median duration between tests was 6 months, and the interquartile range was 3 to 9 months. Among these, 7355 (72.8%) had an estimated glomerular filtration rate (eGFR) $ 60 ml/min per 1.73 m2, 2154 (21.3%) had an eGFR between 30 and 60 ml/min per 1.73 m2 and 590 (5.8%) had an eGFR < 30 ml/min per 1.73 m2 (380 with eGFR 15–29 ml/min per 1.73 m2, 210 with eGFR < 15 ml/min per 1.73 m2). Compared with individuals with eGFR $ 60 ml/min per 1.73 m2, those subjects with an eGFR < 60 ml/min per 1.73 m2 tended to have a higher risk of fracture and were more likely to be older, male, diabetic, overweight, and have increased albuminuria, lower femoral neck T-scores, and increased mortality. There were 226 (2.2%) incident hip fracture events and 772 individuals (7.6%) with 1 or more major osteoporotic fracture (MOF) events (231 forearm, 206 hip, 180 vertebral, and 158 humerus) over a mean follow-up of 4.6 1.1 years, with higher fracture rates in individuals with lower eGFR. Subjects in the eGFR $ 60 ml/min per 1.73 m2 stratum were significantly less likely to experience a hip fracture or MOF compared with those in the eGFR 30 to 60 ml/min per 1.73 m2 (P < 0.001) or eGFR < 30 ml/min per 1.73 m2 (P < 0.001) groups during the observation period (Figure 2).
FRAX as a predictor of hip and MOF fractures
The effect of a higher FRAX score (with and without BMD measurement) and BMD alone on the risk of hip fracture and MOF in patients with or without reduced kidney function is presented in Table 2. For every SD increase in FRAX computed with BMD, the hazard ratio (HR) for hip fracture was 4.54 in individuals with an eGFR $ 60 ml/min per 1.73 m2 (95% CI: 3.57–5.77, P < 0.001), 4.52 in individuals with an eGFR of 30 to 60 ml/min per 1.73 m2 (95% CI: 3.15–6.49, P < 0.001), and 3.10 in individuals with an eGFR < 30 ml/ min per 1.73 m2 (95% CI: 3.10–5.33, P < 0.001). The interaction between hip fracture probability and eGFR was not significant for FRAX with BMD (P ¼ 0.48) and without BMD (P ¼ 0.23). For MOF probability with BMD, the HR was 1.83 in individuals with an eGFR $ 60 ml/min per 1.73 m2 (95% CI: 1.67–2.01, P < 0.001), 2.33 in individuals with an eGFR of 30 to 60 ml/min per 1.73 m2 (95% CI: 1.96–2.77, P < 0.001), and 2.04 in individuals with an eGFR < 30 ml/min per 1.73 m2 (95% CI: 1.58–2.84, P < 0.001). The interaction of MOF probability and eGFR was significant both with and without BMD, with stronger associations of FRAX in patients with CKD than in those without (P < 0.001). Both FRAX without BMD and BMD alone were also significantly associated with fracture risk irrespective of
Table 1 | Characteristics of study population
Baseline Sex (male) Age Body mass index (kg/m2) Rheumatoid arthritis COPD Alcohol or substance abuse diagnosis Recent glucocorticoid use Prior fracture Parental hip fracture Femoral neck T-score FRAX predicted probability of hip fracture (without BMD) FRAX predicted probability of hip fracture (with BMD) FRAX predicted probability of MOF (without BMD) FRAX predicted probability of MOF (with BMD) eGFR (ml/min per 1.73 m2) Urine ACR (mg/mmol)a Diabetes Observed Events Hip fracture Rate of hip fractures (per 1000 person-yr) MOF Rate of MOF (per 1000 person-yr) Mortality Lost to follow-up
All (n [ 10 099)
eGFR ‡ 60 (n [ 7 355)
eGFR 30–60 (n [ 2 154)
eGFR < 30 (n [ 590)
P value
1308 (13.0%) 64.4 13.3 27.3 5.7 579 (5.7%) 1013 (10.0%) 303 (3.0%) 1193 (11.8%) 1768 (17.5%) 1121 (11.1%) –1.3 1.0 3.9 6.0 2.9 5.0 11.7 9.4 10.8 8.2 73 24 1.5 (0.6–5.3) 2125 (21.0%)
794 (10.8%) 62.2 12.7 27.1 5.6 455 (6.2%) 674 (9.2%) 247 (3.4%) 652 (8.9%) 1258 (17.1%) 837 (11.4%) –1.3 1.0 3.1 5.4 2.4 4.4 10.5 8.6 9.8 7.4 84 15 1.2 (0.6–3.1) 1262 (17.2%)
380 (17.6%) 70.9 12.6 27.7 5.8 97 (4.5%) 253 (11.7%) 40 (1.9%) 384 (17.8%) 399 (18.5%) 231 (10.7%) –1.5 1.1 6.1 7.0 4.2 5.7 15.5 10.4 13.4 9.0 47 8 2.1 (0.6–7.6) 605 (28.1%)
134 (22.7%) 67.0 15.3 28.0 6.0 27 (4.6%) 86 (14.6%) 16 (2.7%) 157 (26.6%) 111 (18.8%) 53 (9.0%) –1.6 1.1 5.4 7.2 4.5 7.1 13.9 10.9 13.4 10.5 18 8 11.5 (1.9–79.5) 258 (43.7%)
<0.001 <0.001 <0.001 0.006 <0.001 0.001 <0.001 0.22 0.168 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
