Atrial Fibrillation Is an Independent Determinant of Low Cognitive Function. A Cross-Sectional Study in Elderly Men

Transcription

1 Is an Independent Determinant of Low Cognitive Function A Cross-Sectional Study in Elderly Men Lena Kilander, MD, PhD; Bertil Andrén, MD, PhD; Håkan Nyman, PhD; Lars Lind, MD, PhD; Merike Boberg, MD, PhD; Hans Lithell, MD, PhD Background and Purpose Cerebrovascular disease is increasingly recognized as a cause of dementia and cognitive decline. We have previously reported an association between hypertension and diabetes and low cognitive function in the elderly. Atrial fibrillation is another main risk factor for cerebrovascular disease. The aim of this study was to investigate whether atrial fibrillation is associated with low cognitive function in elderly men with and without previous manifest stroke. Methods This was a cross-sectional study based on a cohort of 952 community-living men, aged 69 to 75 years, in Uppsala, Sweden. Cognitive functions were assessed by the Mini Mental State Examination and the Trail Making Tests, and a composite z score was calculated. The relation between atrial fibrillation and cognitive z score was analyzed, with stroke and other vascular risk factors taken into account. Results All analyses were adjusted for age, education, and occupational level. Men with atrial fibrillation (n 44) had lower mean adjusted cognitive z scores ( ) than men without atrial fibrillation ( ; P ). The exclusion of stroke patients did not alter this relationship; the mean cognitive z score was in the 36 men with atrial fibrillation and in those without atrial fibrillation (P ), corresponding to a difference of 0.4 SDs between groups. Adjustments for 24-hour diastolic blood pressure and heart rate, diabetes, and ejection fraction did not change this relationship. Men with atrial fibrillation who were treated with digoxin (n 27) performed markedly better ( ) than those without treatment (n 9; ; adjusted P ). Previous myocardial infarction was not associated with impaired cognitive results. Conclusions In these community-living elderly men, we found an association between atrial fibrillation and low cognitive function independent of stroke, high blood pressure, and diabetes. Interventional studies are needed to answer the question of whether optimal treatment of atrial fibrillation may prevent or postpone cognitive decline and dementia. (Stroke. 1998;29: ) Key Words: atrial fibrillation cognition risk factors stroke Cognitive decline and dementia are major causes of disability in the elderly. Cerebrovascular disease contributes to a substantial part of all cases of dementia. 1 Hypertension, diabetes, and atrial fibrillation are major risk factors of stroke. We have previously reported that high 24-hour diastolic blood pressure (BP) and diabetes were associated with low cognitive performance in a cohort of stroke-free elderly men. 2 The aim of the present study was to investigate the association between atrial fibrillation and cognition. In the same cohort, we examined relations between atrial fibrillation and cognitive function, taking stroke and possible confounders such as high BP, diabetes, myocardial infarction, and ejection fraction into account. Subjects and Methods A longitudinal health survey focusing on cardiovascular risk factors was started in the early 1970s in Uppsala, Sweden. The original cohort was defined as all 50-year-old men, ie, those born in , who lived in Uppsala at that time. The baseline examination included information on socioeconomic factors; measurements of BP, blood glucose, serum insulin, lipids and lipoproteins; and recording of ECG. 3 The participation rate was 82% (n 2322). Primary preventive treatment was initiated in subjects at high risk for vascular disease. 4 Twenty years later, 448 men had died, and 193 men had moved. Of the 1681 men still alive and living in Uppsala county, 99 men died before being asked and 361 men refused to participate. A total of 1221 men (65.2% of all 1874 survivors) took part in the follow-up at age 70. This examination covered analyses of several variables linked to vascular disease. ECG was recorded in 1136 men, and echocardiography was performed in the first consecutive 584 men. All 1221 participants were invited to a testing of cognitive functions, in which 999 men (82%) participated. For a total of 952 men, both ECGs and cognitive test results were available. Four hundred nineteen men took part both in the cognitive testing, the ECG, and the echocardiographic examination. Received March 12, 1998; final revision received June 8, 1998; accepted June 8, From the Department of Public Health and Social Sciences/Geriatrics (L.K., M.B., H.L.) and Department of Clinical Physiology, Uppsala University, Uppsala (B.A.); Department of Clinical Neurosciences, Karolinska Institute, Stockholm (H.N.); and Department of Internal Medicine, Uppsala (L.L.) (Sweden). Correspondence to Dr Lena Kilander, Department of Public Health and Caring Sciences/Geriatrics, PO Box 609, SE Uppsala, Sweden American Heart Association, Inc. 1816

