ebm self instructional manual 2008[1]

EVIDENCE BASED MEDICINE

Self-Instructional Manual

Noel L. Espallardo, MD, MSc

Department of Clinical Epidemiology

UP College of Medicine and Philippine General Hospital

Copyright, 2008

TABLE OF CONTENTS CONTRIBUTORS

i

ACKNOWLEDGEMENT

i

EVIDENCE BASED MEDICINE Introduction

1

SEARCHING THE MEDLINE Session Briefing Searching the MEDLINE Database

6

DIFFERENTIAL DIAGNOSIS Session Briefing Clinical Decision on Differential Diagnosis Workshop Briefing Appraisal Sheet

15

DIAGNOSTIC TEST Session Briefing Clinical Decision on a Diagnostic Test Workshop Briefing Appraisal Sheet

25

THERAPY OR PREVENTION Session Briefing Clinical Decision on Therapy or Prevention Workshop Briefing Appraisal Sheet

35

TABLE OF CONTENTS (CONT’D) HARMFUL EFFECT Session Briefing Clinical Decision on Harm Workshop Briefing Appraisal Sheet

48

PROGNOSIS Session Briefing Clinical Decision on Prognosis Workshop Briefing Appraisal Sheet

58

HEALTH ECONOMIC ANALYSIS Session Briefing Clinical Decision on Health Economic Analysis Workshop Briefing Appraisal Sheet

69

SYSTEMATIC REVIEW OR META‐ANALYSIS Session Briefing Clinical Decision on Systematic Review or Meta‐analysis Workshop Briefing Appraisal Sheet

80

CLINICAL PRACTICE GUIDELINES Session Briefing How to Use a Clinical Practice Guideline Workshop Briefing Appraisal Sheet

90

Evidence-based Medicine: Self-instructional Manual

Page i

CONTRIBUTORS Dr. Leilani Apostol and Dr. Ma Elinore Alba of the Department of Family and Community Medicine, UP College of Medicine and Philippine General Hospital They were dedicated students in the Master of Science program of the Department of Family and Community Medicine, UP College of Medicine and Philippine General Hospital. Their knowledge and experience in conducting evidence‐based medicine training in Family Medicine have been very helpful in designing this package.

ACKNOWLEDGEMENT Dr. Noel Juban and the Faculty of the Department of Clinical Epidemiology, UP College of Medicine The department is considered as the center of evidence‐based medicine in the Philippines. The staff provided essential advice and feedback in the application of evidence‐based.

Evidence-based Medicine Self-Instructional Manual

Page 1

EVIDENCE BASED MEDICINE INTRODUCTION Medicine is a dynamic endeavor. Everyday challenging problems arise, new modalities of treatment are promoted, disease management done under minimal or far from ideal conditions. Suppose a patient who consulted for cough productive of yellowish phlegm and asked for a prescription of an antibiotic. Will you prescribe it or not? These are common problems that may escape our attention and diminish the quality of care we give if we make inappropriate decisions. In the old practice faced with this question, a physician will just ask a colleague or an expert for the answer or rely on his/her prior knowledge of the disease. He may also prescribe a drug because of the promotional lecture sponsored by the manufacturer. In evidence‐based medicine (EBM) a new paradigm is introduced. Before he makes a decision, the physician will first try to retrieve his latest article about the topic that he kept from his file, appraise the article then makes a decision. Later, he evaluates the effectiveness of his decision. This loop ensures improvement in the quality of care.

DEFINITION OF EBM Evidence‐based medicine is defined as the process of systematically finding, appraising, and using contemporaneous research findings as the basis for clinical decisions (NLM, 2008). Evidence‐based medicine follows four steps:

• formulate a clear clinical question from a patient's problem • search the literature for relevant clinical articles • evaluate (critically appraise) the evidence for its validity and usefulness • implement useful findings in clinical practice.

THE PARADIGM SHIFT The traditional method of answering clinical problems was based on the following assumptions (User’s Guide, 1992):


Page 2

• clinical experience is the way of building and maintaining one's knowledge

• basic mechanisms of disease is a sufficient guide for clinical practice • traditional medical training and common sense is sufficient to allow one

to evaluate new tests and treatment. • clinical experts are a sufficient to generate valid guidelines for clinical

practice The traditional assumptions are now being questioned with the new paradigm. The assumptions of the new paradigm are (User’s Guide, 1992):

• clinical experience are crucial but in the absence of systematic observation one must be cautious in the interpretation of information derived from clinical experience for it may at times be misleading.

• understanding of basic mechanisms of disease are necessary but insufficient guides for clinical practice

• understanding certain rules of evidence is necessary to correctly interpret literature on causation, prognosis, diagnostic tests, and treatment strategy

The new paradigm makes learning new things more self‐directed and less reliant on teachers. Students can gain the skills to make independent assessments of evidence, and thus evaluate the credibility of opinions being offered by experts. The purpose of this course is to introduce the concept of evidence based medicine and the use of these concepts to improve the quality of his/her own practice.

WHY TEACH EBM Medical education faces a problem in a present setting: too much information, too little time, too many students in crowded rooms, and exams that discouraged real life‐long learning (Rangachari, 2007). There is a need to make students asks questions about useful information and try to seek the answer for themselves. The term "evidence based medicine" was coined at McMaster Medical School in Canada in the 1980's to label an “active clinical learning” strategy, which people at the school had been developing for over a decade (NLM, 2008). Randomized controlled trials have shown that evidence‐based medicine learning is more effective than didactic learning among medical interns in family medicine. They provide better care in terms of providing treatment to patients with hypertension (Espallardo, 2006).


Page 3

MEDICAL STUDENTS LEARNING EBM Currently most physicians report a moderate amount of exposure to EBM. But physicians in clinical research careers were more favorable towards EBM than those in the clinical practice careers (Luebbe, 2007). But among those who were already exposed to EBM there was mismatch perceived competence and their actual performance. This suggests that better education in EBM is needed (Caspi, 2006). Starting the training in the undergraduate program (medical students) is the logical approach. Early experience helps medical students learn, helps them develop appropriate attitudes towards their studies and future practice and orientates medical curriculums towards society's needs (Littlewood, 2005).

OBJECTIVES After going through the readings and workshops of this manual, the participants should be able to:

• define and describe the steps in applying evidence‐based medicine into his/her own clinical practice

• appraise and use randomized controlled trials and other types of studies in solving clinical problems in clinical practice

• make an efficient literature search and identify problems and solutions in the application of evidence based medicine

METHODS This is a series of self‐reading materials and group discussions. Allot a fix time for you to read the reading assignment in the manual. Conduct the group discussion with at least 5 of your colleagues. Assign a facilitator and a co‐facilitator/scribe for each discussion. Observe the rules enumerated in the succeeding section. Monitor your progress by reviewing the checklist provided before each section. The pace of learning depends on your time schedule. However, I suggest that you allot one day a week for the reading time and group discussion. You can also try to apply critical appraisal by yourself with other topics of interest.


Page 4

RULES OF DISCUSSION The ground rules that are encouraged to be observe during the group discussion are as follows:

• Honor the established time limits. • Allow one person to talk at a time. • Focus on the topic. Avoid sideline conversation. • Listen to what others have to say. All ideas are valued. • Encourage participation in the discussion for all participants. • Critique on ideas and thoughts, not on the person.

ROLES AND RESPONSIBILITIES FACILITATOR The facilitator serves as the process facilitator. He/she is also responsible for the content and final outcome of the discussion. His/her responsibilities are to:

• Provide the process to achieve the objectives and desired outcomes. • Pose probing but non‐threatening questions to provoke thought, clarify

discussion and bring insight on some points. • Provide balance. Facilitate rather than lead the discussion. • Remain neutral on content and avoid evaluation and decisions on ideas. • Encourage equal participation among group members.

CO‐FACILITATOR/SCRIBE The co‐facilitator helps the facilitator to achieve his/her objectives. He/she may join the group discussion, but must bear in mind of his/her other functions:

• Be a timekeeper to ensure progress. • Contribute ideas to the topic being discussed. • Meet with the facilitator during the break to discuss the process and

ideas on how to proceed. • Record essential information (content) and observation (group process)

for post‐discussion processing and evaluation. PARTICIPANTS The participants are encouraged to actively participate in the discussion. He/she


Page 5

working agreement set by the group. The amount of learning you will get from this course is proportional to the degree of your participation.

REFERENCES Caspi O, McKnight P, Kruse L, Cunningham V, Figueredo AJ, Sechrest L. Evidence‐based medicine: discrepancy between perceived competence and actual performance among graduating medical students. Med Teach. 2006 Jun;28(4):318‐25. Espallardo NL. Effectiveness of Critical Appraisal Workshop as a Method for Disseminating a Clinical Practice Guideline on Hypertension. Fil Fam Phys 2006; 44 (2): 54‐60. Littlewood S, Ypinazar V, Margolis SA, Scherpbier A, Spencer J, Dornan T. Early practical experience and the social responsiveness of clinical education: systematic review. BMJ. 2005 Aug 13;331(7513):387‐91. Luebbe AM, Radcliffe AM, Callands TA, Green D, Thorn BE. Evidence‐based practice in psychology: perceptions of graduate students in scientist‐practitioner programs. J Clin Psychol. 2007 Jul;63(7):643‐55. National Library of Medicine. Medical Subject Headings. www.ncbi.nlm.nih.gov/sites/entrez (May 27, 2008). Rangachari PK. Back to the future? Active learning of medical physiology in the 1900s. Adv Physiol Educ. 2007 Dec;31(4):283‐7. Users' Guides to Evidence‐based Medicine. JAMA. 1992;268(17):2420‐5.


Page 6

READING ASSIGNMENT SEARCHING THE MEDLINE (SESSION BRIEFING)

CHECKLIST Have you read the previous two topics?

Yes No Did I get your interest now?

Yes No Are you ready to work with your group or colleagues?

Yes No If your answer is yes to all of the questions, congratulations! You have the potential of being a quality health care provider. If your answer is yes to only one of the questions, don’t worry you can always go back and read the previous two topics.

ANOTHER CHECKLIST Have you formed a group?

Yes No Has the group elected a facilitator?

Yes No Has the group elected a co‐facilitator/scribe?

Yes No If not yet, what are you waiting for!

ORGANIZE NOW!


Page 7

OBJECTIVES The purpose of the reading assignment is to introduce to the participants how to formulate a question and search the MEDLINE for the answer. At the end of the reading session, you should be able to search the MEDLINE with skill and confidence that you were able to retrieve the right answer to the problem.

INSTRUCTIONS Read the assignment for an article on MEDLINE search. Focus on the use of key terms and how to use the operant words AND or OR. After reading the paper, proceed to the library or internet and try searching in the MEDLINE.


Page 8

READING ASSIGNMENT SEARCHING THE MEDLINE DATABASE CLINICAL SCENARIO Supposing a patient with cough came in to your clinic and asks for an anti‐biotic drug because he wants to be relieved of his cough right away. Will you prescribe it? Patients usually come in to the clinic for problems. Unfortunately these problems are vague and sometimes not clearly stated. To state the problem clearly, you must bear in mind that there are only three important elements that the patient want to know:

• What their diseases are • What treatment they should be given • What is the expected outcome of the treatment

These three important elements in clinical research are basically:

• the patient (P) • the intervention/exposure (I) • the outcome (O)

Sometimes the researcher can add

• method (M) Going back to our scenario, my clearly stated problem will be: “Among patients with cough (P) will anti‐biotic (I) provide symptom relief faster?”

TRADITIONAL METHOD OF LITERATURE SEARCH A recent survey of important knowledge sources that influence clinical practice was conducted among faculty members, fellows and residents of a large teaching tertiary care hospital. The results showed that the most important resources were English journals, text books and experience (Yousefi‐Nooraie, 2007). This dominance of the traditional information resources and experience‐


Page 9

based medicine may be one of the barriers to the dissemination of evidence‐based medicine. Thus educational programs to develop skills of efficiently searching the research literature need to be developed. Brief (two‐hour) instructional intervention on EBM‐based techniques for searching Medline for evidence related to a clinical problem provided to the students have been shown to be effective. With this training, students had fewer search errors and correspondingly higher quality searches. The most common search errors were a lack of Medical Subject Headings (MeSH) explosion, missing MeSH terms, lack of appropriate limits, failure to search for best evidence, and inappropriate combination of all search concepts (Gruppen, 2005).

PUBMED, ENTREZ AND MEDLINE PubMed was developed by the National Center for Biotechnology Information (NCBI) at the National Library of Medicine (NLM), located at the U.S. National Institutes of Health (NIH). PubMed provides access to bibliographic information that includes MEDLINE, as well as:

• out‐of‐scope citations (e.g., articles on plate tectonics or astrophysics) from certain MEDLINE journals, primarily general science and chemistry journals

• citations that precede the date that a journal was selected for MEDLINE indexing.

• additional life science journals that submit full text to PubMed Central and receive a qualitative review by NLM.

