Review
Secondary stroke prevention Graeme J Hankey
Survivors of stroke and transient ischaemic attacks are at risk of a recurrent stroke, which is often more severe and disabling than the index event. Optimum secondary prevention of recurrent stroke needs rapid diagnosis and treatment and prompt identification of the underlying cardiovascular cause. Effective treatments include organised acute assessment and intervention with antithrombotic therapy, carotid revascularisation, and control of causal risk factors, as appropriate. However, effective treatments are not implemented optimally in clinical practice. Recurrent strokes continue to account for 25–30% of all strokes and represent unsuccessful secondary prevention. Immediate and sustained implementation of effective and appropriate secondary prevention strategies in patients with first-ever stroke or transient ischaemic attack has the potential to reduce the burden of stroke by up to a quarter.
Introduction The world is facing a stroke epidemic: between 1990 and 2010, the number of stroke-related deaths increased by 26% (95% CI 14–32) and disability-adjusted life-years by 19% (5–26%), to make stroke the second leading cause of death and third leading contributor to disabilityadjusted life-years in the world.1,2 Complementary strategies to reduce the increasing global burden of stroke include the so-called population/mass and highrisk approaches to prevent first-ever and recurrent stroke, and the treatment and rehabilitation of acute stroke to improve disability-free survival.3 The most relevant of these approaches for stroke physicians are the treatment and rehabilitation of acute stroke and the secondary prevention of recurrent stroke. This Review summarises evidence for optimum secondary prevention of recurrent ischaemic stroke. The rationale for this report is that recurrent strokes constitute a notable proportion (25–30%) of all preventable strokes;4 they are frequently ischaemic and more disabling, fatal, and costly than the first stroke;4,5 and their rates could be reduced by improved translation of evidence-based advances in secondary stroke prevention into clinical practice.6 Since the rate of recurrent stroke and the potential benefits and risks of secondary prevention strategies are higher earlier rather than later after ischaemic stroke, I will describe separately the prognosis and predictors for early and long-term recurrent ischaemic stroke and the strategies to prevent both types of stroke.
Early recurrent stroke Prognosis and predictors The risk of a recurrent stroke is highest early after an ischaemic stroke or transient ischaemic attack (TIA)— about 1% at 6 h, 2% at 12 h, 3% at 2 days, 5% at 7 days, and 10% at 14 days.7–9 Therefore, ischaemic stroke or TIA is a medical emergency that demands immediate diagnosis and treatment. Some clinical features, such as sudden-onset unilateral weakness and speech disturbance lasting for longer than 10 min, predict a high risk of stroke soon after TIA, probably because they distinguish TIA from a myriad of TIA mimics that have a more benign
prognosis.10 These, and other, clinical features have been used to derive a prognostic index: the ABCD2 score (panel 1; appendix pp 4–5).10–12 Additional prognostic information from carotid and brain imaging about the cause of the ischaemic event has been incorporated into the ABCD3–I score19 (panel 1; appendix pp 4–5) and the recurrence risk estimator.14,24 However, low scores on these prognostic indices should not preclude appropriate assessment and management because some patients with a low predicted risk of stroke have treatable causes, such as carotid stenosis and atrial fibrillation, that might warrant urgent intervention.17,18
Lancet Neurol 2013 Published Online December 20, 2013 http://dx.doi.org/10.1016/ S1474-4422(13)70255-2 School of Medicine and Pharmacology, The University of Western Australia, Perth, Australia; and Department of Neurology, Sir Charles Gairdner Hospital, Nedlands, Perth, Australia (Prof G J Hankey MD) Correspondence to: Prof Graeme J Hankey, School of Medicine and Pharmacology, The University of Western Australia, Room 222, Harry Perkins Institute of Medical Research, QQ Block, QEII Medical Centre, 6 Verdun Street, Nedlands, Perth, 6009, Australia [email protected] See Online for appendix For the recurrence risk estimator see http://www.nmr. mgh.harvard.edu/RRE/
Panel 1: The ABCD2 and ABCD3–I scores to predict early recurrent stroke ABCD2 The ABCD2 score (appendix pp 4–5) is a simple scoring system that can be used in primary care and by emergency department clinicians to stratify patients with transient ischaemic attack (TIA) into low, moderate, and high risk of early recurrent stroke.10 The score has been validated in several independent datasets to have good power to predict the rate and severity of recurrent stroke at 7 days.11,12 A limitation of the ABCD2 score is that it was not derived from datasets with information about the cardiovascular lesion(s) causing the ischaemic stroke or TIA, such as carotid stenosis and atrial fibrillation, which also determine early prognosis. Most early recurrent strokes are caused by recurrent thromboembolism from an unstable atherosclerotic plaque that is still unstable (vs recurrent cardiogenic embolism or small vessel occlusion).13–16 Another limitation of the score is that 10% of strokes at 7 days after TIA occur in patients judged to be at low risk (ABCD2 score 0–3), and 50% of strokes occur in patients classed as moderate risk (ABCD2 score 4–5).17,18 Hence, patients with TIA should be assessed immediately, irrespective of ABCD2 score, because some with lower scores have treatable causes, such as carotid stenosis and atrial fibrillation, that might be associated with raised short-term risks of stroke. ABCD3–I The ABCD3–I score has been developed for use in secondary care settings that have access to carotid and brain imaging (appendix pp 4–5).19 It uses the prognostic significance of the cardiovascular lesion causing the ischaemic stroke or TIA, and the presence of recent focal brain ischaemia detected by magnetic resonance diffusion-weighted imaging.20,21 It has been validated externally as having a higher predictive value than the ABCD2 score for assessing risk of early stroke.19,22 Limitations of the ABCD3–I score are that the yield of magnetic resonance diffusion-weighted imaging varies according to the timing of the scan after the TIA; if the procedure is undertaken more than 24 h after symptom onset, the odds of finding a diffusion-weighted imaging lesion could be increased by up to threefold.23
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
1
Review
Secondary prevention Effective strategies to prevent early recurrent ischaemic stroke include assessment and management in an acute specialist unit, immediate antiplatelet therapy, and early carotid revascularisation, as appropriate (table 1, panel 2).
Antiplatelet and anticoagulation therapies As a single antiplatelet therapy, aspirin 160–300 mg daily, started within 48 h of onset of ischaemic stroke in 40 000 patients, and continued for 2–4 weeks, reduced the odds of recurrent ischaemic stroke by 23% (2·4% aspirin vs 3·1% control; odds ratio [OR] 0·77, 95% CI 0·69–0·87), increased the odds of symptomatic intracranial haemorrhage by 22% (1·0% vs 0·8%; OR 1·22, 95% CI 1·00–1·50), and reduced the odds of any recurrent stroke by 12% (3·4% vs 3·9%; OR 0·88, 95% CI 0·79–0·97) compared with control (table 1).25 A meta-analysis of 14 randomised controlled trials comparing dual antiplatelet therapy with antiplatelet Outcome
monotherapy, both started within 3 days of onset of ischaemic stroke or TIA, in 9102 adults showed that dual antiplatelet therapy significantly reduced early recurrent stroke by a third (6·21% dual therapy vs 8·95% monotherapy; risk ratio [RR] 0·69, 95% CI 0·60–0·80, absolute risk reduction 2·74%) at about 3 months’ followup (range 7 days–18 months), compared with monotherapy, and non-significantly increased the risk of major bleeding (0·52% dual therapy vs 0·36% monotherapy; RR 1·35, 95% CI 0·70–2·59, absolute risk increase 0·16%).26 The results of trials comparing different dual combinations of antiplatelet drugs with different single antiplatelet drugs were consistent with the overall result for all trials (table 1).26 However, the results might not apply to patients with major ischaemic stroke (eg, those with a score on the National Institutes of Stroke Scale >4) who were excluded from many trials because of the risk of haemorrhagic transformation of new brain infarction.51
Stroke rate
RRR (95% CI)
Control
Intervention
ARR
NNT and time period
Early recurrent stroke Aspirin (vs no aspirin)25
Stroke at 2–4 weeks
Clopidogrel plus aspirin (vs aspirin)9,26
Stroke at roughly 3 months
3·9%
3·4%
12% (3 to 21)
0·5%
200 over 2–4 weeks
11·1%
7·8%
30% (18 to 41)
3·3%
Aspirin plus dipyridamole (vs aspirin)26
30 over roughly 3 months
Stroke at 3–28 months
7·9%
5·3%
36% (–10 to 63)
2·6%
38 over 3–28 months
Aspirin plus dipyridamole (vs clopidogrel)26
Stroke at 3 months
2·9%
1·6%
44% (–17 to 73)
1·8%
56 over 3 months
Clopidogrel plus aspirin (vs clopidogrel)26
Stroke at 18 months
4·7%
3·9%
17% (–93 to 64)
0·8%
125 over 18 months
Carotid endarterectomy or stent (vs no carotid revascularisation)27,28 70–99% stenosis, symptomatic
Stroke at 5 years
33%
17%
48% (38 to 60)
16%
6 over 5 years
50–69% stenosis, symptomatic
Stroke at 5 years
27%
19%
28% (14 to 42)
8%
12 over 5 years
Stroke at 90 days
10·3%
2·1%
80% (51 to 92)
8·2%
12 over 90 days
7·0%
6·1%
13% (6 to 19)
0·9%
111 over 1 year
All high-vascular-risk patients
5·8%
5·3%
9% (0·3 to 16)
0·5%
200 over 1 year
Ischaemic stroke patients
7·7%
7·1%
7% (–6 to 19)
0·6%
167 over 1 year
Recurrent stroke, per 2·6 years (mean)
11·3%
9·0%
22% (10 to 32)
2·3%
43 over 2·6 years, or 113 over 1 year
Stroke, MI, or VD, per 2·6 years (mean)
15·2%
12·5%
18% (8 to 28)
2·7%
37 over 2·6 years, or 100 over 1 year
Acute specialty units (vs outpatient clinics)29 Longer-term recurrent stroke Aspirin (vs no aspirin)30,31
Stroke, MI, or VD, per year
Clopidogrel (vs aspirin) 32,33
Stroke, MI, or VD, per year
Aspirin and extended-release dipyridamole (vs aspirin)34
Aspirin and extended-release dipyridamole (vs clopidogrel)35
Recurrent stroke, per 2·5 years (mean)
8·8%
9·0%
−1% (–11 to 8)
Stroke, MI, or VD per 2·5 years (mean)
13·1%
13·1%
1% (–7 to 8)
8·8%
6·3%
Cilostazol (vs aspirin)36
Stroke, MI, or VD, per 1·25 years (mean)
Warfarin for atrial fibrillation (vs no warfarin)37,38
Recurrent stroke, per year
New direct oral anticoagulants for atrial fibrillation (vs warfarin)39
Stroke and systemic embolism, per 1·9 years
Lowering of blood pressure (vs no lowering of blood pressure)40,41 Recurrent stroke, per 3 years (roughly) Statins to reduce low-density-lipoprotein cholesterol (vs no statin)42
Recurrent stroke, per 5 years
12%
0·2% (NS) NA 0% (NS)
NA
28% (11 to 43)
2·5%
40 over 1·25 years, or 50 over 1 year
4%
61% (37 to 75)
8%
12 over 1 year
5·3%
4·5%
14% (0 to 26)
0·8%
134 over 1·9 years, or 255 over 1 year
9·9%
8·6%
22% (10 to 32)
1·3%
75 over roughly 3 years, or 225 over roughly 1 year
11·9%
10·5%
12% (1 to 22)
1·4%
74 over 5 years, or 370 over roughly 1 year
See appendix pp 1–3 for explanations of statistical terms. RRR=relative risk reduction (risk ratio). ARR=absolute risk reduction. NNT=number needed to treat. NS=not significant. MI=myocardial infarction. VD=vascular death. NA=not applicable.
