Editors: Norris, John W.; Hachinski, Vladimir
Title: Stroke Prevention, 1st Edition
Copyright 2001 Oxford University Press
> Table of Contents > III - Prevention: Policy and Practice > 15 - From Clinical Trials to Clinical Practice
From Clinical Trials to Clinical Practice
Geoffrey A. Donnan
Be delightfully surprised when any treatment at all is effective, and always assume that a treatment is ineffective unless there evidence to the contrary.
--A. L. Cochrane, 19711
In its endless quest for health into old age, the human race has been constantly searching for forms of therapy that will minimize the impact disease processes. One of the interesting qualities of many therapeutic compounds is that they occur as natural products of our environment (e.g., aspirin), sometimes in a ubiquitous fashion, and need to be identified, concentrated, applied in a highly specific manner (e.g., intravenously) to produce benefit in terms of reducing morbidity or mortality. Other therapies need to be specifically manufactured (e.g., clopidogrel), but often their genesis lies in an understanding of the biological basis of the disease or the effect of a naturally available compound. Regardless their origins, the real test of any form therapy is efficacy. From early times the history of medicine has been riddled with therapies that have sometimes existed for centuries but have ultimately proven to be ineffective. Conversely, other therapies have been discarded without ever being adequately tested. The concept of adequate proof of efficacy has been gradually evolving since early in the twentieth century. Only simple experiments were required for such dramatically effective agents as penicillin, sulphonamides, and insulin, but for less markedly beneficial therapies, more specific proof is required.
One of the champions of a more scientific approach to the introduction therapies in medicine was Archie Cochrane. Before working the National Health
Under the best conditions one would have expected an appreciable mortality; there in the gulag I expected hundreds to die of diptheria alone in the absence of specific therapy. In point of fact there were only four deaths, which three were due to gunshot wounds inflicted by the Germans. This excellent result had, of course, nothing to do with the therapy they received or my clinical school. It demonstrated, on the other hand, very clearly, the relative unimportance of therapy in comparison with the recuperative powers of the human body.
While this point is perhaps a little overstated, it does illustrate the need for making a distinction between the natural history of disease processes and the effect of therapy per se. Understandably, most people (particularly physicians) wish for a beneficial outcome of therapy. This inherent bias has bedeviled observational literature on medical therapies over the centuries. The need to eliminate this positive (and negative) bias led to the simple concept of the randomized controlled trial (RCT) in the United Kingdom during the 1950s, which has been developed into more sophisticated forms during the last 30 years. One of the earliest such trials was published by Daniels and Hill,2 who described three trials of the use of chemotherapy for pulmonary tuberculosis using a randomized approach to test bed rest, streptomycin PAS, or a combination of both chemotherapies. Trial methodology has been further refined and the double blind technique is now considered the gold standard for evidence against which all other forms should be measured. A boom in clinical trials has occurred all branches of medicine and surgery, and, fortunately, stroke medicine has been not immune to its influence; so much so that Caplan called the 1980s the era of clinical trials. 3
This chapter addresses the problem of translating burgeoning mass knowledge about cerebrovascular disease prevention from clinical trials into clinical practice. In doing so, it is necessary to describe how trials are conducted and then discuss the various ways in which information from these trials can be converted into better clinical outcomes (Fig. 15.1). The process must involve physician, patient, and community acceptance of the information contained in trial, so that generalized clinical usage is embraced by all stakeholders. Unfortunately, barriers to these processes abound, as will be shown.
How Trials are Conducted
Types of Trials
A large variety of clinical trial types have been designed for testing either pharmaceutical agents or surgical procedures.4 These range from single-patient studies to mega-trials and include single or multiple groups of patients with parallel,
FIGURE 15.1. The pathway from clinical trial information to clinical outcomes, with the key elements of physician, patient, and community acceptance. Barriers to these elements abound.