Entrez is the text‐based search and retrieval system used at NCBI for services including PubMed. MEDLINE is the largest component of PubMed and is the freely accessible online database of biomedical journal citations and abstracts created by the U.S. National Library of Medicine (NLM). Approximately 5,200 journals published in the United States and more than 80 other countries have been selected and are currently indexed for MEDLINE. A distinctive feature of MEDLINE is that the records are indexed with NLM's controlled vocabulary, the Medical Subject Headings (MeSH). In the internet, you can access MEDLINE through PUBMED or GRATEFULMED or through other organizations.


Page 10

PARTS OF PUBMED HOMEPAGE Side bar

Page header Entrez databases Query box Features tab

BASIC SEARCH STRATEGY The first step in using PubMed is to first develop a search strategy, a plan that helps you look for the information you need. This can be done by doing the following steps:

• Identify the key concepts (should include the PIOM discussed earlier) • Determine alternative terms for these concepts (can be facilitated with

MESH term search) • Refine your search (use limits like publication dates, study subjects, study

designs, patient age, etc) Our clinical question in the previous scenario was “Among patients with cough (P) will anti‐biotic (I) provide symptom relief faster?” The key concepts in my search terms are:

• Cough • Antibiotics • Relief of symptoms

When I type the key term “cough” in the search box the yield was 27,751 articles and it will be impossible for me to browse these articles. When I typed “cough AND antibiotics" the yield was 2,175 and when I typed “cough AND antibiotics AND relief” the yield was 15 articles. Now I can browse through these articles. How did this happen?


Page 11

THE BOOLEAN PRINCIPLE Searching MEDLINETHE BOOLEAN LOGIC

12 3

4 56 7

89 10 11

121314

15 16

A B

C

A OR BA AND BA AND B AND C

(A AND B) OR C A AND (B OR C) A AND (B AND C)

In the Boolean principle, elements labeled such as 1, 2, 3, etc., can belong to set A, B, C, etc. Some of these elements can belong to two or more sets, and some of these sets may contain no elements. In the above example, if you want to combine elements in two sets you use the word OR, i.e. the elements in set A OR set B are 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 and 14. If you want the elements common to the two sets you use the word AND, i.e. the elements that belong to both set A AND set B are 7, 8 and 9. In the MEDLINE, articles are indexed together as set of articles based on their key words. Just like the Boolean principle the main operators in the MEDLINE are also the words AND and OR. As a beginner this may be enough for you.

Searching MEDLINETHE BOOLEAN LOGIC

12 3

4 56 7

89 10 11

121314

15 16

TB RCTs

MENINGITIS

TB OR RCTsTB AND RCTsTB AND RCT AND MENINGITIS


Page 12

In the example above, if I want articles about tuberculosis (TB), I will type “tuberculosis” in the search window and the result will give me 9 articles. But if my concern is only to read randomized controlled trials (RCT) on tuberculosis I will type in the search window “tuberculosis AND randomized controlled trial” and it will give me 3 articles. Reading 3 articles instead of 9 will save me a lot of time. What will you type in the search box if you want randomized controlled trials on tuberculous meningitis? What articles will you get if you typed “tuberculosis AND randomized controlled trial AND meningitis”?

THE ADVANCED SEARCH STRATEGY The advance search page can be accessed by clicking the link Advanced Search (beta) on the right side of the query box. I used the advance search option and got the results shown below. When I type cough AND antibiotic, the yield was 2, 175 articles and when I type cough AND antibiotic relief, the yield was 15 articles. If I have no time to browse through 15 articles, I can limit this further by checking other boxes for relevance such as date of publication, type of studies, age of subjects etc. This can also be done in the basic search but it will take several steps.


Page 13


Page 14

REFERENCES Gruppen LD, Rana GK, Arndt TS. A controlled comparison study of the efficacy of training medical students in evidence‐based medicine literature searching skills. Acad Med. 2005 Oct;80(10):940‐4. National Library of Medicine. PubMed OVerview. www.ncbi.nlm.nih.gov/sites/entrez (May 27, 2008). Yousefi‐Nooraie R, Shakiba B, Mortaz‐Hedjri S, Soroush AR. Sources of knowledge in clinical practice in postgraduate medical students and faculty members: a conceptual map. J Eval Clin Pract. 2007 Aug;13(4):564‐8.


Page 15

READING ASSIGNMENT CRITICAL APPRAISAL OF AN ARTICLE ABOUT DIFFERENTIAL DIAGNOSIS (SESSION BRIEFING)

CHECKLIST Have you read the topic on MEDLINE search?

Yes No Have you gone to library?

Yes No Are you ready to work with your group or colleagues?

Yes No If your answer is yes to all of the questions, congratulations! You are all set for an exciting learning experience!

OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about differential diagnosis. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the assignment for an article about a differential diagnosis. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper. After reading the paper you can proceed to conduct the group workshop.


Page 16

READING ASSIGNMENT CLINICAL DECISION ON DIFFERENTIAL DIAGNOSIS CLINICAL SCENARIO Suppose a 35‐year old female patient came in to your clinic for fever and abdominal pain for a week. There was neither diarrhea nor dysuria. You read in the papers that there was an increased incidence of dengue fever in children. How will you optimize (request only for what is essential) the diagnostic laboratory tests for this patient? Naturally you can do that if you already have an initial diagnosis in mind. Unfortunately there might be many of them. This session will help you trim down the differential diagnosis and request only for the laboratory tests that are essential. Differential diagnosis is the method of limiting the possible causes of the patient’s symptoms before making a final diagnosis. Identifying the right differentials will make patient management more focused and efficient. Differential diagnosis can be arrived at by using the anatomic approach i.e. considering the possibilities based on organs that may be affected within the proximity of the symptom like chest pain may have differential diagnosis like herpes zoster (skin), costochondritis (ribs), pneumonia (lings) or angina (heart). If the symptom is systemic like fever, the differentials can by be pathophysiology i.e. vascular, inflammatory/infectious, neoplastic/neurologic, degenerative, intoxication/idiopathic, congenital, allergic/autoimmune, trauma and endocrine (VINDICATE) (Friedland, 1998). With these approaches however the frequency or probability of each differential will not be known.. If we consider all known causes equally possible (the ‘possibilistic’ approach), then the patient will have unnecessary diagnostic tests performed on them. Instead, we must considering first those that are more common (a ‘probabilistic’ approach), or more more serious if left undiagnosed and untreated (a ‘prognostic’ approach) or more responsive to treatment (Richardson, 1999). And they are important because the probabilities of the individual differential will help us focus our diagnostic strategies as shown in the table below. The disease probabilities can be taken from population prevalence statistics or from original research. Research studies focus more directly on the frequency of diseases that cause symptoms (Kroenke, 1997) are preferred over population survey because they are more associated with presenting symptoms.


Page 17

Probabilities of Differential Diagnosis and Recommended Diagnostic Strategies Differential Diagnosis Diagnostic Tests Treatment Working diagnosis – most possible cause that should be ruled in

Choose test(s) with high specificity and LR+ much

larger than one.

Start empiric treatment

Alternative diagnosis – other possibilities that should be ruled‐out

Choose test(s) with high sensitivity and LR- much

smaller than one.

Start supportive

treatment

Remote diagnosis

None

None

SEARCH You found an unpublished retrospective study in the archives of your department written by a previous resident Santos AR, entitled “Differential diagnosis of typhoid fever in the emergency room”.

CRITICAL APPRAISAL RELEVANCE QUESTION

• Is the objective of the article on differential diagnosis similar to your clinical dilemma?

To answer this question, look at the objective of the study. It is important that your article is relevant to the question you have raised in order for you to make maximum use of the results of the study and be able to apply it to decision making that influences patient care. For differential diagnosis it is important that the focus is to find the cause of symptoms, clinical and laboratory presentation among patients similar to your patient or case scenario.


Page 18

VALIDITY GUIDES

• Did the study patients represent the full spectrum of patients who present with this clinical problem?

Study designs that answer clinical questions like differential diagnosis can be a cross‐sectional study or cohort study. An important element with these designs is how the subjects are recruited so they can represent other patients who may also have the same symptoms i.e. representativeness. The definition of the clinical problem under study describes the population to which the study will be applied. The problem usually is a symptom or an abnormal physical examination such as headache or abdominal mass or a combination of symptoms and abnormal physical findings like headache and facial asymmetry. This is usually defined in the inclusion and exclusion criteria of the study. With the symptom already defined, the other strategies that can assure representativeness are any of the following:

• Random selection – not always possible in clinical setting • Consecutive patient recruitment – most feasible • Recruitment in defined setting – must always be done

The Article by Santos included patients consulting for fever in the emergency room. Although the inclusion criteria were fever alone as the chief complaint, there was a subgroup analysis of patients with fever and abdominal complaint. The total number of patients included in the study was 235. This coincides with your case scenario.

• Were the criteria for each final diagnosis explicit and credible? Determination of final diagnosis must be clearly described, may not necessarily be based on the ultimate reference standard. However the criteria must be explicit enough to make sure that different clinician will arrive at the similar diagnosis (inter‐rater reproducibility). The final diagnosis in Santos’s paper was based on clinical syndromes and criteria. Blood cultures, ultrasound and other tests were not done to establish the final diagnosis in only 48% of the cases.


Page 19

• Was the diagnostic work‐up comprehensive and consistently applied? The set of diagnostic work‐up should be thorough to come up with an accurate diagnosis. Then a minimum set of diagnostic work‐up that includes a thorough history and physical examination and a few initial laboratory tests should have been applied consistently for all patients. This can be answered when the study described a prospective approach in identifying patients in the study. Retrospective approach is usually limited because, records cannot guarantee a standard diagnostic approach for everybody. The diagnostic tests done for 85% of patients in the Santos study were CBC, urinalysis and stool examination. Temperature was measured using a mercury type thermometer and records with “febrile” as reported documentation of fever were excluded.

• For initially undiagnosed patients, was follow‐up to come up with a diagnosis sufficiently long and complete?

Sometimes the diagnosis at the early stage of the disease is really difficult and the patient may be classified as not having the disease or undetermined. To assure ourselves with the eventual diagnosis of undetermined cases, we may have to observe them over time. The Santos study observed the patients for 24 to 48 hours in the emergency room. In most patients antibiotics were started and the patients were sent home without fever. OVERALL, IS THE STUDY VALID? Although the study was a retrospective study, you decided that you can use this article because it is the only available study in your setting. WHAT ARE THE RESULTS? What were the diagnoses and their probabilities? How precise are the estimate of the probabilities? The probabilities of the differential diagnosis are reported as either incidence or prevalence with their 95% confidence interval. In the Santos study, the following top three diseases were the most common diagnosis given to adult patients with fever: a) typhoid fever, 34%; b) urinary tract infection, 32%; and c) acute gastroenteritis, 29%. No confidence intervals were reported.


Page 20

CAN THE RESULTS HELP ME IN CARING FOR MY PATIENTS?

• Are the study patients similar to my own? For a study on differential diagnosis be applied to your patient you have to be assured that the characteristics of your patient is similar to the study’s inclusion criteria. In the Santos study, they included patients consulting in the emergency room but were eventually sent home. The cases seen in this study seemed to be the milder cases similar to your patient.

• Do you think the disease probabilities in the study still apply today? Disease prevalence and incidence change across time. Old disease can be controlled because of effective treatment. Thus a paper on differential diagnosis may still include smallpox for patients with fever and skin lesions in the 1950’s, the probability is almost zero today. The probability of Dengue fever may differ in different times of the year. A little knowledge on epidemiology of disease across time may be necessary to have an accurate answer to this question. However if the disease in question does not vary over time then this is not a problem. The study of Santos was a three‐year retrospective study from January 1988 to December 1991. Seasonal variation may have been accounted for but the study is already 9 years old. Unfortunately you cannot find a more recent one. RESOLUTION OF THE PROBLEM IN THE SCENARIO After appraising the study of Santos you decided that your diagnostic tests will focus on ruling in or ruling out typhoid fever, urinary tract infection and gastroenteritis.

REFERENCES Friedland ed. Evidence‐based Medicine: A framework for clinical practice. Appleton and Lange, 1998. Kroenke K. Symptoms and science: the frontiers of primary care research [Editorial]. J Gen Intern Med 1997; 12: 509 ‐ 510. Richardson WS, Wilson MC, Guyatt GH, Cook DJ, Nishikawa J, and the Evidence Based Medicine Working Group. JAMA, 1999 Apr 7; 281(13):1214‐9.


Page 21

WORKSHOP CRITICAL APPRAISAL OF AN ARTICLE ABOUT DIFFERENTIAL DIAGNOSIS (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making about differential diagnosis. Another objective is to introduce concepts of critical appraisal of an article regarding differential diagnosis focusing on the following:

• validity • interpretation of the results • applicability of the results

INSTRUCTIONS Divide the participants into groups of six to ten persons per group. Assign a case scenario to each group and formulate an answerable problem from the scenario. Establish initial group consensus on how to proceed with the scenario. Ask the group to read the article retrieved to answer the problem in the scenario. Focus on the abstract, methods and results section. Again establish a group consensus on how to proceed with the scenario. Note any change in decisions. Critically appraise the article using the appraisal sheet provided. Answer validity questions, analyze the results and determine the applicability of the results. Establish another group consensus and note any change in decision. Process the exercise. Focus on barriers and solution to the application of the exercise in usual clinic practice.


