Atrial Fibrillation and Stroke
Mortality and Causes of Death After the First Acute Ischemic Stroke
Background and Purpose Atrial fibrillation (AF) is a risk factor for stroke. This study was undertaken to determine the influence of AF on the mortality of stroke patients and on the causes of death after a stroke event.
Methods Patients with first ischemic stroke who were 35 to 74 years old and registered in the FINMONICA stroke register during 1982 through 1992 were analyzed (n=6912). There were 642 patients with AF (9.3%) (mean age, 67 years) and 6270 patients without AF (90.3%) (mean age, 63 years). The association between AF and stroke mortality was investigated by use of logistic regression and Cox proportional hazards models.
Results Mortality was higher in the AF group both at 28 days (19.5% versus 14.4%, P<.001) and 1 year after the attack (30.5% versus 21.8%, P<.001). After adjustment for age and sex, the odds ratio for 28-day case fatality (AF versus non-AF) was 1.27 (95% CI, 1.03 to 1.56; P=.003), and that for 1-year mortality was 1.36 (95% CI, 1.14 to 1.63; P<.001). In the proportional hazards model, AF was a significant independent risk factor for 1-year mortality (hazard ratio, 1.26; 95% CI, 1.09 to 1.46; P=.002). Cardiac causes of death were more common in the AF group at the acute stage. In the course of 1 year, there were no differences in the distributions of causes of death.
Conclusions Stroke patients with AF are at high risk of death both at the acute phase of stroke and during the subsequent year after the first acute stroke event. Mortality from cardiac diseases prevailed in the AF group during the acute phase of stroke. Careful cardiac evaluation and treatment are essential in patients with AF and stroke.
Atrial fibrillation (AF) is known to be a major risk factor for stroke.1 2 3 4 5 6 Once stroke has occurred, it is not clear whether the subsequent short-term and especially the long-term risk of death differ between stroke patients with and without AF. Some studies have suggested that AF-associated stroke carries an increased risk of death,7 8 9 10 11 whereas others do not share this point of view.1 12 13
There are only a few long-term studies on the causes of death of AF patients who have suffered a first ischemic stroke, and the results are contradictory. Some studies have claimed that there is no difference in the causes of death between AF and non-AF patients in the acute or subacute phase after ischemic stroke,7 8 12 14 whereas others suggest that cardiac causes of death predominate in AF patients compared with non-AF patients.10 11 15 It can be argued that cerebrovascular causes of death prevail among AF patients during the acute phase, because embolic stroke relatively often results in large infarcts with an unfavorable outcome.16 17
The aim of this study was to determine the influence of AF on overall mortality and cause-specific mortality of acute stroke during the acute phase and in the course of 1 year.
Subjects and Methods
The FINMONICA Stroke Register was a population-based register and a part of the collaborative WHO MONICA Project.18 The registration methods have been described in detail previously.19 20 21 22 23 During 1982 through 1992, the FINMONICA stroke register operated in three areas of Finland: North Karelia, Kuopio, and Turku-Loimaa, with a total population of 390 000.
The register aimed at the registration of every attack of cerebrovascular disease occurring in people 25 to 74 years old and permanently resident in the target areas. In this study, we included patients 35 to 74 years old, because most of the cases in the age group 25 to 34 years were cerebral hemorrhages.
All patients who were admitted to hospital with symptoms and signs suggesting acute cerebrovascular disease and who were permanent residents of the target areas were evaluated for registration.20 22 The primary source of case finding was the hospital admission lists, but for the sake of completeness the hospital discharge diagnoses and diagnoses in death certificates also were routinely checked. Almost all patients with symptoms and clinical signs suggestive of acute stroke were hospitalized in Finland. Nurses specially trained for the stroke registration filled in the stroke record forms 28 days after the onset of the stroke attack. In cases in which the patient died within 28 days, the cause of death, duration of survival, and autopsy findings were recorded. The FINMONICA Stroke Register data were cross-checked annually at the National Public Health Institute with the computerized countrywide National Death Register for completeness of case ascertainment. Information about cases occurring in the National Death Register but not in the FINMONICA Stroke Register was sent to the stroke teams of respective areas. They retrieved relevant hospital documents and/or autopsy reports and evaluated the case according to the FINMONICA protocol.
Diagnostic Criteria and Quality Assurance
Stroke was defined as a rapidly developing symptom of focal (or global) disturbance of cerebral function lasting >24 hours (unless interrupted by surgery or death) with no apparent nonvascular cause. This category included patients showing clinical signs and symptoms suggestive of subarachnoid hemorrhage, intracerebral hemorrhage, or cerebral ischemic infarct.18
Each suspected stroke event was classified by the local register physician into one of three categories: definite stroke, no stroke, or insufficient information. The same specialist-physician experienced in stroke performed the diagnostic classification in each area during the whole 10-year study period. The category “insufficient information” was used for fatal events, especially for cases of sudden death without necropsy, or for nonfatal events if it was impossible to determine whether the symptoms were of stroke or of other disease. This category represented <1% of all registered events. According to the practice of the MONICA Project, the category “insufficient information” was combined in the analyses with the category “definite stroke.”
