Visual Abstract
Keywords: CKD, clinical epidemiology, epidemiology and outcomes, kidney disease, CKD non-dialysis
Abstract
Key Points
In patients with CKD and atrial fibrillation, we observed no difference in the rates of fracture between initiators of direct oral anticoagulant and warfarin.
However, direct oral anticoagulant use relative to warfarin was associated with a lower risk of all-cause mortality.
Background
Direct oral anticoagulant (DOAC) use has been associated with a lower risk of adverse events relative to warfarin in patients with atrial fibrillation. Little is known about the risk of fracture in association with anticoagulant therapy in patients with CKD.
Methods
We conducted a new user, active comparator cohort study in a United States-based commercial claims database spanning 2013 through 2020 to quantify the comparative risk of fracture associated with select DOACs (apixaban or rivaroxaban) versus warfarin. Individuals were required to have International Classification of Diseases diagnosis codes for CKD (stages 3–5) and atrial fibrillation during the 365-day baseline period before anticoagulant initiation. Primary analyses quantified nonvertebral fracture risk between patients initiating DOACs and warfarin using a 1:1 propensity score-matched design. Cox proportional hazards regression was used to obtain hazard ratios (HRs) and 95% confidence intervals (CIs) of nonvertebral fracture. Secondary analyses evaluated risks of hip fracture and all-cause mortality.
Results
The 1:1 propensity score-matched population included 14,370 DOAC initiators and 14,370 warfarin initiators. The mean age at anticoagulant initiation was 77 years, and 45% were female. The HR for nonvertebral fracture comparing DOACs with warfarin was 1.12 (95% CI, 0.95 to 1.32), and the corresponding incidence rate difference per 1000 person-years was 3.55 (95% CI, −1.67 to 8.76). The HR and incidence rate difference comparing DOACs with warfarin were 0.98 (95% CI, 0.68 to 1.41) and −0.13 (95% CI, −2.52 to 2.25), respectively, for hip fracture and 0.91 (95% CI, 0.85 to 0.98) and −17.23 (95% CI, −29.49 to −4.96), respectively, for all-cause mortality.
Conclusions
In patients with CKD and atrial fibrillation, we did not observe a difference in the rates of fracture between DOAC and warfarin initiators. DOAC use relative to warfarin was associated with a lower risk of all-cause mortality.
Introduction
Osteoporotic fracture is a frequent cause of morbidity and mortality among older adults with atrial fibrillation.1–3 Patients with CKD and atrial fibrillation, in particular, may have a high risk of fracture due to CKD-specific changes in mineral metabolism and bone turnover.4–7 In the United States, age-standardized incidence rates (IRs) of hip fracture are as high as 3.89/1000 persons in patients with ESKD and 1.81/1000 persons in patients with non–dialysis-requiring CKD relative to 1.18/1000 persons in individuals with normal kidney function.7 Anticoagulant initiation may be warranted in patients with CKD and atrial fibrillation to reduce the burden of embolic events, but little is known about the risk of fracture in association with anticoagulant therapy in this population.
Warfarin, a vitamin K antagonist, was previously the only oral anticoagulant available to prevent or treat stroke complications in patients with atrial fibrillation.1 Long-term warfarin use has been associated with lower bone mineral density and greater susceptibility to bone fracture compared with nonuse, which may be more pronounced in patients with CKD.3,8 With the more recent approval of direct oral anticoagulants (DOACs), including dabigatran, apixaban, rivaroxaban, and edoxaban, several comparative studies have identified lower risks of fracture and other adverse clinical outcomes relative to warfarin in patients with atrial fibrillation.1,3,9 However, anticoagulant safety and efficacy data are limited in patients with moderate to advanced CKD.10 Four randomized trials found no significant difference between DOACs and warfarin for the treatment of venous thromboembolism in patients with creatinine clearance (CrCl) 30–50 ml/min.11–14 Other randomized trial subanalyses in patients with CrCl <50 ml/min identified similar risks of stroke or systemic embolism between those initiating DOACs and warfarin.15–18 Findings for major bleeding in patients with CrCl <50 ml/min were mixed because only those taking apixaban and edoxaban had lower risks of major bleeding compared to those taking warfarin.15,17,18 Aside from thrombotic and bleeding events, patients with moderate and advanced CKD have an elevated risk of fracture due to disease-specific changes in bone mineral density.4–7 Accordingly, it is unclear whether all anticoagulants confer the same risk of fracture or whether there may be safer alternatives for patients with CKD and atrial fibrillation.
The objective of this study was to quantify the comparative risks of fracture and other adverse clinical outcomes associated with DOAC versus warfarin initiation in patients with moderate to advanced CKD and atrial fibrillation.
Methods
We conducted a new user, active comparator cohort study to quantify the comparative risk of fracture for select DOACs versus warfarin in patients with moderate to advanced CKD (stages 3–5) and atrial fibrillation (Figure 1).
Figure 1.
New user and active comparator cohort study design to quantify the comparative risk of fracture in patients with CKD and atrial fibrillation initiating DOACs versus warfarin. DOAC, direct oral anticoagulant.
Data Sources
We used data from Optum's deidentified Clinformatics Data Mart Database spanning January 1, 2013, through December 31, 2020. This large United States-based insurance database contains administrative pharmacy and medical claims for commercially insured individuals across all 50 states, including data for retirees with Medicare supplemental insurance paid by employers. Analyses were restricted to these calendar years based on the US Food and Drug Administration approval and marketing of all anticoagulants under study.