2 Kilander et al September Cardiovascular Factors Subjects with a stroke before the cognitive testing were identified by data from the Swedish National Inpatient Register, covering all diagnoses in hospitalized patients from 1970 and onward, and/or by a positive answer to the question Have you had a stroke: cerebrovascular infarct or bleeding? in connection with the testing. The following diagnoses, according to the International Classification of Diseases (ICD)-8 or ICD-9, were used for classification of stroke: intracerebral hemorrhage (431), thromboembolic stroke (433 and 434), transient ischemic attack (435), and acute ill-defined cerebrovascular disease (436). A previous myocardial infarction was identified through the Inpatient Register, ICD diagnoses 410 to 412. As previously described, 2 24-hour ambulatory BP was monitored with Accutracker 2 equipment (Suntech Medical Instruments Inc). An oral glucose tolerance test identified subjects with diabetes according to the classification of the National Diabetes Data Group. 5 Resting standard 12-lead ECGs were obtained. They were read and coded by the same physician (L.L.) with regard to presence or absence of atrial fibrillation. Comprehensive 2-dimensional and Doppler echocardiography was performed with a Hewlett Packard Sonos 1500 cardiac ultrasound unit, as described previously. 6 A 2.5-MHz transducer was used for the majority of 2-dimensional, M-mode, and pulsed-wave Doppler examinations. All subjects were examined by 1 experienced physician (B.A.). As a measurement of left ventricular systolic function, ejection fraction was computed according to the Teichholz M-mode formula. 7 Treatment with agents affecting BP, irrespective of indication (ie, -blockers, calcium antagonists, angiotensin-converting enzyme inhibitors, diuretics, and -blockers) and treatment with digoxin, low-dose aspirin, and warfarin were registered in a questionnaire. Educational level was stratified as low (elementary school only, 6 to 7 years), medium (secondary school), or high (university studies). Main previous occupational level was divided into 3 categories: low (manual workers), medium (foremen, clerks, and salesmen), and high (major professionals, business managers). Cognitive Function The psychometric testing included the Trail Making Tests (TMTs) A (n 998) and B (n 996) 8 and the Mini Mental State Examination (MMSE) 9 (n 891, since it was added to the protocol after the start). The TMTs were selected to assess speed and shifting capacity, which are measurements of subcorticofrontal function, among others. 10 The MMSE is an instrument widely used in the screening for cognitive disorders. Standardized procedures from published manuals were used in the administration and evaluation. The maximum time set for the TMTs was 4 minutes. After a logarithmic transformation of the test results, a z transformation was applied, and a composite cognitive score was calculated for each subject as the sum of the z scores, divided by the number of tests performed. Thus, 0.0 is equal to the mean result in the population; 1.0 is a result 1 SD over the mean score. Informed consent was obtained from the participants after the nature of the procedures had been fully explained. The study was approved by the Ethics Committee at Uppsala University. Statistical Analysis Values are expressed as means and adjusted means or, in case of skewed distributions (MMSE, TMT), as median values. Student s unpaired t test was applied for comparisons between independent groups, the 2 test was used for analysis of relations between categorical variables, and linear relations were examined with Pearson s correlation coefficient. ANCOVA was applied in multivariate models with continuous dependent variables, and logistic regression was used when outcome variables were binary. All analyses were adjusted for age. The analyses of determinants of