Page 22

TIME ALLOTTED The time allotted for this workshop is two hours. The recommended break‐up is:

• 15 minutes to analyze the scenario and develop consensus • 15 minutes to read the article • 45 minutes to appraise the validity • 15 minutes to analyze results • 20 minutes to establish applicability • 10 minutes to summarize the process

DESIRED OUTCOME The participants should make a clinical decision on the scenario based on the critical appraisal of the evidence.


Page 23

Appraisal Sheet CLINICAL DECISION ON DIFFERENTIAL DIAGNOSIS CLINICAL SCENARIO OR QUESTION

SEARCH

CRITICAL APPRAISAL RELEVANCE

Is the objective of the article on differential diagnosis similar to your clinical dilemma?

PRIMARY VALIDITY GUIDES

Did the study patients represent the full spectrum of patients who present with this clinical problem? Definition of the clinical problem or the patient whom the study will be applied. Were the criteria for each final diagnosis explicit and credible? Determination of final diagnosis must be clearly described, may not necessarily be the ultimate reference standard.

SECONDARY VALIDITY GUIDES

Was the diagnostic work‐up comprehensive and consistently applied?


Page 24

For initially undiagnosed patients, was follow‐up to come up with a diagnosis sufficiently long and complete?

OVERALL, IS THE STUDY VALID?

WHAT ARE THE RESULTS?

What were the diagnoses and their probabilities? How precise are the estimates of probabilities?


Are the study patients similar to my own? Inclusion criteria, exclusion criteria, clinical definition Do you think the disease probabilities in the study still apply today? Is the study recent? Could the probabilities change since the study publication?

RESOLUTION OF THE PROBLEM IN THE SCENARIO


Page 25

READING ASSIGNMENT CRITICAL APPRAISAL OF AN ARTICLE ABOUT A DIAGNOSTIC TEST (SESSION BRIEFING)

CHECKLIST Have you read the previous topic on differential diagnosis?

Yes No Have you undergone the workshop on differential diagnosis with your group?

Yes No Did you enjoy the workshop?

Yes No If your answer is no to the last question please state the reasons below and share it to the group before starting the next workshop.

OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about a diagnostic test. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the reading assignment for an article about a diagnostic test. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper. After reading the paper you can proceed to conduct the group workshop.


Page 26

READING ASSIGNMENT CLINICAL DECISION ON A DIAGNOSTIC TEST

CLINICAL SCENARIO A 70 years old female patient came in to the clinic complaining of forgetfulness. She’s afraid that she has dementia just like her sister. She does not want to be subjected to MRI or CT scan. You referred her to the psychiatric resident and she suggested that you perform the Mini‐mental State Examination (MMSE), but you doubt her decision. The next weekend you went to the library and try to learn more about the MMSE.

SEARCH After searching in the MEDLINE you found the article by Mulligan et al entitled “ A comparison of alternative methods of screening for dementia in clinical settings” published in the Archive of Neurology, June 1996. Luckily the full text was also available.


• Was the objective of the paper relevant to your clinical question? Most of the time we read journal articles because the topic is interesting. Because of this application to clinical practice is not ensured. We can only ensure that the results of the article are applied to practice if the objectives of the article are relevant to the clinical problems we see in clinical practice. Thus the objective of the study must determine the accuracy (outcome) of the contemplated diagnostic test (intervention/exposure) among patients (population) similar to your case scenario.


Page 27

VALIDITY GUIDES

• Was there an independent comparison with a reference standard? There are two elements in this guide question i.e. use of a reference standard and independent comparison. A reference standard for a diagnostic test is the test that gives the information nearest to the “truth”. Thus the accuracy of the test should be compared against the standard. If the diagnostic test approximated the standard, that means the test also approximates the “truth”. An independent comparison means that the reader of the reference standard did not know the result of the diagnostic test being evaluated (Jaeschke, 1994). Awareness of the initial test result may lead to increase confirmation with reference standard leading to bias on the accuracy of the diagnostic test being evaluated. Thus the first question you should answer is whether there was a comparison with the reference standard and whether the reference standard used was acceptable to your setting. The second is whether the reader of the reference standard was blinded to result of the diagnostic test being evaluated. In the study by Mulligan et al the reference standard used was the diagnosis of dementia based on the DSM‐III‐R.

• Did the patient sample include an appropriate spectrum of patients to whom the test will be used?

The accuracy of a diagnostic test among patients with low risk for the disease is different from patients with high risk of the disease. The clinical utility of a test can be seen when used among persons who are healthy, patients who are very sick and mostly those in‐between because these are the patients who will be requiring the test. Patients consulting in family practice usually belong to the healthy and in‐between groups while patients consulting in the hospitals are those in the in‐between and more severe groups. The in‐between groups may give an underestimate of the accuracy of the test (but it is the accuracy value to whom the test will be used) while the healthier and more severe may give an overestimate of the accuracy. If all groups are equally represented the average accuracy will be obtained. The study of Mulligan et al included elderly patients consulting in a geriatric hospital and memory clinic. The elderly age group is the population with the highest risk of dementia thus the results from this study may be an overestimate.


Page 28

• Was the reference standard done regardless of the result of the diagnostic test being evaluated?

In some studies, the accuracy of a diagnostic test is examined retrospectively (chart review of actual practice). In actual practice however, physicians request to perform the reference standard based on the initial result of the diagnostic test. The reference standard is used to verify the initial finding i.e. when positive. When this happens most of the data available will be those positive for the diagnostic test and will likely be positive in the reference standard. This will increase the accuracy of the test. This is called verification bias (Jaeschke, 1994). To avoid this, the study must show that the reference standard was done regardless of the result of the diagnostic test being evaluated. It was mentioned in the Mulligan et al study that the comparison with the DSM‐III‐R was blind and independent.

• Were the methods for performing the test described in sufficient detail to permit replication?

This is necessary so that the reader will be able to duplicate the test in his/her own setting and get the same valid result. Description should include preparation for the patient such as diet, drugs to avoid, precautions, ideal conditions for performing the diagnostic test and a step by step description of how the diagnostic test is done and interpreted. There are a lot of papers dealing with instructions on how to administer the MMSE and its interpretation. OVERALL, IS THE STUDY VALID? Yes. You accepted the validity since most of the questions were answered adequately. WHAT ARE THE RESULTS?

• What were the likelihood ratios for the different possible test results? Likelihood ratio indicates by how much a given test result increases the pre‐test probability of the disease. A likelihood ratio of 1 means that the post‐test probability is similar to the pre‐test probability. A likelihood ratio of greater than 1 increase the chance that the disease is present, and the greater the likelihood ratio the greater is the increase in chance.


Page 29

Some papers give the sensitivity and specificity values rather than the likelihood ratio. The formula for computing the likelihood ratio from sensitivity and specificity is shown below: Likelihood ratio of a positive test LR (+) = Sn/1‐Sp Likelihood ratio of a negative test LR (‐) = 1‐Sn/Sp A rough guide in evaluating LR values:

• LRs >10 or < 0.1 generate large, and often conclusive changes from pre‐ to post‐test probability;

• LRs of 5‐10 and 0.1‐0.2 generate moderate shifts in pre‐ to post‐test probability;

• LRs of 2‐5 and 0.5‐0.2 generate small (but sometimes important) changes in probability; and

• LRs of 1‐2 and 0.5‐1 alter probability to a small (and rarely important) degree.

In the Mulligan study the LR (+) MMSE is 2.46 and the LR (‐) is 0.14. The alternative clinical test is the Antisaccadic Eye Movement Test (AEMT) but did not do very well compared with the MMSE. CAN THE RESULTS HELP ME IN CARING FOR MY PATIENTS?

• Will the reproducibility of the test result and its interpretation be satisfactory in my setting?

Even if the test was described very well, the reproducibility of the interpretation is necessary for the test to be adequately applied in your setting. The paper should report measures of agreement between interpreters or raters. If agreement is not reported, decide for yourself. Is the interpretation simple enough? Is the basis of the interpretation clear and specific? The MMSE questionnaire also contains instruction on how to administer the test and interpret the result.

• Are the results applicable to my patient? If your setting is somewhat similar to that in the study, and the inclusion criteria include characteristics of your patient, then you can apply the results to your patient. Sometimes your clinical judgment is required.


Page 30

• Will the results change my management? The usefulness of a diagnostic test result is whether it will help the clinician manage his/her patient. If the diagnostic test will lead the doctor decide to give treatment or not, then the test is helpful and may be requested. However if after applying the LR value to determine the post‐test probability and this did not help in the decision, then you don’t have to perform the test. RESOLUTION OF THE PROBLEM IN THE SCENARIO After looking at the Mulligan et al study, you agreed with the psychiatry resident and even asked her help to administer the MMSE to the patient yourself.

REFERENCES Jaeschke R, Guyatt G, Sackett D, and the Evidence Based Medicine Working Group. How to Use an Article About a Diagnostic Test: Validity Guides. JAMA, 1994; 271(5):389-391. Jaeschke R, Guyatt G, Sackett D, and the Evidence Based Medicine Working Group. How to Use an Article About a Diagnostic Test: Results and Applicability. JAMA, 1994; 271(9):703-707.


Page 31

WORKSHOP CRITICAL APPRAISAL OF AN ARTICLE ABOUT A DIAGNOSISTIC TEST (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making about the usefulness of a diagnostic test. Another objective is to introduce concepts of critical appraisal of an article regarding a diagnostic test focusing on the following:

• Validity • Reference standard • Sensitivity and specificity • Likelihood ratios • Applicability of the results

INSTRUCTIONS Divide the participants into groups of six to ten persons per group. Assign a case scenario to each group and formulate an answerable problem from the scenario. Establish initial group consensus on how to proceed with the scenario. Ask the group to read the article retrieved to answer the problem in the scenario. Advice them to focus on the abstract, methods and results section. Again, establish a group consensus on how to proceed with the scenario. Note any change in decisions. Critically appraise the article using the appraisal sheet provided. Answer validity questions, analyze the results and determine the applicability of the results. Establish another group consensus and note any change in decision. Process the exercise. Focus on barriers and solution to the application of the exercise in usual clinic practice.


Page 32





Page 33

Appraisal Sheet CLINICAL DECISION ON A DIAGNOSTIC TEST CLINICAL SCENARIO OR QUESTION

SEARCH


Is the objective of the study relevant to your clinical question?

VALIDITY GUIDES

Was there an independent and blind comparison with a reference standard? What was the reference standard. Were they assessed independently? Did the patient sample include an appropriate spectrum of patients to whom the test will be used?

Was the reference standard done regardless of the result of the diagnostic test being evaluated?


Page 34

Were the methods for performing the test described in sufficient detail to permit replication?



What were the likelihood ratios for the different possible test results?


Will the reproducibility of the test result and its interpretation be satisfactory in my setting? Are the results applicable to my patient? Will the results change my management?



Page 35

READING ASSIGNMENT CRITICAL APPRAISAL OF AN ARTICLE ABOUT THERAPY OR PREVENTION (SESSION BRIEFING)

CHECKLIST Have you read the previous topic on diagnostic tests?

Yes No Have you undergone the workshop on diagnosis with your group?



OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about therapy or prevention. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the reading assignment for an article about therapy or prevention. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper. After reading the paper you can proceed to conduct the group workshop.


Page 36

READING ASSIGNMENT CLINICAL DECISION ON THERAPY OR PREVENTION

CLINICAL SCENARIO You are working at the out‐patient clinic in your institution when a 55 year‐old, male bank executive came to your clinic for constricting chest pain located at mid‐sternum precipitated by exertion and relieved by rest. His blood pressure was 130/80, heart rate was 85/minute and the respiratory rate was 20/minute. An ECG was done revealing lateral wall ischemia. You sent the patient home on oral nitrates and aspirin. However, further work‐ups revealed hypercholesterolemia with a level of 6.5 mol/liter. You gave dietary advice but the patient claimed that he has been on a high fiber diet with low fat intake for the past 6 months. He is now inquiring if he should take a drug for his elevated cholesterol.

SEARCH You decided to translate this clinical dilemma to an answerable question. You also decided that the article should include a population of patients that has an elevated cholesterol who have already undergone dietary therapy for at least 6 months to be consistent with your patient’s profile. The drug intervention should be compared with placebo. The article must report clinically important outcomes, such as reduction in cardiovascular deaths. Finally, you wanted an article that employed randomization. The article you finally retrieved included 4444 patients randomized to either simvastatin or placebo. However, before applying the results to your patient, you decided to appraise the article using the following guides.

RANDOMIZED CONTROLLED TRIALS Randomized controlled trials are the standards design to prove effectiveness of drugs or other forms of intervention. When done properly, it can provide the best evidence of effectiveness. In this type of design, individuals are randomly assigned (randomization) to either of the two or more groups, one with the


Page 37

intervention the other without the intervention being tested or another intervention. Randomization tries to make the two groups similar for both known and unknown factors that may affect the outcome other than the intervention being tested. Then they are observed forward in time and their outcome compared. The outcome can be the cure of a disease, relief of symptoms or improvement in quality of life (Espallardo, 2000).


• Is the objective of the article comparing therapeutic interventions similar to your clinical dilemma?

