Portal de Periódicos da CAPES

Decision Making in the Face of Uncertainty: The Case of Carotid Endarterectomy

1990; Elsevier BV; Volume: 65; Issue: 5 Linguagem: Inglês

10.1016/s0025-6196(12)65135-0

1942-5546

David B. Matchar,

Health Systems, Economic Evaluations, Quality of Life

The management of patients with cerebrovascular disease has been characterized as ranging from "therapeutic nihilism to enthusiastic treatment with unproven remedies."1Hachinski V Norris JW The Acute Stroke. FA Davis Company, Philadelphia1985Google Scholar Nowhere has this variability in physician practices been more evident than in the use of carotid endarterectomy for prevention of stroke. The positions of experts have ranged from strongly supportive to highly critical of the use of this procedure.2Byer JA Easton JD Therapy of ischemic cerebrovascular disease.Ann Intern Med. 1980; 93: 742-756Crossref PubMed Scopus (22) Google Scholar Community practices mirror this diversity of opinion, which has resulted in substantial variation in its use across physicians and geographic areas. A recent declining trend in the number of carotid endarterectomies may reflect a growing consensus about inappropriate applications of this procedure. Nevertheless, the number of procedures performed and the variability in its use still far exceed what many experts agree is reasonable.3Winslow CM Solomon DH Chassin MR Kosecoff J Merrick NJ Brook RH The appropriateness of carotid endarterectomy.N Engl J Med. 1988; 318: 721-727Crossref PubMed Scopus (382) Google Scholar Determining the appropriate use of carotid endarterectomy has special urgency, in light of evidence that this procedure is performed with less than optimal safety in many settings.4Brott T Thalinger K The practice of carotid endarterectomy in a large metropolitan area.Stroke. 1984; 15: 950-955Crossref PubMed Scopus (159) Google Scholar What are the sources of this variation in practice? Researchers have proposed several explanations. To some extent, variations are appropriate, as they reflect important differences in patient and physician characteristics. For example, some areas may have unusually high stroke rates, and some medical centers may have especially good surgical results. The residual unexplained variation is attributable largely to practice style.5Wennberg JE Dealing with medical practice variations: a proposal for action.Health Aff (Millwood). Summer 1984; 3: 6-32Crossref PubMed Scopus (493) Google Scholar The roots of practice style are numerous, including medical system differences, such as the availability of resources;6Caper P Solving the medical care dilemma (editorial).N Engl J Med. 1988; 318: 1535-1536Crossref PubMed Scopus (29) Google Scholar the nature of existing referral networks; and the influence of leaders with strong opinions.7Leape LL Unnecessary surgery.Health Serv Res. 1989; 24: 351-407PubMed Google Scholar With regard to carotid endarterectomy, however, variations in practice style reflect something more fundamental: to a large extent, variation is related to uncertainty about what constitutes an optimal management strategy. This uncertainty has two main elements-scientific uncertainty and cognitive uncertainty. The objective of carotid endarterectomy is to diminish the risk of stroke. Stroke has an enormous impact on the health of its victims. Currently, 500,000 Americans suffer new strokes each year, and there are 1.8 million stroke survivors.8Wolf PA Kannel WB McGee DL Epidemiology of strokes in North America.in: Barnett HJM Mohr JP Stein BM Yatsu FM Stroke: Pathophysiology, Diagnosis, and Management. Churchill Livingstone, New York1986: 19-29Google Scholar Mayo Clinic data show that almost 60% require some assistance in basic functioning and almost 40% of first stroke survivors are still dependent after 5 years.9Dombovy ML Basford JR Whisnant JP Bergstralh EJ Disability and use of rehabilitation services following stroke in Rochester, Minnesota, 1975-1979.Stroke. 1987; 18: 830-836Crossref PubMed Scopus (143) Google Scholar Carotid endarterectomy is promoted especially for patients with transient ischemic attacks and minor completed strokes, although most such procedures are performed on asymptomatic patients.3Winslow CM Solomon DH Chassin MR Kosecoff J Merrick NJ Brook RH The appropriateness of carotid endarterectomy.N Engl J Med. 1988; 318: 721-727Crossref PubMed Scopus (382) Google Scholar The evidence supporting this approach is based primarily on the single randomized controlled trial performed in the late 1960s, the Joint Study of Extracranial Arterial Occlusion.10Bauer RB Meyer JS Fields WS Remington R Macdonald MC Callen P Joint Study of Extracranial Arterial Occlusion. III. Progress report of controlled study of long-term survival in patients with and without operation.JAMA. 1969; 208: 509-518Crossref PubMed Scopus (140) Google