Evidence-based medicine—does it apply to my particular patient?
Evidence-based medicine—does it apply to my particular patient?
Proponents of evidence-based medicine (EBM) have established a clear, unambiguous requirement for what they consider credible evidence, the randomized controlled trial (RCT), and especially the systematic review of several RCTs. They propose that clinical practice should be dominated by adherence to the ‘evidence’ as they define it.
It would be obtuse to argue that the doctor should ignore evidence from RCTs or aggregates thereof, but taking care of complex patients is very different from care during trials. Many conditions are unsuitable for trials, and many patients are not included in trials. Patients selected are often not representative of the conditions seen in the clinic by practitioners. Furthermore, there are important limitations in trial design and analysis that make the ‘evidence’ not very practically useful in everyday practice.
Instead of basing decisions largely on trial results of homogenized groups of patients, an alternative approach advocated in this chapter emphasizes spending more time at the bedside and in the clinic, finding out exactly what is wrong with each patient, and getting to know each patient and their circumstances, family situations, psychosocial and economic stresses, thoughts, fears, biases, and wishes. Therapeutic decisions are made with, by, and for complex individuals. Therapeutic decisions should be guided by detailed knowledge of the pathology, pathophysiology and circumstances in individual patients. One size does not fit all or most patients.
What is ‘evidence-based medicine’ and how is it new?
The latest medical crusade is to render the care of patients evidence-based. This term has become a shibboleth, a sacrosanct icon almost like motherhood. Who could possibly be against basing decisions on evidence? The Oxford College Dictionary defines evidence as ‘something that furnishes proof; an outward sign; an indication, testimony’. Haven’t doctors always prided themselves in having some evidence behind treatment selection? It is difficult to think of a polite term for actions and decisions not based on any evidence. The change from the past, however, is that advocates of evidence-based medicine have established a clear unambiguous requirement for what they consider credible evidence—the randomized controlled trial and especially the systematic review of several randomized controlled trials. But the almost religious zeal for cloaking all decisions about patient care under the banner of ‘evidence-based’ misses the real problem: that is, how well does the evidence from trials apply to the care of individual patients? Governmental organizations, insurance companies, and other funders embrace this new concept of evidence-based medicine since few treatments meet the strict criteria. They would prefer not to pay. He who pays the piper calls the tune.
In contrast to the situation envisioned by evidence-based medicine zealots, courts of law accept and evaluate many different types of ‘evidence’ and testimony. Judicial decisions depend on how the evidence applies to the individual case being considered.
How do trial data relate to treating individual patients?
Physicians in the free world are not compelled to give all patients with a condition the same defined treatment, as is the case in trials. Randomized trials mandate that numbers of patients with a general condition will be given treatment X and the results will be compared with patients given treatment Y or Z or placebo.
Doctors care for one patient at a time. Caring for individual patients is complicated. Care involves: (1) understanding what is wrong with the patient, and (2) understanding the patient’s risks for disease, and (3) understanding the patient, their background, genetics, socioeconomic milieu, psychology, responsibilities, goals, etc., and (4) understanding the benefits and risks of potential therapeutic strategies to treat the patient’s conditions (often multiple) and to prevent conditions that they are at risk for developing, and (5) communicating with the patient and sometimes family members and friends, listening and conveying information, and teaching. These functions are extensive and often difficult. They require much innate intelligence, experience, sensitivity, and training. They take time, a commodity now often jeopardized by large patient lists, managed care directives, and the need to support oneself and one’s family. ‘Evidence-based medicine’ relates to only one of these doctor functions, number 4 above.
George Thibault said it very well:
We then need to decide which approach in our large therapeutic armamentarium will be most appropriate in a particular patient, with a particular stage of disease and particular coexisting conditions, and at a particular age. Even when randomized clinical trials have been performed (which is true for only a small number of clinical problems), they will often not answer this question specifically for the patient sitting in front of us in the office or lying in the hospital bed.
Do randomized trials have theoretical and practical limitations?
