Thrombolysis for acute ischemic stroke: still a treatment for the few bythe few

The National Institute of Neurological Disorders and Stroke (NINDS) trialwas groundbreaking.¹It demonstrated that stroke patients could be admitted to hospital, scanned,and treated within 3 hours of stroke, a major achievement in the early 1990s.The Food and Drug Administration accepted the NINDS evidence (actuallyconducted as 2 consecutive trials, only 1 of which gave positive results) andlicensed recombinant tissue plasminogen activator (tPA) for the treatment ofstroke.

That was 1995, and stroke trial methods have moved on, particularly inrandomization and follow-up methods. Patients were randomly allocated into theNINDS trial by the selection of a drug “prepack” stored in eachcenter, followed by retrospective notification to the central trial officewithin 2 hours. Thus, it was impossible to balance randomization for baselinefactors such as stroke severity. The 3-month follow-up was by a physician ineach center. It is difficult to know how blind this assessor truly was (thebiologic effects of recombinant tPA are difficult to blind). The opportunitynot to be blind to treatment allocation was obvious. Modern trialscommonly use prospective randomization from a central site and minimization (amethod of balancing important prognostic variables across treatment andcontrol groups) to avoid imbalances in baseline patient characteristics andimprove security (ie, once in the trial, the patient cannot be omitted). Theuse of central telephone or postal follow-up directly to the patient can alsoavoid any potential for local investigator influence.

Could this dated design have influenced the NINDS results? Jeffrey Mann'sarticle is an important comment and a careful piece of detective work thatclearly describes the imbalance in stroke severity between patients allocatedto receive tPA and those given placebo. But how important is this imbalance?We have independently examined the published data from the NINDS investigatorsand agree that the baseline imbalances explain about a third of the apparentbenefits. The baseline imbalance alone would produce the following proportionof excellent outcomes without the tPA: 0 to 90-minute group, 32% tPA versus33% placebo; 91 to 180-minute group, 39% tPA versus 32% placebo; and allpatients randomly allocated into the trial, 0 to 180 minutes, 36% tPA versus32% placebo. This 4% absolute difference, or 40 per 1,000 more independentsurvivors, arose simply because of a chance imbalance in baseline strokeseverity and has been overlooked by neurologists. False-positive results arecommon in small, inevitably unbalanced, trials. The warning, “Don'tIgnore Chance Effects” (or DICE) is an excellent example of how chanceskews trial results, and it should be more widelyrecognized.²

Does this mean that the NINDS result is invalid? No, for two reasons.First, of the 120 per 1,000 more who were alive and independent according tothe NINDS trial, in theory, 80 per 1,000 had improved outcome that might beattributed to the use of the recombinant tPA, not chance. Second, the resultis consistent with the data from patients randomly assigned within 3 hours inthe other seven trials oftPA.³ In fact, theNINDS trial contributes only 21% of the total data available on tPA (8 trials,total of 2,955 patients, up to 6 hours after stroke). Given the citation biasof the positive NINDS trial (43 articles, equivalent to 6% of the sample size,or 14 patients per publication!), it is easy to understand why this has beenoverlooked.

The totality of the data suggests that tPA is doing something, but the keyquestion, and the factor that above all else may have kept tPA from wider use,is how big is that effect and in which patients? The big disappointment isthat the uncritical and optimistic acceptance of tPA has discouraged thestroke community from obtaining more convincing data and actually blocked thefinding of safe ways to give tPA more widely. If tPA is so effective, then itshould be available for most people, not just the few lucky enough to beadmitted to a specialist tertiary referral center. Patients with stroke areoften elderly, frail, and aphasic and thus unable to fight for their rights.This simple fact is reflected in a grossly disproportionate lack of interest,research funding, and medical intervention for stroke compared with far lessdevastatingconditions.⁴

So, 6 years after NINDS, thrombolysis remains a treatment for the few bythe few. How can the stroke community solve this problem? First, implementbasic proven knowledge: organize inpatient stroke services, and give aspirinimmediately to all aspirin-eligible patients. Then, where information islacking, get more data: for those patients for whom the use of tPA may bepromising but is unproved, we suggest randomly allocating them into a trialsuch as the Third International Stroke Trial, which is evaluating the use oftPA up to 6 hours after definite stroke onset. Details are available from theauthors or at the trial's web site(www.dcn.ed.ac.uk/ist3).