Thrombolysis for CVA ==================== þ Comments on studies: - The results overall showed no significant benefit from thrombolysis so they went back and then found the subgroup less than 3 hours to be significant. This is very poor methodology. - Harvey's point about infarction is crucial to the analysis. Patients with infarction as well as those with haemorrhage are contraindicated for lysis. When they went back and reviewed the data a significant proportion of patients had evidence of infarction on the original scan when reviewed retrospectively (ie after they had already been treated). This was referred to as a "protocol violation". Now comes the really good part: on the basis that these patients should not have received treatment an analysis was performed on only those patients who fitted criteria (ie no infarction on the retrospectoscope) and, not surprisingly, the results suddenly looked good. This masssaging of data is just incredible. - There is little doubt our radiology collegues will become more attuned to picking acute infarction in time without the retrospectoscope but, until that happens, I'm not sure we have a lot of solid evidence to go on at the moment. Encouraging data, yes. Convincing evidence, no. --Dr Garry Wilkes MBBS FACEM Department of Emergency Medicine Royal Brisbane Hospital Queensland, Australia Email G.Wilkes@mailbox.uq.oz.au The main finding is: tPA given to all stroke patients (CT negative for hemorrhage) within 3 hours of onset of symptoms, increased the percentage without deficit at 3 months from 30% to 40%. This is the answer to the question they set out to answer in this very well designed and apparently well executed study. Yes they also found that it substantially increased the rate of hemorrhagic conversion, but even taking those into consideration, patients with any type of neurodeficit had a smaller chance of dying and a better chance of having a smaller deficit if they received tPA. Much of the confusion comes from people getting the conclusive results of the NINDS study confused with the confusing and equivocal results of the ECASS (2). The ECASS data could be viewed as supporting tPA as well, but like Dorothy in the Wizard of Oz, you really had to close your eyes and beleive. Dr. Harvey Louzon wrote: > I believe that another exclusion criteria is not only the presence of > bleed on CT scan but evidence for infarction. I believe that this > distiction explains the difference between the ECASS and NINDS > studies, although I don't have them in front of me and I could be > wrong about this. You're wrong about this. NINDS did NOT exclude patients with evidence for infarction. The only CT exclusion criteria was intracerebral hemorrhage, and the radiology reveiw panel rarely disagreed with the initial reading. ECASS did exclude patients with evidence of infarction and this was one of the major methodological flaws that makes it confusing and hard to interpret. Dr Garry Wilkes wrote: > The results overall showed no significant benefit from thrombolysis so > they went back and then found the subgroup less than 3 hours to be > significant. This is very poor methodology. That would be very poor methodology, but that is not the way it was done. The NINDS study never included any patients over 3 hours out. The positive results come from the whole study group. I agree that subgroup analysis is usually garbage, but subgroup analysis from NINDS has not been published (although it almost undoubtly will be). > Harvey's point about infarction is crucial to the analysis. Patients > with infarction as well as those with haemorrhage are contraindicated > for lysis. Not according to the NINDS, you're thinking of ECASS (as well as several of the stroke streptokinase trials). This theoretical contraindication would be a stupid exclusion criteria anyway if you think about it. I mean they should ALL have strokes, and they should all show strokes on CT eventually. Visualizing them early is a function of the CT scan, not the patient. Criteria for treatment should be based on the history and physiology of the patient, not the resolution of the hardware. > When they went back and reviewed the data a significant proportion of > patients had evidence of infarction on the original scan when reviewed > retrospectively (ie after they had already been treated). This was > referred to as a "protocol violation". Now comes the really good part: > on the basis that these patients should not have received treatment an > analysis was performed on only those patients who fitted criteria (ie > no infarction on the retrospectoscope) and, not surprisingly, the > results suddenly looked good. This masssaging of data is just > incredible. This is the ECASS problem, it has nothing to do with NINDS study. The number of protocol violations in ECASS was ridiculous (109 of 620) they included several errors including many "misread" CT scans. They did the right statistics for a trial with this problem, ie., separate analysis by "intention to treat" (what they were really randomized to and received) and by "target population" (what happened when the people who should have been studied, got what they were supposed to). But after all the craziness, who knows what it really means? > There is little doubt our radiology collegues will become more attuned > to picking acute infarction in time without the retrospectoscope but, > until that happens, I'm not sure we have a lot of solid evidence to go > on at the moment. Encouraging data, yes. Convincing evidence, no. The NINDS avoided the whole problem by NOT including signs of stroke on CT as an exclusion criteria. ECASS was encouraging, but NINDS is convincing. Joe Bocka wrote: > Even though we can usually get a CT done and read within 30 mins 24 > hrs per day on ave, we do not see the risks outweighing the benefits. > Waiting to see more than Time and Women's Home Journal being so > positive, especially since we see a large amount of strokes. I think not using the only drug with FDA approval for the therapy of acute stroke in that small number of eligible patients is pretty tenuous. Waiting for more studies is not in those patients best interest. Furthermore you may be waiting for a long long time. Another study will probably NOT be done. NIH already paid big bucks to answer this question, and the NINDS is pretty clear that it feels that it has the answer. Furthermore the NIH and most IRB's will not find another placebo controlled trial ethically permissible. They next study will only occur when another treatment is tested against tPA in a large multicenter way. Probably the best excuse for not providing this therapy is institutional