/ Stroke Cerebrouasc Dis 1994;4:1-3© 1994 National Stroke Association
Editorial: Extracranial-Intracranial Bypass StudyEight Years Later
James 1. Ausman, M.D., Ph.D.
Criticisms of the extracranial-intracranial (EC-IC)bypass study (I) are not popular. It is claimed that therandomized study was outstanding and that the results are unassailable. It is difficult, if not impossible,to have any articles dealing with the subject acceptedfor publication because only a randomized study isdeemed acceptable.
Let us examine the results of this study now 8 yearslater. There are still several major crit icisms that limitthe applicability of the data from the studies. First, thestudy only required bilateral carotid angiography inthe evaluation of patients. [This is a similar flaw in theNASCET Trial (2).] Vertebral angiography was notperformed, and no evaluation of intracranial collateral circulation was reported. Ask a cardiologist toinject only one coronary artery in the evaluation ofischemic cardiopathy and you will be regarded as afool. Collateral circulation is critical in the evaluationof the state of the myocardium. It is also true for thebrain, although as yet there are no easy means toquantitate it or evaluate it. Is a patient with an occluded left internal carotid artery, with a patent contralateral internal carotid, vertebrals, basilar, and withlarge posterior and anterior communicating arteriesat the same risk as one with an isolated hemispherewith no collaterals and highly stenotic ipsilateral carotid siphon? Or as another with an occluded internalcarotid ipsilateral to the symptomatic hemisphere, noposterior communicating arteries, and with only collateral circulation from an 80% stenotic Al and someophthalmic flow? I believe the answer is "no"; theyare not the same. Furthermore, it would be virtually
From the Department of Neurosurgery, University of Illinois at Chicago, Chicago, It, U.S.A.
Address correspondence and reprint requests to Dr. ]. I.Ausman at Department of Neurosurgery, University of Illinois at Chicago, 912 S. Wood Street, Chicago, IL 60612,U.S.A.
impossible to find 100 similar patients to evaluate inany study. Atherosclerosis is a disseminated disease,the lesions of which have an infinite number of permutations and combinations making standardizationimpossible.
The second major flaw in the study is in the execution of the study itself and its selection process. Thestudy conclusions are valid only for a stated population in which bilateral carotid angiography is performed and collaterals not evaluated with the admission symptoms as stated. The conclusions cannot begeneralized to the total population of patients withcerebral ischemia but only to a similar group of patients. Ideally, in an appropriate randomized study, arandom sample from the universe of patients with thedisease criteria being studied is selected. The randomsample is divided into two populations and then subjected to the same treatments with the exception ofone variable dividing the random group into treatment A and treatment B. Both sample populations arefollowed to an end point and the results compared. Inthe EC-IC bypass study, only one of three patients(and one of four in U.S. clinical centers) in the samplefrom the universe being studied was subjected to randomization and followed to an endpoint. Two of three(or, in the United States, three of four) candidateswere preselected for surgery, presumably on a biasthat they were candidates who would benefit betterfrom surgery than medical management (3). This isakin to having a laboratory assistant perform a studyon a random sample of mice from a universal population of mice. The assistant selects only the spottedmice for randomization and does not include thewhite mice for randomization because of his bias. Hereports the results to you. The selected sample ofspotted mice is not a random sample of the universe.Thus, the conclusions from the experiment cannot begeneralized to the entire population of mice and thusmust be reported to be only on a selected sample with
/ STROKE CEREBROVASC DIS, VOL. 4, NO.1, 1994 1
].I.AUSMAN
bias or rejected. Thus, on this basis, the results cannotbe generalized to the population at large. The sampleof patients in the bypass study is a biased and selectedsample of the universe of patients with cerebral ischemia. The results of the study sample cannot be generalized to the ischemic population as a whole as hasbeen done.
The bypass study cost $9-11 million. It is beingdefended by both the government and the investigators. Although I was a participant in the study, I withdrew after 1 year because I could not honestly randomize my patients. I was told that the 10 patientswhom we contributed to the study were similar tothose of others and, thus, would be included, although we had selected patients for surgery outsidethe study whom we thought were excellent candidates for a variety of reasons. To my knowledge, thispattern was duplicated by a number of investigatorsin the trial. Truly, those investigators who broke thecode of the study and operated outside of it are dishonest participants. By the way, when the principalinvestigators were visited by a special committee ofthe AANS and asked to produce data from theirstudy on the number of patients operated outside thetrial and those randomized, the investigators couldnot produce the same data on four separate occasions as was published in the New England Journal ofMedicine (4). Yet, these investigators were rewardedwith a second study, the NA5CET trial. Why? Wouldthis have happened to the average NIH applicant?
