To the Editor: The recent article by Szoeke and colleagues on stroke management expressed the hope that thrombolytic therapy will be licensed for use by specialist units in Australia, based on the “proof” of its benefit demonstrated by the National Institute of Neurological Disorders and Stroke (NINDS) trial.1

The NINDS trial was a small, flawed study in which 312 patients received thrombolytic therapy with tissue plasminogen activator (tPA) for stroke.2 Higher scores for stroke severity in the placebo group could themselves explain the improved outcome attributed to thrombolysis. Further clarification has been thwarted by the investigators’ refusal to release the raw data and allow clarification of uncertainty surrounding the results whereby benefit appears confined to those treated at 0–90 minutes after onset, with no benefit in those treated at 90–180 minutes.2

Reports of the introduction of thrombolysis with tPA into clinical practice consistently document substantial protocol violations and worse outcomes than without thrombolysis. The study by Szoeke et al documents mortality attributed to thrombolysis given when protocol criteria were not met, as well as a 23% protocol violation rate in a presumed “best practice” setting. It should be highlighted that their finding in an audit of 30 patients that outcomes were “consistent with reported trial data” means they were also consistent with a worse outcome, although confidence intervals are not presented.

Thrombolysis with tPA is not endorsed as a standard of care in stroke by the Canadian Association of Emergency Physicians, the American Academy of Emergency Medicine or the American College of Emergency Physicians; nor do any of these organisations advocate its introduction into practice outside research trials.3,4

In contrast, the American Heart Association upgraded its rating for thrombolysis in stroke from a class IIa to a class I recommendation in its 2000 guidelines without any additional data from randomised trials. Conflicts of interest are substantial and not widely disclosed.2 Genentech, the manufacturer of tissue plasminogen activator, has contributed US$11 million to the American Heart Association and paid for its national headquarters. Six of the nine panellists responsible for the guidelines had financial ties to Genentech, which were not disclosed.5 A dissenting panellist had his name removed from the list of contributors, despite previous assurances that his dissenting position would be published.5

There are not many areas where so much has been made of so little; it falls a long way short of proof.

In reply: The comments of Smith may not reflect the consensus of his emergency medicine colleagues. The broad view of the place of tissue plasminogen activator (tPA) is best appreciated from overviews and meta-analysis of all trials of intravenous tPA in acute ischaemic stroke.1-3 Overall, tPA is one of the most powerful biological agents in medicine, with a number needed to treat of about eight to benefit one patient.

The integrity of the investigators of the tPA trials is unquestionable. All results were published in journals of the highest repute (including the New England Journal of Medicine, the Lancet and the Journal of the American Medical Association). The National Institute of Neurological Disorders and Stroke (NINDS) trial4 was investigator-driven and funded by the US National Institutes of Health, the highest standard achievable in trial management.5 NINDS receives unrestricted grants from the pharmaceutical industry with appropriate ethical guidelines, as do many societies worldwide. To suggest a linkage is certainly extending conspiracy theory to its limits.

Several trial-related issues mentioned by Smith deserve comment. In the NINDS trial, as in many randomised controlled trials, the analysis adjusted for minor baseline imbalances in stroke severity, with no significant impact on outcome. The Melbourne study of Szoeke et al was not a randomised controlled trial, but rather an audit of practice in an expert setting.6 The protocol violations were all relatively minor, and the rate of 23% is comparable with rates in other Phase IV studies.7

Although only a small proportion of stroke patients are eligible for tPA, it is one of the most important advances in stroke medicine. Emergency physicians must play a collaborative role with stroke physicians in delivering this benefit.