This study was approved by the Institutional Review Board at Brigham and Women's Hospital.
Study Population
Eligible individuals were identified based on their first dispensing for a study anticoagulant at 50 years or older (cohort entry date) immediately after at least 365 days of continuous health plan enrollment with no anticoagulant use. The cohort was further restricted to patients with at least one International Classification of Diseases (ICD) diagnosis code for CKD, stages 3–5, and atrial fibrillation during the 365-day baseline period (Figure 1). Among this eligible population, each patient was assigned to a single CKD stage based on the most recent ICD diagnosis code before the index date. While eGFR results more accurately classify patients into CKD stages than diagnosis codes used for billing purposes, laboratory data were only available for a limited subset of eligible individuals. Nevertheless, ICD diagnosis codes for CKD have previously been validated and shown to have high specificity and positive predictive value (>80%) in claims data.19 The anticoagulants under study included warfarin, apixaban, and rivaroxaban. Apixaban and rivaroxaban were evaluated as a single group (DOACs) in the primary analysis. Dabigatran and edoxaban were not assessed because of the small number of eligible patients with CKD initiating each drug during the study period. Patients with CKD who experienced fracture during the baseline period or on the cohort entry date were excluded from the study.
Outcome Assessment and Follow-Up
Nonvertebral fracture was defined using a combination of ICD diagnosis and procedure codes and Current Procedural Terminology codes for the following locations: hip and femur; pelvis; radius and ulna; wrist; humerus; and other. The secondary outcome of all-cause mortality was defined based on inpatient encounters.
We performed a 365-day as-treated analysis to evaluate the risk of nonvertebral fracture in the 1-year period after anticoagulant initiation. A 1-year follow-up period was selected to focus on adverse events that may be attributable to anticoagulant use. Follow-up began the day after cohort entry until outcome occurrence, anticoagulant discontinuation, initiation of an anticoagulant from the other group, health plan disenrollment, death, end of the 365-day period immediately after the cohort entry date, or end of the data study period. We incorporated a grace period of 60 days between the end of an anticoagulant days' supply and subsequent anticoagulant dispensing to define continuous treatment. We also performed a 365-day intention-to-treat analysis to account for potential informative censoring in the as-treated analysis. We used identical censoring criteria as in the as-treated analysis, but patients with CKD were not censored at treatment discontinuation or switch.
Covariates
We assessed demographic characteristics; comorbid conditions; osteoporosis-related characteristics, such as bone mineral density scan, osteopenia, and vitamin D deficiency; other medication use; and health care utilization over the 365 days before and on the cohort entry date (see Supplemental Table 2 for a full list of covariates). For the subset of patients with available laboratory results (approximately 40% of patients), we also evaluated the results of laboratory tests, such as serum creatinine and eGFR, to ensure comparability of treatment groups at baseline.
Statistical Analyses
We used 1:1 nearest neighbor propensity score matching without replacement to adjust for covariates.20 This approach has been shown to yield the lowest bias compared with other matching approaches by minimizing the distance between the treatment and reference patient in each matched set.21 By focusing on the subpopulation with the highest comparability, pairwise matching is not subject to the extreme weights and resulting high variance that may be observed with propensity score weighting methods.21,22 A logistic regression model was used to estimate the probability of treatment with DOACs as a function of 82 covariates. Laboratory results, such as serum creatinine and eGFR, were excluded from the propensity score model because these laboratory results were only available for a subset (approximately 40%) of the study population. Upon assigning each individual an estimated propensity score, each patient with CKD and atrial fibrillation initiating a select DOAC was matched to a patient initiating warfarin using a matching caliper of one percentage point. In the matched population, survival curves were constructed using the Kaplan–Meier method. We also fit Cox proportional hazards regression models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) of fracture for DOACs relative to warfarin. IR differences (IRDs) of fracture per 1000 person-years and 95% CIs were calculated using a modified generalized linear model with normal distribution and identity link.23 All analyses were conducted using Aetion Evidence Platform.
Secondary Analyses
We performed several additional analyses. Hip fracture and all-cause mortality were evaluated as secondary outcomes for patients with CKD initiating DOACs versus warfarin in 365-day as-treated analyses. We also repeated all analyses comparing apixaban and rivaroxaban individually to warfarin. Finally, we excluded patients with ICD codes for valvular atrial fibrillation in a sensitivity analysis to evaluate the potential effect of the underlying indication for anticoagulant dispensing on study results.
Results
Study Population
We identified 49,698 DOAC initiators and 18,931 warfarin initiators before matching (Supplemental Table 1). The 1:1 propensity score-matched population included 14,370 DOAC initiators and 14,370 warfarin initiators. The mean age at anticoagulant initiation was 77 years (SD, 8 years), and 45% were female in the matched population (Table 1). Before matching, DOAC initiators were more likely to have diagnosis codes for CKD stage 3, while a larger proportion of those initiating warfarin had diagnosis codes for CKD stages 4 and 5. Occurrence of certain cardiovascular-related comorbid conditions, including congestive heart failure, coronary artery bypass graft surgery, and venous thromboembolism, were higher in warfarin compared with DOAC users (Supplemental Table 2). Other comorbidities, medication use, and health care utilization measures were otherwise similar between the two groups before and after matching. Propensity score distributions before and after matching are presented in Supplemental Figures 1–4.
