When I testify on behalf of an emergency physician accused of medical negligence simply because he or she did not use thrombolytic therapy in a patient with acute ischemic stroke (AIS), I do not (as many attorneys seem to expect) offer the opinion that such treatment is harmful. What I say, instead, is that there is too little available evidence for us to know whether, on balance, it does more good than harm, more harm than good, or has no appreciable effect. Indeed the amount of good data from high-level studies is appallingly small. Furthermore, although one highly cited randomized controlled trial (RCT), the National Institute of Neurologic Disease and Stroke (NINDS) study,1 reported about a 12–13% absolute benefit, compared with placebo, in 90 day outcomes among the small cohort of patients who received the treatment, none of the handful of other RCT has confirmed such benefit, and several of them clearly found overall harm.2–7
I am also often asked what I would do if I, or a member of my family, had a stroke, and ‘qualified’ for thrombolysis, according to the NINDS guidelines. Here, too, I do not have an easy answer, because much though I believe that such drugs may well ultimately be shown to do more harm than good, I also know that it would be tragic to fail to use a therapy which could produce dramatic benefit in an individual patient, for such a devastating disease. I therefore typically answer along the lines of ‘I wish I knew’. I suggest as well that I would try to factor in such things as the severity of the stroke, the likely prognosis without thrombolysis, time from symptom onset, whether treatment would be delivered by an expert etc. Thus, I do not foreclose the possibility of considering such therapy, at least in the abstract. On the other hand, I really do wish I knew better – perhaps if there were more evidence I could give a more reasoned answer, both in general, and in specific. That is to say, with more information, we might be able to define a subset of AIS patients in whom the drug offers the likelihood of more benefit than harm, and another subset in whom it should not be used, because the opposite is true.
I say all this in order to suggest that when we as a community fail to address adequately a potentially critical issue for individual patients, we tolerate a situation that should be considered unacceptable. For more than 10 years, now, since the publication on NINDS, I and others have been pleading for more studies, and more evidence. That plea remains no less important today than it was before.
Some advocates of tissue plasminogen activator (tPA) for AIS claim it would be unethical to do further studies, in light of NINDS. There are a number of flaws in that line of reasoning. First of all, even if there were no other RCT of thrombolysis for AIS, it would be foolhardy to extrapolate from the results of a single study. This is magnified by the fact that NINDS was relatively small,1 with just over 300 patients treated with tPA, as well as by the fact that the absolute benefit was also relatively small. In addition, given that some patients are clearly harmed by the drug, any slight misestimation of the balance between benefit and harm could be of enormous importance. In truth, although, there are multiple other studies of thrombolysis for AIS, and not a single one of them was successful according to its own primary end-points.2–7 It is possible that this is because only NINDS tested the right hypothesis – this particular drug, given in this particular time window, at this precise dose, really is better than any of the regimens used in the multiple other studies. It is, however, equally plausible that what we are seeing simply reflects the expected splay in results that would occur if the same study (without any different doses, or timing etc.) were repeated a handful of times – one of the studies would be expected to find results that are a little better than any of the others (and conversely one would be expected to have outcomes worse than the average, as in the ASK trial, for example).
None of this even considers some of the problems with NINDS itself, including the fact that it recruited half its study group among patients who were able to be treated in less than 90 min from symptom onset; this is certain to have distorted the measured benefit if, as advocates claim, ‘time is brain’.1,8 In real life, one might expect perhaps 5%, at most, of all those who ‘qualify’ for tPA, to arrive early enough to be included in such an extremely early treatment group. Furthermore, there were by chance major differences in baseline stroke severity – a critical determinant of outcome – between patients who received tPA and those who received placebo, in the (later) 91–180 min subgroup. In this group – who represent the vast majority of those likely to receive thrombolytic therapy in a real-world setting – patients randomized to tPA had far less severe stroke at baseline than did those who were randomized to placebo, such that the outcome differences found after treatment could easily be explained by the baseline differences.8
Defenders of the NINDS study point out that sophisticated statistical analysis fails to identify any subgroup that looked statistically different than the overall population, and who therefore should not receive the drug.9 What they fail to acknowledge, however, is that the study was woefully under-powered to test the hypothesis that any or many of the subgroups were indeed different, or that they included results that should warn us away from using this therapy. The failure to find a statistical difference, in a study this small, when it is further divided into multiple subgroups, is in no way equivalent to proving that no such difference exists.
The same type of statistical support supposedly derives from a number of rehashes and re-analyses of old data, including from studies that originally failed to show benefit.2,9,10 Aside from the fact that these re-analyses are selective in terms of the data they include (conveniently ignoring negative data involving streptokinase, which for many reasons is inappropriate), and in terms of the measures retrospectively applied to the original data (magically producing ‘positive’ results from the very same data that were negative according to the original studies themselves) it is simply poor science, and poor mathematics, to use such re-analyses for anything other than generation of future hypotheses.11 Retrospective slicing and dicing of negative data to produce new, marvellously positive results simply is not the equivalent of retesting the precise regimen used in NINDS, to see if it would benefit another cohort of patients, under the same conditions.