To the Editor: In their retrospective audit of stroke patients presenting to a Victorian hospital, Szoeke and colleagues present a one-sided view of the usefulness of thrombolysis with tissue plasminogen activator (tPA) in acute stroke.1

Proof of the efficacy of tPA in acute ischaemic stroke is far from settled. The National Institute of Neurological Disorders and Stroke (NINDS) trial, in which only 312 patients received thrombolytic therapy, remains the only trial demonstrating benefit from intravenous thrombolysis in a primary outcome measure.2

The study design of the NINDS trial required the enrolment of a disproportionate number of patients with very early stroke (within 0–90 minutes of onset). These patients are rarely encountered in everyday clinical practice. Patients in the 91–180 minute group who received placebo were sicker at baseline than those who received tPA, raising significant doubts as to the efficacy of tPA. After further analysis of the results, the NINDS investigators reported that the greatest positive effect of tPA was seen in the 0–90 minute group.3 The positive effect of tPA in the 91–180 minute group, while not specifically reported, can only have been very small. It is interesting that the median time to treatment in Szoeke et al’s study was 2 h 48 min, implying that 50% of patients were treated in the last 12 minutes of the 3-hour window — when benefits of treatment are at their smallest (should they exist at all), but all the risks of therapy remain.

I am also astounded that in the setting of a dedicated stroke unit, with all patients attended to by a team comprising “a stroke neurologist, ‘stroke’ fellow, registrar and nurse”, and a requirement for specific approval for use of tPA to treat stroke in the hospital, that protocol violations occurred in 23% of patients (7/30). That is not the sort of performance that I would want to place in the public domain.

Tiny retrospective “trials” are fraught with potential bias, as non-blinded treatments and outcome measures may reflect the enthusiasm of the authors. In addition, significant publication bias may exist — groups with bad results from thrombolysis may not publish.

I await the publication of further well designed, randomised, placebo-controlled clinical trials, not linked to the manufacturers of tPA, that demonstrate an improvement in a primary outcome measure in patients treated with tPA before deciding that this treatment may have some use outside clinical trials. I will not hold my breath.

In reply: Johnson correctly points out that the National Institute of Neurological Disorders and Stroke (NINDS) trial had two parts.2 Part 1 was designed to test whether tissue plasminogen activator (tPA) had early clinical activity at 24 hours, and Part 2 was designed to assess whether tPA conferred outcome benefits at 3 months. The results of Part 1 of the trial showed a non-significant improvement in neurological score at 24 hours. However, the finding in Part 2 that there was a significant difference in the primary endpoint (a global test statistic at 90 days) was also true for Part 1 and for a combined analysis of Parts 1 and 2.3 Although there was a substantial increase in symptomatic intracerebral haemorrhage in the thrombolysis group, there was no increase in mortality, and the adverse effects were outweighed by the highly significant benefits at outcome.

Johnson’s comment on the choice of endpoints is also of interest. It should be emphasised that the efficacy of tPA in the NINDS trial applied to a range of standard outcome evaluations, including the NIH Stroke Scale, Glasgow Outcome Scale, modified Rankin Scale and Barthel Index, as well as the new global test statistic, which incorporates these scores.2

Many of the issues raised by Bailey have been covered by Donnan et al,4 but some additional comments are warranted. Based on the NINDS trial and meta-analyses of all the intravenous tPA data, this therapy has been licensed for stroke in the United States, Canada, Europe (including the United Kingdom) and many other parts of the world. It is being considered for licensing in Australia. The minor baseline disparities between the tPA and placebo groups have been subject to further rigorous analysis by an independent review committee commissioned by NINDS. This analysis confirmed the statistically significant benefit of tPA within 3 hours.5

The benefits of tPA are indeed time-linked, as shown by further analysis of the NINDS data, but are highly significant right up to the end of the 3-hour window.6 The odds ratio for favourable 3-month outcome with tPA was 2.11 (95% CI, 1.33–3.35) for treatment at 0–90 minutes and 1.69 (95% CI, 1.09–2.62) for treatment at 91–180 minutes. Furthermore, meta-analysis indicates benefit beyond the 3-hour window,7 but there is consensus that further trials are needed to extend the current window, and that tPA should not be used after 3 hours, except in clinical trials.

We again emphasise that our audit was not a “trial” and that our protocol violations were generally minor and in line with other expert experience.1 We do not apologise for emphasising the importance of a well-resourced acute stroke team. This is integral to the expert setting required for tPA administration. In 2003, would anyone suggest that patients with acute myocardial infarction should be treated without optimal resources and expert care? Why should acute stroke patients, with high mortality and disability rates, be the poor relations?