Services on Demand
Print version ISSN 0034-7094
Rev. Bras. Anestesiol. vol.54 no.4 Campinas July/Aug. 2004
Evidence-based anesthesiology: what is it and how to practice it*
Anestesiología fundamentada en evidencias: lo que es y como practicar
Rodrigo Mariano da Costa de Angelis, TSA, M.D.I; Álvaro Avezum Júnior, M.D.II; Alexandre Biasi Cavalcanti, M.D.III; Rogério Teixeira de Carvalho, M.D.IV
Hospital Israelita Albert Einstein, São Paulo, SP
IICoordenador Executivo do Centro Brasileiro de Estudos Clínicos, Instituto Dante Pazzanese de Cardiologia; Coordenador de Centro de Pesquisa Clínica do Instituto de Ensino e Pesquisa do Hospital Israelita Albert Einstein, São Paulo, SP
IIIEspecialista em Cardiologia
IVMembro da Sociedade Brasileira de Ortopedia, Membro da Sociedade Brasileira de Cirurgia do Joelho
BACKGROUND AND OBJECTIVES:
The principles of evidence-based medicine have been applied to Anesthesiology.
The purpose of this article is to review the basis for the practice of evidence-based
medicine and give some examples of how its principles may be incorporated to
the clinical practice. We couldn't find any article about this issue in
the Brazilian medical literature and decided to perform it.
CONTENTS: This study is a review focusing on a new medical paradigm and how it can be applied to Anesthesiology.
CONCLUSIONS: Evidence-Based Anesthesiology is the integration of best scientific evidence, anesthesiologists' clinical experience and patients' expectations, aiming at improving patients' care.
Key Words: ANESTHESIOLOGY: evidence-based anesthesia; EPIDEMIOLOGY, Clinical
JUSTIFICATIVA Y OBJETIVOS:
Los principios de la Medicina fundamentados en evidencias han sido aplicados
a la Anestesiología. La propuesta de esa revisión es abordar los fundamentos
para la práctica de la Medicina fundamentados en evidencias y proporcionar
ejemplos de como eses principios pueden ser incorporados a la práctica
diaria. Hasta el momento de la conclusión de ese artículo, no fue
encontrado ningún trabajo sobre Anestesiología fundamentados en evidencias
en la literatura médica brasileña, lo que determinó su realización.
CONTENIDO: Este artículo consiste en una revisión que aborda un nuevo paradigma de la Medicina y como él puede ser aplicado a la Anestesiología.
CONCLUSIONES: La Anestesiología fundamentada en evidencias constituye la integración de la mejor evidencia científica disponible a la experiencia clínica del anestesiologista y a las expectativas del paciente, con la finalidad de la mejoría y cuidado con él.
Evidence-based medicine, expression used for the first time in Brazil by Avezum1, the origin of which is France, 19th Century, is the conscious, explicit and judicious use of the best evidence aiming at decision making for the treatment of individual patients2. Major evidence-based medicine principle is the effective use of medical literature as a guide for daily practice3. There are three key-elements: best scientific evidence, physicians' clinical experience and patients' preferences.
The expression evidence-based medicine was introduced in the medical literature approximately 20 years ago4, and in spite of being initially related to Internal Medicine, it has been incorporated to Anesthesiology. The first textbook on evidence-based Anesthesiology was published in 2000, in the UK5.
The last decade has witnessed an increasing interest on the subject, attributed by Sackett et al.6 to four basic reasons: the increasing need for knowledge about diagnosis, prognosis, therapy and prevention; limitations of traditional information sources, such as textbooks (often outdated), specialists (very often giving inaccurate opinions based on inadequate scientific grounds) and continued medical education (most not so effective); disparity between physicians' diagnostic ability and clinical reasoning, which tend to improve with age and experience, and their scientific knowledge and clinical performance, which tend to decrease with age and lack of time for studying, very often resulting in only 30 minutes/week dedicated to searching and assimilating knowledge4.
It is estimated that 3 hours and 19 minutes are needed every day for physicians to remain updated on their specialty.
Societies of Anesthesiology have tried to apply evidence-based Medicine principles to improve patients' care. As result, anesthesiologists are considered pioneers in the development and implementation of guidelines, which seem to have contributed to the sustained and disseminated decrease in anesthesia-related mortality and morbidity7. However it is difficult to establish a strong causal relationship between the activities of those Societies and the decrease in morbidity and mortality.
