Opening argument: NO
ECASS III is Just a New Spin on a Bad Idea
by Kevin Klauer, DO
Here’s the history. For those in support of t-PA, the basis for this support is the NINDS trial from 1995. This study should have only provoked more investigation and not prompted widespread acceptance of t-PA administration in stroke. So, how many patients did this landmark study test, “proving” t-PA’s benefits: 100,000? 50,000? 20,000? How about 1000? None of the above. NINDS conclusions are based on the treatment of 624 patients, with only about half receiving t-PA. If this study had adequate power, then why do we keep asking the same question in follow-up studies? The results suggest that patients treated with t-PA are 30% more likely to experience complete or nearly complete neurological recovery. At what expense? The expense to the patient is the risk of intracranial hemorrhage. Although this risk is substantial, pro-t-PA investigators frequently minimize it. NINDS discovered a 6.4% “symptomatic intracranial hemorrhage” rate for those treated with t-PA. What isn’t often mentioned is the additional 4.4% that are “asymptomatic,” totaling a 10.8% intracranial hemorrhage rate. So, for a 30% chance of resolving or partially resolving your stroke symptoms, 10.8% of patients will suffer a major complication. When you consider the possibility that some of the patients claimed by NINDS, as successes from t-PA, must have been TIAs that would have resolved anyway and strokes that may have substantial improvement without the administration of t-PA, the 30% more likely to improve number is overstated.
So, if NINDS, unveiled, is not so convincing, why did the American Heart Association give t-PA in stroke a glowing endorsement, a Class I recommendation, in its 2000 guidelines, which they further stated “Saves lives”? Following the money often brings clarity to things that don’t make sense. Genentech is known to have donated more than $10 million over 10 years to the AHA, including 2.5 million dollars to build a new headquarters in Dallas, Texas. The AHA maintained that the panel of nine physicians that drafted the recommendation were not biased and had submitted conflict-of-interest forms. Unfortunately, the AHA does not publish those forms for public review. Eight of the panelists endorsed this recommendation and one, Jerry Hoffman, dissented.
Interestingly, six of the eight supporters had financial ties to
Genentech, and Dr. Hoffman’s testimony in dissent was not included in
the guidelines and his name was subsequently removed from the group of
panelists. Of further interest is that Genentech attempted to duplicate
the NINDS data with the ATLANTIS trial. In part A of the trial, there
was no benefit and the 30-day mortality rate was 18% vs. 4% for those
not treated with t-PA. A 0-6 hour time window was used. However, a 0-3
hour cohort was studied as well. Interestingly, analysis of that
subgroup was never included in the final publication. The final
publication, released in 1999, was published six years after completion
of the part A data (Jeanne Lenzer. Alteplase for stroke: money and
optimistic claims buttress the “brain attack” campaign, BMJ. 2002 March
23; 324(7339): 723–729.).Am I a t-PA skeptic? You bet I am. There are
external validity issues with the NINDS trial. This is probably why
more studies are being done by t-PA supporters to prove its efficacy.
If the question has been answered, why do we keep studying the concept?
There are many studies casting doubt on the safety of t-PA in stroke
patients. The Cleveland Area Experience, JAMA 2000, had to be stopped
due to a 17% intracranial hemorrhage rate. Furthermore, two large
registries, one of nearly 250,000 stroke victims and the other of
nearly 100,000, report that those treated with t-PA have almost twice
the mortality rate as those not treated with the drug, 11.4% v. 6.8%
and 10.1% v. 5.8%. (Bateman, B.T., et al, Stroke 37:440, February 2006;
Dubinsky, R., et al, Neurology 66:1742, June 2006). These registries
also concur that about 1% of stroke victims, 13% of those eligible,
actually receive t-PA.The goal of the ECASS investigators was to extend
t-PA to more patients by expanding the treatment window to 4.5 hours.
In short, ECASS 3, the European Cooperative Acute Stroke Study 3, in my
opinion, is a new spin on a bad idea. What hasn’t worked and has harmed
patients at 3 hours, will certainly pose similar risks at 4.5. This
study really hasn’t proved safety or efficacy any better than any of
its predecessors.
The investigators report that expanding the time window provided benefit to “substantially more patients,” without increasing the rate of intracranial hemorrhage (ICH). The odds ratio for a “favorable outcome” at 90 days was 1.34. An odds ratio of 1 would mean that the treatment group and the placebo group had the same chance to of experiencing the favorable outcome at 90 days. Thus, 1.34 really isn’t much benefit over placebo at all. Although they reported that the intracranial hemorrhage rate was higher in the treatment group, they stated that it was the same as noted in previous studies for patients treated within three hours. What is not widely publicized is that the investigators changed the definition of symptomatic ICH. They stated that in order to be considered symptomatic, it must be identified as the predominant cause of the deterioration. Nonetheless, they report an overall ICH rate of 27.0% vs. 17.6% for placebo. The change in definition artificially reduced the symptomatic group to 2.4%. Thus, we aren’t comparing apples to apples, with respect to prior data. More importantly, is the fact that 27% of patients bled in their head! This ICH rate is 10% higher that in the placebo group. At least this data is consistent with most recently published articles on this topic (Lansberg MG, Bluhmki E, Thijs VN Efficacy and safety of tissue plasminogen activator 3 to 4.5 hours after acute ischemic stroke: a metaanalysis.Stroke. 2009 Jul;40(7):2438-41.).
However, If this data is true, and ischemic stroke patients are so likely to experience ICH, then why would we consider giving a systemic thrombolytic to any of them?