Scholar In that study, patients with transient ischemic attacks and minor completed strokes experienced approximately half as many strokes in a 42-month follow-up as did patients treated with what was then conventional medical therapy. After accounting for surgical mortality and morbidity, however, the results were not statistically significant. In a follow-up to the Joint Study, Hass and Jonas11Hass WK Jonas S Caution falling rock zone: an analysis of the medical and surgical management of threatened stroke.Proc Inst Med Chgo. 1980; 33: 80-84PubMed Google Scholar suggested that had the surgical mortality and morbidity been as low as 3%, the difference in cumulative stroke and death rate between surgical and nonsurgical groups would have reached statistical significance. Although this article was probably intended as a caveat to those who might overinterpret the results of the Joint Study, the flurry of surgical reports of low perioperative complication rates suggests that the analysis by Hass and Jonas was taken less as a warning than as a call to arms. Lack of a consensus after the Joint Study has spawned a series of clinical trials now in progress.12Marler JR Carotid endarterectomy clinical trials (editorial).Mayo Clin Proc. 1989; 64: 1026-1029Abstract Full Text Full Text PDF PubMed Scopus (2) Google Scholar Even if scientific understanding of stroke were perfect, physicians face another overwhelming source of uncertainty. With no clear guidelines defining when it is appropriate to perform carotid endarterectomy, each physician faces the complex problem of synthesizing evidence on the value of the procedure. Although much of the debate about carotid endarterectomy in the past 2 decades has focused on the relative efficacy of the procedure versus its risks, the question of whether to perform carotid endarterectomy is a much more subtle analytic problem. Physicians must consider a host of factors, including not only the efficacy and risk of surgical intervention but also the native risk of stroke, the risk of succumbing to nonstroke diseases, the durability of carotid endarterectomy, and the relative severity of cerebrovascular symptoms (ranging from transient or minor symptoms to a severely debilitating deficit). In addition, the timing of outcomes must be considered. Clearly, a stroke today is worse than a stroke several years hence. This issue is especially important in the case of a prophylactic procedure because such a procedure subjects a patient to an early risk in the hope of substantially diminishing a later risk. Recent work has focused on the use of decision analysis to reduce cognitive uncertainty by including each of these factors in a simulation model.13Matchar DB Pauker SG Transient ischemic attacks in a man with coronary artery disease: two strategies neck and neck.Med Decis Making. 1986; 6: 239-249Crossref PubMed Scopus (15) Google Scholar, 14Matchar DB Pauker SG Endarterectomy in carotid artery disease: a decision analysis.JAMA. 1987; 258: 793-798Crossref PubMed Scopus (65) Google Scholar, 15Feussner JR Matchar DB When and how to study the carotid arteries.Ann Intern Med. 1988; 109: 805-818Crossref PubMed Scopus (61) Google Scholar This type of analysis indicates that three crucial factors influence the relative preference for carotid endarterectomy: the native risk of stroke; the risk of surgical treatment; and the efficacy of operative intervention in terms of proportionate reduction of stroke events among surgical survivors. The results of this model suggest that even among high-risk patients (that is, those with a 5% annual risk of stroke) subjected to a procedure with low risk (a perioperative mortality of 1% and associated morbidity of 4%) and high potential gain (a diminution of subsequent risk of stroke by 50% for 10 years), the improvement in life expectancy would be approximately 1 month (or 3 months if one measures life expectancy in terms of quality-adjusted life-years). Two principal lessons can be learned from decision modeling. First, unless the procedure can be performed with relative safety, it cannot be justified. For example, when the procedure is performed with 4% mortality and 8% morbidity, patients with an underlying annual risk of stroke of 5% would not benefit. Second, it is difficult to support the use of a prophylactic procedure with even a modest risk when the underlying risk of disease is low. This is true even when the procedure has a definite and substantial efficacy. When the native risk of stroke is even lower than for patients with transient ischemic attacks, such as for asymptomatic patients, the risks outweigh the benefits (Fig. 1). Unless the surgical risk is reduced to negligible levels or asymptomatic patients with an especially high risk for unheralded stroke can be identified, the expansion of the use of carotid endarterectomy beyond the group of symptomatic patients cannot be justified. Although some medical centers, including the Mayo Clinic (as discussed by Sundt and colleagues in this issue of the Proceedings, pages 625 to 635), have achieved low levels of surgical risk, it continues to be difficult to predict which asymptomatic patients are at especially high risk for stroke. If the decision to perform a carotid endarterectomy involved simply a series of analytic judgments about efficacy and risk, the continuing controversy would be difficult to explain. In the absence of scientific support, we might expect a clinician to avoid such a potentially risky intervention, especially in asymptomatic patients. As an advocate of such a conservative approach, I have had many opportunities to interact with those who favor broader use of carotid endarterectomy. The most common reason cited for favoring a more aggressive approach is that carotid endarterectomy seems to work. Reliance on such clinical perception to support carotid endarterectomy is an especially slippery slope because personal experiences may lead to gross errors in judgment as they are heavily weighted in favor of recent experience.16Tversky A Kahneman D Judgment under uncertainty: heuristics and biases.Science. 1974; 185: 1124-1131Crossref PubMed Scopus (19270) Google Scholar In the case of carotid endarterectomy, clinical judgment serves only to reinforce preexisting biases. The self-reinforcing nature of the decision to perform carotid endarterectomy is best seen by an illustrative example. Let us assume that carotid endarterectomy has no influence on long-term risk of stroke. If the true risk of perioperative mortality were as high as 5%, what is the probability that a physician would have one poor outcome in a series of five patients? On the basis of simple probability theory, the likelihood that no patient will experience perioperative complications is 0.955Wennberg JE Dealing with medical practice variations: a proposal for action.Health Aff (Millwood). Summer 1984; 3: 6-32Crossref PubMed Scopus (493) Google Scholar or 77%. Thus, the probability that at least one patient would have a poor outcome is only 23%. In other words, the odds that a physician's recent experience would contradict his belief that the procedure is safe is less than 1:3. The chance of having a single poor outcome does not exceed even odds (50:50) until the number of patients exceeds 13. This bias in judgment can also be seen in the follow-up period. If the native risk of stroke is 5%, the chance of having at least one patient with a stroke within a 6-month follow-up period does not exceed even odds until the number of patients exceeds 28. This is so even if the procedure is not at all effective in reducing the risk of stroke. A second reason that a preference for carotid endarterectomy is self-reinforcing is that should a poor outcome occur postoperatively, it is natural to attribute this result to some unpredictable patient factor. After all, the patient would not have been selected for the procedure had he not been judged to be at high risk for stroke. Thus, because good outcomes are common and poor outcomes are easily explained, the clinician who is already inclined to support the use of carotid endarterectomy will find that personal experience only serves to reinforce that preexisting bias. An additional factor that bears on the decision to perform carotid endarterectomy is the relative preference some physicians have for "aggressive" versus "conservative" management strategies. This relative preference can be explained in part by the concept of retrospective regret or "the chagrin factor."17Feinstein AR The 'chagrin factor' and qualitative decision analysis.Arch Intern Med. 1985; 145: 1257-1259Crossref PubMed Scopus (151) Google Scholar Retrospective regret is experienced by a decision maker when a decision is associated with a poor outcome. Some physicians judge a poor outcome to be less regrettable when it results from an active or "aggressive" intervention. In other words, it is judged to be better to have tried and lost than never to have tried at all (and lost). Faced with a patient perceived to be at high risk for stroke, a physician may tend to choose (all other things being equal) to minimize his retrospective regret. Despite the diminishing frequency of use, carotid endarterectomy remains a common procedure. Ongoing clinical trials should help to clarify which subgroups of patients and what surgical conditions must exist for the procedure to be an appropriate consideration. Although these trials will decrease our scientific uncertainty, the decision to perform carotid endarterectomy will continue to be made with a measure of cognitive uncertainty. In this regard, decision modeling can be useful as a method for integrating the results of clinical studies with data from local experience.