Timothy John Kleinig, PhD, Head of Stroke Unit; Leonid Churilov, PhD, Professor and Head of Statistics and Decision Analysis Platform; Cathrin Sibylle Parsch, MBBS, FACEM, Senior Emergency Consultant; Helen M Dewey, PhD, Associate Professor and Head of Stroke Unit; P Alan Barber, PhD, Professor of Neurology and Head of Stroke Unit.
Stroke thrombolysis and the third international stroke trial: Examining ‘the totality of the evidence’
Article first published online: 7 APR 2013
© 2013 The Authors. EMA © 2013 Australasian College for Emergency Medicine and Australasian Society for Emergency Medicine
Emergency Medicine Australasia
Volume 25, Issue 2, pages 107–109, April 2013
How to Cite
Kleinig, T. J., Churilov, L., Parsch, C. S., Dewey, H. M. and Barber, P. A. (2013), Stroke thrombolysis and the third international stroke trial: Examining ‘the totality of the evidence’. Emergency Medicine Australasia, 25: 107–109. doi: 10.1111/1742-6723.12046
- Issue published online: 7 APR 2013
- Article first published online: 7 APR 2013
In the October issue of Emergency Medicine Australasia, the editorials of Fatovich and Hoffman and Cooper discuss stroke thrombolysis in the light of the third international stroke trial (IST-3). The authors mention several flaws, which were all documented by trial authors.[3, 4] These flaws did not stem from inexperience with trial design or lack of scientific rigour but with restricted trial funding. The authors laudably chose to continue an academically funded trial with acknowledged limitations rather than to terminate it. The editorials raise the possibility of systematic biasing of outcome from ‘P.R.O.B.E.’ design; however, this seems unlikely given the lack of benefit in many patient groups. Mortality is the ultimate ‘hard end-point’. IST-3's neutral mortality result defuses criticism that alteplase-favourable unblinding concealed net harm.
Neither editorial critically appraised the population included in IST-3. Patients were eligible by virtue of the ‘uncertainty principle’, that is, patients were only enrolled if the clinical team were uncertain of the benefit of thrombolysis. For most triallists, this denoted an ever-shrinking pool of eligible patients. IST-3 was conceived in 1998, and enrolment commenced in 2000. However, alteplase was licensed in Europe in 2003. In 2004, a meta-analysis demonstrating time-dependent benefit up to 4.5 h was published. In 2007, the Safe Implementation of Thrombolysis in Stroke-Monitoring Study registry data confirmed benefit in ‘real world’ practice. In 2008, the European Cooperative Acute Stroke Study III (ECASS III) confirmed benefit to 4.5 h. In 2010, an analysis suggesting benefit in patients aged over 80 was published. The trial therefore suffered from ‘reverse cherry-picking’.
Trial organisers estimated that 2% of IST-3 patients met 2003 European Union alteplase licensing criteria. Following expansion of the treatment window to 4.5 h, 12.7% would have met the criteria. IST-3 is therefore largely a trial of alteplase in patients with various alteplase relative contraindications. For example, 2300 of 3035 participants were aged over 80 years or treated between 4.5 and 6 h.
It should be acknowledged that IST-3's primary outcome was negative. Alteplase should not be used indiscriminately in patients presenting under 6 h. With the benefit of hindsight, this was unsurprising. However, important new data were obtained from IST-3, none of which subvert current thrombolysis practice. First, and arguably most important, IST-3 enrolled more patients aged over 80 years and treated within 3 h than the whole National Institute of Neurological Disorders and Stroke rt-PA Stroke Study Group (NINDS) trial. Treatment in this group was highly beneficial. Second, patients with clinically severe strokes (usually considered a relative contraindication) seemed to benefit most from treatment. Third, a 3.7% risk of early death (predominantly from intracerebral haemorrhage) was neutralised by reduction in late deaths, negating an overall mortality trend seen in previous meta-analyses. IST-3 confirmed that unselected thrombolysis 4.5–6 h post-onset is not beneficial.
The editorials consequentially challenge the validity of stroke thrombolysis. Fatovich claims alteplase advocates ‘believe in one study’. Both papers seek to sway readers by tallying ‘positive’ and ‘negative’ thrombolysis trials. But it is not logical to argue against ECASS III (alteplase administered 3–4.5 h) by citing the Desmoteplase in Acute Ischaemic Stroke 2 (desmoteplase administered mostly 6–9 h). A total determination of risk and benefit is determined by data meta-analysis and not by a ledger of positive and negative trials. The relevant meta-analysis, including 6887 patients from 12 alteplase trials) was published together with IST-3, but not mentioned by either editorial. This is somewhat surprising as Fatovich states that ‘the totality of the evidence’ should be considered. In all trials of alteplase-treated patients within 3 h of onset, regardless of age, the number needed to treat for ‘cure’ (no symptoms or symptoms with no disability) is 11 (95% confidence intervals [CI] 8–22, P < 0.0001.)
The issue Fatovich raises of NINDS trial baseline stroke severity imbalance has been dealt with comprehensively by an independent committee that reviewed the trial data at NINDS' behest, following criticisms raised by the American Emergency Physicians. The trials' positive findings were unaltered by adjustment for baseline imbalances. Similarly, the supposed lack of benefit demonstrated by graphical analysis of Δ-National Institutes of Health Stroke Scale (NIHSS) at 90 days is highly contestable.
Regardless, functional outcome is a valid and meaningful outcome measure, and Δ-NIHSS is not. Patients are unconcerned whether their NIHSS is 3 or 4, but they care greatly if they can or cannot walk independently (modified Rankin scores [mRS] of 3 and 4). We agree with Fatovich that depending on the research question, dichotomising mRS of 0–1 versus 2–6 can miss important alteplase trial information. Although this dichotomisation examines an absolute threshold of excellent functional outcome (no symptoms or symptoms with no disability), an alternative approach – ‘shift analysis’ – detects harm and benefit across the whole mRS spectrum. This method suggests a ‘number needed to functionally benefit’ from tissue plasminogen activator of 3–4 in patients treated within 3 h. Incidentally, this method was highly statistically significant in a pre-specified analysis of IST-3 (P < 0.001).
Fatovich demands that the NINDS trial be repeated, but IST-3 actually fulfilled this wish, showing clear evidence of benefit in the 849 randomised within 3 h. It will not be repeated again; the comprehensive meta-analysis now demonstrates a clinically relevant treatment effect with a less than 0.01% chance of being spurious. This is sufficiently convincing to all professional societies representing stroke physicians worldwide and most doctors who routinely treat stroke patients. The future of acute stroke research does not lie in attempts to further tighten 95% CIs, but in the testing of more effective thrombolytic agents, in finding ways to decrease onset-treatment times, in examining additional recanalisation techniques, in testing penumbral imaging-based patient selection beyond 4.5 h and in using individual patient meta-analysis and advanced neuroimaging to refine selection criteria. Stroke physicians and neurologists from around the country welcome the assistance of emergency physicians in helping us perform these important studies. Additionally, enhanced cooperation between our specialties will enable more patients to be treated earlier, when most agree that benefits are greatest.
TJK wrote the initial draft. All additional authors revised it critically for important intellectual content and gave final approval of the version to be published.
There are no relevant financial interests to declare. HMD enrolled patients in IST-3 but was not involved in either study design or interpretation.