Table 1.
Select baseline characteristics of patients with CKD and atrial fibrillation initiating direct oral anticoagulants or warfarin before and after 1:1 propensity score matching (2013–2020)
| Baseline Characteristic | Unmatched | 1:1 Propensity Score Matched | ||||
|---|---|---|---|---|---|---|
| DOACs | Warfarin | ASD | DOACs | Warfarin | ASD | |
| n | 49,698 | 18,931 | 14,370 | 14,370 | ||
| Apixaban, n (%) | 38,741 (77.9) | 0 (0) | 10,102 (70.3) | 0 (0) | ||
| Rivaroxaban, n (%) | 10,990 (22.1) | 0 (0) | 4268 (29.7) | 0 (0) | ||
| Demographic characteristics, n (%) | ||||||
| Age, yr, mean (SD) | 77.65 (8.01) | 76.54 (8.17) | 0.14 | 76.88 (8.24) | 76.89 (8.01) | 0.001 |
| Sex | 0.08 | 0.02 | ||||
| Female | 24,074 (48.4) | 8434 (44.6) | 6431 (44.8) | 6566 (45.7) | ||
| Male | 25,624 (51.6) | 10,497 (55.4) | 7939 (55.2) | 7804 (54.3) | ||
| Race | 0.05 | 0.01 | ||||
| Asian | 1157 (2.3) | 452 (2.4) | 331 (2.3) | 342 (2.4) | ||
| Black | 6233 (12.5) | 2170 (11.5) | 1638 (11.4) | 1652 (11.5) | ||
| Other | 2405 (4.8) | 891 (4.7) | 699 (4.9) | 699 (4.9) | ||
| White | 34,632 (69.7) | 13,580 (71.7) | 10,313 (71.8) | 10,285 (71.6) | ||
| Ethnicity | 0.05 | 0.01 | ||||
| Hispanic | 5271 (10.6) | 1838 (9.7) | 1389 (9.7) | 1392 (9.7) | ||
| Unknown | 44,427 (89.4) | 17,093 (90.3) | 12,981 (90.3) | 12,978 (90.3) | ||
| Region | 0.27 | 0.01 | ||||
| Midwest | 8917 (17.9) | 4428 (23.4) | 3195 (22.2) | 3211 (22.3) | ||
| Northeast | 4234 (8.5) | 2202 (11.6) | 1533 (10.7) | 1500 (10.4) | ||
| South | 22,345 (45.0) | 6102 (32.2) | 4984 (34.7) | 4955 (34.5) | ||
| West | 14,202 (28.6) | 6199 (32.7) | 4658 (32.4) | 4704 (32.7) | ||
| Calendar year | 1.07 | 0.02 | ||||
| 2013 | 817 (1.6) | 2954 (15.6) | 817 (5.7) | 782 (5.4) | ||
| 2014 | 1632 (3.3) | 2446 (12.9) | 1479 (10.3) | 1545 (10.8) | ||
| 2015 | 2404 (4.8) | 2765 (14.6) | 1978 (13.8) | 1961 (13.6) | ||
| 2016 | 3838 (7.7) | 2745 (14.5) | 2336 (16.3) | 2336 (16.3) | ||
| 2017 | 6330 (12.7) | 2881 (15.2) | 2690 (18.7) | 2667 (18.6) | ||
| 2018 | 10,442 (21.0) | 2182 (11.5) | 2135 (14.9) | 2140 (14.9) | ||
| 2019 | 12,187 (24.5) | 1708 (9.0) | 1709 (11.9) | 1694 (11.8) | ||
| 2020 | 12,048 (24.2) | 1250 (6.6) | 1226 (8.5) | 1245 (8.7) | ||
| Comorbidities, n (%) | ||||||
| CKD stage 3 | 42,893 (86.3) | 14,612 (77.2) | 0.25 | 11,633 (81.0) | 11,575 (80.5) | 0.01 |
| CKD stage 4 | 5542 (11.2) | 3164 (16.7) | 0.25 | 2098 (14.6) | 2131 (14.8) | 0.01 |
| CKD stage 5 | 1263 (2.5) | 1155 (6.1) | 0.25 | 639 (4.4) | 664 (4.6) | 0.01 |
| Hypertension | 47,852 (96.3) | 17,945 (94.8) | 0.07 | 13,717 (95.5) | 13,653 (95.0) | 0.02 |
| Hyperlipidemia | 42,525 (85.6) | 15,724 (83.1) | 0.07 | 12,012 (83.6) | 11,990 (83.4) | 0.004 |
| Ischemic heart disease | 29,408 (59.2) | 11,564 (61.1) | 0.04 | 8747 (60.9) | 8694 (60.5) | 0.01 |
| Ischemic stroke | 13,083 (26.3) | 5503 (29.1) | 0.06 | 3994 (27.8) | 4006 (27.9) | 0.002 |
| Combined comorbidity score, mean (SD) | 7.36 (3.09) | 7.56 (2.98) | 0.07 | 7.50 (3.16) | 7.48 (3.01) | 0.004 |
| Frailty score, mean (SD) | 0.13 (0.06) | 0.18 (0.09) | 0.65 | 0.16 (0.07) | 0.16 (0.08) | 0.004 |
| Falls | 9167 (18.4) | 2584 (13.6) | 0.13 | 2199 (15.3) | 2158 (15.0) | 0.01 |
| Osteoporosis without fractures | 5380 (10.8) | 1721 (9.1) | 0.06 | 1362 (9.5) | 1368 (9.5) | 0.001 |
| Osteoporosis-related characteristics, n (%) | ||||||
| Bone mineral density scan | 3477 (7.0) | 1181 (6.2) | 0.03 | 930 (6.5) | 917 (6.4) | 0.004 |
| Osteoarthritis | 20,016 (40.3) | 6256 (33.0) | 0.15 | 5059 (35.2) | 5093 (35.4) | 0.01 |
| Vitamin D deficiency | 11,845 (23.8) | 4034 (21.3) | 0.06 | 3068 (21.4) | 3125 (21.7) | 0.01 |