226 (2.2%) 4.57 772 (7.6%) 16.10 1205 (11.9%) 201 (2.0%)
122 (1.7%) 3.38 514 (7.0%) 14.68 592 (8.0%) 158 (2.1%)
81 (3.8%) 7.72 198 (9.2%) 19.47 402 (18.7%) 31 (1.4%)
23 (3.9%) 8.01 60 (10.2%) 21.76 211 (35.8%) 12 (2.0%)
<0.001 <0.001 <0.001 <0.001 <0.001 0.117
ACR, albumin-to-creatinine ratio; BMD, bone mineral density; COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate; FRAX, Fracture Risk Assessment Tool; MOF, major osteoporotic fracture. a Urine ACR obtained for a total of 1210 (11.9%) patients: 727 (9.9%) for eGFR $ 60, 426 (19.5%) for eGFR of 30 to 60, and 147 (23.8%) for eGFR < 30 ml/min per 1.73 m2. Median and interquartile range are presented.
448
Kidney International (2019) 95, 447–454
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
Paents with at least 1 DXA during the study period (n = 42,551)
Exclusions: • No SCr measurement (n = 32,188) • < 20 years of age (n = 150) • On RRT at baseline (n = 114)
Paents included in the cohort for analysis (n = 10,099)
eGFR ≥ 60 ml/min per 1.73 m2 (n = 7,355)
eGFR 30 — 60 ml/min per 1.73 m2
(n = 2,154)
eGFR < 30 ml/min per 1.73 m2 (n = 590)
Figure 1 | Cohort selection. DXA, dual-energy X-ray absorptiometry; eGFR, estimated glomerular filtration rate; RRT, renal replacement therapy (dialysis or transplant); SCr, serum creatinine.
kidney function. However, the magnitude of the hazard ratios as well as model fit and calibration statistics (Brier scores and Akaike Information Criterion) for FRAX with BMD were superior in every eGFR group for both hip fractures and MOFs (Supplementary Table S1). In our cohort, the observed 5-year cumulative probability of incident hip fracture and MOF risk was slightly but consistently higher than the predicted fracture risk across all eGFR strata (Figures 3 and 4). For MOF risk, the observed 5year cumulative fracture percentage was on average approximately 2.5% higher than the predicted fracture risk. For hip fractures, the observed 5-year cumulative fracture percentage was on average approximately 1% higher than the predicted fracture risk (Figure 5). Supplementary analyses
The association between FRAX and MOF or hip fracture, or the interaction between FRAX, eGFR, and fractures was unchanged in the subgroup of participants older than 40 years of age and also when adjusting for bisphosphonate use. FRAX with BMD showed very good discrimination for predicting hip fractures (c-statistic ¼ 0.82) and moderate discrimination for predicting MOF (c-statistic ¼ 0.68) (Supplementary Table S2). After adjusting for clinical risk factors, BMD, and diabetes, eGFR was not independently associated with MOF (HR per 10 ml/min per 1.73 m2 decrease: 1.01, 95% CI: 0.98–1.05, P ¼ 0.39) (Supplementary Table S3). Regarding the
Kidney International (2019) 95, 447–454
association with hip fractures, once adjusted for clinical risk factors, BMD, and diabetes, we found that eGFR was weakly associated with hip fracture (HR per 10 ml/min per 1.73 m2 decrease: 1.07, 95% CI: 1.00–1.14, P ¼ 0.035). Sex was not associated with hip or MOF in these models. Urine albumin-to-creatinine ratio (ACR) measurements were available in 1206 individuals. Of these, 85 MOF and 19 hip fractures were observed during the follow-up period. Urine ACR was not associated with either MOF (P ¼ 0.95) or hip fracture (P ¼ 0.53). DISCUSSION
In this population-based study of more than 10,000 individuals, we found that the FRAX score with or without BMD measurement significantly discriminated fracture risk in patients with nondialysis CKD. Risk for hip fracture did not appear to be influenced by CKD stage, and although FRAX was more strongly predictive of MOF in patients with reduced eGFR, the differences in the hazard ratios between eGFR groups were small, and no linear trend was evident. These findings support the use of FRAX to identify patients with moderate to severe CKD who are at high risk for fracture events, similar to its use in the general population. To our knowledge, this is the largest study to examine fracture risk in a population with advanced CKD using DXA assessment and the FRAX tool. Our assertion that FRAX is valid in patients with CKD is consistent with other smaller studies in the literature. In 2014 Jamal et al.24 used a
