Abstract
New staging systems for CKD account for both reduced eGFR and albuminuria; whether each measure associates with greater risk of hemorrhage is unclear. In this retrospective cohort study (2002–2010), we grouped 516,197 adults ≥40 years old by eGFR (≥90, 60 to <90, 45 to <60, 30 to <45, 15 to <30, or <15 ml/min per 1.73 m2) and urine albumin-to-creatinine ratio (ACR; >300, 30–300, or <30 mg/g) to examine incidence of hemorrhage. The 3-year cumulative incidence of hemorrhage increased 20-fold across declining eGFR and increasing urine ACR groupings (highest eGFR/lowest ACR: 0.5%; lowest eGFR/highest ACR: 10.1%). Urine ACR altered the association of eGFR with hemorrhage (P<0.001). In adjusted models using the highest eGFR/lowest ACR grouping as the referent, patients with eGFR=15 to <30 ml/min per 1.73 m2 had adjusted relative risks of hemorrhage of 1.9 (95% confidence interval [95% CI], 1.5 to 2.4) with the lowest ACR and 3.7 (95% CI, 3.0 to 4.5) with the highest ACR. Patients with the highest eGFR/highest ACR had an adjusted relative risk of hemorrhage of 2.3 (95% CI, 1.8 to 2.9), comparable with the risk for patients with the lowest eGFR/lowest ACR. The associations attenuated but remained significant after adjustment for anticoagulant and antiplatelet use in patients ≥66 years old. The risk of hemorrhage differed by urine ACR in high risk subgroups. Our data show that declining eGFR and increasing albuminuria each independently increase hemorrhage risk. Strategies to reduce hemorrhage events among patients with CKD are warranted.
Keywords: chronic kidney disease, clinical epidemiology, vascular disease, albuminuria, hemorrhage, bleeding risk
The number of patients with CKD is increasing worldwide, with an estimated prevalence ranging from 23% to 36% in adults ≥64 years of age.1 CKD and albuminuria are well recognized risk factors for thrombosis.2 The risk of thrombosis undergoes a stepwise increase with declining eGFR and increasing albuminuria.3 Paradoxically, CKD has also been associated with an increased risk of hemorrhage.4–11 With ESRD, one in seven patients on dialysis will experience a major hemorrhage within 3 years of dialysis initiation.12 Recent evidence suggests that albuminuria itself, as a marker of vascular injury and stress, is associated with cerebral hemorrhage independent of kidney function.13–15 However, whether albuminuria is associated with an increased risk of hemorrhage in other end organs or an effect modifier of the association between kidney function and hemorrhage remains uncertain. Characterization of the risk of hemorrhage across stages of CKD with varying levels of albuminuria and among high-risk subgroups will aid in the development of appropriate screening and preventative strategies. In this regard, we examined the association of hemorrhage with eGFR and albuminuria in a large population–based sample.
Results
Baseline Characteristics
Of 516,197 patients who met the inclusion and exclusion criteria (Supplemental Figure 1), 58% (n=300,646) of adults ages ≥40 years old had reduced kidney function (eGFR<90 ml/min per 1.73 m2). The baseline characteristics for patients within each level of albuminuria are presented in Table 1. Compared with an albumin-to-creatinine ratio (ACR) <30 mg/g, individuals with an ACR≥30 mg/g were more commonly men and older. Hypertension and cardiovascular disease (coronary artery disease, myocardial infarction, congestive heart failure, coronary revascularization, ischemic stroke, and amputation) were all more prevalent among those with a higher ACR. A prior history of hemorrhage, atrial fibrillation, deep vein thrombosis, and pulmonary embolism all increased with increasing ACR. In individuals with eGFR>60 ml/min per 1.73 m2, there was a lower prevalence of ACR>30 mg/g. Medication information was available for 176,608 (34%) individuals ≥66 years of age. Among these older individuals, patients with a higher ACR were more likely to be prescribed oral anticoagulants, antiplatelet agents, and proton pump inhibitors.
Table 1.
Baseline characteristics of the study cohort
| Characteristics | ACR<30 mg/g | ACR=30–300 mg/g | ACR>300 mg/g |
|---|---|---|---|
| Total, N (%) | 408,125 (79.1) | 87,872 (17.0) | 20,200 (3.9) |
| Age at index, yr, mean (SD) | 59.5 (11.8) | 63.9 (13.0)a | 63.9 (13.1)a |
| Age ≥75 yr old, N (%) | 50,875 (12.5) | 21,452 (24.4)a | 4954 (24.5)a |
| Men, N (%) | 203,256 (49.8) | 46,466 (52.9) | 11,910 (59.0)a |
| Income, lowest quintile, N (%) | 76,911 (18.8) | 19,371 (22.0) | 4798 (23.8) |
| Residential status, rural, N (%) | 32,726 (8.0) | 6770 (7.7) | 1553 (7.7) |
| Comorbidities | |||
| Prior major hemorrhage, N (%) | 9893 (2.4) | 2674 (3.0) | 797 (3.9) |
| Gastrointestinal endoscopy,b N (%) | 100,561 (24.6) | 19,600 (22.3) | 4377 (21.7) |
| Diabetes mellitus, N (%) | 175,985 (43.1) | 53,743 (61.2)a | 13,676 (67.7)a |
| Atrial fibrillation/flutter, N (%) | 6585 (1.6) | 3676 (4.2)a | 982 (4.9)a |
| Coronary artery disease, N (%) | 83,185 (20.4) | 25,630 (29.2)a | 6911 (34.2)a |
| Myocardial infarction, N (%) | 8205 (2.0) | 3037 (3.5) | 989 (4.9)a |
| Congestive heart failure, N (%) | 16,655 (4.1) | 8200 (9.3)a | 2893 (14.3)a |
