|Home | About | Journals | Submit | Contact Us | Français|
Atrial fibrillation (AF) substantially increases the risk of ischemic stroke but this risk varies among individual patients with AF. Existing risk stratification schemes have limited predictive ability. Chronic kidney disease is a major cardiovascular risk factor, but whether it independently increases the risk for ischemic stroke in persons with AF is unknown.
We examined how chronic kidney disease (reduced glomerular filtration rate or proteinuria) affects risk of thromboembolism off anticoagulation in patients with AF. We estimated glomerular filtration rate (eGFR) using the Modification of Diet in Renal Disease equation and proteinuria from urine dipstick results found in laboratory databases. Patient characteristics, warfarin use, and thromboembolic events were ascertained from clinical databases, with validation of thromboembolism by chart review.
During 33,165 person-years off anticoagulation among 10,908 patients with atrial fibrillation, we observed 676 incident thromboembolic events. After adjustment for known risk factors for stroke and other confounders, proteinuria increased the risk of thromboembolism by 54% (relative risk [RR] 1.54, 1.29 to 1.85) and there was a graded, increased risk of stroke associated with progressively lower level of eGFR compared with eGFR ≥60 (in units of ml/min/1.73 m2): RR 1.16 (95% CI: 0.95−1.40) for eGFR 45−59 and RR 1.39 (95% CI: 1.13−1.71) for eGFR <45 (P=0.0082 for trend).
. Chronic kidney disease increases the risk of thromboembolism in AF independent of other risk factors. Knowing the level of kidney function and presence of proteinuria may improve risk stratification for decision-making about the use of antithrombotic therapy for stroke prevention in AF.
Atrial fibrillation (AF) currently affects more than 2.2 million Americans1, 2 and independently increases the risk of ischemic stroke by approximately fourfold.3 This risk varies according to several demographic and clinical characteristics.4 However, existing stroke risk stratification schemes for AF are limited in their predictive ability.5-7 Identification of additional incremental risk factors for ischemic stroke in this population is needed to improve the anticoagulation decision in patients with AF.4, 7
End-stage renal disease (ESRD) requiring renal replacement therapy is associated with progressive vascular atherosclerosis, metabolic abnormalities, and prothrombotic tendencies,8 and it is considered a potent risk factor for ischemic stroke in the general population based on limited data.9 Previous studies have yielded conflicting data on the impact of AF on outcomes in patients with ESRD.10, 11 Chronic kidney disease (CKD), defined as reduced glomerular filtration rate and/or proteinuria,12 markedly increases the risk for cardiovascular events in the general population,13-16 although conflicting data exist for its association specifically with ischemic stroke in the absence of AF.14, 17, 18 Even less is known about the incremental effect of CKD and its severity on the risk of ischemic stroke in the setting of AF.
To address these issues, we examined the independent effect of two measures of CKD, reduced glomerular filtration rate and proteinuria, on the risk of thromboembolism off anticoagulation therapy in a large cohort of adults with nonvalvular AF.
Assembly of the AnTicoagulation and Risk Factors In Atrial Fibrillation (ATRIA) cohort has been described in detail previously.19 Briefly, we identified all patients aged 18 years or older with diagnosed AF within Kaiser Permanente of Northern California, a large integrated healthcare delivery system, based on physician-assigned diagnoses of AF found in ambulatory visit (International Classification of Diseases, Ninth Edition [ICD-9] code 427.31) and electrocardiographic databases between July 1, 1996 and December 31, 1997. We focused on patients with presumed chronic nonvalvular AF by excluding patients with diagnosed mitral stenosis or valvular repair or replacement, transient perioperative AF, or concomitant hyperthyroidism.19
This study was approved by institutional review boards of the collaborating institutions. Waiver of informed consent was obtained given the nature of the study.