Trials, the core of evidence-based medicine, have important theoretical and practical limitations. They are expensive, time consuming, and require enormous resources. To provide statistically valid results, randomized trials must contain large numbers of patients with enough end points for analysis. Sufficient endpoints must be obtained in a relatively short period. The condition studied must either be acute and cause adverse endpoints or rapid improvement within a short time. Chronic conditions must be severe enough to cause clear end points within 1 to 5 years of follow-up. Many medical conditions are unsuitable for study by trials. Less common, heterogeneous, and chronic conditions are difficult to study in trials. Patients who are too ill, too old, too young, female and ‘of childbearing age’, incapable of giving informed consent, too complex, or too full of coexisting illnesses are often not included in trials. But these are just the patients who visit doctors in their office and are under their care in the hospital.
The major theoretical limitation of trials is the issue of numbers vs specificity. For trials to yield statistically valid results, they must include many patients—numbers. For the results to be useful to practising physicians, the data must specifically apply to individual patients with the condition studied. To include enough patients, the condition to be studied must be common, and usually multiple physicians at multiple centres must be used. A single doctor or medical centre would have too few patients or would take an unacceptably long time to accrue the number of patients needed. To achieve numbers, a lumping strategy must predominate over splitting. For example, to study the effectiveness of a treatment to prevent embolism in patients with mitral valve prolapse, a study would not be able to obtain enough patients with mitral valve prolapse, mitral regurgitation, and mitral valve fibrinoid degeneration who had prior brain or systemic emboli and congestive heart failure even though this group is at highest risk and would be most likely to respond to prophylaxis. The study would have to include all patients with mitral valve prolapse to recruit enough patients.
The sample size of a trial will increase if: the projected effectiveness of the treatment (the percentage reduction in adverse outcomes) is low (10–15%); a number of treatments will be studied; the follow-up period is short (otherwise many patients will be lost to follow-up, withdraw, or become noncompliant); the anticipated outcome event rate is low; and a high power of protection from type I and type II errors is desired. As an example of the extraordinary numbers of patients needed for some studies, the authors of a meta-analysis of randomized controlled trials of agents that decrease platelet aggregation for the secondary prevention of stroke calculated that 13 000 patients would be needed to detect, with 90% power, an observed reduction of 15% in endpoints with aspirin.
The greater the numbers of patients required, the more pressure there is to adopt a lumping strategy. The more a study lumps diverse subgroups, the more general are the results and their applicability to specific patients declines. For practising physicians, treatment must be very specific. Physicians are faced with individual patients for whom they must make therapeutic decisions. To be useful, trial results must help physicians treat individual patients in given situations. Subgroups can be managed either by prospective stratification, that is, by randomizing patients using predetermined criteria (e.g. sex, race, age) to ensure that subgroups will be relatively equally represented in the different treatment groups, or by analysing the treatment results by subgroup determinants that have been prospectively defined. But the subgroups must also be very large to satisfy statisticians.
A confounding issue is the number of treatments and agents that patients receive. For example, many trials consider secondary stroke prevention after an initial transient ischaemic attack or stroke. Patients are enrolled even 3 to 6 months after their initial cerebrovascular event. These trials have not taken into consideration the effect on the outcome of the initial treatment before randomization. Experience and outcome data show that treatments that are effective during the initial event (e.g. thrombolytics, mechanical clot extraction, anticoagulants, antiplatelets) often have durability and greatly influence the occurrence of later events. If initial treatments are taken into consideration, a much larger number of patients would be required to assess the results. Similarly, many patients with vascular disease are treated with polypharmacy prior to enrolment. It is difficult to balance the control and various treatment groups for all potential drugs and combinations of drugs.
There are also many practical problems confronting trialists. The logistics of performing randomized therapeutic trials can be problematical. Trials are big operations. Multiple centres require many physicians and clerical staff. Computer hardware and software and statistical skills are needed to record, manage, interpret, and analyse data. The personnel and equipment are very costly. Money for funding comes either from governmental, or private sources, most often pharmaceutical or device companies. Much time and effort is expended in writing grants and many pilot data are required. Government funding is becoming scarcer all over the world. Alternatively, private industry may be interested in funding grants if their products are being studied. Potential problems arise from involvement of private enterprises that have much to gain and much to lose, depending on trial results. Many companies strive to dictate trial methodology and/or play a role in analysis and publication of the results as a condition for funding studies. Companies have bailed out of studies depending on company finances and goals. Many worthwhile trials go unfunded.