| Coronary revascularization, N (%) | 11,746 (2.9) | 3683 (4.2) | 1082 (5.4) |
| Hypertension, N (%) | 240,339 (58.9) | 63,062 (71.8)a | 15,669 (77.6)a |
| Heart valve replacement, N (%) | 896 (0.2) | 359 (0.4) | 89 (0.4) |
| Deep vein thrombosis, N (%) | 681 (0.2) | 269 (0.3) | 103 (0.5) |
| Pulmonary embolism, N (%) | 849 (0.2) | 272 (0.3) | 77 (0.4) |
| Lower leg amputation, N (%) | 354 (0.1) | 330 (0.4) | 219 (1.1) |
| Ischemic stroke, N (%) | 2766 (0.7) | 1384 (1.6) | 447 (2.2) |
| eGFR, 60 ml/min per 1.73 m2, mean (SD) | 84.6 (18.2) | 78.2 (23.3)a | 66.8 (27.8)a |
| eGFR≥60 ml/min per 1.73 m2, N (%) | 365,470 (89.5) | 67,786 (77.1)a | 11,623 (57.5)a |
| eGFR=45–59 ml/min per 1.73 m2, N (%) | 29,880 (7.3) | 10,908 (12.4)a | 3418 (16.9)a |
| eGFR=30–44 ml/min per 1.73 m2, N (%) | 10,497 (2.6) | 6566 (7.5)a | 2986 (14.8)a |
| eGFR=15–29 ml/min per 1.73 m2, N (%) | 2193 (0.5) | 2405 (2.7)a | 1817 (9.0)a |
| eGFR<15 ml/min per 1.73 m2, N (%) | 85 (0.0) | 207 (0.2) | 356 (1.8)a |
| Medication use among individuals ≥66 years of age (total =176,608) | |||
| Age ≥66 yr old, N (%) | 126,554 (71.7) | 40,692 (23.0) | 9362 (5.3) |
| Oral anticoagulants, N (%) | 7030 (5.6) | 4150 (10.2)a | 1056 (11.3)a |
| Oral antiplatelets, N (%) | 15,007 (11.9) | 6549 (16.1) | 1763 (18.8)a |
| Proton pump inhibitors, N (%) | 25,159 (19.9) | 8022 (19.7) | 1987 (21.2) |
Data presented as numbers (percentages), except for age and eGFR, which are presented as means (SDs). Total =516,197. Standardized differences were used to compare the CKD stage under investigation with the referent group ACR<30 mg/g. Standardized differences are less sensitive to sample size than traditional hypothesis tests. They provide a measure of the difference between groups divided by the pooled SD; a value >10% is interpreted as a meaningful difference between groups.
Significant difference (>10%).
Includes either of upper endoscopy or colonoscopy.
Hemorrhage Events and Risk of All–Cause Major Hemorrhage by eGFR and Urine ACR
Among 516,197 adults, 6153 (1.2%) experienced a hemorrhage event during the study period. The association of eGFR and urine ACR was examined by treating both eGFR and ACR as continuous variables (Table 2). When adjusted for demographics, year of cohort entry, comorbidities, and albuminuria, the risk of hemorrhage decreased by 9% (95% confidence interval [95% CI], 8% to 11%) per each 10-ml/min per 1.73 m2 increase in eGFR. In an analysis adjusted for demographics, year of cohort entry, comorbidities, and eGFR, the risk of hemorrhage increased by 2% (95% CI, 1% to 2%) per each 88.5-mg/g (10 mg/mmol) increase in urine ACR. Similar results were seen in an analysis restricted to individuals ages ≥66 years old adjusted for medication use (Table 2).
Table 2.
Association of eGFR and urine ACR with hemorrhage when examined as a continuous variable
| Cohort and Variable | Unadjusted RR (95% CI)a | Adjusted RR (95% CI) | P Value |
|---|---|---|---|
| Total cohort | |||
| eGFR, ml/min per 1.73 m2 | 0.73 (0.72 to 0.74) | 0.91 (0.89 to 0.92)b | <0.001 |
| Urine ACR, mg/g | 1.02 (1.01 to 1.02) | 1.02 (1.01 to 1.02)c | <0.001 |
| Age ≥66 yr old | |||
| eGFR, ml/min per 1.73 m2 | 0.81 (0.80 to 0.83) | 0.91 (0.89 to 0.93)b,d | <0.001 |
| Urine ACR, mg/g | 1.02 (1.02 to 1.03) | 1.02 (1.01 to 1.02)c,d | <0.001 |
RR reported for a 10-U increase of eGFR or 88.5-U increase of ACR. For ACR, to convert from milligrams per gram to milligrams per millimole, multiply by 0.113.
Adjusted for age (per year), sex, income quintile (lowest referent), ischemic stroke, myocardial infarction, coronary artery disease, coronary revascularization, deep venous thrombosis, atrial fibrillation, hypertension, congestive heart failure, diabetes, prior hemorrhage, residential status, year of index date (2002 referent), and urine ACR (continuous).
Adjusted for age (per year), sex, income quintile (lowest referent), ischemic stroke, myocardial infarction, coronary artery disease, coronary revascularization, deep venous thrombosis, atrial fibrillation, hypertension, congestive heart failure, diabetes, prior hemorrhage, residential status year of index date (2002 referent), and eGFR (continuous).
Analysis restricted to individuals ages ≥66 years old. Also adjusted for anticoagulant, antiplatelet, and proton pump inhibitor use.
The interaction between eGFR and urine ACR with hemorrhage was highly significant (multiplicative interaction term; P<0.001; additive interaction term, P=0.003), indicating that urine ACR was an effect modifier of the association of eGFR and hemorrhage.
The 3-year cumulative incidence, incidence rate (IR), and crude and adjusted relative risks (RRs) of hemorrhage by eGFR and urine ACR groupings are presented in Table 3 and Supplemental Table 1. Individuals in the lowest eGFR (<15 ml/min per 1.73 m2) and highest albuminuria (>300 mg/g) grouping had the highest risk of hemorrhage over 3 years (cumulative incidence, 10.1%; 95% CI, 7.0 to 13.2; IR, 42.4 per 1000 person- years; 95% CI, 30.9 to 58.5; adjusted RR, 5.5; 95% CI, 3.9 to 7.6; eGFR>90 ml/min per 1.73 m2 and ACR<30 mg/g as the referent group).
