Famfly Practice © Oxford University Press 1992
Vol. 9, No. 2 Printed in Great Britain
Selections from Current Literature: Prevention of Stroke in Nonrheumatic Atrial Fibrillation and Carotid Artery Stenosis ROBERT KELLY
NON-RHEUMATIC ATRIAL FIBRILLATION It is well known that patients with atrial fibrillation and rheumatic valvular heart disease should receive long-term anticoagulant therapy. 'Non-rheumatic' atrial fibrillation is more common, is associated with a 5-fold higher incidence of ischaemic stroke compared to patients in normal sinus rhythm, and accounts for an increasing proportion of strokes as age increases. Results of earlier studies of anticoagulation treatment for this condition, particularly the Copenhagenbased Atrial Fibrillation, Aspirin, Anticoagulation (AFASAK) study,2-3 demonstrated a favourable trend but were not conclusive. Stroke Prevention in Atrial Fibrillation Investigators. Preliminary report of the Stroke Prevention in Atrial Fibrillation Study. N EnglJ Med 1990; 322: 863-868. The SPAF study was a randomized trial to test the effect of warfarin or aspirin on the incidence of ischaemic stroke and systemic thromboembolism in patients with atrial fibrillation unrelated to rheumatic valvular disease. Patients with constant or intermittent atrial fibrillation, and no echocardiographic evidence Department of Family Medicine, State University of New York at Stony Brook, Stony Brook, NY 11794, USA.
231
of rheumatic mitral stenosis were recruited at 15 clinical centres in the United States. Details of the protocol have been published.4 Patients with prior cerebral ischaemia, thromboembolism, severe mitral regurgitation, or dilated cardiomyopathy were excluded. Patients eligible to receive warfarin (Group 1) based on clinical estimation were randomized to receive open-label warfarin (dose adjusted to achieve 1.3-1.8 times control prothrombin time), 325 mg enteric-coated aspirin in a double-blind fashion, or matching placebo in a double-blind fashion. Those not eligible to receive warfarin (Group 2) were randomized to treatment with aspirin or placebo. Patients were monitored every 3 months for complications of therapy and for the development of stroke or embolism. The sample was 71 % male with a mean age of 67 and a prevalence of intermittent atrial fibrillation of 34%. Group 1 and 2 patients were quite similar demographically, though Group 2 patients were 3 years older on average. After 18 months, 1244 patients had been enrolled and followed for a mean of 1.13 years; 47% had been judged as eligible for warfarin therapy and assigned to Group 1. At that time, the Safety Monitoring Committee for the study terminated the placebo arm of Group 1, because of evidence of superior results with either active agent. The study will continue to enrol patients to allow comparison of warfarin and aspirin treatment in Group 1, and comparison of aspirin and placebo in Group 2. Data were analysed using an intention to treat approach based on initial group assignment. In Group 1, 25 patients experienced primary events; 24 had ischaemic strokes while only one had systemic embolism. Patients given active treatment with either warfarin or aspirin had an incidence of 1.6 events/100 person-years, while those receiving placebo experienced 8.3 events/100 person-years. This 81% risk reduction was highly significant statistically. The rate of transient ischaemic attack (TIA) was reduced by 50% in the active treatment group, but this difference was
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
In the most recent installment of Selections from Current Literature, the effect of hypertension treatment on stroke incidence was discussed as an important outcome of the Systolic Hypertension in the Elderly Program (SHEP).1 Stroke is often a devastating event, leaving the patient, their family, and their physician with a feeling of helplessness and a dread of what the future holds. None would argue that effective generic or condition-specific prevention of stroke is a worthwhile goal. For this article, therefore, I will review a number of recent papers that describe attempts to prevent stroke in patients with one of two common clinical syndromes: non-rheumatic atrial fibrillation and carotid artery stenosis.
232
FAMILY PRACTICE—AN INTERNATIONAL JOURNAL
The Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation. NEnglJMed 1990; 323: 1505-1511. This was an unblinded randomized trial of lowdose warfarin in patients with non-rheumatic atria] fibrillation. The target prothrombin time for those assigned to warfarin treatment was 1.2-1.5 times control value, a lower range than used in the SPAF study. Prothrombin times were measured every 3 weeks to allow for adjustment of dosage. Control group patients were allowed to take aspirin as they chose. A sample of 420 patients was randomized and followed for an average of 2.3 years. The enrollment, exclusion criteria, and demographic make-up of the sample were very similar to the SPAF study. Fewer patients had intermittent atrial fibrillation (17%), and a larger
number had no clinical heart disease other than atrial fibrillation (48%). The principal end-point was ischaemic stroke; record reviewers who determined end-point status were blinded to treatment assignment. Just as in the SPAF study, the trial was terminated earlier than planned because of evidence favouring treatment. A total of 15 patients had strokes, two in the warfarin group and 13 in the control group (0.4 versus 3.0 events/100 person-years). This represents an 86% risk reduction and is highly significant statistically. Overall mortality was also reduced by 62% in the warfarin group compared to controls (11 versus 26 deaths, P < 0.005). This difference was observed despite the fact that 46% of the control group patients were taking aspirin regularly. The authors did not find a beneficial effect of aspirin in control group patients. They report instead that eight of 13 control group patients who suffered strokes were taking aspirin (seven took ^ 25 mg/day) and that half of these patients were under 75 years of age. Three 'major' and 38 'minor' bleeding events occurred in the warfarin group, compared to two major and 21 minor events in the control group. Rates of hospital admission and transfusion were quite similar between the two groups. The authors conclude that "low-dose warfarin is effective in preventing stroke in patients with nonrheumatic atrial fibrillation and that it can be quite safe."