inability to get the patient to CT and have the CT read in the very short window of opportunity. I have been trying to understand why so many physicians seem so slow to believe and incorporate these results. I think some of the reasons that skepticism seems to run so deep are expressed by Dr. John Meade who wrote: > I, too, remain unconvinced regarding the use of tPA for CVA... > I have yet to see a study which shows any significant benefit to > anyone other than Genentech, for the use of any thrombolytic in the > setting of CVA. Two issues here. Many people saw the confusing and inconclusive ECASS results in JAMA last October (2), but did not get around to thoroughly reviewing the powerful results of the NINDS study in the NEJM last December (1). Second, Genetech has seemed so greedy and exploitative of data in the past that everyone is skeptical of results that benefit them. The NINDS study was funded by NIH without any drug company money or support. ECASS had Boehringer Ingelheim money, but nothing from Genetech. I think people just have to open their minds to the idea that even though they don't like Genetech, tPA just may be a good drug for some patients. > I also see the unfortunate sequelae of the news people latching on to > something for a headline, with many of us feeling the need to "do > something", for fear of either missing something beneficial to > patients, or appearing to be a bad clinician stuck in the mud of time. Another potent bias against this data for many physicians has been backlash to its delivery in the popular press. Since they always blow everything out of proportion, this must be another nothing study suffering gross exaggeration. I don't like the sensationalistic lay reporting of most "advances" either, but we must evaluate data as presented only in our literature. Sometimes it really will be compelling, as I think it is in this case. Sorry for going on like this, but I think as Emergency Physicians we really need to step up to the challenge and start treating stroke as an emergency. 1) NEJM 1995;333(24):1581-87 2) JAMA 1995;274(13):1017-25 Robert, ---------------------------------------------------------------------------- Robert Silbergleit, MD office 202 994-3921 Research Fellow lab 202 994-0927 Department of Emergency Medicine fax 202 994-3924 George Washington University home 202 333-7557 ---------------------------------------------------------------------------- Let me note at the outset, tho, that this is too large a topic to summarize in a couple of paragraphs -- which is why Rick Bukata and I have written a 2-part essay about it in EMA (11 and 12/96). Nevertheless, here are a few points I would make: 1) NINDS is far from "convincing," given that it is the only one of 5 randomized trials to find any benefit from the use of lytics. This may be because it's the only one to use the right protocol/ drug/ dosage, etc., but it may also be because when you do multiple trials of even a useless therapy, occasionally one will appear to be positive. In addition, the "30 to 40%" (relative) benefit is actually an ABSOLUTE benefit of 12%, which means you need to treat more than 8 patients for one to get anything from it. Since there are probably more than 100 stroke patients for every one who qualifies, by the NINDS criteria (they've informally "estimated" that up to 5% qualified in their trial [which raises the question as to why they don't have, or report, the exact #s], but most other data supports a 0.5-1% rate, or even less, and certainly one would expect it to be less in community hospitals not participating in a funded study than whatever it was in NINDS), that means you might benefit 1 stroke patient out of 150-1,500 (or, I'd bet, out of even more). Hardly the stuff that dreams are made of! Particularly when you consider that "mistakenly" giving it to (and harming) just a few patients who don't qualify could easily overwhelm the tiny benefits found in NINDS (IF you accept its results AT FACE VALUE). This would include patients treated after 3 hours (assoc. with dramatically increased mortality in all 4 studies in which such pts were included), as well as "stroke mimics" (initially thought to have stroke but actually with tox/metabolic, infectious, or p-ictal cause of presentation), who may comprise up to 20% of ED-diagnosed "strokes" (even though the diagnoses in the study were made by a SPECIALIZED STROKE INVESTIGATION TEAM! - Libman, Arch Neurol 11/95). 2) ECASS is "encouraging" only if you're a mortician. Their post-hoc manipulation of their own terribly negative data is incredible, and far from being "the right statistics for a trial with this problem," was actually an extraordinary violation of the most basic rules of research. "Intention to treat" refers to results according to TREATMENT ASSIGNMENT, which for various reasons is not always the same as the ACTUAL TREATMENT RECEIVED. In this study that distinction was never an issue -- the authors' data manipulating had to do with distinguishing both TREATMENT ASSIGNMENT/ ACTUAL TREATMENT RECEIVED (which were the same), from "WHEN WE DISCOVERED THEY DIED WE WENT BACK AND FOUND A REASON TO EXCLUDE THEM." !!!! And, by the way, these preposterous after-the-fact exclusions were based not only on "misreading" of the initial CTs (by study neuroradiologists, in fact, who obviously aren't as competent at reading CTs as your average radiologist or emerg physician), but on "presence of signs of infarction"; if (as Dr. Silbergleit states) this is a "stupid exclusion criteria" (sic - he meant 'criterion'), then even their incredible excuse (for saying their results weren't so bad) is out the window. So on what possible basis do they (as they do) or Dr. Silbergleit (as he does) conclude that ECASS is "encouraging"? !!!??? Obviously there are other issues that could be addressed (such as the retrospective distorted sub-group analysis by the ASK investigators, to which Garry Wilkes was apparently referring, and the distortions in the popular press, and the precipitous approval by the FDA, and the unjustified assumption that tPA must be better than STK, and etc. etc.), but let me end by quoting the MAST-E and MAST-I authors, who agreed with Dr. Silbergleit that perhaps "another placebo controlled trial (wouldn't be) ethically permissible" -- for a somewhat different reason, though. They think it might well be unethical to subject patients to the risk of being randomized to the treatment arm, given so much negative data already available! I, for one, will "step up to the challenge and start treating stroke as an emergency" as soon as there's good evidence that I have something to treat it with. Jerry Hoffman