Table 3.
Risk of all–cause major hemorrhage by eGFR and urine ACR
| CKD Stage by eGFR Categories | Albuminuria Categories, mg/g | ||
|---|---|---|---|
| <30 | 30–300 | >300 | |
| ≥90 ml/min per 1.73 m2 | |||
| IR per 1000 person-yr (95% CI) | 1.6 (1.5 to 1.7)a | 3.1 (2.7 to 3.4)b | 4.5 (3.6 to 5.8)c |
| Adjusted RRd (95% CI) | Referent | 1.6 (1.4 to 1.8)b | 2.3 (1.8 to 2.9)c |
| 60 to <90 ml/min per 1.73 m2 | |||
| IR per 1000 person-yr (95% CI) | 3.1 (2.9 to 3.2)a | 6.7 (6.2 to 7.2)b | 10.7 (9.3 to 12.2)c |
| Adjusted RRd (95% CI) | 1.0 (0.9 to 1.1)a | 1.6 (1.4 to 1.8)b | 2.5 (2.2 to 3.0)c |
| 45 to <60 ml/min per 1.73 m2 | |||
| IR per 1000 person-yr (95% CI) | 7.1 (6.6 to 7.7)b | 10.4 (9.4 to 11.6)c | 15.0 (12.7 to 17.6)e |
| Adjusted RRd (95% CI) | 1.4 (1.2 to 1.6)b | 1.7 (1.4 to 1.9)c | 2.6 (2.2 to 3.2)e |
| 30 to <45 ml/min per 1.73 m2 | |||
| IR per 1000 person-yr (95% CI) | 11.4 (10.3 to 12.7)c | 18.2 (16.3 to 20.2)e | 19.3 (16.5 to 22.5)e |
| Adjusted RRd (95% CI) | 1.7 (1.5 to 2.0)c | 2.3 (2.0 to 2.6)e | 2.8 (2.3 to 3.3)e |
| 15 to <30 ml/min per 1.73 m2 | |||
| IR per 1000 person-yr (95% CI) | 17.1 (14.1 to 20.8)e | 21.6 (18.3 to 25.5)e | 29.5 (25.0 to 34.7)e |
| Adjusted RRd (95% CI) | 1.9 (1.5 to 2.4)e | 2.4 (1.9 to 2.9)e | 3.7 (3.0 to 4.5)e |
| <15 ml/min per 1.73 m2 | |||
| IR per 1000 person-yr (95% CI) | 29.3 (13.3 to 64.4)e | 35.1 (22.0 to 56.0)e | 42.4 (30.9 to 58.5)e |
| Adjusted RRd (95% CI) | 3.0 (1.3 to 6.6)e | 3.5 (2.2 to 5.6)e | 5.5 (3.9 to 7.6)e |
Categories of eGFR and ACR on the basis of the 2012 Kidney Disease: Improving Global Outcomes (KDIGO) nomenclature, which classifies adults into four categories by CKD prognosis (low, moderate, high, or very high risk). ACR was determined by a random spot urine albumin-to-creatinine ratio.
KDIGO CKD risk group: low.
KDIGO CKD risk group: moderate.
KDIGO CKD risk group: high.
Adjusted for age (per year), sex, income quintile (lowest referent), ischemic stroke, myocardial infarction, coronary artery disease, coronary revascularization, deep venous thrombosis, atrial fibrillation, hypertension, congestive heart failure, diabetes, prior hemorrhage, residential status and year of index date (2002 referent).
KDIGO CKD risk group: very high.
For individuals with a urine ACR <30 mg/g, the cumulative incidence of hemorrhage increased 14.2-fold (0.5% versus 7.1%; P for trend <0.001) and for individuals with a urine ACR >300 mg/g, the cumulative incidence of hemorrhage increased 7.8-fold (1.3% versus 10.1%) as eGFR declined. Among individuals with normal kidney function (eGFR≥90 ml/min per 1.73 m2), the cumulative incidence of hemorrhage increased 2.6- fold (0.5% versus 1.3%) and among individuals with kidney failure (eGFR<15 ml/min per 1.73 m2), the cumulative incidence of hemorrhage increased 1.4-fold (7.1% versus 10.1%) as urine ACR increased.
Similar results were seen for the IR of hemorrhage per 1000 person-years, wherein the IR increased more with declining eGFR in the lowest compared with the highest albuminuria grouping (18.3-fold; IR, 29.3 versus 1.6 per 1000 person-years for eGFR<15 versus ≥90 ml/min per 1.73 m2 and ACR<30 mg/g; 9.4-fold; IR, 42.4 versus 4.5 per 1000 person-years for eGFR<15 versus ≥90 ml/min per 1.73 m2 and ACR>300 mg/g). The IR of hemorrhage increased more with increasing albuminuria in the higher eGFR groupings (2.8-fold; IR, 4.5 versus 1.6 per 1000 person-years for ACR>300 versus <30 mg/g and eGFR>90 ml/min per 1.73 m2; 1.4-fold; IR, 42.4 versus 29.3 for ACR>300 versus <30 mg/g and eGFR<15 ml/min per 1.73 m2).