Comment It is noteworthy that these studies both showed strong beneficial effects of treatment, replicating the less conclusive findings of the earlier AFASAK study.2'3 It seems safe to conclude that those patients with nonrheumatic atrial fibrillation who are able to take warfarin will benefit from it. Outside of study protocols, in the real world of practice, there may not be the same degree of patient compliance or supervision of anticoagulation, leading to potentially higher complication rates. It is encouraging that relatively low doses of warfarin (target prothrombin time of 1.2-1.5 times control) can be used. The Boston Area study evaluated the effect of warfarin treatment on quality of life.3 The investigators report that "no significant differences between warfarin-treated and control patients were found on well-validated measures of functional status, well-being, and health perceptions", except in patients who had bleeding complications, and conclude that "negative effects of warfarin treatment on health perceptions may be balanced by confidence in its protective effects." In contrast to warfarin, the potential role of aspirin is less clear, given the mixed findings available at this time. It seems unlikely to be effective in those patients over 75 years of age, and the AFASAK study and the Boston Area study's secondary analyses raise doubts about its general effectiveness. We will need to wait for further data from the SPAF study to shed more light on this question. In an editorial accompanying the
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
not statistically significant due to the small numbers of patients experiencing TIAs. Total mortality was not significantly reduced (3.5% for placebo; 3.1% for active treatment). There were not enough patients experiencing primary events in the active treatment arms of Group 1 to permit detection of a statistically significant difference between warfarin and aspirin. Group 1 and Group 2 patients were combined to compare aspirin and placebo. In this analysis, 52 of 57 primary events were ischaemic strokes. Patients given aspirin had an incidence of 3.2 events/100 personyears, while those receiving placebo experienced 6.3 events/100 person-years, a statistically significant 49% risk reduction in primary endpoints. TIA incidence was reduced by 61% (P = 0.06) and total mortality was reduced by 19% (P = 0.11). Subgroup analysis revealed that aspirin was not apparently effective in the 238 patients over 75 years of age, and that the 51 patients without cardiovascular disease other than atrial fibrillation experienced no primary endpoints. In patients receiving warfarin, haemorrhagic events requiring hospitalization, transfusion, or surgery occurred with an incidence of 1.7/100 patient-years; intracranial haemorrhage, including cerebral haemorrhage and subdural hematoma, occurred at a rate of 0.9/100 person-years. These rates can be compared to rates of 0.9% and 0.2% for aspirin and 1.2% and 0.2% for placebo. The authors conclude that aspirin and warfarin are separately "effective in preventing ischaemic stroke and systemic embolism and comparatively safe for patients with atrial fibrillation unrelated to rheumatic valvular disease who are acceptable candidates for anticoagulant therapy." They point out that the absolute risk reduction of 6.8%/year in Group 1 exceeded the incremental risk of clinically important haemorrhage of < 1%/year. They note the limitations of their relatively short duration of follow-up and the absence of demonstrated benefit for aspirin in patients over 75 years of age.
SELECTIONS FROM CURRENT LITERATURE 6
Boston Area study, Chesebro et al. point out that atrial fibrillation may not itself be a risk factor for stroke, but rather a "risk marker for other associated cardiovascular disease." They feel that patients with stasis-related thrombi will benefit most from anticoagulation, whereas platelet inhibition with aspirin may most benefit those with cerebrovascular atherosclerosis. They further propose out that the somewhat variable results of existing trials relate to the relative frequencies of stasis and atherosclerosis as causes of stroke, and the makeup of the samples in each study. The logical extension of this reasoning is that clinicians should choose warfarin or aspirin based on their suspicions or knowledge of patients' risk for stasis and atherosclerosis.
European Carotid Surgery Trialists' Collaborative Group. MRC European Carotid Surgery Trial: Interim results for symptomatic patients with severe (70-99%) or with mild (0-29%) carotid stenosis. Lancet 1991; 337: 1235-1243. This paper reports data from a clinical trial of carotid endarterectomy in patients with a stenotic carotid artery lesion, who had experienced a non-disabling, ipsilateral carotid territory ischaemic stroke, TIA, or retinal infarct. By angiography, patients were characterized as having mild (< 30%), moderate (30-69%), or severe (70-99%) carotid artery stenosis preoperatively. Patients whose physicians were 'substantially uncertain' whether or not to recommend carotid endarterectomy were eligible to be randomized to recommendation for 'immediate surgery' (60%) or 'no immediate surgery' (40%). All patients received the best available medical care in other respects, usually including management of hypertension, aspirin therapy, and advice to stop smoking.
Over a ten-year period, 2518 patients were randomized at 80 centres in 14 countries. The paper reports data regarding 1152 with mild or severe stenosis who were followed for an average of 3 years after surgery. Data collection for patients with moderate stenosis will continue and has not yet been reported. The principal endpoints were fatal or disabling stroke, surgeryassociated deaths (within 30 days of surgery), and nondisabling strokes lasting more than 7 days. The sample was predominantly male (70%) with a mean age of 60 years. Most had experienced a TIA in the past (78%) and half had experienced a prior stroke. For patients with severe stenosis, total mortality was lower in those patients who were allocated to surgery (9.9 versus 12.7%); this difference was not statistically significant. Surgery resulted in a rate of death or disabling stroke within 30 days of 3.7%. Those with high blood pressure (systolic > 160 mm Hg) or rapid surgery Casting less than one hour) were most likely to have an adverse outcome at 30 days. The rest of the patients assigned to immediate surgery, who had an 'apparently successful' procedure, had an eight-fold reduction in ipsilateral fatal or disabling ischaemic strokes (1.1 versus 8.4%) and a six-fold reduction in non-disabling strokes compared to the group who were not operated on. Surgery conferred an absolute risk reduction over 3 years of 9.6% (12.3 versus 21.9%) when all endpoints were combined (any stroke lasting more than 7 days; surgical death). The data also show that most of the benefit of surgery occurred in the first year for death and disabling stroke, and in the first 2 years for non-disabling stroke. For patients with mild stenosis, total mortality was actually higher in those allocated to surgery (11.4 versus 7.7%) but this difference was not statistically significant. The immediate consequences of surgery within 30 days was a rate of death or disabling stroke of 2.3 %. There was no benefit of surgery demonstrated on subsequent rates of death, disabling stroke, or nondisabling stroke, even after 3 years of follow-up. The authors conclude that carotid endarterectomy is associated with serious risks and that "successful surgery avoids the majority of the excess risk of ipsilateral ischaemic stroke associated with severe (70-99%) carotid artery stenosis. Clearly, the lower the institutional surgical risk, the more this risk will be worth taking." North American Symptomatic Carotid Endarterectomy Trial Collaborators. Beneficial effect of carotid endarterectomy in symptomatic patients with high-grade carotid stenosis. N EnglJ Med 1991; 325: 445-453. This ongoing trial in the United States and Canada took a very similar approach to the question of carotid endarterectomy in symptomatic patients with carotid stenosis of 30-99%. Entry criteria, sample demographics, and outcome measures were quite comparable to the European trial. For the group of patients with severe stenosis, the trial has been stopped because of
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
ENDARTERECTOMY FOR CAROTID ARTERY STENOSIS Carotid endarterectomy is a common surgical procedure, performed well over 100000 times each year, chiefly in North America.7 In the United Kingdom, carotid endarterectomy is performed 27 times less often per capita than in the United States, and seven times less frequently than in Canada.8 The procedure has been considered an option for patients who have experienced a stroke, TIA, or other symptom of presumed carotid origin as well as for asymptomatic patients with clearly identified anatomical carotid disease. Since the procedure carries with it a significant risk of mortality and morbidity, it should only be performed when its benefit exceeds its risk. Preoperative evaluation with arteriography is not itself without risk, adding to the complexity of decision-making for physicians. Recently, results from several large clinical trials have become available. These results address the key question of subgroup differences: Which types of patients will clearly benefit from the procedure?