| Other drug dispensings, n (%) | ||||||
| ACE inhibitors and ARBs | 33,251 (66.9) | 11,931 (63.0) | 0.08 | 9273 (64.5) | 9239 (64.3) | 0.01 |
| Antiarrhythmics | 10,623 (21.4) | 3912 (20.7) | 0.02 | 2968 (20.7) | 2952 (20.5) | 0.003 |
| Antiplatelet agents | 10,918 (22.0) | 3935 (20.8) | 0.03 | 3003 (20.9) | 2992 (20.8) | 0.002 |
| β blockers | 38,802 (78.1) | 14,645 (77.4) | 0.02 | 11,199 (77.9) | 11,179 (77.8) | 0.003 |
| Calcium channel blockers | 26,127 (52.6) | 9259 (48.9) | 0.07 | 7159 (49.8) | 7133 (49.6) | 0.004 |
| Diuretics | 34,100 (68.6) | 13,528 (71.5) | 0.06 | 10,157 (70.7) | 10,189 (70.9) | 0.01 |
| Proton pump inhibitors | 18,223 (36.7) | 6977 (36.9) | 0.004 | 5220 (36.3) | 5190 (36.1) | 0.004 |
| Statins | 35,456 (71.3) | 13,081 (69.1) | 0.05 | 10,064 (70.0) | 9988 (69.5) | 0.01 |
| Health care utilization, n (%) | ||||||
| Emergency room visit | 31,013 (62.4) | 11,848 (62.6) | 0.004 | 8974 (62.4) | 8947 (62.3) | 0.004 |
| Hospitalization | 32,714 (65.8) | 13,370 (70.6) | 0.10 | 9812 (68.3) | 9870 (68.7) | 0.01 |
| Length of stay, d, mean (SD) | 8.82 (14.93) | 12.10 (19.85) | 0.19 | 10.67 (19.03) | 10.68 (16.96) | 0.001 |
| No. of distinct medications, mean (SD) | 15.66 (7.17) | 15.34 (7.23) | 0.04 | 15.27 (7.04) | 15.29 (7.26) | 0.002 |
| Visit to cardiologist in the past 365 d | 43,034 (86.6) | 15,983 (84.4) | 0.06 | 12,216 (84.9) | 12,239 (85.0) | 0.004 |
| Visit to nephrologist in the past 365 d | 19,600 (39.4) | 8920 (47.1) | 0.16 | 6296 (43.7) | 6343 (44.1) | 0.01 |
| Laboratory results, median (IQR) | ||||||
| Serum creatinine (mg/dl) | 1.47 (1.17–2.04) | 1.36 (1.12–1.74) | 0.21 | 1.41 (1.15–1.86) | 1.44 (1.16–1.94) | 0.05 |
| eGFR (ml/min per 1.73 m2) | 43.63 (32.42–54.15) | 40.33 (26.85–52.38) | 0.18 | 42.30 (29.89–53.41) | 41.05 (28.25–52.69) | 0.07 |
| Missing, n (%) | 24,742 (49.78) | 11,360 (60.01) | 6612 (46.0) | 7209 (50.2) | ||
ACE, angiotensin-converting enzyme; ARB, angiotensin II, receptor blocker; ASD, absolute standardized difference; DOAC, direct oral anticoagulant; GI, gastrointestinal; IQR, interquartile range; NSAID, nonsteroidal anti-inflammatory drug.
As-Treated Analysis
In the matched population, 556 fracture events were captured across DOAC and warfarin initiators in the 365-day as-treated analysis. The IR of fracture per 1000 person-years was 33.14 among patients with CKD using DOACs and 29.59 in those using warfarin. Cumulative incidence curves in the matched population indicated similar risks of fracture for DOAC and warfarin initiators (Supplemental Figure 5). The HR of fracture comparing DOACs with warfarin was 1.12 (95% CI, 0.95 to 1.32). The IRD of fracture per 1000 person-years was 3.55 (95% CI, −1.67 to 8.76) for the same comparison (Table 2).
Table 2.
Hazard ratios of nonvertebral fracture in patients with CKD and atrial fibrillation initiating direct oral anticoagulants versus warfarin in the 1:1 propensity score-matched population
| Analysis | Treatment | No. of Patients | Median Follow-Up Time, d (IQR) | No. of Non-Vertebral Fracture Events | IR per 1000 Person-Years | HR (95% CI) | IRD (95% CI) |
|---|---|---|---|---|---|---|---|
| 365-d as-treated | DOACs | 14,370 | 237 (98–365) | 301 | 33.14 | 1.12 (0.95 to 1.32) | 3.55 (−1.67 to 8.76) |
| Warfarin | 14,370 | 209 (91–365) | 255 | 29.59 | Ref | Ref | |
| 365-d intention-to-treat | DOACs | 14,387 | 237 (103–365) | 299 | 32.66 | 1.11 (0.94 to 1.31) | 3.21 (−1.93 to 8.35) |
| Warfarin | 14,387 | 218 (107–365) | 262 | 29.45 | Ref | Ref |
CI, confidence interval; DOAC, direct oral anticoagulant; HR, hazard ratio; IQR, interquartile range; IR, incidence rate; IRD, incidence rate difference.
