Stroke 3
Stroke 3
Stroke 3
n e w e ng l a n d j o u r na l
of
m e dic i n e
original article
A BS T R AC T
BACKGROUND
Both atrial fibrillation and chronic kidney disease increase the risk of stroke and
systemic thromboembolism. However, these risks, and the effects of antithrombotic
treatment, have not been thoroughly investigated in patients with both conditions.
METHODS
Using Danish national registries, we identified all patients discharged from the
hospital with a diagnosis of nonvalvular atrial fibrillation between 1997 and 2008.
The risk of stroke or systemic thromboembolism and bleeding associated with
nonend-stage chronic kidney disease and with end-stage chronic kidney disease
(i.e., disease requiring renal-replacement therapy) was estimated with the use of
time-dependent Cox regression analyses. In addition, the effects of treatment with
warfarin, aspirin, or both in patients with chronic kidney disease were compared
with the effects in patients with no renal disease.
RESULTS
Of 132,372 patients included in the analysis, 3587 (2.7%) had nonend-stage chronic
kidney disease and 901 (0.7%) required renal-replacement therapy at the time of inclusion. As compared with patients who did not have renal disease, patients with
nonend-stage chronic kidney disease had an increased risk of stroke or systemic
thromboembolism (hazard ratio, 1.49; 95% confidence interval [CI], 1.38 to 1.59;
P<0.001), as did those requiring renal-replacement therapy (hazard ratio, 1.83; 95%
CI, 1.57 to 2.14; P<0.001); this risk was significantly decreased for both groups of
patients with warfarin but not with aspirin. The risk of bleeding was also increased
among patients who had nonend-stage chronic kidney disease or required renalreplacement therapy and was further increased with warfarin, aspirin, or both.
CONCLUSIONS
Chronic kidney disease was associated with an increased risk of stroke or systemic
thromboembolism and bleeding among patients with atrial fibrillation. Warfarin
treatment was associated with a decreased risk of stroke or systemic thromboembolism among patients with chronic kidney disease, whereas warfarin and aspirin
were associated with an increased risk of bleeding. (Funded by the Lundbeck Foundation.)
625
The
n e w e ng l a n d j o u r na l
626
of
m e dic i n e
ME THODS
REGISTRY DATA SOURCES
In this Danish cohort study, we linked individuallevel data from national registries, using the personal registration number provided to all Danish
residents. The data were obtained from the Central
Population Registry, the National Patient Registry,
the Registry of Medicinal Product Statistics, the
National Registry on Regular Dialysis and Transplantation, and the National Registry of Causes of
Death.17-22 The details of the information contained
in each of these registries, as well as the diagnoses,
surgical procedures, and pharmacotherapy used for
defining the study population, coexisting conditions, and outcomes, are provided in the Supplementary Appendix, available with the full text of
this article at NEJM.org.
The Lundbeck Foundation provided grant support but had no role in the conduct of the study.
The study was approved by the Danish Data Protection Agency. Approval by an ethics committee
and written informed consent are not required for
retrospective registry studies in Denmark. The
first author vouches for the integrity of the data
and the accuracy of the data analysis.
STUDY POPULATION
We identified all patients discharged from the hospital with a diagnosis of nonvalvular atrial fibrillation during the study period, from 1997 through
2008.17,19 Pharmacotherapy was determined by
means of filled prescriptions and, because treatment may have been changed or intensified during or immediately after hospitalization, the baseline assessment and follow-up period began 7 days
after discharge. Patients were excluded if they
died, had a thromboembolic event, or had major
bleeding during the 7 days before the baseline
assessment (Fig. 1).17-19
CHRONIC KIDNEY DISEASE AND RENALREPLACEMENT THERAPY
To investigate the risk associated with the severity of nonend-stage chronic kidney disease,
patients were stratified in a time-dependent manner according to the treatment dose of loop diuretics, because high doses are frequently used in patients with severe renal failure or the nephrotic
syndrome. We studied the influence of the underlying renal disease in nonend-stage chronic
kidney disease by comparing the following diagnostic groups: autosomal dominant polycystic kidney disease, chronic glomerulonephritis, diabetic
nephropathy, chronic tubulointerstitial nephropathy, hypertensive nephropathy, and other causes.
PHARMACOLOGIC TREATMENT
627
The
n e w e ng l a n d j o u r na l
The predicted risk of stroke or systemic thromboembolism for all patients was assessed with the use
of the CHA2DS2-VASc score,17,24 which reflects
the risk of stroke among patients with atrial fibrillation who are not receiving anticoagulant
therapy, with values ranging from 0 to 9 and with
higher scores indicating greater risk (see Table 1
in the Supplementary Appendix). The predicted
risk of bleeding was assessed with the use of the
HAS-BLED score,18,25 which reflects the risk of
major bleeding among patients with atrial fibrillation who are receiving anticoagulant therapy,
with values ranging from 0 to 9 and with higher
scores indicating greater risk (see Table 2 in the
Supplementary Appendix). Two risk factors typically included in the HAS-BLED score were not
included in this analysis: abnormal renal function (since chronic kidney disease was the subject
of the study) and labile international normalized
ratios (because these data were not available).