233
234
FAMILY PRACTICE—AN INTERNATIONAL JOURNAL
The authors conclude that patients with a recent TIA or minor stroke without an obvious cardiac cause should be screened with non-invasive ultrasonographic techniques, and subsequently studied with arteriography if they have moderate to severe carotid artery narrowing. "Patients with high grade stenosis (7099%) should be considered for referral to institutions and surgeons who practice vigorous quality control and have the low rates of perioperative morbidity and mortality that have characterized the centers and physicians in this trial." However, the authors point out in their discussion that readers should not apply these conclusions too broadly. They mention several caveats. If serious surgical morbidity or mortality approaches 10% for a particular institution or surgical team, then the potential benefit of surgery is lost. A less obvious limitation is that this study provides no data regarding patients who suffered a TIA or stroke more than 120 days prior to surgery. Such patients were not selected for study and might be expected to benefit less from surgery because they have already survived the most dangerous period for repeat events. The study similarly provides no data about patients whose first stroke is a more devastating one, or who have major failure of other organs or embolus-generating heart disease.
Finally, it is important to adjust benefit of surgery downward to the extent that serious morbidity or mortality occurs as a result of cerebral arteriography in a particular institution. Comment The replication of findings in these two studies is very convincing evidence that there is a surgically modifiable, graded risk of stroke associated with increasing degrees of carotid stenosis in symptomatic patients. The investigators demonstrated a highly significant beneficial effect of surgery in the patients with the greatest degrees of stenosis. In the MRC European trial, the data suggest that performing 20 procedures "will cause one death or disabling stroke while preventing two." The limitations of the data bear review, however. Endarterectomy has only been shown to be of benefit if four conditions are met: Firstly, the patient must be symptomatic (recent minor stroke or TIA); secondly there must be 70-99% carotid stenosis on the side consistent with symptoms; thirdly the surgery must be performed quickly, within weeks of the symptomatic event; lastly the record of mortality and morbidity for the surgical team must be excellent. In addition, patients should be made aware that the long term benefits come at the expense of an increased short term surgical risk. The CASANOVA Study Group. Carotid surgery versus medical therapy in asymptomatic carotid stenosis. Stroke 1991; 22: 1229-1235. The CASANOVA Study (Carotid Artery Stenosis with Asymptomatic Narrowing: Operation Versus Aspirin), in contrast to the two papers discussed above, focused on patients without symptoms. It is the first completed randomized trial of carotid endarterectomy in asymptomatic patients. Between 1982 and 1988, the trial enrolled 410 patients with unilateral or bilateril carotid stenosis of 50-90%, based on Doppler sonography with angiographic confirmation. Patients were asymptomatic, based on physical examination and past medical history. These patients were randomized into two groups. In Group A, patients were operated on unilaterally or bilaterally for all 50-90% stenosis lesions. In Group B, patients with bilateral stenosis had an endarterectomy performed on the most severe side; patients with unilateral stenosis did not undergo surgery. Patients in both groups received aspirin 330 mg and dipyridamole 75 mg three times a day. In addition, during a 3-year follow-up period, surgery was performed for any patients who developed a stenosis > 90%, or who developed bilateral stenoses > 50%, or who had a TIA in a region associated with a > 50% stenosis, or who developed restenosis or new stenosis of > 50% if in Group A. The exclusion of patients with > 90% stenosis at entry, the provision for surgery during the follow-up period, and the provision for surgery in Group B patients with bilateral
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
the weight of evidence favouring surgery. The authors report data regarding 659 patients with carotid stenosis of 70-99%, randomized to surgery or medical therapy, and followed for an average of 18 months. Enrollment continues for patients with moderate stenosis. The sample was 70% male, and all had suffered either a TIA (69%) or stroke (31 %) within 120 days of enrollment. Medical therapy for all patients included aspirin, 1300 mg per day, if tolerated. Total mortality was nonsignificantly lower in the surgical group (4.6 versus 6.3%). Surgery resulted in a rate of death or disabling stroke within 30 days of 2.1%, versus 0.9% in the medical group over the same period. Surgery conferred an absolute risk reduction over 2 years of 17% (9 versus 26%) when all endpoints were combined (any stroke or death). Survival curves demonstrate that the disadvantages of surgical mortality and morbidity were overcome within 3 months by its benefits. The absolute increase in major stroke or death of 1.2% in the first 30 days was accompanied by an absolute risk reduction of 10.6% in the subsequent 2 years. Subgroup analysis indicated a graded effect of the degree of stenosis on absolute risk reduction for all ipsilateral stroke after surgery, ranging from 12% for stenosis of 70-89% to 26% for stenosis of 90-99%. In the medically treated group, risk could be stratified by a count of other risk factors, such as age greater than 70 years, male sex, elevated systolic (> 160 mm Hg) or diastolic ( > 90 mm Hg) blood pressure, evidence of carotid ulceration, etc. These variables could identify patients with a risk of ipsilateral stroke over 2 years of 17% in the lowest risk group and 39% in the highest risk group. The prognosis of surgical patients did not vary significantly among the same risk groups.