Intention-to-Treat Analysis
In the 365-day intention-to-treat analysis, we identified 561 fracture events across DOAC and warfarin initiators in the matched population. The IR of fracture per 1000 person-years was 32.66 among patients with CKD using DOACs and 29.45 in those using warfarin. Cumulative incidence curves indicated similar risks of fracture for DOAC and warfarin initiators in the matched population (Supplemental Figure 6). The HR of fracture comparing DOACs with warfarin was 1.11 (95% CI, 0.94 to 1.31). The IRD of fracture per 1000 person-years was 3.21 (95% CI, −1.93 to 8.35) for the same comparison (Table 2).
Secondary Analyses
When evaluating hip fracture, we identified 118 events across DOAC and warfarin initiators. The IR of hip fracture per 1000 person-years was 6.55 among patients with CKD using DOACs and 6.68 among those using warfarin. Cumulative incidence curves indicated similar risks of hip fracture for DOAC and warfarin initiators in the matched population (Supplemental Figures 7 and 8). The HR of hip fracture comparing DOACs with warfarin was 0.98 (95% CI, 0.68 to 1.41; Figure 2). The IRD of hip fracture per 1000 person-years was −0.13 (95% CI, −2.52 to 2.25).
Figure 2.
HRs and 95% CIs of secondary outcomes comparing DOACs and warfarin in the 1:1 propensity score-matched population. CI, confidence interval; HR, hazard ratio.
We captured 3122 all-cause mortality events in inpatient settings across DOAC and warfarin initiators in the matched population. The IR of all-cause mortality per 1000 person-years was 166.17 among patients with CKD using DOACs and 183.40 among those using warfarin. Cumulative incidence curves indicated a lower risk of all-cause mortality for DOAC versus warfarin initiators (Supplemental Figures 9 and 10). The HR of all-cause mortality comparing DOACs to warfarin was 0.91 (95% CI, 0.85 to 0.98) in the 365-day as-treated analysis (Figure 2). The IRD of all-cause mortality per 1000 person-years was −17.23 (95% CI, −29.49 to −4.96). Results were similar after excluding patients with ICD codes for valvular atrial fibrillation (Supplemental Table 3).
After re-estimation of the propensity score, we retained 12,174 patients in the apixaban and warfarin arms, respectively, and 8557 patients in the rivaroxaban and warfarin arms, respectively, after 1:1 matching (Supplemental Tables 4 and 5). Similar to the primary analysis, no significant differences in fracture and hip fracture risks were observed when evaluating apixaban and rivaroxaban individually relative to warfarin. However, point estimates suggested a slight increase in fracture risk for apixaban and a minor decrease in risk for rivaroxaban compared with warfarin (Supplemental Table 6). Results for all-cause mortality indicated a no statistically significant association for apixaban and rivaroxaban versus warfarin initiation, respectively (Supplemental Table 6).
Discussion
In this 1:1 propensity score-matched cohort study of patients with CKD and atrial fibrillation, we observed similar rates of nonvertebral fracture and hip fracture in patients initiating DOACs and warfarin. DOAC use was also associated with a lower risk of all-cause mortality compared with warfarin in this population. Results for all-cause mortality were similar when evaluating apixaban and rivaroxaban individually to warfarin, and no significant differences in nonvertebral fracture and hip fracture risks were observed for each comparison. However, point estimates suggested a slightly higher fracture risk association for apixaban and a slightly lower fracture risk association for rivaroxaban relative to warfarin, respectively. Findings for nonvertebral fracture and hip fracture may not be interpretable because of the small numbers of events in these analyses.
Our cohort study is the first to evaluate an important clinical question in patients with CKD, who comprise at least 14% of the US population and are often excluded from studies because of their high burden of comorbidity.24 Prior studies evaluating fracture risk have only compared DOACs and warfarin in patients with atrial fibrillation and without a focus on patients with CKD.2,9 In addition, fracture risk is of particular concern in the CKD population with their high propensity for mineral and bone disorders (i.e., CKD-mineral and bone disorder), a complex condition characterized by the confluence of reduced kidney function; dysregulated calcium, phosphate, and vitamin D homeostasis; abnormal parathyroid hormone regulation; and other factors.25 A US claims-based cohort study found that the use of alternative DOACs was associated with a lower risk of incident hip fracture relative to warfarin.9 Another study conducted using electronic health record data in Hong Kong found similar results in patients newly diagnosed with atrial fibrillation.2 Our results may differ from those observed in previous studies because we evaluated the comparative risk of fracture in a high-risk group of patients with moderate to severe kidney damage. The baseline risk of fracture is already elevated in these patients, particularly in those who initiated DOACs, which may have contributed to the slightly higher fracture risk in this group. Indeed, patients who initiated DOACs versus warfarin were older and had a higher history of falls, osteoarthritis, obesity, and overweight, which have been shown to be strong risk factors for fracture.26 Nevertheless, the higher risk may not be clinically meaningful such that the initiation of DOACs versus warfarin may not substantially change the risk of the outcome, or both agents may increase outcome risk to the same extent in this population. Overall, we had limited power to detect changes in adverse event risk due to the few observed nonvertebral fracture and hip fracture events in this small group of patients with moderate to advanced CKD.