The increased hemorrhage risk seen with declining eGFR attenuated; however, it still persisted on adjustment for demographic factors, year of cohort entry, and comorbidities. For example, the unadjusted and adjusted RRs for hemorrhage with declining eGFR in the urine ACR <30 mg/g grouping were RR, 1.9; 95% CI, 1.7 to 2.0 and adjusted RR, 1.0; 95% CI, 0.9 to 1.1 for eGFR=60 to <90 ml/min per 1.73 m2 and RR, 14.7; 95% CI, 6.8 to 31.8 and adjusted RR, 3.0; 95% CI, 1.3 to 6.6 for eGFR<15 ml/min per 1.73 m2 (eGFR≥90 ml/min per 1.73 m2 and ACR<30 mg/g as the referent group). For the association of hemorrhage with albuminuria, less attenuation was observed on adjustment. For example, the unadjusted and adjusted RRs for hemorrhage with increasing urine ACR in the eGFR≥90 ml/min per 1.73 m2 grouping were RR, 2.8; 95% CI, 2.2 to 3.6 and adjusted RR, 2.3; 95% CI, 1.8 to 2.9 for ACR>300 mg/g (ACR<30 mg/g and eGFR≥90 ml/min per 1.73 m2 as the referent group). Similar results were seen in an analysis restricted to individuals ages ≥66 years old adjusted for medications (oral anticoagulants, antiplatelets, and proton pump inhibitors; ACR<30 mg/g grouping: unadjusted RR, 3.5; 95% CI, 2.7 to 4.4 and adjusted RR, 1.6; 95% CI, 1.2 to 2.0; eGFR=15 to <30 ml/min per 1.73 m2 [eGFR≥90 ml/min per 1.73 m2 and ACR<30 mg/g as the referent group]; eGFR≥90 ml/min per 1.73 m2 grouping: unadjusted RR, 3.0; 95% CI, 1.9 to 4.7 and adjusted RR, 2.5; 95% CI, 1.6 to 4.0 [ACR>300 mg/g]; ACR<30 mg/g and eGFR≥90 ml/min per 1.73 m2 as the referent group) (Supplemental Table 2).
Subgroup Analyses
The adjusted risk of hemorrhage increased with older age, a previous history of hemorrhage, diabetes, atrial fibrillation, ischemic stroke, and anticoagulant use (Table 4). Risk of hemorrhage increased in a graded fashion with increasing albuminuria grouping (adjusted RR, 3.2; 95% CI, 2.8 to 3.7) among individuals ages 40 to <66 years old (ACR>300 versus <30 mg/g), comorbidities associated with antiplatelet and anticoagulant use (adjusted RR, 2.3; 95% CI, 2.1 to 2.5) among individuals without a history of atrial fibrillation (ACR>300 versus <30 mg/g), and anticoagulant use (adjusted RR, 2.1; 95% CI, 1.9 to 2.3) among individuals not on anticoagulants (ACR>300 versus <30 mg/g) (Table 4).
Table 4.
The association of urine ACR and hemorrhage in selected risk groups
| Urine ACR, mg/g | Adjusted RR (95% CI) of All-Cause Hemorrhagea | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Age, yr | Diabetes | History of Hemorrhage | Atrial Fibrillation | Ischemic Stroke | Anticoagulant Useb | ||||||||
| 40 to <66 | 66–80 | >80 | No | Yes | No | Yes | No | Yes | No | Yes | No | Yes | |
| <30 | Referent | 2.67 (2.47 to 2.89) | 4.49 (4.03 to 5.00) | Referent | 1.28 (1.20 to 1.37) | Referent | 3.03 (2.72 to 3.38) | Referent | 1.59 (1.40 to 1.82) | Referent | 1.63 (1.37 to 1.94) | Referent | 1.77 (1.56 to 2.02) |
| 30–300 | 1.89 (1.70 to 2.11) | 3.85 (3.50 to 4.25) | 5.37 (4.76 to 6.05) | 1.51 (1.38 to 1.66) | 1.84 (1.79 to 1.99) | 1.49 (1.40 to 1.58) | 3.89 (3.39 to 4.47) | 1.50 (1.41 to 1.60) | 1.90 (1.65 to 2.19) | 1.48 (1.39 to 1.57) | 2.00 (1.65 to 2.43) | 1.37 (1.27 to 1.47) | 2.17 (1.89 to 2.49) |
| >300 | 3.18 (2.75 to 3.68) | 5.17 (4.53 to 5.91) | 7.00 (5.90 to 8.32) | 2.37 (2.06 to 2.73) | 2.64 (2.39 to 2.93) | 2.36 (2.17 to 2.58) | 3.33 (2.64 to 4.20) | 2.26 (2.07 to 2.47) | 2.47 (2.01 to 3.04) | 2.23 (2.05 to 2.43) | 2.07 (1.51 to 2.85) | 2.08 (1.87 to 2.31) | 2.21 (1.77 to 2.76) |
Adjusted for age (per year), sex, income quintile (lowest referent), ischemic stroke, myocardial infarction, coronary artery disease, coronary revascularization, deep venous thrombosis, atrial fibrillation, hypertension, congestive heart failure, diabetes, prior hemorrhage, residential status year of index date (2002 referent), and eGFR (≥90 ml/min per 1.73 m2 as the referent).
Also adjusted for anticoagulant, antiplatelet, and proton pump inhibitor use (analysis restricted to individuals ≥66 years of age).
Effect Modification of Risk Factors for Hemorrhage
The association of ACR and hemorrhage differed significantly by age (P<0.001 for interaction), previous history of hemorrhage (P<0.001 for interaction), atrial fibrillation (P=0.003 for interaction), ischemic stroke (P<0.01 for interaction), and anticoagulant use (P<0.001). Diabetes status did not alter the association of ACR and hemorrhage (P=0.25). Increasing categories of urine ACR were associated with larger relative increases in the adjusted risk of hemorrhage among younger individuals and those with an absence of a previous history of hemorrhage, atrial fibrillation, ischemic stroke, or anticoagulant use (Supplemental Table 3).
Discussion
In this administrative database study of a large adult CKD cohort, we found that the risk of major hemorrhage increased in a graded fashion with declining eGFR and increasing albuminuria. Abnormal kidney function and albuminuria both independently increased hemorrhage risk, and albuminuria was an effect modifier of the association between eGFR and hemorrhage. Albuminuria was more strongly associated with hemorrhage risk among younger individuals and individuals without other risk factors for hemorrhage, such as anticoagulant use. Our data illustrate the high risk of major hemorrhage among individuals with CKD defined by either eGFR or albuminuria.