Before going any further, first ascertain if the objective of the study addresses the clinical problem you face. Appraising an irrelevant article would not be helpful to your clinical dilemma and will be a waste of time. Below are a few tips that will help you decide on relevance:

• Population of the study (P )– should be similar to the characteristic of your patient.

• Intervention/comparative intervention/exposure (I) – should include the therapeutic intervention you want to test.

• Outcome of the study (O) – one of the outcomes measured should be the goal you and your patient wish to work for.

VALIDITY GUIDES

• Was the assignment of patients to treatment randomized? In order to answer this question, the reader is advised to look into the article’s abstract or methodology section. For the 4S study, randomization was done and was written both in the abstract and methods section. The strength of randomization is that if the sample size is sufficiently large, it assures that both known and unknown determinants of outcome are evenly distributed between the treatment and control groups. In the absence of randomization, these factors might be difficult to control and might be the one strongly influencing outcome rather than the treatment itself (Guyatt, 1994).


Page 38

At times though because of the rarity of the disease and small patient sample size, randomization might not be feasible. In these cases, a clinician must rely on weaker studies but should be aware of its potentials for errors. Were all patients who entered the trial properly accounted for and attributed at its conclusion?

• Was follow‐up complete? This is best checked by looking at the number of patients enrolled at the outset and comparing this with the number of patients reported in the results table. Every patient who entered the trial should be accounted for at its conclusion. If substantial numbers are “lost to follow‐up”, the validity of the conclusions are open to question. A drop‐out rate of 20% or more is usually declared as substantial. If the number lost to follow‐up is less than this the reader can decide if this affects the conclusion by assuming a “worst case scenario”. This means that the numbers lost in the treatment group are assumed to have bad outcomes and the numbers lost in the control group are assumed to have been cured and if the conclusions differ, a substantial number was lost to follow‐up. Another way of deciding whether follow‐up was complete is to check whether an intention to treat analysis was done. If this is reported one can safely assume that follow‐up was complete.

• Were patients analyzed in the groups to which they were randomized? It simply means that all those belonging to the control group or treatment group are analyzed from beginning to end in this same grouping including those who were dropped or withdrawn or changed treatment. No crossing over treatment modalities were done as this would likely lead to biased results. Excluding non‐compliant patients from the analysis leaves behind those who may be destined to have a better outcome and destroys the unbiased comparison provided by randomization. This principle of attributing all patients to the group to which they were randomized results in an "intention‐to‐treat" analysis. This strategy preserves the value of randomization: prognostic factors that we know about, and those we don't know about, will be, on average, equally distributed in the two groups, and the effect we see will be just that due to the treatment assigned (Guyatt, 1994).


Page 39

• Were patients, their clinicians, and study personnel "blind" to treatment?

To answer this question the reader is again advised to look into the abstract or the methodology section. Blinding is the process by which the intervention being given is concealed from the patient, the clinicians and the one who analyzes the data. Patients, clinicians and data analysts are likely to have an opinion regarding the experimental treatment. These opinions, whether optimistic or pessimistic, can systematically distort reporting of treatment outcomes. As to avoid these “reporter and observer” bias, blinding is necessary.

• Were the groups similar at the start of the trial? To answer this question, one should look for a report of the comparison of the baseline characteristics of the experimental and control group. For most studies, this is labeled as table 1. In the 4S trial, baseline characteristics were similar. For reassurance about the study’s validity, readers would like to be informed that the treatment and control groups were similar for all the factors that determine clinical outcomes of interest save for the experimental therapy. The greater the similarity between known prognostic factors for the control and experimental group, the more likely that the results can be attributed to the intervention, rather than due to the differences in these factors.

• Aside from the experimental intervention, were the groups treated equally?

Interventions other than the treatment under study, when differentially applied to the treatment and control groups, are called “co‐interventions”. This might distort the results since they in themselves might cause changes in reported outcomes. OVERALL, IS THE STUDY VALID? If you want to be strict about it, you should answer yes in all 5 questions. However, you as the user of the journal can make the decision. A simple rule might be to answer yes to at least, one primary guide and two secondary guides. The 4S study yielded yes to all the appraisal questions, hence you decided that over‐all the study was valid. You now proceed to analyze the results.


Page 40


• How large was the treatment effect? Most randomized controlled trials report outcomes either as treatment success or treatment failures and adverse effects. Examples of outcomes include cure rates, side effects or death. Patients either do or do not suffer these events and the article frequently reports the proportion of patients who develop such events. In the 4S study, 11.5% died in the placebo group and 8.2% died in the simvastatin group. By eyeballing these figures, simvastatin seems better in reducing deaths. But how else could this figures be compared? The following simple computations could help: Risk in Control (Rc) = Death in control/N patients in the control Risk in Treatment (Rt) = Death in treatment/N patients in treatment Absolute Risk Reduction (ARR) = Rc – Rt = 0.115 ‐ 0.082 = 0.033 Relative Risk (RR) = Rt/Rc = 0.082/0.115 = 0.71 Relative Risk Reduction (RRR) = 1 – RR = 1‐ 0.71 = 0.29 (29%) Absolute risk reduction is the absolute difference between the proportion who died in the placebo group compared to the simvastatin group. Relative risk is the risk of events in the simvastatin group or new treatment relative to the placebo or control group. The most useful measure to use in explaining the benefit of treatment to patients is the relative risk reduction. In this case, simvastatin treatment reduces deaths 29% more than placebo.

• How precise was the estimate of treatment effect? To decide regarding precision, one should look at the reported 95% confidence interval. The closer these values, the more precise your estimates. If this is not reported check the p‐value, anywhere from </= .10 is acceptable. Ideally, the reported minimum and maximum values of this interval should all be positive or all be negative for the absolute risk reduction, all above below one for the relative risk and relative risk reduction to be considered precise.


Page 41


• Can the results be applied to my patient care? If your patient meets all of the inclusion criteria and none of the exclusion criteria, the applicability of the study’s results to your patient is without question. It is rare however that we get a patient who conforms to all the characteristics of the study subjects. In these cases, we should decide if the reason is compelling enough not to apply the results of our study to our particular patient.

• Were all clinically important outcomes considered? Clinically important outcomes may range form decreasing mortality, morbidity, improving quality of life. These are outcomes that are important to the patients and will lead directly to reducing symptoms or decreasing death. Some studies might report improvement in cholesterol levels, improvement in PFTs but these are what might be labeled as surrogate endpoints. That is, the researchers have substituted these physiologic measures for important outcomes we have just mentioned. For reduction in these laboratory parameters does not always translate into decrease in morbidity and mortality. A dramatic example of the danger of substitute endpoints was found in the evaluation of the usefulness of clofibrate as anti‐cholesterol drug. It shown to decrease serum cholesterol but was shown to increase all‐cause mortality. Similar findings were noted in an anti‐arrhythmia trial hen the investigators had to stop the trials when they discovered that mortality was substantially higher in patients receiving antiarrhythmic treatment than in those receiving placebo.

• Are the likely treatment benefits worth the potential harm and costs? Computation for cost effectiveness and checking for side effects might be done to check if the treatment benefits are worth the potential harm and costs. RESOLUTION OF THE PROBLEM IN THE SCENARIO The article was valid and giving simvastatin reduces all deaths by 29% compared to placebo. However when we look into costs, giving this drug treatment is fairly expensive. However due to the marked cardiac risks for this patient you decide to give simvastatin since your patient could also afford drug treatment. At this point, we hope that you are encouraged to adapt this new paradigm in your medical decision making. The steps are relatively simple. First, define the problem clearly. Second, use one of several search strategies to come up with


Page 42

relevant articles. Third, appraise the article. Assess the results and the applicability of your article to your patient. Guided by this, you then decide on the action to take. This may sound like a tedious process but as they say practice makes perfect. And if each of us will continue asking, searching, we will continue learning and hopefully improving our patient care.

REFERENCES Espallardo, NL. Research Protocol Development for Resident Physicians. Family Medicine Research Group, Inc. Manila, 2000. Guyatt GH, Sackett D, Cook DJ, for the Evidence Based Medicine Working Group. How to Use an Article About Therapy or Prevention: Validity. JAMA, 1993;270(21):2598‐2601. Guyatt GH, Sackett D, Cook DJ, for the Evidence Based Medicine Working Group. How to Use an Article About Therapy or Prevention: Results and Apllicability. JAMA, 1994;271(1):59‐63.


Page 43

WORKSHOP CRITICAL APPRAISAL OF AN ARTICLE ABOUT THERAPY OR PREVENTION (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making about treatment. Another objective is to introduce concepts of critical appraisal of an article regarding treatment focusing on the following:

• validity • randomization • intention‐to‐treat • interpretation of the results • applicability of the results • clinically relevant endpoints • cost‐effectiveness

INSTRUCTIONS Divide the participants into groups of six to ten persons per group. Assign a case scenario to each group and formulate an answerable problem from the scenario. Establish initial group consensus on how to proceed with the scenario. Ask the group to read the article retrieved to answer the problem in the scenario. Focus on the abstract, methods and results section. Again, establish a group consensus on how to proceed with the scenario. Note any change in decisions. Critically appraise the article using the appraisal sheet provided. Answer validity questions, analyze the results and determine the applicability of the results. Establish another group consensus and note any change in decision. Process the exercise. Focus on barriers and solution to the application of the exercise in usual clinic practice.


Page 44





Page 45

Appraisal Sheet CLINICAL DECISION ON THERAPY OR PREVENTION CLINICAL SCENARIO OR QUESTION

SEARCH


Is the objective of the article comparing therapeutic interventions similar to your clinical dilemma?

VALIDITY GUIDES

Was the assignment of patients to treatment randomized? Randomization vs. random selection Were all patients who entered the trial properly accounted for and attributed at its conclusion? Was follow‐up complete? Dropouts, withdrawals Were patients analyzed in the groups to which they were randomized? Intention‐to‐treat analysis


Page 46

Were patients, their clinicians, and study personnel "blind" to treatment?



How large was the treatment effect? Risk in control, risk in treatment, relative risk, relative risk reduction, absolute risk reduction How precise was the estimate of treatment effect? 95% confidence interval, p value


Can the results be applied to my patient care? Inclusion criteria, exclusion criteria Were all clinically important outcomes considered? Outcome, results


Page 47

Are the likely treatment benefits worth the potential harm and costs? Side effects, NNT, costs



Page 48

READING ASSIGNMENT CRITICAL APPRAISAL OF AN ARTICLE ABOUT HARM (SESSION BRIEFING)

CHECKLIST Have you read the previous topic on treatment?

Yes No Have you undergone the workshop on treatment with your group?



OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about harm. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the reading assignment for an article about harm. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper. After reading the paper you can proceed to conduct the group workshop.


Page 49

READING ASSIGNMENT CLINICAL DECISION ON HARM

CLINICAL SCENARIO You are the resident physician covering for the consultant in his clinic. You met one of his patient, 45 year old male who is on anti‐cholesterol medication. You noted that his cholesterol is now only 4.35 mmol/L but the patient was still prescribed with anti‐cholesterol drug from his last week visit. You approach the consultant and mentioned it to him. He told you its okay; it’s the high cholesterol that we should be concerned with anyway. Still doubtful, you went to the library and look for the harmful effect of very low cholesterol.

SEARCH You found the article by Zureik et al, entitled “Serum cholesterol concentration and death from suicide in men: Paris prospective study 1” published in the British Medical Journal, September 1996.

COHORT AND CASE-CONTROL STUDIES Harmful effects of drugs or other exposure can best be studied with cohort or case‐control study designs. A cohort is any group of individuals who share the same characteristics. In a cohort study, selection of subjects starts with identifying individuals who have the same characteristics or presence or absence of a particular cause or exposure. They are then divided into two groups, those with the characteristics or causes and those without the characteristics. They are then observed forward in time and determine who among them develop the outcome or effect (Espallardo, 2000). Cohort studies can also be prospective or retrospective depending on the manner of patient recruitment. If recruitment is being done forward in time it is prospective, but if the cohort already existed in the past and data gathering is being done by reviewing existing clinical records then it is retrospective (Espallardo, 2000).


Page 50

In a case‐control study, inclusion of subjects starts with defining or selecting those who have the outcome or effect. These are considered as cases. Then this group is compared with subjects who don’t have the outcome or effect. These are considered as the controls. Both the cases and the control should be taken from within the same population. Then the two groups are investigated as to the presence or absence of hypothesized causes or risk factors for the outcome (Espallardo, 2000). Case‐control study can be prospective or retrospective depending on the manner of patient recruitment. If recruitment is being done as cases develop forward in time it is prospective, but if the cases have already developed in the past and patient recruitment is being done by reviewing existing clinical records then it is retrospective (Espallardo, 2000).