Table 1: Summary of effective strategies to prevent recurrent stroke
2
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
Panel 2: Recommendations for early secondary prevention Acute specialty units • Patients with suspected transient ischaemic attack (TIA) should be assessed and managed urgently in an acute specialty unit, such as a dedicated TIA clinic.29,43 Investigations usually comprise immediate imaging of the brain, arteries to the brain, and heart; electrocardiograph; and measurement of fasting blood glucose and lipids. Interventions usually include antiplatelet therapy, blood pressure-lowering, statins, and anticoagulation and carotid endarterectomy or stenting, as appropriate (evidence level B). Antiplatelet therapy • All patients with acute TIA and ischaemic stroke should be given at least 160 mg of aspirin or acetylsalicylic acid immediately as a single loading dose (evidence level A).25 • In dysphagic patients, aspirin can be given by enteral tube or by rectal suppository. • Chinese patients with acute (160 mg) was used on day 1 to rapidly inhibit platelet aggregation—initiation with a daily dose of 75 mg
• Patients with atrial fibrillation and acute transient ischaemic attack can begin oral anticoagulation (warfarin, dabigatran, rivaroxaban, or apixaban) immediately because the risk of intracranial haemorrhage is probably low (ie, there is no fresh brain infarction to become haemorrhagic). Carotid revascularisation • Patients with recently symptomatic carotid territory transient ischaemic attack or non-disabling ischaemic stroke and ipsilateral 50–99% internal carotid artery stenosis (measured by two concordant non-invasive imaging modalities) who are fit and willing for surgery should be offered carotid endarterectomy as soon as possible, ideally within the first few days and up to 1 week after the ischaemic event and when the patient is clinically stable (evidence level A).15,27 • Carotid endarterectomy should be done by a surgeon with an audited perioperative morbidity and mortality of less than 5% (evidence level A).27 • Carotid endarterectomy is more appropriate than carotid stenting for patients older than 70 years of age who are otherwise fit for surgery because stenting carries a higher short-term risk of stroke and death (evidence level A).27,50 Carotid stenting might be as safe as endarterectomy in patients less than 70 years of age.28,50 • Carotid stenting could also be considered for patients who are not candidates for carotid endarterectomy because of technical, anatomical, or medical reasons (evidence level A).28 Interventionalists should have expertise in carotid procedures and an expected risk of periprocedural morbidity and mortality of less than 5%. Levels of evidence • Level A=evidence from meta-analyses of randomised controlled trials or directly from randomised controlled trials. Desirable effects clearly outweigh undesirable effects, or vice versa. • Level B=single randomised controlled trial or well-designed observational study with strong evidence; or well-designed cohort or case–control analytical study; or several time series or dramatic results of an uncontrolled experiment. Desirable effects closely balanced with undesirable effects. • Level C=at least one well-designed, non-experimental descriptive study (eg, comparative studies, correlation studies, case studies) or expert committee reports, opinions and/or experience of respected authorities, including consensus from development and/or reviewer groups.
of clopidogrel and aspirin takes several days to produce maximum inhibition of platelet aggregation.9,45 Most of the benefit of clopidogrel plus aspirin was seen within the first few days after TIA and ischaemic stroke, when the risk of recurrence was highest.9 The safety and efficacy of dual and triple antiplatelet therapy in non-Chinese populations with acute TIA and mild ischaemic stroke is
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
3
Review
being tested in two large phase 3 trials.52,53 Newer and more potent antiplatelet drugs than aspirin and clopidogrel, such as prasugrel and ticagrelor, that have shown efficacy in atherothrombotic acute coronary syndromes are being considered for clinical trials in acute atherothromboembolic TIA and mild ischaemic stroke trials.54 In 23 748 patients with acute ischaemic stroke of presumed arterial origin, anticoagulation started within 48 h of stroke reduced early recurrent ischaemic stroke (OR 0·76, 95% CI 0·65–0·88) but increased symptomatic intracranial haemorrhage (2·55, 1·95–3·33), and achieved no net benefit in reducing any recurrent stroke (0·97, 0·85–1·11).46 The results were consistent for all types of heparin (intravenous and subcutaneous unfractionated heparin, heparinoids, and low-molecularweight heparin) and in patients at raised risk of thrombotic events and reduced risk of haemorrhagic events, which suggests that routine and selective use of heparin in ischaemic stroke has no net benefit.46,47 The new direct oral anticoagulants, one of which (rivaroxaban) has shown efficacy in acute coronary syndromes, could be assessed as potential treatments for acute TIA and mild ischaemic stroke of presumed arterial origin.55 In 4624 patients with acute ischaemic stroke of presumed cardiac origin, anticoagulation started within 48 h of cardioembolic ischaemic stroke did not significantly reduce early recurrent ischaemic stroke (OR 0·7, 95% CI 0·4–1·1) but increased symptomatic intracranial haemorrhage (2·9, 1·2–7·0), with no net effect on any recurrent stroke (1·18, 0·74–1·88).48 Whether or not early anticoagulation is safe and effective in selected patients at high risk of early recurrent cardioembolic stroke and low risk of haemorrhagic transformation of recently infarcted brain remains uncertain.
Revascularisation The perioperative risk of stroke or death associated with carotid endarterectomy for symptomatic carotid stenosis is about 7% (95% CI 6·2–8·0).27 Perioperative risks are increased in women; in patients with hemispheric events, contralateral carotid occlusion, irregular or ulcerated plaque, diabetes, chronic obstructive pulmonary disease, ischaemic heart disease, heart failure, peripheral arterial disease, and renal impairment warranting dialysis; and when carotid endarterectomy is undertaken within 48 h after ischaemic stroke.27,56,57 For long-term outcomes, the addition of carotid endarterectomy to best medical treatment reduces the risk of stroke or death at 5 years by half in patients with recently symptomatic 70–99% carotid stenosis, and by a quarter in those with 50–69% stenosis (table 1).27 The benefits of the procedure are greatest in men, older patients, those with recent (rather than later) ischaemic events of the brain (vs the eye), those with increasing 4
carotid stenosis, and in patients with ulcerated (vs smooth) carotid plaque.15 A systematic review compared carotid endarterectomy with carotid artery stenting in 7572 patients with recent symptomatic carotid stenosis enrolled in 16 trials.28 Compared with carotid endarterectomy, carotid artery stenting was associated with a lower risk of periprocedural access site haematoma (0·9% carotid artery stenting vs 2·7% carotid endarterectomy; OR 0·37, 95% CI 0·18–0·77, I²=27%), cranial nerve injury (0·3% vs 5·5%; OR 0·08, 95% CI 0·05–0·14, I²=0%), and myocardial infarction (0·4% vs 1·0%; OR 0·44, 95% CI 0·23–0·87, I²=0%). However, carotid artery stenting was associated with an increased risk of periprocedural stroke (7·91% carotid artery stenting vs 4·56% carotid endarterectomy; OR 1·81, 95% CI 1·40–2·34, I²=12%), stroke or death (8·18% vs 5·00%; OR 1·72, 95% CI 1·29–2·31, I²=27%), and stroke, myocardial infarction, or death (8·42% vs 6·03%; OR 1·44, 95% CI 1·15–1·80, I²=7%).28 The rate of perioperative stroke or death was twice as high with carotid artery stenting than with carotid endarterectomy in patients 70 years of age or older (OR 2·20, 95% CI 1·47–3·29), whereas it was similar in patients younger than 70 years of age (1·16, 0·80–1·67; p value for interaction=0·02).50 The perioperative risks of carotid artery stenting compared with carotid endarterectomy were also greatest in patients treated within 7 days of symptom onset.58 In patients at high surgical risk for carotid endarterectomy, the perioperative risks of carotid artery stenting are raised in elderly patients with recent myocardial infarction, renal impairment requiring dialysis, longer carotid plaque, a tortuous carotid arterial system, and need for cardiac surgery.59 In the longer term, compared with carotid endarterectomy, carotid artery stenting was associated with an increased risk of severe carotid restenosis (10·6% vs 4·4%; OR 2·41, 95% CI 1·28–4·53, I²=55%), but not stroke (5·6% vs 4·5%; OR 1·26, 95% CI 0·77–2·04, I²=33%), after the periprocedural period in 2319 participants with symptomatic carotid stenosis followed up beyond 18 months in six trials.28 In the CREST trial, rates of restenosis and occlusion were similar at 2 years (6·0% carotid artery stenting vs 6·3% carotid endarterectomy, hazard ratio [HR] 0·90, 95% CI 0·63–1·29).60 An observational study of patients with recent symptomatic carotid stenosis led to the hypothesis that immediate treatment with a loading dose of aspirin and clopidogrel, followed by continued treatment with aspirin, clopidogrel, and a statin, reduces the risk of early recurrent ischaemic stroke sufficiently to potentially obviate the need for urgent carotid revascularisation.61 The substantial improvements in efficacy of medical treatments to prevent recurrent carotid atherothromboembolic ischaemic stroke since the late 1980s, when the controlled trials of carotid endarterectomy were done, also suggest that the absolute rates of recurrent
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
stroke with best possible medical treatment are now notably lower than they were 25 years ago. In the absence of similar substantial reductions in perioperative risks of carotid endarterectomy and stenting in the past 25 years,28 the benefits of adding carotid revascularisation to aggressive medical treatment, as used in the SAMMPRIS trial (see later), might be more marginal in 2014 than the published evidence quoted above from 1991–92 suggests.15,27,62,63 This hypothesis is being assessed in the second European Carotid Surgery Trial (ECST-2).62 For a recent (3 months) of aspirin and clopidogrel combined is not recommended because of the cumulative risks of bleeding (evidence level B).92–95 Anticoagulation • Patients with transient ischaemic attack and ischaemic stroke who have atrial fibrillation should be treated with anticoagulation, not antiplatelet therapy (evidence level A).37,38,96,97 • Warfarin (evidence level A),37,38 dabigatran (evidence level B),98,99 apixaban (evidence level B),100–102 and rivaroxaban (evidence level B)103,104 are all indicated in non-valvular atrial fibrillation. • The selection of an anticoagulant agent should be individualised on the basis of renal and hepatic function, potential for drug interactions, patient preference, cost, tolerability, and other clinical characteristics, including time in international normalised ratio therapeutic range if the patient has been taking warfarin. • For some patients, the individual’s preferences, level of disability, prognosis, and overall clinical status might preclude oral anticoagulation. • Apixaban 2·5 mg twice daily should be considered as an alternative to aspirin in stroke patients with non-valvular atrial fibrillation who are judged unsuitable for vitamin K antagonist therapy if their creatinine clearance is >25 mL per min (evidence level B).101 • Concomitant antiplatelet and anticoagulant therapy is not recommended in patients with atrial fibrillation unless there is
a specific indication, such as a mechanical heart valve, recent acute coronary syndrome, or coronary stent (evidence level B).105,106 Blood pressure • Gradual, sustained lowering of blood pressure is recommended in all stroke patients, but care is needed, particularly in patients with carotid or vertebrobasilar occlusive disease (evidence level A).40,41 • The ideal time to start lowering of blood pressure after stroke is uncertain, but it should be started before discharge from hospital (evidence level B).40,41 • The optimum blood pressure-lowering drugs depend on patient comorbidities. The combination of an angiotensin-converting enzyme inhibitor (or angiotensin II receptor antagonist) and calcium channel blocker (or diuretic in the elderly) might be preferable because the latter drugs (ie, calcium channel blockers and diuretics) also reduce blood pressure variability (evidence level C).41,107 • The target systolic blood pressure is lower than 130 mm Hg in patients with lacunar stroke (evidence level B).108 Lipids • Low-density-lipoprotein cholesterol concentration should be reduced in patients with ischaemic stroke or transient ischaemic attack caused by atherosclerosis by means of diet and lifestyle modification, and statin therapy (evidence level A).42,109,110 • The target low-density-lipoprotein cholesterol concentration is lower than 2 mmol/L (evidence level B).42,109,111 Lifestyle behaviours Smoking • Patients should stop smoking (evidence level B).112,113 • A combination of pharmacological (nicotine replacement therapy, bupropion, cytisine, or varenicline) and behavioural therapy should be considered (evidence level A).114
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
(Continues on next page)
7
Review
(Continued from previous page)
Alcohol consumption • Alcohol consumption should be limited to less than two standard drinks per day; less than 14 drinks per week for men; and less than nine drinks per week for women (evidence level C).112 Physical activity • Routine activities of daily living should be supplemented by moderate physical exercise—walking (ideally briskly), jogging, cycling, swimming, or other dynamic exercise—for 30–60 min on 4–7 days per week. High-risk patients (eg, those with cardiac disease) should participate in medically supervised exercise programmes (evidence level B).112 Bodyweight • The body-mass index (BMI) should be maintained at 18·5–24·9 kg/m², and waist circumference less than 80 cm for women and less than 94 cm for men (evidence level B).112 Sodium • The recommended adequate daily sodium intake for people aged 9–50 years is 1500 mg, decreasing to 1300 mg for individuals 50–70 years of age and to 1200 mg for those older than 70 years. A daily upper consumption limit of
Xa inhibitors rivaroxaban103 and apixaban100—are at least as efficacious and safe as warfarin. Apixaban101 is superior to, and as safe as, aspirin for preventing stroke among patients with atrial fibrillation. Compared with warfarin, the novel oral anticoagulants collectively reduce recurrent stroke or systemic embolism and major bleeding by about 14% and intracranial haemorrhage by about half (figure 1, table 1).39,99,102,104 Indirect comparisons of the relative effects of each novel oral anticoagulant, compared with warfarin, are consistent for stroke and systemic embolism (I²=0%), major bleeding (I²=6%), and all-cause mortality (I²=0%), but are not consistent for intracranial bleeding (I²=69%) or major gastrointestinal bleeding (I²=35%; figure 1).39 When modelling is used to subtract the hazards from the benefits of the novel oral anticoagulants compared with warfarin observed in the clinical trials and applied to real-world populations, all novel oral anticoagulants have a greater net clinical benefit than warfarin in patients with atrial fibrillation, particularly those at highest risk of stroke (CHA2DS2-VASc score >2).124 Figure 1 shows the effects of novel direct anticoagulants compared with warfarin in selected clinical trials. Of these trials, RE-LY is divided into two studies—RE-LY 110 involves participants randomly allocated to dabigatran 110 mg twice daily or warfarin, and RE-LY 150 enrolled participants who were randomly allocated to dabigatran 150 mg twice daily or warfarin. In the ROCKET AF trial, participants were randomly assigned to recieve rivaroxaban 20 mg daily or warfarin. In the final trial included in figure 1—ARISTOTLE—participants were randomly allocated to apixaban 5 mg twice daily or 8
2300 mg should not be exceeded by any age group (evidence level B).115,116 Healthy balanced diet • Eat a diet low in saturated fat, cholesterol, and sodium; and high in fresh fruits, vegetables, low-fat dairy products, dietary and soluble fibre, whole grains, and protein from plant sources (evidence level B).116 Levels of evidence • Level A=evidence from meta-analyses of randomised controlled trials or directly from randomised controlled trials. Desirable effects clearly outweigh undesirable effects, or vice versa. • Level B=single randomised controlled trial or well-designed observational study with strong evidence; or well-designed cohort or case–control analytical study; or several time series or dramatic results of an uncontrolled experiment. Desirable effects closely balanced with undesirable effects. • Level C=at least one well-designed, non-experimental descriptive study (eg, comparative studies, correlation studies, case studies) or expert committee reports, opinions and/or experience of respected authorities, including consensus from development and/or reviewer groups.