In secondary analyses, individual comparisons suggested a slightly higher risk of fracture for apixaban and minor lower fracture risk for rivaroxaban when compared with warfarin. These results may not be interpretable because of the small number of patients, particularly in the rivaroxaban analysis, that may have led to unstable estimates with poor precision. In addition, there may be an imbalance in baseline fracture risk based on a higher prevalence of risk factors for fracture, including older age, history of falls, osteoarthritis, obesity, and overweight, in apixaban initiators that was not observed to such an extent in the rivaroxaban group.26
Initiation of DOACs was associated with a lower risk of all-cause mortality relative to warfarin. These findings are consistent with prior studies that evaluated mortality and other outcomes in association with anticoagulant therapy.27,28 We also found that, while DOAC initiators were slightly older, cardiovascular-related comorbidities were higher in patients with CKD initiating warfarin before matching. It is possible that these conditions may be strong risk factors for subsequent mortality. Advanced CKD is associated with several cardiovascular outcomes, including cardiovascular mortality, which reflects underlying elevated risk in this population due to diabetes, hypertension, bone mineral metabolism, and other factors.29 Future studies may seek to evaluate major adverse cardiovascular events as an outcome or incorporate information on cause of death associated with DOAC initiation to better elucidate comparative risks of outcome events of interest in this population.
Our results have important implications for clinical practice. In this commercially insured population, most patients had diagnosis codes for CKD stages 3 and 4. These patients may already have frequent encounters with healthcare providers, and further discussions regarding the use anticoagulant therapy may be warranted to reduce the risk of adverse events. For patients with moderate to severe kidney damage and atrial fibrillation, close monitoring may be needed at the time of anticoagulant initiation. Based on findings from our secondary analyses, patients initiating warfarin may benefit from greater observation with respect to other risk factors for mortality.
Our study has several limitations. We defined exposure to anticoagulants based on dispensing dates and days' supply, but we were unable to assess if patients took these medications as prescribed. Patients were classified into CKD stages based on ICD diagnosis codes rather than eGFR because laboratory results were missing for 50–60% of eligible patients. Patients with CKD initiating DOACs and warfarin were slightly different from one another with respect to cardiovascular-related comorbid conditions before matching, which may reflect residual confounding in favor of DOACs in the comparative analyses. Further adjustment for important covariates that may inform fracture risk, such as calcium, phosphate, and parathyroid hormone levels, may be needed to the extent that these data are available. All-cause mortality was defined based on events captured in inpatient encounters, and it is possible that mortality events outside of these settings were not captured well. However, outcome misclassification is unlikely to be differential across the treatment groups, and relative measures of association are not expected to be biased. Although we could rule out large changes in adverse event risk, we may have been underpowered to detect minor increases or decreases in the risks of nonvertebral fracture and hip fracture because of the small number of events in the 1:1 matched analysis. Alternative confounding adjustment approaches, such as propensity score weighting or n:1 propensity score matching, may be considered to retain more outcome events at the expense of a potential increase in bias. Evaluation of long-term anticoagulant use beyond 1 year may be needed to accrue a greater number of events to quantify comparative fracture risks in this population of patients with CKD. Anticoagulant use and fracture risk may be of more clinical interest in patients with CKD stages 4 and 5 because guidelines for anticoagulation are unclear and few studies have evaluated comparative safety and efficacy of DOACs and warfarin in this population.30 While we were underpowered to conduct subgroup analyses by CKD stage, future studies may seek to understand risks in patients with more advanced CKD.
The results of this nationwide cohort study suggest that, in patients with moderate to advanced CKD and atrial fibrillation, initiation of DOACs and warfarin have similar risks of non-vertebral fracture and hip fracture. However, DOAC use relative to warfarin is associated with a lower risk of all-cause mortality in this population. Close monitoring of patients with CKD and atrial fibrillation may be warranted at the time of anticoagulant initiation to prevent adverse clinical outcomes.
Supplementary Material
Acknowledgments
The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclosures
Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C101.
Funding
N.F. Khan: National Institute of Arthritis and Musculoskeletal and Skin Diseases (T32AR05585). J.M. Paik: National Institute of Arthritis and Musculoskeletal and Skin Diseases (AR075117). S.C. Kim: National Institutes of Health (K24 AR078959). K. Bykov: National Institutes of Health (K01AG068365).
Author Contributions
Conceptualization: Nazleen F. Khan, Seoyoung C. Kim, Su Been Lee.
Formal analysis: Nazleen F. Khan.
Methodology: Nazleen F. Khan.
Supervision: Katsiaryna Bykov, Seoyoung C. Kim, Julie M. Paik.
Writing – original draft: Nazleen F. Khan.
Writing – review & editing: Katsiaryna Bykov, Nazleen F. Khan, Seoyoung C. Kim, Su Been Lee, Julie M. Paik.
Data Sharing Statement
Data cannot be shared. A data use agreement is required for each of the data sources included in our study. Our data use agreements do not permit us to share patient-level source data or data derivatives with individuals and institutions not covered under the data use agreements. The databases used in this study are accessible to other researchers by contacting the data providers and acquiring data use agreements/licenses.