PRACTICE OF EVIDENCE-BASED MEDICINE
The practice of evidence-based medicine implies the integration of individual clinical experience and best available external evidences produced by clinical research8, and consists of five basic steps: translate information needs into a well structured clinical question; look for the best evidence to answer the question; evaluate the evidence in terms of validity and applicability to the clinical practice; integrate the best evidence to personal experience and patients' preferences to put it into practice; and evaluate the effectiveness of measures adopted in search for continuous improvement9.
How can evidence-based medicine be applied to Anesthesiology? It may be incorporated to our daily practice and also to review meetings, case discussions and teaching of the specialty. Anesthesiologists may replace review meetings and case discussions based on non-systematized reviews, by systematized reviews with clinical cases and literature evaluation.
CLINICAL QUESTION FORMULATION
A well-structured clinical question is the first step for decision making in evidence-based medicine. A good clinical question should be strictly relevant to the problem and asked in a way to help the search for the answer. Well-structured questions have four key-elements: (1) the patient or problem of interest; (2) major intervention, that is, exposure, diagnostic test, prognostic factor or treatment regimen; (3) comparison between exposure interventions, if applicable; (4) clinical outcome of interest6.
Questions related to therapy, diagnosis and prognosis are of special interest for anesthesiologists. Some examples of clinical Anesthesiology questions are: should patients submitted to total hip replacement receive general or regional anesthesia? Which is the best strategy to prevent postoperative vomiting? Which is the probability of one patient with restrictive pulmonary disease, whose preoperative vital capacity is 30% of normal, need postoperative mechanical ventilation? For patients who, in previous exposure to anesthesia, there has been suspicion of malignant hyperthermia, what is the ability of the in vitro contracture test to predict who will develop the problem? Would low molecular weight heparin for a female patient aged 80 years submitted to hip surgery improve her prognosis? Which is the risk for bronchospasm in a patient receiving rapacuronium as neuromuscular blocker?10,11
After developing a well-structured question the next step is to look for the best possible evidence.
SEARCH FOR THE BEST EVIDENCE
There are basically four ways to obtain information: ask somebody, check a textbook, find an article related to the subject in your personal files or use a database. Asking a professor or a colleague is a highly effective way; however it only makes sense when it is a rare situation or case which the professional will seldom face again. Check a textbook may be interesting, although they are in general outdated as from the moment of their publication. Refer to a personal file article implies the risk that data were not maintained updated and complete.
Currently, electronic databases are the fastest and more updated means to find information. Largest and most widely used data source is MedLine, which counts on more than 3500 indexed journals with approximately 11 million references from 1966 to date. Medical subject heading, MeSH, is the name of the standard vocabulary used to index MedLine references. The hierarchic database organization follows a branched structure which allows the user to use more specific or more generic terms to change the strategy of the search. MeSH query allows the search by subject, providing a better result as compared to searching just by words present on the text. MedLine may be accessed by the PubMed of the National Library of Medicine, by Ovid Technologies, MdConsult and other electronic information sources.
Specifically for Anesthesiology, information is spread throughout general medicine journals of multiple sub-specialties, including Anesthesiology itself, Pediatrics and Surgery, in addition to basic sciences.
The search for MedLine answers should be based on the four already mentioned clinical question elements.
MedLine provides access to most published biomedical literature. However, it does not critically evaluate database references. Obtained information analysis is the next step for the practice of evidence-based medicine. Physicians who read a lot but do not evaluate the validity and applicability of the source are well informed, but have not learned and will not have adequate subsidies to argue in favor of their management.
Five questions have to be asked for treatment studies12,13: (1) are study results valid? (2) Which were the outcomes? (3) Have treatment benefits overcome damages and costs? (4) Will outcomes help improve my patients' care? (5) Were all major clinical outcomes taken into consideration?
Are Study Results Valid?
