THE LANCET
Articles
The International Stroke Trial (IST): a randomised trial of aspirin, subcutaneous heparin, both, or neither among 19 435 patients with acute ischaemic stroke International Stroke Trial Collaborative Group*
Summary Background Only a few small trials have compared antithrombotic therapy (antiplatelet or anticoagulant agents) versus control in acute ischaemic stroke, and none has been large enough to provide reliable evidence on safety or efficacy. Methods The International Stroke Trial (IST) was a large, randomised, open trial of up to 14 days of antithrombotic therapy star ted as soon as possible after stroke onset. The aim was to provide reliable evidence on the safety and efficacy of aspirin and of subcutaneous heparin. Half the patients were allocated unfractionated heparin (5000 or 12 500 IU bd [twice daily]) and half were allocated “avoid heparin”; and, in a factorial design, half were allocated aspirin 300 mg daily and half “avoid aspirin”. The primar y outcomes were death within 14 days and death or dependency at 6 months. 19 435 patients with suspected acute ischaemic stroke entering 467 hospitals in 36 countries were randomised within 48 hours of symptom onset. Results Among heparin-allocated patients, there were non-significantly fewer deaths within 14 days (876 [9·0%] heparin vs 905 [9·3%] no heparin), corresponding to 3 (SD 4) fewer deaths per 1000 patients. At 6 months the percentage dead or dependent was identical in both groups (62·9%). Patients allocated to heparin had significantly fewer recurrent ischaemic strokes within 14 days (2·9% vs 3·8%) but this was offset by a similar-sized increase in haemorrhagic strokes (1·2% vs 0·4%), so the difference in death or non-fatal recurrent stroke (11·7% vs 12·0%) was not significant. Heparin was associated with a significant ex cess of 9 (SD 1) transfused or fatal ex tracranial bleeds per 1000. Compared with 5000 IU bd heparin, 12 500 IU bd heparin was associated with significantly more transfused or fatal ex tracranial bleeds, more haemorrhagic strokes, and more deaths or non-fatal strokes within 14 days (12·6% vs 10·8%). Among aspirinallocated patients there were non-significantly fewer deaths within 14 days (872 [9·0%] vs 909 [9·4%]), corresponding to 4 (SD 4) fewer deaths per 1000 patients. At 6 months there was a non-significant trend towards a smaller percentage of the aspirin group being dead or dependent (62·2% vs 63·5%, 2p=0·07), a difference of 13 (SD 7) per 1000; after adjustment for baseline prognosis the benefit from aspirin was *Details of members of the collaborating group, writing committee, and sources of suppor t are given in the acknowledgment section
Correspondence to: Dr PAG Sandercock, Depar tment of Clinical Neurosciences, Western General Hospital, Edinburgh EH4 2XU, UK (e-mail:
[email protected])
Vol 349 • May 31, 1997
significant (14 [SD 6] per 1000, 2p=0·03). Aspirinallocated patients had significantly fewer recurrent ischaemic strokes within 14 days (2·8% vs 3·9%) with no significant ex cess of haemorrhagic strokes (0·9% vs 0·8%), so the reduction in death or non-fatal recurrent stroke with aspirin (11·3% vs 12·4%) was significant. Aspirin was associated with a significant ex cess of 5 (SD 1) transfused or fatal ex tracranial bleeds per 1000; in the absence of heparin the ex cess was 2 (SD 1) and was not significant. There was no interaction between aspirin and heparin in the main outcomes. Interpretation Neither heparin regimen offered any clinical advantage at 6 months. The results suggest that if heparin is given in routine clinical practice, the dose should not ex ceed 5000 IU subcutaneously twice daily. For aspirin, the IST suggests a small but wor thwhile improvement at 6 months. Taking the IST together with the comparably large Chinese Acute Stroke Trial, aspirin produces a small but real reduction of about 10 deaths or recurrent strokes per 1000 during the first few weeks. Both trials suggest that aspirin should be star ted as soon as possible after the onset of ischaemic stroke; previous trials have already shown that continuation of low-dose aspirin gives protection in the longer term.