SELECTIONS FROM CURRENT LITERATURE
Comment It is not entirely clear how patients came to be in this study. Subjects were "recruited by the ultrasound laboratory in one of the participating centers," but we are not told why they were there in the first place. Presumably they WCTC being evaluated for cervical bruits, but this is not stated. They were certainly referred for Doppler ultrasound studies because of some concern on the part of the referring physician, and therefore may or may not be representative of all patients with asymptomatic carotid disease. This possible selection bias would seem to favour surgery, allowing for more confidence in the study's negative conclusions about surgical benefits. When the study began, the medical climate in the United States favoured carotid surgery, and the investigators had to deal with a number of patients who insisted on surgery after assignment to medical treatment. By the end of the study, 6 years later, "it became more and more difficult to convince patients, and their referring physicians, that endarterectomy still could be a useful procedure." Recruitment into the trial was stopped for this reason, even though originally a larger sample had been planned. Due to the size of the sample, the 95% confidence limits for the relative risk of endpoints for Group A versus Group B were 0.57 and 1.98. The authors correctly, but rather optimistically, interpret this to mean that they have not excluded up to a 43% reduction due to surgery. In presenting this apparent pro-surgery bias, they fail to point out that up to a 98% increase in risk due to surgery is just as likely. They then extrapolate that a
larger study (or a meta-analysis of several studies) could find that prophylactic effects of endarterectomy might be 25-33% after 5 years. This could indeed happen, but findings of a 50-66% increase in risk due to surgery is just as likely, given the data presented. It seems clear that there is no current evidence of a benefit from surgery in asymptomatic patients with carotid artery stenosis of 50-90%. Ongoing trials*"12 that have enrolled patients with stenosis of > 90% should provide important data about this potentially higher risk group, as well as expand the database for patients with lesser degrees of stenosis. As always, when one question seems answered, others arise. The available studies of non-rheumatic atrial fibrillation and symptomatic carotid artery stenosis demonstrate replication of findings to a great degree, which should make us more confident of the trends in the data. However, we should not extrapolate too far beyond a narrow interpretation. For instance, while warfarin was of benefit to those with atrial fibrillation as a group, what about those patients who are candidates for warfarin but who do not have any apparent risk for stasis-related thrombus formation? Would this group do just as well with aspirin, or with nothing? There were not enough such subjects to answer this question, although their stroke rate was low with or without active treatment. In symptomatic carotid stenosis, the potential benefits of surgery are clear for those with severe stenosis (70-99%), the potential risk of surgery is clear for those with mild stenosis (< 30%), but it is not clear what should be done with those who have moderate stenosis (3069%). It would also be premature to operate on most asymptomatic patients, given the CASANOVA study data.
REFERENCES SHEP Cooperative Research Group. Prevention of stroke by antihypertensive drug treatment in older persons with isolated systolic hypertension. Final results of the Systolic Hypertension in the Elderly Program (SHEP). J Am MedAssoc 1991; 265: 3255-3264. 2 Petersen P, Boysen G, Godtfredsen J, Andersen ED, AndeTsen B. Warfarin to prevent thromboembolism in chronic atrial fibrillation. Lancet 1989; i: 670. 3 Petersen P, Godtfredsen J, Boysen G, Andersen ED, Andersen B. Placebo-controlled, randomized trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: The Copenhagen AFASAK Study. Wancet 1989; I: 175-179. 4 The Stroke Prevention in Atrial Fibrillation Investigators. Design of a multicenter randomized trial for the Stroke Prevention in Atrial Fibrillation Study. Stroke 1990; 21: 538-545. 5 Lancaster TR, Singer DE, Sheehan MA, et al. The impact of long-term warfarin therapy on quality of life. Evidence from a randomized trial. Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators. Arch Intern Med 1991; 151: 1944-1949. 1
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
stenoses, were based on advice from the ethics committee in 1981 when the study began. Endpoints were 'minor stroke' (deficit lasting 24 hours to 4 weeks), 'stroke' (deficit lasting > 4 weeks), and death due to surgery or stroke from the time of randomization, including complications of angiography or surgery. All endpoints were analysed for the 3-year follow-up period, independent of the vascular territory or cerebral hemisphere involved. In this primary analysis, 10.7"% of patients in Group A and 11.3 "Vo of patients in Group B experienced endpoints, a non-significant difference. More detailed analysis showed that patients in Group A experienced more endpoints related to surgery and patients in Group B experienced more spontaneous events. The risk of death from surgery was 1.2% and the risk of major complications (including death, all strokes, pulmonary embolism, and myocardial infarction) was 6.9%. The results of four other studies of asymptomatic patients are yet to be reported.'-12 The CASANOVA investigators point out that their study would have been designed differently in the light of today's knowledge. They conclude that "carotid endarterectomy for asymptomatic patients with a stenosis of < 90% is not recommended at this time except as part of a controlled clinical trial."
235
236 6
7 8 9
FAMILY PRACTICE—AN INTERNATIONAL JOURNAL
Chesebro JH, Fuster V, Halperin JL. Atrial fibrillationrisk marker for stroke. New Engl J Med 1990; 323: 1556-1558. Dyken ML. Carotid endarterectomy studies: A glimmering of science. Stroke 1986; 17: 355-358. Barnett HJ. Symptomatic carotid endarterectomy trials. Stroke 1990; 21 (guppl HI): III-2-IH-5. The Asymptomatic Cervical Bruit Study Group. Natural history and effectiveness of aspirin in asymptomatic patient with cervical bruits. Arch Neurol 1991; 48: 683-686.
10
11
1J
Veterans Administration Cooperative Study: Role of carotid endarterectomy in asymptomatic carotid stenosis. Stroke 1986; 17: 534-539. Mayo Asymptomatic Carotid Endarterectomy Study Group. Effectiveness of carotid endarterectomy for asymptomatic carotid stenosis: Design of a clinical trial. Mayo Clin Proc 1989; 64: 897-904. Asymptomatic Carotid Atherosclerosis Study Group. Study design for randomized prospective trial of carotid endarterectomy for asymptomatic atherosclerosis. Stroke 1989; 20: 844-849.