Only patients with first ischemic stroke were included in the present study. This subtype of stroke can be distinguished fairly reliably in the FINMONICA Stroke Register data.19 23 Until 1986, the classification of events by the subtype of stroke in cases in which CT or necropsy was not done was based on clinical symptoms and signs, existence of blood in cerebrospinal fluid, and other investigations, such as isotope brain scan. If not even lumbar puncture was performed, the patient was diagnosed as having an unspecified stroke. In 1987, more stringent requirements for the diagnosis of hemorrhagic stroke were adopted by the WHO MONICA Stroke Study, and positive signs in angiography or CT became necessary. The unspecified stroke category was not included in the present study.
The stroke was classified as the first event if there was no evidence of a previous stroke event in the patient's history. If the patient suffered from a new acute cerebrovascular attack within 28 days from the onset of the first event, the attack was considered to belong to the same event. If the episode occurred >28 days after the first event, it was classified as a recurrent stroke and a new register form was filled in.
An ECG was performed in 92% of the nonfatal cases and 82% of the fatal cases during the attack period. AF patients were identified by this ECG. If ECG was not performed during the attack period, then previous medical records were checked.
Since the main goal of the stroke register was to assess the trends in stroke incidence and mortality, the data on potential risk factors for stroke were limited. Data on the history of hypertension and myocardial infarction (MI) were obtained during the entire registration period. Both recent MI, ie, one occurring during the 28-day period preceding the stroke event, and previous MI, ie, one occurring >28 days before the stroke, were recorded during the whole study period. The stroke register form was modified over the years, and data on an earlier transient ischemic attack and diabetes mellitus were included in 1988 and 1990, respectively. Thus, the effect of transient ischemic attack and that of diabetes mellitus on 1-year mortality after stroke were not analyzed.
CT or MRI of the brain was done in 4378 of all 8677 stroke patients (50%) and in 3198 of 6912 ischemic stroke patients (46%).
Autopsy was performed in 31% of the fatal ischemic strokes and in 45% of all fatal strokes that occurred during 28 days from the onset of the attack. Data on the causes of death came from autopsy findings or from the underlying cause of death on death certificates. The causes of death were defined as follows: (1) stroke (first or recurrent), (2) cardiac (MI or heart failure), and (3) other causes.
One-year follow-up of mortality was carried out by computerized record linkage of the stroke register data with the National Death Register on the basis of the personal identification number, unique to every citizen in Finland. Cause-specific mortality should be viewed with some caution, since the data are not based on strictly standardized coding. Our experience in the framework of the FINMONICA Stroke and Myocardial Infarction Registers, however, indicates that the cardiovascular causes of death are fairly reliable in the National Death Register.
Standard statistical analyses were used to describe the patient sample. The significance of differences between expected and observed proportions was assessed by means of χ2 and Fisher's exact tests. Age- and sex-standardized odds ratios of acute and 1-year mortality were analyzed by logistic regression analysis. The Cox proportional hazards model with other prognostic factors as covariates was used for estimating the independent effect of AF on 1-year mortality. All statistical analyses were performed with SAS statistical software.24
There were 6912 patients who suffered their first ischemic stroke attack. Their mean age was 63.7±8.5 (±SD) years, and the proportion of men was 56%. AF was present in 642 patients (9.3% of all ischemic strokes). The proportion of ischemic stroke patients having AF increased with advancing age (Fig 1⇓). The mean age of the AF patients was 67.4±6.3 years.
The comparison group (n=6270; mean age, 63.4±8.6 years) was made up of stroke patients with no AF. The differences between the AF group and the non-AF group were studied in regard to age, sex, history of hypertension, diabetes mellitus, and previous MI and transient ischemic attack. The groups did not differ in any other respect but age and sex. The patients in the AF group were older and more often female (Table 1⇓).
The case fatality of stroke within 28 days was considerably higher in stroke patients with AF than in those without: 19.5% versus 14.4%, respectively (P<.001) (Table 2⇓). AF did not increase mortality in patients from 35 to 64 years old, whereas in patients 65 to 74 years old, those with AF had increased 28-day case fatality (P=.038). The odds ratio for the 28-day case fatality of AF adjusted for age and sex in a logistic regression model was 1.27 (95% CI, 1.03 to 1.56; P=.029).