449
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
1
FRACTION WITHOUT MOF
0.97
0.94
0.91
eGFR ≥ 60 eGFR 30–60
0.88
Log-rank test P value Comparison eGFR 30–60 vs. eGFR ≥ 60 < 0.001 eGFR < 30 vs. eGFR ≥ 60 < 0.001 eGFR < 30 vs. eGFR 30—60 0.53
eGFR < 30
0.85 0
1
3
2
4
5
YEARS
Figure 2 | Kaplan-Meier curve: time to major osteoporotic fracture. eGFR, estimated glomerular filtration rate; MOF, major osteoporotic fracture.
cross-sectional convenience sample of 353 CKD patients with mean eGFR of 28 ml/min per 1.73 m2 to demonstrate that FRAX was able to discriminate prevalent fractures in the CKD population, but offered no additional discrimination compared with BMD measurement. Because of the crosssectional nature of the study, the ability of FRAX to predict incident fracture risk was not assessed. More recently, a study by Naylor et al.26 used a larger Canadian prospective cohort stratified by eGFR to show that FRAX could accurately predict MOF in patients with reduced kidney function. However, of the 2107 patients in the study, only 320 had an eGFR < 60 ml/min per 1.73 m2, and only 13 had eGFR < 30 ml/min per 1.73 m2. This small sample size limited their ability to assess FRAX prediction for fracture in those with CKD. Notably, they showed that FRAX calibration was not improved when CKD was added as a secondary cause of osteoporosis in those patients with eGFR < 60 ml/min per 1.73 m2. Our data, based on a much larger sample size with adequate representation of patients with eGFR of 30 to 60 and < 30 ml/min per 1.73 m2, clearly demonstrated that FRAX can predict fracture risk in patients with CKD. Previous studies have examined whether CKD, diminished eGFR, or albuminuria are independent risk factors for
fracture, and found conflicting results. Investigators from the Cardiovascular Health Study measured cystatin C in more than 4000 older adults, and found that eGFR based on cystatin C was associated with a higher risk of incident fractures. Although the investigators adjusted for several possible confounders, they did not have access to BMD measurements and did not adjust for all possible FRAX risk factors.27 In contrast, investigators from the Alberta Kidney Disease Network studied nearly 1.8 million individuals and found that age- and sex-adjusted rates of osteoporotic and hip fracture did not differ in individuals with CKD versus those without.23 While the Alberta study included nearly 130,000 individuals with CKD, no data on BMD or other FRAX risk factors was available. A study of 2982 individuals from the Osteoporotic Fractures in Men cohort found no association between urine albumin levels and risk of incident clinical fracture.28 The investigators did adjust for confounders and had access to BMD measurements, but no women were included in the study. Our findings suggest that a lower eGFR or presence of albuminuria is not associated with MOF risk after adjustment for BMD and FRAX clinical risk factors. There was a weak association between eGFR and hip fracture risk, but this was attenuated after adjustment for diabetes, a leading cause of
Table 2 | Gradient of risk (per SD increase) for incident fractures in groups stratified by kidney function eGFR ‡ 60 (n [ 7 355)
HR (95% CI) Probability of Probability of BMD alonea Probability of Probability of BMD alonea
hip fracture from FRAX with BMD hip fracture from FRAX without BMD MOF from FRAX with BMD MOF from FRAX without BMD
4.54 3.55 3.41 1.83 1.68 1.75
(3.57–5.77) (2.83–4.45) (2.79–4.18) (1.67–2.01) (1.52–1.84) (1.60–1.92)
eGFR 30–60 (n [ 2 154) 4.52 4.34 2.85 2.33 2.31 2.00
(3.15–6.49) (2.95–6.38) (2.20–3.69) (1.96–2.77) (1.91–2.80) (1.71–2.34)
eGFR < 30 (n [ 590) 3.10 1.98 2.48 2.04 1.87 1.87
(1.80–5.33) (1.25–3.15) (1.62–3.78) (1.58–2.84) (1.42–2.46) (1.44–2.42)
Interaction eGFR*FRAX P ¼ 0.47 P ¼ 0.23 P ¼ 0.109 P < 0.001 P < 0.001 P ¼ 0.048
BMD, bone mineral density; CI, confidence interval; eGFR, estimated glomerular filtration rate; FRAX, Fracture Risk Assessment Tool; HR, hazard ratio; MOF, major osteoporotic fracture. a BMD is measured by femoral neck T-score, and its HR is reported per SD decrease.