The eGFR-independent effect of urine ACR on the risk of hemorrhage is a novel finding. Previous studies suggest that albuminuria is associated with an increased risk of hemorrhagic stroke.16–19 Albuminuria may represent an early marker of endothelial dysfunction and vascular damage, especially in strain vessels. These vessels are responsible for providing strong vascular tone and may be uniquely susceptible to hypertensive injury with subsequent endothelial disruption and hemorrhage.20 Because hypertension and hypertensive vascular injury are almost ubiquitous in the hemodialysis population, this presents a plausible mechanism of injury. Our study found that albuminuria independently increases hemorrhage risk and significantly modifies the association of eGFR with hemorrhage. Thus, reduction in albuminuria may provide a novel therapeutic target for preventing hemorrhage, an avenue of future investigation.
The increased risk of hemorrhage among patients with reduced eGFR or kidney failure has been previously reported. However, prior studies have been limited by variation in the definition of CKD and hemorrhage, the use of nonvalidated outcomes, and small or highly specific study populations. For example, in highly select patients with atrial fibrillation on oral anticoagulants, a component of the hypertension, abnormal renal and liver function, stroke, Bleeding, labile INRs, elderly, drugs or alcohol risk score for bleeding has been developed using CKD defined by the arbitrary serum creatinine >200 μmol/L.4,5,21–24 Other studies examining gastrointestinal hemorrhage in CKD used CKD diagnostic codes to define CKD, which are known to have limited sensitivity and specificity.10,25–28 Our study defined CKD on the basis of direct laboratory values of serum creatinine and urine ACR. Furthermore, the hemorrhagic outcomes used in our study were validated International Classification of Diseases (ICD) diagnostic codes and clinically meaningful, because our definition was restricted to events severe enough to require hospitalization. Several highly specific population studies have shown an increased risk of hemorrhage associated with CKD among patients prescribed antiplatelets after an acute coronary syndrome7,8,29 and an increased risk of hemorrhagic stroke6,9,30–32 and upper gastrointestinal hemorrhage associated with CKD.10,11,25–27 Furthermore, previous studies have been limited by the inclusion of few patients with higher stages of CKD6 or the sole inclusion of patients with kidney failure on dialysis.31,32 To our knowledge, our study is the first to report the risk of all–cause major hemorrhage across all stages of CKD at a broad population level, and our findings agree with and expand on previous findings by including a much larger sample size (n=516,197),11 patients with milder kidney dysfunction, and various subtypes of hemorrhage.
The relative contributions of eGFR and albuminuria to the risk of hemorrhage seem to differ. With increasing CKD stages, there were observed increases in mean age, the number of comorbidities, and anticoagulant/antiplatelet prescription, all of which are well recognized risk factors for hemorrhage.23,24 Adjustment for these factors attenuated the RR of hemorrhage associated with eGFR considerably. However, a significant graded increase still remained present. Increases in the RR associated with albuminuria did not attenuate to the same degree with adjustment. For example, with ACR<30 mg/g, the unadjusted RR of hemorrhage for eGFR=15–30 ml/min per 1.73 m2 is 9.6, which attenuated to 1.9 with full adjustment (referent: ACR<30 mg/g and eGFR≥90 ml/min per 1.73 m2). In comparison, with eGFR≥90 ml/min per 1.73 m2, the unadjusted RR of hemorrhage for ACR>300 mg/g is 2.8, which attenuated to 2.3 with full adjustment (referent: ACR<30 mg/g and eGFR≥90 ml/min per 1.73 m2). This illustrates the important contribution of age and comorbidities to hemorrhage risk with declining eGFR. Furthermore, the adjusted risk of hemorrhage for patients with normal eGFR and ACR>300 mg/g is surprisingly high (RR, 2.3) and comparable with a patient with an eGFR<15 ml/min per 1.73 m2 and minimal albuminuria (RR, 3.0; referent: ACR<30 mg/g and eGFR≥90 ml/min per 1.73 m2). The degree of risk for patients with high-level albuminuria and preserved eGFR is likely underappreciated and highly clinically relevant.
Our study has important limitations. Although we were able to adjust for a number of important confounders, residual confounding is always a possibility, and therefore, the association observed between hemorrhage and eGFR and albuminuria may not be causal. We were unable to adjust for serum albumin and anemia, which may increase bleeding risk.11,33 We did not have data on filled prescriptions for patients <65 years old and therefore, were unable to adjust for anticoagulant or antiplatelet use in younger patients. However, we adjusted for conditions associated with antiplatelet usage, such as myocardial infarction, ischemic stroke, coronary revascularization, atrial fibrillation, and deep venous thrombosis, which would partially reduce confounding by antiplatelet or anticoagulant use. Furthermore, the adjusted results in the primary cohort and the subcohort of patients ≥66 years of age, for whom we had medication information, were consistent. We lacked information on nonprescription acetyl-salicylic acid use in our databases. However, there is evidence to suggest that aspirin use alone may not be a significant contributor to increased bleeding risk in patients with CKD.11
In conclusion, our data show that the risk of hemorrhage increases in a graded fashion with declining eGFR and that albuminuria is an independent effect modifier. Albuminuria, as an easy, readily accessible, and inexpensive test, could provide useful information on hemorrhagic risk in patients initiating anticoagulation.
Concise Methods
Design and Setting
We conducted a population–based retrospective cohort study in the province of Ontario, Canada using health care databases housed at the Institute for Clinical Evaluative Sciences (ICES). Residents of Ontario have universal access to hospital care and physician services, and individuals ≥65 years of age have universal prescription coverage. The study was conducted according to a prespecified protocol that was approved by the Research Ethics Board at Sunnybrook Health Sciences Centre (Toronto, ON, Canada). The reporting of this study follows the STrengthening the Reporting of OBservational studies in Epidemiology guidelines for observational studies34 (Supplemental Appendix 1).
