Head To Head

Do risks outweigh benefits in thrombolysis for stroke?

BMJ 2013; 347 doi: http://dx.doi.org/10.1136/bmj.f5215 (Published 29 August 2013) Cite this as: BMJ 2013;347:f5215
  1. Simon G A Brown, professor in emergency medicine12,
  2. Stephen P J Macdonald, associate professor in emergency medicine13,
  3. Graeme J Hankey, Winthrop professor of neurology45
  1. 1School of Primary, Aboriginal and Rural Health Care, University of Western Australia, Perth, Australia
  2. 2Emergency Department, Royal Perth Hospital, Perth, WA 6847, Australia
  3. 3Emergency Department, Armadale Health Service, Armadale, WA 6992, Australia
  4. 4School of Medicine and Pharmacology, University of Western Australia
  5. 5Department of Neurology, Sir Charles Gairdner Hospital, Nedlands, Perth 6009, Australia
  1. Correspondence to: S G A Brown simon.brown{at}uwa.edu.au, G J Hankey graeme.hankey{at}uwa.edu.au

Simon Brown and Stephen Macdonald argue that patients with stroke should not be given thrombolysis outside clinical trials, but Graeme Hankey says the benefits are clear in carefully selected patients

Yes—Simon G A Brown and Stephen P J Macdonald

Emergency physicians are strong advocates of thrombolysis for myocardial infarction because a series of studies in large numbers of patients have clearly and consistently shown that the benefits outweigh the risks. Thrombolysis for stroke does not receive the same unanimous support because the risks are higher and the evidence of benefit is not yet convincing. Of 12 controlled trials on the use of alteplase (recombinant tissue plasminogen activator) for stroke, only two found a benefit as defined by primary outcome measures.1 2 Two were stopped early because of harm,3 4 and the remaining studies had negative findings. This pattern is typical for a treatment that does not work.

Evidence of harm is clear

Randomised controlled trials have consistently found that thrombolysis for stroke is associated with a higher risk of intracranial haemorrhage and early death compared with placebo. For alteplase, excess haemorrhages and deaths in the first seven days have been calculated to be 58 per 1000 cases treated (95% confidence interval 49 to 68) and 25 (11 to 39), respectively, although by 3-6 months death rates are similar whether treated with alteplase or not.5

Overall benefit is unclear

The critical question is whether there is any long term benefit from thrombolysis that outweighs the early hazard. A systematic review has found that giving alteplase within six hours of stroke reduces the composite outcome of death or dependency by 42/1000 people treated (95% confidence interval 19 to 66), with a greater effect if given within three hours (90/1000 people treated, 46 to 135).5 Nevertheless, there is persistent concern about the positive studies informing this meta-analysis.6 7 8

Patients randomised to placebo in the National Institute of Neurological Disorders (NINDS) study,1 which provides the majority of patients treated under three hours, had more severe strokes at baseline. The results were challenged, leading to a reanalysis and statistical manipulations.9 A graphical analysis of the original data showed that the improved final outcome for alteplase patients paralleled the greater number of alteplase patients who had had a very mild stroke.10 This graphical analysis was criticised for ignoring the non-linearity of the National Institutes for Health stroke scale,11 but the problem of imbalanced severity of stroke remains.

The placebo group in the third European Cooperative Acute Stroke Study (ECASS-III)2 also had more severe strokes as well as more patients with a history of previous stroke (14.1% versus 7.7%) than the treatment group. The P value for this was originally miscalculated and dismissed as insignificant, but subsequently corrected to 0.003.12 ECASS-III also used an unusual primary outcome with mild disability (a modified Rankin scale score of 2) being grouped with dead (score of 6). If those with a score of 0-2 are compared with those with a score of 3-6, there is no significant difference between alteplase and placebo.13