450
Kidney International (2019) 95, 447–454
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
Figure 3 | Kaplan-Meier curve: time to hip fracture. eGFR, estimated glomerular filtration rate.
CKD. Whether CKD is truly an independent risk factor for fracture remains to be fully understood, and our study calls for better recognition and classification of fracture risk in patients with CKD.29 There are important clinical and research implications of our work. We believe that physicians can consider FRAX scores valid in patients with nondialysis CKD and use them to inform treatment decisions. Bisphosphonates are a frequently used therapy for the general population with osteoporosis and have been shown to be effective in fracture risk reduction.30 This class of medication is not well characterized in CKD and generally not recommended once eGFR is <30 ml/min per 1.73 m2; however, early research into the use of these and other medications for fracture prevention in patients with CKD has shown potential benefits.31–33 Given the morbidity from fractures, and the older CKD population, wider use of
risk assessment tools like FRAX and therapies including bisphosphonates may have a beneficial clinical impact. Although the c-statistics for FRAX with or without BMD were similar to BMD alone in our study population, it is important to note that c-statistics are relatively insensitive to the inclusion of additional risk factors and should not discourage efforts to identify newer biological and/or radiological markers of fracture risk that can markedly improve discrimination and calibration. The strengths of our study include the use of large, robust regional databases that capture all bone densitometry and nearly all fracture events, and include a large number of patients with reduced kidney function. Only 200 patients (2.0%) migrated from Manitoba before the end of the study period, resulting in minimal loss to follow-up. The available BMD and FRAX scores from nearly 600 patients with advanced
15.0%
Fracture percentage
12.0%
9.0%
6.0% 10.3%
9.3% 7.1%
3.0%
6. 7% 4.9%
7.7%
6.7%
6.9%
5.2%
0.0% eGFR ≥ 60 Observed 5-year MOF
eGFR 30 –60
eGFR < 30
Predicted 5-year hip fracture MOF (with BMD)
Predicted 5-year hip fracture MOF (without BMD)
Figure 4 | Observed versus predicted 5-year major osteoporotic fracture probability. BMD, bone mineral density; eGFR, estimated glomerular filtration rate; MOF, major osteoporotic fracture. Note: Error bars represent 95% confidence intervals. Kidney International (2019) 95, 447–454
451
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
10.0%
Fracture percentage
8.0%
6.0%
4.0%
2.0%
3.9%
3.8% 1.7%
1.2%
1.6%
3.0%
2.7%
2.1%
2.2%
eGFR 30 – 60
eGFR < 30
0.0% eGFR ≥ 60 Observed 5-year hip fractures
Predicted 5-year hip fracture FRAX (with BMD)
Predicted 5-year hip fracture FRAX (without BMD)
Figure 5 | Observed versus predicted 5-year hip fracture probability. BMD, bone mineral density; eGFR, estimated glomerular filtration rate. Note: Error bars represent 95% confidence intervals.