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
Vol. 9, No. 2 Printed in Great Britain
Selections from Current Literature: Prevention of Stroke in Nonrheumatic Atrial Fibrillation and Carotid Artery Stenosis ROBERT KELLY
NON-RHEUMATIC ATRIAL FIBRILLATION It is well known that patients with atrial fibrillation and rheumatic valvular heart disease should receive long-term anticoagulant therapy. 'Non-rheumatic' atrial fibrillation is more common, is associated with a 5-fold higher incidence of ischaemic stroke compared to patients in normal sinus rhythm, and accounts for an increasing proportion of strokes as age increases. Results of earlier studies of anticoagulation treatment for this condition, particularly the Copenhagenbased Atrial Fibrillation, Aspirin, Anticoagulation (AFASAK) study,2-3 demonstrated a favourable trend but were not conclusive. Stroke Prevention in Atrial Fibrillation Investigators. Preliminary report of the Stroke Prevention in Atrial Fibrillation Study. N EnglJ Med 1990; 322: 863-868. The SPAF study was a randomized trial to test the effect of warfarin or aspirin on the incidence of ischaemic stroke and systemic thromboembolism in patients with atrial fibrillation unrelated to rheumatic valvular disease. Patients with constant or intermittent atrial fibrillation, and no echocardiographic evidence Department of Family Medicine, State University of New York at Stony Brook, Stony Brook, NY 11794, USA.
231
of rheumatic mitral stenosis were recruited at 15 clinical centres in the United States. Details of the protocol have been published.4 Patients with prior cerebral ischaemia, thromboembolism, severe mitral regurgitation, or dilated cardiomyopathy were excluded. Patients eligible to receive warfarin (Group 1) based on clinical estimation were randomized to receive open-label warfarin (dose adjusted to achieve 1.3-1.8 times control prothrombin time), 325 mg enteric-coated aspirin in a double-blind fashion, or matching placebo in a double-blind fashion. Those not eligible to receive warfarin (Group 2) were randomized to treatment with aspirin or placebo. Patients were monitored every 3 months for complications of therapy and for the development of stroke or embolism. The sample was 71 % male with a mean age of 67 and a prevalence of intermittent atrial fibrillation of 34%. Group 1 and 2 patients were quite similar demographically, though Group 2 patients were 3 years older on average. After 18 months, 1244 patients had been enrolled and followed for a mean of 1.13 years; 47% had been judged as eligible for warfarin therapy and assigned to Group 1. At that time, the Safety Monitoring Committee for the study terminated the placebo arm of Group 1, because of evidence of superior results with either active agent. The study will continue to enrol patients to allow comparison of warfarin and aspirin treatment in Group 1, and comparison of aspirin and placebo in Group 2. Data were analysed using an intention to treat approach based on initial group assignment. In Group 1, 25 patients experienced primary events; 24 had ischaemic strokes while only one had systemic embolism. Patients given active treatment with either warfarin or aspirin had an incidence of 1.6 events/100 person-years, while those receiving placebo experienced 8.3 events/100 person-years. This 81% risk reduction was highly significant statistically. The rate of transient ischaemic attack (TIA) was reduced by 50% in the active treatment group, but this difference was
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
In the most recent installment of Selections from Current Literature, the effect of hypertension treatment on stroke incidence was discussed as an important outcome of the Systolic Hypertension in the Elderly Program (SHEP).1 Stroke is often a devastating event, leaving the patient, their family, and their physician with a feeling of helplessness and a dread of what the future holds. None would argue that effective generic or condition-specific prevention of stroke is a worthwhile goal. For this article, therefore, I will review a number of recent papers that describe attempts to prevent stroke in patients with one of two common clinical syndromes: non-rheumatic atrial fibrillation and carotid artery stenosis.
232
FAMILY PRACTICE—AN INTERNATIONAL JOURNAL
The Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators. The effect of low-dose warfarin on the risk of stroke in patients with nonrheumatic atrial fibrillation. NEnglJMed 1990; 323: 1505-1511. This was an unblinded randomized trial of lowdose warfarin in patients with non-rheumatic atria] fibrillation. The target prothrombin time for those assigned to warfarin treatment was 1.2-1.5 times control value, a lower range than used in the SPAF study. Prothrombin times were measured every 3 weeks to allow for adjustment of dosage. Control group patients were allowed to take aspirin as they chose. A sample of 420 patients was randomized and followed for an average of 2.3 years. The enrollment, exclusion criteria, and demographic make-up of the sample were very similar to the SPAF study. Fewer patients had intermittent atrial fibrillation (17%), and a larger
number had no clinical heart disease other than atrial fibrillation (48%). The principal end-point was ischaemic stroke; record reviewers who determined end-point status were blinded to treatment assignment. Just as in the SPAF study, the trial was terminated earlier than planned because of evidence favouring treatment. A total of 15 patients had strokes, two in the warfarin group and 13 in the control group (0.4 versus 3.0 events/100 person-years). This represents an 86% risk reduction and is highly significant statistically. Overall mortality was also reduced by 62% in the warfarin group compared to controls (11 versus 26 deaths, P < 0.005). This difference was observed despite the fact that 46% of the control group patients were taking aspirin regularly. The authors did not find a beneficial effect of aspirin in control group patients. They report instead that eight of 13 control group patients who suffered strokes were taking aspirin (seven took ^ 25 mg/day) and that half of these patients were under 75 years of age. Three 'major' and 38 'minor' bleeding events occurred in the warfarin group, compared to two major and 21 minor events in the control group. Rates of hospital admission and transfusion were quite similar between the two groups. The authors conclude that "low-dose warfarin is effective in preventing stroke in patients with nonrheumatic atrial fibrillation and that it can be quite safe."