The most recent study of alteplase, and the largest to date (International Stroke Trial, IST-3), has been claimed to show positive results but it found no effect on the study’s primary outcome, the proportion of people alive and independent at 6 months.14 A secondary ordinal analysis found a shift in Oxford handicap scale (OHS) score in favour of alteplase. The scores were collected by postal survey and telephone rather than objective face to face assessment, and because IST-3 was an open label study a substantial bias strongly favoured alteplase.15 16 This concern applies equally to a recently reported 18 month outcome analysis showing no net difference in mortality but improved self reported function.17

Imbalances in baseline group characteristics (NINDS and ECASS-3) and open label design (IST-3) all favour alteplase and seriously compromise the systematic reviews and pooled analyses purporting to show a benefit of earlier treatment. Post-hoc attempts to statistically correct for baseline imbalances are imperfect and subject to bias, so the only way to resolve the question is to perform further, properly designed trials.18 Large industry funded registry datasets have been used to claim a benefit with earlier alteplase treatment but are heavily confounded by selection bias and prove nothing.

Reliance on dichotomous outcome measures may conceal important treatment effects. However, alternatives such as shift analysis of disability groups (modified Rankin scale or OHS) are subjective and limited by inter-observer reliability.19 Furthermore, if a trial is adequately powered it is implausible for significant improvements in neurological function not to affect a simple objective primary outcome of death or dependency.

Too many uncertainties

We can say for certain that thrombolysis harms some patients early, including some who might otherwise have made a good recovery. However, the quality of evidence for benefit is poor. Stroke, like the brain, is a complex entity—much more so than myocardial infarction. Effective treatment requires more than trying to open a blocked artery.

Patients need treatments that are based on sound science. This is especially true in an emergency, when they are incapable of full participation in decision making that may be strongly influenced by the beliefs of the treating physician. For now, thrombolysis for stroke should be restricted to high quality, placebo controlled clinical trials aiming to resolve the question of whether, and in which patients, it provides net benefit that justifies the risk.

No—Graeme J Hankey

Thrombolysis aims to restore perfusion and function of the ischaemic brain but may cause bleeding. Although the latest systematic review shows that thrombolysis given within six hours of ischaemic stroke increases the risk of symptomatic intracranial haemorrhage and death within seven days, at final follow-up it is associated with an absolute increase in survival free of dependency and favourable outcomes without excess mortality.5 If patients are treated early, within three hours, the net benefit of thrombolysis is a 9% (95% confidence interval 4.6% to 13.5%) increase in survival free of dependency and 8.7% (4.6% to 12.8%) increase in favourable outcome.5 Various prediction models and risk scores are available to help identify those patients who may bleed and those who may benefit, and when correctly targeted, the benefits of alteplase clearly outweigh the risks.

What does the research show?

The best method of evaluating the risks and benefits of thrombolysis is a systematic review and meta-analysis of all randomised controlled trials.20 A recent updated systematic review and meta-analysis of 12 randomised trials of alteplase given within six hours of ischaemic stroke in 7012 patients indicated that in the first seven days, compared with control, alteplase significantly increased symptomatic intracranial haemorrhage (absolute increase 5.8%, 95% confidence interval 4.9% to 6.8%), fatal intracranial haemorrhage (2.9%, 2.3% to 3.6%), and death (2.5%, 1.1% to 3.9%).5 However, at final follow-up, one to six months after randomisation, alteplase significantly increased survival free of dependency (4.2%, 1.9% to 6.6%) and a favourable outcome (5.5%, 3.3% to 7.7%) on the modified Rankin scale with no significant difference in mortality (19.1% alteplase versus 18.5% control, odds ratio 1.06, 0.94 to 1.20).5 There was no significant heterogeneity between trials for any of these outcomes. The results were consistent among older and younger patients.

In the one trial of 2348 patients with long term follow-up, alteplase significantly increased survival free of handicap (35% v 31.4%; adjusted odds ratio 1.28, 1.03 to 1.57) and overall self reported health related quality of life (adjusted mean difference in EuroQoL utility index 0.06; P=0.019) at 18 months, compared with control.17

Large observational studies, such as the Safe Implementation of Thrombolysis in Stroke International Stroke Thrombolysis Register (29 618 patients, of whom 25 279 were treated within three hours21 22 23) and Canadian Alteplase for Stroke Effectiveness Study (2247 patients, 1135 treated within three hours 24 25) suggest that the efficacy of thrombolysis reported in clinical trials translates into comparable effectiveness and acceptable safety of thrombolysis in clinical practice.