Beyond even all of this there is the fact that ‘efficacy’ of a treatment given in the carefully controlled conditions of a study should never be assumed to translate directly into ‘effectiveness’ with widespread community use.12–14 This is particularly worrisome with a drug that has the potential to do major harm. It is further worrisome given the fairly high prevalence of stroke mimics, which seem to represent somewhere between 5 and 20% of all patients diagnosed in the ED as having AIS.15 Because most such mimics – Todd's paralysis following a seizure, or an endocrine abnormality such as hypoglycaemia – are likely to arrive at an ED very soon after the onset of symptoms, whereas only a very small proportion of true AIS patients do so, their representation among the group of patients who ‘qualify’ for thrombolysis is likely to be even substantially higher. Finally, throw in the pressure on physicians to make an extremely rapid diagnosis so as to begin therapy as soon as possible, it is not surprising that studies evaluating community-wide use of tPA for AIS find an extraordinarily high proportion of ‘protocol violations’. Because a stroke mimic can certainly not expect to benefit from thrombolysis, but has every chance of being harmed, inclusion of even a few extra such patients in community use (more than occurred in the expert RCT) would easily tip the balance from benefit (even if it really does exist in the best of circumstances) to substantial harm.
Although it would indeed be tragic to withhold a potentially important therapy from someone who, without it, might go on to a devastating outcome, it is no less tragic to convert what would have a perfectly good outcome into a terrible one (as can occur with tPA-induced intracerebral haemorrhage). We can never be certain, beforehand, what will happen in an individual, either with or without any particular course of treatment, so we must make decisions based on best estimates of the likely benefit versus harm with any proposed action, were we to use it among a group of patients similar to the one for whom we are caring. Unless and until we know far more about this likely balance with tPA, we run the real risk of mistreating our patients, no matter which decision we make.
Although I have tried to emphasize that I do not believe there is nearly enough evidence to know with confidence what type of balance would occur with the widespread use of tPA for presumed AIS, I suspect it is unlikely that it would produce substantially more good than harm, or substantially more harm than good. Not for individual patients, of course, who may have tremendously positive or tremendously negative results. But in the AIS population at large it is almost inconceivable that tPA would have as much impact on overall outcome as the use of a single aspirin.16,17 If that is the case, one has to ask why so much attention has been paid to this treatment . . . and at the same time why some of us have spent so much effort raising objections. If it really is not likely to cause lots of harm, why put up so much resistance – even if it does not produce lots of benefit? There are two good answers, I believe. First, as noted above, only with more information will we be able to avoid doing the wrong thing for individual patients. Although the population overall may not gain or suffer markedly, I do not want to use a drug that is more likely than not to hurt the patient in front of me. At the same time, and equally important, I do not wish to deny it to someone who stands a real chance of reaping important benefit.
There is perhaps an even more important reason to resist the steamroller efforts by advocates of thrombolysis for AIS. I believe scientific questions should be answered by the best available science, and that we should not allow misplaced zeal, not to mention politics, or money, to stand in the place of such science. Among the organizations whose guidelines have trumpeted tPA for AIS as a ‘proven’ therapy, some have substantial conflicts of interest, having received major financial support from the drug's manufacturers.18 The same is true of many of the experts chosen to sit on guideline panels, who make these recommendations.18,19
It is equally distasteful to use extra-medical venues to try to influence practice, when debate and argument have failed to do so. Advocates of thrombolysis for AIS sometimes dismiss sceptics – including the large number of emergency physicians who have chosen not to adopt this therapy – as either greedy, or simply ignorant.20 In addition, there have been efforts to force changes in behaviour through regulatory agencies (such as the Joint Commission on the Accreditation of Healthcare Organizations in the USA), legislation (numerous bills have been proposed at the local, state and national level in the USA), and the financial and emotional penalties associated with claims of malpractice.
Which takes us back to the beginning … lawsuits against emergency physicians for not having used this drug in AIS. Although the details of these cases – and the overall care provided – vary enormously, such that my personal feelings about them is also very varied, I believe it would be tragic – in every one of these cases – if we allowed the threat of legal action to stampede us, as individuals and as a community, into adopting an approach about which a huge number of us so obviously remain unconvinced. It is ultimately for this reason that I personally am willing to put so much time and effort (for which I receive no personal financial recompense) into this issue – even though I myself suspect that the impact of thrombolysis for AIS will ultimately prove to be small. It is because I also believe that the impact of this larger question – how do we as a medical community allow policy to be made, in the absence of scientific rigor – is likely to be far, far more substantial.