Risk Factors for Stroke in Patients with Nonrheumatic Atria1 Fibrillation: A Case-Control Study ANNEW. MOULTON, M.D., DANIELE. SINGER,M.D., JENNIFERS. HAAS,M.D., Providence,
Rhode Island, and Boston, Massachusetts
Randomized controlled trials have demonstrated that anticoagulant therapy is very effective at preventing stroke among patients with nonrheumatic atrial fibrillation. However, these trials have reported too few strokes for powerful risk factor analy&. Observational studies may provide additional information. The purpose of this study was to identify risk factors in a larger number of patients with stroke and nonrheumatic atrial fibrillation, using case-control methodology. PATIENTS AND METHODS We identified all patients discharged from one hospital over an 8year period who met our case definition of nonrheumatic atrial fibrillation and ischemic stroke (n = 134), and compared them with contemporaneous control subjects who were discharged with nonrheumatic atrial fibrillation without stroke (n = 131). RESULT& Cases and controls were similar in terms of duration of atrial fibrillation; proportion with paroxysmal atrial fibrillation; percentage with a past medical history of angiaa, myocardial infarction, congestive heart failure, diabetes, or smoking; and mean left atrial sire. In contrast, cases were significantly older than controls (78.5versus 74.8years, p = 0.002) and more likely to have a history of hypertension (55% versus 3&I%,p = 0.0093). The relative odds for stroke was 1.91 for patients with hypertension, 1.73 for patients older than 75 years, and 326 for patients with both factors. PURPOSE:
CONCLUSIONS:
k
analysis
SUggeStS
that
age
and hypertension should be considered when deciding upon long-term anticoagulant therapy to prevent stroke in patients with nonrheumatic atrial fibrillation. From the Division of General Internal Medicine, Rhode Island Hospital, Providence, Rhode Island; the General Medicine Unit, Massachusetts General Hospital, Boston, Massachusetts; and the Division of General Medicine, Brigham and Women’s Hospital, Boston, Massachusetts. Partial support for this study was provided by the Harvard General Medicine Fellowship (Dr. Moulton), the Charles Dana Fellowship for Harvard Medical School (Dr. Haas). and the Henry J. Kaiser Family Foundation via a Faculty Scholar Award (Dr. Singer). Requests for reprints should be addressed to Anne W. Moulton, M.D.. Division of General Internal Medicine, Rhode Island Hospital, 593 Eddy Street, Providence, Rhode Island 02903. Manuscript submitted March 11, 1991. and accepted April 26, 1991.
156
August 1991
The AmericanJournal of Medicine
Volume 91
onrheumatic atrial fibrillation increasesthe N risk of stroke fivefold to approximately 5%per year Becauseof the high prevalenceof atrial [l-4].
fibrillation with increasingage,approximately 15% of all strokesin patients older than 60canbe attributed to atrial fibrillation [5]. Recent randomized controlled trials haveshownconclusivelythat warfarin is very effective in preventing strokes in patients with nonrheumatic atrial fibrillation [6-8]. However,becausewarfarin is a risky and inconvenient therapy, it would be ideal to restrict its useto subgroupsat greatestrisk for stroke. The randomized trials werestopped after only a small number of strokesoccurred.As a result,thesetrials havenot beenable to provide powerful assessmentsof risk factors. Case-controlstudies can assemblea larger number of strokesand may provide more efficient assessmentof risk factors.We report the results of sucha case-controlstudy to identify risk factorsfor ischemicstroke amongpatients with nonrheumatic atrial fibrillation. PATIENTSAND METHODS Using a dischargelog of consecutiveadmissions to the MassachusettsGeneralHospital during the time period January 1979 to December 1987,we identified 417individual recordswith an ICD-9 dischargediagnosisof atrial fibrillation (427.31)and ischemicstroke (436.0,434.0,434.1,434.9). Records for thesepatients werereviewedto ascertainthat a stroke had truly occurred.Stroke was defined as a sudden developmentof a major neurologicdeficit that lasted more than 24 hours and correlatedwith a cerebrovascularterritory. In order to addressthe anticoagulation decision for patients with chronicnonrheumaticatrial fibrillation who had not had a previous stroke, we excluded (1) patients with a history of a previous stroke or a previous transient ischemic attack requiring hospitalization; (2) patients with evidence of mitral stenosison echocardiogramor a history of rheumatic heart diseaseor a physical examination result consistent with mitral stenosis;(3) patients with a history of prior embolic phenomena;(4) patients with evidenceof another cardiac sourcefor emboli (e.g.,left ventricular thrombus on echocardiogram); (5) patients whoseatrial fibrillation was
STROKE
secondary to a transient etiology, e.g., an acute myocardial infarction; and (6) patients with a documented history of severe carotid disease or previous carotid surgery (in order to focus more specifically on cardioembolic stroke). Using these criteria, we identified 134 cases of ischemic stroke in patients with nonrheumatic atrial fibrillation. Cases were more precisely classified as embolic or nonembolic using criteria developed by us prior to the chart review. A definite embolic stroke required autopsy or angiographic evidence of an embolus (n = 5). Probable embolic stroke was defined if the patient had a clinical picture and computed tomographic scan that were consistent with an embolus but no angiographic or autopsy evidence (n = 78). Control patients were randomly selected from a discharge log of all patients with an ICD-9 discharge diagnosis of atrial fibrillation (427.31) and without a diagnosis of stroke, admitted during the same time period January 1979 to December 1987. If controls were admitted more than once, only the first admission was included. Patients were excluded for the same medical reasons noted for the cases. One hundred thirty-one control subjects were identified. Selected Relevant Definitions ATRIAL, FIBRILLATION: Sustained atrial fibrillation was defined as at least two electrocardiograms (EKGs) demonstrating atrial fibrillation tbroughout the time of admission with no interim EKG showing normal sinus rhythm or, in the absence of that information, a history of sustained atrial fibrillation from the admitting physician and at least one EKG demonstrating atrial fibrillation during the admission. Paroxysmal atrial fibrillation was defined as at least one episode of atrial fibrillation documented by EKG separated in time by two EKGs showing normal sinus rhythm. The onset of atrial fibrillation was dated according to physicians’ notes or the date of the first EKG on which atrial fibrillation was present, whichever was earlier. Sixty cases and 51 controls were missing data about duration of atrial fibrillation. HYPERTENSION: Hypertension was defined as probable if the patient had such a diagnosis listed in the medical record and as definite if the patient had at least two documented blood pressure readings with a systolic blood pressure of 160 mm Hg or greater or a diastolic reading of 95 mm Hg or greater or if the patient was currently receiving antihypertensive medication. Hypertension was classified as definite in the majority of both cases (95%) and controls (86%). There was no significant change in an analysis considering only definite hypertension,
AND NONRHEUMATIC
ATRIA1
FIBRILLATION
/ MOULTON
ET AL
and all subsequent analyses combine probable and definite hypertension. Left Atrial Size Echocardiography data were available for 81 cases and 64 controls. Of these, measurements of left atrial anteroposterior diameter were available for 74% of cases (n = 60) and 64% of controls (n = 41) and details about valvular pathology, other than mitral stenosis, were available for 86% (n = 70) of cases and 95% (n = 61) of controls. M-mode techniques were used prior to 1981 and two-dimensional techniques afterwards. Measurements were made using standard inner-edge methods [9]. Statistical Analysis Univariate comparisons between cases and controls were assessed by chi-square with Yates’ correction and Student’s t-test [lo]. Confidence intervals (CIs) for univariate relative risks used Miettinen’s [ll] test-based method. The effect of multiple covariates was analyzed by multiple logistic regression [ 121.