Mortality continued to be increased in patients with AF during 1 year after acute stroke, although more clearly so during the first 2 months after the stroke (Fig 2⇓). Thus, cumulative 1-year mortality after the stroke attack was 30.5% in patients with AF and 21.8% in patients without AF (P<.001). After adjustment for age and sex, the odds ratio for 1-year mortality was 1.36 (AF versus no AF; 95% CI, 1.14 to 1.63; P<.001) among those alive at 28 days after the onset of acute stroke.
In the Cox proportional hazards model, 1-year mortality attributable to AF was 1.26 (95% CI, 1.09 to 1.46; P=.002). When the model was run by age group, the hazard ratio was statistically significant in the age group 65 to 74 years (1.23; 95% CI, 1.04 to 1.45; P=.016) (Table 3⇓). In the age group 35 to 64 years, the result was not statistically significant, because of the smaller number of events, but the point estimate of the hazard ratio of AF was similar to that in the older age group (hazard ratio, 1.29; 95% CI, 0.92 to 1.81; P=.14). In both age groups, the most important risk factor predicting 1-year mortality was the history of recent MI. A positive history of previous MI also significantly influenced the 1-year prognosis, in particular in the older age group.
During the first 28 days, ie, the acute stage, the distribution of the underlying causes of death was different in regard to cardiac causes, which were more common in the AF group (Table 4⇓). There was no difference concerning stroke as the underlying cause of death. From 28 days up to 1 year, the mortality rate was higher in the AF group than in the non-AF group. However, there was no difference in the distributions of the underlying causes of death between the two groups.
The prospective follow-up of the Framingham Study cohort failed to observe an association between AF at baseline and the subsequent risk of fatal stroke.1 In the Oxfordshire Community Stroke Project,9 AF-related stroke was related to increased mortality during the acute phase of stroke, but in the long term, AF was only a weak predictor of total mortality. It is likely that prospective cohort studies have a classification bias, because data on the exposure are usually available at the baseline only, and not at the time of the event. If the outcome event is death, as in the Framingham study, such a bias cannot be detected or corrected for.
Only a few hospital-based studies have found no association between AF and excess stroke mortality.12 13 In most hospital-based studies, AF has contributed to stroke mortality, but the follow-up time in these studies has usually been short,7 8 and in some studies, the case ascertainment has been incomplete when long-term mortality was studied.15 16 In our study, the complete coverage of mortality surveillance was ensured by comparing the FINMONICA stroke register with the data from the National Death Register by use of personal identification codes of the deceased.
Data on anticoagulation therapy were not available, but the use of this treatment was probably limited at the time of our study. We did not have good enough data to distinguish between chronic and paroxysmal AF nor information on the duration of AF. According to the Stroke Prevention in Atrial Fibrillation investigators, the duration and intermittency of AF were not predictors of thromboembolism in univariate analysis.25
The early case-fatality among stroke patients with AF was 19.5%, and thus in the lower range of the 18% to 38% found in the previous studies.7 8 9 15 26 One-year mortality was also increased among AF patients, and it was 30.5% in our study. In a Swedish study,10 1-year mortality in the AF group was considerably higher than in our study (49% in AF and 18% in the sinus rhythm group), but their patients were older and probably suffered more severe strokes, because they were all stroke unit patients.
The hazard ratio of 1-year mortality was 1.26 for AF in our study. It is close to that observed in the community-based Oxfordshire study9 but lower than the 1.55 for 6-month mortality in an Italian study.8 Broderick et al11 reported a hazard ratio of 1.7 for AF, but they did not have any upper age limit, and their follow-up time was 5 years. In an Australian stroke study,27 the hazard ratio of 1-year mortality for AF was as high as 2.3.
During a longer follow-up beyond the initial hospital discharge after acute stroke, there are only a few, contradictory reports on the causes of death among AF patients who have recovered from their first acute stroke, and these studies are all hospital-based.8 10 12 15 Candelise et al8 followed patients for 6 months and Friedman12 for 18 months, and they did not see any differences in the causes of death between AF and non-AF patients. The study by Sage and van Uittert15 dealt with 87 AF patients who had survived stroke, of whom only 59 could be traced during the follow-up time of 9 years. Only 12% of their patients died of a recurrent stroke and 34% of cardiac disease. Gustafsson and Britton10 followed ischemic stroke patients for 5 years. Cardiac disease was the underlying cause of death in approximately half of the nonvalvular AF patients, whereas in the sinus rhythm group, complications of the initial stroke or a recurrent acute stroke were the predominant causes of death. In one small autopsy study,14 there were no significant differences between the distributions of causes of death in embolic and nonembolic infarctions during 3 months.