Inclusion and exclusion criteria are designed so that patients entered will be ‘pure breed’ and can be followed until study completion. Most severe intercurrent diseases exclude patients, as do relative contraindications to treatments studied. Comorbid conditions such as alcoholism, cancer, liver, lung, blood, and renal disease are exclusions. The plethora of exclusions often make it difficult to recruit enough patients to meet sample size requirements. Estimates of the number of patients a centre predicts it will recruit are usually at least two or three times more than they actually manage to enter once the trial begins. In some trials, patients who are eligible under the inclusion/exclusion rules of a trial are not entered by physicians who feel that a particular patient needs the treatment and should not be randomized.
Eligible patients are not always easy to enrol in trials. Many patients decline because they don’t want to be guinea-pigs and view trials as something others do, especially charity cases. Some patients are put off by the acknowledged lack of a scientific basis for treatment and cannot accept that a flip of the coin will decide treatment. Some are disturbed that neither they nor their physicians will know what treatment they receive. Especially disconcerting for many is the prospect that they may receive a placebo. They believe their problem is serious and warrants active treatment. With time and patience, some of these patients can be enrolled, but with much effort. Alas, some who have enrolled will be dissuaded later by their all-knowing friend or relative, and will drop out. To document that all procedures have been followed and all necessary examinations and evaluations have been performed, most studies require mountains of paper. Completion of forms takes time. Often, the filling out of forms is delegated to a clerk, a resident, or the most junior investigator. The validity of the data is thus jeopardized. The results are only valid if the data are reliable and accurate. Senior experienced clinicians should have seen all patients and personally reviewed the forms to ensure accuracy, but this is often not done.
To compare the effectiveness of different treatments, outcomes must be measured and quantified. Trials that study stroke prevention are an example. This is simple if large events such as death or new stroke are used, but are all strokes equal? In nonfatal diseases, other criteria, e.g. severity of deficits, disability, or other objective measures, must be used. Especially in neurology, severity and disability scores are problematic. How can aphasia be compared with diplopia, ataxia, facial numbness, and limb weakness? How are weights assigned to various abnormalities? For some patients, a hemianopia that makes reading difficult but does not effect daily living poses no major problem but to a physician, editor, or surveyor the same deficit is devastating.
Are some recommendations based on trial results useful?
Randomized trials that study common, relatively homogeneous, specific, acute conditions have been quite helpful to practising physicians. A randomized trial of high concentration oxygen therapy given to premature infants showed that blindness due to retrolental fibroplasia was an important complication of this treatment. The results prevented innumerable cases of blindness. Before this trial, renowned professors of paediatrics had embraced this treatment. Many trials and analyses have clarified the indications for carotid artery surgery in symptomatic patients with various severities of arterial narrowing. Many trials and analyses have shown that warfarin anticoagulation is superior in preventing strokes in patients with atrial fibrillation than aspirin and placebo treatment. Unfortunately, many doctors have not followed the recommendation to anticoagulate patients with atrial fibrillation unless there are important contraindications. Cardiac revascularization conveys more frequent and better survival from cardiac shock than medical stabilization. There are many such examples of very useful trials that have promoted changes in treatment of patients.
Are other trial-based guidelines and recommendations less useful?