Comment It is noteworthy that these studies both showed strong beneficial effects of treatment, replicating the less conclusive findings of the earlier AFASAK study.2'3 It seems safe to conclude that those patients with nonrheumatic atrial fibrillation who are able to take warfarin will benefit from it. Outside of study protocols, in the real world of practice, there may not be the same degree of patient compliance or supervision of anticoagulation, leading to potentially higher complication rates. It is encouraging that relatively low doses of warfarin (target prothrombin time of 1.2-1.5 times control) can be used. The Boston Area study evaluated the effect of warfarin treatment on quality of life.3 The investigators report that "no significant differences between warfarin-treated and control patients were found on well-validated measures of functional status, well-being, and health perceptions", except in patients who had bleeding complications, and conclude that "negative effects of warfarin treatment on health perceptions may be balanced by confidence in its protective effects." In contrast to warfarin, the potential role of aspirin is less clear, given the mixed findings available at this time. It seems unlikely to be effective in those patients over 75 years of age, and the AFASAK study and the Boston Area study's secondary analyses raise doubts about its general effectiveness. We will need to wait for further data from the SPAF study to shed more light on this question. In an editorial accompanying the
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
not statistically significant due to the small numbers of patients experiencing TIAs. Total mortality was not significantly reduced (3.5% for placebo; 3.1% for active treatment). There were not enough patients experiencing primary events in the active treatment arms of Group 1 to permit detection of a statistically significant difference between warfarin and aspirin. Group 1 and Group 2 patients were combined to compare aspirin and placebo. In this analysis, 52 of 57 primary events were ischaemic strokes. Patients given aspirin had an incidence of 3.2 events/100 personyears, while those receiving placebo experienced 6.3 events/100 person-years, a statistically significant 49% risk reduction in primary endpoints. TIA incidence was reduced by 61% (P = 0.06) and total mortality was reduced by 19% (P = 0.11). Subgroup analysis revealed that aspirin was not apparently effective in the 238 patients over 75 years of age, and that the 51 patients without cardiovascular disease other than atrial fibrillation experienced no primary endpoints. In patients receiving warfarin, haemorrhagic events requiring hospitalization, transfusion, or surgery occurred with an incidence of 1.7/100 patient-years; intracranial haemorrhage, including cerebral haemorrhage and subdural hematoma, occurred at a rate of 0.9/100 person-years. These rates can be compared to rates of 0.9% and 0.2% for aspirin and 1.2% and 0.2% for placebo. The authors conclude that aspirin and warfarin are separately "effective in preventing ischaemic stroke and systemic embolism and comparatively safe for patients with atrial fibrillation unrelated to rheumatic valvular disease who are acceptable candidates for anticoagulant therapy." They point out that the absolute risk reduction of 6.8%/year in Group 1 exceeded the incremental risk of clinically important haemorrhage of < 1%/year. They note the limitations of their relatively short duration of follow-up and the absence of demonstrated benefit for aspirin in patients over 75 years of age.
SELECTIONS FROM CURRENT LITERATURE 6
Boston Area study, Chesebro et al. point out that atrial fibrillation may not itself be a risk factor for stroke, but rather a "risk marker for other associated cardiovascular disease." They feel that patients with stasis-related thrombi will benefit most from anticoagulation, whereas platelet inhibition with aspirin may most benefit those with cerebrovascular atherosclerosis. They further propose out that the somewhat variable results of existing trials relate to the relative frequencies of stasis and atherosclerosis as causes of stroke, and the makeup of the samples in each study. The logical extension of this reasoning is that clinicians should choose warfarin or aspirin based on their suspicions or knowledge of patients' risk for stasis and atherosclerosis.
European Carotid Surgery Trialists' Collaborative Group. MRC European Carotid Surgery Trial: Interim results for symptomatic patients with severe (70-99%) or with mild (0-29%) carotid stenosis. Lancet 1991; 337: 1235-1243. This paper reports data from a clinical trial of carotid endarterectomy in patients with a stenotic carotid artery lesion, who had experienced a non-disabling, ipsilateral carotid territory ischaemic stroke, TIA, or retinal infarct. By angiography, patients were characterized as having mild (< 30%), moderate (30-69%), or severe (70-99%) carotid artery stenosis preoperatively. Patients whose physicians were 'substantially uncertain' whether or not to recommend carotid endarterectomy were eligible to be randomized to recommendation for 'immediate surgery' (60%) or 'no immediate surgery' (40%). All patients received the best available medical care in other respects, usually including management of hypertension, aspirin therapy, and advice to stop smoking.
Over a ten-year period, 2518 patients were randomized at 80 centres in 14 countries. The paper reports data regarding 1152 with mild or severe stenosis who were followed for an average of 3 years after surgery. Data collection for patients with moderate stenosis will continue and has not yet been reported. The principal endpoints were fatal or disabling stroke, surgeryassociated deaths (within 30 days of surgery), and nondisabling strokes lasting more than 7 days. The sample was predominantly male (70%) with a mean age of 60 years. Most had experienced a TIA in the past (78%) and half had experienced a prior stroke. For patients with severe stenosis, total mortality was lower in those patients who were allocated to surgery (9.9 versus 12.7%); this difference was not statistically significant. Surgery resulted in a rate of death or disabling stroke within 30 days of 3.7%. Those with high blood pressure (systolic > 160 mm Hg) or rapid surgery Casting less than one hour) were most likely to have an adverse outcome at 30 days. The rest of the patients assigned to immediate surgery, who had an 'apparently successful' procedure, had an eight-fold reduction in ipsilateral fatal or disabling ischaemic strokes (1.1 versus 8.4%) and a six-fold reduction in non-disabling strokes compared to the group who were not operated on. Surgery conferred an absolute risk reduction over 3 years of 9.6% (12.3 versus 21.9%) when all endpoints were combined (any stroke lasting more than 7 days; surgical death). The data also show that most of the benefit of surgery occurred in the first year for death and disabling stroke, and in the first 2 years for non-disabling stroke. For patients with mild stenosis, total mortality was actually higher in those allocated to surgery (11.4 versus 7.7%) but this difference was not statistically significant. The immediate consequences of surgery within 30 days was a rate of death or disabling stroke of 2.3 %. There was no benefit of surgery demonstrated on subsequent rates of death, disabling stroke, or nondisabling stroke, even after 3 years of follow-up. The authors conclude that carotid endarterectomy is associated with serious risks and that "successful surgery avoids the majority of the excess risk of ipsilateral ischaemic stroke associated with severe (70-99%) carotid artery stenosis. Clearly, the lower the institutional surgical risk, the more this risk will be worth taking." North American Symptomatic Carotid Endarterectomy Trial Collaborators. Beneficial effect of carotid endarterectomy in symptomatic patients with high-grade carotid stenosis. N EnglJ Med 1991; 325: 445-453. This ongoing trial in the United States and Canada took a very similar approach to the question of carotid endarterectomy in symptomatic patients with carotid stenosis of 30-99%. Entry criteria, sample demographics, and outcome measures were quite comparable to the European trial. For the group of patients with severe stenosis, the trial has been stopped because of
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
ENDARTERECTOMY FOR CAROTID ARTERY STENOSIS Carotid endarterectomy is a common surgical procedure, performed well over 100000 times each year, chiefly in North America.7 In the United Kingdom, carotid endarterectomy is performed 27 times less often per capita than in the United States, and seven times less frequently than in Canada.8 The procedure has been considered an option for patients who have experienced a stroke, TIA, or other symptom of presumed carotid origin as well as for asymptomatic patients with clearly identified anatomical carotid disease. Since the procedure carries with it a significant risk of mortality and morbidity, it should only be performed when its benefit exceeds its risk. Preoperative evaluation with arteriography is not itself without risk, adding to the complexity of decision-making for physicians. Recently, results from several large clinical trials have become available. These results address the key question of subgroup differences: Which types of patients will clearly benefit from the procedure?