RESULTS Cases Of the 134 cases of stroke, the majority of patients presented with a major neurologic deficit at the onset of their symptoms (84%). Most patients presented with a motor and/or sensory deficit (81%). The remainder of patients presented with aphasia and/or a visual disturbance and/or a gait disturbance (19%). Sixty-two (46%) patients had carotid noninvasive studies and all but one revealed the absence of or minimal stenosis. Only four patients underwent angiography and three of these demonstrated an embolus in the distribution of the deficit (data not shown). By our criteria, there were 83 cases of probable or definite embolic stroke. Analyses restricted to such cases of embolic stroke were not different from analyses using all 134 cases of ischemic stroke. All reported comparisons are based on the entire set of stroke cases. Controls Primary discharge diagnoses for the 131 controls with atrial fibrillation and without stroke included atrial fibrillation itself (25%), cardiovascular disease (17%), cancer (15%), surgical procedures (8%), chronic lung disease (4%), pneumonia (3%), and miscellaneous (28%) (e.g., seizure, syncope, gout). Comparison of Clinical Characteristics The clinical characteristics of cases and controls are compared in Table I. There were no significant
August
1991
The American
Journal
of Medicine
Volume
91
157
STROKE
AND NONRHEUMATIC
ATRIA1
FlBRlLLAllON
/ MOULTON
ET AL
TABLE I Clinical Characteristics:Caseswith StrokeCompared with ControlsWithout Stroke
cases Age
Mean (median) Age > 75 years Gender Female Past medical history Angina Myocardial infarction Congestive heart failure Hypertension Diabetes Smoking (ever smoked) Warfarin treatment (current) Aspirin treatment (current)
Controls
OddsRatio
(n = 134)
fn = 131)
(95% Cl)
78.5 (79.0)* 65% (87/134)+
51% (67/131)
1.76 (1.08,2.89)
54% (731134)
43% (56/131)
1.60 fO.99,2.66)
24% (32/132) 14% (19/132) 29% (39/134) 55% t73/133j*
15% (19/131) 13% (17/129) 20% (26/131) 38% (50/131) 15% (20/131) 46% (57/125) 7% (9/130) 6% (8025)
1.89(1.01,3.54) 1.11 (0.54,2.28)
74.8 (76.0)
18% (24/131)
38% (44/l 15) 2% (3/134) 13% (16/126)
1.66 (0.94,2.93)
1.97 (1.18,3.28) 1.24 (0.65,2.35)
0.74 (0.44, 1.23) 0.31 (0.09, 1.09) 2.12 (0.89, 5.03)
Cl = confidence interval. ‘p t0.01.
tp t0.05.
differences between the two groups in gender or in past medical history of angina, myocardial infarction, congestive heart failure, diabetes, or smoking. In contrast, cases were older (78.5 versus 74.8 years, p = 0.002) and significantly more likely to have a history of hypertension (55% versus 38%, p = 0.009). The relative odds for stroke among those older than 76 years was 1.76 (95% CI: 1.08 to 2.89); the relative odds for those with hypertension was 1.97 (95% CI: 1.18 to 3.28). The effect of hypertension was very similar when examined only among the 83 cases of probable or definite embolic stroke (odds ratio = 1.79). Too few subjects were taking warfarin or aspirin for a meaningful assessment of their therapeutic effect. There were no features of atrial fibrillation itself that were clearly associated with stroke (Table II). There were no significant differences between cases and controls in proportions of patients with sustained versus paroxysmal atrial fibrillation, or in the duration of atrial fibrillation. There was no sig-
nificant difference in the percentage of patients in each group who presented with new-onset atrial fibrillation. However, if we excluded the 17 patients in the control group who were specifically admitted for a diagnosis of new-onset atrial fibrillation, there was a difference in the percent of patients in whom new-onset atrial fibrillation was noted secondarily on admission (cases 29% versus controls 16%, p = 0.025; odds ratio = 2.14, 95% CI = 1.10 to 4.16). Comparison of Echocardiographic Features Mean left atrial size did not differ between cases and controls: 43.7 versus 43.3 mm (Table III). Although extreme left atrial size (greater than 50 mm) occurred more commonly in cases than in controls (23% versus lo%), this difference did not achieve statistical significance. There was no difference in coexistent valvular disease among the two groups (63% versus 61%). Some level of mitral regurgitation (with or without other valvular disease) was seen on echocardiograms of 48% of cases and in 27% of con-
TABLE II Characteristicsof Atrial Fibrillation: Caseswith Stroke Comparedwith ControlsWithout Stroke*
Sustained atrial fibrillation Paroxysmal atrial fibrillation Duration of atrial fibrillation (yea&t Mean Median Range New-onset atrial fibrillation
Cases (n = 134)
(n = 131)
OddsRatio (95% Cl)
60% (79/131) 40% (52/131)
56% (74/131) 44% (57/131)
1.17 (0.72, 1.91) 0.86 (0.53, 1.39)
?: O.OZi-25
it O.Oi-23 27% (35/129)
29% (37/128)
Cl = confidence interval. *All diirences statistically nonsignikcant (p > 0.05). tDuration specified in 74 cases and 80 controls. tMierance in the mean.