Data Sources
We ascertained patient characteristics, laboratory data, outcome data, and drug use from linked databases. Cerner and Gamma-Dynacare Databases were used to obtain outpatient laboratory data. Cerner is a hospital network in southwestern Ontario housing data from 11 hospitals. Gamma-Dynacare is a laboratory service provider that has outpatient laboratory information for individuals who had bloodwork drawn at any of their 148 collection sites in Ontario. Demographics and vital status information were obtained from the Ontario Registered Persons Database. Diagnostic and procedural information from all hospitalizations was determined using the Canadian Institute for Health Information Discharge Abstract Database (CIHI-DAD). Information was also obtained from the Ontario Health Insurance Plan database, which contains all health claims for inpatient and outpatient physician services. Prescription drug use was determined using the Ontario Drug Benefit Plan database, which contains highly accurate records of all outpatient prescriptions dispensed to individuals ages ≥65 years with an error rate <1%.35 We identified patients with a history of renal transplant (exclusion criteria) using the Canadian Organ Replacement Register. These datasets were linked using uniquely encoded identifiers and analyzed at the ICES. We have previously used these databases to research renal health outcomes and health services.12,36,37 Whenever possible, we defined patient characteristics and outcomes using validated codes (Supplemental Appendix 2).
Study Cohorts
The databases were reviewed from April 1, 2002 to March 31, 2010 for patient accrual. To be included, patients had to have at least one eGFR value and a urine ACR test performed 0–12 months before the date of the eGFR. For the urine ACR, the most recent measurement performed in the 12 months before the eGFR was taken. Urine ACR values after the eGFR measurement were not considered to eliminate potential immortal time bias. The median time between the eGFR and urine ACR was 0 days (interquartile range, 0–0). The date of the eGFR value was taken as the index date. We used the Chronic Kidney Disease Epidemiology Collaboration equation to calculate eGFR.39 We defined groupings of eGFR and urine ACR using thresholds described in the 2012 Kidney Disease: Improving Global Outcomes Guidelines (eGFR: ≥90 [normal kidney function], 60 to <90, 45 to <60, 30 to <45, 15 to <30, and <15 ml/min per 1.73 m2; urine ACR: <30, 30–300, and >300 mg/g).40 Cohort inclusion criteria included use of the first eligible inclusion date (for individuals with multiple dates) and adults ages ≥40 years old because of the small sample size of younger individuals. Adults who received a kidney transplant or chronic dialysis before their index date were excluded.
Known demographics and comorbid risk factors for hemorrhage were captured in the 5 years before the index date, including hypertension, prior stroke, prior major hemorrhage, diabetes, and cardiovascular disease.23 Because prescription medications information is only available for patients >65 years old in the Ontario Drug Benefit Plan database, we created a separate cohort of adults ≥66 years of age at index date. In Ontario, acetyl-salicylic acid is often available without a prescription and therefore, would not be universally captured in the Ontario Drug Benefit Plan. A lookback period of 120 days was used for baseline medication use. Specific medications are outlined in Supplemental Appendix 3.
Outcomes
We followed up all enrolled adults until there was evidence of hemorrhage associated with a hospital admission or death or a follow-up duration of 3 years after the index date (last potential follow-up date: March 31, 2013). A 3-year observation period was selected as a period of follow-up that could be reasonably adopted in a future interventional trial testing interventions to prevent hemorrhage. The following validated types of hemorrhage were included in the outcome of all hemorrhage: upper or lower gastrointestinal, intracerebral, subarachnoid, and other nontraumatic intracranial. Hospitalizations with a diagnosis of hemorrhage were identified using validated ICD revisions 9 (pre-2002) and 10 codes in the CIHI-DAD (Supplemental Appendix 2).41
Statistical Analyses
We used standardized differences to assess differences in baseline characteristics between adults with ACR<30 mg/g (referent) and those with lower stages of eGFR. Standardized differences describe differences between group means relative to the pooled SD and are less sensitive to large sample sizes than traditional hypothesis testing. A difference >10% is considered significant.42 We calculated the 3-year cumulative incidence of hospitalization with a diagnosis of hemorrhage (defined as the proportion of patients who experienced the event at least once within the 3 years of follow-up) and the IR (defined as the rate per 1000 person-years of follow-up). For the 3-year cumulative incidence and IRs of all–cause major hemorrhage, individuals were censored at the time of their first applicable hemorrhage event. When examining different types of hemorrhage, recurrent events were included if they contributed to a different subtype of hemorrhage. For example, an individual with an upper gastrointestinal hemorrhage and a subsequent subarachnoid hemorrhage would contribute one event for total hemorrhage but two events for each type of hemorrhage. The association of eGFR, urine ACR, and their interaction (eGFR×ACR) with hemorrhage was examined as both continuous and categorical variables in unadjusted and adjusted models using modified Poisson regression.43 The RRs of hemorrhage for each eGFR and urine ACR grouping were calculated using eGFR≥90 ml/min per 1.73 m2 and urine ACR<30 mg/g as the referent category. We adjusted for the following factors: age (per year), sex (referent: men), income quintile (lowest quintile referent), ischemic stroke, myocardial infarction, coronary artery disease, coronary revascularization, deep venous thrombosis, atrial fibrillation, hypertension, congestive heart failure, diabetes, prior hemorrhage, residential status, year of index date (referent: 2002), and medication usage (age >66 years old; n=176,608). We conducted all analyses with SAS software, version 9.4 (SAS Institute Inc., Cary, NC).
Disclosures
None.
Supplementary Material
Acknowledgments
This study was supported by the Institute for Clinical Evaluative Sciences (ICES) Western Site. The ICES is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). Core funding for the ICES Western Site is provided by the Academic Medical Organization of Southwestern Ontario (AMOSO), the Schulich School of Medicine and Dentistry (SSMD), Western University, and the Lawson Health Research Institute (LHRI). This project was conducted with members of the provincial ICES Kidney, Dialysis and Transplantation Research Program (www.ices.on.ca/kdt), which receives programmatic grant funding from the Canadian Institutes of Health Research (CIHI). A.O.M. received support from the Kidney Research Scientist Core Education and National Training Foundation. A.X.G. is supported by the Dr. Adam Linton Chair in Kidney Health Analytics. M.M.S. is supported by the Jindal Research Chair for the Prevention of Kidney Disease.
Parts of this material are on the basis of data and information compiled and provided by the CIHI. The opinions, results, and conclusions are those of the authors and independent from the funding/data sources. No endorsement by the ICES, the AMOSO, the SSMD, the LHRI, the CIHI, or the MOHLTC is intended or should be inferred.