cognitive function were adjusted for age and educational (low, medium, high) and occupational (low, medium, high) levels. Results Representativity The nonparticipating survivors (n 653) had significantly higher concentrations of blood glucose, serum insulin, and triglycerides at the baseline examination at age 50 years than the participants in the follow-up (n 1221). BP levels, rates of atrial fibrillation, and socioeconomic factors were equal between the groups. The 448 men who died during follow-up had a lower educational level and markedly higher BP, blood glucose, serum insulin, and lipids at baseline than the survivors. The 952 men who took part in the cognitive tests and the ECG did not differ from the other men investigated (n 269) with regard to rates of myocardial infarction, stroke, atrial fibrillation, treatment, previous occupational level, or mean 24-hour diastolic BP. In nonparticipants, diabetes was more frequent (21.5%) than in participants (13.1%; P 0.001), and slightly more nonparticipants than participants had low levels of education (66% versus 54%; P NS). Description of the Study Cohort Characteristics of the participants (n 952) are presented in Table 1. Eighty-four men had suffered a stroke before the testing, and 44 men (4.6%) had atrial fibrillation. One third of the cohort were treated with antihypertensive agents, irrespective of indication. The median score in the MMSE was 29 points, and 45 men (5.3%) scored 26 points. Measurements of ejection fraction were available in 358 men. They were equal to the others regarding all variables, except that they were older. Characteristics of Men With Stroke and As shown in Table 2, men with a previous stroke (n 84) had a higher rate of atrial fibrillation and myocardial infarction and lower ejection fraction than the stroke-free participants. Twenty-four-hour diastolic BP and the rate of diabetes did not differ significantly between groups. Characteristics of men with atrial fibrillation (n 44) are presented in Table 3. TABLE 1. Description of the Study Cohort (n 952). (%) Mean SD (Range) Low education 510 (53.6%) Stroke 84 (8.8%) Atrial fibrillation 44 (4.6%) Myocardial infarction 92 (9.7%) Diabetes (n 930, OGTT) 122 (13.1%) Antihypertensive treatment 330 (34.7%) Treatment with digoxin 65 (6.8%) Treatment with 147 (15.4%) aspirin/warfarin Age at cognitive testing, y (69 to 75) Cognitive score ( 5.2 to 2.0) 24-h diastolic BP, mm Hg 76 8 (55 to 109) (n 906) Ejection fraction (n 358) (0.24 to 0.87) MMSE, points 29* (21 to 30) TMT A, s 43* (17 to 240) TMT B, s 110* (43 to 240) *Median values.

3 1818 and Cognition TABLE 2. Characteristics of Participants With Stroke TABLE 4. Cognitive Performance in Stroke Stroke P n Low education 53.5% 54.8% Atrial fibrillation 9.5% 4.2% h diastolic BP, mm Hg 77 (8) 76 (8) h heart rate, bpm 70 (9) 70 (10) Ejection fraction (n 33;325) 0.58 (0.13) 0.65 (0.10) Diabetes (OGTT) 14.8% 13.0% Myocardial infarction 20.2% 8.6% Antihypertensive treatment 53.6% 32.8% Treatment with aspirin/warfarin 61.9% 10.9% Treatment with digoxin 15.5% 6.0% Values are mean (SD). P values are adjusted for age. Double n values refer to subjects with and without stroke. They had higher 24-hour diastolic BP and heart rate and a higher rate of diabetes than the other men. Confounding Factors As previously described, age, education, occupation, high 24-hour diastolic BP, and diabetes were independent determinants of low cognitive function. 2 Participants with antihypertensive treatment performed equal to the rest of the cohort. Neither previous myocardial infarction (n 92) nor low ejection fraction was associated with impaired cognitive results independent of atrial fibrillation (data not shown). Cognitive Function in and Stroke Stroke patients had lower results on the MMSE and the TMTs; the adjusted mean cognitive score was versus in stroke-free participants (P ), TABLE 3. Characteristics of Participants With P n Low education 63.6% 53.1% Stroke 18.2% 8.4% h diastolic BP, mm Hg 79 (10) 76 (8) h heart rate, bpm 76 (12) 69 (10) Ejection fraction (n 18;340) 0.60 (0.13) 0.65 (0.10) Diabetes (OGTT) 23.8% 12.6% Myocardial infarction 4.6% 9.9% Antihypertensive treatment 65.9% 33.2% Treatment with aspirin/warfarin 50.0% 13.8% Treatment with digoxin 75.0% 3.5% Values are mean (SD). P values are adjusted for age. Double n values refer to subjects with and without atrial fibrillation. Stroke n MMSE, points (n 78;770) 28 (25 to 30) 29 (27 to 30) TMT A, s 50 (32 to 92) 