158
August 1991 The American Journal of Medicine
Volume 91
COfttfOlS
-0.80 (-2.5,0.92)t 1.09 (0.64, 1.86)
STROKE
AND NONRHEUMATIC
ATRIAL
FIBRILLATION
/ MOULTON
ET AL
TABLE III EchocardiographyAssessment:Caseswith Stroke Compared with ControlsWithout Stroke*
Echocardiograms performed Left atrial size Mean (median), mm > 40 mm (%) > 50 mm (%I Valvular disease Any valvular disease Mitral regurgitation+ Mitral valve prolapse Mitral annular calcification Other vakular disease
Cases (n = 134)
Controls (n = 131)
60% (81/134)
49% (64/131)
43.7 (43.5) 60% (36/60) 23% (14/60)
43.3 (45.0) 61% (25/41)
63% (44/70) 48% (21/44)
61% (37/61) 27% (10/37)
11% (5/44)
19% (7/371
27% (12/44) 14% (6/44)
32% (12/37) 22% (8137)
Odds Ratio (95% Cl)
0.96 (0.42,2.17) 2.82 (0.88,8.93)
10% (4/41)
1.10 (0.54,2.26) 2.47 0.55 0.78 0.57
(0.97,6.27) (0.16, 1.89) (0.30,2.03) (0.18, 1.82)
Cl = confidenceinterval. ‘All comparisons were nonsignificant.
tlncludes patients with other coexistent valvular disease.
trols, but this difference did not achieve statistical significance. There was no difference in mitral annular calcification. Multivariate Analyses The increased risk of stroke conferred by age and the increased risk of stroke conferred by hypertension were largely independent. Table IV presents these risks via a stratified analysis in which patients were grouped according to whether they had no risk factors, one risk factor, or both. Note that patients who were both older than 75 and had a history of hypertension had an odds ratio for stroke of 3.26. Analyzed by a multiple logistic regression equation including hypertension and age (dichotomized at less than or equal to 75, greater than 75 years), both factors had coefficients that were statistically significant (p 75, no HTN Age 5 75, yes HTN Age > 75, yesHTN
1991 The American
Journal
z3
1.91 3.26
of Medicine
Volume
91
159
STROKE AND NONRHEUMATIC ATRIA1 FIBRILLATION / MOULTON ET AL
Additional possible risk factors for stroke in nonrheumatic atrial fibrillation have been previously reported. The Framingham Study analysis also identified congestive heart failure and coronary artery disease, features not conferring excess risk in our study. Other reports have suggested an increased risk of stroke early in the course of atrial fibrillation [4,22]. Our data provide some support for this finding. Control subjects admitted specifically for new-onset atrial fibrillation provide a biased comparison and removing such controls reveals an increased risk of stroke at first diagnosis of atrial fibrillation. However, conclusions about the role of duration of atrial fibrillation on stroke are limited by inaccuracy in assessing time of onset of atrial fibrillation. Embolism is a well-known complication of electrical cardioversion of atria1 fibrillation, which might suggest a greater risk with paroxysmal atrial fibrillation [23]. However, several studies have found that paroxysmal atrial fibrillation actually confers a smaller risk of embolism than sustained atrial fibrillation [24-261. We found no difference in risk between paroxysmal and sustained atria1 fibrillation. Studies of left atrial size and stroke in patients with atrial fibrillation have produced conflicting results [8,27-321. In our study, left atrial size was not associated with stroke, although there was a suggestion that the presence of an extremely large left atrium (greater than 50 mm) increased risk. Recent echocardiographic analyses have suggested that mitral annular calcification in some way predisposes to stroke [8,33]. Our study does not support this role for mitral annular calcification. Potential methodologic limitations of this study merit discussion. In a retrospective study, missing data are not retrievable. Data were essentially complete for our major risk factors-age and hypertension. However, many subjects did not have echocardiography measurements. In a case-control study, there is also concern about bias in the sampling of the cases and controls. Our cases of hospitalized stroke represented the more severe end of the clinical spectrum and our conclusions may not bear on the risk of milder or even “silent” stroke [34]. The controls in a case-control study should be a fair sample of the population that gives rise to the cases. In this study, the theoretically ideal control group would be a random sample of patients with nonrheumatic atrial fibrillation who would come to our study hospital if they had a stroke. In practice, such a group is difficult to define and survey. Other hospitalized patients, the group we chose, are an attractive alternative source of controls, so long as their reasons for hospitalization were not related to the exposures being explored. With regard to our 160
August 1991 The American Journal of Medicine
major findings, that age and hypertension were associated with stroke, we would anticipate that if hospitalized controls introduce any bias, it would be towards older age and more hypertension-biases that could not explain our findings.
ACKNOWLEDGMENT We are indebted to Dr. Steven J. Skates for stahstical assistance, Helen O’Connor for preparation of the manuscript, Tina Desai for data entry, and Hilda Sequeira for help with the hospital records.