STUDY OUTCOMES
of
m e dic i n e
The risk of stroke or systemic thromboembolism associated with nonend-stage chronic kidney disease or disease requiring renal-replacement
therapy was estimated by means of time-dependent Cox proportional-hazards models, with adjustment for changes in renal status or antithrombotic treatment during follow-up. Cox analyses
were adjusted for risk factors in the CHA2DS2-VASc
score (congestive heart failure, hypertension, age
>75 years, diabetes mellitus, history of stroke or
thromboembolism, vascular disease, age 65 to
74 years, female sex), antithrombotic treatment,
and year of inclusion. The total study population
was included in the initial analysis, and subsequent
analyses were restricted to patients with nonendstage chronic kidney disease and to those requiring renal-replacement therapy.
Similar analyses were conducted to assess the
risk of bleeding, with the use of the HAS-BLED
risk factors (hypertension, abnormal liver function,
history of stroke or thromboembolism, history of
bleeding, age 65 years, use of nonsteroidal antiinflammatory drugs, and unhealthy alcohol use).
Finally, the risks of stroke or systemic thromboembolism, bleeding, myocardial infarction, and
death from any cause were estimated by means of
time-dependent Cox proportional-hazards models,
with adjustment for all baseline characteristics.
A two-sided P value of less than 0.05 was
considered to indicate statistical significance. All
analyses were performed with the use of SAS
software, version 9.2 (SAS Institute), and Stata
software, version 11.0 (StataCorp).
No Renal Disease
(N=127,884)
NonEnd-Stage
Chronic Kidney
Disease
(N=3587)
Disease Requiring
Renal-Replacement
Therapy
(N=901)
P Value
73.212.9
76.511.0
66.811.7
<0.001
22,073 (17.3)
1284 (35.8)
171 (19.0)
<0.001
Hypertension
53,917 (42.2)
1952 (54.4)
486 (53.9)
<0.001
75 yr
66,675 (52.1)
2277 (63.5)
267 (29.6)
<0.001
6574 yr
31,245 (24.4)
809 (22.6)
293 (32.5)
<0.001
Diabetes mellitus
10,920 (8.5)
885 (24.7)
129 (14.3)
<0.001
17,928 (14.0)
644 (18.0)
133 (14.8)
<0.001
Vascular disease
18,174 (14.2)
1034 (28.8)
248 (27.5)
<0.001
Female sex
59,930 (46.9)
1472 (41.0)
303 (33.6)
<0.001
53,917 (42.2)
1952 (54.4)
486 (53.9)
<0.001
Characteristic
Age yr
Risk factors for stroke or thromboembolism no. (%)
Age
2,070 (1.6)
106 (3.0)
36 (4.0)
<0.001
17,928 (14.0)
644 (18.0)
133 (14.8)
<0.001
8,969 (7.0)
584 (16.3)
137 (15.2)
<0.001
Age 65 yr
95,418 (74.6)
3035 (84.6)
533 (59.2)
<0.001
Use of NSAIDs
26,592 (20.8)
843 (23.5)
99 (11.0)
<0.001
Alcohol abuse
4,552 (3.6)
145 (4.0)
43 (4.8)
0.05
Warfarin only
36,638 (28.6)
609 (17.0)
178 (19.8)
<0.001
Aspirin only
23,952 (18.7)
879 (24.5)
153 (17.0)
<0.001
10,745 (8.4)
290 (8.1)
45 (5.0)
<0.001
11,720 (9.2)
70 (2.0)
42 (4.7)
16,926 (13.2)
251 (7.0)
165 (18.3)
99,238 (77.6)
3266 (91.1)
694 (77.0)
CHA2DS2-VASc score
<0.001
HAS-BLED score
<0.001
0 or 1
51,262 (40.1)
883 (24.6)
390 (43.3)
46,159 (36.1)
1336 (37.2)
312 (34.6)
30,463 (23.8)
1368 (38.1)
199 (22.1)
* Plusminus values are means SD. NSAIDs denotes nonsteroidal antiinflammatory drugs.
Scores on the CHA2DS2-VASc,17,24 which reflect the risk of stroke in patients with atrial fibrillation who are not receiving anticoagulant therapy, range from 0 to 9; a score of 0 indicates low risk, a score of 1 intermediate risk, and a score
of 2 or more high risk (Table 1 in the Supplementary Appendix).