• Is the objective of the article on harm similar to your clinical dilemma? Your formulated clinical question must be addressed by the objective of the study. For a decision to stop the harmful exposure the article must be designed to determine the harmful effect (outcome) of the a drug, chemical or environmental substances (intervention/exposure) to the patient (population). This must be clearly stated by the objective of the study. VALIDITY GUIDES

• Were there clearly identified comparison groups? Just like studies of effectiveness of treatment, studies of harm require that a control group for comparison should be done. Unfortunately, randomized controlled trials to prove harm is not ethical, so studies of weaker design like cohort or case‐control studies may be relied upon. Both designs has a control group for comparison, in the former controls are chosen by absence of exposure and in the latter by absence of the outcome or disease. In a cohort study, a group of patients are observed. They are divided into those with the exposure and those without the exposure. Then the outcome is observed forward in time. In a case‐control study, patients with the outcome are gathered. Then a group of patient without the outcome matched to the preceding group for certain characteristics other than the exposure is also


Page 51

gathered. The presence of the exposure in both groups is then ascertained (Levine, 1994). In a cross‐sectional study, patients are grouped and analyzed with respect to their outcome and exposure. This study design can also be a basis for establishing harmful effect, but the temporal relationship of the exposure occurring before the outcome cannot be established. The study by Zureik observed 6,728 men who had measurements of serum cholesterol. They grouped and categorized cholesterol levels into low (<4.78 mmol/L), normal (4.78 to 6.21 mmol/L) and high (>6.21 mmol/L). The changes were also categorized. These were the comparison groups.

• Were the exposures and outcomes measured in the same way in the groups compared?

Measurement of outcomes must be similar in both groups. In cohort study, the investigators must show that diligent observation for the outcome was done in the groups with the exposure (high risk) as well as the groups without the exposure (low risk). When the patient with the exposure were observed more diligently, there will be a higher detection rate of the outcome leading to increase incidence of the disease in the exposed group. This is called surveillance bias (Levine, 1994). This must be avoided. This bias may also occur in case‐control studies, when the detection of exposure is more diligent in the group with the disease or outcome. Suicide data were taken from the national databases and death certificates in all groups in the Zureik study.

• Was follow‐up sufficiently long and complete? The length of follow‐up must be sufficiently long enough to detect the outcome. If follow‐up is short, the chance of underestimating the effect of the exposure is high. When the relation between asbestos and lung cancer was being investigated the relative risk was only 1.4 in the early years of observation compared to the subsequent relative risk of 18.2 when the years of observation was extended to 15 years and beyond. The Zureik study observed their patients for 4 years after enrolment. They had 95% follow‐up.


Page 52

• Is the temporal relationship between the exposure and outcome correct and dose response gradient present?

For an exposure to cause an effect, two important criteria may be considered. First the exposure must be present before the outcome and second there must be a dose response gradient, i.e. the higher the dose the higher is the probability of the outcome. Cross‐sectional studies usually cannot establish temporal relationship but it can establish a dose response gradient and a comparison between groups. The Zureik study measured serum cholesterol before the event of suicide so the temporal relationship was correct. OVERALL, IS THE STUDY VALID? Since all the answers to the validity guides, the study can be considered valid. WHAT ARE THE RESULTS?

• What is the magnitude of the association between exposure and outcome? Was the estimate of the risk precise?

The relative risk is the incidence of the adverse effect in the group with the exposure divided by the incidence of adverse effect in the group without the exposure. If the relative risk is more than 1, then the exposure is causing harm and if less than 1 the exposure reduces harm. Relative risk is usually computed when the design is a cohort study. In a case control study, the odds ratio is computed. The odds ratio approximates the relative risk, especially when the disease is rare. The values of 95% confidence interval should be greater than 1 to say that the exposure really causes harm. If one value of the 95% confidence interval is less than 1 and the other is more than 1, then the effect of the exposure is uncertain. The Zureik study showed that those with low cholesterol had increased risk of suicide with a relative risk of 3.16 with a 95% CI of 1.38 to 7.22. The analysis was also adjusted for age, smoking and mean corpuscular volume.


Page 53


• Are the study patients similar to my own? Just like in an article about a beneficial intervention, for the harmful effect to be extrapolated to your patient you have to be assured that the characteristics of your patient is similar to the study’s inclusion criteria. The Zureik study recruited men between 43 to 52 years old with similar demographic characteristics with our patient.

• Should I attempt to stop the exposure? In answering this question you should consider the following:

• How large and precise is the risk of harm? • What are the consequences if I withdraw the exposure? • Do I have any alternative for the exposure?

Decision is simple when the answers to these questions are clear. For example cigarette smoking has been associated with increase incidence of lung cancer and cardiac deaths, but withdrawing smoking may lead to “decrease quality of life” for smokers. But recently an alternative like nicotine patch has been shown to decrease withdrawal discomfort and eventually improve smoking cessation. So the decision to withdraw smoking for every patient consulting in the clinic is warranted. RESOLUTION OF THE PROBLEM IN THE SCENARIO Based on the appraisal you decided to go back to your consultant and inform him about the study that you found. He thanked you for the information and promised to withdraw the anti‐cholesterol drug when the patient comes back.

REFERENCES Espallardo, NL. Research Protocol Development for Resident Physicians. Family Medicine Research Group, Inc. Manila, 2000. Levine M, Walter S, Lee H, Haines T, Holbrook A, Moyer V, for the Evidence Based Medicine Working Group. How to Use an Article about Harm. JAMA, 1994;271(20):1615‐1619.


Page 54

WORKSHOP CRITICAL APPRAISAL OF AN ARTICLE ABOUT HARM (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making about harmful effect of a substance or drug. Another objective is to introduce concepts of critical appraisal of an article regarding a harmful effect of a substance or drug focusing on the following:

• Validity • Cohort study, case‐control study, case series and case report • Interpretation of the results • Applicability of the results



Page 55





Page 56

Appraisal Sheet CLINICAL DECISION ON HARM CLINICAL SCENARIO OR QUESTION

SEARCH


Is the objective of the article on harm similar to your clinical dilemma?


Were there clearly identified comparison groups? Were the exposures and outcomes measured in the same way in the groups compared? Was follow‐up sufficiently long and complete? Is the temporal relationship between the exposure and outcome correct and dose response gradient present?


Page 57



What is the magnitude of the association between exposure and outcome? Was the estimate of the risk precise?


Are the study patients similar to my own? Should I attempt to stop the exposure?



Page 58

READING ASSIGNMENT CRITICAL APPRAISAL OF AN ARTICLE ABOUT PROGNOSIS (SESSION BRIEFING)

CHECKLIST Have you read the previous topic on harm?

Yes No Have you undergone the workshop on harm?



OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about prognosis. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the reading assignment for an article about prognosis. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper. After reading the paper you can proceed to conduct the group workshop.


Page 59

READING ASSIGNMENT CLINICAL DECISION ON PROGNOSIS

CLINICAL SCENARIO Your brother consulted you because of what happened to his wife lately. She just had a miscarriage (after 6 months of pregnancy) last week and she had not taken her meals lately. They are already in their five years of marriage and they don’t have a child yet. Your brother is asking you if they should take a vacation despite his being very busy at work and he is trying to save for the house mortgage. What will you advice him? If you advice him to take a vacation, they might loss their house or even his job. If you advice to continue working and let time heal his wife’s grief, the chance of a psychological problem worsening is great. Since you have attended a workshop on evidence based family practice, you formulated the question “What is the chance that my brother’s wife will go further into grief after the miscarriage considering that they still don’t have any child yet?” and went to the library.

SEARCH You typed the combination of the terms, pregnancy loss and prognosis and grief and you were able to get across the article of Janssen et al. study entitled “A prospective study of risk factors predicting grief intensity following pregnancy loss” publish in the Archive of General Psychiatry last January 1997. Now you proceed to see if the information in this study can be used to answer your brother’s question.

WHAT IS PROGNOSIS Prognosis refers to the development of possible “outcome” of disease i.e. death in patient with cancer. Prognostic factors are characteristics of a particular patient can be used to predict that patient's eventual outcome i.e. patients advanced TNM cancer stage may have more death than those with less advance


Page 60

TNM cancer stage. Prognostic factors need not necessarily cause the outcomes but just predict their development. Thus prognosis is a prediction of the probable outcome of a disease based on a individual's condition and the usual course of the disease as seen in similar situations. Risk factors on the other hand are patient characteristics associated with the development of the disease rather than the outcome of the disease. The study designs for prognostic and risk factors are cohort study and case‐control study. Cross‐sectional studies do not give valid conclusions about prognostic or risk factors because temporal relationship between factors and outcome is not established. A cohort study follows one or more groups (cohorts) of individuals who have not yet suffered an adverse event and monitor the number of outcome events over time. An ideal cohort study consists of well defined sample of subjects representative of the population of interest, and uses objective outcome criteria. Investigators can also collect "cases" of individuals who have already suffered the outcome event (death due to cancer) and compare them to "controls" who have not (cancer patients who are alive). In these "case‐control" studies the investigators count the number of individuals in each group with a particular prognostic factor (advance or less advance TNM cancer stage). To be valid studies on prognosis, these observational studies must be conducted and reported with information that address this appraisal guide (Von Elm, 2007)


• Is the objective of the article on prognosis similar to your clinical dilemma?

Your formulated clinical question must be addressed by the objective of the study. The PIO can still be applied in this type of article. The objective of the study must clearly state that it is determining the prognosis (outcome) of some patients with the prognostic factor (intervention/exposure) among patients with the disease being studied (population).


Page 61

VALIDITY GUIDES

• Was there a representative sample of patients without the outcome at the start of observation?

The authors must specify how they defined or diagnosed the patients included in the study. The authors should also specify at what stage of the disease they started observing their patients. If these were not done bias can distort the result of the study. If the study included patients who are more severe, the prognosis will naturally be poor and if they include patients who are mild, the prognosis will be good. However if you mix these patients in one study without subgroup analysis, the results will be mixed and biased. If this is not explicit in the study, you can look at the inclusion criteria or examine the setting where the study was done. The inclusion criteria may give the basis for the diagnosis, and the setting may give the stage of the disease i.e. outpatient setting may have included patients in the earlier stage and hospital setting may be patients in the late stage. In the Janssen et al study, 221 women were recruited through a magazine add. They had a stable marriage and reported a recent pregnancy loss. So these women may have been recruited at a relatively similar stage.

• Was follow‐up sufficiently long and complete? Just like in the paper about harm, the length of follow‐up must be sufficiently long enough to detect the outcome. If follow‐up is short, the chance of arriving at a good prognosis is high because few will develop the outcome resulting to false hopes for the patient. If it is too long, the prognosis will be poor because everybody will eventually die in the long term. Measuring prognosis over a given period is usually acceptable i.e. 5 year survival for chronic diseases, 6‐24 months survival for cancer, 30 days survival after ICU admission etc. The number of lost to follow‐up will also lead to bias results especially when the outcome is unknown. If patients were lost to follow‐up because they felt bad about the outcome, prognosis will be better if they are excluded in the analysis. If they were lost to follow‐up because they felt better and the investigators assumed the worse scenario, the prognosis will look bad. In the Janssen et al study, follow‐up was 94%, a relatively high rate.


Page 62

• Were the criteria for determining the prognostic factor and outcome explicit and credible?

The criteria for determining the outcome in study about prognosis are usually straightforward I.e. mortality. Mortality or survival can be taken from death certificates and other medical records; morbidity can be taken from hospitalization records, etc. In some cases outcomes are recurrence of disease or disease progression in which case this must be clearly defined. Definitions can be taken from the NLM MESH definitions or ICD 10 classification of the WHO. The Janssen et al study, measured the following prognostic factors; a) Symptom Checklist‐90, b) Dutch Personality Inventory, and c) information about quality of partnership, education, religion, social support etc. using existing records and surveys. The outcomes were measured using the Perinatal Grief Scale immediately after pregnancy loss and at 6, 12 and 18 months.

• Was there adjustment for other prognostic factors? Age and sex are factors that can affect prognosis but something we cannot do about. Thus many prognostic studies look at the effect of other modifiable prognostic factors by adjusting for age and sex. Doing subgroup analysis does this. In subgroup analysis, the results of the study are presented for each subgroup i.e. age and sex. Thus the prognosis of different TNM stage for cancer can be presented in different age group or in different sex. Another method is through multivariate analysis or regression model approach. In this approach the basic variables included in the model are the prognostic factor and the outcome. To adjust for age and sex, these variables are included into the model. This is usually described in the analysis section of the methodology. In the Janssen et al study, multivariate analysis was done. OVERALL, IS THE STUDY VALID? Since all the validity questions were fulfilled, the study can be considered to be valid.


Page 63


• How large is the likelihood of outcome to occur in those with the prognostic factor in a specified period of time? Was it statistically significant?

The relative risk is the incidence of the outcome in the group with the prognostic factor divided by the incidence of the outcome in the group without the prognostic factor. If the outcome being measured is death and the relative risk is more than 1, then the factor results into poor prognosis and if less than 1 the factor causes good prognosis. Relative risk is usually computed when the design is a cohort study. In a case control study, the odds ratio is computed. The odds ratio approximates the relative risk, especially when the disease is rare. If the relative risk or odds risk is 1.11, it means the chance of developing the outcome is just slightly higher if the patient has the prognostic factor. If the relative risk or odds risk is 1.99 it means the chance of developing the outcome is almost two times (2x) and if the relative risk or odds risk is 9.89 the chance is almost ten times (10x). The question of “how large is the chance” involves preferential judgment from the patient and the physician. To be statistically significant, the upper and lower values of 95% confidence interval should be greater than 1 to say that the factor gives a bad prognosis when the outcome is death. If one value of the 95% confidence interval is less than 1 and the other is more than 1, then the effect of the prognostic factor is uncertain. In some cases, studies report the p value for statistical significance i.e p <0.05 as significant. The Janssen et al study reported that grief intensity was higher for a) women who had been pregnant longer, b) pre‐loss neurotic personalities, c) pre‐loss psychiatric symptoms, and d) did not have any living children. All these factors were significant at p <0.05. CAN THE RESULTS HELP ME IN CARING FOR MY PATIENTS?