REFERENCES 1. Alpert JS. Petersen P, Gotdfredsen J. Atrial fibrillation: natural history, complications and management. Am Rev Med 1988; 39: 41-52. 2. Dunn M. Alexander J. deSilva R. Hildner F. Antithrombolic therapy in atrial fibrillation. Chest 1989; 95 (suppl): 118S-27s. 3. Wolf PA, Dawber TR. Thomas HE, Kannel WE. Epidemiologic assessment of chronic atrial fibrillation and risk of stroke: the Framingham Study. Neurology 1978; 28: 973-7. 4. Wolf PA, Kannel WB. McGee DL. Duration of atrial fibrillation and imminence of stroke: the Framingham Study. Stroke 1983; 14: 664-7. 5. Wolf PA, Abbott RD. Kannel WB. Atrial fibrillation: a major contributor to stroke in the elderly: the Framingham Study. Arch Intern Med 1987; 147: 1561-4. 6. Petersen P, Godtfredsen J. Boysen G. Andersen E, Andersen B. Placebocontrolled, randomized trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK Study. Lancet 1989; 1: 175-9. 7. Stroke Prevention in Atrial Fibrillation Study Group Investigators. Preliminary report of the Stroke Prevention in Atrial Fibrillation Study. N Engl J Med 1990; 322: 863-8. 8. The Boston Area Anticoagulation Trial in Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with non-rheumatic atrial fibrillation. N Engl J Med 1990; 323: 1505-l 1. 9. Weyman AE. Cross-sectional echocardiography. Philadelphia: Lea & Febiger. 1982. 10. Crunch Statistical Package. Oakland, California: Crunch Software Corporation, 1987. 11. Miettinen OS. Estimability and estimation in case-referent studies. Am J Epidemiol 1976; 103: 226-35. 12. Royal Statistical Society, London 1985; GLIM 3.77 update 1. 13. Flegel KM, Shipley MJ. Rose G. Risk of stroke in nonrheumatic atrial fibrillation Lancet 1987; 1: 526-9. 14. Flegel KM, Hanley J. Risk factors for stroke and other embolic events in patients with nonrheumatic atrial fibrillation. Stroke 1989; 20: lOCO-4. 15. Aronow WS. Gutstein H, Hsieh FY. Risk factors for thromboembolic stroke in elderly patients with chronic atrial fibrillation. Am J Cardiol 1989; 63: 366-7. 16. Petersen P. Kastrup J. Helweg-Larsen S, Boysen G. Godtfredsen J. Risk factors for thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK Study. Arch Intern Med 1990; 150: 819-21. 17. Landefeld CS, Goldman L. Major bleeding in outpatients treated with warfarin: incidence and prediction by factors known at the start of outpatient therapy. Am J Med 1989; 87: 144-52. 18. Gurwitz JH. Goldberg RJ, Holden A, Knapic N. Ansell J. Age-related risks of long-term oral anticoagulant therapy. Arch Intern Med 1988; 148: 1733-6. 19. Sixty-Plus Reinfarction Study Research Group. Risks of long-term oral anticoagulant therapy in elderly patients after myocardial infarction. Lancet 1982; 1: 64-8. 20. Levine MN, Raskob G, Hirsh J. Hemorrhagic complications of long-term anticoagulant therapy. Chest 1989; 95 (suppl): 26S-36s. 21. Wintzen AR, deJonge H. Loeliger EA, Bots GTAM. The risk of intracerebral hemorrhage during oral anticoagulant treatment: a population study. Ann Neurol 1984; 16: 553-8. 22. Hart RG. Easton JD, Sherman DG. Duration of non-valvular atrial fibrillation and stroke [letter]. Stroke 1983; 14: 827. 23. Bjerkeland CJ, Orning OM. The efficacy of anticoagulant therapy in preventing embolism related to D.C. electrical conversion of atrial fibrillation. Am J Cardiol 1969; 23: 208-16. 24. Peterson P, Godtfredsen J. Embolic complications lation. Stroke 1986; 17: 622-6.
Volume 91
in paroxysmal
atrialfibril-
STROKE 25. Treseder AS, Sastry BSD, Thomas TPL, Yates MA, Pathy MSJ. Atrial fibrillation and stroke in elderly hospitalized patients. Age Ageing 1986; 15: 89-92. 26. Britton FM, Gustafson C. Non-rheumatic atrial fibrillation as a risk factor for stroke. Stroke 1985; 16: 182-8. 27. Caplan LR, D’Cruz I, Hier DB. Reddy H, Shah S. Atrial srze. atrial fibrillation and stroke. Ann Neural 1986; 19: 158-61. 26. Weiner I. Clinical and echocardiographic correlates of systemic embolization in nonrheumatic atrial fibrillation, Am J Cardiol 1987; 59: 177. 29. Ruocco NA, Most AS. Clinical and echocardiographic risk factors for systemic embolization in patients with atrial fibrillation in the absence of mitral stenosis [abstract]. J Am Coll Cardiol 1986; 7: 165A. 30. Cabin HS, Clubb S, Hall C, Perlmutter RA. Feinstein AR. Risk of systemic
AND NONRHEUMATIC
ATRIAL
FIBRILLATION
/ MOULTON
ET AL
embolization of atrial fibrillation without mitral stenosis. Am J Cardiol 1990; 65: 1112-6. 31. Tegeler CH. Hart RG, Sherman DG. Quiroga ES, Easton JD. Atrial fibrillation -stroke study [abstract]. Stroke 1986; 17: 131A. 32.Petersen P. Kastrup J, Helweg-Larson S, Boysan G, Gotfredsen J. Risk factors for thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK Study. Arch Intern Med 1990; 150: 819-21. 33. Aronow WS, Koenigsberg M. Kronzon I, Gutstein H. Association of mitral annular calcium with new thromboembolic stroke and cardiac events at 39month follow-up in elderly patients. Am J Cardiol 1990; 65: 1511-2. 34. Peterson P. Madsen EB, Brun B. Peterson F. Boysen G. Silent cerebral infarction in atrial fibrillation. Stroke 1987; 18: 1098-l 100.