Supplemental Material
This article contains the following supplemental material online at http://links.lww.com/CJN/C100.
Supplemental Table 1. Eligibility criteria for inclusion in new user and active comparator cohort study.
Supplemental Table 2. Baseline characteristics of patients with CKD and atrial fibrillation initiating DOACs or warfarin before and after 1:1 propensity score matching (2013–2020).
Supplemental Table 3. As-treated results comparing DOACs with warfarin after exclusion of patients with valvular atrial fibrillation.
Supplemental Table 4. Baseline characteristics of patients with CKD and atrial fibrillation initiating apixaban or warfarin before and after 1:1 propensity score matching (2013–2020).
Supplemental Table 5. Baseline characteristics of patients with CKD and atrial fibrillation initiating rivaroxaban or warfarin before and after 1:1 propensity score matching (2013–2020).
Supplemental Figure 1. Propensity score distributions before matching in as-treated analysis.
Supplemental Figure 2. Propensity score distributions after matching in as-treated analysis.
Supplemental Figure 3. Propensity score distributions before matching in intention-to-treat analysis.
Supplemental Figure 4. Propensity score distributions after matching in intention-to-treat analysis.
Supplemental Figure 5. Cumulative incidence curves for fracture in as-treated analysis.
Supplemental Figure 6. Cumulative incidence curves for fracture in intention-to-treat analysis.
Supplemental Figure 7. Cumulative incidence curves for hip fracture in as-treated analysis.
Supplemental Figure 8. Cumulative incidence curves for hip fracture in intention-to-treat analysis.
Supplemental Figure 9. Cumulative incidence curves for all-cause mortality in as-treated analysis.
Supplemental Figure 10. Cumulative incidence curves for all-cause mortality in intention-to-treat analysis.
Supplemental Table 6. Results comparing individual DOACs with warfarin for all outcomes.
References
- 1.Huang HK Peng CC Lin SM, et al. Fracture risks in patients treated with different oral anticoagulants: a systematic review and meta-analysis. J Am Heart Assoc. 2021;10(7):e019618. doi: 10.1161/JAHA.120.019618 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Lau WCY Cheung CL Man KKC, et al. Association between treatment with apixaban, dabigatran, rivaroxaban, or warfarin and risk for osteoporotic fractures among patients with atrial fibrillation: a population-based cohort study. Ann Intern Med. 2020;173(1):1–9. doi: 10.7326/M19-3671 [DOI] [PubMed] [Google Scholar]
- 3.Xie X Liu Y Li J, et al. Fracture risks in patients with atrial fibrillation treated with different oral anticoagulants: a meta-analysis and systematic review. Age Ageing. 2022;51(1):afab264. doi: 10.1093/ageing/afab264 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Hsu CY, Chen LR, Chen KH. Osteoporosis in patients with chronic kidney diseases: a systemic review. Int J Mol Sci. 2020;21(18):6846. doi: 10.3390/ijms21186846 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Khairallah P, Nickolas TL. Management of osteoporosis in CKD. Clin J Am Soc Nephrol. 2018;13(6):962–969. doi: 10.2215/CJN.11031017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Alem AM Sherrard DJ Gillen DL, et al. Increased risk of hip fracture among patients with end-stage renal disease. Kidney Int. 2000;58(1):396–399. doi: 10.1046/j.1523-1755.2000.00178.x [DOI] [PubMed] [Google Scholar]
- 7.Kim SM, Long J, Montez-Rath M, Leonard M, Chertow GM. Hip fracture in patients with non-dialysis-requiring chronic kidney disease. J Bone Miner Res. 2016;31(10):1803–1809. doi: 10.1002/jbmr.2862 [DOI] [PubMed] [Google Scholar]
- 8.Marietta M, Coluccio V, Boriani G, Luppi M. Effects of Anti-vitamin k oral anticoagulants on bone and cardiovascular health. Eur J Intern Med. 2020;79:1–11. doi: 10.1016/j.ejim.2020.05.032 [DOI] [PubMed] [Google Scholar]
- 9.Lutsey PL Norby FL Ensrud KE, et al. Association of anticoagulant therapy with risk of fracture among patients with atrial fibrillation. JAMA Intern Med. 2020;180(2):245–253. doi: 10.1001/jamainternmed.2019.5679 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Parker K Hartemink J Saha A, et al. A systematic review of the efficacy and safety of anticoagulants in advanced chronic kidney disease. J Nephrol. 2022;35(8):2015–2033. doi: 10.1007/s40620-022-01413-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Agnelli G Buller HR Cohen A, et al. Oral apixaban for the treatment of acute venous thromboembolism. N Engl J Med. 2013;369(9):799–808. doi: 10.1056/NEJMoa1302507 [DOI] [PubMed] [Google Scholar]
- 12.Bauersachs RM Lensing AW Prins MH, et al. Rivaroxaban versus enoxaparin/vitamin K antagonist therapy in patients with venous thromboembolism and renal impairment. Thromb J. 2014;12:25. doi: 10.1186/1477-9560-12-25 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Goldhaber SZ Schulman S Eriksson H, et al. Dabigatran versus warfarin for acute venous thromboembolism in elderly or impaired renal function patients: pooled analysis of RE-COVER and RE-COVER II. Thromb Haemost. 2017;117(11):2045–2052. doi: 10.1160/TH17-03-0176 [DOI] [PubMed] [Google Scholar]