Kidney function was assessed in two ways: level of estimated glomerular filtration rate (eGFR) and presence of proteinuria. To assess level of eGFR, we identified all outpatient serum creatinine tests (range: 0.1−20.0 mg/dl) performed among cohort members found in health plan laboratory databases between January 1, 1995 and September 30, 2003. Baseline level of eGFR was defined based on the nearest measurement up to 12 months before the index date, or up to 90 days after index date if no prior measurement within 12 months was available. All Kaiser Permanente of Northern California laboratories used the same serum creatinine assay during the study period, with 85% of tests performed by a single regional laboratory. Because we were interested in the absolute level of kidney function associated with the risk of thromboembolism, we calibrated serum creatinine test results with the Cleveland Clinic Laboratory used to develop the Modification of Diet in Renal Disease (MDRD) glomerular filtration rate estimating equation.20, 21 We next calculated at baseline and throughout follow-up each patient's eGFR using the abbreviated four-variable MDRD equation: eGFR (ml/min/1.73m2) = 186 × [serum Cr (mg/dL)] −1.154 × [age] −0.203 × [0.742 if female] × [1.212 if black].20, 21 Using a modification of the National Kidney Foundation classification for CKD,12, 13 level of eGFR was categorized as: ≥60, 45−59, and <45 ml/min/1.73 m2.
Proteinuria was defined as a urine dipstick protein result of ≥1+ (approximately 30 mg/dL or higher) in the absence of potential urinary tract infection (i.e., concomitant positive urine nitrite or esterase) found in laboratory databases.13 Receipt of renal replacement therapy (peritoneal dialysis, hemodialysis, or kidney transplant) was identified from a health plan ESRD treatment registry.13 We excluded patients with prior kidney transplant since we were interested in the effect of de novo CKD.
We used administrative databases for information on patient age, sex, and self-reported race/ethnicity updated through August 2008. U.S. Census block data from the 2000 survey were used to classify cohort members’ educational attainment and annual household income status.13 As previously described,1, 19 we searched automated inpatient, outpatient, laboratory, and pharmacy databases during the five years before index date to identify the following known or putative risk factors for ischemic stroke in AF: previous ischemic stroke, diagnosed heart failure, known coronary heart disease, and hypertension. We also used a validated longitudinal health plan diabetes registry to identify patients with diabetes mellitus.1, 19 We have previously validated the use of these approaches to ascertain these selected diagnoses in this cohort.19 Each of these characteristics was also updated during follow-up through September 30, 2003 using the same methods.
To examine the role of kidney function measures on risk of thromboembolism, we focused on periods of follow-up off anticoagulation. To identify periods off warfarin therapy, we assigned use of warfarin based on a combination of information from prescriptions and outpatient international normalized ratio (INR) measurements found in pharmacy and laboratory databases, respectively. Longitudinal warfarin exposure was based on number of days supply per prescription and intervening INRs. For any two consecutive prescriptions with a gap of up to 60 days, a patient was considered continually on warfarin. For gaps longer than 60 days, we considered the patient continually on warfarin if there were intervening INR measurements at least every 42 days. Otherwise, the patient was considered off warfarin from day 31 after the end date of the first prescription until the start date of the next prescription. This grace period of 30 days at the end of each warfarin period was given since changes in warfarin dose are common. We previously demonstrated the validity of this approach based on chart review.22
From index date through September 30, 2003, we prospectively searched hospitalization and billing claims databases for potential thromboembolic events based on relevant ICD-9 codes for stroke and peripheral embolism found in the primary discharge diagnosis position as previously described.22 Potential events were adjudicated by a Clinical Outcomes Committee composed of physicians based on medical records review, with a final decision made by a consulting neurologist if consensus was not reached by the committee. A valid ischemic stroke was defined as a documented acute neurological deficit lasting >24 hours that was not explained by other etiologies (i.e., primary hemorrhage, trauma, infection, or vasculitis). A peripheral thromboembolic event was considered valid if an embolus was demonstrated by radiographic imaging, intra-operative exam, or pathological findings in the absence of underlying atherosclerotic disease.