August
1991
The American
Journal
of Medicine
Volume
91
161
Rhode Island, and Boston, Massachusetts
Randomized controlled trials have demonstrated that anticoagulant therapy is very effective at preventing stroke among patients with nonrheumatic atrial fibrillation. However, these trials have reported too few strokes for powerful risk factor analy&. Observational studies may provide additional information. The purpose of this study was to identify risk factors in a larger number of patients with stroke and nonrheumatic atrial fibrillation, using case-control methodology. PATIENTS AND METHODS We identified all patients discharged from one hospital over an 8year period who met our case definition of nonrheumatic atrial fibrillation and ischemic stroke (n = 134), and compared them with contemporaneous control subjects who were discharged with nonrheumatic atrial fibrillation without stroke (n = 131). RESULT& Cases and controls were similar in terms of duration of atrial fibrillation; proportion with paroxysmal atrial fibrillation; percentage with a past medical history of angiaa, myocardial infarction, congestive heart failure, diabetes, or smoking; and mean left atrial sire. In contrast, cases were significantly older than controls (78.5versus 74.8years, p = 0.002) and more likely to have a history of hypertension (55% versus 3&I%,p = 0.0093). The relative odds for stroke was 1.91 for patients with hypertension, 1.73 for patients older than 75 years, and 326 for patients with both factors. PURPOSE:
CONCLUSIONS:
k
analysis
SUggeStS
that
age
and hypertension should be considered when deciding upon long-term anticoagulant therapy to prevent stroke in patients with nonrheumatic atrial fibrillation. From the Division of General Internal Medicine, Rhode Island Hospital, Providence, Rhode Island; the General Medicine Unit, Massachusetts General Hospital, Boston, Massachusetts; and the Division of General Medicine, Brigham and Women’s Hospital, Boston, Massachusetts. Partial support for this study was provided by the Harvard General Medicine Fellowship (Dr. Moulton), the Charles Dana Fellowship for Harvard Medical School (Dr. Haas). and the Henry J. Kaiser Family Foundation via a Faculty Scholar Award (Dr. Singer). Requests for reprints should be addressed to Anne W. Moulton, M.D.. Division of General Internal Medicine, Rhode Island Hospital, 593 Eddy Street, Providence, Rhode Island 02903. Manuscript submitted March 11, 1991. and accepted April 26, 1991.
156
August 1991
The AmericanJournal of Medicine
Volume 91
onrheumatic atrial fibrillation increasesthe N risk of stroke fivefold to approximately 5%per year Becauseof the high prevalenceof atrial [l-4].
fibrillation with increasingage,approximately 15% of all strokesin patients older than 60canbe attributed to atrial fibrillation [5]. Recent randomized controlled trials haveshownconclusivelythat warfarin is very effective in preventing strokes in patients with nonrheumatic atrial fibrillation [6-8]. However,becausewarfarin is a risky and inconvenient therapy, it would be ideal to restrict its useto subgroupsat greatestrisk for stroke. The randomized trials werestopped after only a small number of strokesoccurred.As a result,thesetrials havenot beenable to provide powerful assessmentsof risk factors. Case-controlstudies can assemblea larger number of strokesand may provide more efficient assessmentof risk factors.We report the results of sucha case-controlstudy to identify risk factorsfor ischemicstroke amongpatients with nonrheumatic atrial fibrillation. PATIENTSAND METHODS Using a dischargelog of consecutiveadmissions to the MassachusettsGeneralHospital during the time period January 1979 to December 1987,we identified 417individual recordswith an ICD-9 dischargediagnosisof atrial fibrillation (427.31)and ischemicstroke (436.0,434.0,434.1,434.9). Records for thesepatients werereviewedto ascertainthat a stroke had truly occurred.Stroke was defined as a sudden developmentof a major neurologicdeficit that lasted more than 24 hours and correlatedwith a cerebrovascularterritory. In order to addressthe anticoagulation decision for patients with chronicnonrheumaticatrial fibrillation who had not had a previous stroke, we excluded (1) patients with a history of a previous stroke or a previous transient ischemic attack requiring hospitalization; (2) patients with evidence of mitral stenosison echocardiogramor a history of rheumatic heart diseaseor a physical examination result consistent with mitral stenosis;(3) patients with a history of prior embolic phenomena;(4) patients with evidenceof another cardiac sourcefor emboli (e.g.,left ventricular thrombus on echocardiogram); (5) patients whoseatrial fibrillation was
STROKE
secondary to a transient etiology, e.g., an acute myocardial infarction; and (6) patients with a documented history of severe carotid disease or previous carotid surgery (in order to focus more specifically on cardioembolic stroke). Using these criteria, we identified 134 cases of ischemic stroke in patients with nonrheumatic atrial fibrillation. Cases were more precisely classified as embolic or nonembolic using criteria developed by us prior to the chart review. A definite embolic stroke required autopsy or angiographic evidence of an embolus (n = 5). Probable embolic stroke was defined if the patient had a clinical picture and computed tomographic scan that were consistent with an embolus but no angiographic or autopsy evidence (n = 78). Control patients were randomly selected from a discharge log of all patients with an ICD-9 discharge diagnosis of atrial fibrillation (427.31) and without a diagnosis of stroke, admitted during the same time period January 1979 to December 1987. If controls were admitted more than once, only the first admission was included. Patients were excluded for the same medical reasons noted for the cases. One hundred thirty-one control subjects were identified. Selected Relevant Definitions ATRIAL, FIBRILLATION: Sustained atrial fibrillation was defined as at least two electrocardiograms (EKGs) demonstrating atrial fibrillation tbroughout the time of admission with no interim EKG showing normal sinus rhythm or, in the absence of that information, a history of sustained atrial fibrillation from the admitting physician and at least one EKG demonstrating atrial fibrillation during the admission. Paroxysmal atrial fibrillation was defined as at least one episode of atrial fibrillation documented by EKG separated in time by two EKGs showing normal sinus rhythm. The onset of atrial fibrillation was dated according to physicians’ notes or the date of the first EKG on which atrial fibrillation was present, whichever was earlier. Sixty cases and 51 controls were missing data about duration of atrial fibrillation. HYPERTENSION: Hypertension was defined as probable if the patient had such a diagnosis listed in the medical record and as definite if the patient had at least two documented blood pressure readings with a systolic blood pressure of 160 mm Hg or greater or a diastolic reading of 95 mm Hg or greater or if the patient was currently receiving antihypertensive medication. Hypertension was classified as definite in the majority of both cases (95%) and controls (86%). There was no significant change in an analysis considering only definite hypertension,
AND NONRHEUMATIC
ATRIA1
FIBRILLATION
/ MOULTON
ET AL
and all subsequent analyses combine probable and definite hypertension. Left Atrial Size Echocardiography data were available for 81 cases and 64 controls. Of these, measurements of left atrial anteroposterior diameter were available for 74% of cases (n = 60) and 64% of controls (n = 41) and details about valvular pathology, other than mitral stenosis, were available for 86% (n = 70) of cases and 95% (n = 61) of controls. M-mode techniques were used prior to 1981 and two-dimensional techniques afterwards. Measurements were made using standard inner-edge methods [9]. Statistical Analysis Univariate comparisons between cases and controls were assessed by chi-square with Yates’ correction and Student’s t-test [lo]. Confidence intervals (CIs) for univariate relative risks used Miettinen’s [ll] test-based method. The effect of multiple covariates was analyzed by multiple logistic regression [ 121.