The third major flaw is obvious in the analogy to adrug trial. In the EC-IC bypass study, a single operation, superficial temporal artery (5TA) to middlecerebral artery (MCA) bypass, was performed as anexample of a single dose of a drug. Quantitation ofthe filling from the intracranial vessels was not published. We do not know whether the bypass filled allof the MCA vessels or only a few. What happened 3months to 1 year after the bypass to the function ofthe bypasses angiographically? Would large vein bypass grafts delivering higher flows or superficial temporal artery bypasses to the proximal branches of theMCA have provided "other doses" to be tested in thisdose-response curve? Do we discard a drug after atrial of one dose in which it is shown to be ineffective? Does this mean that the drug is useless in thetreatment of ischemia as was concluded and stated?No, and again, an improper conclusion was drawnfrom the data. How does the amount of flow delivered relate to the collateral circulation or its deficit?How can we quantify these two variables? At presentwe do not have a means to do this.
The fourth major flaw is with the scientific conclusions drawn from the data by the investigators. The
2 ] STROKE CEREBROVASC DIS, VOL. 4, NO.1, 1994
conclusion from the selected population of patients,who were chosen with presumed bias, was thatSTA-MCA bypass surgery was no better than medical management for cerebral ischemia in the carotidor MCA territories in the total (biased) randomsample or the subgroups. For the study to be repeated today with the same criteria, carotid angiography, and selection bias, 95 times out of 100 theresults would be the same. However, we have beentold that there are no indications for the use of 5TAMCA bypass in cerebral ischemia. Medicare funding for the procedure was stopped for this reason.Randomized studies have been used to mesmerizethe physicians into abandoning common sense andhave been promoted as the ultimate answer to theclinical problem while disparaging all other observations as anecdotes. Conclusions from a selectedsample of patients are only valid for that sample andnot generalizable to the population at large withoutthe appropriate qualifications stated in this editorialor unless selected on a truly random basis and followed to the end point.
The concept of randomization is fine, but a randomized study is only as good as the data enteredand the conclusions drawn, which all depend on theinvestigators doing the study. There is nothingwrong with computers, only the data entered andretrieved. Furthermore, the investigators must stillevaluate the outcome as sensible or not. Eight yearslater, it is difficult for neurosurgeons and neurologists around the world to accept the conclusions ofthe bypass study because the results do not conformto "anecdotal observations," which are obviouslydisregarded as valueless when compared to randomization by the investigators. However, when compared to the flawed conduct and conclusions of thisrandomized study, the observations assume evengreater value. Moreover, the multiple observationsby neurosurgeons and neurologists of naturally occurring collateral channels, "bypasses" (which are
. seen on angiography and which provide collateral ofobvious value in supporting cerebral circulation),cannot be dismissed. The conclusions of the bypassstudy only make sense if they are restricted to thepopulation selected. They cannot be generalized.
So, although it is unpopular to criticize this randomized study, the facts are as above. As a result ofthe flawed conclusions, many young investigatorshave been discouraged from working in the area. Anarea of negativism prevails around the world on surgery for ischemia. Medicare funding for the surgicalprocedure has been withdrawn and patients havebeen deprived of potentially valuable treatment forischemia. The public has been convinced that a
needed operation has been stopped and money hasbeen saved in the escalating costs of health care.Have all of these steps been based on valid conclusions? I believe not.
References1. The EC/IC Bypass Study Group. Failure of extracra
nial-intracranial arterial bypass to reduce the risk ofischemic stroke. Results of an international randomized trial. N Engl / Med 1985;313:1191-1200.
EC-IC BYPASS STUDY EIGHT YEARS LATER
2. North American Symptomatic Carotid Endarterectomy Trial Collaborators. Beneficial effect of carotidendarterectomy in symptomatic patients with highgrade carotid stenosis. N Ellgl / Med 1991;325:44553.
3. Sundt TM Jr. Was the international randomized trialof extra cranial-intracranial bypass representative ofthe population at risk? N Engl / Med 1987;316:8146.
4. ReIman AS. The extra cranial-intracranial arterialbypass study. What have we learned? N Engl J Med1987;316:809-10.
/ STROKE CEREBROVASC DIS, VOL. 4, NO.1, 1994 3