warfarin. Since the RE-LY trial compared two treatments (dabigatran 110 mg vs 150 mg twice daily) against one control (warfarin) group, the effect of dabigatran 110 mg twice daily versus warfarin and the effect of dabigatran 150 mg twice daily are not independent of each other. This fact needed to be taken into account in computation of the variance. To calculate a summary effect across all three studies (RE-LY, ROCKET AF, and ARISTOTLE), the weighted mean of the effects of dabigatran 110 mg compared with warfarin and of dabigatran 150 mg versus warfarin and the variance were computed, taking into account the correlations among the different treatments. Efforts are continuing to develop rapid, accurate, and widely available measures of the anticoagulant effects of the novel oral anticoagulants (eg, the Hemoclot test for dabigatran and anti-factor Xa assays for rivaroxaban and apixaban) and reversal strategies for major bleeding. Clinical trials are assessing an antidote for dabigatran (aDabi-Fab), which is a humanised antibody fragment that binds dabigatran more strongly than its affinity for thrombin and rapidly reverses its anticoagulant effects in vitro and in vivo.125 r-Antidote (PRT064445) is a recombinant protein in development that binds factor Xa inhibitors and reverses the anticoagulant effects of apixaban and rivaroxaban.126 For patients with atrial fibrillation who have had a stroke but in whom oral anticoagulation is contraindicated, the left atrial appendage can be occluded by the WATCHMAN device (a self-expanding cage placed in the left atrial appendage via a transeptal approach with femoral access). In 707 patients with atrial fibrillation randomly allocated
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
to the WATCHMAN or long-term warfarin treatment, the rates of stroke, cardiovascular death, or systemic embolism were similar (RR 0·62, 95% CI 0·35–1·25) but the rates of NOAC Events Total
A
Stroke and systemic embolism 1195 55
procedure-related complications were higher with the WATCHMAN device (1·69, 1·01–3·19).127 Definitive randomised trials of this device are needed.
Warfarin Events Total
Fixed effects odds ratio (95% CI)
Fixed odds ratio (95% CI)
65
1195
0·84 (0·58–1·21)
RE-LY 150
51
1233
65
1195
0·75 (0·52–1·09)
ROCKET AF
179
3754
187
3714
0·94 (0·77–1·17)
ARISTOTLE
73
1694
98
1742
0·76 (0·56–1·03)
358
7876
350
6651
0·86 (0·74–1·00)
RE-LY 110
Total
Heterogeneity: χ2=1·7, df=2, p=0·43, I2=0% Test for overall effect: Z=1·95, p=0·05 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
B
Major bleeding
RE-LY 110
65
1195
97
1195
0·65 (0·48–0·9)
RE-LY 150
102
1233
97
1195
1·02 (0·76–1·36)
ROCKET AF
178
3754
183
3714
0·96 (0·78–1·19)
ARISTOTLE
77
1694
106
1742
0·74 (0·55–0·99)
422
7876
386
6651
0·86 (0·75–1·00)
Total
Heterogeneity: χ2=2·13, df=2, p=0·34, I2=6% Test for overall effect: Z=1·97, p=0·05 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
C
Intracranial bleeding 6
1195
30
1195
0·26 (0·13–0·5)
RE-LY 150
13
1233
30
1195
0·43 (0·24–0·79)
ROCKET AF
34
3754
46
3714
0·73 (0·47–1·13)
ARISTOTLE
15
1694
41
1742
0·40 (0·24–0·68)
Total
68
7876
117
6651
0·45 (0·26–0·78)
RE-LY 110
Heterogeneity: χ2=6·41, df=2, p=0·04, I2=69% Test for overall effect: Z=2·83, p=0·005 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
D
Major gastrointestinal bleeding 1195 33
33
1195
1·00 (0·61–1·63)
RE-LY 150
57
1233
33
1195
1·68 (1·11–2·57)
ROCKET AF
0
0
0
0
ARISTOTLE
18
1694
22
1742
0·84 (0·45–1·57)
108
4122
55
2937
1·18 (0·85–1·63)
RE-LY 110
Total
Not estimable
Heterogeneity: χ2=1·53, df=1, p=0·22, I2=35% Test for overall effect: Z=0·97, p=0·33 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
E
Death from any cause 77
1195
107
1195
0·70 (0·52–0·95)
RE-LY 150
108
1233
107
1195
0·98 (0·74–1·29)
ROCKET AF
288
3754
294
3714
0·97 (0·82–1·14)
ARISTOTLE
129
1694
150
1742
0·88 (0·69–1·12)
Total
602
7876
551
6651
0·91 (0·81–1·03)
RE-LY 110
Heterogeneity: χ2=1·02, df=2, p=0·60, I2=0% Test for overall effect: Z=1·48, p=0·14 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
0·2
0·5
0·7
Favours NOAC
1
1·5
2
Favours warfarin
Figure 1: Forest plots showing meta-analyses of randomised controlled trials of the effects of NOACs versus warfarin on several outcomes The trials assessed the effects of NOACs versus warfarin on (A) stroke or systemic embolism, (B) major bleeding, (C) intracranial bleeding, (D) major gastrointestinal bleeding, and (E) death from any cause. Participants were patients with atrial fibrilliation and previous stroke or transient ischaemic attack, enrolled in the RE-LY,98,99 ROCKET AF,103,104 and ARISTOTLE100,102 trials.39 A fixed effects model was used to pool the data for all outcomes because no substantial heterogeneity occurred between the trials, except for intracranial haemorrhage (C), in which a random effects model was used because of the substantial heterogeneity between the trials (I²=69%; appendix pp 1–3).123 df=degrees of freedom. NOAC=novel direct oral anticoagulant.
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
9
Review
Ischaemic stroke from paradoxical embolism Patients with cryptogenic ischaemic stroke or TIA and a patent foramen ovale have a similar rate of recurrent ischaemic stroke (1·6% per year, 95% CI 1·1–2·1%) as patients without a patent foramen ovale (RR 1·1, 95% CI 0·8–1·5).128 However, an additional atrial septal aneurysm increases the risk of recurrent stroke (HR for both patent foramen ovale and atrial septal aneurysm vs neither=4·2, 95% CI 1·5–12).129 Possible strategies to reduce recurrent stroke from paradoxical embolism include antiplatelet drugs, anticoagulation, and percutaneous closure of the patent foramen ovale with a device. Three randomised controlled trials comparing patent foramen ovale closure with medical therapy showed that the device was implanted successfully in 81–92%130–132 of patients who underwent patent foramen ovale closure. Overall, patent foramen ovale closure was associated with a non-significant reduction in the primary outcome and recurrent stroke (figure 2) and a significant increase in new-onset atrial fibrillation (3·8% closure vs 1·0% medical; RR 3·5, 95% CI 1·5–8·3).130–133 The frequency of atrial fibrillation was increased by eight times with the STARFlex device in CLOSURE (5·7% vs 0·7%)130 and by two-to-three times with the Amplatzer device in RESPECT (3·0% vs 1·5%)131 and the PC trial (2·9% vs 1·0%).132 The trial results are compromised by treatment crossovers and short and incomplete follow-up over 2–4 years. Three randomised controlled trials of patent foramen ovale closure are PFO closure Events Total
A
Medical therapy Events Total
ongoing (CLOSE [NCT00562289], DEFENSEPFO [NCT01550588], and REDUCE [NCT00738894]).