RESULTS Cases Of the 134 cases of stroke, the majority of patients presented with a major neurologic deficit at the onset of their symptoms (84%). Most patients presented with a motor and/or sensory deficit (81%). The remainder of patients presented with aphasia and/or a visual disturbance and/or a gait disturbance (19%). Sixty-two (46%) patients had carotid noninvasive studies and all but one revealed the absence of or minimal stenosis. Only four patients underwent angiography and three of these demonstrated an embolus in the distribution of the deficit (data not shown). By our criteria, there were 83 cases of probable or definite embolic stroke. Analyses restricted to such cases of embolic stroke were not different from analyses using all 134 cases of ischemic stroke. All reported comparisons are based on the entire set of stroke cases. Controls Primary discharge diagnoses for the 131 controls with atrial fibrillation and without stroke included atrial fibrillation itself (25%), cardiovascular disease (17%), cancer (15%), surgical procedures (8%), chronic lung disease (4%), pneumonia (3%), and miscellaneous (28%) (e.g., seizure, syncope, gout). Comparison of Clinical Characteristics The clinical characteristics of cases and controls are compared in Table I. There were no significant
August
1991
The American
Journal
of Medicine
Volume
91
157
STROKE
AND NONRHEUMATIC
ATRIA1
FlBRlLLAllON
/ MOULTON
ET AL
TABLE I Clinical Characteristics:Caseswith StrokeCompared with ControlsWithout Stroke
cases Age
Mean (median) Age > 75 years Gender Female Past medical history Angina Myocardial infarction Congestive heart failure Hypertension Diabetes Smoking (ever smoked) Warfarin treatment (current) Aspirin treatment (current)
Controls
OddsRatio
(n = 134)
fn = 131)
(95% Cl)
78.5 (79.0)* 65% (87/134)+
51% (67/131)
1.76 (1.08,2.89)
54% (731134)
43% (56/131)
1.60 fO.99,2.66)
24% (32/132) 14% (19/132) 29% (39/134) 55% t73/133j*
15% (19/131) 13% (17/129) 20% (26/131) 38% (50/131) 15% (20/131) 46% (57/125) 7% (9/130) 6% (8025)
1.89(1.01,3.54) 1.11 (0.54,2.28)
74.8 (76.0)
18% (24/131)
38% (44/l 15) 2% (3/134) 13% (16/126)
1.66 (0.94,2.93)
1.97 (1.18,3.28) 1.24 (0.65,2.35)
0.74 (0.44, 1.23) 0.31 (0.09, 1.09) 2.12 (0.89, 5.03)
Cl = confidence interval. ‘p t0.01.
tp t0.05.
differences between the two groups in gender or in past medical history of angina, myocardial infarction, congestive heart failure, diabetes, or smoking. In contrast, cases were older (78.5 versus 74.8 years, p = 0.002) and significantly more likely to have a history of hypertension (55% versus 38%, p = 0.009). The relative odds for stroke among those older than 76 years was 1.76 (95% CI: 1.08 to 2.89); the relative odds for those with hypertension was 1.97 (95% CI: 1.18 to 3.28). The effect of hypertension was very similar when examined only among the 83 cases of probable or definite embolic stroke (odds ratio = 1.79). Too few subjects were taking warfarin or aspirin for a meaningful assessment of their therapeutic effect. There were no features of atrial fibrillation itself that were clearly associated with stroke (Table II). There were no significant differences between cases and controls in proportions of patients with sustained versus paroxysmal atrial fibrillation, or in the duration of atrial fibrillation. There was no sig-
nificant difference in the percentage of patients in each group who presented with new-onset atrial fibrillation. However, if we excluded the 17 patients in the control group who were specifically admitted for a diagnosis of new-onset atrial fibrillation, there was a difference in the percent of patients in whom new-onset atrial fibrillation was noted secondarily on admission (cases 29% versus controls 16%, p = 0.025; odds ratio = 2.14, 95% CI = 1.10 to 4.16). Comparison of Echocardiographic Features Mean left atrial size did not differ between cases and controls: 43.7 versus 43.3 mm (Table III). Although extreme left atrial size (greater than 50 mm) occurred more commonly in cases than in controls (23% versus lo%), this difference did not achieve statistical significance. There was no difference in coexistent valvular disease among the two groups (63% versus 61%). Some level of mitral regurgitation (with or without other valvular disease) was seen on echocardiograms of 48% of cases and in 27% of con-
TABLE II Characteristicsof Atrial Fibrillation: Caseswith Stroke Comparedwith ControlsWithout Stroke*
Sustained atrial fibrillation Paroxysmal atrial fibrillation Duration of atrial fibrillation (yea&t Mean Median Range New-onset atrial fibrillation
Cases (n = 134)
(n = 131)
OddsRatio (95% Cl)
60% (79/131) 40% (52/131)
56% (74/131) 44% (57/131)
1.17 (0.72, 1.91) 0.86 (0.53, 1.39)
?: O.OZi-25
it O.Oi-23 27% (35/129)
29% (37/128)
Cl = confidence interval. *All diirences statistically nonsignikcant (p > 0.05). tDuration specified in 74 cases and 80 controls. tMierance in the mean.