CKD (eGFR < 30 ml/min per 1.73 m2) is a unique feature of this study, and extends our confidence in FRAX risk assessment to all patients with more advanced CKD. Our study also has limitations. Only adult Canadians who had both serum creatinine and BMD measured within 1 year were included, and as a consequence our cohort was predominantly female. As such, the applicability of our findings to male patients with CKD and to those living in other regions is unknown and needs further validation. Our classification of subjects into eGFR categories was based upon a single serum creatinine, which may not represent the subject’s true baseline renal function. However, measurement error should bias results toward the null; therefore our results are likely to be conservative. In addition, although the discrimination of the FRAX score in our population was similar (cstatistics within 0.01) to what was reported for hip and MOF events in the general population,34 calibration may require adjustment because the observed incidence of hip and MOF events was slightly higher than predicted across all eGFR strata. However, this may not be due to CKD, and may reflect characteristics intrinsic to our cohort, who may have had a higher fracture risk compared with the general population based on our selection criteria, as it was also seen in those with normal eGFR. In other populations, such as patients with type 2 diabetes where FRAX has underestimated risk, inclusion of rheumatoid arthritis as a proxy for type 2 diabetes has been proposed to improve calibration.35,36 Whether a similar substitution would be beneficial for patients with CKD is unknown. We also cannot rule out that confounding by unmeasured variables may have affected fracture risk: for example, we did not account for race, hyperparathyroidism, vitamin D deficiency, or other changes of renal osteodystrophy, as these variables were not routinely available in our data sets. Finally, although the FRAX score was initially designed to evaluate 10-year fracture risk, we have used half 452
the 10-year risk to estimate 5-year risk, which is a valid approach based on previous literature.26,37 In summary, our findings support the use of BMD testing and use of the FRAX tool to determine fracture risk in patients with nondialysis CKD. Additional studies synthesizing the evidence and comparing the effectiveness of newer agents for fracture risk reduction in patients with CKD are needed to inform treatment decisions. METHODS Study design and population We conducted a retrospective cohort study that used anonymously linked administrative health databases from Manitoba, a Canadian province with a population of w1.3 million, where all residents are covered through a single-payer universal health care system. All databases used in the analysis were from the Data Repository housed at the Manitoba Centre for Health Policy at the University of Manitoba. The study was approved by the Health Research Ethics Board for the University of Manitoba, and data access was granted by the Health Information Privacy Committee, Diagnostic Services of Manitoba, and the Winnipeg Regional Health Authority. Our cohort consisted of all individuals age 20 years and older who underwent both a DXA examination and a serum creatinine test within 12 months between the years 2005 and 2010. The DXA appointment date was used as the index date. Individuals in the cohort were stratified into 3 groups according to their eGFR as calculated by the Chronic Kidney Disease Epidemiology Collaboration equation38: eGFR $ 60 ml/min per 1.73 m2 (normal kidney function or CKD stage 1 and 2), eGFR of 30 to 60 ml/min per 1.73 m2 (CKD stage 3), and eGFR < 30 ml/min per 1.73 m2 (CKD stages 4 and 5). The lowest eGFR value within 12 months of the index date was used. Serum creatinine was obtained through the Diagnostic Services of Manitoba laboratory database, which serves >70% of the province. Patients on dialysis at the time of the index date were excluded from the analysis because the sample size was too low to yield enough incident fracture events. Patients known to have a renal transplant were also excluded.
Kidney International (2019) 95, 447–454
clinical investigation
RH Whitlock et al.: Utility of FRAX in CKD
Clinical risk factor collection Results from the DXA examination were captured in the Manitoba BMD database. In addition to the BMD measurements obtained by DXA, the Manitoba BMD database also captured clinical risk factors for fracture. Height and weight were measured with a wall-mounted stadiometer and scale at the time of DXA appointment, and were used to calculate body mass index. Parental hip fracture was selfreported. Age and sex were captured from the Manitoba Health Insurance Registry. Recent glucocorticoid use (>90 days dispensed in the year prior to DXA testing) was obtained through the provincewide Drug Prescription Information Network. A diagnosis of rheumatoid arthritis, prior fracture, alcohol and/or substance abuse, or chronic obstructive pulmonary disease (the latter two being proxies for high alcohol intake and smoking, respectively) were obtained through hospital abstracts and physician claims databases using methods that have been previously described.34 Probability of fracture Ten-year probability of an MOF and hip fracture were calculated for each subject by the Canadian FRAX tool (version 3.11, FRAX Desktop Multi-Patient Entry) using the clinical risk factors described above and femoral neck BMD expressed as a T-score (the number of SDs above or below young adult mean BMD based on white female reference data from the National Health and Nutrition Examination Survey III survey).39 The Canadian FRAX tool has been independently validated for fracture risk prediction in our population.34,40 Outcomes of interest The presence of hip, clinical vertebral, forearm, or humerus fracture codes (collectively designated as MOFs) was determined through a combination of hospital discharge abstracts and physician billing claims using validated algorithms.41,42 A hip fracture was identified if it was the primary diagnosis code for a hospitalization. A vertebral fracture was identified if it was the primary diagnosis code for a hospitalization or there was a single physician visit. For forearm and humerus fractures, at least 1 hospitalization (with a primary diagnosis code) or 2 separate physician visits (within 90 days) were required. Incident fractures were ascertained for up to 5 years after the index date. In order to exclude possible recurrent fractures, we required that a fracture of the same type did not occur in the previous 6 months. To increase accuracy for an acute event, the study required hip and forearm fractures to be accompanied by a sitespecific procedural code. Fractures associated with high trauma codes were excluded.