233
234
FAMILY PRACTICE—AN INTERNATIONAL JOURNAL
The authors conclude that patients with a recent TIA or minor stroke without an obvious cardiac cause should be screened with non-invasive ultrasonographic techniques, and subsequently studied with arteriography if they have moderate to severe carotid artery narrowing. "Patients with high grade stenosis (7099%) should be considered for referral to institutions and surgeons who practice vigorous quality control and have the low rates of perioperative morbidity and mortality that have characterized the centers and physicians in this trial." However, the authors point out in their discussion that readers should not apply these conclusions too broadly. They mention several caveats. If serious surgical morbidity or mortality approaches 10% for a particular institution or surgical team, then the potential benefit of surgery is lost. A less obvious limitation is that this study provides no data regarding patients who suffered a TIA or stroke more than 120 days prior to surgery. Such patients were not selected for study and might be expected to benefit less from surgery because they have already survived the most dangerous period for repeat events. The study similarly provides no data about patients whose first stroke is a more devastating one, or who have major failure of other organs or embolus-generating heart disease.
Finally, it is important to adjust benefit of surgery downward to the extent that serious morbidity or mortality occurs as a result of cerebral arteriography in a particular institution. Comment The replication of findings in these two studies is very convincing evidence that there is a surgically modifiable, graded risk of stroke associated with increasing degrees of carotid stenosis in symptomatic patients. The investigators demonstrated a highly significant beneficial effect of surgery in the patients with the greatest degrees of stenosis. In the MRC European trial, the data suggest that performing 20 procedures "will cause one death or disabling stroke while preventing two." The limitations of the data bear review, however. Endarterectomy has only been shown to be of benefit if four conditions are met: Firstly, the patient must be symptomatic (recent minor stroke or TIA); secondly there must be 70-99% carotid stenosis on the side consistent with symptoms; thirdly the surgery must be performed quickly, within weeks of the symptomatic event; lastly the record of mortality and morbidity for the surgical team must be excellent. In addition, patients should be made aware that the long term benefits come at the expense of an increased short term surgical risk. The CASANOVA Study Group. Carotid surgery versus medical therapy in asymptomatic carotid stenosis. Stroke 1991; 22: 1229-1235. The CASANOVA Study (Carotid Artery Stenosis with Asymptomatic Narrowing: Operation Versus Aspirin), in contrast to the two papers discussed above, focused on patients without symptoms. It is the first completed randomized trial of carotid endarterectomy in asymptomatic patients. Between 1982 and 1988, the trial enrolled 410 patients with unilateral or bilateril carotid stenosis of 50-90%, based on Doppler sonography with angiographic confirmation. Patients were asymptomatic, based on physical examination and past medical history. These patients were randomized into two groups. In Group A, patients were operated on unilaterally or bilaterally for all 50-90% stenosis lesions. In Group B, patients with bilateral stenosis had an endarterectomy performed on the most severe side; patients with unilateral stenosis did not undergo surgery. Patients in both groups received aspirin 330 mg and dipyridamole 75 mg three times a day. In addition, during a 3-year follow-up period, surgery was performed for any patients who developed a stenosis > 90%, or who developed bilateral stenoses > 50%, or who had a TIA in a region associated with a > 50% stenosis, or who developed restenosis or new stenosis of > 50% if in Group A. The exclusion of patients with > 90% stenosis at entry, the provision for surgery during the follow-up period, and the provision for surgery in Group B patients with bilateral
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
the weight of evidence favouring surgery. The authors report data regarding 659 patients with carotid stenosis of 70-99%, randomized to surgery or medical therapy, and followed for an average of 18 months. Enrollment continues for patients with moderate stenosis. The sample was 70% male, and all had suffered either a TIA (69%) or stroke (31 %) within 120 days of enrollment. Medical therapy for all patients included aspirin, 1300 mg per day, if tolerated. Total mortality was nonsignificantly lower in the surgical group (4.6 versus 6.3%). Surgery resulted in a rate of death or disabling stroke within 30 days of 2.1%, versus 0.9% in the medical group over the same period. Surgery conferred an absolute risk reduction over 2 years of 17% (9 versus 26%) when all endpoints were combined (any stroke or death). Survival curves demonstrate that the disadvantages of surgical mortality and morbidity were overcome within 3 months by its benefits. The absolute increase in major stroke or death of 1.2% in the first 30 days was accompanied by an absolute risk reduction of 10.6% in the subsequent 2 years. Subgroup analysis indicated a graded effect of the degree of stenosis on absolute risk reduction for all ipsilateral stroke after surgery, ranging from 12% for stenosis of 70-89% to 26% for stenosis of 90-99%. In the medically treated group, risk could be stratified by a count of other risk factors, such as age greater than 70 years, male sex, elevated systolic (> 160 mm Hg) or diastolic ( > 90 mm Hg) blood pressure, evidence of carotid ulceration, etc. These variables could identify patients with a risk of ipsilateral stroke over 2 years of 17% in the lowest risk group and 39% in the highest risk group. The prognosis of surgical patients did not vary significantly among the same risk groups.