First we have to investigate whether the study represents an unbiased estimate of the effects of the approach or if there is some type of bias, that is, systematic error in an epidemiologic study resulting in incorrect estimate of the association between exposure and outcome. This question may be answered by the following questions: (1) Were patients randomly selected for each treatment group? Random distribution decreases biases and assures that clinical outcomes determinants are homogeneously distributed among groups; (2) Were all selected patients computed? Have all patients starting the study finished it? Were patients who have not concluded the study included in the final analysis? (3) Were patients analyzed according to the groups they were randomly allocated to in the beginning of the study (treating intention principle)? (4) Is it a double-blind study? In a double-blind study, patients and investigators do not know the treatment option; (5) With the exception of the intervention being investigated, were patients treated equally throughout the study? This latter question is closely related to the previous one: in open studies, in which physicians and patients know the treatment being administered, both patients' behavior after knowing the actual treatment they are receiving and the way clinicians manage the case may be modified. End result will be that study results will not reflect just the effects of the treatment being tested, but also the effects of other treatments which are different among groups. Adopting the double-blind scheme may prevent this problem.
An example of this type of study related to Anesthesiology is a randomized double-blind study comparing drugs A and B as to their efficacy in preventing perioperative vomiting, such as the study performed by Tang et al., who have compared the efficacy of ondansetron and droperidol to control perioperative vomiting of elective gynecological surgeries14.
Which were the Results?
Two questions should be asked with regard to results: (1) Which has been the magnitude of observed therapeutic effect? (2) What is the accuracy of therapeutic effect estimate?
Consider that in the above-mentioned study about perioperative vomiting, the incidence of nausea and vomiting in the group of patients allocated to drug A has been 20% and that in group receiving drug B this has been observed in 15% of cases. The risk for perioperative nausea and vomiting in group B as compared to group A is 0.15/0.20 = 0.75. This is called relative risk (RR). Relative risk is the ratio of events rate in the treated group divided by events rate in control group (TER/CER). Another important concept is relative risk reduction (RRR) which is control group events rate less treated group events rate divided by control group events rate (CER - TER)/CER or (1 - RR) x 100. RRR is expressed in percentage. The new therapy (drug B) has decreased the risk for nausea and vomiting in 25% as compared to conventional therapy (drug A). But which is the accuracy of this 25% decrease? This depends on the number of patients enrolled in the study, that is, sample size. Let's say that the study included 100 patients; there would be 20 vomiting episodes in group A and 15 in group B, a difference of just 5 episodes. If the number of patients increased to 1000, there would be 200 episodes in group A and 150 in group B, a difference of 50 episodes. RRR is the same, but confidence interval will be narrower, indicating a higher accuracy of results. This happens because the higher number of events decreases the possibility of results having been reached by chance.
The confidence interval informs values within which the true effect should be located. It combines information about the strength of an association (such as relative risk reduction) with information about effects of chance on the observed result.
By convention, confidence interval used is 95%, meaning that there is 95% probability that actual population value is within the limit found. The larger the sample size, the higher will be the number of events observed and more convinced readers may be that the result is a true and more accurate representation of values observed in the general population.
Confidence intervals interpretation is relatively simple. In our previous example there are two situations. The study involving 100 patients had a confidence interval for RRR of -38% (increased incidence of nausea and vomiting with drug B) to 59% (decreased incidence of vomiting with drug B). Since 95% confidence interval includes zero, it is considered that the studied effect is not statistically significant. In the study involving 1000 patients, confidence interval was 9% to 41%, meaning that readers may be 95% sure that vomiting decrease with drug B is between 9% and 41%, strongly suggesting that drug B is better than drug A to decrease perioperative vomiting.
Jung14 has studied 161 female patients physical status ASA I, submitted to outpatient gynecologic surgeries. Patients were randomly allocated to receive one out of four regimens: placebo, droperidol (0.625 mg), droperidol (1.25 mg) or ondansetron (4 mg). This study was also double blind, that is, patients and investigators were blind to the drug each patient was receiving. Patients were evaluated within the groups they were initially allocated to. With the exception of the investigated intervention, patients were similarly treated throughout the study and p < 0.05 was considered statistically significant. The incidence of vomiting and the need for rescue anti-emetics has been significantly lower in the ondansetron and droperidol groups as compared to placebo (p < 0.005), but has not differed between droperidol (0.625 mg and 1.25 mg) and ondansetron groups.
Have Treatment Benefits overcome Damages and Costs?