Additionally, FRAX was modeled for the entire cohort alongside eGFR as a continuous variable. In these models we tested for an eGFR*FRAX interaction to examine whether eGFR modified the effect of FRAX on the risk of fracture. Hip fractures and MOF outcomes were modeled separately. Patients were censored in the analysis at the time of death or if they emigrated from Manitoba as determined by the Manitoba Health Insurance Registry. Loss to follow-up due to migration was less than 2% and did not differ between groups. There were no missing data. The observed 5-year cumulative fracture probability was derived using a modified Kaplan-Meier method with competing mortality framework whereby individuals who died before experiencing a fracture event were not censored at the time of death and instead were assigned a follow-up time of 5 years.43 This approach is equivalent to other competing risk approaches (e.g., that of Satagopan et al.).44 The observed values among kidney disease groups were compared with the log-rank test. The 5-year predicted fracture probability (one-half of the 10-year probability) was compared with observed values to assess FRAX calibration.37 Supplementary analyses We also conducted several subgroup and sensitivity analyses. First, we excluded individuals younger than 40 years of age, and examined hazard ratios for differences and P values for interaction between the FRAX score and the observed outcomes of hip and MOF. Second, we adjusted each model for past bisphosphonate use (at least 90 days in the year before the index date) and examined for the same differences. Third, we also calculated concordance statistics (c- statistics) among kidney disease groups as an alternative measure of the ability of FRAX and BMD alone to discriminate fracture risk. Furthermore, we examined the association of measures of CKD and fracture risk. Cox proportional hazards models of eGFR as a predictor of time to fracture were modeled univariately and in the presence of the FRAX clinical risk factors, femoral neck BMD, and diabetes. Diabetes was included as it is the primary cause of CKD in the majority of individuals, and has been shown to be independently associated with fracture risk.45–47 A diagnosis of diabetes was captured through hospital discharge abstracts, physician claims, or a prescription of insulin or anti-diabetes medication. Cox proportional hazards models of urine ACR as a predictor of fracture were modeled univariately. Urine ACR was captured through the Diagnostic Services of Manitoba database and was tested by using the highest value within 12 months of the index date. Both eGFR and urine ACR were modeled as continuous variables. DISCLOSURE
Data analysis All analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, NC). Descriptive statistics consisted of baseline demographic information, clinical risk factors, FRAX scores, and outcomes of interest. Continuous variables were presented with mean and SD and were compared using analysis of variance or Mann-Whitney U Test. Categorical variables were presented as frequency and percentage and were compared with the chi-square test. Cox proportional hazards models were used to test for associations between FRAX score and the outcomes of interest (incident fracture within 5 years). FRAX score (with and without BMD) was log-transformed and modeled in all 3 stratified groups with the hazard ratios presented per SD increase. Model fit statistics and calibration were assessed using the Akaike Information Criterion and the Brier Score. Kidney International (2019) 95, 447–454
Our study was funded by Diagnostic Services Manitoba (DSM). DSM had no role in the study design, collection, analysis and interpretation of data, writing the report, or decision to submit the report for publication. NT has received honoraria or research support from AstraZeneca, Otsuka Inc., and Tricida Inc. All the other authors declared no competing interests. ACKNOWLEDGMENTS