For trials of pharmaceutical compounds, the process is usually tiered from phase I through phase IV, as shown in Table 15.1.5 In some instances, pharmaceutical companies may conduct a number of phases simultaneously or there be some overlap between phases. Trials of surgical procedures (for example,
TABLE 15.1. Phases of Clinical Trials of Drug Therapy
Sample Size and Statistical Power Considerations
A key issue in the design of any clinical trial is that sample size. A fault of many the earlier randomized controlled trials was that inadequate power to test the hypotheses proposed. Directly or indirectly, this has led to the phenomenon of pooled analyses, or meta-analyses, to generate greater power and avoid of type II error. The relationship between power and therapeutic effect is
Funding for Clinical Trials
An inevitable accompaniment of the growth clinical trials in medicine has been an explosion of costs associated with their conduct. The two broad groups involved in the clinical trial industry are academic clinical investigators and pharmaceutical companies. Each group has produced studies of varying quality as study design and trial conduct have improved over time. The agenda of the two
FIGURE 15.2. The relationship between effect size and power of trials therapy. (Reprinted by permission of
Kraemer, H.C. Sample size: When is enough? Am J Med Sci. 269:61;1986.11)
Categories of these relationships are shown in Table 15.2, and, as can be seen, the most highly regarded trials are those in which adequate funding is achieved through national research organizations or a collaborative effort between investigators and the pharmaceutical industry. When such collaborations occur, the ideal arrangement is for the academic investigating group to be hired to conduct
TABLE 15.2. Funding for Clinical Trials: Models of Relationships Between Academic and Commercial Groups
Interestly, it has been shown that randomized trials conducted with commercial sponsorship are more likely to report statistically significant advances.12 Further, some trials may be stopped by commercial sponsors without explanation.13,14
Standards of Trial Design and Reporting
As a part of the effort to improve the standards study trial design, several important initiatives have recently been developed. The journal Lancet now accepts protocols for review, so that alterations can be made to defective protocols before study commencement.15 Another was the development of the Consolidated Standards of Reporting Trials (CONSORT) guidelines.1,18 These guidelines outline a set of minimum criteria that should be followed when investigators submit manuscripts to journals for publication. Although the guidelines have been adopted by only a small number of high-profile journals to date (including Lancet, JAMA and BMJ), they should gradually become more widespread, so that there will be more consistency in the design, analysis, and reporting of clinical trials.18
As stated, clinical trial methodology has changed considerably over the past 20 years, with improvements and quality control mechanisms now in place that may allow observers to place greater credence in published results. However, this higher-quality evidence must be translated into clinical practice and, ultimately, into improved clinical outcomes. For this to occur, physician, patient, and community acceptance of the information must occur.
Clinical Trials: From Evidence to Clinical Practice
For physicians to accept new clinical trial information, it must first be gleaned and synthesized by the target group. Sources of information include original research articles, pharmaceutical company material, expert reviews, scientific meetings, and standard texts. The technique of meta-analysis has been increasingly used to advantage, so that data from a number of clinical trials can be reviewed. Information from these and other sources may then be used to produce clinical practice guidelines. Interpretation of data from a variety sources may be improved by adhering to the principles of evidence based medicine (EBM).
While the boom in clinical trials has been of enormous benefit to stroke physicians and their patients, it has also created problems. How can the mass of data generated from these trials be assimilated and the information disseminated to
The power of meta-analysis is best illustrated by the technique cumulative meta-analysis.21 Using this method, the results of new trials are sequentially included in the analysis to provide a new point estimate and a steadily diminishing 95% confidence interval because of an incrementally increasing sample size. When this approach was retrospectively applied to the use of streptokinase in acute myocardial infarction, it was found that a statistically significant reduction in mortality was achieved in 1973 after only eight trials involving 2432 patients (Fig. 15.4). The results of the 25 subsequent trials, which enrolled an additional 34,542 patients, had little or no effect on the odds ratio for mortality, but simply narrowed the confidence interval. When these results were compared to recommendations of reviewers (articles and textbooks) the same era, significant discrepancies were found between the information available from trials and current recommendations22 (Fig. 15.5). One can conclude that, without metaanalyses to aid in trial aggregation and synthesis, significant delays in the introduction of important therapies into routine practice may occur.
One of the most important meta-analyses performed in the area of stroke prevention was carried out by the Antiplatelet Trialists Group (APT).7,8,9 By aggregating data from some 140,000 patients with increased risk of vascular disease from more than 300 trials, this group gained information on a variety of outcomes using this form of therapy.7,8,9 Specifically, the relative risk reduction about 22% for patients with prior stroke or transient ischemic attack of developing
FIGURE 15.3. The Cochrane Collaboration logo showing how pooling data reveals the significance of treatment effects.
FIGURE 15.4. Conventional and cumulative meta-analysis of trials intravenous streptokinase for acute myocardial infarction. The odds ratio and 95% confidence interval an effect of treatment on mortality are shown a logarithmic scale. From Lau et al.21
However, the enthusiasm for meta-analyses is far from uniform. Critics emphasize that this approach involves the mixing of apples and oranges, often causing clinical and statistical heterogeneity that may make correct interpretation of the results more difficult. On occasion, therefore, meta-analyses may provide misleading information, creating the need for more definitive proof from a megatrial . An example of this was the use of magnesium in myocardial infarction. Meta-analyses of a large number smaller trials had suggested that this approach might be effective.23 However, a megatrial subsequently proved, to everyone's reasonable satisfaction, that this was not the case.24 Issues such as variable protocols of differing quality and publication bias were probably responsible. In a formal comparison of meta-analyses and large trials, large trials differed from meta-analyses 10% to 23% of the time.25 In other words, meta-analyses should not be taken as the absolute gold standard, but merely guides based on, among other things, the quality of the input data.
FIGURE 15.5. Results of meta-analysis and recommendations reviewers (articles textbooks of the same era). For each treatment for acute myocardial infarction, the cumulative meta-analyses by year of publication randomized control trials (RCTs) are presented on the left. The cumulative number of trials and patients (PTs) are also presented. On the right, the recommendations of the clinical expert reviewers are presented in two-year segments (number of recommendations). The letter M indicates that least one meta-analysis was published that year. From Antman, et al.22
Even megatrials themselves are not without their critics. Charleton argued that megatrials tend to use simple protocols in multiple research centers rather than rigorous protocols in more clinically homogenous, smaller groups.26 This tradeoff between rigorous control of homogenous groups and large size to account for random fluctuations in trial entry criteria is a matter of ongoing debate. Further, megatrials may underestimate differences among groups ( null bias ), mainly due to poor compliance, but also perhaps due to poorly controlled additional therapies that can become unbalanced among groups, a phenomen that may have occurred in coronary artery disease studies.27 To counter these problems, population-based strategies have been advocated, such as that instigated in the north of England, termed Population Adjusted Clinical Epidemiology (PACE). Here, concern had been expressed that progress in therapy among adults with
Interpreting clinical trials and evidence-based medicine
Health policy is decided upon from a variety of sources, including observational studies, entrenched existing policies, and political expediency (often responding to high profile pressure groups). It has been estimated that less than 20% of clinical policies are based on randomized control trials.28,29 A number of guidelines are available on how to interpret clinical trials. All follow the general principle of using a series checkpoints to determine the nature of the population from which subjects are drawn, the randomization process, the stipulation of endpoints, and analysis techniques. Other guides have been published on how to decide on the applicability of clinical trial results to individual patients.30
The EBM approach to the interpretation of clinical information was developed during the 1990s.31,32 It represents a sensible crystallization of an approach searching for, and subsequently interpreting, data in the environment of clinicians. By structuring levels of evidence (Table 15.3), clinicians can quantify more readily the importance and reliability of the data they are examining. Increasingly, this approach is being used throughout the English-speaking world, although this process, also, is not without its critics.33 The most common criticisms of EBM are that the process is not new, that information derived too heavily from randomized controlled trials and meta-analyses, that it excludes specific patient problems. Further, the authoritative aura given to the collection of evidence under EBM may lead to major abuses that produce inappropriate guidelines or doctrinaire dogmas for clinical practice.
Clinical practice guidelines
A flurry of activity has occurred in the production clinical practice guidelines, including those for stroke prevention.34,35,36,37,38,39 While these have been extremely useful and widely published, particularly by the Scottish Intercollegiate Guidelines Network (SIGN), which receives government funding,37,38,39 major problems remain in disseminating these guidelines among medical practitioners.
TABLE 15.3. Levels of Evidence
Clinical trial interpretation, EBM, and clinical practice guidelines are intimately connected with the issue of continuing medical education (CME) and its effectiveness. Concerns have been raised about the effectiveness of traditional continuing medical educational strategies, an issue that has been examined in recent studies. For example, Davis et al.40 reviewed educational approaches used from 1975 to 1994. Reassuringly, they showed that about two-thirds of the studies (70%) displayed a change in physician performance, while almost half (48%) of interventions produced changes in healthcare outcomes. The most effective strategies were communityand practice-based methods, such as reminders patient-mediated strategies with multiple interventions. Poorer outcomes were demonstrated by audit and educational materials. The weakest approach was formal CME conferences. Fortunately, an increasing amount of research is being directed toward monitoring and improving the effectiveness of translating original research data into better clinical outcomes.
Patient and Community Acceptance
Even though the medical fraternity may accept a particular form of therapy, patient and community acceptance must be generated for favorable outcomes to be achieved. For example, when the evidence for aspirin as an effective antiplatelet agent emerged in the 1970s and 1980s, patient community resistance to its use had to be overcome. The community needed to be convinced that an agent that had been used as a simple analgesic for 70 years would be effective in its new role as an anticoagulant. The main mechanism by which patient and community acceptance of new forms therapy occurs is by education via the print and electronic media. Acceptance may be facilitated when there is ease of oral
Barriers to Physician Acceptance of Clinical Trial Information
While most worthy new information is introduced eventually into regular clinical practice, the rate and extent of adoption new approaches varies enormously. This may be due to a number of factors, including the following.41
Failure of information transfer, based on the system outlined in the preceeding sections (Figure 15.1).
Misleading use of risk reductions by pharmaceutical companies or others (relative vs. absolute risk reduction).
Relatively small benefit balanced by high risk.
Trials not relevant to the general population because of restrictive inclusion/exclusion criteria.
A particularly important issue in interpreting clinical trials is the difference between relative and absolute risk. Relative risk is always much greater often quoted by pharmaceutical companies in their promotional material to emphasize a positive clinical effect to the unwary physician. Absolute risk reduction is a better index of the effect of a drug. It can be translated readily into number needed to treat (NNT), a useful conceptual figure for physicians when assessing whether a particular form of therapy may be worthwhile. The method of calculating these figures from a hypothetical trial of stroke prevention conducted over five years is shown in Table 15.4.
Some good examples in the stroke prevention literature of importance understanding the relationship among relative risk, absolute and NNT may be found in the Clopidogrel versus Aspirin Patients at Risk of Ischaemic Events (CAPRIE)42 and Asymptomatic Carotid Artery Surgery (ACAS) 43 trial reports. In the CAPRIE trial, a relative risk reduction of 8.6% was achieved in vascular endpoints for clopidogrel vs. aspirin. However, the absolute risk reduction was only 0.5% hence the NNT to obtain benefit was 200. It can be seen that relative risk reduction of 8.6% was statistically significant, but it is open to discussion whether this was biologically significant. About a 50% relative risk reduction was achieved in vascular endpoints for patients treated with carotid endarterectomy in the ACAS trial. However, because the control population had only about a 2% stroke risk per year, the absolute reduction was only about 1% and the NNT 100. When put in this context, the 50% relative risk reduction does not appear to be as biologically important: it must be debated whether endarterectomy is appropriate on an individual patient basis.
TABLE 15.4. Calculation Method for Relative and Absolute Risk Reductions and Number Needed to Treat (NNT) in a Hypothetical Trail of Stroke Prevention Conducted Over Five Years
Risks and benefits
The issues of risk vs. benefit and the generalizability of trial results are well illustrated in the use of anticoagulants as a form of primary stroke prevention patients with atrial fibrillation. Atrial fibrillation (AF) has been investigated fairly thoroughly over the last few decades and has been established as one of the more important risk factors for stroke. The stroke in subjects with nonvalvular atrial fibrillation (NVAF) is about five times that of those who have no NVAF.44 As discussed in Chapter 6, this risk is stratified depending on associated factors. The risk of stroke per year is only about 1% to 2% among those with lone AF, but rises to around 5% with the addition of one risk factor (e.g., hypertension, diabetes, age greater than 65 years, or recent-onset cardiac failure thromboembolism elsewhere). With two or more risk factors, the stroke per year rises to around 7% to 8%.45 Effective forms of therapy for the prevention stroke in patients with atrial fibrillation are warfarin and aspirin. Based on pooled analyses, treatment with warfarin is associated with a relative risk reduction of stroke of around 70% (absolute reduction 3.1%, NNT 32).45 Sudlow et al.46 screened a random sample of 4843 people from a community aged 65 years and older for atrial fibrillation. Participants who were found to have AF received further investigations to stratify their eligibility for anticoagulation based on the pooled analysis of all trials warfarin therapy for NVAF,45 the SPAF47 and III48 trials. Of the participants, 228 (4.7%) had AF. Of these, 61% would have benefiled
TABLE 15.5. Percentage of Patients with Atrial Fibrillation Eligible for Anticoagulation and Those Anticoagulated by Risk Classification (adapted from Sudlow et al.46)
The reasons for the low level of penetration this form anticoagulants in primary prevention of stroke risk are complex and probably relate to the difficulty in detecting cases (the majority are asymptomatic), together with a reticence by physicians to use a therapy that, although in for around 40 years, is perceived to be difficult manage (repeated INR estimations required) and is associated with significant risk of bleeding. While the latter is certainly an important issue, in the pooled analysis the rate of major bleeding was only 1.0% among controls and 1.3% among patients with AF. This figure is likely to be higher among nontrial patients in the community. While the ratio of benefit to risk is certainly in favor of therapy, physicians are clearly less willing to use warfarin unless they are convinced of its efficacy (and low risk).