Footnotes
Published online ahead of print. Publication date available at www.jasn.org.
This article contains supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2015050535/-/DCSupplemental.
References
- 1.Zhang QL, Rothenbacher D: Prevalence of chronic kidney disease in population-based studies: Systematic review. BMC Public Health 8: 117, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Wattanakit K, Cushman M, Stehman-Breen C, Heckbert SR, Folsom AR: Chronic kidney disease increases risk for venous thromboembolism. J Am Soc Nephrol 19: 135–140, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mahmoodi BK, Gansevoort RT, Næss IA, Lutsey PL, Brækkan SK, Veeger NJ, Brodin EE, Meijer K, Sang Y, Matsushita K, Hallan SI, Hammerstrøm J, Cannegieter SC, Astor BC, Coresh J, Folsom AR, Hansen JB, Cushman M: Association of mild to moderate chronic kidney disease with venous thromboembolism: Pooled analysis of five prospective general population cohorts. Circulation 126: 1964–1971, 2012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Abdelhafiz AH, Myint MP, Tayek JA, Wheeldon NM: Anemia, hypoalbuminemia, and renal impairment as predictors of bleeding complications in patients receiving anticoagulation therapy for nonvalvular atrial fibrillation: A secondary analysis. Clin Ther 31: 1534–1539, 2009 [DOI] [PubMed] [Google Scholar]
- 5.Apostolakis S, Guo Y, Lane DA, Buller H, Lip GY: Renal function and outcomes in anticoagulated patients with non-valvular atrial fibrillation: The AMADEUS trial. Eur Heart J 34: 3572–3579, 2013 [DOI] [PubMed] [Google Scholar]
- 6.Bos MJ, Koudstaal PJ, Hofman A, Breteler MM: Decreased glomerular filtration rate is a risk factor for hemorrhagic but not for ischemic stroke: The Rotterdam Study. Stroke 38: 3127–3132, 2007 [DOI] [PubMed] [Google Scholar]
- 7.Cuschieri JR, Drawz P, Falck-Ytter Y, Wong RC: Risk factors for acute gastrointestinal bleeding following myocardial infarction in veteran patients who are prescribed clopidogrel. J Dig Dis 15: 195–201, 2014 [DOI] [PubMed] [Google Scholar]
- 8.Fischer MJ, Ho PM, McDermott K, Lowy E, Parikh CR: Chronic kidney disease is associated with adverse outcomes among elderly patients taking clopidogrel after hospitalization for acute coronary syndrome. BMC Nephrol 14: 107, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Holzmann MJ, Aastveit A, Hammar N, Jungner I, Walldius G, Holme I: Renal dysfunction increases the risk of ischemic and hemorrhagic stroke in the general population. Ann Med 44: 607–615, 2012 [DOI] [PubMed] [Google Scholar]
- 10.Huang KW, Leu HB, Luo JC, Chan WL, Hou MC, Lin HC, Lee FY, Kuan YC: Different peptic ulcer bleeding risk in chronic kidney disease and end-stage renal disease patients receiving different dialysis. Dig Dis Sci 59: 807–813, 2014 [DOI] [PubMed] [Google Scholar]
- 11.Liang CC, Wang SM, Kuo HL, Chang CT, Liu JH, Lin HH, Wang IK, Yang YF, Lu YJ, Chou CY, Huang CC: Upper gastrointestinal bleeding in patients with CKD. Clin J Am Soc Nephrol 9: 1354–1359, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Sood MM, Bota SE, McArthur E, Kapral MK, Tangri N, Knoll G, Zimmerman D, Garg AX: The three-year incidence of major hemorrhage among older adults initiating chronic dialysis. Can J Kidney Health Dis 1: 21, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ninomiya T, Perkovic V, Verdon C, Barzi F, Cass A, Gallagher M, Jardine M, Anderson C, Chalmers J, Craig JC, Huxley R: Proteinuria and stroke: A meta-analysis of cohort studies. Am J Kidney Dis 53: 417–425, 2009 [DOI] [PubMed] [Google Scholar]
- 14.Ovbiagele B, Wing JJ, Menon RS, Burgess RE, Gibbons MC, Sobotka I, German L, Shara NM, Fernandez S, Jayam-Trouth A, Edwards DF, Kidwell CS: Association of chronic kidney disease with cerebral microbleeds in patients with primary intracerebral hemorrhage. Stroke 44: 2409–2413, 2013 [DOI] [PubMed] [Google Scholar]
- 15.Toyoda K: Cerebral small vessel disease and chronic kidney disease. J Stroke 17: 31–37, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Cho BH, Kim JT, Chang J, Choi KH, Park MS, Cho KH: Prediction of hemorrhagic transformation in acute ischaemic stroke by micro- and macroalbuminuria after intravenous thrombolysis. Eur J Neurol 20: 1145–1152, 2013 [DOI] [PubMed] [Google Scholar]
- 17.Rodríguez-Yáñez M, Castellanos M, Blanco M, Millán M, Nombela F, Sobrino T, Lizasoain I, Leira R, Serena J, Dávalos A, Castillo J: Micro- and macroalbuminuria predict hemorrhagic transformation in acute ischemic stroke. Neurology 67: 1172–1177, 2006 [DOI] [PubMed] [Google Scholar]
- 18.Mahmoodi BK, Yatsuya H, Matsushita K, Sang Y, Gottesman RF, Astor BC, Woodward M, Longstreth WT Jr., Psaty BM, Shlipak MG, Folsom AR, Gansevoort RT, Coresh J: Association of kidney disease measures with ischemic versus hemorrhagic strokes: Pooled analyses of 4 prospective community-based cohorts. Stroke 45: 1925–1931, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Aguilar MI, O’Meara ES, Seliger S, Longstreth WT Jr., Hart RG, Pergola PE, Shlipak MG, Katz R, Sarnak MJ, Rifkin DE: Albuminuria and the risk of incident stroke and stroke types in older adults. Neurology 75: 1343–1350, 2010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ito S, Nagasawa T, Abe M, Mori T: Strain vessel hypothesis: A viewpoint for linkage of albuminuria and cerebro-cardiovascular risk. Hypertens Res 32: 115–121, 2009 [DOI] [PubMed] [Google Scholar]