This work was supported through funding provided by the Department of Health of the Province of Manitoba to the University of Manitoba (HIPC no. 2015/2016-12). The results and conclusions are those of the authors, and no official endorsement by Manitoba Health was intended or should be inferred. Data used in this study are from the Population Health Research Data Repository housed at the Manitoba Centre for Health Policy, University of Manitoba, and were derived from data provided by Manitoba Health, Winnipeg Regional 453
clinical investigation
Health Authority, and Diagnostic Services Manitoba. This article has been reviewed and approved by the members of the Manitoba Bone Density Program Committee. SUPPLEMENTARY MATERIAL Table S1. Akaike Information Criterion and Brier scores for Fracture Risk Assessment Tool (FRAX) with bone mineral density (BMD) measurement and BMD alone as a predictor of fracture. Table S2. C-statistics for Fracture Risk Assessment Tool (FRAX) as a predictor of fracture within 5 years. Table S3. Hazard ratios and 95% confidence intervals for estimated glomerular filtration rate as a risk factor for fracture within 5 years. Supplementary material is linked to the online version of the paper at www.kidney-international.org. REFERENCES 1. Johnell O, Kanis JA. An estimate of the worldwide prevalence and disability associated with osteoporotic fractures. Osteoporosis Int. 2006;17:1726–1733. 2. Hopkins RB, Burke N, Von Keyserlingk C, et al. The current economic burden of illness of osteoporosis in Canada. Osteoporosis Int. 2016;27: 3023–3032. 3. Center JR, Nguyen TV, Schneider D, et al. Mortality after all major types of osteoporotic fracture in men and women: an observational study. Lancet. 1999;353:878–882. 4. Ioannidis G, Papaioannou A, Hopman WM, et al. Relation between fractures and mortality: results from the Canadian Multicentre Osteoporosis Study. Can Med Assoc J. 2009;181:265–271. 5. Morin S, Lix LM, Azimaee M, et al. Mortality rates after incident nontraumatic fractures in older men and women. Osteoporos Int. 2011;22: 2439–2448. 6. Tosteson AN, Gottlieb DJ, Radley DC, et al. Excess mortality following hip fracture: the role of underlying health status. Osteoporos Int. 2007;18: 1463–1472. 7. Johnell O, Kanis JA, Oden A, et al. Mortality after osteoporotic fractures. Osteoporos Int. 2004;15:38–42. 8. Hallberg I, Rosenqvist AM, Kartous L, et al. Health-related quality of life after osteoporotic fractures. Osteoporosis Int. 2004;15:834–841. 9. Tarride J, Burke N, Leslie WD, et al. Loss of health related quality of life following low-trauma fractures in the elderly. BMC Geriatrics. 2016;16:84. 10. Al-Sari UA, Tobias J, Clark E. Health-related quality of life in older people with osteoporotic vertebral fractures: a systematic review and metaanalysis. Osteoporos Int. 2016;27:2891–2900. 11. Borgstrom F, Lekander I, Ivergard M, et al. The International Costs and Utilities Related to Osteoporotic Fractures Study (ICUROS)–quality of life during the first 4 months after fracture. Osteoporos Int. 2013;24:811–823. 12. Oden A, McCloskey EV, Kanis JA, et al. Burden of high fracture probability worldwide: secular increases 2010-2040. Osteoporos Int. 2015;26:2243– 2248. 13. Gullberg B, Johnell O, Kanis JA. World-wide projections for hip fracture. Osteoporosis Int. 1997;7:407–413. 14. Schneider EL, Guralnik JM. The aging of America. Impact on health care costs. JAMA. 1990;263:2335–2340. 15. Go AS, Chertow GM, Fan D, et al. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med. 2004;351: 1296–1305. 16. Tonelli M, Wiebe N, Culleton B, et al. Chronic kidney disease and mortality risk: a systematic review. J Am Soc Nephrol. 2006;17:2034–2047. 17. Arora P, Vasa P, Brenner D, et al. Prevalence estimates of chronic kidney disease in Canada: results of a nationally representative survey. Can Med Assoc J. 2013;185:E417–E423. 18. Hill NR, Fatoba ST, Oke JL, et al. Global prevalence of chronic kidney disease - a systematic review and meta-analysis. PLoS One. 2016;11, e0158765. 19. Klawansky S, Komaroff E, Cavanaugh PF, et al. Relationship between age, renal function and bone mineral density in the US population. Osteoporosis Int. 2003;14:570–576. 20. Kanis JA, Harvey NC, Cooper C, et al. A systematic review of intervention thresholds based on FRAX: a report prepared for the National Osteoporosis Guideline Group and the International Osteoporosis Foundation. Arch Osteoporos. 2016;11:25.