The important finding in this study was the higher mortality rate of stroke patients with AF, which remained after adjustment for age. The unfavorable outcome among the AF patients seemed to be largely due to the existing cardiac disease, most often advanced coronary atherosclerosis, causing the arrhythmia, because cardiac causes of death prevailed among the AF patients during the acute phase after the stroke. Our study showed that the cardiac causes of death were more common in the AF group than in the non-AF group in acute stroke. Cerebrovascular causes of death are generally expected to predominate among AF patients during the acute phase of stroke. Thus, our results emphasize not only the importance of secondary prevention of stroke by anticoagulation but also the treatment of coexisting cardiac disease and cardiovascular risk factors in stroke patients with AF.
We are grateful to the hospitals, primary healthcare centers, and health personnel of the registration areas for supporting the local data collection. Much of the practical work in case findings and data collection was done by the nurses Elvi Surakka in North Karelia; Anu Mononen in Turku; Hannele Kastarinen, Sinikka Röynä, Eva Talasniemi, and Markku Kalinen in Kuopio; and Aino-Maija Kantee in Loimaa.
- Received August 29, 1996.
- Revision received October 24, 1996.
- Accepted October 25, 1996.
- Copyright © 1997 by American Heart Association
Wolf PA, Kannel WB, McGee DL, Meeks SL, Bharucha NE, McNamara PM. Duration of atrial fibrillation and imminence of stroke: the Framingham study. Stroke. 1983;14:664-667.
Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham study. Stroke. 1991;22:983-988.
Britton M, Gustafsson C. Non-rheumatic atrial fibrillation as a risk factor for stroke. Stroke. 1985;16:182-187.
Candelise L, Pinardi G, Morabito A, and the Italian Acute Stroke Study Group. Mortality in acute stroke with atrial fibrillation. Stroke. 1991;22:169-174.
Sandercock P, Bamford J, Dennis M, Burm J, Slattery J, Jones L, Boonyakarnkul S, Warlow C. Atrial fibrillation and stroke: prevalence in different types of stroke and influence on early and long term prognosis (Oxfordshire Community Stroke Project). BMJ. 1992;305:1460-1465.
Broderick JP, Phillips SJ, O'Fallon WM, Frye RL, Whisnant JP. Relationship of cardiac disease to stroke occurrence, recurrence and mortality. Stroke. 1992;23:1250-1256.
Friedman PJ. Atrial fibrillation after stroke in the elderly. Stroke. 1991;22:209-214.
Censori B, Camerlingo M, Casto L, Ferraro B, Gazzaniga GC, Cesana B, Mamoli A. Prognostic factors in first-ever stroke in the carotid artery territory seen within 6 hours after onset. Stroke. 1993;24:532-535.
Sage JI, van Uittert RL. Risk of recurrent stroke in patients with atrial fibrillation and non-valvar heart disease. Stroke. 1983;14:537-540.
Caplan LR, Hier DB, D'Cruz I. Cerebral embolism in the Michael Reese Stroke Registry. Stroke. 1983;14:530-536.
Sarti C, Tuomilehto J, Sivenius J, Kaarsalo E, Narva E, Salmi K, Torppa J, Salomaa V. Declining trends in incidence, case-fatality and mortality of stroke in three geographical areas in Finland during 1983 to 1989: results from the FINMONICA Stroke Register. J Clin Epidemiol. 1994;47:1259-1269.
Tuomilehto J, Sarti C, Narva E, Salmi K, Sivenius J, Kaarsalo E, Salomaa V, Torppa J. The FINMONICA Stroke Register: community based stroke registration and analysis of stroke incidence in Finland, 1983-1985. Am J Epidemiol. 1992;135:1259-1270.
Sarti C, Tuomilehto J, Sivenius J, Kaarsalo E, Narva E, Salmi K, Salomaa V, Torppa J. Stroke mortality and case fatality rates in three geographical areas of Finland from 1983 to 1986. Stroke. 1993;24:1140-1147.
Tuomilehto J, Rastenyte D, Sivenius J, Sarti C, Immonen-Räihä P, Kaarsalo E, Kuulasmaa K, Narva E, Salomaa V, Salmi K, Torppa J. Ten year trends in stroke incidence and mortality in the FINMONICA Stroke Study. Stroke. 1996;27:825-832.
Sarti C, Tuomilehto J, Salomaa V, Sivenius J, Kaarsalo E, Narva E, Salmi K, Torppa J. Epidemiology of subarachnoid hemorrhage in Finland during 1983 to 1985. Stroke. 1991;22:848-853.
SAS Institute Inc. SAS/Stat Users Guide, Version 6. 4th ed. Cary, NC: SAS Institute, Inc; 1990:vol 1 and 2.
The Stroke Prevention in Atrial Fibrillation Investigators. Predictors of thromboembolism in atrial fibrillation: clinical features of patients at risk. Ann Intern Med. 1992;116:1-5.
Anderson CS, Jamrozik KD, Broadhurst RJ, Stewart-Wynne EG. Predicting survival for 1 year among different subtypes of stroke. Stroke. 1994;25:1935-1944.