Treatment with alteplase also seems to be cost effective. Compared with no thrombolysis, administration of alteplase within 4.5 hours of ischaemic stroke has been estimated to have incremental cost effectiveness ratios of $A2377 (£1400; €1600; $2150) per life year saved and $A1478 per quality adjusted life year saved.26 27

Selecting patients most likely to benefit

Thrombolysis carries serious and potentially fatal risks in a minority of patients but despite these risks the evidence shows it is clinically and cost effective in appropriately selected patients. The challenge is to reliably identify which patients are likely, and unlikely, to benefit from thrombolysis, and to minimise the risk of symptomatic intracranial haemorrhage.

Time to treatment seems to be critical. The benefit of alteplase is greatest in patients treated within three hours after stroke onset. A meta-analysis of six randomised controlled trials of alteplase given 0-3 hours after ischaemic stroke in 1779 patients indicated that, compared with control, alteplase increased survival free of dependency by 9.0% (40.7% v 31.7%; odds ratio 1.53, 95% confidence interval 1.26 to 1.86). In contrast, a meta-analysis of seven randomised controlled trials of alteplase given 3-6 hours after ischaemic stroke in 4971 patients indicated that, compared with control, alteplase increased survival free of dependency by 1.8% (47.5% v 45.7%; 1.07, 0.96 to 1.20) . The difference in the odds ratios between the subgroups treated within 3 hours and in 3-6 hours was significant (P=0.002).5 Observational data from 58 353 patients treated with alteplase within 4.5 hours of ischaemic stroke in 1395 US hospitals also show that shorter time between onset and treatment (in 15 minute increments) is associated with reduced symptomatic intracranial haemorrhage (odds ratio 0.96, 0.95 to 0.98), reduced in-hospital mortality (0.96; 0.95 to 0.98), increased independent ambulation at discharge (1.04; 1.03 to 1.05), and increased discharge home (1.03; 1.02 to 1.04).28

Observational studies have identified several other factors that predict a favourable outcome with alteplase. These include younger age, good pre-stroke function, less severe stroke, normal blood glucose or glycated haemoglobin concentration and renal function, and absence of early infarct signs or hyperdense cerebral artery on computed tomography.29 30 31 32 33 34 35 These factors are included in various prediction models and risk scores that help determine which patients are most, and least, likely to benefit from use of alteplase.29 30 31 32 33 34 35

Ongoing analysis of individual patient data from randomised trials to determine the effect of alteplase in subgroups of patients with different comorbidities and subtypes, severities, and imaging features of stroke should help refine our identification of patients likely to benefit from alteplase.5 36 Trials are also investigating whether lower doses of alteplase, active lowering of blood pressure, and concurrent use of minocycline can help reduce intracranial haemorrhage.

Meanwhile clinicians should urgently assess all patients presenting with suspected acute stroke and consider thrombolysis in those who meet all the inclusion criteria in the relevant randomised trials and clinical guidelines for alteplase in ischaemic stroke and who consent to accepting the risks of treatment for potentially greater benefit.


Cite this as: BMJ 2013;347:f5215


  • Competing interests and funding: The authors have read and understood the BMJ Group policy on declaration of interests and declare SB is supported by a National Health and Medical Research Council (Australia) Career Development Fellowship and a Royal Perth Hospital Medical Research Foundation grant to conduct mechanistic research into the immunological mechanisms and management of serious illness and injury, including stroke. GJH was co-national coordinator for Australia and a member of the steering committee for the International Stroke Trial (IST)-3. He has received payments for advisory board consultancies and speaking at educational events on novel anticoagulants in atrial fibrillation.

  • Provenance and peer review: Commissioned; not externally peer reviewed.