- 21.Kooiman J, van Rein N, Spaans B, van Beers KA, Bank JR, van de Peppel WR, del Sol AI, Cannegieter SC, Rabelink TJ, Lip GY, Klok FA, Huisman MV: Efficacy and safety of vitamin K-antagonists (VKA) for atrial fibrillation in non-dialysis dependent chronic kidney disease. PLoS One 9: e94420, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Limdi NA, Beasley TM, Baird MF, Goldstein JA, McGwin G, Arnett DK, Acton RT, Allon M: Kidney function influences warfarin responsiveness and hemorrhagic complications. J Am Soc Nephrol 20: 912–921, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY: A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: The Euro Heart Survey. Chest 138: 1093–1100, 2010 [DOI] [PubMed] [Google Scholar]
- 24.Suzuki M, Matsue Y, Nakamura R, Matsumura A, Hashimoto Y: Improvement of HAS-BLED bleeding score predictive capability by changing the definition of renal dysfunction in Japanese atrial fibrillation patients on anticoagulation therapy. J Cardiol 64: 482–487, 2014 [DOI] [PubMed] [Google Scholar]
- 25.Kuo CC, Kuo HW, Lee IM, Lee CT, Yang CY: The risk of upper gastrointestinal bleeding in patients treated with hemodialysis: A population-based cohort study. BMC Nephrol 14: 15, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Luo JC, Leu HB, Hou MC, Huang KW, Lin HC, Lee FY, Chan WL, Lin SJ, Chen JW: Nonpeptic ulcer, nonvariceal gastrointestinal bleeding in hemodialysis patients. Am J Med 126: 264.e25–264.e32, 2013 [DOI] [PubMed] [Google Scholar]
- 27.Luo JC, Leu HB, Huang KW, Huang CC, Hou MC, Lin HC, Lee FY, Lee SD: Incidence of bleeding from gastroduodenal ulcers in patients with end-stage renal disease receiving hemodialysis. CMAJ 183: E1345–E1351, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Vlasschaert ME, Bejaimal SA, Hackam DG, Quinn R, Cuerden MS, Oliver MJ, Iansavichus A, Sultan N, Mills A, Garg AX: Validity of administrative database coding for kidney disease: A systematic review. Am J Kidney Dis 57: 29–43, 2011 [DOI] [PubMed] [Google Scholar]
- 29.Palmer SC, Di Micco L, Razavian M, Craig JC, Perkovic V, Pellegrini F, Copetti M, Graziano G, Tognoni G, Jardine M, Webster A, Nicolucci A, Zoungas S, Strippoli GF: Effects of antiplatelet therapy on mortality and cardiovascular and bleeding outcomes in persons with chronic kidney disease: A systematic review and meta-analysis. Ann Intern Med 156: 445–459, 2012 [DOI] [PubMed] [Google Scholar]
- 30.Lee JG, Lee KB, Jang IM, Roh H, Ahn MY, Woo HY, Hwang HW: Low glomerular filtration rate increases hemorrhagic transformation in acute ischemic stroke. Cerebrovasc Dis 35: 53–59, 2013 [DOI] [PubMed] [Google Scholar]
- 31.Iseki K, Kinjo K, Kimura Y, Osawa A, Fukiyama K: Evidence for high risk of cerebral hemorrhage in chronic dialysis patients. Kidney Int 44: 1086–1090, 1993 [DOI] [PubMed] [Google Scholar]
- 32.Masson P, Kelly PJ, Craig JC, Lindley RI, Webster AC: Risk of stroke in patients with ESRD. Clin J Am Soc Nephrol 10: 1585–1592, 2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Hedges SJ, Dehoney SB, Hooper JS, Amanzadeh J, Busti AJ: Evidence-based treatment recommendations for uremic bleeding. Nat Clin Pract Nephrol 3: 138–153, 2007 [DOI] [PubMed] [Google Scholar]
- 34.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP STROBE Initiative : The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. Lancet 370: 1453–1457, 2007 [DOI] [PubMed] [Google Scholar]
- 35.Levy AR, O'Brien BJ, Sellors C, Grootendorst P, Willison D: Coding accuracy of administrative drug claims in the Ontario Drug Benefit database. Can J Clin Pharmacol 10: 67–71, 2003 [PubMed] [Google Scholar]
- 36.Molnar AO, Coca SG, Devereaux PJ, Jain AK, Kitchlu A, Luo J, Parikh CR, Paterson JM, Siddiqui N, Wald R, Walsh M, Garg AX: Statin use associates with a lower incidence of acute kidney injury after major elective surgery. J Am Soc Nephrol 22: 939–946, 2011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Naylor KL, McArthur E, Leslie WD, Fraser LA, Jamal SA, Cadarette SM, Pouget JG, Lok CE, Hodsman AB, Adachi JD, Garg AX: The three-year incidence of fracture in chronic kidney disease. Kidney Int 86: 810–818, 2014 [DOI] [PubMed] [Google Scholar]
- 38.Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF 3rd, Feldman HI, Kusek JW, Eggers P, Van Lente F, Greene T, Coresh J CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) : A new equation to estimate glomerular filtration rate. Ann Intern Med 150: 604–612, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.KDIGO : KDIGO 2012 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease. Kidney Int Suppl 3(1): 1–150, 2013 [DOI] [PubMed] [Google Scholar]
- 40.Arnason T, Wells PS, van Walraven C, Forster AJ: Accuracy of coding for possible warfarin complications in hospital discharge abstracts. Thromb Res 118: 253–262, 2006 [DOI] [PubMed] [Google Scholar]
- 41.Austin PC: Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat Theory Methods 38: 1228–1234, 2009 [Google Scholar]
- 42.Zou G: A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol 159: 702–706, 2004 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.