Re: Do risks outweigh benefits in thrombolysis for stroke?
Sandercock, Wardlaw and Dennis claim we have refused to consider the evidence in a balanced way, and that this will deny our patients with acute ischaemic stroke an effective treatment. This is hard to defend, other than to say that we think otherwise and that the proponents of this treatment are exposing their patients to potential harm without sufficient evidence of benefit. We don’t know who is right, because some basic scientific principles have not been applied. These include: (i) elimination of bias, (ii) replication of findings, and; (iii) informed debate.
Sandercock et al. think we over-emphasise the effect of study group imbalances (allocation bias) in NINDS  and ECASS-3 , but offer no evidence to substantiate their view. Allocation bias in ECASS-III was highly significant  and a nearly identical study (ATLANTIS) that did not have the same bias found significant harm without any benefit in patients given rtPA (alteplase) [4,5]. This stark comparison suggests that allocation bias is, in fact, a critical issue.
With regard to the only other positive rtPA trial (NINDS) Hoffman and Schriger make a compelling case that the “treatment effect” observed in NINDS is due largely to allocation bias with a greater number of more severe strokes in the placebo group . A subsequent paper by Saver demonstrates some potential flaws in Hoffman and Schriger’s approach  but really does not deal adequately with the issue of baseline imbalance in stroke severity. Corrections for allocation bias were made during a reanalysis of the NINDS data , but we must recognise that such post-hoc corrections are in fact subject to investigator bias, because the investigator is in the position to choose the statistical method that provides the best result. Hence it is recommended that: (i) randomisation algorithms include a system for minimising imbalances in the first place (one of the good things that was done in the IST-3 trial) and/or; (ii) any statistical corrections for inadvertent imbalances are predefined, before the trial starts.
History is replete with examples of drug approvals being withdrawn and treatment guidelines superseded after more considered analyses or when new evidence emerges. A recent example is the use of drotrecogin alpha (Xigris, Eli Lilly) in severe sepsis, another high-mortality condition . Although Xigris was approved for clinical use after one positive trial result, a further trial to replicate these findings was performed. This turned out to be negative and the drug was voluntarily withdrawn.
By comparison, for tPA in stroke there has been only one positive study of treatment <3 hours - NINDS. This was a relatively small study (n=333) that triggered a heated debate between highly capable and respected clinical scientists that continues to this day. However, a methodologically rigorous (i.e. double-blind) trial to replicate the results of this single positive study has not been done. This lack of replication (a standard applied to many other areas of medicine, including thrombolysis for myocardial infarction) is of particular concern given the potential for substantial harm.
Reanalysis, post-hoc statistical manipulations, complex meta analyses, registry studies and flawed open label trials like IST-3 simply cannot resolve these issues . Sandercock et al. think it would be unethical to conduct further placebo-controlled trials. We think the reverse is true - that the current state of the evidence and the potential for great harm from thrombolysis in stroke makes it unethical to push this treatment as a standard of care without sufficient evidence from another double-blind, placebo-controlled trial.
We wholeheartedly agree that patients with stroke and their families need to be partners when deciding on treatment options. This should be informed by a realistic appraisal of the risks and potential benefits of thrombolysis, which at present should be considered an unproven therapy with a substantial associated risk of harm.
1. The National Institute of Neurological Disorders and Stroke rt-PA Stroke Study Group. Tissue plasminogen activator for acute ischemic stroke. N Engl J Med 1995;333:1581-7.
2. Hacke W, Kaste M, Bluhmki E, Brozman M, Dávalos A, Guidetti D, et al. Thrombolysis with alteplase 3 to 4.5 hours after acute ischemic stroke. N Engl J Med 2008;359:1317-29.
3. Shy BD. Implications of ECASS III Error on Emergency Department Treatment of Ischemic Stroke. J Emerg Med 2012.
4. Clark WM, Wissman S, Albers GW, Jhamandas JH, Madden KP, Hamilton S. Recombinant tissue-type plasminogen activator (Alteplase) for ischemic stroke 3 to 5 hours after symptom onset. The ATLANTIS Study: a randomized controlled trial. . JAMA 1999;282:2019-26.
5. Clark WM, Madden KP. Keep the three hour TPA window: the lost study of Atlantis. J Stroke Cerebrovasc Dis 2009;18:78-9.
6. Hoffman JR, Schriger DL. A graphic reanalysis of the NINDS Trial. Ann Emerg Med 2009;54:329-36, 336 e1-35.
7. Saver JL, Gornbein J, Starkman S. Graphic reanalysis of the two NINDS-tPA trials confirms substantial treatment benefit. Stroke 2010;41:2381-90.
8. Ingall TJ, O'Fallon WM, Asplund K, Goldfrank LR, Hertzberg VS, Louis TA, Christianson TJ. Findings from the reanalysis of the NINDS tissue plasminogen activator for acute ischemic stroke treatment trial. Stroke 2004;35:2418-24.
9. Martí-Carvajal AJ, Solà I, Lathyris D, Cardona AF. Human recombinant activated protein C for severe sepsis. Cochrane Database Syst Rev 2012;3:CD004388.
10. Radecki RP, Chathampally YG, Press GM. rt-PA and Stroke: Does IST-3 Make It All Clear or Muddy the Waters?Answers to the November 2012 Journal Club Questions. Ann Emerg Med 2013;61:489-98.
Competing interests: No competing interests