Follow-up occurred through September 30, 2003, disenrollment, or death. Membership gaps lasting greater than 90 days without evidence of interim medical care was considered disenrollment from the health plan, and patients were censored at the last known membership date. Disenrollment due to death through September 30, 2003 was ascertained from hospital databases, health plan member reporting, Social Security Administration vital status files, and the California state death certificate registry.23
All analyses were conducted using SAS, version 9.1 (Cary, N.C.). Continuous variables were compared across levels of eGFR using the Spearman correlation coefficient, and categorical variables were compared using the Cochran-Armitage test for trend. Event rates were initially calculated using log-linear (Poisson regression) models with a generalized estimating equations approach to account for the same subjects contributing person-years off warfarin and at different levels of eGFR and/or the presence or absence of documented proteinuria over time.
To evaluate the independent association of level of eGFR and proteinuria with risk of thromboembolism, we performed multivariable Poisson regression models with generalized estimating equations that adjusted for sociodemographic characteristics and known or putative risk factors for stroke in AF (i.e., prior ischemic stroke, heart failure, hypertension, diabetes mellitus, and coronary heart disease).24 We included eGFR in the model as a time-varying categorical variable (≥60 [reference], 45−59, and <45 ml/min/1.73 m2) and documented proteinuria as a time-dependent dichotomous variable. We also examined for a possible interaction between eGFR level and proteinuria on risk of thromboembolism in a separate multivariable model. Analyses using only baseline kidney function showed a similar trend for eGFR and proteinuria, so only results from models using time-varying kidney function status are reported
The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
Among 13,535 adults with nonvalvular AF and no prior kidney transplant, mean age was 71.6 years at entry and 42.8% were women. A total of 85.2% of the cohort had available information on baseline eGFR (Table 1). At baseline, 18.5% of men and 25.9% of women of the subset with known kidney function at entry had reduced eGFR of 45−59 ml/min/1.73 m2, and 10.6% men and 12.9% women had an eGFR <45 ml/min/1.73 m2. Only 0.9% of men and 0.7% of women of the subset with known baseline kidney function met criteria for ESRD at study entry, defined as eGFR <15 ml/min/1.73 m2 or receiving maintenance dialysis. A total of 6444 subjects had available data on urine dipstick protein assessments on or before index date.
Patients with lower levels of eGFR at baseline were older than those with higher eGFR, and the distribution of race and gender across levels of eGFR varied significantly (Table 2). At progressively lower baseline levels of eGFR, the prevalence was higher for prior ischemic stroke, diagnosed heart failure, diagnosed hypertension, diabetes, and known coronary disease. Proteinuria was documented in a higher proportion of patients with lower eGFR level. Of note, baseline warfarin use was slightly lower with lower level of eGFR (Table 2).
The median number of outpatient serum creatinine results per patient from baseline through the end of follow-up was 8 (interquartile range 4 to 14) and 98.7% of the patients had at least one known eGFR during follow-up. Overall, 9200 (68.0% of the cohort) members had one or more urine dipstick protein assessments during follow-up. There were a total of 33,165 person-years of follow-up off warfarin therapy contributed by 10,908 patients (80.6% of the cohort). Among the 10,908 patients who contributed time off warfarin therapy, 10,614 (97.3%) had at least one eGFR result and 8,746 (80.2%) had at least one available urine dipstick proteinuria result at baseline or during follow-up. A total of 1604 subjects were censored due to disenrollment from the health plan and 6 patients due to incident kidney transplant during follow-up.
During follow-up, there were a total of 676 validated thromboembolic events (637 ischemic strokes, 39 other thromboembolism) during periods off warfarin therapy. This included 344 events among individuals with eGFR ≥60 ml/min/1.73 m2, 168 events among those with eGFR 45−59 ml/min/1.73m2, 149 events in patients with eGFR <45 ml/min/1.73 m2 and 15 events with unknown kidney function. The rate of thromboembolism off warfarin increased significantly with lower eGFR, ranging from 1.63 per 100 person-years for eGFR ≥60 ml/min/1.73 m2 to 4.22 per 100 person-years for eGFR <45 ml/min/1.73 m2 (Figure). Rates of thromboembolism were higher with documented proteinuria at every level of eGFR (Table 3). Of note, for the 676 observed thromboembolic events that occurred off warfarin, the median (interquartile range) time between measurement of eGFR and event was 61 (10−173) days; the median (interquartile range) time between measurement of dipstick proteinuria and event was 298 (93−646) days.
In multivariable analysis, compared with eGFR ≥60 ml/min/1.73 m2, there was a higher adjusted rate of thromboembolism associated with declining kidney function ranging from a 16% increased rate for eGFR 45−59 ml/min/1.73 m2 up to 39% for eGFR <45 ml/min/1.73 m2 (P=0.0082 for trend across eGFR categories) (Table 4). Results were not significantly different when patients with ESRD were excluded from the subgroup of patients with eGFR <45 ml/min/1.73 m2. In addition, documented proteinuria was associated with a 54% increased adjusted rate of thromboembolism, even after controlling for level of eGFR, sociodemographic characteristics and known risk factors for stroke in AF (Table 4). There was no significant interaction between level of eGFR and proteinuria (P=0.082).
Within a large, ambulatory cohort of adults with AF, we found that lower level of eGFR was associated with a graded, increased risk of ischemic stroke and other systemic embolism independent of known risk factors in AF. Documented proteinuria also significantly increased the risk of thromboembolism even after accounting for level of eGFR and other confounders.. The magnitude of association for having dipstick proteinuria and for levels of eGFR below 45 ml/min/1.73 m2 were in the range seen for other known thromboembolic risk factors in AF such as older age (per decade), heart failure, hypertension, diabetes, and female gender.4, 25
Previous studies involving CKD and AF have primarily focused on the epidemiology and impact of AF on outcomes in patients with ESRD treated with chronic dialysis rather than the role of kidney dysfunction on thromboembolism in the setting of AF. In patients with ESRD, the rate of ischemic stroke is high overall,9 and risk factors for stroke such as hypertension, diabetes mellitus, coronary disease, and heart failure are common.26 Furthermore, a study of 215 ESRD patients on chronic dialysis with no history of AF or cardiovascular disease observed that 33% of subjects had left atrial appendage thrombi identified by transesophageal echocardiography.27 In patients receiving chronic dialysis, AF appears to be a common acute and chronic complication,11, 28 and some but not all studies suggest that AF further increases the risk of ischemic stroke significantly after accounting for other known risk factors.10, 29 Even less is known about the epidemiology and outcomes of AF in patients with CKD not yet requiring renal replacement therapy. Thus, our study provides novel insights in demonstrating that the presence and severity of CKD (as reflected by reduced eGFR and proteinuria) are associated with a higher risk of ischemic stroke and other thromboembolism in patients with AF, independent of known risk factors for stroke in AF.