Some other randomized trials are much less useful to practising physicians. In this category are the stroke prevention trials of drugs that decrease platelet aggregation. Some studies showed in full group analyses a benefit for aspirin, aspirin combined with sulfinpyrazone or dipyridamole, ticlopidine, and clopidogrel. Unfortunately, the mixture of patients treated with antiplatelet aggregants or placebo was probably not representative of patients in the community presenting with transient ischaemic attacks or minor strokes. In none of these studies was clarification of the nature and severity of the causative vascular and cardiac lesions required for entry. Patients with lesions thought favourable for carotid surgery were often operated on and were ineligible. Patients with ‘surgical’ lesions deemed unfit for surgery, and those unfit for angiography, were included in medical treatment groups. Some patients with detected cardiac sources of emboli were not entered. No systematic evaluation for carotid artery or cardiac disease was mandated. Subgroup analysis was only by sex and tempo of ischaemia (transient ischaemic attack, reversible ischaemic neurological deficits, minor stroke). The tempo of ischaemia does not predict the nature, severity, or location of causative vascular lesions. Since cardiac studies were not required, the groups must also have contained patients with cardiac-origin brain embolism as the cause of their brain ischaemia. A meta-analysis of randomized control trials of antiplatelet agents in the secondary prevention of stroke found that for aspirin compared with placebo there was a nonsignificant reduction in stroke of 15% and a trend in reduction of stroke for any regimen containing aspirin. The results of these studies are difficult for physicians to apply to individual stroke patients with identified stroke mechanisms, e.g stenosis of the vertebral artery, arterial dissection, fibromuscular dysplasia, cardiogenic embolism. In defence of the studies cited, the technology now available—high quality duplex ultrasound scans, pulsed and continuous wave Doppler ultrasound, transcranial Doppler ultrasound, CT angiography, brain MRI and MR angiography, and echocardiography—was not widely available when the studies were designed. To recruit enough patients, the decision was made not to require angiography for entry (the numbers vs specificity issue). The result is that, despite enormous expense, the data are not very useful for physicians treating patients with the conditions studied in the trials. Future trials of antiplatelet aggregants should be conceived differently and have sufficient subgroup data related to the presence and severity of vascular lesions to be meaningful to practising physicians.
Guidelines for thrombolysis in acute stroke patients are still based on a single study funded by the United States government, planned nearly two decades ago and published more than 10 years ago. Release of the results of the National Institute of Neurological Diseases and Stroke (NINDS) trial gave momentum to a movement to quickly introduce intravenous thrombolysis widely into the community. During the summer of 1996, about 6 months after the publication of the NINDS trial, the United States Food and Drug Administration approved the use of recombinant tissue plasminogen activator (rt-PA) for the treatment of stroke patients when the drug was given within the first 3 h. The NINDS trial did not include patients treated after 3 h. The American Heart Association and American Academy of Neurology published treatment guidelines that recommended intravenous administration of rt-PA according to the NINDS protocol. The recommendations suggested that a CT scan performed before thrombolysis should not show major infarction, mass effect, oedema, or haemorrhage. The guidelines did not require or suggest MRI or vascular tests before treatment, despite the fact that stroke is of course a vascular disease. The inclusions and exclusions of the NINDS trial were copied in the recommendation. Patients who had minor deficits, were improving, awakened with deficits, or had seizures were not recommended for intravenous thrombolysis. The recommendations have never been updated.
Before and since the NINDS trial, many thousands of patients have been given stroke thrombolyis throughout the world. Early studies involved angiography before and after intravenous or intra-arterial thrombolysis. These studies showed convincingly that outcome correlated highly and consistently with opening of arterial occlusions and reperfusion. Many patients treated after 3 h improved after thrombolysis, depending on the presence and amount of brain infarction and the nature and location of the arterial occlusion. These studies were only observational, since controls were seldom used and patients were not randomized, but successive patients meeting protocol requirements were treated. Unfortunately the results of these early studies were not included in the publications of the results of the NINDS and randomized European thrombolytic trials.
Many patients who awaken with deficits can be studied using modern brain and vascular imaging to determine the potential salvageability of brain tissue by thrombolysis. Patients with minor or improving deficits frequently worsen or crash in the ensuing 24 h after onset. Some patients with seizures have acute vascular occlusions as the cause. Modern brain and vascular imaging can be done safely and quickly and can yield the key information needed to logically choose those patients who are likely to benefit from thrombolysis, those who are likely to be harmed, and the best route of administration of the agent or the potential of device-engendered reperfusion.
The present thrombolytic recommendations are an embarrassment. They prevent many potentially eligible patients from being treated. They expose other patients who do not have occlusive arterial lesions to needless risks. Less than 5% of potentially eligible acute stroke patients are given thrombolysis. It should be obvious to even the most naive person that patients do not change from good candidates (queens) to no candidate (pumpkins) when the clock strikes four. Use of a clock and a plain CT scan to screen candidates in 2009 is archaic and is completely outmoded except in regions where skills and technological equipment are lacking. Organizations have failed to update the currently outmoded recommendations because the important information (evidence) gleaned from decades of experience is not ‘evidence-based’ according to their very strict criteria.