Lowering of blood pressure to prevent all types of recurrent stroke Sustained lowering of blood pressure by 5·1 mm Hg systolic and 2·5 mm Hg diastolic reduces recurrent stroke by about a fifth (table 1).40 Larger reductions in blood pressure—by 10 mm Hg systolic and 5 mm Hg diastolic— are associated with larger reductions in recurrent stroke of about a third (RR 0·66, 95% CI 0·56–0·79).41 The effect is consistent, irrespective of previous hypertension and most subtypes of stroke. For patients who have survived for more than 2 weeks after lacunar stroke, gradual targeting of systolic blood pressure to lower than 130 mm Hg seems to be safe and shows a non-significant reduction in the rate of recurrent stroke by about a fifth compared with a target systolic blood pressure of 130–149 mm Hg (HR 0·81, 95% CI 0·64–1·03).108 Caution is needed in older patients and those with severe, bilateral carotid or vertebrobasilar occlusive disease. Blood pressure can be reduced by lifestyle interventions, such as regular physical exercise for 30 min daily, alcohol reduction, a low-salt diet, and increased potassium intake, and by medication with any class of antihypertensive drugs.41,115,134,135 In addition to reducing mean blood pressure, reducing visit-to-visit variability in systolic blood pressure, by means of calcium-channel blockers and diuretics, rather than β blockers, is also likely to reduce stroke risk.41,107,136 Peto fixed effects odds ratio (95% CI)
Weight
Peto fixed odds ratio (95% CI)
Primary outcome 23
447
29
462
54·1%
0·81 (0·46–1·42)
RESPECT
9
499
16
481
26·9%
0·54 (0·25–1·20)
PC trial
7
204
11
210
19·0%
0·65 (0·25–1·67)
1153
100·0%
CLOSURE-1
1150
Total Total events
0·70 (0·46–1·05)
56
39
Heterogeneity: χ2=0·69, df=2, p=0·71, I2=0% Test for overall effect: Z=1·72, p=0·09 0·2
0·5
Favours PFO closure
B
1·0
2·0
5·0
Favours medical therapy
Stroke 12
447
13
462
44·5%
0·95 (0·43–2·11)
RESPECT
9
499
16
481
44·6%
0·54 (0·25–1·20)
PC trial
1
204
5
210
10·8%
0·27 (0·05–1·34)
1153
100·0%
CLOSURE-1
1150
Total Total events
0·65 (0·38–1·10)
34
22
Heterogeneity: χ2=2·26, df=2, p=0·32, I2=12% Test for overall effect: Z=1·62, p=0·11 0·02
0·1 Favours PFO closure
1·0
10·0
50·0
Favours medical therapy
Figure 2: Forest plots of a pooled analysis, with a fixed effects model, of the published intention-to-treat data from three randomised controlled trials of PFO closure The trials were of PFO closure with the STARFlex device in the CLOSURE-1 trial130 or the Amplatzer PFO occluder (St Jude Medical) in the RESPECT131 and PC132 trials, plus medical therapy versus medical therapy alone (aspirin or warfarin or both). The pooled analysis shows that PFO closure was associated with a non-significant trend towards a reduction in the primary outcome (OR 0·70, 95% CI 0·46–1·05) and recurrent stroke (OR 0·65, 95% CI 0·38–1·10). PFO=patent foramen ovale.
10
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
Atherosclerotic ischaemic stroke Lowering of low-density lipoprotein (LDL) cholesterol concentration by about 1 mmol/L with statins reduces the risk of recurrent stroke by about 12% (table 1) and all stroke by about 21% (95% CI 6–33%).42 Greater reductions in LDL cholesterol concentration, by 2–3 mmol/L, are associated with greater reductions in stroke risk of 40–50%.42,109 The effect is consistent across noncardioembolic ischaemic stroke subtypes.110 However, the optimum target LDL cholesterol concentration is uncertain. Subgroup analyses raise the hypothesis that targeting of an LDL-C concentration below 1·8 mmol/L might be associated with a greater (28%) reduction in recurrent atherosclerotic stroke risk compared with a target of 2·59 mmol/L.111 This hypothesis is being tested in the Treat Stroke to Target (TST) trial (NCT01252875). A meta-analysis of 31 randomised controlled trials showed no difference in the incidence of intracerebral haemorrhage in 91 588 participants allocated statins and 91 215 controls (OR 1·08, 95% CI 0·88–1·32), and no relation between intracerebral haemorrhage risk and LDL cholesterol concentrations or reduction.137 However, the two largest trials of statins in patients with previous stroke both showed that statins were associated with a small absolute, but large relative, increase in risk of haemorrhagic stroke (67% [95% CI 9–156]) and 91% [–8 to 395], respectively).42,137 Allocation to atorvastatin was an independent predictor of risk of haemorrhagic stroke,138 but the risk of haemorrhagic stroke was not related to the degree of LDL cholesterol reduction.139 Much of the risk of haemorrhagic stroke associated with atorvastatin was in patients with a qualifying diagnosis of stroke because of small vessel disease.138 This fact negated the benefits of atorvastatin in reducing ischaemic events in patients with previous haemorrhagic small vessel disease but not in patients with ischaemic small vessel disease.110 For patients with stroke who do not achieve a low enough LDL cholesterol concentration, the addition of ezetimibe 10 mg, a cholesterol absorption inhibitor, to statin therapy could produce greater reductions in LDL cholesterol concentration than statin monotherapy.140 However, in 720 patients with familial hypercholesterolaemia, combined treatment with ezetimibe 10 mg daily and simvastatin 80 mg daily did not significantly change carotid artery intima-media thickness compared with simvastatin alone, despite significantly reducing LDL cholesterol concentrations by 16%.141 PCSK9 is a protein secreted by hepatocytes that regulates surface expression of LDL cholesterol receptors by targeting them for lysosomal degradation. Loss-of-function mutations in the PCSK9 gene produce lifelong low concentrations of LDL cholesterol and reduce cardiovascular risk. Statins enhance PCSK9 synthesis, thereby blunting their effectiveness in lowering LDL cholesterol concentrations. Human monoclonal antibodies to PCSK9, such as REGN727/SAR236553 (REGN727) and AMG 145, administered subcutaneously every 2 or 4 weeks to patients
with hypercholesterolaemia have been well tolerated; LDL-C concentrations have been lowered by 40–70%, despite statins and ezetimibe; and apolipoprotein B, triglycerides, lipoprotein (a), and high-density lipoprotein C concentrations have been favourably altered.142–144 PCSK9 inhibitors are being tested in phase 3 trials. Raising of high-density lipoprotein (HDL) cholesterol concentrations, by inhibiting cholesteryl ester transfer protein with dalcetrapib, does not reduce recurrent cardiovascular events in patients with acute coronary syndrome,145 but its effect in patients with ischaemic stroke is uncertain. The addition of extended-release niacin (1500–2000 mg per day) to simvastatin (40–80 mg per day), plus ezetimibe (10 mg per day) if needed, increased median HDL cholesterol concentration (from 0·91 to 1·08 mmol/L) and lowered LDL cholesterol concentration (from 1·91 to 1·60 mmol/L), but had no significant effect on ischaemic stroke (HR 1·78, 95% CI 1·00–3·17) or the primary outcome of major vascular events (1·02, 0·87–1·21) compared with placebo in 3414 patients with atherosclerotic cardiovascular disease and LDL cholesterol concentration lower than 1·81 mmol/L.146 The addition of extended-release niacin 2 g plus laropiprant 40 mg daily to simvastatin 40 mg plus, if needed, ezetimibe 10 mg daily, had no effect on major vascular events compared with placebo (RR 0·96, 95% CI 0·90–1·03) but significantly increased diabetic complications, new-onset diabetes, infection, bleeding, haemorrhagic stroke, and myopathy after 3·9 years’ (median) follow-up in 25 673 patients with occlusive arterial disease.147,148 A meta-analysis of eight randomised controlled trials in patients with type 2 diabetes showed that, compared with a standard regimen (n=15 844), intensive treatment with oral hypoglycaemic drugs or insulin to reduce blood glucose concentration (n=17 742) over a mean of 5 years was associated with no reduction in stroke (RR 0·96, 99% CI 0·83–1·13; I²=0%) or mortality (RR 1·04, 99% CI 0·91–1·19; I²=42%).149 Although intensive treatment reduced non-fatal myocardial infarction (RR 0·85, 99% CI 0·74–0·96) and microalbuminuria (0·90, 0·85–0·96), it increased severe hypoglycaemia (2·33, 1·62–3·36). More randomised controlled trials are needed to establish the best approach for blood glucose management in people with type 2 diabetes. Lowering total plasma homocysteine with folic acid, vitamin B6, and vitamin B12 does not prevent recurrent stroke (RR 0·92, 95% CI 0·81–1·06) or any stroke (0·91, 0·82–1·00, I²=11%) compared with placebo in populations with established or increasing intake of folate and vitamin B12.150 However, a reduction in total plasma homocysteine might effectively prevent stroke in people with a low intake of folate or vitamin B12 (eg, Asian people),151 intracranial small vessel disease,152 and those not taking or who are unable to take antiplatelet therapies, such as aspirin.153
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
11
Review
Dietary supplementation with the antioxidant vitamins β-carotene, vitamin C, and vitamin E does not prevent stroke.154,155 Whether or not vitamin D supplementation can prevent stroke is uncertain.154 A meta-analysis of 18 randomised controlled trials of hormone replacement therapy involving 36 523 women showed that this treatment increased the risk of stroke by a third (OR 1·32, 95% CI 1·14–1·53) compared with controls.156 A subsequent smaller trial showed that hormone replacement therapy, if started early and used for more than 10 years, did not increase stroke (14 strokes in the control group vs 11 strokes in the hormone replacement therapy group; HR 0·77, 95% CI 0·35–1·70),157 but this study was not powered to reliably exclude the one-third increased relative risk of stroke in the earlier meta-analysis of randomised controlled trials.156 Epidemiological data suggest that a substantial proportion of strokes can be attributed to unhealthy lifestyle behaviours (table 4).112 Cessation of smoking and alcohol abuse and adherence to careful diets that are aligned to the Mediterranean or dietary approaches to stop hypertension, and are low in salt and added sugars, high in potassium, and meet, but do not exceed, energy requirements, are associated with reduced risks of stroke.112,116
Translation of evidence into practice Convincing evidence outlined in this Review suggests that several interventions are efficacous in reducing recurrent stroke under ideal circumstances (ie, explanatory clinical trial settings). The combination of five of these strategies— aspirin, an antihypertensive drug, a statin, exercise, and dietary modification—could reduce recurrent stroke by 80%.158 However, evidence exists that these interventions
Prevalence (%)
Odds ratio (99% CI)
Population-attributable risk (99% CI)
Controls
Cases
History of hypertension
32%
56%
2·6 (2·3–3·1)
35% (30–39)
Regular physical activity
12%
8%
0·7 (0·5–0·9)
28% (14–48)
Waist-to-hip ratio (T3 vs T1)
41%
33%
1·6 (1·4–2·0)
26% (19–36)
Ratio of apolipoprotein B to apolipoprotein A-I (T3 vs T1)
33%
47%
1·9 (1·5–2·4)
25% (16–37)
Current smokers
24%
36%
2·1 (1·7–2·5)
19% (15–23)
Diet risk score (T3 vs T1)
35%
30%
1·4 (1·1–1·6)
19% (11–30)
Cardiac causes*
5%
12%
2·4 (2·8–3·2)
7% (5–9)
Diabetes
12%
19%
1·4 (1·1–1·7)
5% (3–9)
Depression
14%
20%
1·4 (1·1–1·6)
5% (3–10)
Psychosocial stress
15%
20%
1·3 (1·1–1·6)
5% (2–10)
Alcohol intake >30 drinks per month
11%
16%
1·5 (1·2–1·9)
4% (1–14)
Control participants were matched with cases for age and sex, and were assessed in 22 countries between 2007 and 2010. The multivariate model was adjusted for age, sex, and region. T3=highest tertile. T1=lowest tertile. *Cardiac causes include atrial fibrillation or flutter, previous myocardial infarction, rheumatic valve disease, or prosthetic valve disease.
Table 4: Risk factors for all types of stroke in 3000 cases of acute first stroke compared with 3000 controls with no history of stroke in the INTERSTROKE study112
12
are effective (ie, they do actually work, and produce more good than harm when used under normal circumstances in clinical practice) and are cost effective (ie, their effects in relation to the resources they consume are worth it) is less convincing. Several studies have shown that the effectiveness of secondary stroke prevention is insufficient because of the widespread failure to appropriately translate evidence-based recommendations into clinical practice.159–165 The most recent global study reported that in 2292 adults with previous stroke who were living in 628 urban and rural communities in three high-income, seven upper-middle-income, three lower-middleincome, and four low-income countries, few took antiplatelet drugs (24·3%), blood pressure-lowering drugs (40·0%), or statins (9·0%).164 Use of these drugs was highest in high-income countries (antiplatelet drugs 53·1%, blood pressure-lowering drugs 60·6%, and statins 51·6%), lowest in low-income countries (3·8%, 13·0%, and 0·6%, respectively), and decreased with reduction in country economic status (p value for trend
Secondary stroke prevention Graeme J Hankey
Survivors of stroke and transient ischaemic attacks are at risk of a recurrent stroke, which is often more severe and disabling than the index event. Optimum secondary prevention of recurrent stroke needs rapid diagnosis and treatment and prompt identification of the underlying cardiovascular cause. Effective treatments include organised acute assessment and intervention with antithrombotic therapy, carotid revascularisation, and control of causal risk factors, as appropriate. However, effective treatments are not implemented optimally in clinical practice. Recurrent strokes continue to account for 25–30% of all strokes and represent unsuccessful secondary prevention. Immediate and sustained implementation of effective and appropriate secondary prevention strategies in patients with first-ever stroke or transient ischaemic attack has the potential to reduce the burden of stroke by up to a quarter.