• Are the study patients similar to my own? Just like in an article about harm, for the prognostic factor to be extrapolated to your patient you have to be assured that the characteristics of your patient is similar to the study’s inclusion criteria. The setting may also be important. Patients being observed in setting where the facilities are advanced and complete may have better prognosis than among patients who are being observed in resource poor setting even though they have the same prognostic factor.


Page 64

The subjects in the Janssen et al study were women who reported recent pregnancy loss with stable marital relationship. The subjects were similar to the sister‐in‐law’s case.

• Can I use the results to decide on the intervention or reassure my patient?

Prognostic data should be used in decisions about therapy. Knowing the probability of the outcome based on the prognostic factors present in the patient should influence the decision to give or withhold treatment. For example surgical excision for cancer with the hope of improving survival should be withheld in favor of palliation treatment if the prognosis of the patient is very poor. Prognosis data may also be helpful in reassuring anxious patients about their outcome. For example some patients with dyspepsia may become too worried about the chronic epigastric symptom and can be reassured and counseled about the low prognosis of dyspepsia leading to bleeding ulcer or cancer. RESOLUTION OF THE PROBLEM IN THE SCENARIO Based on the Janssen et al study, you would rather advice your brother to take a vacation, because his wife’s grief may even intensify based on the results of Janssen et al study.

REFERENCES Andreas Laupacis, George Wells, W. Scott Richardson, Peter Tugwell for the Evidence‐Based Medicine Working Group. How to Use an Article about Prognosis. JAMA. 1994;272(3):234‐237. National Library of Medicine. Medical Subject Headings. www.ncbi.nlm.nih.gov/sites/entrez (May 27, 2008). Von Elm E, Altman D, Egger M, Pocock S, Gøtzsche P, Vandenbroucke J. STROBE Initiative. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ 2007; 335: 806‐808.


Page 65

WORKSHOP CRITICAL APPRAISAL OF AN ARTICLE ABOUT PROGNOSIS (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making using an article about prognosis. Another objective is to introduce concepts of critical appraisal of an article regarding prognosis:

• Validity • Representative sample • Interpretation of the results • Applicability of the results




Page 66




Page 67

Appraisal Sheet CLINICAL DECISION ON PROGNOSIS CLINICAL SCENARIO OR QUESTION

SEARCH




Was there a representative sample of patients without the outcome at the start of observation? Was follow‐up sufficiently long and complete?

Were the criteria for determining the prognostic factor and outcome explicit and credible? Was there adjustment for other prognostic factors?


Page 68



How large is the chance of the outcome to occur in a specified period of time? How precise were they?


Are the study patients similar to my own? Can I use the results to decide on the intervention or reassure my patient?



Page 69

READING ASSIGNMENT CRITICAL APPRAISAL OF AN ARTICLE ABOUT HEALTH ECONOMIC ANALYIS (SESSION BRIEFING)

CHECKLIST Have you read the previous topic on prognosis?

Yes No Have you undergone the workshop on prognosis with your group?


Yes No Please state the reasons below and share it to the group before starting the next workshop.

OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about a health economic analysis. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the reading assignment for an article about a health economic analysis. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper. After reading the paper you can proceed to conduct the group workshop.


Page 70

READINGS CLINICAL DECISION USING AN ARTICLE ON HEALTH ECONOMIC ANALYSIS

CLINICAL SCENARIO Your father is the mayor of one of the town in the Visayas. His budget for health care is minimal but he wants to start a program for pregnant women. Although most of the pregnancies in the area were low risk, most patients still go to the nearest obstetrician who is in the city 50 kilometers away. He is planning to set‐up a maternity hospital with a trained obstetrician in your hometown. He is now asking you for a recommendation. What is your recommendation?

SEARCH Having attended an EBM workshop you thought that the best way to convince your father is to show cost‐effectiveness of handling low risk deliveries with different types of trained physicians. You were able to get an article by Ratcliffe and Tucker entitled “The costs of alternative types of routine antenatal care for low‐risk women: obstetrician‐led shared care vs care by general practitioners and midwives” published in the Journal of Health Services and Policy, 1996. Appraising the article and conveying the information to your father seemed to be a good strategy, so you proceeded in appraising the article.

HEALTH ECONOMIC ANALYSIS Health economic analysis is a formal, quantitative methods used to compare alternative strategies with respect to their cost and their expected outcomes (Eisenberg, 1989). Its purpose is to inform decisions on resource allocation. It can potentially inform decisions in institutions like hospitals and in regional or national health policy (Russell, 1996). In this design the cost of a particular intervention is estimated. Estimation include direct and indirect costs. There are three types of economic analysis depending on the type of outcome. If the


Page 71

outcome being considered is effectiveness of treatment, it is called cost‐effectiveness analysis. If the outcome is savings in terms of monetary units it is called cost‐benefit analysis. If the outcomes are equal and the cost is the only one being compared it is called cost minimization.


• Is the objective of the article on economic analysis similar to your clinical dilemma?

Your scenario must be addressed by the objective of the study. The perspective or “point of view” in economic analysis usually refers to the one who will pay for the intervention. Often, the point of view of the economic analysis is stated in the objective and this is important to determine its relevance to your case scenario. If you are using an economic analysis for policy decision like the government pay for this kind of drug or facility, then the point of view must be the societal point of view. If your scenario is to assist a patient to make a decision on an intervention in a “pay‐for‐service” setting, then the perspective must be from the patient or “payer” perspective. Health insurance perspective is also a payer perspective. VALIDITY GUIDES

• Did the analysis provide a full economic comparison of health care strategies?

Physicians usually choose between two alternatives. The range of alternative strategies examined must include at least the currently accepted standard and the new alternative (O’Brien, 1997). Another alternative is the “do nothing” alternative but may not be realistic in some cases because of ethical issues. When we compare the cost of giving each of the two alternatives, this is cost analysis. When we use this to make a decision, we only give the alternative with the lowest cost that may not be necessarily effective. When we use comparison of effectiveness such as a randomized controlled trial in making a decision, we give an effective alternative that the patient may not be able to afford. Thus it makes sense to consider cost and effectiveness when making clinical decisions. A full economic analysis compares not only the cost of the two alternatives but also integrate information about efficacy of the alternatives. Thus cost and


Page 72

outcomes should both be analyzed for each alternative strategies being compared. This can be achieved if the study design is a cost‐benefit or a cost effectiveness approach. The Ratcliffe and Tucker study was a full economic valuation of the cost of tests, investigations, personnel, cost incurred by the patients of pregnancies delivered by obstetricians vs family physicians. The study included patients enrolled in a randomized controlled trial to answer the effectiveness outcome.

• Were the costs and outcomes properly measured and valued? Cost pertains to resources used and this must be differentiated from charges or prices of commodities. The point of view of costing refers to the one who will pay for the cost and this may differ. For example cost to government hospital or funder may be different from the point of view of the patient. What may be cost saving for the funder may actually be an increased cost for the patient. The ideal point of view is from the society’s view, but this is difficult to measure. Thus proper measurement of cost may differ from the health care system. In a system where the payment is a fee‐for‐service set‐up, an economic analysis on the point of view of the paying patient may be a good basis for making decision. In a system where health care is being paid for by the government, an economic analysis from the point of view of society may be a good basis. Thus in measuring cost, the point of view of the analysis must be established (O’Brien, 1997). Outcomes in health care must be an outcome that is of value to the patient and society. It should be something that can be appreciated by the patient. For example in making a decision about the treatment for hypertension, outcomes like decrease in incidence of mortality or stroke, decrease in hospitalization or myocardial infarction instead of just the lowering of blood pressure should be the outcome to be considered. In addition these outcomes must be measured in the best possible designs i.e. randomized controlled trials for treatment, controlled comparison for complex intervention etc. Systematic reviews or meta‐analysis of these interventions are better methods for establishing outcomes. Lastly, the cost and outcome must be expressed as a ratio i.e. cost per outcome (cost per life‐year gained, or cost per death avoided etc.). This expression of result will give the most relevant information for decision making. In the Ratcliffe and Tucker study costs were extracted from clinical data that came from a randomized controlled trial. They included cost per patient, staffing cost, non‐health services cost and mean societal cost. Cost data was available in about 94% of subjects included in the trial.


Page 73

• Was appropriate allowance made for uncertainties in the analysis? Economic analysis usually depends on analysis of secondary data and since not all data are available for each alternative, some assumptions need to be made. There are uncertainties to these assumptions. A good economic analysis is one that recognizes these uncertainties and look at how it affects their findings. This method is called sensitivity analysis (O’Brien, 1997). In sensitivity analysis the estimates for key variables in costs are changed in order to assess their impact they on the results. The changes can be based on variation in price index, opportunity costs, geographical price differences etc. In terms of outcomes, the variation can be from confidence intervals or from lowest and highest effect noted from the studies that were reviewed. OVERALL, IS THE STUDY VALID? Since the study was a full economic comparison and wide perspective of cost was considered you decided that the study was valid. WHAT ARE THE RESULTS?

• What were the costs and outcomes of each strategy? Economic analysis papers should have tables that report the costs of resources used in the alternative intervention such as drugs, personnel services, facilities, supplies etc. It should also contain tables about the outcome of each alternative. Lastly, this is expressed as a cost‐effectiveness ratio or cost‐benefit. The table below is a good guide to help decide the alternative to choose. “C” will be the best choice since it is more effective and less cost. “B” is not a good choice because it is less effective but more expensive. “A” is more effective but more expensive as well. Effectiveness

High Low Cost High A B

Low C D Sometimes, an alternative may be more effective but also more expensive. To make a decision in this scenario, an incremental analysis should be done. Incremental cost is the amount we pay for the added effectiveness of the


Page 74

alternative. Usually the availability of resources and personal judgment may be needed to decide whether the incremental cost is worth it. Ratcliffe and Tucker showed that the total societal mean cost for GP or midwife care was lower by P 2,178 and was statistically significant.

• How much does allowance for uncertainty change the result? Looking at how uncertainties affect the results is called sensitivity analysis. A direct approach for doing this is by computing for the cost effectiveness ratio using the lower and upper limit of the 95% confidence interval of the effectiveness outcome. If the cost‐effectiveness values are reversed with sensitivity analysis then the results are considered to be soft and its reliability is less. CAN THE RESULTS HELP ME IN CARING FOR MY PATIENTS?

• Could my patients expect similar outcomes? The inclusion criteria of the cited clinical trial or other studies reviewed to determine the outcome and the setting from which the trial was done can be duplicated in your setting will play an important factor. If the patients in the study are similar to your patient and setting for which the intervention was given can be duplicated, then you can expect the same outcome (O’Brien, 1997).

• Could my patients expect similar costs? The health care system may be different from the setting where the economic analysis was done. This difference may lead to difference in costing once applied for decision making in your setting. Cost data may be different for two reasons: 1) clinical practice vary in resource consumption associated with the treatment and 2) prices for resources differ from those used in the study (O’Brien, 1997). This can only be answered if the analysis presented the detailed cost so the reader can decide whether the costing in the study can also be applied in his/her own setting. Countries may differ with respect to the value they place on health benefits. If $50,000 per life‐year is an acceptable cost‐effectiveness threshold for the US it may not be affordable in the Philippines. Countries vary in their willingness to pay for health care (O’Brien, 1997).


Page 75

RESOLUTION OF THE PROBLEM IN THE SCENARIO Since most pregnancies in your hometown were low risk, you showed to your father that the town will save more money if they hire family physicians or midwives rather than an obstetrician. You called up your father one month later and the town council opted to hire more midwives.

REFERENCES Espallardo, NL. Research Protocol Development for Resident Physicians. Family Medicine Research Group, Inc. Manila, 2000. Eisenberg JM. Clinical economics. A guide to the economic analysis of clinical practices. JAMA, 1989; 262:2879‐86. O'Brien B, Heyland D, Richardson WS, Levine M, Drummond M, for the Evidence‐Based Medicine Working Group. How to use an Article on Economic Analysis of Clinical Practice: Validity Guides. JAMA, 1997; 277(19):1552‐1557. O'Brien B, Heyland D, Richardson WS, Levine M, Drummond M, for the Evidence‐Based Medicine Working Group. How to use an Article on Economic Analysis of Clinical Practice: Results and Applicability. JAMA, 1997; 277(22):1802‐1806. Russell LB, Gold MR, Siegel JE, Daniels N, Weinstein MC. The role of the cost‐effectiveness analysis in health and medicine. JAMA, 1996; 276(1)‐1172‐7.