The investigators report that expanding the time window provided benefit to “substantially more patients,” without increasing the rate of intracranial hemorrhage (ICH). The odds ratio for a “favorable outcome” at 90 days was 1.34. An odds ratio of 1 would mean that the treatment group and the placebo group had the same chance to of experiencing the favorable outcome at 90 days. Thus, 1.34 really isn’t much benefit over placebo at all. Although they reported that the intracranial hemorrhage rate was higher in the treatment group, they stated that it was the same as noted in previous studies for patients treated within three hours. What is not widely publicized is that the investigators changed the definition of symptomatic ICH. They stated that in order to be considered symptomatic, it must be identified as the predominant cause of the deterioration. Nonetheless, they report an overall ICH rate of 27.0% vs. 17.6% for placebo. The change in definition artificially reduced the symptomatic group to 2.4%. Thus, we aren’t comparing apples to apples, with respect to prior data. More importantly, is the fact that 27% of patients bled in their head! This ICH rate is 10% higher that in the placebo group. At least this data is consistent with most recently published articles on this topic (Lansberg MG, Bluhmki E, Thijs VN Efficacy and safety of tissue plasminogen activator 3 to 4.5 hours after acute ischemic stroke: a metaanalysis.Stroke. 2009 Jul;40(7):2438-41.).
However, If this data is true, and ischemic stroke patients are so likely to experience ICH, then why would we consider giving a systemic thrombolytic to any of them?
Based on the ECASS 3 definitions, the symptomatic ICH rate is no
greater than that previously reported. However, what they really
proved, once again, is that t-PA causes a lot of intracranial
hemorrhages with marginal benefit at best.
Kevin Klauer, DO, is the Editor in Chief of Emergency Physicians Monthly
Continue to next page for the counter argument by Chris Carpenter, MD, MSc
Counter argument: YES
In ECASS III, Benefits Outweigh Risk
by Christopher Carpenter, MD, MSc
For ischemic stroke care, a practice change did occur in 1995 based upon a single trial of 624 subjects (NINDS). Certainly, practice-changing results need to be verified in heterogeneous patient and practice environments. However, would it truly be ethical to subject 100,000 patients to a study-intervention when a priori power calculations suggest that 320 subjects are sufficient for 95% power as was the case with NINDS? I think not and research ethicists would concur (Howard, Detsky, Schriger). What about asymptomatic and symptomatic ICH? The authors should and do report both rates, but is an asymptomatic ICH truly significant to patients or physicians? Personally, I’d rather have neither, but given the choice of a 30% reduced chance of good functional recovery from an ischemic stroke with an increased risk of a bleed into my brain that I never even know occurred versus poor functional recovery with a fairly high risk of intracranial hemorrhage anyway – I’ll take the risk. In fact a study published in July 2009 re-analyzing the ECASS III data suggests that the NNT and NNH may favor t-PA even more strongly than described above (NNT = 6, NNH = 37) and a separate meta-analysis of all randomized trial data evaluating treatment within the 3- to 4.5-hour window also favors t-PA. The ECASS III authors did change the definition of symptomatic ICH. They also painstakingly reported their results using the definition of symptomatic ICH from ECASS II, SITS-MOST, and NINDS (NNH = 32, 58, 22 respectively) so that readers can determine the impact of this altered definition.
Were a portion of ECASS III recoveries among those who received t-PA simply misdiagnosed TIA’s? Perhaps, but in a randomized controlled trial equal numbers of subjects in the treatment and control group would experience this recovery as long as they were appropriately randomized. If the same proportion of patients in the t-PA and placebo groups experience spontaneous recovery (i.e. not as a result of the t-PA) then the resulting difference cannot be attributed to TIA-related recovery. The NINDS trial reported that 2% of the placebo group had complete resolution of symptoms suggesting that few cases of TIA were mislabeled as a stroke.
Did the negative ATLANTIS trial delay publication for 6-years? In fact, the ATLANTIS trial recruited patients from December 1993 through July 1998 and published their findings in the 3- to 5-hour window group (including 31/613 subjects treated in under 3-hours) in December 1999 not 6-years after the trial completion. While prior studies have failed to demonstrate a significant benefit for thrombolysis in the 3-6 hour window, further analysis of these negative trials consistently suggests a reasonable benefit in the 3-4.5 hour window. The Cleveland experience was one of the negative t-PA trials, but after initiating a stroke quality improvement program they lowered their institutional intracranial hemorrhage rates to those reported in NINDS. Would another Cleveland trial now demonstrate significantly better outcomes in appropriate stroke patients treated with t-PA within 4.5 hours? What about community hospitals lacking an ever-ready stroke team? Even in the absence of an acute stroke team, EPs can and everyday do treat stroke patients with t-PA resulting in rare (7%; 95% CI 3%-12%) misdiagnosis rates or hemorrhagic complications. Furthermore, the Safe Implementation of Thrombolysis in Stroke-Monitoring (SITS-MOST) trial demonstrated that alteplase is efficacious (safe and equally effective) in real-world (non-research community hospital) settings outside the research realm. Strong collaborative relationships between community and academic centers can optimize safe and effective acute stroke management in appropriate patients. Cunningham recently developed an Excel-based decision aid to facilitate risk-benefit discussions with stroke patients and families that includes an outcomes wheel.
In 2009, stroke remains the leading cause of long-term morbidity in the US and the third most common cause of mortality. Lacking effective alternatives, EM providers need to be cognizant but wary of clinical nihilism and skewed prognostication in deciding to accept or reject current research findings and professional society recommendations. Optimally, the ECASS III results will be expeditiously replicated in future studies, but this currently represents the best-evidence by which to treat (otherwise untreatable) ischemic strokes and ought to be discussed with patients and families in conjunction with a local neurological emergencies protocol.
Christopher Carpenter is EPM’s Chief Clinical Editor