Scores on the HAS-BLED,18,25 which reflect the risk of major bleeding in patients with atrial fibrillation who are receiving anticoagulant therapy, range from 0 to 9; a score of 0 or 1 indicates low risk, a score of 2 intermediate risk, and a
score of 3 or more high risk (Table 2 in the Supplementary Appendix).
629
The
n e w e ng l a n d j o u r na l
of
m e dic i n e
tients (0.4%) after a median of 217 days (interquartile range, 33 to 681). Of the 1378 patients requiring renal-replacement therapy during the study
period, 1074 (77.9%) were receiving hemodialysis,
212 (15.4%) were receiving peritoneal dialysis,
and 92 (6.7%) underwent kidney transplantation.
The outcomes did not differ significantly according to the type of renal-replacement therapy.
Rates of stroke or systemic thromboembolism,
bleeding, myocardial infarction, and death from
any cause are shown according to renal status in
Table 2. All event rates were markedly increased
in both groups of patients with renal disease. The
distribution between the three types of stroke or
thromboembolic outcomes and the four types of
bleeding outcomes is shown in Table 4 in the
Supplementary Appendix.
RISK OF BLEEDING
Event
No. of
Person-yr
No. of
Events
461,734
16,648
3.61 (3.553.66)
13,078
842
6.44 (6.026.89)
2,922
164
5.61 (4.826.54)
457,605
16,195
3.54 (3.483.59)
12,515
1,097
8.77 (8.269.30)
2,734
243
480,745
9,037
1.88 (1.841.92)
13,500
784
5.81 (5.416.23)
2,925
175
5.98 (5.166.94)
493,305
55,297
11.21 (11.1211.30)
14,052
5,431
38.65 (37.6339.69)
3,114
914
29.35 (27.5131.32)
Stroke or thromboembolism
No renal disease
Nonend-stage CKD
Disease requiring renalreplacement therapy
Bleeding
No renal disease
Nonend-stage CKD
Disease requiring renalreplacement therapy
8.89 (7.8410.08)
Myocardial infarction
No renal disease
Nonend-stage CKD
Disease requiring renalreplacement therapy
Death
No renal disease
Nonend-stage CKD
Disease requiring renalreplacement therapy
Table 3 shows the results from Cox regression The risk of bleeding was higher among patients
models of the risk of stroke or systemic throm- with nonend-stage chronic kidney disease and
boembolism, with adjustment for CHA2DS2-VASc among those requiring renal-replacement thera630
0.70 (0.650.75)
1.06 (1.031.09)
Hypertension
<0.001
<0.001
1.12 (1.081.15)
1.12 (1.071.16)
3.24 (3.143.35)
1.32 (1.251.39)
2.03 (1.922.16)
3.56 (3.383.76)
1.05 (1.021.09)
1.03 (0.991.08)
0.69 (0.640.74)
1.10 (1.061.14)
0.59 (0.560.61)
1.00
1.00
Hazard Ratio
(95% CI)
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.002
0.11
<0.001
<0.001
<0.001
P Value
No Renal Disease
(N=127,884)
1.06 (0.921.22)
0.89 (0.761.05)
2.71 (2.343.15)
1.16 (0.991.36)
1.52 (1.181.94)
1.87 (1.482.36)
1.13 (0.981.30)
0.98 (0.841.14)
0.76 (0.561.03)
1.25 (1.071.47)
0.84 (0.691.01)
1.00
1.49 (1.381.59)
Hazard Ratio
(95% CI)
0.44
0.17
<0.001
0.07
0.001
<0.001
0.10
0.78
0.08
0.01
0.07
<0.001
P Value
* Results from the time-dependent Cox regression analyses were adjusted for year of inclusion.
Numbers of patients are from baseline data. Because the analyses were time-dependent, these numbers changed during follow-up.
The reference group for the hazard ratio for each thromboembolism risk factor is the group of all patients in the study without that risk factor.