Page 76

WORKSHOP CRITICAL APPRAISAL OF AN ARTICLE ON HEALTH ECONOMIC ANALYSIS (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making about an economic analysis. Another objective is to introduce concepts of critical appraisal of an article regarding an economic analysis focusing on the following:

• Validity • Costs in health care • Interpretation of the results • Incremental costs • Applicability of the results



Page 77





Page 78

Appraisal Sheet CLINICAL DECISION ON ECONOMIC ANALYSIS CLINICAL SCENARIO OR QUESTION

SEARCH


Is the objective of the article on economic analysis similar to your clinical dilemma?

VALIDITY GUIDES

Did the analysis provide a full economic comparison of health care strategies? Were the costs and outcomes properly measured and valued?

Was appropriate allowance made for uncertainties in the analysis? Are estimates of costs and outcomes related to the baseline risk in the treatment?


Page 79



What were the incremental costs and outcomes of each strategy? Do incremental costs and outcomes differ between subgroups? How much does allowance for uncertainty change the result?


Could my patients expect similar outcomes? Could my patients expect similar costs?



Page 80

READING ASSIGNMENT CRITICAL APPRAISAL OF AN ARTICLE ABOUT SYSTEMATIC REVIEWS OR META-ANALYSIS (SESSION BRIEFING)

CHECKLIST Have you read the previous topic on economic analysis?

Yes No Have you undergone the workshop on economic analysis with your group?



OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about systematic reviews or meta‐analysis. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the assignment for an article about systematic reviews or meta‐analysis. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper. After reading the paper you can proceed to conduct the group workshop.


Page 81

READING ASSIGNMENT CLINICAL DECISION ON SYSTEMATIC REVIEW OR META-ANALYSIS

CLINICAL SCENARIO Your grandmother had a history of fall a week ago after taking a bath. She was brought to the hospital for treatment of minor bruises in her knees. The x‐rays were normal. Your mother asked you if she needs a walker or cane to prevent falls and subsequent injury.

SEARCH You ask a colleague from Rehabilitation Medicine and she gave you an article from the NHS Center for Reviews and Dissemination she got from the internet entitled “Preventing falls and subsequent injury in older people”.

REVIEWS, SYSTEMATIC REVIEWS AND META-ANALYSIS A review is secondary study design that integrates findings of two or more studies that discuss similar topic usually defined by PIO. There may be some bias when the reviewer subjectively decides which studies to include or exclude in the review. A systematic review is similar to review but has a way to systematically search the literature searching and has systematic rules in combining the studies to be reviewed. The results of systematic reviews are more often objective than a review. Meta‐analysis is like a systematic review but applies some statistical analysis to the results. A meta‐analysis is a procedure that integrates and combine the results of two or more primary studies that are similar in the population enrolled the intervention used and the outcome measured. The pooled result is then subjected to a statistical analysis. A well conducted meta‐analysis allows a more objective appraisal of the existing evidence about a problem than a traditional review or systematic review. Meta‐analysis may also be biased owing to the inclusion or exclusion of some irrelevant or relevant studies respectively (Espallardo, 2000).


Page 82

Sample of meta‐analysis result


• Is the objective of the article on meta‐analysis similar to your clinical dilemma?

Your formulated clinical question must be addressed by the objective of the study. The objective of an appropriately done meta‐analysis is often a focused clinical objective with PIO and the method being clearly defined. This is often used for the systematic literature search and basis for inclusion or exclusion. VALIDITY GUIDES

• Did the review address a focused clinical problem? Systematic reviews of the medical literature try to summarize publications related to a similar topic. Because several articles are combined together, sometimes the purpose of the review is not clear or very broad. It therefore becomes difficult to determine what the review is trying to achieve.


Page 83

In order to know what the objective of the review is, the clinical problem must be focused. There must be a clear description of the patient and its relation with an exposure or an outcome. The NHS study specifically stated that they tried to identify strategies that prevent falls and subsequent injury in older people. The focused objective seemed to apply to your problem.

• Were the criteria for searching and selecting articles for inclusion and exclusion explicit and credible?

In conducting systematic reviews, a systematic search and appraisal of the literature should be done in order to:

• ensure that no relevant articles were missed • studies were included because they are good studies and not because

they agree with the authors opinion • studies were excluded because they are bad studies and not because

they disagree with the author’s opinion. Thus paper should describe the method of searching for the medical literature. Statements like “an electronic search of published articles in the MEDLINE using the terms . . . from 1966 to 2000 was done” must be found somewhere in the methodology section. This assures the readers that the findings of the study were based on a wide range of literature source and represent the most current and complete information about the clinical problem. The paper should also describe how they include or exclude retrieved articles. The paper must contain statements like “all retrieved abstracts were reviewed by three independent reviewers and articles that were randomized controlled trial on . . . using the intervention . . .” somewhere in the method section. This assures the readers that the articles used in the study were objectively chosen and not because they agree with the authors opinion. The NHS study identified trials published in computerized databases like Social Science Citation Index, PSYCHLIT, EMBASE, RCN database, AMED and UNCOVER. The citations also identified reviews and peer contribution from reviewers and other experts in the field.

• Was the validity of included studies appraised and the appraisal reproducible?

Even if all included studies are randomized controlled trials, there may be some small differences among different trials that might affect the results of the study. Thus a standard appraisal of each article must be done. Peer review may not be


Page 84

a reliable method of appraisal. Differences in peer perception and interest may lead to differences in the result of the appraisal. While there are no agreed standards to evaluate validity, the review must have at least developed a checklist of criteria that focused on the methodology of the study being appraised. The NHS study included only randomized controlled trials that evaluated strategies to prevent falls. OVERALL, IS THE STUDY VALID? Overall the study is valid. WHAT ARE THE RESULTS?

• What are the overall results of the systematic review? When looking at the results of a systematic review or meta‐analysis, you should look for clinically relevant presentation like lower mortality rates in one group compared to the other, or the difference in quality of life scores between the two groups. Sometimes subgroup analysis i.e. patients with high risk and patients with low risk, to see the different effect of an exposure may also be helpful. In a meta‐analysis, the confidence interval of the overall results can also be computed and this can provide information about the precision of the results. Thirty‐six randomized controlled trials were included in the meta‐analysis. The results showed that 10‐24 weeks of exercise including balance training showed an effective risk reduction by as much as 37% with an adjusted fall incidence ratio of 0.90 and with a 95% CI of 0.81 to 0.99. CAN THE RESULTS HELP ME IN CARING FOR MY PATIENTS?

• Are the study patients similar to my own? In a systematic review, the patient’s characteristics are varied because they came from different studies. Application to patients therefore becomes wider. When variation in patient inclusion may influence the effect, subgroup analysis between different patient characteristics may also help decide what kind of patient will benefit from the intervention or will be affected by the exposure.


Page 85

This information can be seen in the methodology section where the researchers describe the type of patients in the literature search and inclusion criteria of the studies. The NHS meta‐analysis included only studies done on elderly.

• Are the results of the review relevant to my patient? When the clinical question of the review is focused, the answer to this question becomes evident. You only need to focus on the outcome measured and decide whether this is relevant to your patient. When the outcome differs between studies, they are combined and this is reported in systematic reviews or meta‐analysis as effect size. This is a difficult situation because the outcomes are combined and you cannot easily decide whether the outcome is relevant or not. In this case you can look at the results of individual studies and choose studies that give relevant outcomes and use them to make a decision. RESOLUTION OF THE PROBLEM IN THE SCENARIO Based on the review, balance training rather than walker devices will help prevent further fall and subsequent injury.

REFERENCES Espallardo, NL. Research Protocol Development for Resident Physicians. Family Medicine Research Group, Inc. Manila, 2000. Oxman A, Cook D, Guyatt G, for the Evidence Based Medicine Working Group. How to Use an Overview. JAMA, 1994;272(17):1367‐71.


Page 86

WORKSHOP CRITICAL APPRAISAL OF AN ARTICLE ABOUT A SYSTEMATIC REVIEW OR META-ANALYSIS (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making using an article about a systematic review or meta‐analysis. Another objective is to introduce concepts of critical appraisal of an article regarding differential diagnosis focusing on the following:

• Validity • Review, systematic review or overview, meta‐analysis • Interpretation of the results • Applicability of the results



Page 87





Page 88

Appraisal Sheet CLINICAL DECISION ON SYSTEMATIC REVIEW OR META‐ANALYSIS CLINICAL SCENARIO OR QUESTION

SEARCH




Did the review address a focused clinical problem? Were the criteria for searching and selecting articles for inclusion and exclusion explicit and credible? Was the validity of included studies appraised and the appraisal reproducible?


Page 89



What are the overall results of the systematic review?


Are the study patients similar to my own? Are the results of the review relevant to my patient?



Page 90

READING ASSIGNMENT CRITICAL APPRAISAL OF AN A CLINICAL PRACTICE GUIDELINE (SESSION BRIEFING)

CHECKLIST Have you read the previous topic on systematic reviews?

Yes No Have you undergone the workshop on systematic reviews with your group?



OBJECTIVES The purpose of the reading assignment is to introduce to the participants the concept of medical decision making using an article about a clinical practice guideline. At the end of the reading session, you should be able to answer the user guides questions for the workshop.

INSTRUCTIONS Read the reading assignment for an article about a clinical practice guideline. Focus on the critical appraisal questions, why they are asked and how to get the answers from the paper.


Page 91

READING ASSIGNMENT HOW TO USE A CLINICAL PRACTICE GUIDELINE

CLINICAL PRACTICE GUIDELINES "Clinical Practice Guidelines are systematically developed statements to assist practitioner decisions about appropriate health care for specific clinical circumstances."(Field, 1990). Guidelines were developed to:

• Make evidence‐based management explicit. • Make clinical decision making more objective and scientific. • Assess professional performance. • Educate the patients and practitioners about current "best practice."

If guidelines are to improve practice, they need to be developed by the people who are actually going to have to apply them. Guidelines, like so much else in healthcare today are no longer as immutable as the Laws of the Medes, you will have to periodically check that they are working, that they are getting to the people who are going to use them and that they are up to date. In general, good topics for guidelines are those which:

• contribute a high workload • poor treatment risks disastrous outcomes • change is practical with the resources you have available • there is variation in current management (i.e. there is uncertainty about

the best management strategy), but some hope of reaching a consensus • the changes you wish to implement will be acceptable to patients • there is good evidence to back up the protocols

It might be useful to consider the criteria used to select illnesses for screening programs (how common, how serious, how preventable, how acceptable? Clinical practice guidelines, which have been defined as "systematically developed statements to assist practitioner and patient decisions about appropriate health care for specific clinical circumstances," represent an attempt to distill a large body of medical knowledge into a convenient, readily useable format. Like overviews, they gather, appraise and combine evidence. Guidelines, however, go beyond most overviews in attempting to address all the issues


Page 92

relevant to a clinical decision and all the values that might sway a clinical recommendation. Like decision analyses, guidelines refine clinical questions and balance trade‐offs. Guidelines differ from decision analyses in relying more on qualitative reasoning and in emphasizing a particular clinical context.

CASE SCENARIO At the end of a busy day in your clinic, you are glad to find that your last patient is a 45/male previously diagnosed to have hypertension. He claims that his highest blood pressure for the past month was 160/90 and his usual blood pressure was 130/90 and he has been taking a beta‐blocker for the past year. He was relatively symptom free save for occasional headache and nape pains during BP spikes. Thinking that this was just another run of the mill hypertensive patient you were ready to refill his prescription and give your usual advice regarding diet and exercise. As you were about to do just that, your patient began asking you the relative benefits of the alternative drugs available in the market. He was also asking if he needed to have an ECG done together with blood chemistries, urinalysis and a 2D‐Echo since his friend who consulted another physician was advised to do that. Since he was not overweight he was asking if a regular exercise program would add any additional benefit in controlling his symptoms. You gave him the usual advice you knew based on your knowledge of pathophysiology and pharmacokinetics. However, as you were finally closing your clinic you decided that you were not satisfied with the answers you gave him and decided to do a search for the best available evidence.

SEARCH FOR CPG You decide that in order to find relevant answers to a variety of clinical questions, a clinical practice guideline would be the best article to retrieve. You initially searched ww.guidelines.gov and found numerous guidelines for hypertension. However, you did not have sufficient time to download and print the full text version of these guidelines. You then remembered a copy of the Philippine compendium sent to you by mail 2 years ago and taught of the Philippine Clinical Practice Guidelines on the Detection and Management of Hypertension. You then go home, sit at your desk and decide to review this article in order to be more prepared for your next patient encounter.


Page 93


• Is the objective of the article on clinical practice guideline similar to your clinical dilemma?

Your formulated clinical question must be addressed by the objective of the clinical practice guideline. Usually guideline objectives are broad i.e. answers questions about the best diagnostic test, the recommended treatment, the expected outcome or prognosis etc. VALIDITY GUIDES

• Were all important options and outcomes considered? Guidelines aid us in our decision making skills and we make better judgment calls if we know all the alternative options open to us and the relative harm and benefits of each choice. Guideline developers then should present most of the reasonable options seen in practice and their corresponding outcomes. In the case of the Philippine Clinical practice guideline on the detection and management of hypertension, several treatment options were presented ranging from beta‐blockers, diuretics, ACE inhibitors, calcium channel blockers with recommendations of the best alternative for hypertensive patients with co‐morbid conditions. The guideline however did not include the newer generation anti‐hypertensives. As important as presenting all the options, the corresponding outcomes such as morbidity and mortality data, prevention of complications and other measures that improve health related quality of life should be reported. Inasmuch as all these will be helpful and clinically relevant to individual patients. In this hypertension guideline, mortality and morbidity data together with prevention of hypertensive complications were the outcomes given emphasis. However costs and side effects were not well mentioned.