42 (28 to 69) TMT B, s 124 (82 to 234) 109 (66 to 193) Composite cognitive score * Range 10% to 90% 1.65 to to 0.95 Adjusted mean (composite score). Double n values refer to subjects with and without stroke. *Stroke, P ; age, P ; education, P ; occupation, P In a model including age, education, occupation, 24-h diastolic BP, 24-h heart rate, diabetes, antihyperetensive treatment, and ejection fraction, the P value for stroke was ie, a difference of 0.31 SD (Table 4). Stroke is a possible intermediate in the relation between atrial fibrillation and low cognitive function. Therefore, separate analyses were made with and without stroke patients. In the total population, men with atrial fibrillation had lower adjusted cognitive score ( ) than men without atrial fibrillation TABLE 5. Cognitive Performance in Total cohort n MMSE, points (n 37;811) 28 (25 to 30) 29 (26 to 30) TMT A, s 50 (36 to 88) 43 (28 to 69) TMT B, s 134 (82 to 240) 109 (67 to 196) Composite cognitive score * Range 10% to 90% 1.54 to to 0.94 Adjusted mean Stroke patients excluded n MMSE, points (n 29;741) 28 (24 to 30) 29 (27 to 30) TMT A, s 49 (36 to 99) 42 (28 to 69) TMT B, s 134 (83 to 240) 108 (66 to 192) Composite cognitive score Range 10% to 90% 1.79 to to 0.98 Adjusted mean (composite score). Double n values refer to subjects with and without atrial fibrillation. *Atrial fibrillation, P ; age, P ; education, P ; occupation, P Atrial fibrillation, P ; age, P ; education, P ; occupation, P In a model including age, education, occupation, 24-h diastolic BP, 24-h heart rate, diabetes, antihypertensive treatment, and ejection fraction (n 313), the P value for atrial fibrillation was

4 Kilander et al September TABLE 6. Digoxin Treatment in and Cognitive Performance Digoxin Treatment n 27 9 MMSE, points (n 20;9) 28 (26 to 30) 28 (21 to 29) TMT A, s 46 (35 to 65) 65 (44 to 240) TMT B, s 116 (79 to 177) 203 (92 to 240) Composite cognitive score * Range 10% to 90% 1.23 to to 0.11 Adjusted mean (composite score). Stroke patients are excluded. Double n values refer to subjects with and without digoxin treatment. *Digoxin treatment, P 0.003; age, P 0.175; education, P 0.794; occupation, P In a model including age, education, occupation, 24-h diastolic BP, 24-h heart rate, and diabetes, the P value for digoxin treatment was ( ; P ) (Table 5). Exclusion of participants with stroke did not change this relationship; cognitive score was in the 36 men with atrial fibrillation and in men without atrial fibrillation (P ), ie, a difference of 0.41 SD (Table 5). Results in the MMSE and in the TMTs were equally affected. The relationship between atrial fibrillation and low cognitive score was similar after adjustment for 24-hour diastolic BP and heart rate, diabetes, ejection fraction, and antihypertensive treatment. Men with atrial fibrillation and concomitant antihypertensive treatment (n 29) performed equal to the other 15 men with atrial fibrillation ( 0.37 versus 0.48; P NS). Eight men had the combination of stroke and atrial fibrillation. Their adjusted mean cognitive score was , and the mean score in stroke patients without atrial fibrillation was (P 0.755). Only 3 subjects had atrial fibrillation at age 50 years; their mean cognitive score was 0.27 (P NS compared with the rest of the cohort). Digoxin Treatment and Cognitive Function Men with atrial fibrillation who were treated with digoxin (n 27) performed markedly better than those without treatment (n 9) (Table 6). The difference remained highly significant after adjustment for 24-hour diastolic BP, heart rate, and diabetes. Ejection fraction did not differ between treated and untreated men and was not included in the model because of the small number. Test results did not differ in men without atrial fibrillation with regard to treatment with digoxin. In those with atrial fibrillation, test results were equal between men treated with low-dose aspirin or warfarin (n 17) and untreated men (n 21) and between men with and without antihypertensive treatment. Discussion In these elderly men, those with atrial fibrillation had lower cognitive function independent of socioeconomic factors, 24-hour diastolic BP, diabetes, and ejection fraction, even after exclusion of subjects with a previous stroke. The difference in the mean adjusted cognitive z score between stroke-free men with and without atrial fibrillation was 0.41 SD, ie, somewhat higher than the difference between men with and without a previous manifest stroke (0.31 SD). In men with atrial fibrillation, digoxin treatment was associated with better cognitive performance. This cohort consisted of community-living men with few cases of cognitive impairment; only 5.3% had a score of 26 points on the MMSE, which is a low rate compared with other populations, eg, the Rotterdam Study cohort. 