SELECTIONS FROM CURRENT LITERATURE
Comment It is not entirely clear how patients came to be in this study. Subjects were "recruited by the ultrasound laboratory in one of the participating centers," but we are not told why they were there in the first place. Presumably they WCTC being evaluated for cervical bruits, but this is not stated. They were certainly referred for Doppler ultrasound studies because of some concern on the part of the referring physician, and therefore may or may not be representative of all patients with asymptomatic carotid disease. This possible selection bias would seem to favour surgery, allowing for more confidence in the study's negative conclusions about surgical benefits. When the study began, the medical climate in the United States favoured carotid surgery, and the investigators had to deal with a number of patients who insisted on surgery after assignment to medical treatment. By the end of the study, 6 years later, "it became more and more difficult to convince patients, and their referring physicians, that endarterectomy still could be a useful procedure." Recruitment into the trial was stopped for this reason, even though originally a larger sample had been planned. Due to the size of the sample, the 95% confidence limits for the relative risk of endpoints for Group A versus Group B were 0.57 and 1.98. The authors correctly, but rather optimistically, interpret this to mean that they have not excluded up to a 43% reduction due to surgery. In presenting this apparent pro-surgery bias, they fail to point out that up to a 98% increase in risk due to surgery is just as likely. They then extrapolate that a
larger study (or a meta-analysis of several studies) could find that prophylactic effects of endarterectomy might be 25-33% after 5 years. This could indeed happen, but findings of a 50-66% increase in risk due to surgery is just as likely, given the data presented. It seems clear that there is no current evidence of a benefit from surgery in asymptomatic patients with carotid artery stenosis of 50-90%. Ongoing trials*"12 that have enrolled patients with stenosis of > 90% should provide important data about this potentially higher risk group, as well as expand the database for patients with lesser degrees of stenosis. As always, when one question seems answered, others arise. The available studies of non-rheumatic atrial fibrillation and symptomatic carotid artery stenosis demonstrate replication of findings to a great degree, which should make us more confident of the trends in the data. However, we should not extrapolate too far beyond a narrow interpretation. For instance, while warfarin was of benefit to those with atrial fibrillation as a group, what about those patients who are candidates for warfarin but who do not have any apparent risk for stasis-related thrombus formation? Would this group do just as well with aspirin, or with nothing? There were not enough such subjects to answer this question, although their stroke rate was low with or without active treatment. In symptomatic carotid stenosis, the potential benefits of surgery are clear for those with severe stenosis (70-99%), the potential risk of surgery is clear for those with mild stenosis (< 30%), but it is not clear what should be done with those who have moderate stenosis (3069%). It would also be premature to operate on most asymptomatic patients, given the CASANOVA study data.
REFERENCES SHEP Cooperative Research Group. Prevention of stroke by antihypertensive drug treatment in older persons with isolated systolic hypertension. Final results of the Systolic Hypertension in the Elderly Program (SHEP). J Am MedAssoc 1991; 265: 3255-3264. 2 Petersen P, Boysen G, Godtfredsen J, Andersen ED, AndeTsen B. Warfarin to prevent thromboembolism in chronic atrial fibrillation. Lancet 1989; i: 670. 3 Petersen P, Godtfredsen J, Boysen G, Andersen ED, Andersen B. Placebo-controlled, randomized trial of warfarin and aspirin for prevention of thromboembolic complications in chronic atrial fibrillation: The Copenhagen AFASAK Study. Wancet 1989; I: 175-179. 4 The Stroke Prevention in Atrial Fibrillation Investigators. Design of a multicenter randomized trial for the Stroke Prevention in Atrial Fibrillation Study. Stroke 1990; 21: 538-545. 5 Lancaster TR, Singer DE, Sheehan MA, et al. The impact of long-term warfarin therapy on quality of life. Evidence from a randomized trial. Boston Area Anticoagulation Trial for Atrial Fibrillation Investigators. Arch Intern Med 1991; 151: 1944-1949. 1
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
stenoses, were based on advice from the ethics committee in 1981 when the study began. Endpoints were 'minor stroke' (deficit lasting 24 hours to 4 weeks), 'stroke' (deficit lasting > 4 weeks), and death due to surgery or stroke from the time of randomization, including complications of angiography or surgery. All endpoints were analysed for the 3-year follow-up period, independent of the vascular territory or cerebral hemisphere involved. In this primary analysis, 10.7"% of patients in Group A and 11.3 "Vo of patients in Group B experienced endpoints, a non-significant difference. More detailed analysis showed that patients in Group A experienced more endpoints related to surgery and patients in Group B experienced more spontaneous events. The risk of death from surgery was 1.2% and the risk of major complications (including death, all strokes, pulmonary embolism, and myocardial infarction) was 6.9%. The results of four other studies of asymptomatic patients are yet to be reported.'-12 The CASANOVA investigators point out that their study would have been designed differently in the light of today's knowledge. They conclude that "carotid endarterectomy for asymptomatic patients with a stenosis of < 90% is not recommended at this time except as part of a controlled clinical trial."
235
236 6
7 8 9
FAMILY PRACTICE—AN INTERNATIONAL JOURNAL
Chesebro JH, Fuster V, Halperin JL. Atrial fibrillationrisk marker for stroke. New Engl J Med 1990; 323: 1556-1558. Dyken ML. Carotid endarterectomy studies: A glimmering of science. Stroke 1986; 17: 355-358. Barnett HJ. Symptomatic carotid endarterectomy trials. Stroke 1990; 21 (guppl HI): III-2-IH-5. The Asymptomatic Cervical Bruit Study Group. Natural history and effectiveness of aspirin in asymptomatic patient with cervical bruits. Arch Neurol 1991; 48: 683-686.
10
11
1J
Veterans Administration Cooperative Study: Role of carotid endarterectomy in asymptomatic carotid stenosis. Stroke 1986; 17: 534-539. Mayo Asymptomatic Carotid Endarterectomy Study Group. Effectiveness of carotid endarterectomy for asymptomatic carotid stenosis: Design of a clinical trial. Mayo Clin Proc 1989; 64: 897-904. Asymptomatic Carotid Atherosclerosis Study Group. Study design for randomized prospective trial of carotid endarterectomy for asymptomatic atherosclerosis. Stroke 1989; 20: 844-849.
Downloaded from http://fampra.oxfordjournals.org/ at Karolinska Institutet on July 23, 2015