We need more and better randomized therapeutic trials designed by clinicians to answer clinically relevant specific therapeutic problems. We need more critical reviews of trials and therapeutic dilemmas by experienced senior clinicians. Inexpert reviews by young academics often miss nuances and frequently lack clinical perspective and experience. All available information, not just that gleaned from randomized trials, should be included.
The panacea and saviour for medical therapeutics is not, and will not be, randomized trials or evidence-based reviews or meta-analyses. There are too many situations that cannot be clarified by trials. In other conditions general results are hard to apply to complex patients. Some envisage that the bulk of medical care will be delivered by primary care physicians who will spend much time at the computer reviewing evidence bases to guide therapeutic decisions. The role of specialists who have extensive experience and training in treating patients within their fields of expertise is minimized. After all, specialists are thought to be more expensive than primary care physicians (although no credible data proves this assumption). What a nightmare for present patients and for we doctors who ultimately will also become patients. Instead, I suggest that more time should be spent by general physicians and specialists at the bedside and in the clinic finding out exactly what is wrong with each patient, and getting to know each patient and their circumstances, family situations, psychosocial and economic stresses, thoughts, fears, biases, and wishes. Therapeutic decisions are made with, by, and for complex individuals. They cannot be readily homogenized without losing the essence of what being a doctor is all about.
Antithrombotic Trialists’ Collaboration (2002). Collaborative meta-analysis of randomised trials of antiplatelet therapy for prevention of death, myocardial infarction, and stroke in high-risk patients. BMJ, 324, 71–86.Find this resource:
Biller J, et al. (1998). Guidelines for carotid endarterectomy. A statement of healthcare professionals from a special writing group of the Stroke Council, American Heart Association and the American Academy of Neurology. Stroke, 29, 554–62.Find this resource:
Caplan LR (1988). TIAs—we need to return to the question, ‘What is wrong with Mr Jones?’ Neurology, 38, 791–3. [Strongly advocates thorough diagnosis of the cause in each patient with brain ischaemia.]Find this resource:
Caplan LR (2000). Caplan’s Stroke, a clinical approach, 3rd edition. Butterworth-Heinemann, Boston, MA. [Contains reviews of the results of trials and other data regarding treatment of patients with minor strokes and TIAs, carotid endarterectomy, atrial fibrillation, and stroke thrombolysis.]Find this resource:
Caplan LR (2001). Evidence-based medicine. Concerns of a clinical neurologist. J Neurol Neurosurg Psych, 71, 569–76.Find this resource:
Chimowitz MI, et al. (2005). Comparison of warfarin and aspirin for symptomatic intracranial arterial stenosis. N Engl J Med, 352, 1305–16.Find this resource:
Hochman JS, et al. (2006). Early revascularization and long-term survival in cardiogenic shock complicating acute myocardial infarction. JAMA, 295, 2511–15.Find this resource:
Mohr JP, et al. for the Warfarin-Aspirin Recurrent Stroke Study Group (2001). A comparison of warfarin and aspirin for the prevention of recurrent ischemic stroke. N Engl J Med, 345, 1444–51.Find this resource:
National Institute of Neurological Disorders and Stroke rt-PA Study Group (1995). Tissue plasminogen activator for acute ischemic stroke. New Engl J Med, 333, 1581–7.Find this resource:
Quality Standards Subcommittee of the American Academy of Neurology (1996). Practice advisory: Thrombolytic therapy for acute ischemic stroke—summary statement. Neurology, 47, 835–9.Find this resource:
Sackett DL, et al. (1996). Evidence-based medicine. How to practice and teach EBM. Churchill Livingstone, Edinburgh.Find this resource:
Stroke Prevention in Atrial Fibrillation Investigators (1991). The stroke prevention in atrial fibrillation study: final results. Circulation, 84, 527–39.Find this resource:
Thibault GE (1993). Clinical problem solving: Too old for what? N Engl J Med, 328, 946–50.Find this resource: