Extracranial-Intracranial Bypass To Reduce the Risk of Ischemic Stroke: Health Technology Assessment Reports, 1990 Number 6

Foreword

The Office of Health Technology Assessment (OHTA) evaluates the risks, benefits, and clinical effectiveness of new or unestablished medical technologies that are being considered for coverage under Medicare. These assessments are performed at the request of the Health Care Financing Administration (HCFA). They are the basis for recommendations to HCFA regarding coverage policy decisions under Medicare.

Questions about Medicare coverage for certain health care technologies are directed to HCFA by such interested parties as insurers, manufacturers, Medicare contractors, and practitioners. Those questions of a medical, scientific, or technical nature are formally referred to OHTA for assessment.

OHTA's assessment process includes a comprehensive review of the medical literature and emphasizes broad and open participation from within and outside the Federal Government. A range of expert advice is obtained by widely publicizing the plans for conducting the assessment through publication of an announcement in the Federal Register and solicitation of input from Federal agencies, medical specialty societies, insurers, and manufacturers. The involvement of these experts helps assure inclusion of the experienced and varying viewpoints needed to round out the data derived from individual scientific studies in the medical literature.

After OHTA receives information from experts and the scientific literature, the results are analyzed and synthesized into an assessment report. Each report represents a detailed analysis of the risks, clinical effectiveness, and uses of new or unestablished medical technologies considered for Medicare coverage. These Health Technology Assessment Reports form the basis for the Public Health Service recommendations to HCFA and are disseminated widely. Individual reports are available to the public once HCFA has made a coverage decision regarding the subject technology.

OHTA is one component of the Agency for Health Care Policy and Research (AHCPR), Public Health Service, Department of Health and Human Services.

• Thomas V. Holohan, M.D.

• Director

• Office of Health Technology Assessment

• J. Jarrett Clinton, M.D.

• Acting Administrator

• Copies may be obtained at no charge from:

• Division of Information and Publications

• AHCPR

• 5600 Fishers Lane

• Parklawn Building, Room 18-12

• Rockville, MD 20857; (301) 443-4100

Introduction and Rationale

Extracranial-intracranial arterial bypass (EC-IC bypass), occasionally referred to as superficial temporal artery-middle cerebral artery bypass (STA-MCA bypass), is a surgical procedure in which the superficial temporal artery is anastomosed to the middle cerebral artery on the surface of the brain; it was developed and initially described by Yasargil in 1969.(1) Atheromata in cerebral vessels and vessels leading to the brain are considered to be a significant cause of reduced perfusion predicated to result in transient ischemic attack (TIA), reversible ischemic neurologic deficit (RIND), and stroke. Occlusive lesions high in the internal carotid system or in the middle cerebral artery are inaccessible to direct surgical intervention, and the EC-IC bypass procedure has been advocated as a prophylactic intervention to increase perfusion of the brain thereby minimizing the likelihood of cerebral ischemia with resultant stroke. The surgery has been variously employed for patients with TIA, RIND, stroke-in-evolution, and completed stroke with mild or severe neurologic deficit.

Review of Available Information and Discussion

Although a number of studies of EC-IC bypass have been published, only one of the reports(2). included randomized control groups. Table 1 contains data from 13 studies reported in the medical literature that are representative of those supporting the effectiveness of the operation.(3-15) These reports provide information on a total of 1,464 patients subjected to EC-IC bypass surgery. One of these studies included a nonrandomized control group of unoperated patients; the remaining 12 studies reported results in surgical case series, with some of the authors comparing their results with historical data.

Determination of the incidence of complications of surgery as well as the occurrence of stroke during postoperative observation periods is critical to the evaluation of the clinical utility and prophylactic effectiveness of surgical Such information permits a logical conclusion as to the benefit of surgical intervention compared with alternative treatment or no treatment. This is particularly important regarding EC-IC bypass surgery in view of the fact that published criticisms of the single prospective randomized trial(2). of this procedure created significant controversy and uncertainty in the medical community. In this assessment, the information from various surgical case series of EC-IC bypass is compared with that reported in the EC-IC prospectively randomized cooperative trial.(2) That trial defined study "events" as deaths from all causes that occurred between randomization and 30 days following bypass, and fatal or nonfatal stroke occurring during the entire postoperative observation period.(16) Operative deaths from all causes were recorded so as to include any excess perioperative risk. For purposes of uniformity, the term "events" will be used in the same sense as in the EC-IC cooperative trial, i.e., all operative deaths plus all strokes.