Introduction The world is facing a stroke epidemic: between 1990 and 2010, the number of stroke-related deaths increased by 26% (95% CI 14–32) and disability-adjusted life-years by 19% (5–26%), to make stroke the second leading cause of death and third leading contributor to disabilityadjusted life-years in the world.1,2 Complementary strategies to reduce the increasing global burden of stroke include the so-called population/mass and highrisk approaches to prevent first-ever and recurrent stroke, and the treatment and rehabilitation of acute stroke to improve disability-free survival.3 The most relevant of these approaches for stroke physicians are the treatment and rehabilitation of acute stroke and the secondary prevention of recurrent stroke. This Review summarises evidence for optimum secondary prevention of recurrent ischaemic stroke. The rationale for this report is that recurrent strokes constitute a notable proportion (25–30%) of all preventable strokes;4 they are frequently ischaemic and more disabling, fatal, and costly than the first stroke;4,5 and their rates could be reduced by improved translation of evidence-based advances in secondary stroke prevention into clinical practice.6 Since the rate of recurrent stroke and the potential benefits and risks of secondary prevention strategies are higher earlier rather than later after ischaemic stroke, I will describe separately the prognosis and predictors for early and long-term recurrent ischaemic stroke and the strategies to prevent both types of stroke.
Early recurrent stroke Prognosis and predictors The risk of a recurrent stroke is highest early after an ischaemic stroke or transient ischaemic attack (TIA)— about 1% at 6 h, 2% at 12 h, 3% at 2 days, 5% at 7 days, and 10% at 14 days.7–9 Therefore, ischaemic stroke or TIA is a medical emergency that demands immediate diagnosis and treatment. Some clinical features, such as sudden-onset unilateral weakness and speech disturbance lasting for longer than 10 min, predict a high risk of stroke soon after TIA, probably because they distinguish TIA from a myriad of TIA mimics that have a more benign
prognosis.10 These, and other, clinical features have been used to derive a prognostic index: the ABCD2 score (panel 1; appendix pp 4–5).10–12 Additional prognostic information from carotid and brain imaging about the cause of the ischaemic event has been incorporated into the ABCD3–I score19 (panel 1; appendix pp 4–5) and the recurrence risk estimator.14,24 However, low scores on these prognostic indices should not preclude appropriate assessment and management because some patients with a low predicted risk of stroke have treatable causes, such as carotid stenosis and atrial fibrillation, that might warrant urgent intervention.17,18
Lancet Neurol 2013 Published Online December 20, 2013 http://dx.doi.org/10.1016/ S1474-4422(13)70255-2 School of Medicine and Pharmacology, The University of Western Australia, Perth, Australia; and Department of Neurology, Sir Charles Gairdner Hospital, Nedlands, Perth, Australia (Prof G J Hankey MD) Correspondence to: Prof Graeme J Hankey, School of Medicine and Pharmacology, The University of Western Australia, Room 222, Harry Perkins Institute of Medical Research, QQ Block, QEII Medical Centre, 6 Verdun Street, Nedlands, Perth, 6009, Australia [email protected] See Online for appendix For the recurrence risk estimator see http://www.nmr. mgh.harvard.edu/RRE/
Panel 1: The ABCD2 and ABCD3–I scores to predict early recurrent stroke ABCD2 The ABCD2 score (appendix pp 4–5) is a simple scoring system that can be used in primary care and by emergency department clinicians to stratify patients with transient ischaemic attack (TIA) into low, moderate, and high risk of early recurrent stroke.10 The score has been validated in several independent datasets to have good power to predict the rate and severity of recurrent stroke at 7 days.11,12 A limitation of the ABCD2 score is that it was not derived from datasets with information about the cardiovascular lesion(s) causing the ischaemic stroke or TIA, such as carotid stenosis and atrial fibrillation, which also determine early prognosis. Most early recurrent strokes are caused by recurrent thromboembolism from an unstable atherosclerotic plaque that is still unstable (vs recurrent cardiogenic embolism or small vessel occlusion).13–16 Another limitation of the score is that 10% of strokes at 7 days after TIA occur in patients judged to be at low risk (ABCD2 score 0–3), and 50% of strokes occur in patients classed as moderate risk (ABCD2 score 4–5).17,18 Hence, patients with TIA should be assessed immediately, irrespective of ABCD2 score, because some with lower scores have treatable causes, such as carotid stenosis and atrial fibrillation, that might be associated with raised short-term risks of stroke. ABCD3–I The ABCD3–I score has been developed for use in secondary care settings that have access to carotid and brain imaging (appendix pp 4–5).19 It uses the prognostic significance of the cardiovascular lesion causing the ischaemic stroke or TIA, and the presence of recent focal brain ischaemia detected by magnetic resonance diffusion-weighted imaging.20,21 It has been validated externally as having a higher predictive value than the ABCD2 score for assessing risk of early stroke.19,22 Limitations of the ABCD3–I score are that the yield of magnetic resonance diffusion-weighted imaging varies according to the timing of the scan after the TIA; if the procedure is undertaken more than 24 h after symptom onset, the odds of finding a diffusion-weighted imaging lesion could be increased by up to threefold.23
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
1
Review
Secondary prevention Effective strategies to prevent early recurrent ischaemic stroke include assessment and management in an acute specialist unit, immediate antiplatelet therapy, and early carotid revascularisation, as appropriate (table 1, panel 2).
Antiplatelet and anticoagulation therapies As a single antiplatelet therapy, aspirin 160–300 mg daily, started within 48 h of onset of ischaemic stroke in 40 000 patients, and continued for 2–4 weeks, reduced the odds of recurrent ischaemic stroke by 23% (2·4% aspirin vs 3·1% control; odds ratio [OR] 0·77, 95% CI 0·69–0·87), increased the odds of symptomatic intracranial haemorrhage by 22% (1·0% vs 0·8%; OR 1·22, 95% CI 1·00–1·50), and reduced the odds of any recurrent stroke by 12% (3·4% vs 3·9%; OR 0·88, 95% CI 0·79–0·97) compared with control (table 1).25 A meta-analysis of 14 randomised controlled trials comparing dual antiplatelet therapy with antiplatelet Outcome
monotherapy, both started within 3 days of onset of ischaemic stroke or TIA, in 9102 adults showed that dual antiplatelet therapy significantly reduced early recurrent stroke by a third (6·21% dual therapy vs 8·95% monotherapy; risk ratio [RR] 0·69, 95% CI 0·60–0·80, absolute risk reduction 2·74%) at about 3 months’ followup (range 7 days–18 months), compared with monotherapy, and non-significantly increased the risk of major bleeding (0·52% dual therapy vs 0·36% monotherapy; RR 1·35, 95% CI 0·70–2·59, absolute risk increase 0·16%).26 The results of trials comparing different dual combinations of antiplatelet drugs with different single antiplatelet drugs were consistent with the overall result for all trials (table 1).26 However, the results might not apply to patients with major ischaemic stroke (eg, those with a score on the National Institutes of Stroke Scale >4) who were excluded from many trials because of the risk of haemorrhagic transformation of new brain infarction.51
Stroke rate
RRR (95% CI)
Control
Intervention
ARR
NNT and time period
Early recurrent stroke Aspirin (vs no aspirin)25
Stroke at 2–4 weeks
Clopidogrel plus aspirin (vs aspirin)9,26
Stroke at roughly 3 months
3·9%
3·4%
12% (3 to 21)
0·5%
200 over 2–4 weeks
11·1%
7·8%
30% (18 to 41)
3·3%
Aspirin plus dipyridamole (vs aspirin)26
30 over roughly 3 months
Stroke at 3–28 months
7·9%
5·3%
36% (–10 to 63)
2·6%
38 over 3–28 months
Aspirin plus dipyridamole (vs clopidogrel)26
Stroke at 3 months
2·9%
1·6%
44% (–17 to 73)
1·8%
56 over 3 months
Clopidogrel plus aspirin (vs clopidogrel)26
Stroke at 18 months
4·7%
3·9%
17% (–93 to 64)
0·8%
125 over 18 months
Carotid endarterectomy or stent (vs no carotid revascularisation)27,28 70–99% stenosis, symptomatic
Stroke at 5 years
33%
17%
48% (38 to 60)
16%
6 over 5 years
50–69% stenosis, symptomatic
Stroke at 5 years
27%
19%
28% (14 to 42)
8%
12 over 5 years
Stroke at 90 days
10·3%
2·1%
80% (51 to 92)
8·2%
12 over 90 days
7·0%
6·1%
13% (6 to 19)
0·9%
111 over 1 year
All high-vascular-risk patients
5·8%
5·3%
9% (0·3 to 16)
0·5%
200 over 1 year
Ischaemic stroke patients
7·7%
7·1%
7% (–6 to 19)
0·6%
167 over 1 year
Recurrent stroke, per 2·6 years (mean)
11·3%
9·0%
22% (10 to 32)
2·3%
43 over 2·6 years, or 113 over 1 year
Stroke, MI, or VD, per 2·6 years (mean)
15·2%
12·5%
18% (8 to 28)
2·7%
37 over 2·6 years, or 100 over 1 year
Acute specialty units (vs outpatient clinics)29 Longer-term recurrent stroke Aspirin (vs no aspirin)30,31
Stroke, MI, or VD, per year
Clopidogrel (vs aspirin) 32,33
Stroke, MI, or VD, per year
Aspirin and extended-release dipyridamole (vs aspirin)34
Aspirin and extended-release dipyridamole (vs clopidogrel)35
Recurrent stroke, per 2·5 years (mean)
8·8%
9·0%
−1% (–11 to 8)
Stroke, MI, or VD per 2·5 years (mean)
13·1%
13·1%
1% (–7 to 8)
8·8%
6·3%
Cilostazol (vs aspirin)36
Stroke, MI, or VD, per 1·25 years (mean)
Warfarin for atrial fibrillation (vs no warfarin)37,38
Recurrent stroke, per year
New direct oral anticoagulants for atrial fibrillation (vs warfarin)39
Stroke and systemic embolism, per 1·9 years
Lowering of blood pressure (vs no lowering of blood pressure)40,41 Recurrent stroke, per 3 years (roughly) Statins to reduce low-density-lipoprotein cholesterol (vs no statin)42
Recurrent stroke, per 5 years
12%
0·2% (NS) NA 0% (NS)
NA
28% (11 to 43)
2·5%
40 over 1·25 years, or 50 over 1 year
4%
61% (37 to 75)
8%
12 over 1 year
5·3%
4·5%
14% (0 to 26)
0·8%
134 over 1·9 years, or 255 over 1 year
9·9%
8·6%
22% (10 to 32)
1·3%
75 over roughly 3 years, or 225 over roughly 1 year
11·9%
10·5%
12% (1 to 22)
1·4%
74 over 5 years, or 370 over roughly 1 year
See appendix pp 1–3 for explanations of statistical terms. RRR=relative risk reduction (risk ratio). ARR=absolute risk reduction. NNT=number needed to treat. NS=not significant. MI=myocardial infarction. VD=vascular death. NA=not applicable.