- 14.Buller HR Decousus H Grosso MA, et al.; Hokusai-VTE Investigators. Edoxaban versus warfarin for the treatment of symptomatic venous thromboembolism. N Engl J Med. 2013;369(15):1406–1415. doi: 10.1056/NEJMoa1306638 [DOI] [PubMed] [Google Scholar]
- 15.Bohula EA Giugliano RP Ruff CT, et al. Impact of renal function on outcomes with edoxaban in the ENGAGE AF-TIMI 48 trial. Circulation. 2016;134(1):24–36. doi: 10.1161/CIRCULATIONAHA.116.022361 [DOI] [PubMed] [Google Scholar]
- 16.Fox KA Piccini JP Wojdyla D, et al. Prevention of stroke and systemic embolism with rivaroxaban compared with warfarin in patients with non-valvular atrial fibrillation and moderate renal impairment. Eur Heart J. 2011;32(19):2387–2394. doi: 10.1093/eurheartj/ehr342 [DOI] [PubMed] [Google Scholar]
- 17.Hijazi Z Hohnloser SH Oldgren J, et al. Efficacy and safety of dabigatran compared with warfarin in relation to baseline renal function in patients with atrial fibrillation: a RE-LY (Randomized Evaluation of Long-term Anticoagulation Therapy) trial analysis. Circulation. 2014;129(9):961–970. doi: 10.1161/CIRCULATIONAHA.113.003628 [DOI] [PubMed] [Google Scholar]
- 18.Hohnloser SH Hijazi Z Thomas L, et al. Efficacy of apixaban when compared with warfarin in relation to renal function in patients with atrial fibrillation: insights from the ARISTOTLE trial. Eur Heart J. 2012;33(22):2821–2830. doi: 10.1093/eurheartj/ehs274 [DOI] [PubMed] [Google Scholar]
- 19.Paik JM Patorno E Zhuo M, et al. Accuracy of identifying diagnosis of moderate to severe chronic kidney disease in administrative claims data. Pharmacoepidemiol Drug Saf. 2022;31(4):467–475. doi: 10.1002/pds.5398 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399–424. doi: 10.1080/00273171.2011.568786 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Rassen JA, Shelat AA, Myers J, Glynn RJ, Rothman KJ, Schneeweiss S. One-to-many propensity score matching in cohort studies. Pharmacoepidemiol Drug Saf. 2012;21(suppl 2):69–80. doi: 10.1002/pds.3263 [DOI] [PubMed] [Google Scholar]
- 22.Sturmer T Webster-Clark M Lund JL, et al. Propensity score weighting and trimming strategies for reducing variance and bias of treatment effect estimates: a simulation study. Am J Epidemiol. 2021;190(8):1659–1670. doi: 10.1093/aje/kwab041 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Xu Y, Cheung YB, Lam KF, Tan SH, Milligan P. A simple approach to the estimation of incidence rate difference. Am J Epidemiol. 2010;172(3):334–343. doi: 10.1093/aje/kwq099 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.United States Renal Data System. 2022 Annual Data Report, 2023. Accessed January 31, 2023. https://usrds-adr.niddk.nih.gov/2022/chronic-kidney-disease/1-ckd-in-the-general-population [Google Scholar]
- 25.Moe SM, Drueke T, Lameire N, Eknoyan G. Chronic kidney disease-mineral-bone disorder: a new paradigm. Adv Chronic Kidney Dis. 2007;14(1):3–12. doi: 10.1053/j.ackd.2006.10.005 [DOI] [PubMed] [Google Scholar]
- 26.Rinonapoli G Pace V Ruggiero C, et al. Obesity and bone: a complex relationship. Int J Mol Sci. 2021;22(24):13662. doi: 10.3390/ijms222413662 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Lopez-Lopez JA Sterne JAC Thom HHZ, et al. Oral anticoagulants for prevention of stroke in atrial fibrillation: systematic review, network meta-analysis, and cost effectiveness analysis. BMJ. 2017;359(8132):j5058. doi: 10.1136/bmj.j5058 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Alcusky M, Tjia J, McManus DD, Hume AL, Fisher M, Lapane KL. Comparative safety and effectiveness of direct-acting oral anticoagulants versus warfarin: a national cohort study of nursing home residents. J Gen Intern Med. 2020;35(8):2329–2337. doi: 10.1007/s11606-020-05777-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Matsushita K, Ballew SH, Wang AY, Kalyesubula R, Schaeffner E, Agarwal R. Epidemiology and risk of cardiovascular disease in populations with chronic kidney disease. Nat Rev Nephrol. 2022;18(11):696–707. doi: 10.1038/s41581-022-00616-6 [DOI] [PubMed] [Google Scholar]
- 30.Rogula S, Gasecka A, Mazurek T, Navarese EP, Szarpak L, Filipiak KJ. Safety and efficacy of DOACs in patients with advanced and end-stage renal disease. Int J Environ Res Public Health. 2022;19(3):1436. doi: 10.3390/ijerph19031436 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data cannot be shared. A data use agreement is required for each of the data sources included in our study. Our data use agreements do not permit us to share patient-level source data or data derivatives with individuals and institutions not covered under the data use agreements. The databases used in this study are accessible to other researchers by contacting the data providers and acquiring data use agreements/licenses.