158
August 1991 The American Journal of Medicine
Volume 91
COfttfOlS
-0.80 (-2.5,0.92)t 1.09 (0.64, 1.86)
STROKE
AND NONRHEUMATIC
ATRIAL
FIBRILLATION
/ MOULTON
ET AL
TABLE III EchocardiographyAssessment:Caseswith Stroke Compared with ControlsWithout Stroke*
Echocardiograms performed Left atrial size Mean (median), mm > 40 mm (%) > 50 mm (%I Valvular disease Any valvular disease Mitral regurgitation+ Mitral valve prolapse Mitral annular calcification Other vakular disease
Cases (n = 134)
Controls (n = 131)
60% (81/134)
49% (64/131)
43.7 (43.5) 60% (36/60) 23% (14/60)
43.3 (45.0) 61% (25/41)
63% (44/70) 48% (21/44)
61% (37/61) 27% (10/37)
11% (5/44)
19% (7/371
27% (12/44) 14% (6/44)
32% (12/37) 22% (8137)
Odds Ratio (95% Cl)
0.96 (0.42,2.17) 2.82 (0.88,8.93)
10% (4/41)
1.10 (0.54,2.26) 2.47 0.55 0.78 0.57
(0.97,6.27) (0.16, 1.89) (0.30,2.03) (0.18, 1.82)
Cl = confidenceinterval. ‘All comparisons were nonsignificant.
tlncludes patients with other coexistent valvular disease.
trols, but this difference did not achieve statistical significance. There was no difference in mitral annular calcification. Multivariate Analyses The increased risk of stroke conferred by age and the increased risk of stroke conferred by hypertension were largely independent. Table IV presents these risks via a stratified analysis in which patients were grouped according to whether they had no risk factors, one risk factor, or both. Note that patients who were both older than 75 and had a history of hypertension had an odds ratio for stroke of 3.26. Analyzed by a multiple logistic regression equation including hypertension and age (dichotomized at less than or equal to 75, greater than 75 years), both factors had coefficients that were statistically significant (p 75, no HTN Age 5 75, yes HTN Age > 75, yesHTN
1991 The American
Journal
z3
1.91 3.26
of Medicine
Volume
91
159
STROKE AND NONRHEUMATIC ATRIA1 FIBRILLATION / MOULTON ET AL
Additional possible risk factors for stroke in nonrheumatic atrial fibrillation have been previously reported. The Framingham Study analysis also identified congestive heart failure and coronary artery disease, features not conferring excess risk in our study. Other reports have suggested an increased risk of stroke early in the course of atrial fibrillation [4,22]. Our data provide some support for this finding. Control subjects admitted specifically for new-onset atrial fibrillation provide a biased comparison and removing such controls reveals an increased risk of stroke at first diagnosis of atrial fibrillation. However, conclusions about the role of duration of atrial fibrillation on stroke are limited by inaccuracy in assessing time of onset of atrial fibrillation. Embolism is a well-known complication of electrical cardioversion of atria1 fibrillation, which might suggest a greater risk with paroxysmal atrial fibrillation [23]. However, several studies have found that paroxysmal atrial fibrillation actually confers a smaller risk of embolism than sustained atrial fibrillation [24-261. We found no difference in risk between paroxysmal and sustained atria1 fibrillation. Studies of left atrial size and stroke in patients with atrial fibrillation have produced conflicting results [8,27-321. In our study, left atrial size was not associated with stroke, although there was a suggestion that the presence of an extremely large left atrium (greater than 50 mm) increased risk. Recent echocardiographic analyses have suggested that mitral annular calcification in some way predisposes to stroke [8,33]. Our study does not support this role for mitral annular calcification. Potential methodologic limitations of this study merit discussion. In a retrospective study, missing data are not retrievable. Data were essentially complete for our major risk factors-age and hypertension. However, many subjects did not have echocardiography measurements. In a case-control study, there is also concern about bias in the sampling of the cases and controls. Our cases of hospitalized stroke represented the more severe end of the clinical spectrum and our conclusions may not bear on the risk of milder or even “silent” stroke [34]. The controls in a case-control study should be a fair sample of the population that gives rise to the cases. In this study, the theoretically ideal control group would be a random sample of patients with nonrheumatic atrial fibrillation who would come to our study hospital if they had a stroke. In practice, such a group is difficult to define and survey. Other hospitalized patients, the group we chose, are an attractive alternative source of controls, so long as their reasons for hospitalization were not related to the exposures being explored. With regard to our 160
August 1991 The American Journal of Medicine
major findings, that age and hypertension were associated with stroke, we would anticipate that if hospitalized controls introduce any bias, it would be towards older age and more hypertension-biases that could not explain our findings.
ACKNOWLEDGMENT We are indebted to Dr. Steven J. Skates for stahstical assistance, Helen O’Connor for preparation of the manuscript, Tina Desai for data entry, and Hilda Sequeira for help with the hospital records.