Table 1: Summary of effective strategies to prevent recurrent stroke
2
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
Panel 2: Recommendations for early secondary prevention Acute specialty units • Patients with suspected transient ischaemic attack (TIA) should be assessed and managed urgently in an acute specialty unit, such as a dedicated TIA clinic.29,43 Investigations usually comprise immediate imaging of the brain, arteries to the brain, and heart; electrocardiograph; and measurement of fasting blood glucose and lipids. Interventions usually include antiplatelet therapy, blood pressure-lowering, statins, and anticoagulation and carotid endarterectomy or stenting, as appropriate (evidence level B). Antiplatelet therapy • All patients with acute TIA and ischaemic stroke should be given at least 160 mg of aspirin or acetylsalicylic acid immediately as a single loading dose (evidence level A).25 • In dysphagic patients, aspirin can be given by enteral tube or by rectal suppository. • Chinese patients with acute (160 mg) was used on day 1 to rapidly inhibit platelet aggregation—initiation with a daily dose of 75 mg
• Patients with atrial fibrillation and acute transient ischaemic attack can begin oral anticoagulation (warfarin, dabigatran, rivaroxaban, or apixaban) immediately because the risk of intracranial haemorrhage is probably low (ie, there is no fresh brain infarction to become haemorrhagic). Carotid revascularisation • Patients with recently symptomatic carotid territory transient ischaemic attack or non-disabling ischaemic stroke and ipsilateral 50–99% internal carotid artery stenosis (measured by two concordant non-invasive imaging modalities) who are fit and willing for surgery should be offered carotid endarterectomy as soon as possible, ideally within the first few days and up to 1 week after the ischaemic event and when the patient is clinically stable (evidence level A).15,27 • Carotid endarterectomy should be done by a surgeon with an audited perioperative morbidity and mortality of less than 5% (evidence level A).27 • Carotid endarterectomy is more appropriate than carotid stenting for patients older than 70 years of age who are otherwise fit for surgery because stenting carries a higher short-term risk of stroke and death (evidence level A).27,50 Carotid stenting might be as safe as endarterectomy in patients less than 70 years of age.28,50 • Carotid stenting could also be considered for patients who are not candidates for carotid endarterectomy because of technical, anatomical, or medical reasons (evidence level A).28 Interventionalists should have expertise in carotid procedures and an expected risk of periprocedural morbidity and mortality of less than 5%. Levels of evidence • Level A=evidence from meta-analyses of randomised controlled trials or directly from randomised controlled trials. Desirable effects clearly outweigh undesirable effects, or vice versa. • Level B=single randomised controlled trial or well-designed observational study with strong evidence; or well-designed cohort or case–control analytical study; or several time series or dramatic results of an uncontrolled experiment. Desirable effects closely balanced with undesirable effects. • Level C=at least one well-designed, non-experimental descriptive study (eg, comparative studies, correlation studies, case studies) or expert committee reports, opinions and/or experience of respected authorities, including consensus from development and/or reviewer groups.
of clopidogrel and aspirin takes several days to produce maximum inhibition of platelet aggregation.9,45 Most of the benefit of clopidogrel plus aspirin was seen within the first few days after TIA and ischaemic stroke, when the risk of recurrence was highest.9 The safety and efficacy of dual and triple antiplatelet therapy in non-Chinese populations with acute TIA and mild ischaemic stroke is
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
3
Review
being tested in two large phase 3 trials.52,53 Newer and more potent antiplatelet drugs than aspirin and clopidogrel, such as prasugrel and ticagrelor, that have shown efficacy in atherothrombotic acute coronary syndromes are being considered for clinical trials in acute atherothromboembolic TIA and mild ischaemic stroke trials.54 In 23 748 patients with acute ischaemic stroke of presumed arterial origin, anticoagulation started within 48 h of stroke reduced early recurrent ischaemic stroke (OR 0·76, 95% CI 0·65–0·88) but increased symptomatic intracranial haemorrhage (2·55, 1·95–3·33), and achieved no net benefit in reducing any recurrent stroke (0·97, 0·85–1·11).46 The results were consistent for all types of heparin (intravenous and subcutaneous unfractionated heparin, heparinoids, and low-molecularweight heparin) and in patients at raised risk of thrombotic events and reduced risk of haemorrhagic events, which suggests that routine and selective use of heparin in ischaemic stroke has no net benefit.46,47 The new direct oral anticoagulants, one of which (rivaroxaban) has shown efficacy in acute coronary syndromes, could be assessed as potential treatments for acute TIA and mild ischaemic stroke of presumed arterial origin.55 In 4624 patients with acute ischaemic stroke of presumed cardiac origin, anticoagulation started within 48 h of cardioembolic ischaemic stroke did not significantly reduce early recurrent ischaemic stroke (OR 0·7, 95% CI 0·4–1·1) but increased symptomatic intracranial haemorrhage (2·9, 1·2–7·0), with no net effect on any recurrent stroke (1·18, 0·74–1·88).48 Whether or not early anticoagulation is safe and effective in selected patients at high risk of early recurrent cardioembolic stroke and low risk of haemorrhagic transformation of recently infarcted brain remains uncertain.
Revascularisation The perioperative risk of stroke or death associated with carotid endarterectomy for symptomatic carotid stenosis is about 7% (95% CI 6·2–8·0).27 Perioperative risks are increased in women; in patients with hemispheric events, contralateral carotid occlusion, irregular or ulcerated plaque, diabetes, chronic obstructive pulmonary disease, ischaemic heart disease, heart failure, peripheral arterial disease, and renal impairment warranting dialysis; and when carotid endarterectomy is undertaken within 48 h after ischaemic stroke.27,56,57 For long-term outcomes, the addition of carotid endarterectomy to best medical treatment reduces the risk of stroke or death at 5 years by half in patients with recently symptomatic 70–99% carotid stenosis, and by a quarter in those with 50–69% stenosis (table 1).27 The benefits of the procedure are greatest in men, older patients, those with recent (rather than later) ischaemic events of the brain (vs the eye), those with increasing 4
carotid stenosis, and in patients with ulcerated (vs smooth) carotid plaque.15 A systematic review compared carotid endarterectomy with carotid artery stenting in 7572 patients with recent symptomatic carotid stenosis enrolled in 16 trials.28 Compared with carotid endarterectomy, carotid artery stenting was associated with a lower risk of periprocedural access site haematoma (0·9% carotid artery stenting vs 2·7% carotid endarterectomy; OR 0·37, 95% CI 0·18–0·77, I²=27%), cranial nerve injury (0·3% vs 5·5%; OR 0·08, 95% CI 0·05–0·14, I²=0%), and myocardial infarction (0·4% vs 1·0%; OR 0·44, 95% CI 0·23–0·87, I²=0%). However, carotid artery stenting was associated with an increased risk of periprocedural stroke (7·91% carotid artery stenting vs 4·56% carotid endarterectomy; OR 1·81, 95% CI 1·40–2·34, I²=12%), stroke or death (8·18% vs 5·00%; OR 1·72, 95% CI 1·29–2·31, I²=27%), and stroke, myocardial infarction, or death (8·42% vs 6·03%; OR 1·44, 95% CI 1·15–1·80, I²=7%).28 The rate of perioperative stroke or death was twice as high with carotid artery stenting than with carotid endarterectomy in patients 70 years of age or older (OR 2·20, 95% CI 1·47–3·29), whereas it was similar in patients younger than 70 years of age (1·16, 0·80–1·67; p value for interaction=0·02).50 The perioperative risks of carotid artery stenting compared with carotid endarterectomy were also greatest in patients treated within 7 days of symptom onset.58 In patients at high surgical risk for carotid endarterectomy, the perioperative risks of carotid artery stenting are raised in elderly patients with recent myocardial infarction, renal impairment requiring dialysis, longer carotid plaque, a tortuous carotid arterial system, and need for cardiac surgery.59 In the longer term, compared with carotid endarterectomy, carotid artery stenting was associated with an increased risk of severe carotid restenosis (10·6% vs 4·4%; OR 2·41, 95% CI 1·28–4·53, I²=55%), but not stroke (5·6% vs 4·5%; OR 1·26, 95% CI 0·77–2·04, I²=33%), after the periprocedural period in 2319 participants with symptomatic carotid stenosis followed up beyond 18 months in six trials.28 In the CREST trial, rates of restenosis and occlusion were similar at 2 years (6·0% carotid artery stenting vs 6·3% carotid endarterectomy, hazard ratio [HR] 0·90, 95% CI 0·63–1·29).60 An observational study of patients with recent symptomatic carotid stenosis led to the hypothesis that immediate treatment with a loading dose of aspirin and clopidogrel, followed by continued treatment with aspirin, clopidogrel, and a statin, reduces the risk of early recurrent ischaemic stroke sufficiently to potentially obviate the need for urgent carotid revascularisation.61 The substantial improvements in efficacy of medical treatments to prevent recurrent carotid atherothromboembolic ischaemic stroke since the late 1980s, when the controlled trials of carotid endarterectomy were done, also suggest that the absolute rates of recurrent
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
stroke with best possible medical treatment are now notably lower than they were 25 years ago. In the absence of similar substantial reductions in perioperative risks of carotid endarterectomy and stenting in the past 25 years,28 the benefits of adding carotid revascularisation to aggressive medical treatment, as used in the SAMMPRIS trial (see later), might be more marginal in 2014 than the published evidence quoted above from 1991–92 suggests.15,27,62,63 This hypothesis is being assessed in the second European Carotid Surgery Trial (ECST-2).62 For a recent (3 months) of aspirin and clopidogrel combined is not recommended because of the cumulative risks of bleeding (evidence level B).92–95 Anticoagulation • Patients with transient ischaemic attack and ischaemic stroke who have atrial fibrillation should be treated with anticoagulation, not antiplatelet therapy (evidence level A).37,38,96,97 • Warfarin (evidence level A),37,38 dabigatran (evidence level B),98,99 apixaban (evidence level B),100–102 and rivaroxaban (evidence level B)103,104 are all indicated in non-valvular atrial fibrillation. • The selection of an anticoagulant agent should be individualised on the basis of renal and hepatic function, potential for drug interactions, patient preference, cost, tolerability, and other clinical characteristics, including time in international normalised ratio therapeutic range if the patient has been taking warfarin. • For some patients, the individual’s preferences, level of disability, prognosis, and overall clinical status might preclude oral anticoagulation. • Apixaban 2·5 mg twice daily should be considered as an alternative to aspirin in stroke patients with non-valvular atrial fibrillation who are judged unsuitable for vitamin K antagonist therapy if their creatinine clearance is >25 mL per min (evidence level B).101 • Concomitant antiplatelet and anticoagulant therapy is not recommended in patients with atrial fibrillation unless there is
a specific indication, such as a mechanical heart valve, recent acute coronary syndrome, or coronary stent (evidence level B).105,106 Blood pressure • Gradual, sustained lowering of blood pressure is recommended in all stroke patients, but care is needed, particularly in patients with carotid or vertebrobasilar occlusive disease (evidence level A).40,41 • The ideal time to start lowering of blood pressure after stroke is uncertain, but it should be started before discharge from hospital (evidence level B).40,41 • The optimum blood pressure-lowering drugs depend on patient comorbidities. The combination of an angiotensin-converting enzyme inhibitor (or angiotensin II receptor antagonist) and calcium channel blocker (or diuretic in the elderly) might be preferable because the latter drugs (ie, calcium channel blockers and diuretics) also reduce blood pressure variability (evidence level C).41,107 • The target systolic blood pressure is lower than 130 mm Hg in patients with lacunar stroke (evidence level B).108 Lipids • Low-density-lipoprotein cholesterol concentration should be reduced in patients with ischaemic stroke or transient ischaemic attack caused by atherosclerosis by means of diet and lifestyle modification, and statin therapy (evidence level A).42,109,110 • The target low-density-lipoprotein cholesterol concentration is lower than 2 mmol/L (evidence level B).42,109,111 Lifestyle behaviours Smoking • Patients should stop smoking (evidence level B).112,113 • A combination of pharmacological (nicotine replacement therapy, bupropion, cytisine, or varenicline) and behavioural therapy should be considered (evidence level A).114
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
(Continues on next page)
7
Review
(Continued from previous page)
Alcohol consumption • Alcohol consumption should be limited to less than two standard drinks per day; less than 14 drinks per week for men; and less than nine drinks per week for women (evidence level C).112 Physical activity • Routine activities of daily living should be supplemented by moderate physical exercise—walking (ideally briskly), jogging, cycling, swimming, or other dynamic exercise—for 30–60 min on 4–7 days per week. High-risk patients (eg, those with cardiac disease) should participate in medically supervised exercise programmes (evidence level B).112 Bodyweight • The body-mass index (BMI) should be maintained at 18·5–24·9 kg/m², and waist circumference less than 80 cm for women and less than 94 cm for men (evidence level B).112 Sodium • The recommended adequate daily sodium intake for people aged 9–50 years is 1500 mg, decreasing to 1300 mg for individuals 50–70 years of age and to 1200 mg for those older than 70 years. A daily upper consumption limit of
Xa inhibitors rivaroxaban103 and apixaban100—are at least as efficacious and safe as warfarin. Apixaban101 is superior to, and as safe as, aspirin for preventing stroke among patients with atrial fibrillation. Compared with warfarin, the novel oral anticoagulants collectively reduce recurrent stroke or systemic embolism and major bleeding by about 14% and intracranial haemorrhage by about half (figure 1, table 1).39,99,102,104 Indirect comparisons of the relative effects of each novel oral anticoagulant, compared with warfarin, are consistent for stroke and systemic embolism (I²=0%), major bleeding (I²=6%), and all-cause mortality (I²=0%), but are not consistent for intracranial bleeding (I²=69%) or major gastrointestinal bleeding (I²=35%; figure 1).39 When modelling is used to subtract the hazards from the benefits of the novel oral anticoagulants compared with warfarin observed in the clinical trials and applied to real-world populations, all novel oral anticoagulants have a greater net clinical benefit than warfarin in patients with atrial fibrillation, particularly those at highest risk of stroke (CHA2DS2-VASc score >2).124 Figure 1 shows the effects of novel direct anticoagulants compared with warfarin in selected clinical trials. Of these trials, RE-LY is divided into two studies—RE-LY 110 involves participants randomly allocated to dabigatran 110 mg twice daily or warfarin, and RE-LY 150 enrolled participants who were randomly allocated to dabigatran 150 mg twice daily or warfarin. In the ROCKET AF trial, participants were randomly assigned to recieve rivaroxaban 20 mg daily or warfarin. In the final trial included in figure 1—ARISTOTLE—participants were randomly allocated to apixaban 5 mg twice daily or 8
2300 mg should not be exceeded by any age group (evidence level B).115,116 Healthy balanced diet • Eat a diet low in saturated fat, cholesterol, and sodium; and high in fresh fruits, vegetables, low-fat dairy products, dietary and soluble fibre, whole grains, and protein from plant sources (evidence level B).116 Levels of evidence • Level A=evidence from meta-analyses of randomised controlled trials or directly from randomised controlled trials. Desirable effects clearly outweigh undesirable effects, or vice versa. • Level B=single randomised controlled trial or well-designed observational study with strong evidence; or well-designed cohort or case–control analytical study; or several time series or dramatic results of an uncontrolled experiment. Desirable effects closely balanced with undesirable effects. • Level C=at least one well-designed, non-experimental descriptive study (eg, comparative studies, correlation studies, case studies) or expert committee reports, opinions and/or experience of respected authorities, including consensus from development and/or reviewer groups.
warfarin. Since the RE-LY trial compared two treatments (dabigatran 110 mg vs 150 mg twice daily) against one control (warfarin) group, the effect of dabigatran 110 mg twice daily versus warfarin and the effect of dabigatran 150 mg twice daily are not independent of each other. This fact needed to be taken into account in computation of the variance. To calculate a summary effect across all three studies (RE-LY, ROCKET AF, and ARISTOTLE), the weighted mean of the effects of dabigatran 110 mg compared with warfarin and of dabigatran 150 mg versus warfarin and the variance were computed, taking into account the correlations among the different treatments. Efforts are continuing to develop rapid, accurate, and widely available measures of the anticoagulant effects of the novel oral anticoagulants (eg, the Hemoclot test for dabigatran and anti-factor Xa assays for rivaroxaban and apixaban) and reversal strategies for major bleeding. Clinical trials are assessing an antidote for dabigatran (aDabi-Fab), which is a humanised antibody fragment that binds dabigatran more strongly than its affinity for thrombin and rapidly reverses its anticoagulant effects in vitro and in vivo.125 r-Antidote (PRT064445) is a recombinant protein in development that binds factor Xa inhibitors and reverses the anticoagulant effects of apixaban and rivaroxaban.126 For patients with atrial fibrillation who have had a stroke but in whom oral anticoagulation is contraindicated, the left atrial appendage can be occluded by the WATCHMAN device (a self-expanding cage placed in the left atrial appendage via a transeptal approach with femoral access). In 707 patients with atrial fibrillation randomly allocated
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
to the WATCHMAN or long-term warfarin treatment, the rates of stroke, cardiovascular death, or systemic embolism were similar (RR 0·62, 95% CI 0·35–1·25) but the rates of NOAC Events Total
A
Stroke and systemic embolism 1195 55
procedure-related complications were higher with the WATCHMAN device (1·69, 1·01–3·19).127 Definitive randomised trials of this device are needed.
Warfarin Events Total
Fixed effects odds ratio (95% CI)
Fixed odds ratio (95% CI)
65
1195
0·84 (0·58–1·21)
RE-LY 150
51
1233
65
1195
0·75 (0·52–1·09)
ROCKET AF
179
3754
187
3714
0·94 (0·77–1·17)
ARISTOTLE
73
1694
98
1742
0·76 (0·56–1·03)
358
7876
350
6651
0·86 (0·74–1·00)
RE-LY 110
Total
Heterogeneity: χ2=1·7, df=2, p=0·43, I2=0% Test for overall effect: Z=1·95, p=0·05 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
B
Major bleeding
RE-LY 110
65
1195
97
1195
0·65 (0·48–0·9)
RE-LY 150
102
1233
97
1195
1·02 (0·76–1·36)
ROCKET AF
178
3754
183
3714
0·96 (0·78–1·19)
ARISTOTLE
77
1694
106
1742
0·74 (0·55–0·99)
422
7876
386
6651
0·86 (0·75–1·00)
Total
Heterogeneity: χ2=2·13, df=2, p=0·34, I2=6% Test for overall effect: Z=1·97, p=0·05 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
C
Intracranial bleeding 6
1195
30
1195
0·26 (0·13–0·5)
RE-LY 150
13
1233
30
1195
0·43 (0·24–0·79)
ROCKET AF
34
3754
46
3714
0·73 (0·47–1·13)
ARISTOTLE
15
1694
41
1742
0·40 (0·24–0·68)
Total
68
7876
117
6651
0·45 (0·26–0·78)
RE-LY 110
Heterogeneity: χ2=6·41, df=2, p=0·04, I2=69% Test for overall effect: Z=2·83, p=0·005 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
D
Major gastrointestinal bleeding 1195 33
33
1195
1·00 (0·61–1·63)
RE-LY 150
57
1233
33
1195
1·68 (1·11–2·57)
ROCKET AF
0
0
0
0
ARISTOTLE
18
1694
22
1742
0·84 (0·45–1·57)
108
4122
55
2937
1·18 (0·85–1·63)
RE-LY 110
Total
Not estimable
Heterogeneity: χ2=1·53, df=1, p=0·22, I2=35% Test for overall effect: Z=0·97, p=0·33 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
E
Death from any cause 77
1195
107
1195
0·70 (0·52–0·95)
RE-LY 150
108
1233
107
1195
0·98 (0·74–1·29)
ROCKET AF
288
3754
294
3714
0·97 (0·82–1·14)
ARISTOTLE
129
1694
150
1742
0·88 (0·69–1·12)
Total
602
7876
551
6651
0·91 (0·81–1·03)
RE-LY 110
Heterogeneity: χ2=1·02, df=2, p=0·60, I2=0% Test for overall effect: Z=1·48, p=0·14 RE-LY 110 and RE-LY 150 have been combined to find the pooled effect
0·2
0·5
0·7
Favours NOAC
1
1·5
2
Favours warfarin
Figure 1: Forest plots showing meta-analyses of randomised controlled trials of the effects of NOACs versus warfarin on several outcomes The trials assessed the effects of NOACs versus warfarin on (A) stroke or systemic embolism, (B) major bleeding, (C) intracranial bleeding, (D) major gastrointestinal bleeding, and (E) death from any cause. Participants were patients with atrial fibrilliation and previous stroke or transient ischaemic attack, enrolled in the RE-LY,98,99 ROCKET AF,103,104 and ARISTOTLE100,102 trials.39 A fixed effects model was used to pool the data for all outcomes because no substantial heterogeneity occurred between the trials, except for intracranial haemorrhage (C), in which a random effects model was used because of the substantial heterogeneity between the trials (I²=69%; appendix pp 1–3).123 df=degrees of freedom. NOAC=novel direct oral anticoagulant.
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
9
Review
Ischaemic stroke from paradoxical embolism Patients with cryptogenic ischaemic stroke or TIA and a patent foramen ovale have a similar rate of recurrent ischaemic stroke (1·6% per year, 95% CI 1·1–2·1%) as patients without a patent foramen ovale (RR 1·1, 95% CI 0·8–1·5).128 However, an additional atrial septal aneurysm increases the risk of recurrent stroke (HR for both patent foramen ovale and atrial septal aneurysm vs neither=4·2, 95% CI 1·5–12).129 Possible strategies to reduce recurrent stroke from paradoxical embolism include antiplatelet drugs, anticoagulation, and percutaneous closure of the patent foramen ovale with a device. Three randomised controlled trials comparing patent foramen ovale closure with medical therapy showed that the device was implanted successfully in 81–92%130–132 of patients who underwent patent foramen ovale closure. Overall, patent foramen ovale closure was associated with a non-significant reduction in the primary outcome and recurrent stroke (figure 2) and a significant increase in new-onset atrial fibrillation (3·8% closure vs 1·0% medical; RR 3·5, 95% CI 1·5–8·3).130–133 The frequency of atrial fibrillation was increased by eight times with the STARFlex device in CLOSURE (5·7% vs 0·7%)130 and by two-to-three times with the Amplatzer device in RESPECT (3·0% vs 1·5%)131 and the PC trial (2·9% vs 1·0%).132 The trial results are compromised by treatment crossovers and short and incomplete follow-up over 2–4 years. Three randomised controlled trials of patent foramen ovale closure are PFO closure Events Total
A
Medical therapy Events Total
ongoing (CLOSE [NCT00562289], DEFENSEPFO [NCT01550588], and REDUCE [NCT00738894]).