Several points regarding the data as represented in Table 1 are instructive. These surgical series all reported perioperative death and nonfatal stroke rates; those rates ranged from 0-7.7, and 0-10.3, respectively. Using that data, the mean perioperative death rate for all 13 studies may be calculated as 1.8, and the mean perioperative nonfatal stroke rate, 2.5. Only six of the 13 studies, comprising a total of 254 patients, included followup data that permitted calculation of postoperative stroke rates; those studies reported durations of followup varying from 12 -54 months. The annualized rates for operative death plus all strokes (events) for each of these studies were calculated by dividing the ratio of events to study sample size by the mean duration of followup specified, and averaging on a 12-month basis. The rates so obtained ranged from 2-27.5. A combined annualized event rate for all six studies that included followup data was calculated based upon reported durations of followup, sample size in each study, and the total number of operative deaths plus all strokes (Table 2). The mean number of patient-months of followup for each study (mean number of months of followup times sample size) was summed over all six reports to obtain a cumulative weighted average patient-months of observation. Division of the total number of events by this figure gives the rate of events per patient-month, which may then be annualized. There were 32 operative deaths plus strokes recorded over 9,262 patient-months of observation, which is equivalent to an annualized event rate of 4.1.

The results represented in Table 1 were frequently claimed by the various authors to be superior to historical controls in which annual stroke rates in symptomatic patients were variously cited as 3-4, (6,9). 5-15, (5,6,8,9). or as high as 25-60.(15) However, several advocates of the operation pointed out that the variability of the natural history of cerebrovascular disease and the impact of medical therapy made it difficult to retrospectively evaluate the effectiveness of EC-IC bypass.(8,9) Others noted that a prospective randomized trial would be necessary to determine the clinical utility of this surgery.(5,6,8,9).

None of the 13 surgical series described in Table 1 included a randomized nonoperated control group, and therefore are potentially subject to unspecifiable sources of selection bias. The durations of followup were often irregularly documented and in seven studies were not reported or were insufficiently precise to reliably determine the annual postoperative stroke rate. Patient selection criteria for the most part were not presented in sufficient detail to allow for critical analyses of the indications for selection of surgery.

There were little data provided regarding the concomitant use of acetylsalicylic acid, which has been acknowledged to have some efficacy in reducing the incidence of stroke in patients with TIA, (17). and it is impossible to determine from the data the contribution of prophylactic surgery versus medical therapy. The separation of the effects of operation from those of antiplatelet medication is a difficult yet central question, particularly when benefit is judged relative to historical controls who may or may not have received the drug. However, only four of the 15 reports specified what fraction of surgical patients were concomitantly treated with antiplatelet drugs;(3,8,11,14). five commented that "some," "many," or "most" patients were so treated, and four did not address the question at all. Acetylsalicylic acid, sulfinpyrazone, and dipyridamole were variously employed.

A complicating factor in attempting to draw conclusions from the data is that four of the reports(5,6,8,9). representing 680 cases and covering the period 1977-85 may have included patients from a single medical center. It is unclear whether or not some patients were multiply reported.

The study by Powers(3). cited in Table 1, which included nonrandomized control groups, described followup periods for operated cases of 12 months. The data indicated a higher annual stroke rate for the surgical than for the control patients (27.5 versus 0).

Several reports included information that could legitimately raise questions regarding the ultimate clinical utility of EC-IC bypass. Massimo et al(4). reviewed cases operated for symptoms persisting despite medical therapy. Although they concluded that surgery should be reserved for recurrence of symptoms while patients were under medical therapy, they were unable to determine why vessel occlusions occurring after bypass were not associated with neurological impairment. They also noted that direct comparison of their results with nonoperative therapy was not possible. Heilbrun et al(18). studied symptomatic cases with internal carotid artery occlusion and concluded operation was not indicated for the majority of such patients. They noted that if new ischemic events in this group were in fact less common than had been postulated on the basis of historical evidence (cited by the authors as 5-10), benefit of EC-IC bypass "should appear to be statistically significant when in fact a sham operation would appear as beneficial."