This question should be answered through the concept of the number needed to treat (NNT). NNT quantifies the number of patients to be treated to obtain one additional favorable outcome or one unfavorable outcome less. NNT takes into account not only relative risk reduction but also the baseline risk for the adverse outcome such intervention is attempting to decrease. In mathematical terms, we have: NNT = 1/probability of adverse outcomes in the control group - probability of adverse outcome in the treatment group or (1/Absolute Risk Reduction (ARR). NNT is the reciprocate of absolute risk reduction (ARR), which is the events rate in controls less events rate in treated. RRR, ARR and NNT are different ways to summarize relative effects of two treatments15. Let's once more consider drug B with its 25% perioperative vomiting decrease. Let's admit that there is a group of patients at high risk for vomiting (40%) and one at low risk (5%). The decrease of 25% in the high risk group represents absolute risk decrease from 40% to 30%, that is, 10%. NNT will be (1/10) x 100 = 10. This means that 10 patients will have to be treated with drug B in the high risk group to prevent one postoperative vomiting episode. In the low risk group, absolute risk decreases from 5% to 3.75%, or a difference of 1.25%. NNT is (1/1.25) x 100 = 80. So, it is necessary to treat 80 low risk patients to prevent one vomiting episode. Before starting a treatment, physicians should consider patients' risk for the adverse event, because in general, the higher the baseline risk, the higher will be the benefits of the treatment.
Will outcomes Help Improve my Patients' Care?
This question could be answered by another question: which is the similarity of my patient and patients included in the study? Clinical trials are controlled experiments, with strict inclusion and exclusion criteria. Studies involving treatments determine the efficacy of an intervention. Efficacy means that the treatment decreases the probability of an adverse effect, strictly within study conditions. It is necessary to determine the effectiveness of the treatment, or how does the treatment work in practice. For such, physicians should consider inclusion and exclusion criteria and ask themselves whether there is a reason why study results should not be applied to their patients, that is, patients are much different than patients participating in the study.
Were all Major Outcomes taken into Consideration?
Treatment should impact outcomes, which are clinical events most interesting patients and physicians. There are basically 5 clinical outcomes: end (death), disease (symptoms, physical signs), distress (pain, nausea, dyspnea), functional impairment (limited ability to perform usual activities), unhappiness (emotional reaction to disease or care)15. Examples of clinical outcomes are: decreased myocardial infarction, prevent admission for congestive heart failure or decrease dyspnea of air for daily activities.
One alternative to clinical outcome is substitute outcome, which is a measure or clinical sign used to replace clinical outcome15. The advantage of the substitute outcome is that fewer patients may be involved due to the fact that such outcomes are in general selected for being more common and easy to be observed. That is, the substitute outcome decreases study costs and duration.
The basis for the substitute outcome is that changes on substitute outcome promoted by treatment should be reflected in changes in clinical outcome. However, this is often not true and is what has been seen in studies such as Cardiac Arrhythmia Suppression Trial, which has evaluated the use of anti-arrhythmics after acute myocardial infarction16. In preliminary studies, it has been observed that encainide and flecainide decrease the incidence of early ventricular contractions after myocardial infarction. It has been thought that they would also decrease the incidence of life-threatening severe arrhythmias. However, it has been observed that patients receiving such drugs had higher mortality rates as compared to patients receiving placebo.
Biological outcomes cannot be adequate substitutes for clinical outcomes without direct evidence that both are inter-related. The use of substitute outcomes in Anesthesiology is a special problem17. Currently, mortality as direct result of anesthesia is highly decreased. Few outcomes are directly attributed to anesthesia. In addition, Anesthesiology is a unique specialty in the sense that most interventions are not exactly therapeutic, but aimed at helping other procedures. Since the incidence of anesthesia-induced severe adverse events is low, many studies in the area use substitute outcomes. However, the latter should not definitively replace clinical outcomes. Clinical outcomes in Anesthesiology have increased, as anesthesiologists become physicians acting throughout the perioperative period. Economic outcomes, patients' satisfaction and improved quality of life have also been used as outcomes in clinical trials in Anesthesiology.
Three important questions should be asked18: (1) Are study results valid? (2) Which were the study results? (3) Will results help me improve patients' care?
Are Study Results Valid?
(1) Was an independent and blind comparison performed between the proposed diagnostic test and the widely used reference test?
(2) Does sample size include a spectrum of patients in whom the diagnostic test will be applied in the clinical practice?