Lowering of blood pressure to prevent all types of recurrent stroke Sustained lowering of blood pressure by 5·1 mm Hg systolic and 2·5 mm Hg diastolic reduces recurrent stroke by about a fifth (table 1).40 Larger reductions in blood pressure—by 10 mm Hg systolic and 5 mm Hg diastolic— are associated with larger reductions in recurrent stroke of about a third (RR 0·66, 95% CI 0·56–0·79).41 The effect is consistent, irrespective of previous hypertension and most subtypes of stroke. For patients who have survived for more than 2 weeks after lacunar stroke, gradual targeting of systolic blood pressure to lower than 130 mm Hg seems to be safe and shows a non-significant reduction in the rate of recurrent stroke by about a fifth compared with a target systolic blood pressure of 130–149 mm Hg (HR 0·81, 95% CI 0·64–1·03).108 Caution is needed in older patients and those with severe, bilateral carotid or vertebrobasilar occlusive disease. Blood pressure can be reduced by lifestyle interventions, such as regular physical exercise for 30 min daily, alcohol reduction, a low-salt diet, and increased potassium intake, and by medication with any class of antihypertensive drugs.41,115,134,135 In addition to reducing mean blood pressure, reducing visit-to-visit variability in systolic blood pressure, by means of calcium-channel blockers and diuretics, rather than β blockers, is also likely to reduce stroke risk.41,107,136 Peto fixed effects odds ratio (95% CI)
Weight
Peto fixed odds ratio (95% CI)
Primary outcome 23
447
29
462
54·1%
0·81 (0·46–1·42)
RESPECT
9
499
16
481
26·9%
0·54 (0·25–1·20)
PC trial
7
204
11
210
19·0%
0·65 (0·25–1·67)
1153
100·0%
CLOSURE-1
1150
Total Total events
0·70 (0·46–1·05)
56
39
Heterogeneity: χ2=0·69, df=2, p=0·71, I2=0% Test for overall effect: Z=1·72, p=0·09 0·2
0·5
Favours PFO closure
B
1·0
2·0
5·0
Favours medical therapy
Stroke 12
447
13
462
44·5%
0·95 (0·43–2·11)
RESPECT
9
499
16
481
44·6%
0·54 (0·25–1·20)
PC trial
1
204
5
210
10·8%
0·27 (0·05–1·34)
1153
100·0%
CLOSURE-1
1150
Total Total events
0·65 (0·38–1·10)
34
22
Heterogeneity: χ2=2·26, df=2, p=0·32, I2=12% Test for overall effect: Z=1·62, p=0·11 0·02
0·1 Favours PFO closure
1·0
10·0
50·0
Favours medical therapy
Figure 2: Forest plots of a pooled analysis, with a fixed effects model, of the published intention-to-treat data from three randomised controlled trials of PFO closure The trials were of PFO closure with the STARFlex device in the CLOSURE-1 trial130 or the Amplatzer PFO occluder (St Jude Medical) in the RESPECT131 and PC132 trials, plus medical therapy versus medical therapy alone (aspirin or warfarin or both). The pooled analysis shows that PFO closure was associated with a non-significant trend towards a reduction in the primary outcome (OR 0·70, 95% CI 0·46–1·05) and recurrent stroke (OR 0·65, 95% CI 0·38–1·10). PFO=patent foramen ovale.
10
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
Review
Atherosclerotic ischaemic stroke Lowering of low-density lipoprotein (LDL) cholesterol concentration by about 1 mmol/L with statins reduces the risk of recurrent stroke by about 12% (table 1) and all stroke by about 21% (95% CI 6–33%).42 Greater reductions in LDL cholesterol concentration, by 2–3 mmol/L, are associated with greater reductions in stroke risk of 40–50%.42,109 The effect is consistent across noncardioembolic ischaemic stroke subtypes.110 However, the optimum target LDL cholesterol concentration is uncertain. Subgroup analyses raise the hypothesis that targeting of an LDL-C concentration below 1·8 mmol/L might be associated with a greater (28%) reduction in recurrent atherosclerotic stroke risk compared with a target of 2·59 mmol/L.111 This hypothesis is being tested in the Treat Stroke to Target (TST) trial (NCT01252875). A meta-analysis of 31 randomised controlled trials showed no difference in the incidence of intracerebral haemorrhage in 91 588 participants allocated statins and 91 215 controls (OR 1·08, 95% CI 0·88–1·32), and no relation between intracerebral haemorrhage risk and LDL cholesterol concentrations or reduction.137 However, the two largest trials of statins in patients with previous stroke both showed that statins were associated with a small absolute, but large relative, increase in risk of haemorrhagic stroke (67% [95% CI 9–156]) and 91% [–8 to 395], respectively).42,137 Allocation to atorvastatin was an independent predictor of risk of haemorrhagic stroke,138 but the risk of haemorrhagic stroke was not related to the degree of LDL cholesterol reduction.139 Much of the risk of haemorrhagic stroke associated with atorvastatin was in patients with a qualifying diagnosis of stroke because of small vessel disease.138 This fact negated the benefits of atorvastatin in reducing ischaemic events in patients with previous haemorrhagic small vessel disease but not in patients with ischaemic small vessel disease.110 For patients with stroke who do not achieve a low enough LDL cholesterol concentration, the addition of ezetimibe 10 mg, a cholesterol absorption inhibitor, to statin therapy could produce greater reductions in LDL cholesterol concentration than statin monotherapy.140 However, in 720 patients with familial hypercholesterolaemia, combined treatment with ezetimibe 10 mg daily and simvastatin 80 mg daily did not significantly change carotid artery intima-media thickness compared with simvastatin alone, despite significantly reducing LDL cholesterol concentrations by 16%.141 PCSK9 is a protein secreted by hepatocytes that regulates surface expression of LDL cholesterol receptors by targeting them for lysosomal degradation. Loss-of-function mutations in the PCSK9 gene produce lifelong low concentrations of LDL cholesterol and reduce cardiovascular risk. Statins enhance PCSK9 synthesis, thereby blunting their effectiveness in lowering LDL cholesterol concentrations. Human monoclonal antibodies to PCSK9, such as REGN727/SAR236553 (REGN727) and AMG 145, administered subcutaneously every 2 or 4 weeks to patients
with hypercholesterolaemia have been well tolerated; LDL-C concentrations have been lowered by 40–70%, despite statins and ezetimibe; and apolipoprotein B, triglycerides, lipoprotein (a), and high-density lipoprotein C concentrations have been favourably altered.142–144 PCSK9 inhibitors are being tested in phase 3 trials. Raising of high-density lipoprotein (HDL) cholesterol concentrations, by inhibiting cholesteryl ester transfer protein with dalcetrapib, does not reduce recurrent cardiovascular events in patients with acute coronary syndrome,145 but its effect in patients with ischaemic stroke is uncertain. The addition of extended-release niacin (1500–2000 mg per day) to simvastatin (40–80 mg per day), plus ezetimibe (10 mg per day) if needed, increased median HDL cholesterol concentration (from 0·91 to 1·08 mmol/L) and lowered LDL cholesterol concentration (from 1·91 to 1·60 mmol/L), but had no significant effect on ischaemic stroke (HR 1·78, 95% CI 1·00–3·17) or the primary outcome of major vascular events (1·02, 0·87–1·21) compared with placebo in 3414 patients with atherosclerotic cardiovascular disease and LDL cholesterol concentration lower than 1·81 mmol/L.146 The addition of extended-release niacin 2 g plus laropiprant 40 mg daily to simvastatin 40 mg plus, if needed, ezetimibe 10 mg daily, had no effect on major vascular events compared with placebo (RR 0·96, 95% CI 0·90–1·03) but significantly increased diabetic complications, new-onset diabetes, infection, bleeding, haemorrhagic stroke, and myopathy after 3·9 years’ (median) follow-up in 25 673 patients with occlusive arterial disease.147,148 A meta-analysis of eight randomised controlled trials in patients with type 2 diabetes showed that, compared with a standard regimen (n=15 844), intensive treatment with oral hypoglycaemic drugs or insulin to reduce blood glucose concentration (n=17 742) over a mean of 5 years was associated with no reduction in stroke (RR 0·96, 99% CI 0·83–1·13; I²=0%) or mortality (RR 1·04, 99% CI 0·91–1·19; I²=42%).149 Although intensive treatment reduced non-fatal myocardial infarction (RR 0·85, 99% CI 0·74–0·96) and microalbuminuria (0·90, 0·85–0·96), it increased severe hypoglycaemia (2·33, 1·62–3·36). More randomised controlled trials are needed to establish the best approach for blood glucose management in people with type 2 diabetes. Lowering total plasma homocysteine with folic acid, vitamin B6, and vitamin B12 does not prevent recurrent stroke (RR 0·92, 95% CI 0·81–1·06) or any stroke (0·91, 0·82–1·00, I²=11%) compared with placebo in populations with established or increasing intake of folate and vitamin B12.150 However, a reduction in total plasma homocysteine might effectively prevent stroke in people with a low intake of folate or vitamin B12 (eg, Asian people),151 intracranial small vessel disease,152 and those not taking or who are unable to take antiplatelet therapies, such as aspirin.153
www.thelancet.com/neurology Published online December 20, 2013 http://dx.doi.org/10.1016/S1474-4422(13)70255-2
11
Review
Dietary supplementation with the antioxidant vitamins β-carotene, vitamin C, and vitamin E does not prevent stroke.154,155 Whether or not vitamin D supplementation can prevent stroke is uncertain.154 A meta-analysis of 18 randomised controlled trials of hormone replacement therapy involving 36 523 women showed that this treatment increased the risk of stroke by a third (OR 1·32, 95% CI 1·14–1·53) compared with controls.156 A subsequent smaller trial showed that hormone replacement therapy, if started early and used for more than 10 years, did not increase stroke (14 strokes in the control group vs 11 strokes in the hormone replacement therapy group; HR 0·77, 95% CI 0·35–1·70),157 but this study was not powered to reliably exclude the one-third increased relative risk of stroke in the earlier meta-analysis of randomised controlled trials.156 Epidemiological data suggest that a substantial proportion of strokes can be attributed to unhealthy lifestyle behaviours (table 4).112 Cessation of smoking and alcohol abuse and adherence to careful diets that are aligned to the Mediterranean or dietary approaches to stop hypertension, and are low in salt and added sugars, high in potassium, and meet, but do not exceed, energy requirements, are associated with reduced risks of stroke.112,116
Translation of evidence into practice Convincing evidence outlined in this Review suggests that several interventions are efficacous in reducing recurrent stroke under ideal circumstances (ie, explanatory clinical trial settings). The combination of five of these strategies— aspirin, an antihypertensive drug, a statin, exercise, and dietary modification—could reduce recurrent stroke by 80%.158 However, evidence exists that these interventions
Prevalence (%)
Odds ratio (99% CI)
Population-attributable risk (99% CI)
Controls
Cases
History of hypertension
32%
56%
2·6 (2·3–3·1)
35% (30–39)
Regular physical activity
12%
8%
0·7 (0·5–0·9)
28% (14–48)
Waist-to-hip ratio (T3 vs T1)
41%
33%
1·6 (1·4–2·0)
26% (19–36)
Ratio of apolipoprotein B to apolipoprotein A-I (T3 vs T1)
33%
47%
1·9 (1·5–2·4)
25% (16–37)
Current smokers
24%
36%
2·1 (1·7–2·5)
19% (15–23)
Diet risk score (T3 vs T1)
35%
30%
1·4 (1·1–1·6)
19% (11–30)
Cardiac causes*
5%
12%
2·4 (2·8–3·2)
7% (5–9)
Diabetes
12%
19%
1·4 (1·1–1·7)
5% (3–9)
Depression
14%
20%
1·4 (1·1–1·6)
5% (3–10)
Psychosocial stress
15%
20%
1·3 (1·1–1·6)
5% (2–10)
Alcohol intake >30 drinks per month
11%
16%
1·5 (1·2–1·9)
4% (1–14)
Control participants were matched with cases for age and sex, and were assessed in 22 countries between 2007 and 2010. The multivariate model was adjusted for age, sex, and region. T3=highest tertile. T1=lowest tertile. *Cardiac causes include atrial fibrillation or flutter, previous myocardial infarction, rheumatic valve disease, or prosthetic valve disease.
Table 4: Risk factors for all types of stroke in 3000 cases of acute first stroke compared with 3000 controls with no history of stroke in the INTERSTROKE study112
12
are effective (ie, they do actually work, and produce more good than harm when used under normal circumstances in clinical practice) and are cost effective (ie, their effects in relation to the resources they consume are worth it) is less convincing. Several studies have shown that the effectiveness of secondary stroke prevention is insufficient because of the widespread failure to appropriately translate evidence-based recommendations into clinical practice.159–165 The most recent global study reported that in 2292 adults with previous stroke who were living in 628 urban and rural communities in three high-income, seven upper-middle-income, three lower-middleincome, and four low-income countries, few took antiplatelet drugs (24·3%), blood pressure-lowering drugs (40·0%), or statins (9·0%).164 Use of these drugs was highest in high-income countries (antiplatelet drugs 53·1%, blood pressure-lowering drugs 60·6%, and statins 51·6%), lowest in low-income countries (3·8%, 13·0%, and 0·6%, respectively), and decreased with reduction in country economic status (p value for trend