Two reports studied cerebral blood flow (CBF) utilizing (133)Xenon inhalation before and after EC-IC bypass. Vorstrup et al(7). found only two of 22 patients exhibited improved CBF following surgery. Younkin et al(19). believed that their results in selected patients were "probably the best possible with EC-IC surgery," but were unable to show a statistically significant increase in CBF in patients operated for TIA. They reported that the postoperative incidence of TIA and stroke was lower than expected on the basis of referenced epidemiologic studies, but pointed out that if such clinical benefit accrued from surgery it did so without a concomitant change in CBF. A prospective investigation by Tanahashi et al, (20). also using (133)Xenon tracer, compared CBF following bypass in 38 operated patients with 22 medically treated nonrandomized controls with similar arteriographic findings and clinical symptoms. The authors found that although flow transiently increased following operation and was significantly higher for surgery patients than controls at 24 months, there were no significant differences at 3, 6, 12, 36, or 48 months. They also noted that clinical analyses did not indicate any significant differences in outcome between the two groups. The authors were pessimistic regarding long-term benefits for patients undergoing bypass, but believed that such questions would be answered by the then ongoing cooperative randomized multicenter trial.(2)

The only published, prospective, randomized, controlled trial assessing the effectiveness of EC-IC bypass to reduce the risk of ischemic stroke was begun in 1977 and the final report published in 1985.(2) The investigation was a multicenter, international, cooperative trial incorporating data from 71 medical centers. The targets for sample size and duration of the study were selected so as to permit the demonstration of a 33 net surgical reduction in the 5-year risk of fatal and nonfatal stroke based upon a significance level (alpha) of 0.05, and a statistical power, or probability of obtaining a true positive result, of 0.90 (power = 1.0 -beta, where beta was selected as 0.10). Those criteria were chosen following the agreement that such a degree of risk reduction constituted a clinically significant benefit that would justify advocating the surgery. The protocol specified that eligible patients were those who within 3 months prior to entry had experienced one or more TIAs or minor completed strokes and had radiographically documented stenosis or occlusion of the middle cerebral artery, stenosis of the internal carotid artery above the C-2 vertebral body, or occlusion of the internal carotid artery. Randomization to operative or nonoperative treatment was accomplished through centralized methods centers. Acetylsalicylic acid was prescribed in a dose of 325 mg four times daily for all patients throughout the trial unless such medication was contraindicated or not tolerated. The primary study events were the occurrence of fatal or nonfatal stroke; nonfatal strokes were rated on a previously published stroke severity scale.(16)

Of the 1,377 patients entered in the trial, 714 were randomized to nonoperative (medical) therapy and 663 to surgery. Ninety-three percent had minimal or no impairment at the time of randomization. The authors noted that there were no significant initial differences between treatment arms in terms of TIA or minor stroke incidence, medications, blood pressure levels, or location of vascular lesions. The mean duration of followup was 55.8 months, with a range of 28-90 months. Of the 663 patients randomized to surgery, 98 underwent the operation; 92 of those patients also underwent postoperative angiography, which revealed a graft patency rate of 96. Acetylsalicylic acid in the protocol dosage was taken by 75 of the medical cases and 74 of the surgery patients.