(1) Have test results influenced the decision of performing the standard reference test? A weak test may be compared to other weak standard test considered the validity standard for being widely used or consensus among experts. If a new test is compared to an old but inaccurate test, the new test may seem worse even if being better. If the new test is more sensitive than the standard test, new patients additionally identified by the new test will be considered false-positives as compared to the old test. A classic example is what has happened in the study by Bartrum et al.19 who have compared dynamic ultrasound and oral cholecystography to detect billiary stones. Ultrasound was positive in 5 patients in whom no stones were detected by oral cholecystographies. Afterwards, two of these patients were operated on and stones were found, showing that the reference test was less accurate than the new test;
(2) Were methods used for diagnostic test described in enough detail to allow replication? When authors reach a conclusion on how to use a certain diagnostic test, they are to explain how to perform it.
Which were Study Results?
Once study results are considered valid, next step is to establish diagnostic test accuracy. For such, the concepts of sensitivity, specificity, positive and negative predictive values and likelihood ratio should be understood, because they are major measurements obtained by diagnostic studies.
Sensitivity is the percentage of individuals with the disease with a positive test. A sensitive test is chosen when there is major penalty for not diagnosing a disease, such as severe or transmissible diseases like tuberculosis, syphilis or malignant neoplasias. These are also useful tests to decrease the large number of diagnostic possibilities during early phases of an investigation15.
Specificity is the percentage of individuals without the disease with a negative test. A specific test will seldom misclassify healthy people as sick people and is useful to confirm diagnosis suggested by other data. A highly specific test is seldom positive in the absence of disease, that is, gives few false-positive results, being of interest when such results may physically, emotionally or financially hurt the patient.
Sensitivity and specificity are tests properties taken into account when deciding about asking for a test. With results in hands, sensitivity and specificity are not so relevant. As from this point, it has to be determined whether the patient has or not the disease given tests results. The probability of disease given test results is called test predictive value. Positive predictive value is the probability of disease in a patient with positive result. Negative predictive value is the probability of not having the disease when results are negative (normal).
The predictive value of a test is not a property unique to the test; it depends on the prevalence of the disease among studied population, being prevalence the number of people with the studied condition in a defined population in a certain moment.
Another important concept is likelihood ratio (odds ratio) for a certain value of a diagnostic test, defined as the probability of such result in people with the disease, divided by the probability of the same result in healthy people. Likelihood ratio expresses how many times more (or less) likely it is to find a test result in sick people as compared to healthy people.
Will Results help Improving my Patients' Care?
The third step is to decide whether the test is applicable to a specific individual and to daily practice patients. Questions to be asked are: may study results be generalized, that is, can the physician apply them to patients most frequently seen in his/her daily practice? Does this diagnostic test supply additional information to the history and physical evaluation? Is the new proposed test cheaper and more accessible than other available tests for the same aim?
Prognostic refers to possible outcomes of a disease and the frequency in which they may occur. Often, the characteristics of a certain patient may be used to more accurately predict the outcome. These characteristics are called prognostic factors and may be of several types: demographic (e.g., age, gender), specific of the disease (e.g. tumor stage) or co-morbidity (e.g., other conditions following the disease).
In general, it is anti-ethic to randomly act with patients for them to be exposed to different prognostic factors. So, the best study design to identify risks and determine their increase associated to a prognostic factor is the cohort study. In this study, one or more groups of individuals who have still not suffered an adverse event, or cohorts, are followed for a time period in which events have occurred within groups. As from these studies, it is possible to compare events rate in the cohort exposed to a certain factor to events rate in the cohort not exposed, and obtain risk estimates, such as relative risk (events rate among exposed people/events rate among non-exposed people). The optimal cohort study is that consisting of a representative sample of general population using objective criteria to define outcomes.
Another way to study prognostic factors is to compare groups of individuals who have already presented the clinical outcome to a control group without clinical outcome. In this type of study (case-control), investigators count the number of individuals in each group with a certain prognostic factor. The potential for sampling biases, as well as the retrospective nature of data collection, limits the level of interference clinicians may obtain from those studies. However, case-control studies are often the only feasible way to answer questions about etiology, especially because they help the study of rare diseases.
Cohort and case-control studies are observations of risk factors. Three major questions should be asked for the interpretation of prognostic studies: (1) Are study results valid? (2) Which were the results? (3) Results obtained will help me improve patients' care?
Are Study Results Valid?