REFERENCES 1. Alpert JS. Petersen P, Gotdfredsen J. Atrial fibrillation: natural history, complications and management. Am Rev Med 1988; 39: 41-52. 2. Dunn M. Alexander J. deSilva R. Hildner F. Antithrombolic therapy in atrial fibrillation. Chest 1989; 95 (suppl): 118S-27s. 3. Wolf PA, Dawber TR. Thomas HE, Kannel WE. Epidemiologic assessment of chronic atrial fibrillation and risk of stroke: the Framingham Study. Neurology 1978; 28: 973-7. 4. Wolf PA, Kannel WB. McGee DL. Duration of atrial fibrillation and imminence of stroke: the Framingham Study. Stroke 1983; 14: 664-7. 5. Wolf PA, Abbott RD. Kannel WB. Atrial fibrillation: a major contributor to stroke in the elderly: the Framingham Study. Arch Intern Med 1987; 147: 1561-4. 6. Petersen P, Godtfredsen J. Boysen G. Andersen E, Andersen B. Placebocontrolled, randomized trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK Study. Lancet 1989; 1: 175-9. 7. Stroke Prevention in Atrial Fibrillation Study Group Investigators. Preliminary report of the Stroke Prevention in Atrial Fibrillation Study. N Engl J Med 1990; 322: 863-8. 8. The Boston Area Anticoagulation Trial in Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with non-rheumatic atrial fibrillation. N Engl J Med 1990; 323: 1505-l 1. 9. Weyman AE. Cross-sectional echocardiography. Philadelphia: Lea & Febiger. 1982. 10. Crunch Statistical Package. Oakland, California: Crunch Software Corporation, 1987. 11. Miettinen OS. Estimability and estimation in case-referent studies. Am J Epidemiol 1976; 103: 226-35. 12. Royal Statistical Society, London 1985; GLIM 3.77 update 1. 13. Flegel KM, Shipley MJ. Rose G. Risk of stroke in nonrheumatic atrial fibrillation Lancet 1987; 1: 526-9. 14. Flegel KM, Hanley J. Risk factors for stroke and other embolic events in patients with nonrheumatic atrial fibrillation. Stroke 1989; 20: lOCO-4. 15. Aronow WS. Gutstein H, Hsieh FY. Risk factors for thromboembolic stroke in elderly patients with chronic atrial fibrillation. Am J Cardiol 1989; 63: 366-7. 16. Petersen P. Kastrup J. Helweg-Larsen S, Boysen G. Godtfredsen J. Risk factors for thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK Study. Arch Intern Med 1990; 150: 819-21. 17. Landefeld CS, Goldman L. Major bleeding in outpatients treated with warfarin: incidence and prediction by factors known at the start of outpatient therapy. Am J Med 1989; 87: 144-52. 18. Gurwitz JH. Goldberg RJ, Holden A, Knapic N. Ansell J. Age-related risks of long-term oral anticoagulant therapy. Arch Intern Med 1988; 148: 1733-6. 19. Sixty-Plus Reinfarction Study Research Group. Risks of long-term oral anticoagulant therapy in elderly patients after myocardial infarction. Lancet 1982; 1: 64-8. 20. Levine MN, Raskob G, Hirsh J. Hemorrhagic complications of long-term anticoagulant therapy. Chest 1989; 95 (suppl): 26S-36s. 21. Wintzen AR, deJonge H. Loeliger EA, Bots GTAM. The risk of intracerebral hemorrhage during oral anticoagulant treatment: a population study. Ann Neurol 1984; 16: 553-8. 22. Hart RG. Easton JD, Sherman DG. Duration of non-valvular atrial fibrillation and stroke [letter]. Stroke 1983; 14: 827. 23. Bjerkeland CJ, Orning OM. The efficacy of anticoagulant therapy in preventing embolism related to D.C. electrical conversion of atrial fibrillation. Am J Cardiol 1969; 23: 208-16. 24. Peterson P, Godtfredsen J. Embolic complications lation. Stroke 1986; 17: 622-6.
Volume 91
in paroxysmal
atrialfibril-
STROKE 25. Treseder AS, Sastry BSD, Thomas TPL, Yates MA, Pathy MSJ. Atrial fibrillation and stroke in elderly hospitalized patients. Age Ageing 1986; 15: 89-92. 26. Britton FM, Gustafson C. Non-rheumatic atrial fibrillation as a risk factor for stroke. Stroke 1985; 16: 182-8. 27. Caplan LR, D’Cruz I, Hier DB. Reddy H, Shah S. Atrial srze. atrial fibrillation and stroke. Ann Neural 1986; 19: 158-61. 26. Weiner I. Clinical and echocardiographic correlates of systemic embolization in nonrheumatic atrial fibrillation, Am J Cardiol 1987; 59: 177. 29. Ruocco NA, Most AS. Clinical and echocardiographic risk factors for systemic embolization in patients with atrial fibrillation in the absence of mitral stenosis [abstract]. J Am Coll Cardiol 1986; 7: 165A. 30. Cabin HS, Clubb S, Hall C, Perlmutter RA. Feinstein AR. Risk of systemic
AND NONRHEUMATIC
ATRIAL
FIBRILLATION
/ MOULTON
ET AL
embolization of atrial fibrillation without mitral stenosis. Am J Cardiol 1990; 65: 1112-6. 31. Tegeler CH. Hart RG, Sherman DG. Quiroga ES, Easton JD. Atrial fibrillation -stroke study [abstract]. Stroke 1986; 17: 131A. 32.Petersen P. Kastrup J, Helweg-Larson S, Boysan G, Gotfredsen J. Risk factors for thromboembolic complications in chronic atrial fibrillation: the Copenhagen AFASAK Study. Arch Intern Med 1990; 150: 819-21. 33. Aronow WS, Koenigsberg M. Kronzon I, Gutstein H. Association of mitral annular calcium with new thromboembolic stroke and cardiac events at 39month follow-up in elderly patients. Am J Cardiol 1990; 65: 1511-2. 34. Peterson P. Madsen EB, Brun B. Peterson F. Boysen G. Silent cerebral infarction in atrial fibrillation. Stroke 1987; 18: 1098-l 100.
August
1991
The American
Journal
of Medicine
Volume
91
161