Surgery was accomplished within an average of 9 days following randomization. The operative mortality and perioperative nonfatal major stroke rates were 1.1 and 3.4 respectively (seven fatalities, 23 nonfatal major strokes), were major stroke was defined as one that incurs major impairment in one or more of the domains of swallowing, self-care, ambulation, communication, or comprehension. (16). During the same time period (39 days from randomization), 1.3 of the medically treated patients suffered major strokes (nine strokes); no stroke-related deaths occurred. Over the entire period of observation fatal and nonfatal major and minor strokes occurred earlier and with greater frequency in the patients assigned to surgery. Baseline factors which were shown to have no effect on the likelihood of stroke included the geographic region and size of the participating center; the site, type, and location of the vascular lesion(s); the time from first symptom to treatment; history of prior stroke; and comorbid conditions. Secondary analyses or subgroup analyses were specified prior to data collection, and were performed to identify subgroup trends for formulation of hypotheses for further study. These revealed no subcategory of patients for whom surgery was statistically beneficial; however, patients with severe middle cerebral artery stenoses and internal carotid occlusion with continuing symptoms fared significantly worse with surgical versus medical therapy. EC-IC bypass also appeared to have no beneficial effect upon the final functional outcome of surviving patients. The investigators reanalyzed data eliminating all strokes that occurred after randomization but prior to surgery and patients randomized to surgery who were never operated. In what they termed the "best case scenario" for surgical therapy, there was no statistically significant benefit to surgery. Further, there was no apparent benefit of a well-functioning bypass versus a low-flow or nonpatent anastomosis. The effect of surgery upon the frequency of TIAs was evaluated by comparing the frequency of attacks in the 3-month period prior to entry with the frequency recorded in the last 3-month period of the first year of followup. Excluding those patients with stroke or death in the first year of followup (13.8 of medical patients and 20.3 of surgical patients) the frequency of TIAs did not appear to be reduced by operation; 80 of the medical and 77 of the surgical patients exhibited at least a halving of the frequency of TIAs in the first year following randomization. In addition, surgery was not superior to medical therapy in reducing the stroke rate of patients with frequent or "crescendo" TIAs.

Over the course of the EC-IC cooperative study, 205 fatal and nonfatal strokes occurred in the group of 714 patients treated medically and 205 in the 663 operated patients. Using a reported average duration of observation of 56 months, the annualized event rates may then be calculated as 6.2 for the medically treated patients and 6.6 for surgical patients.

Therefore, the cooperative multicenter trial of EC-IC bypass provided no evidence that the surgical procedure was of benefit for the prevention of ischemic stroke as had been claimed by proponents of the procedure. Indeed, patients tended to fare worse when subjected to surgery than when provided medical treatment alone. This large-scale prospective study was hailed by experts in the field soon after its publication(21,22). who called it "a remarkable piece of clinical investigation," and "a landmark trial." However, analyses critical of the study's methodology and conclusions soon began to appear in the literature. Awad and Spitzler(23). postulated that patients failing medical therapy might be more likely to benefit from surgery. In addition, they claimed that patients with hemodynamic compromise were not addressed by the study, while admitting that there were no "totally reliable criteria" to identify such groups. Further, they criticized the nonblinded character of the study, the fact that preoperative strokes were counted against the surgery arm, and the reporting of secondary analyses by the investigators. Finally, they questioned whether randomized patients adequately represented the population with symptomatic cerebrovascular/extracranial disease. Three case reports were appended to support their contentions. Day et al(24). argued that the number of preoperative strokes in the surgical patients indicated a particularly unstable group for whom a 9-day delay of surgery was unacceptable. They added their belief that the procedure had a sound basis and was indicated for an unspecified subset of patients. Ausman and Diaz(25). claimed the operation had been shown to be beneficial prior to the EC-IC trial, and that the trial entered patients at particularly high risk, citing case series providing lower postoperative stroke rates. They argued that still newer surgical techniques would provide superior results and criticized the trial report's discussion of surgical results in subgroups of patients (secondary analyses).

The principal investigators responded to these and similar criticisms in detail, (26,27). and several of their comments were particularly pertinent in view of subsequent critiques. Barnett et al(26). pointed out that all other published series were uncontrolled, and the trial provided comparable results to surgical series reports in terms of adverse surgical outcomes and anastomotic patency rates. They further noted that of the preoperative ischemic events occurring in the trial, seven were major strokes and three were fatal strokes; three of the strokes were secondary to repeated angiography performed at the surgeon's request, two followed endarterectomy done as a prelude to bypass, one followed an "urgent" nonvascular surgical procedure, and three strokes occurred the evening before or morning of the scheduled bypass surgery. Even when the preoperative strokes were eliminated from analysis, there was no demonstrable benefit of surgery. Concerning secondary analyses, the authors pointed out that prior to intiation of the trial, the Executive Committee (of which Dr. Ausman was a member) agreed that such analyses would serve only to identify trends in subgroups for the purpose of formulating hypotheses for further testing.

Subsequently, Sundt(28). questioned whether the patients enrolled were in fact representative of the population at risk for ischemic stroke. He based his critique on information obtained from his own telephone survey of participants, claiming that in 57 of 71 centers randomizing 601 cases to surgery, 2,572 patients were operated outside of the trial.