1) Was there a representative and well-defined sample of patients at similar stages of the disease? Many biases related to patients' selection might distort results. Authors should describe the stage of the disease in which patients entered the study;
2) Was patients' follow-up sufficiently long and complete? The smaller the number of patients available for follow up, the lower the accuracy to estimate the risk of an adverse outcome. It is important to consider the ratio between the percentage of patients not available for follow up and patients suffering an outcome of interest. The higher the number of patients with unknown outcome as compared to the number suffering an event, the higher the threat to the validity of the study. On the other hand, if the reasons for patients' disappearance is unrelated to clinical outcome and if they are similar to those completing the study, the reader may be more comfortable as to study results validity.
1) Were outcome criteria objective and unbiased? Investigators should supply a clear and scientifically adequate definition of adverse outcomes before starting the study. To decrease bias, the individual defining the outcome should not have access to patients' prognostic factors;
2) Was an adjustment performed for major prognostic factors? In comparing the prognosis of two groups of patients, investigators should check whether their clinical characteristics are similar and adjust the analysis for differences found.
Which were the Results?
Quantitative prognostic or risk studies results are the number of events occurring in a certain period.
It is important to describe major effect measures obtained in studies about etiology and prognosis: relative risk obtained in cohort studies and odds ratio obtained in case-control studies.
In a cohort study, susceptible population is initially divided in two groups: exposed and not exposed to a certain factor. How to decide whether there is increased risk? Relative Risk should be calculated and corresponds to the incidence of disease in exposed patients divided by the incidence of disease in non-exposed people.
Case-control studies start with the selection of a group with the disease and a control group. There is no way to know the incidence of the disease because groups are selected according to investigators' selection criteria. So, it is impossible to obtain the relative risk. In this case, relative frequency of people exposed to the factor in cases and controls may be calculated. The comparison of the exposure between cases and controls gives a risk measurement conceptually and mathematically similar to relative risk, which is odds ratio. Odds ratio, also called unlikelihood ratio, is defined as the chances of a case being exposed divided by the chances of a control being exposed. If exposure frequency is higher among cases, odds ratio will be higher than 1, indicating increased risk. So, the more intense the association between exposure and disease, the higher will be the odds ratio. So, the meaning of odds ratio is similar to that of relative risk. However, odds ratio is approximately equal to relative risk only when there is low incidence of the disease. So, as already mentioned, case-control studies are indicated for the study of rare diseases.
Will Results help me Improve Patients' Care?
Three questions should be asked: (1) are studied patients similar to my patient? Authors should describe studied patients in detail, enough to allow a comparison with your patients. (2) Will results lead me to select or exclude a therapy? Prognostic data are often the basis for deciding the therapy. Knowing the potential clinical course of your patient's disease may help decide whether a treatment should or should not be used. (3) Are results useful to comfort or counsel patients?
DOES EVIDENCE BASED MEDICINE IMPROVE PATIENTS' CARE?
Emerging literature suggests that the practice of evidence-based medicine improves patients' care. In the last 40 years, knowledge provided by randomized clinical trials has been the basis for the evolution of medical therapy concepts21,22. There are no evidences from randomized clinical trials that evidence-based medicine is related to better clinical outcomes, but there are evidences from effectiveness studies that patients receiving evidence-based therapy have better clinical outcomes as compared to those not receiving it, and that residents trained according to evidence-based medicine principles use these principles and adequate treatments more often23.
In addition, medical decision-making computer systems, strongly based on current scientific data, may change physician's behavior and improve patients' management24,25.
Evidence-based medicine is a new paradigm emphasizing critical and investigative reasoning. Although initially related to internal Medicine, evidence-based medicine practice has been incorporated to Anesthesiology and its integration to anesthesiologists' daily practice may improve patients' management.