Why would kidney dysfunction increase the risk of stroke in AF? AF itself leads to disorganized contraction of the atria, with a decrease in atrial blood flow and a resultant increase in blood pooling and stasis, especially in the left atrial appendage.30 AF may also lead to a hypercoagulable state through various metabolic pathways, although delineation of these specific factors and their contribution to thromboembolism are less clear.31-36 Collectively, these effects predispose to thrombus formation and subsequent systemic emboli. ESRD treated with chronic dialysis is associated with a prothrombotic state, including increased levels of various endothelium-related factors such as plasminogen activator inhibitor-1,37-39 and von Willebrand factor,40, 41 as well as abnormalities in various coagulation factor levels and activity (e.g., fibrinogen; fibrinopeptide A; thromboplastin; and factors VII, VIII, IX-XII)8 and inflammation (e.g., C-reactive protein and interleukin-6).42-44 ESRD is further associated with greater arterial media calcification45, 46 and arterial stiffness,47, 48 which are associated with higher cardiovascular event rates. Procoagulant and inflammatory pathways are also abnormally affected in mild-to-moderate CKD independent of other coexisting illnesses and vascular risk factors15, 49 along with alterations in arterial compliance.50 Of interest, AF-associated thrombus is predominantly composed of fibrin rather than platelets and therefore tends to resemble that found in venous thromboembolic disease rather than typical arterial thromboses. Among 19,073 middle-aged and elderly subjects followed for mean 11.8 years, baseline eGFR 15−59 ml/min/1.73 m2 was associated with a significantly increased risk of venous thromboembolism (adjusted hazard ratio 1.71, 95% CI: 1.18−2.49) after adjustment for age, gender, race, diabetes, hypertension, body mass index and factor VIIIc, while there was no association with serum cystatin C.51 With regards to proteinuria, among 298 consecutive patients with nephrotic syndrome (proteinuria ≥3.5 g/d), compared with proteinuria 3.5−4.8 g/d, urinary protein excretion of ≥8.2 g/d was associated with a significantly increased risk of venous thromboembolism (hazard ratio 5.2, 95% CI: 1.1−23.0), although other potential confounders were not accounted for in the analysis.52 Thus, CKD may contribute to an increased risk of ischemic stroke and other thromboembolism in patients with AF by augmenting the underlying prothrombotic state through several different pathophysiological pathways.
In addition to the large, diverse cohort of patients with AF, our study was strengthened by long-term, longitudinal information on kidney function, ability to characterize periods off anticoagulants, as well as accounting for the presence of other known risk factors for stroke during follow-up. This allowed for more careful evaluation of the independent contribution of eGFR and proteinuria to thromboembolic risk off anticoagulation. Furthermore, all thromboembolic events were validated by medical records review using standardized criteria and blinded to kidney function status. Our study also had several limitations. A small proportion of subjects did not have known kidney function, and information on use of aspirin therapy over time was not available. We also did not characterize the subtype of ischemic stroke, although the majority of strokes in the setting of AF are cardioembolic.53 The use of the MDRD equation for estimating GFR has not been validated in non-Black ethnic minorities, so misclassification may be present in such patients in our study sample. Even though we controlled for the presence or absence of hypertension and diabetes, systematic data were unavailable on the severity of these conditions, which may be associated with level of kidney function and also can increase the risk of ischemic stroke. Finally, although our study was conducted within an insured sample of patients in Northern California that is very representative of the local and statewide population, our results may not be fully generalizable to all other health care settings and populations in the U.S.
In sum, we found that proteinuria and reduced eGFR were associated with a higher rate of thromboembolism in patients with nonvalvular AF independent of other stroke risk factors in this setting. Our study suggests that patients with AF should be evaluated for both level of eGFR and presence of proteinuria, as this information may help to improve risk stratification and decision-making about the use of antithrombotic therapy to prevent stroke in patients with AF.
Dr. Go has received relevant research grant support from the National Institute on Aging (R01 AG15478) , National Heart, Lung and Blood Institute (U19 HL091179), and Johnson & Johnson. Dr. Fang has received relevant research support from the National Institute on Aging (K23 AG028978). Dr. Singer has received relevant research grant support from the National Institute on Aging (R01 AG15478) and Daiichi Sankyo; he has been a consultant for Boehringer Ingelheim, Bayer Healthcare, AstraZeneca, Sanofi Aventis, Daiichi Sankyo and Johnson & Johnson; and he has received speaking honoraria from Pfizer and Bristol Meyers Squibb. The other authors have no relevant disclosures.
Supported by: Public Health Services research grant AG15478 from the National Institute on Aging, and the Edith and Eliot B. Shoolman Fund of Massachusetts General Hospital