Lancet 1997; 349: 1569–81 See Commentar y page 1564
Introduction Every year, several million people worldwide are treated for acute ischaemic stroke1. If some widely practicable therapies could be reliably shown to prevent death or dependence for “just” 10 or 20 of every 1000 patients, it would, for every million stroke patients so treated, ensure that an extra 10 000 were alive and independent. If such benefits exist, they must not, therefore, be overlooked. Reliable assessment of them may, however, require randomised trials with tens of thousands of patients.2,3 Two such candidates are heparin and aspirin. Most strokes are caused by acute occlusion of a cerebral artery. Anticoagulants are widely used,4-6 to facilitate early clot lysis, to inhibit clot propagation in the cerebral arteries, and to prevent early arterial re-embolisation and venous thromboembolism originating in immobile limbs.2 However, there is little randomised evidence on the balance of risks and benefits of heparin in acute ischaemic stroke. Low-dose regimens of around 5000 IU twice daily of subcutaneous unfractionated heparin (or equivalent low-molecularweight heparins or heparinoids) do reduce the risk of deep venous thrombosis,2,7 but the effects of these and higher-dose regimens on the cerebral arterial circulation and on the risk of intracranial haemorrhage are 1569
THE LANCET
unclear.2,7 A medium-dose regimen of 12 500 IU subcutaneous unfractionated heparin twice daily prevents left-ventricular wall thrombus in acute myocardial infarction (AMI),8,9 and proved feasible in two large trials,10,11 but has not been tested in acute stroke.2,7 Antiplatelet drugs such as aspirin can be effective in the secondary prevention of “serious vascular events” (stroke, AMI, and vascular death).12 Taken for a few years after a myocardial infarction, ischaemic stroke, or transient ischaemic attack (TIA), antiplatelet therapy typically avoids about 40 serious vascular events per 1000 patients treated.12 It also reduces the incidence of deep venous thrombosis and pulmonary embolism in high risk patients,13 and is effective in the treatment of AMI, preventing about 40 serious vascular events per 1000 patients treated for just one month.12 In acute ischaemic stroke there is substantial platelet activation, which can be inhibited by aspirin.14,15 But there is no large-scale randomised evidence on aspirin in acute ischaemic stroke.2,16 Aspirin and heparin have different mechanisms of action, so the combination might be more effective than either alone, although this might be offset by a greater risk of bleeding. In AMI, the combination of aspirin and heparin was not significantly more effective than aspirin alone,17,18 but no trials have made this comparison in acute stroke.2,7,16 Given the uncertain balance of risk and benefit for heparin and for aspirin in acute stroke,2,7,12,16 the uncertainty of many physicians about the safety and efficacy of these drugs in this setting4-6 and the wide variation in clinical practice,4-6 the IST was designed to assess the separate and combined effects of subcutaneous heparin (in twice daily doses of 5000 IU or 12 500 IU) and of aspirin (300 mg daily). Large numbers of patients were included in order to provide a reliable estimate of their effects on death and other major clinical events during the first 14 days after acute ischaemic stroke, and on death and dependency in activities of daily living at 6 months, as well as any adverse effects on intracranial haemorrhage and on transfused (or fatal) extracranial bleeds.
Methods Patients Eligibility A patient was eligible if, in the view of the responsible physician, there was evidence of an acute stroke (irrespective of severity) with onset less than 48 h previously, no evidence of intracranial haemorrhage (see below), and no clear indications for, or contraindications to, heparin or aspirin. The fundamental criterion for eligibility was simply that the physician was uncertain whether or not to administer either or both of the trial treatments to that particular patient. Ex clusions Possible reasons not to include a patient were either only a small likelihood of worthwhile benefit (eg, the symptoms seemed likely to resolve completely within a few hours or the patient was severely disabled before the stroke) or a high risk of adverse effects (eg, hypersensitivity to aspirin, active peptic ulceration or recent gastrointestinal bleeding; or already on long-term oral anticoagulants). Computed tomographic (CT) scanning All patients were to be CT scanned to exclude intracranial haemorrhage, before randomisation where possible, and in comatose patients a CT was mandatory. However, if there was likely to be a long delay in getting the CT scan and if, on clinical grounds, the physician considered the stroke very likely to be ischaemic, a non-
1570
comatose patient could be randomised before CT. For those allocated active treatment, the initial doses could be given while the CT was being arranged, but treatment was stopped if intracranial haemorrhage was found (which it rarely was). Classification and predicted prognosis Neurological deficits recorded at telephone randomisation were used to classify the infarct as: total anterior circulation, partial anterior circulation, posterior circulation, lacunar, or other, by means of a computer algorithm.19 Prognosis was predicted by a validated model,20 using baseline data to calculate the probability of a poor outcome (see statistical methods).
Planned inter ventions and their timing After informed consent had been obtained (the procedure varied to conform to local ethical committee requirements), patients were entered by telephoning the central randomisation service at the Clinical Trial Service Unit (CTSU), Oxford, UK. After the baseline data had been entered and checked for range and consistency, the computer allocated the study treatment(s) (a minimisation algorithm was used to reduce any imbalance in recorded prognostic features between treatment groups).21 At that point the patient was irrevocably included in the study. Clinicians could depart from the randomly allocated treatment if a clear reason to do so had arisen, but they could not withdraw a patient from follow-up even if the patient turned out not to have had a stroke or did not receive the treatment allocated. Heparin Half of the patients were randomly allocated to receive subcutaneous unfractionated heparin (one-quarter 5000 IU twice daily [low-dose] and one-quarter 12 500 IU twice daily [medium-dose]); and half were allocated to “avoid heparin”. Heparin was supplied by the hospital, and sodium heparin could be used if calcium heparin was not available. In more than 30 000 patients with AMI (most of whom also received fibrinolytic therapy and aspirin) 12 500 IU twice daily had been fairly safe without routine monitoring of coagulation times,10,11 so such monitoring was optional in the IST. Aspirin Using a factorial design, half of all patients were allocated to 300 mg aspirin daily and half to “avoid aspirin”. If a patient could not swallow safely, aspirin was to be given by nasogastric tube, per rectum (300 mg suppository), or intravenously (100 mg lysine salt in 100 mL normal saline over 60 min). Aspirin was supplied by each hospital. Patients could continue any prescribed nonsteroidal antiinflammatory drugs, but analgesics containing aspirin were to be avoided in patients allocated to “avoid aspirin”, unless the physician considered that a clear indication for them had developed. Timing Patients allocated active treatment (heparin, or aspirin, or both) were to receive the first dose(s) immediately after randomisation, and treatment was to continue for 14 days or until prior discharge. At discharge, clinicians were to consider giving all patients long-term aspirin.12 All other aspects of treatment were determined by the responsible clinician.
Events and outcomes Events in hospital A single-sided form was to be completed from the patient’s medical records at 14 days or at hospital discharge or death, if sooner. This included contact details, final diagnosis of the index event, compliance with allocated treatment, use of other drugs, nature and timing of any clinical events in hospital, including the likely cause of death, and aspirin use at discharge. Patients allocated active treatment were considered compliant if they missed no more than two of their scheduled doses. Patients allocated to avoid heparin or to avoid aspirin were considered non-compliant if they received any heparin or any antiplatelet therapy respectively in hospital within 14 days of randomisation. Outcome at 6 months (death, dependency, and incomplete recover y) In a few countries, the randomising physician collected 6-month follow-up data in an outpatient clinic. Elsewhere, the country coordinating centre mailed a validated questionnaire22,23 to the patient (or relative or other proxy) or
Vol 349 • May 31, 1997
THE LANCET
Patients randomised (n=19 435) Allocations Aspirin 300 + Heparin 12 500
Aspirin 300 + Heparin 5000
Aspirin 300 + No Heparin
No aspirin + Heparin 12 500
No aspirin + Heparin 5000
No aspirin + No Heparin
2430
2432
4858
2426
2429
4860
2426 (100·0%)
2429 (100·0%)
4859 (99·99%)
No (% randomised) with mortality follow-up 14 days 2430 (100·0%)
2431 (99·99%)
4858 (100·0%)
(0-3, 3-6, 6-12, 12-24, 24-48 h); and whether any effects of aspirin and of heparin were additive. These comparisons were also to be made for the secondary outcomes. It was prospectively planned to consider the IST results with those of the only comparably large trial, the 20 000 patient Chinese Acute Stroke Trial of aspirin.24
Statistical methods
Most comparisons involved simple analyses of total numbers of patients 6 months affected.12 Proportional reductions were 2413 2410 4816 2411 2407 4828 expressed as odds reductions (with 95% (99·3%) (99·1%) (99·1%) (99·4%) (99·1%) (99·3%) CI for main analyses and 99% CI for subgroup analyses). Absolute Figure 1: Trial randomisation flow chart differences were calculated as benefits per 1000 patients treated, and were generally given together with their SD. interviewed them on the telephone. The different methods of Estimates of treatment effects adjusted for severity of initial follow-up were valid.3 Where it was done centrally, follow-up stroke were calculated as follows. The prognosis was estimated was blind to treatment allocation. using a model which predicted the probability of death or Protocol-specified primar y outcomes These were: (a) death dependency at 6 months from data recorded at randomisation.20 from any cause within 14 days and (b) death or dependency (ie, For each stratum of severity (see footnote, figure 2) the needing help from another person with daily activities) at 6 observed (O) minus expected (E) value and its variance (V) months.22,23 were calculated and these were summed to give the overall Protocol-specified secondar y outcomes These were: (a) (O⫺E) and V. The statistic k was defined as 1000(T + C)/(TC), symptomatic intracranial haemorrhage (haemorrhagic stroke) where T and C are the total numbers of treated and control within 14 days (including any recurrent stroke definitely due to patients: both for heparin and for aspirin k = 0.207. The haemorrhage or symptomatic haemorrhagic transformation of adjusted benefit per 1000 is then k(O⫺E), with standard the original infarct) that was confirmed by CT, magnetic deviation k公僒 V. resonance imaging, or necropsy; (b) ischaemic stroke within 14 Two-sided p values (2p) are cited throughout. days (including any recurrent stroke of ischaemic or unknown The independent data monitoring committee reviewed type); (c) major extracranial haemorrhage (including any bleed interim analyses of major outcome events about once a year that required transfusion or caused death within 14 days); (d) during the recruitment phase, and allowed the study to run its other major events within 14 days such as pulmonary full course. embolism; (e) death from any cause by 6 months; and (f) dependence in activities of daily living, or incomplete recovery from the stroke, at 6 months. Only events in hospital within 14 days of randomisation in patients who went on to survive at least 14 days were considered “non-fatal”. Characteristic Characteristic Planned study size The protocol considered, as an example, the Delay (h from No (%) Systolic BP (mm Hg) No (%) situation where 10% would die without treatment and where symptoms) 180 5408 (28%) 13–24 5568 (29%) conventionally significant (eg, 440 vs 490 deaths). This chance Stroke syndrome 25–48 6588 (34%) of failing to recognise a treatment that really does reduce Total anterior 4638 (24%) Age (yr) mortality by 15% did not seem a reasonable risk to take so the Partial anterior 7912 (40%) 80 5125 (26%) Absent 4502 (23%) the most important complication is fatal or disabling Sex Not assessable 255 (1%) haemorrhagic transformation, the study “must have sufficient F 9028 (46%) CT scan statistical power to detect even moderate increases in this rare M 10 407 (54%) Before randomisation 13 020 (67%) but often serious event”. Onset
Analyses Main analyses The protocol specified two main analyses for the primary outcomes—namely, “immediate heparin” (low or medium dose) vs “avoid heparin” and “immediate aspirin” vs “avoid aspirin”. Subsidiar y analyses For the primary outcomes, the main subsidiary analyses were: medium vs low dose heparin; aspirin vs “avoid aspirin”, subdivided by aspirin use in the previous 3 days, recorded at randomisation; heparin vs “avoid heparin”, subdivided by heparin use in the previous 24 h, recorded at randomisation; subdivision of the main analyses by whether CT was first done before or after randomisation and by the number of hours between symptom onset and randomisation
Vol 349 • May 31, 1997
Awake During sleep
13 750 (71%) 5685 (29%)
Conscious level Unconscious Drowsy Alert
260 (1%) 4254 (22%) 14 921 (77%)
Cardiac rhythm Sinus rhythm, not AF* Atrial fibrillation (AF)
16 266 (84%) 3169 (16%)
After randomisation Not done/not known
5569 (29%) 846 (4%)
Appearance of prerandomisation CT Infarct visible No infarct visible
6415 (49%) 6605 (51%)
Pre-randomisation antithrombotic therapy Aspirin within previous 3 days Heparin within previous 24 hours
3940 (20%) 436 (2%)
*Information on presence of atrial fibrillation at entr y not sought during pilot phase, so cardiac rhythm at entry not known for 984 patients.
Table 1: Characteristics at randomisation for all treatment groups combined
1571
THE LANCET
Results Recruitment and follow-up In the pilot phase 984 patients were recruited between January, 1991, and February, 1993,3 with 18 456 recruited between March, 1993, and May, 1996, in the main trial. 5 were entered in error, so no data were collected on these patients. Thus 19 435 patients were randomised, by 467 hospitals in 36 countries. Outcome data were 99.99% complete for 14-day outcome and 99.2% complete for 6-month outcome (figure 1). Characteristics of patients Large numbers and central randomisation (with minimisation21) ensured good balance between the six treatment groups for the main recorded (and, presumably, unrecorded) prognostic factors. Table 1 shows baseline characteristics for all treatment groups combined.† The median time to randomisation was 19 h (4% within 3 h, 16% within 6 h, 37% within 12 h, and 66% within 24 h). At randomisation, 61% were aged over 70, 23% had impaired consciousness and 16% were known to be in atrial fibrillation. 67% had had a CT scan before randomisation (49% of these scans showed an infarction, no infarct lesion being visible in the remainder); a further 29% were scanned after randomisation; and 4% never had a CT scan. †A
breakdown by treatment group is obtainable from The Lancet
Outcome
Fatal or non-fatal events within 14 days (lower left, table 2) Heparin-allocated patients had significantly fewer recurrent ischaemic strokes within 14 days (2·9% vs 3·8%, 2p=0·005) but this benefit was completely offset by a similar-sized increase in haemorrhagic stroke Aspirin vs no aspirin
No heparin
No randomised
9717
No with 14 day data
9716 (99·99%) 584(6·0%) 79 (0·8%) 28 (0·3%) 64 (0·7%) 36 (0·4%) 12 (0·1%) 57 (0·6%) 16 (0·2%) 876 (9·0%)
613 (6·3%) 98 (1·0%) 15 (0·2%) 79 (0·8%) 39 (0·4%) 3 (0·0%) 44 (0·5%) 14 (0·1%) 905 (9·3%)
283 (2·9%) 120 (1·2%) 396 (4·1%) 1136 (11·7%) 53 (0·5%) 129 (1·3%)
370 (3·8%) 41 (0·4%) 411 (4·2%) 1171 (12·0%) 81 (0·8%) 37 (0·4%)
Fatal and non-fatal events Recurrent ischaemic stroke Haemorrhagic stroke (HS) Recurrent ischaemic stroke or HS Death or non-fatal stroke Pulmonar y embolism Transfused or fatal extracranial haemorrhage
Heparin versus avoid heparin Deaths within 14 days (upper left, table 2) Among heparin-allocated patients there were non-significantly fewer deaths within 14 days (9·0% vs 9·3%, corresponding to an absolute reduction of 3 [SD 4] per 1000 patients). When causes of death were examined separately, the only significant difference was the increased numbers, with heparin, of deaths attributed to haemorrhagic stroke (28 vs 15, 2p=0·04) and to extracranial bleeding (12 vs 3, 2p=0·02).
Heparin vs no heparin Heparin
Deaths and likely causes Initial stroke Recurrent ischaemic stroke Haemorrhagic stroke Coronar y heart disease Pulmonar y embolism Extracranial haemorrhage Other vascular Non-vascular Total (any cause)
Compliance Compliance was good. Low-dose heparin was received throughout the scheduled treatment period by 90% and medium-dose heparin by 88% of those allocated it; no heparin was received by 94% of those allocated to avoid it. Aspirin was taken throughout the scheduled treatment period by 92%, and no antiplatelet therapy was taken by 93% of those allocated to avoid it. For compliant patients, the mean duration of active treatment was 11 days both for heparin and for aspirin. The chief reason for stopping treatment before 14 days was early discharge from hospital.
Events prevented per 1000 (SD)
Aspirin
No aspirin
9718
9720
9715
9717 (99·99%)
9719 (99·99%)
9714 (99·99%)
592 (6·1%) 88 (0·9%) 23 (0·2%) 72 (0·7%) 31 (0·3%) 9 (0·1%) 40 (0·4%) 17 (0·2%) 872 (9·0%)
605 (6·2%) 89 (0·9%) 20 (0·2%) 71(0·7%) 44(0·5%) 6 (0·1%) 61 (0·6%) 13 (0·1%) 909 (9·4%)
1 (3) 0 (1) 0 (1) 0 (1) 1 (1) 0 (0) 2 (1)* 0 (1) 4 (4)
275 (2·8%) 87 (0·9%) 361 (3·7%) 1099 (11·3%) 57 (0·6%) 109 (1·1%)
378 (3·9%) 74 (0·8%) 446 (4·6%) 1208 (12·4%) 77 (0·8%) 57 (0·6%)
11 (3)*** ⫺1 (1) 9 (3)** 11 (5)* 2 (1) ⫺5 (1)•••
⫺3 (3) ⫺2 (1) ⫺1 (1)* ⫺2 (1) ⫺0 (1) ⫺1 (0)* ⫺1 (1) ⫺0 (1) ⫺3 (4) ⫺9 (3)** ⫺8 (1)**** ⫺2 (3) ⫺4 (5) ⫺3 (1)* ⫺9 (1)****
Events prevented per 1000 (SD)
HS=haemorrhagic stroke (ie, symptomatic intracranial haemorrhage or symptomatic haemorrhagic transformation or infarct) confirmed by CT scan, MRI, or necropsy. Negative numbers indicate that more events of this type occurred in patients allocated to receive active treatment. SD=standard deviation. *2p