01. Avezum A - Cardiologia baseada em evidências e a avaliação crítica da literatura cardiológica: princípios de epidemiologia clínica aplicada à cardiologia. Rev Soc Cardiol Estado de São Paulo, 1996;3:241-259. [ Links ]
02. Logan AS - Investigation of Toronto General Practioner's Treatment of Patients with Hypertension. Toronto: Canadian Facts, 1978. [ Links ]
03. Schulman SR, Schardt C, Erb T - Evidence-based Medicine in Anesthesiology. Current Opinion in Anesthesiology, 2002;15: 661-668. [ Links ]
04. Evidence-based Medicine Working Group. Evidence-based medicine: a new approach to teaching the practice of medicine. JAMA,1992;268:2420-2425. [ Links ]
05. Tramer M - Evidence-based resource in anesthesia and analgesia. London: BMJ Books, 2000. [ Links ]
06. Sackett DL, Straus SE, Richardson WS et al - Evidence-Based Medicine: How to Practice and Teach EBM. New York, Churchill Livingstone, 2000. [ Links ]
07. Gaba D, Howard SK, Fish KJ - Crisis Management in Anesthesiology. New York, Churchill Livingstone, 1994. [ Links ]
08. Cook DJ, Levy MM - Evidence-based medicine. A tool for enhancing critical care practice. Crit Care Clin, 1998;14:353-358. [ Links ]
09. Oxman AD, Sackett DL, Guyatt GH - Users guides to the medical literature: How to get started. JAMA, 1993;270:2093-2095. [ Links ]
10. Rajchert DM, Pasquariello CA, Watcha MF et al - Rapacuronium and the risk of bronchospasm in pediatric patients. Anesth Analg, 2002;94:488-493. [ Links ]
11. Schulman SR - Rapacuronium redux. Anesth Analg, 2002;94:483-484. [ Links ]
12. Guyatt GH, Sackett DL, Cook DJ - Users guides to the medical literature: II. How to use an article about therapy or prevention: A. Are the results of the study valid? JAMA, 1993;270: 2598-2601. [ Links ]
13. Guyatt GH, Sackett DL, Cook DJ - Users guides to the medical literature: II. How to use an rticle about therapy or prevention: B. What were the results and will they help me in caring for my patients? JAMA, 1994;271:59-63. [ Links ]
14. Tang J, Watcha MF, White PF - A comparison of costs and efficacy of ondansetron and droperidol as prophylactic antiemetic therapy for elective outpatient gynecologic procedures. Anesth Analg, 1996;83:304-313. [ Links ]
15. Fletcher RH, Fletcher SW, Wagner EH - Epidemiologia Clínica: Elementos Essenciais 3ª Ed, Porto Alegre, Artmed, 1996. [ Links ]
16. Cardiac Arrhythmia Suppression Trial (CAST) Investigators. Preliminary report: effect of encainide and flecainide on mortality in a randomized trial of arrhythmia suppression after myocardial infarction. N Engl J Med, 1989;321:406-412. [ Links ]
17. Fisher DM - Surrogate outcomes: meaningful not! Anesthesiology, 1999;90:355-356. [ Links ]
18. Jaeschke R, Guyatt GH, Sackett DL - Users guides to the medical literature: III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? JAMA, 1994;271:703-707. [ Links ]
19. Bartrum Jr RJ, Crow HC, Foote SR - Ultrasonic and radiographic cholecystography. N Engl J Med, 1977;296:538-541. [ Links ]
20. Laupacis A, Wells G, Richardson WS et al - Users guides to the medical literature: V. How to use an article about prognosis. JAMA, 1994;272:243-237. [ Links ]
21. Yusuf S - Randomised controlled trials in cardiovascular medicine: past achievements, future challenges. BMJ, 1999;319: 564-568. [ Links ]
22. Baum M, Houghton J - Contribution of randomised controlled trials to understanding and management of early breast cancer. BMJ, 1999;319:568-571. [ Links ]
23. Kitchens JM, Pfeifer MP - Teaching residents to read the medical literature: a controlled trial of a curriculum in critical appraisal/clinical epidemiology. J Gen Intern Med, 1989;4:384-387. [ Links ]
24. Durieux P, Nizard R, Ravaud P et al - A clinical decision support system for prevention of venous thromboembolism: effect on physician behavior. JAMA, 2000;283:2816-2821. [ Links ]
25. Randolph AG, Haynes RB, Wyatt J et al - Users guides to the Medical Literature: XVIII. How to use an article evaluating the clinical impact of a computer-based clinical decision support system. JAMA, 1999;282:67-74. [ Links ]
Dr. Rodrigo Mariano da Costa de Angelis
Address: Rua José Maria Lisboa, 155/84
ZIP: 01423-000 City: São Paulo, Brazil
Submitted for publication August
Accepted for publication November 6, 2003
* Received from Instituto de Ensino e Pesquisa do Hospital Israelita Albert Einstein, São Paulo, SP