11 Our cohort is likely to be healthier than a general population since nonresponders had a lower educational level and higher serum concentrations of glucose, insulin, and triglycerides at baseline. Furthermore, special efforts to treat vascular risk factors had been taken since the baseline examination. Thus, our participants probably were selected with regard to both a decreased vascular risk and a higher cognitive function, ie, factors with negative relation to the outcome. For this reason the relationship between atrial fibrillation and low cognitive function may be different in a general population. In the Rotterdam Study, stroke-free women with atrial fibrillation had a higher risk of dementia and of cognitive impairment without dementia, defined as performance in the MMSE 26 points. 11 This is in concordance with our results, but the difference in the MMSE score between men with and without atrial fibrillation was quite small. Measurements of subcorticofrontal functions, such as the TMTs, may be even more sensitive than the MMSE in detecting vascular cognitive decline. This has been indicated in neuroimaging studies in which white matter lesions were associated with impaired results in the TMTs. 10,12,13 We did not find any associations between impaired cognitive performance and other markers of cardiac disease, such as previous myocardial infarction or low ejection fraction. This is in contrast to other reports in which myocardial infarction has been linked to poorer cognitive performance, 14 to dementia in elderly women, 15 and to development of dementia after stroke in subjects with multiple cerebral infarctions. 16 However, in the Cardiovascular Health Study, in which 3301 subjects were examined with MRI, white matter lesions were related to high systolic BP but not to coronary heart disease. 17 In the absence of manifest cerebrovascular disease, how may atrial fibrillation be linked to impaired cognitive function? Are they parallel markers of exposure to atherosclerotic factors, or is there a causal relationship? A reverse causation does not seem plausible. These cross-sectional findings can only be used as a basis for speculations on possible pathophysiological mechanisms. The relation between atrial fibrillation and manifest stroke is well established. 18 Atrial fibrillation has also been associated with silent lacunar infarcts in elderly patients without previous stroke or carotid lesions, 19 and it may be a more common cause of silent embolic cerebrovascular lesions than previously assumed. Myocardial dysfunction may also cause periodic disability in maintaining cerebral blood flow, with subsequent ischemic lesions. Results from a neuropathological study suggested that a failing heart may contribute to dementia through episodes of insufficient cerebral blood flow in the combination of small-vessel

5 1820 and Cognition stenosis and intermittent BP fall. 20 However, in our population, neither low systolic function, low BP, nor extreme nocturnal dipping 2 was associated with lower cognitive performance. Interestingly, digoxin treatment in atrial fibrillation was associated with better performance. This relation is unlikely to be due to biased recall, since cognitive performance did not differ between users and nonusers of other medications. However, differences in baseline cognitive level might determine type of treatment and adherence to medical controls. Speculatively, regularization of cardiac rhythm might prevent minor cerebral ischemic lesions, but this hypothesis is a subject for a randomized trial. Impaired cognitive function is the most sensitive and specific outcome measurement of cerebral target organ damage. Our results add support to a relation between vascular risk factors and cognitive impairment. This may have implications not only for vascular dementia but also for neurodegenerative dementia disorders. In the Rotterdam Study, atrial fibrillation was linked to Alzheimer s disease as well as to vascular dementia. 11 Alzheimer s disease and vascular dementia in the very elderly may have common etiologic pathways, eg, atrial fibrillation and hypertension. 