An effect converse to the problem of delayed and limited acceptance information from clinical trials was seen after the Extra-Cranial to Intra-Cranial (EC/IC) bypass trial results were published.49 This procedure was logical in that blood was bypassed to a presumably under-perfused or hemodynamically at-risk cerebral hemisphere by anastomosing the external carotid artery branches to the middle cerebral artery branches through a skull burr hole. However, when formally tested by a randomized controlled trial, the procedure was found to be ineffective in minimizing subsequent stroke or death. A procedure that had become increasingly accepted among neurosurgeons over the preceding decade plummeted in frequency almost as soon the trial results were published. Indeed, EC/IC bypass is now rarely performed. The transfer of clinical trial evidence to clinical practice was rapid and effective, probably because of the discrete, small number of clinicians involved (neurosurgeons), excellent communication within the group, and the relatively small number of procedures performed at baseline.
Barriers to Patient and Community Acceptance of Clinical Trial Evidence
As shown in Figure 15.1, issues that need to be considered in regard to patient and community acceptance of clinical trial data include community education
Screening for risk factors may seem a logical exercise, in that many risk factors are asymptomatic and may remain undetected unless some form of screening is undertaken. Examples in stroke prevention include screening for AF and asymptomatic carotid artery stenosis. In the latter case, a cost-benefit analysis was performed and the view was expressed that, given current screening costs and the level of effectiveness surgery, routine screening programs should not be introduced.35 This does not preclude the ongoing use of opportunistic screening, which occurs on many occasions when patients visit their doctor.
Poor compliance remains a difficult issue for both physicians and their patients. The reasons are complex and involve drug side effects as well cultural and societal resistance to taking medication. In a study of hypertension as risk factor for intracerebral hemorrhage (ICH), cessation of antihypertensive medication (for whatever reason) was shown to increase the risk of ICH by more than two-fold (odds ratio 2.45, 95% CI 1.13 5.77).50 Hence, compliance problems may significantly contribute to poor clinical outcomes.
A further barrier to therapy may relate quality-of-life issues. While there is an increasing trend to include quality-of-life measures in clinical trials reinforce (or otherwise) the efficacy of therapy, this was not always so. It is alleged that in some prophylactic trials, the clinical side effects (which affect quality of life) may have been measured inadequately. For example, when one antihypertensive agent was shown to be associated with a better quality of life than another, it was found that the investigators had performed an intention to treat analysis that did not evaluate compliance with therapy.51,52 It has been suggested that with many antihypertensive agents, quality of life can be improved merely by stopping the drug.51 While this might exaggerate issue, point is made.
What to do in the Absence of Evidence
While the emphasis of this chapter has been on the transfer information from clinical trials to clinicians, patients, and the community as a whole, the problem remains that evidence from randomized, controlled trials (and other sources of evidence) does not provide guidance for every clinical situation. In these circumstances what does one do? In some areas, medical practice is ahead of clinical trial evidence. In these cases, lower levels of evidence, as well common
1. Cochrane AL. Effectiveness and Efficiency: Random reflections on Health Services. Cambridge: Cambridge University Press, 1971.
2. Daniels M, Hill A. Chemotherapy of pulmonary tuberculosis in young adults. An analysis of the combined results three medical research council trials. BMJ 1952:1;1162.
3. Caplan LR. How should a Clinician choose treatment for preventing stroke when therapeutic trials are not available to guide the situation? Eur Neural 1995:35;293 303.
4. Spilker B. Guide to Clinical Trials. New York: Raven Press, 1991.
5. Berkovitz BA. Drug discovery and evaluation. In: Katzung BG, ed. Basic and Clinical Pharmacology. Lange Medical Publication, 1982.
6. Kraemer HC. Sample size: When is enough? Am J Med Sci 1986:296;361.
7. Antiplatelet Trialists' Collaboration. Secondary prevention of vascular disease by prolonged antiplatelet treatment. BMJ 1988:296;320.
8. Underwood MJ. Aspirin benefits with vascular disease and those undergoing revascularisation. BMJ 1994:308;71.
9. Antiplatelet Trialists' Collaboration. Collaborative overview of randomized trials of antiplatelet therapy - 1: Prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients. BMJ 1994:308;81.
10. Candelise L, et al. Randomized trial of aspirin and sulfinprazone in patients with TIA (AITIAIS). Stroke 1982:13;175 179.
11. Canadian Cooperative Study Group. A randomized trial of aspirin and sulfinpyrazone in threatened stroke. N Engl J Med 1978:299;53 59.
12. Davidson RA. Source of funding: An outcome clinical trials. J Gen Int Med 1986:1;155 158.
13. Good manners for the pharmaceutical industry (editorial). Lancet 1997:349; 1635.
14. A curious stopping rule from Hoechst Marion Roussel (editorial). Lancet 1997:350; 155.
15. Horton R. Pardonable revisions and protocol reviews. Lancet 1997:349;6.
16. Begg C, Cho M, Eastwod S, Horton R, Moher D, Olkin I, et al. Improving the quality of reporting randomized control trials. The CONSORT Statement (editorial). JAMA 1996:276;637 639.
17. Altman DG. Better reporting of randomized control trials: The CONSORT Statement (editorial. BMJ 1996:313;570 571.
18. CONSORT: An important step towards evidence based healthcare (editorial). Ann Int Med 1997:126;81 83.
19. Chalmers I, Dickensin K, et al. Getting to grips with trachie Cochranes Agenda. BMJ 1992:305;786 788.
20. Cochrane Database of Systematic Reviews. Oxford The Library, Update Software Limited, issue 1, 1999.
21. Lau J, et al. Cumulative meta-analysis of therapeutic trials for myocardial infarction. JV Eng J Med 1992:327;248 254.
22. Antman EM, et al. A comparison of results of meta-analysis randomized control trials and recommendations of clinical experts. JAMA 1992:268;240 248.
23. Yusuf S, et al. Intravenous magnesium in acute myocardial infarction. An effective, safe, simple, and expensive intervention. Circulation 1993:897;2043 2046.
24. ISIS (Fourth International Study of Infarct Survival. Collaborative Group. ISIS-4: A randomized factorial trial assessing early oral captopril, oral mononitrate, and intravenous magnesium sulphate in 58,050 patients with suspected acute myocardial infarction. Lancet 1995:345;669 684.
25. Ioannidis JPA, Cappelleri JC, Lau J. Issues in comparisons between meta-analyses and large trials. JAMA 1998:279; 1089 1093.
26. Charleton BG. The future of clinical research: From mega trials towards methodological rigour and representative sampling. J Eval Clin Pract 1996:2(3);159 169.
27. Woods KL. Mega trials and management of acute myocardial infarction. Lancet 1995:346;611.
28. Committee for Evaluating Medical Technologies in Clinical Use, Division of Health Sciences Policy, Division of Health Promotion and Disease Prevention, Institute Medicine. Assessing medical technologies. Washington, DC: National Academy Press, 1985.
29. Hornberger J, Wrone E. When to base clinical policies on observational versus randomized trial data. Ann Int Med 1997:127;697 703.
30. Dans AL, Dans LF, Guyatt GH, Richardson S, for the Evidence-Based Medicine Working Group. Users guide to the medical literature XIV: How to decide on the applicability of clinical trial results to your patient. JAMA 1998:279;545 549.
31. Sackett DL, Rosenberg WMC, Gray JAM, Haynes RB, Richardson WS. Evidence based medicine: What it is and what isn't. BMJ 1996:312;71 72.
32. Sackett DL, Richardson WS, Rosenberg WMC, Haynes RB. Evidence Based Medicine: How to Practice and Teach Evidence Based Medicine. Churchill-Livingstone, 1997.
33. Feinstein AR, Horwitz RI. Problems in the evidence of evidence based medicine. Am J Med 1997:103;529 535.
34. National Health and Medical Research Council. Prevention of stroke: The role anticoagulants, anti-platelet agents and carotid endarterectomy. Australian Government Publishing Service, Canberra: Commonwealth of Australia 1997.
35. Prevention of Stroke: Clinical Practice Guidelines. National Health and Medical Research Council, Canberra December 1996.
36. Lechtenberg R, Schutter AHS, Neurology Practice Guidelines. New York: Marcel Dekker, 1998.
37. Clinical Guidelines: Criteria for Appraisal for National Use. Edinburgh: Scottish Intercollegiate Guidelines Network (SIGN), September 1995.
38. Clinical Guidelines: Management of Patients with Stroke. I: Assessment, Investigation, Immediate Management and Secondary Prevention. Edinburgh: Scottish Intercollegiate Guidelines Network (SIGN), May 1997.
39. Clinical Guidelines: Management of Patients with Stroke. II: Management of Carotid Stenosis and Carotid Endarterectomy. Edinburgh: Scottish Intercollegiate Guidelines Network (SIGN), May 1997.
40. Davis D. Does CME work? An analysis of the effect of educational activities on physician performance or healthcare outcomes. Int J Psychiatr Med 1998:28;21 39.
41. Channer KS. Translating clinical trials into practice. Lancet 1997:349;645.
42. CAPRIE Steering Committee. A randomized, blinded trial of clopidogrel versus aspirin in patients at risk of ischaemic events. Lancet 1996:348;1329 1339.
43. Executive Committee for the Asymptomatic Carotid Atherosclerosis Study. Endarterectomy for asymptomatic carotid artery stenosis. JAMA 1995:273; 1421.
44. Wolf PA, et al. Epidemiologic assessment of chronic atrial fibrllation and risk of stroke: The Framingham study. Neurology 1978:28;973 977.
45. Atrial Fibrillation Investigators. Risk factors for stroke and efficacy of antithrombotic therapy in atrial fibrillation: Analysis of pooled data from five randomized controlled trials. Arch Intern Med 1994:154;1449 1457.
46. Sudlow M, Thomson R, Thwaites B, Rodgers H, Kenny RA. Prevalence of atrial fibrillation and eligibility for anti-coagulants in the community. Lancet 1998:352; 167 1171.
47. Stroke Prevention in Atrial Fibrillation Investigators. Stroke Prevention in Fibrillation Study: Final results. Circulation 1991;84:527
48. Stroke Prevention in Atrial Fibrillation Investigators. Adjusted-dose warfarin versus low-intensity, fixed dose warfarin plus aspirin for high-risk patients with atrial fibrillation: Stroke Prevention in Atrial Fibrillation III randomized clinical trial. Lancet 1996:384;633 638.
49. The EC/IC Bypass Study Group. Failure of extracranial-intracranial bypass to reduce the risk of ischaemic stroke: Results of an intracranial randomized trial. N Eng J Med 1985:313; 191 1200.
50. Thrift AG, McNeil JJ, Forbes A, Donnan GA, for the Melbourne Risk Factor Study Group. Three important subgroups of hypertensive persons at greater risk of intracerebral haemorrhage. Hypertension 1998:31; 1223 1229.
51. Croog SH, Levin S, Testa M, et al. The effect of antihypertensive therapy on the quality of life. New Engl J Med 1986:314;1643 1664.
52. Hansson L, et al. Effect of angiotensin-converting-enzyme in inhibition compared with conventional therapy on cardiovascular morbidity and mortality in hypertension: The Captopril Prevention Project (CAPP. randomized trial. Lancet 1999:353;611 616.
53. Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure. The sixth report of the Joint national Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (MNC VI). Arch Intern Med 1997: 157;2413 2146.
54. Diener HC, et al. European Stroke Prevention Study 2: Dipyridamole and acetylsalicylic acid in the secondary prevention stroke. J Neurol Sci 1996:143; 1 13.
55. Jones EJ, et al. Proximal aortic atheroma: an independent risk factor for cerebral ischaemia. Stroke 1995:26;218 224.
56. Amarenco P, et al. Atherosclerotic disease of the aortic arch and the risk of ischaemic stroke. N Engl J Med 1994:331; 1474 1479.