1.10 (1.061.15)
1.12 (1.081.15)
Vascular disease
Female sex
<0.001
<0.001
1.32 (1.261.38)
3.20 (3.103.31)
Diabetes
2.02 (1.912.14)
<0.001
3.48 (3.313.66)
75 yr
<0.001
<0.001
6574 yr
Age
1.03 (0.991.07)
<0.001
1.11 (1.071.15)
<0.001
0.59 (0.570.62)
<0.001
P Value
Aspirin
1.00
Hazard Ratio
(95% CI)
Warfarin
None
Antithrombotic therapy
All participants
Characteristic
Total Population
(N=132,372)
1.34 (0.971.85)
1.11 (0.781.58)
1.99 (1.362.91)
1.41 (0.952.10)
2.18 (1.463.24)
2.46 (1.603.79)
1.05 (0.761.45)
0.96 (0.641.43)
0.82 (0.371.80)
0.88 (0.591.32)
0.44 (0.260.74)
1.00
1.83 (1.572.14)
Hazard Ratio
(95% CI)
0.08
0.57
<0.001
0.09
<0.001
<0.001
0.78
0.84
0.62
0.54
0.002
<0.001
P Value
631
The
n e w e ng l a n d j o u r na l
DISCUSSION
In a large cohort study, we found that among
patients with atrial fibrillation, nonend-stage
chronic kidney disease and disease requiring renalreplacement therapy were both associated with
increased risks of stroke or systemic thromboembolism and bleeding. Among patients with
nonend-stage chronic kidney disease, the risk
of stroke or systemic thromboembolism was not
influenced by the severity of the renal disease (as
632
of
m e dic i n e
2.15 (2.042.26)
<0.001
<0.001
1.23 (1.181.28)
2.44 (2.332.55)
2.09 (2.002.17)
1.12 (1.081.16)
1.40 (1.301.52)
History of bleeding
Age 65 yr
Use of NSAIDs
Alcohol abuse
1.43 (1.321.56)
1.12 (1.081.17)
2.12 (2.032.20)
2.54 (2.422.67)
1.24 (1.191.30)
1.40 (1.251.57)
1.01 (0.981.04)
2.18 (2.072.30)
1.21 (1.161.26)
1.28 (1.231.33)
1.00
1.00
Hazard Ratio
(95% CI)
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
0.58
<0.001
<0.001
<0.001
P Value
No Renal Disease
(N=127,884)
1.01 (0.731.39)
1.10 (0.961.26)
1.61 (1.351.92)
1.70 (1.451.99)
1.04 (0.891.22)
1.31 (0.901.91)
0.99 (0.871.11)
1.63 (1.322.02)
1.12 (0.961.30)
1.36 (1.171.59)
1.00
2.24 (2.102.38)
Hazard Ratio
(95% CI)
0.97
0.19
<0.001
<0.001
0.62
0.16
0.81
<0.001
0.14
<0.001
<0.001
P Value
NonEnd-Stage Chronic
Kidney Disease (N=3587)
* Results from the time-dependent Cox regression analyses were adjusted for year of inclusion. NSAIDs denotes nonsteroidal antiinflammatory drugs.
Numbers of patients are from baseline data. Because the analyses were time-dependent, these numbers changed during follow-up.
The reference group for the hazard ratio for each bleeding risk factor is the group of all patients in the study without that risk factor.
<0.001
<0.001
<0.001
<0.001
1.01 (0.981.04)
1.37 (1.231.52)
Hypertension
<0.001
1.21 (1.161.26)
<0.001
1.28 (1.231.33)
<0.001
P Value
Warfarin
1.00
Hazard Ratio
(95% CI)
Total Population
(N=132,372)
Aspirin
None
Antithrombotic therapy
All participants
Characteristic
1.33 (0.702.54)
0.91 (0.621.33)
1.36 (1.031.80)
2.09 (1.502.91)
0.93 (0.631.36)
0.74 (0.341.64)
0.92 (0.711.20)
1.71 (0.982.99)
1.63 (1.182.26)
1.27 (0.911.77)
1.00
2.70 (2.383.07)
Hazard Ratio
(95% CI)
0.39
0.63
0.03
<0.001
0.70
0.46
0.55
0.06
0.003
0.15
<0.001
P Value
633
The
n e w e ng l a n d j o u r na l
though the positive predictive value of the diagnosis of atrial fibrillation is very high (99%),29 the
inclusion of only hospitalized patients with atrial
fibrillation is likely to have resulted in an overestimate of the proportion of patients who were
at increased risk for thromboembolism and bleeding. The bleeding outcome was restricted to hospitalization or death related to gastrointestinal
bleeding, intracranial bleeding, bleeding from the
urinary tract, and airway bleeding, and the results cannot be applied to the risk of other types
of bleeding.20,21 The outcomes of stroke or systemic thromboembolism and myocardial infarction are well validated, and the information on
all-cause mortality is accurate,26,27 as is the information on renal-replacement therapy.22 Since we
did not have brain-imaging data for the patients
in the study, some strokes that were reported as
ischemic may actually have been hemorrhagic,
but Krarup and colleagues did not find this potential bias to be of importance.26 Despite the
accuracy of filled prescriptions as a measure of
medication use,30 aspirin can also be bought over
the counter in Denmark, and the use of aspirin
may therefore be underestimated. It is also possible that the increased risk of stroke or systemic
thromboembolism that was associated with aspirin in our analysis was due to confounding by
indication. Finally, some of the dialysis centers in
Denmark have reported that they provide warfarin without the use of prescriptions for patients
who require renal-replacement therapy (primarily
those receiving hemodialysis).
of
m e dic i n e
REFERENCES
1. USRDS 2006 annual data report: atlas
634
635