21 We conclude that further studies are needed to investigate whether treatment of atrial fibrillation 22 and other risk factors might prevent or postpone dementia. Acknowledgments This study was supported by grants from the Swedish Medical Research Council (grant 5446), the King Gustav V and Queen Victoria Foundation, the Foundation of Old Servants, the Alzheimer Foundation, the Dementia Foundation in Sweden, the Swedish Stroke Foundation, and the Swedish Hypertension Society. We are grateful to Gun-Britt Ångman for excellent coordination and care of the study participants. References 1. Skoog I, Nilsson L, Palmertz B, Andreasson L-A, Svanborg A. A population-based study of dementia in 85-year-olds. N Engl J Med. 1993; 328: Kilander L, Nyman H, Boberg M, Hansson L, Lithell H. Hypertension is related to cognitive impairment: a 20-year follow-up of 999 men. Hypertension. 1998;31: Hedstrand H. A Study of Middle-Aged Men With Particular Reference To Risk Factors for Cardiovascular Disease [thesis]. Uppsala, Sweden: Uppsala University; Lithell H, Åberg H, Selinus I, Hedstrand H. The Primary Preventive Study in Uppsala: fatal and non-fatal myocardial infarction during a 10-year follow-up of a middle-aged male population with treatment of high-risk individuals. Acta Med Scand. 1984;215: National Diabetes Data Group. Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes. 1979;28: Andrén B. Left Ventricular Morphology and Function in a Population Sample of Elderly Men [thesis]. Uppsala, Sweden: Uppsala University; Teichholz LE, Kreulen T, Herman MV, Gorlin R. Problems in echocardiographic volume determinations: echocardiographic-angiographic correlations in the presence or absence of asynergy. Am J Cardiol. 1976; 37: Lezak M. Neuropsychological Assessment. 3rd ed. New York, NY: Oxford University Press; Folstein MF, Folstein SE, McHugh PR. Mini-Mental State : a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12: Breteler M, van Amerongen N, van Swieten J, Claus J, Grobbee D, van Gijn J Hofman A, van Harskamp F. Cognitive correlates of ventricular enlargement and cerebral white matter lesions on magnetic resonance imaging. Stroke. 1994;25: Ott A, Breteler M, de Bruyne M, van Harskamp F, Grobbee D, Hofman A. Atrial fibrillation and dementia in a population-based study: the Rotterdam Study. Stroke. 1997;28: Schmidt R, Fazekas F, Offenbacher H, Dusek T, Zach E, Reinhart B, Grieshofer P, Freidl W, Eber B, Schumacher M, Koch M, Lechner H. Neuropsychologic correlates of MRI white matter hyperintensities: a study of 150 normal volunteers. Neurology. 1993;43: DeCarli C, Murphy DGM, Tranh M, Grady CL, Haxy JV, Gillette JA, Salerno JA, Gonzales-Aviles A, Horwitz B, Rapoport SI, Schapiro MB. The effect of white matter hyperintensity volume on brain structure, cognitive performance, and cerebral metabolism of glucose in 51 healthy adults. Neurology. 1995;45: Breteler M, Claus J, Grobbee D, Hofman A. Cardiovascular disease and distribution of cognitive function in elderly people: the Rotterdam Study. BMJ. 1994;308: Aronson MK, Ooi WL, Morgenstern H, Hofner A, Masur D, Crystal H, Frishman W, Fisher D, Katzman R. Women, myocardial infarction, and dementia in the very old. Neurology. 1990;40: Gorelick PB, Brody JA, Cohen DC, Freely S, Levy P, Dollear W, Forman H, Harris Y. Risk factors for dementia associated with multiple cerebral infarcts a case-control analysis in predominantly African-American hospital-based patients. Arch Neurol. 1993;50: Longstreth WT, Manolio TA, Arnold A, Burke GL, Bryan N, Jungreis CA, Enright PL, O Leary D, Fried L. Clinical correlates of white matter findings on cranial magnetic resonance imaging of 3301 elderly people: the Cardiovascular Health Study. Stroke. 1996;27: Wolf PA, Dawber TR, Thomas HE, Kannel WB. Epidemiologic assessment of chronic atrial fibrillation and risk of stroke: the Framingham Study. Neurology. 1978;28: Zito M, Muscari A, Marini E, De Iorio A, Puddu GM, Abate G. Silent lacunar infarcts in elderly patients with chronic nonvalvular atrial fibrillation. Aging Clin Exp Res. 1996;8: Englund E, Brun A, Gustafson L. A white-matter disease in dementia of Alzheimer s type: clinical and neuropathological correlates. Int J Geriatr Psychiatry. 1989;4: Skoog I, Lernfelt B, Landahl S, Palmertz B, Andreasson L-A, Nilsson L, Persson G, Odén A, Svanborg A. 15-year longitudinal study of blood pressure and dementia. Lancet. 1996;347: O Connell JE, Gray CS. Atrial fibrillation and cognitive impairment. Lancet. 1992;340:1169. Letter.