454
RH Whitlock et al.: Utility of FRAX in CKD
21. Kanis JA, Johansson H, Oden A, et al. Worldwide uptake of FRAX. Arch Osteoporos. 2014;9:166. 22. Bucur RC, Panjwani DD, Turner L, et al. Low bone mineral density and fractures in stages 3-5 CKDCKD: an updated systematic review and metaanalysis. Osteoporosis Int. 2015;26:449–458. 23. Elliott MJ, James MT, Quinn RR, et al. Estimated GFR and fracture risk: a population-based study. Clin J Am Soc Nephrol. 2013;8:1367–1376. 24. Jamal SA, West SL, Nickolas TL. The clinical utility of FRAX to discriminate fracture status in men and women with chronic kidney disease. Osteoporos Int. 2014;25:71–76. 25. Naylor KL, McArthur E, Leslie WD, et al. The three-year incidence of fracture in chronic kidney disease. Kidney Int. 2014;86:810–818. 26. Naylor KL, Garg AX, Zou G, et al. Comparison of fracture risk prediction among individuals with reduced and normal kidney function. Clin J Am Soc Nephrol. 2015;10:646–653. 27. Fried LF, Biggs ML, Shlipak MG, et al. Association of kidney function with incident hip fracture in older adults. J Am Soc Nephrol. 2007;18: 282–286. 28. Fink HA, Vo TN, Langsetmo L, et al. Association of increased urinary albumin with risk of incident clinical fracture and rate of hip bone loss: the Osteoporotic Fractures in Men Study. J Bone Miner Res. 2017;32: 1090–1099. 29. Moe SM. Renal osteodystrophy or kidney-induced osteoporosis? Curr Osteoporos Rep. 2017;15:194–197. 30. Crandall CJ, Newberry SJ, Diamant A, et al. Comparative effectiveness of pharmacologic treatments to prevent fractures: an updated systematic review. Ann Intern Med. 2014;161:711. 31. Miller PD, Roux C, Boonen S, et al. Safety and efficacy of risedronate in patients with age-related reduced renal function as estimated by the Cockcroft and Gault method: a pooled analysis of nine clinical trials. J Bone Miner Res. 2005;20:2105–2115. 32. Jamal SA, Bauer DC, Ensrud KE, et al. Alendronate treatment in women with normal to severely impaired renal function: an analysis of the Fracture Intervention Trial. J Bone Miner Res. 2007;22:503–508. 33. Wilson LM, Rebholz CM, Jirru E, et al. Benefits and harms of osteoporosis medications in patients with chronic kidney disease: a systematic review and meta-analysis. Ann Intern Med. 2017;166:649–658. 34. Leslie WD, Lix LM, Johansson H, et al. Independent clinical validation of a Canadian FRAX tool: Fracture prediction and model calibration. J Bone Miner Res. 2010;25:2350–2358. 35. Leslie WD, Rubin MR, Schwartz AV, Kanis JA. Type 2 diabetes and bone. J Bone Miner Res. 2012;27:2231–2237. 36. Leslie WD, Hough S. Fracture risk assessment in diabetes. Diabetic Bone Disease. 2016:45–69. 37. Leslie WD, Majumdar SR, Morin SN, et al. FRAX for fracture prediction shorter and longer than 10 years: the Manitoba BMD registry. Osteoporos Int. 2017;28:2557–2564. 38. Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med. 2009;150:604–612. 39. Looker AC, Wahner HW, Dunn WL, et al. Updated data on proximal femur bone mineral levels of US adults. Osteoporos Int. 1998;8:468–489. 40. 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. 41. Lix LM, Azimaee M, Osman BA, et al. Osteoporosis-related fracture case definitions for population-based administrative data. BMC Public Health. 2012;12:301. 42. O’Donnell S. Canadian Chronic Disease Surveillance System (CCDSS) Osteoporosis Working Group. Use of administrative data for national surveillance of osteoporosis and related fractures in Canada: results from a feasibility study. Arch Osteoporos. 2013;8:143. 43. Leslie WD, Lix LM, Wu X, Manitoba Bone Density Program. Competing mortality and fracture risk assessment. Osteoporos Int. 2013;24:681–688. 44. Satagopan JM, Ben-Porat L, Berwick M, et al. A note on competing risks in survival data analysis. Br J Cancer. 2004;91:1229–1235. 45. Giangregorio LM, Leslie WD, Lix LM, et al. FRAX underestimates fracture risk in patients with diabetes. J Bone Miner Res. 2012;27:301–308. 46. Jha V, Garcia-Garcia G, Iseki K, et al. Chronic kidney disease: global dimension and perspectives. Lancet. 2013;382:260–272. 47. Schwartz AV, Vittinghoff E, Bauer DC, et al. Association of BMD and FRAX score with risk of fracture in older adults with type 2 diabetes. JAMA. 2011;305:2184–2192.
Kidney International (2019) 95, 447–454