The American Association of Neurological Surgeons (AANS) appointed a committee to examine the study.(29) While the committee concluded that within the trial proper the operated patients did no better than the nonoperated patients, they were unable to determine the number of patients subjected to EC-IC bypass outside the trial, citing a number of 1,900 provided by the trial's principal investigators and "as many as" 3,000 reported by Sundt. The committee reported that it was not possible to determine with certainty whether the conclusions of the trial were thus compromised, and consequently, it was premature to conclude that the operation was never indicated "in stroke patients" or that the surgery "should no longer be reimbursed by third-party payers."

The principal investigators of the cooperative study addressed the committee's remarks in a companion article.(30) They pointed out that the issue was not the proportion of eligible patients entered in the trial, but whether subgroups that might particularly benefit from surgery were disproportionately excluded. The calculated that in the 52 centers contributing 86 of the total cases reported, the ratio of nonrandomized to randomized patients was 1:1, and that 75 of those not randomized were in fact ineligible for the study. They also reported that the otherwise eligible patients operated outside the trial represented less than two patients per center every 5 years. A second group (10 centers) provided 11 of the study patients but more than one-third of patients operated outside the trial, and the remaining nine centers had been inactive early in the trial, providing only 2 of the 1,377 study patients. When the study analysis was repeated for the first group of 52 centers alone, the outcome was unchanged. The principal investigators also reanalyzed the trial results for subcategories of patients believed by Sundt to be most likely to benefit from bypass (e.g., frequent ischemic events). In none of these subcategories did the surgical patients have outcomes superior to those of nonoperated patients. They concluded that in the sample from which the overwhelming proportion of patients comprising the study were selected, few otherwise eligible patients were operated outside the trial, and that reanalysis failed to identify any specific subgroups for which surgery was demonstrably superior to medical management.

However, the editor of the New England Journal of Medicine commented that although there was no evidence that the patients not randomized had different outcomes than those included in the trial, the question of generalizability of the trial results had not been resolved.(31) These dialogues(28-31). were published 16 months after the initial report of the cooperative multicenter trial, and such controversy created some degree of uncertainty in the medical community as to the applicability of the findings to clinical practice. It is important to note that little clinical data were available concerning the potentially eligible but nonrandomized patients; assumptions or assertions concerning their comparability with randomized patients appear to have been totally speculative.

The basis for much of the objection to the EC-IC trial appeared to be the proponents' belief in the merits of bypass surgery based on clinical experience reported in surgical series represented by Table 1. As noted above, those data allow some calculations of measures of central tendency, and the weighted mean surgical mortality may be calculated as 1.7, and the perioperative stroke rate as 2.5. The perioperative surgical mortality and stroke rates reported in the EC-IC trial were 1.1 and 3.4 respectively. The comparability of these figures and the graft patency rate of 96 in the EC-IC trial indicate that the technical quality of the surgery in the cooperative study was not inferior to that achieved in the best case series. The annualized rates for adverse events (calculated from time of randomization for the entire period of followup) for medical and surgical patients in the EC-IC trial were 6.2 and 6.6 respectively. The comparable mean annualized rate in the six surgical studies reporting postoperative events was 4.1 with a range of 2-27.5 (Table 1.). These figures do not permit the conclusion that there exists a clinically significant difference in clinical outcome between the surgical case series and the cooperative trial. Moreover, such difference in event rates as exists may be more apparent than real. Followup in the EC-IC trial included examination by a neurologist every 3 months for 5 years, and each suspected event was independently reviewed by both a neurologist and neurosurgeon. None of the reports of Table 1 included such careful postoperative evaluation. For example, several reports(9,11,13,14). stated that followup after postoperative discharge was by telephone or letter for patients who were unable to return for evaluation; the fraction of patients in whom followup was so accomplished could not be determined by the reader. Consequently, it is quite unlikely that stroke events occurred but were not detected in the EC-IC trial, whereas the same confidence cannot be placed in the cited case series.

A large surgical series report (400 patients) included the claim that EC-IC bypass improved the quality of life in operated patients, since the frequency of TIA was reduced following surgery.(6) However, the cooperative trial revealed that while 77 of the operated patients had at least a 50 reduction in TIA frequency at 1-year followup, 80 of the medical patients exhibited the same improvement.

The largest case series included 403 patients(14). and was published in conjunction with an accompanying paper.(32) That latter report was written by three of the same authors and was quite restrained in its assessment of the effectiveness of surgery. The authors selected 239 of the 403 operated patients and excluded those "whom we thought might have a less satisfactory outcome." This select group was compared with 130 patients previously seen at the same institution and subjected to angiography but not operated; the expected stroke rates for both groups were based upon those observed in the regional population during the period of the study. Even in this group of patients, judged most suitable for EC-IC bypass and retrospectively selected from a surgical case series, the authors concluded that "the probability is low that the risk of stroke is less overall in patients who have STA-MCA bypass than in those who do not have surgical treatment." They also noted that "the patients in the 35-through 74-year age groups who underwent STA-MCA bypass did not do as well as the patients who did not have surgical intervention."

It is apparent, then, that the overall perioperative mortality and stroke rates as reported in surgical case series are not superior to those that were achieved in the EC-IC cooperative trial. Moreover, the annualized rate of adverse events in the cooperative trial is comparable to summary data from the surgical series, and the postoperative followup of patients in the cooperative trial was far more assiduous than in any of the surgical series reports. The reduction in the incidence of TIA was essentially equivalent in the surgical and medical arms of the cooperative trial and comparable to the best results claimed in published surgical case series. Therefore, the weight and quality of available evidence do not allow a logical and defensible conclusion that the results reported in published case series were superior to the results in either the surgical or medical arms of the EC-IC cooperative trial.

The issue of nonrandomization of eligible patients who were operated upon outside the trial engendered uncertainty concerning the applicability of the trial results to the clinical population of patients with ischemic symptoms and vascular disease. However, no objective data that clearly indicate that any particular subgroup of patients have reduced rates of stroke following surgery have been provided since publication of the EC-IC trial. This is not unexpected, since the surgical series supporting EC-IC bypass (Table 1). presumably represent the best possible results in patients whose clinical parameters were judged consistent with particular suitability for surgical therapy. There is no evidence that the results reported in the surgical arm of the EC-IC trial are inferior to those compiled from the surgical series. That data diminish the relevance of claims that patients uniquely responsive to bypass were excluded from randomization, thereby biasing the results to the disadvantage of the surgical arm.

Neither subsequent reports in the literature(33-35). nor published correspondence(36-43). provided further evidence in support of the operation based on prospective trials. Jennett(44). reviewed the quality of the evidence in support of EC-IC bypass and found it wanting. He cautioned that all medical therapy is in the end empirical, and should be tested by empirical means. Comparison of the prospective randomized trial of EC-IC bypass with the many uncontrolled clinical case series supports his contention that the objectives of treatment should not be merely the restoration of anatomic or physiologic function but, rather, the improvement of outcome for the patient.

The National Institutes of Health (NIH) has advised OHTA (written communication, February 14, 1986) that the EC-IC bypass study failed to demonstrate any value of surgical bypass over medical management in the prevention of stroke in patients with atherosclerotic arterial disease of the middle cerebral or internal carotid arteries. In addition, the NIH noted that the clinical trial had identified subgroups in whom the procedure appeared to result in a significantly worsened outcome than did medical therapy.

The Committee of the American Academy of Neurology recommended (written communication, March 13, 1986) that EC/IC bypass for the treatment or prevention of stroke be considered investigative.

The American Association of Neurologic Surgeons (AANS) provided several comments to OHTA. The Association stated (written communication, February 25, 1986) that the role of the operation was controversial. The Chairman of the Association's Section of Cerebrovascular Surgery indicated (written communication, March 24, 1986) that they desired to investigate this type of surgery in greater detail. In mid-1986, the Association informed OHTA (written communication, April 14, 1986) that they believed that optimal patient care required continued payment for surgery "when the operation seems indicated." Finally, in 1987 the Association indicated (written communication, January 12, 1987) that the special committee of the AANS had submitted their report for publication and enclosed a summary statement(29). of that committee.

OHTA received numerous correspondence from neurosurgeons; a number expressed the view that patients who have persistent or continued TIAs in spite of optimum medical management might be candidates for EC-IC bypass, and that this group should be investigated further. Some correspondents felt that the quality of life was improved for surgical patients, and that such factors should be considered in evaluating outcome in addition to stroke and death rates.

Summary and Conclusions

Extracranial-intracranial bypass surgery is an operative procedure in which the superficial temporal artery is anastomosed to the middle cerebral artery (hence the alternative term, STA-MCA bypass). The operation, first described in 1969, was employed to circumvent otherwise surgically inaccessible atherosclerotic lesions high in the internal carotid system or in the middle cerebral artery. Such lesions were believed to be proximate causes of reduced cerebral perfusion resulting in transient or permanent neurologic impairment, and the procedure was predicated to reduce the incidence of such sequelae in operated patients. Proponents believed that the anatomic and physiologic rationale for the surgery in conjunction with the relatively low surgical morbidity and mortality was sufficient justification for advocating its use. As the frequency of the operation increased, reports appeared in the literature describing successful accomplishment of the surgery, detailing operative complication rates, and occasionally comparing observed postoperative stroke rates with historical stroke rate data in unoperated patients. Advocates of operative intervention believed that the uncontrolled surgical series so reported established the clinical effectiveness of this intervention in the prevention of stroke. Residual skepticism regarding the ultimate utility of the surgery was based on the observation that all published series used historical controls, if any, and led to the design of a prospective, randomized, controlled multicenter trial of EC-IC. Many advocates of EC-IC bypass enthusiastically supported this study.(6,7,10,11,17) Patient entry was begun in 1977, and completed in 1982; the final report was published in 1985.(2) This study of 1,377 patients concluded that there was no evidence that the surgery reduced the incidence of stroke and stroke-related death; such events occurred earlier and with higher frequency in operated than in medically treated patients. Criticisms of the trial and its findings appeared in the medical literature soon after publication of the study's results. The most significant of these, and the one that was to require elaborate and detailed replies from the principal investigators, was the allegation that a large but indeterminate number of otherwise eligible patients seen at the various study centers were not randomized, but operated outside the trial.(28,29) This was construed to indicate that the results of the trial were not generalizable to the population at risk who might be considered for the operation. The trial team's response indicated that 1,439 patients were ineligible because they did not meet the protocol requirements, 475 did not provide informed consent, and only 95 patients were considered "possibly eligible" but were not randomized.(30) They further stated that of the 71 centers, 52 provided 1,191 of the trial patients. This group had only 66 "possibly eligible" patients operated without randomization. Results from these 52 centers were analyzed, and the findings were virtually identical to the original trial findings; that is, no evidence of surgical effectiveness in preventing stroke or stoke-related deaths.

Taken as a whole, the surgical series ( Table 1). provided insufficient information to precisely characterize the operated patients, and most contained inadequate followup data for accurate determination of the postoperative stroke rate. The historical annual stroke rates to which the surgical results were compared in several of the reports varied between as little as 3 to as much as 25, and the concomitant effect of antiplatelet medication was not adequately addressed. Operative stroke and mortality reported in these series was not superior to that observed in the cooperative trial. Moreover, one cannot conclude that the stroke rates following surgery as reported in the series cited above are superior to those observed in either the medically or surgically treated groups in the cooperative trial. Consequently, when the findings of the cooperative trial are compared with those of the multiple surgical series that support prophylactic operative intervention, it cannot be reasonably concluded that substantial evidence exists that would allow a conclusion that EC-IC bypass is a clinically effective procedure that reduces the incidence of fatal and nonfatal stroke in patients with occlusive vascular lesions suitable for that surgery.

The multicenter cooperative trial of EC-IC bypass was not a perfect study; it is unfortunate that 19 of the 71 centers were unable or unwilling to comply with the protocol to which they had agreed, and that such information was not made known by the principal investigators at an earlier date. However, none of the criticisms of the cooperative trial has provided objective evidence that such flaws as existed in the study render the results suspect, nor has any substantive evidence contrary to the conclusions of that research been provided since the trial results were published. In the absence of reliable objective evidence that there exists a group of patients in whom EC-IC bypass reduces the frequency of stroke in operated patients compared with those managed by medical means, the results of the single prospectively randomized controlled trial must be accepted as the best quality of medical evidence available. The burden of proof rests on those who advocate the prophylactic clinical utility of this surgery for the reduction of the incidence of stroke.