• Was an explicit and sensible process used to identify, select, and combine evidence?

Guideline developers must allow the reader to know how the evidence has been tracked, reviewed, appraised and combined in order to allow them to ascertain the validity of the gathered evidence. Developers should specify a focused


Page 94

question, search the literature for available evidence, critically appraise this evidence and summarize the results in an easy to understand material. The Philippine Clinical Practice Guidelines for Hypertension was not very clear regarding how they searched the literature and what database they used. Mention of tracking, retrieving and appraising was done in Phase 1 of the introduction section, but the complete way on how this was done was not mentioned. However, summary on how the articles were reviewed and graded was provided.

• Is the guideline likely to account for important recent developments? You should look for two important dates: the publication date of the most recent evidence considered and the date on which the final recommendations were made. Some authorities also identify important studies in progress and new information that could change the guideline. Ideally, these considerations may be used to qualify guidelines as "temporary" or "provisional," to specify dates for expiration or review, or to identify key research priorities. For most guidelines, however, you must scan the bibliography to get an impression of how current a particular guideline may be. Once you are confident that the clinical practice guideline addresses your clinical question and is based on a rigorous up‐to‐date assessment of the relevant evidence, you can review the recommendations to determine how useful they will be in your practice. The reader is advised to check the bibliography section of the guideline and check the dates of the most recent articles included. Ideally, the evidence should be within the last 2 years before the guideline was published. Since medical knowledge rapidly transforms, this will ensure that our recommendations will not be outdated. Hence, there is a need to revise guidelines periodically. This guideline on hypertension was released in 1995 and the latest evidence upon looking at the bibliography was in that same year. Being at present in the year 2000 and with the rapid developments in antihypertensive medications, the guideline should be due for review and revision.

• Has the guideline been subjected to peer review and testing? People may interpret evidence differently and their values as to what important options and outcomes are may differ. As such, a guideline that has been subjected to scrutiny by external reviewers and tested in an actual clinical practice setting and found acceptable might be easier to use.


Page 95

OVERALL, IS THE GUIDELINE VALID? Once you are confident that the guideline meets at least 2 out of the 3 requirements above, you can review the recommendations and its applicability to our individual patients. At present, the Philippine Clinical Practice Guidelines on Detection and Management of hypertension although a little bit outdated at the present time will do. WHAT ARE THE RECOMMENDATIONS

• Are practical, clinically important, recommendations made? To be useful guidelines should give practical, unambiguous advice addressing a particular clinical situation. Recommendations should be simple and specific at the same time comprehensive enough to allow the reader a chance to assess the benefits and costs of following the particular recommendation. In this hypertension guideline, recommendations are divided into every aspect of any encounter with a hypertensive patient. The following 6 questions are addressed by the guideline:

• How should blood pressure be measured? • How should hypertension be diagnosed? • How should hypertension be worked up? • What advice should hypertensive patients receive regarding lifestyle

modification? • How should hypertension be treated? • How can hypertension be prevented among normotensives?

This then allows the clinician to answer aspects regarding diagnosis, laboratory work‐ups, treatment options, non‐pharmacologic advice and preventive measures.

• How strong are the recommendations? The "strength," "grade," "confidence," or "force" of a recommendation should be informed by multiple considerations:

• the quality of the investigations which provide the evidence for the recommendations


Page 96

• the magnitude and consistency of positive outcomes relative to negative outcomes (adverse effects, burdens to the patient and the health care system, costs)

• the relative value placed upon different outcomes. Thus, grading of the recommendations are based on the methodological soundness of the available evidence, the number of positive outcomes in relation to negative ones and the consistency of findings across different evidences available. It is not enough to look into the fact that randomized controlled trials were used as evidence but also if findings across different trials were consistent. Inconsistent findings are at times the reason why different guideline developers have different recommendations regarding certain clinical issues. It is also important to note that different guideline developers use different standards for grading their recommendations and that this should explicitly be placed in the guideline for ease of understanding. The Philippine Clinical Practice Guidelines for Hypertension used a system adopted by the Canadian Hypertension Society. Therapy wise, a lot of Grade A recommendation meaning that evidence is based from well‐conducted trials was made. WILL THE RECOMMENDATIONS HELP YOU IN CARING FOR YOUR PATIENTS?

• Is the primary objective of the guideline consistent with your objectives?

The purpose of the guideline developers for coming up with recommendations may vary from your own. Guidelines may be disseminated to assist physicians in decision making (clinical algorithms), to evaluate their practice and the standard of care they give to their patient (quality assurance) or to set limits for physician choices (reimbursements, recertification). In any case, in order to find recommendations most suited to your needs, the purpose of the guideline should be in line with your intended objective. This hypertension guideline was made to ensure the availability of a local guideline for the detection and management of hypertension. Since your questions dealt with management issues, this guideline is appropriate for your purpose.


Page 97

• Are the recommendations applicable to your patients? You must determine if the kind of patients you have are similar to those patients targeted by the guideline. If your patients have a different prevalence or risk of disease, if the diagnostic and therapeutic options recommended are not available in your area, the guideline might not apply. The advantage of reviewing and applying a Philippine practice guideline is that it takes into account the characteristics of our setting and hopefully allows it to be more responsive. RESOLUTION OF THE PROBLEM IN THE SCENARIO Having deemed that the aforementioned guideline is valid, you would still opt to give this patient with uncomplicated hypertension a beta‐blocker. In terms of diagnostics, you would request for an FBS, Serum Creatinine, Serum potassium and urinalysis. You would request for these tests since they would have an effect on the antihypertensive you would choose. Furthermore they would provide the following additional benefits: a)detection and early treatment of diabetes, b)detection of asymptomatic renal disease, c)detection of possible secondary hypertension. You would explain to your patient that since he has no symptoms of any cardiac disease, performing an ECG and Echo is not routinely recommended. In terms of non‐pharmacologic advice, you would encourage him to go with regular aerobic exercise such as walking, jogging or cycling 30 minutes per day 3‐4x/week since regular physical activity reduces blood pressure and results in a decrease in all cause mortality. A lot of decision making, considerations of options and outcomes came into play for a relatively simple case of hypertension. And as medical knowledge improves, more options will be made available. As such the need for relevant, well‐constructed and tested guidelines to improve our clinical decision making will always be there. Again, we as clinicians should be able to adapt guidelines that are valid and whose recommendations will be most appropriate and feasible in our respective settings.

REFERENCES Hayward R, Wilson M, Tunis S, Bass E, Guyatt G for the Evidence Based Medicine Working Group. How to Use a Clinical Practice Guideline: Validity. JAMA, 1995;274(7):570-4


Page 98

Hayward R, Wilson M, Tunis S, Bass E, Guyatt G for the Evidence Based Medicine Working Group. How to Use a Clinical Practice Guideline: Recommendations and Applicability. JAMA, 1995;274(20):1630-2.


Page 99

WORKSHOP CRITICAL APPRAISAL OF CLINICAL PRACTICE GUIDELINE (SESSION BRIEFING)

OBJECTIVES The purpose of the workshop is to introduce to the participants the concept of medical decision making about multiple problems related to disease management. Another objective is to introduce concepts of critical appraisal of a clinical practice guideline:

• validity • recommended options and outcomes • up‐to‐date recommendations • relevance and certainty of the recommendations • applicability of the recommendations

INSTRUCTIONS Divide the participants into groups of six to ten persons per group. Assign a case scenario to each group and formulate an answerable problem from the scenario. Establish initial group consensus on how to proceed with the scenario. Ask the group to read the article retrieved to answer the problem in the scenario. Focus on the abstract, methods and results section. Again, establish a group consensus on how to proceed with the scenario. Note any change in decisions. Critically appraise the article using the appraisal sheet provided. Answer validity questions, analyze the recommendations and determine the applicability of the recommendations. Establish another group consensus and note any change in decision. Process the exercise. Focus on barriers and solution to the application of the exercise in usual clinic practice.


Page 100

TIME ALLOTTED The time allotted for this workshop is one hour. The recommended break‐up is:

• 5 minutes to analyze the scenario and develop consensus • 10 minutes to read the clinical practice guideline • 20 minutes to appraise the validity • 10 minutes to analyze the recommendations • 5 minutes to establish applicability • 10 minutes to summarize the process

DESIRED OUTCOME The participants should make a clinical decision on the scenario based on the critical appraisal of the clinical practice guideline.


Page 101

Appraisal Sheet HOW TO USE A CLINICAL PRACTICE GUIDELINE CASE SCENARIO OR QUESTION

SEARCH


Is the objective of the article on clinical practice guideline similar to your clinical dilemma?

VALIDITY GUIDES

Were all important options and outcomes considered? Alternatives, expected results Was an explicit and sensible process used to identify, select, and combine evidence? Is the guideline likely to account for important recent developments? Last update Has the guideline been subjected to peer review and testing?


Page 102

OVERALL, IS THE GUIDELINE VALID?

WHAT ARE THE RECOMMENDATIONS

Are practical, clinically important, recommendations made? How strong are the recommendations? Grading

WILL THE RECOMMENDATIONS HELP YOU IN CARING FOR YOUR PATIENTS?

Is the primary objective of the guideline consistent with your objectives? Are the recommendations applicable to your patients and setting?


AJR:180, January 2003

17

Accuracy of MR Imaging of the Knee in Adolescents

OBJECTIVE.

A report in the orthopedics literature states that MR imaging for internal de-rangement of the knee has a lower accuracy in adolescents than in adults and may even providespurious information that alters clinical management. This assertion has not been specificallyaddressed in the radiology literature. The purpose of our study was to determine the accuracy ofMR imaging in adolescents with regard to injury of the cruciate ligaments and menisci.

MATERIALS AND METHODS.

A database search of our institution’s records from Janu-ary 1998 to July 2000 yielded 2140 MR examinations of the knee, all of which had been per-formed with a standard knee protocol on a 1.5-T magnet. Of these 2140 examinations, 156included patients younger than 18 years. Fifty-nine of these patients underwent surgery, and theorthopedic surgeons’ operative reports were used as the gold standard with which the MR imag-ing results were compared. Thirty-four boys and 25 girls who ranged in age from 11 to 17 years(mean age, 15 years) were examined. The clinical notes for the remaining 97 patients were evalu-ated for information about management and clinical improvement.

RESULTS

. The sensitivity and specificity values for MR imaging of the menisci and cruciateligaments in adolescents were as follows: medial meniscus, 92% sensitivity and 87% specificity;lateral meniscus, 93% sensitivity and 95% specificity; anterior cruciate ligament, 100% sensitiv-ity and 100% specificity; and posterior cruciate ligament, 0% sensitivity and 100% specificity.

CONCLUSION.

Our data suggest that MR imaging of the knee in adolescents is sensi-tive, specific, and accurate.

n the adult population, MR imag-ing has been accepted as the imag-ing gold standard for detection of

internal derangement of the knee [1–3]. How-ever, when we discussed MR imaging of theknee of an adolescent patient with the orthope-dic surgeons at our institution, the surgeons ex-pressed concern that MR imaging would beless useful in our adolescent patient than in anadult patient because of the purported loweraccuracy of MR imaging for the detection ofinternal derangement of the knee in adoles-cents. This notion had apparently been ac-cepted as truth by our orthopedic surgeons, sothey sometimes forego an MR imaging ex-amination because they believe that the infor-mation from the study would not be usefuland could even be misleading. This assertionhas been made in the orthopedics literature[4] but, to our knowledge, has not been spe-cifically addressed in the radiology literature.The purpose of our study was to determinethe accuracy of MR imaging in adolescentscompared with that in adults with regard to

the detection of injuries to the cruciate liga-ments and menisci.

Materials and Methods

A retrospective search of the musculoskeletalMR imaging database for knee examinations per-formed from January 1998 to July 2000 was con-ducted. Our search yielded 2140 knee examinations.Of these 2140 examinations, 156 included patientswho were younger than 18 years. Fifty-nine of these156 patients proceeded to surgery, and the orthope-dic surgeons’ operative reports were used as the goldstandard with which we compared the MR imagingresults. The study group included 34 boys and 25girls who ranged in age from 11 to 17 years (meanage, 15 years). The clinical notes about the remain-ing 97 patients were evaluated for management andimprovement. Of the 97 patients, 49 were boys and48 were girls, with ages ranging from 13 to 17 years(mean age, 15 years).

One of five musculoskeletal radiologists from ourinstitution had prospectively evaluated each MR im-aging examination. For each patient, the radiolo-gist’s interpretations of the menisci and cruciateligaments were compared with the arthroscopic sur-

Nancy M. Major

1

L. Neal Beard, Jr.Clyde A. Helms

Received January 21, 2002; accepted after revision June 25, 2002.

Presented at the annual meeting of the American Roentgen Ray Society, Seattle, April–May 2001.

1

All authors: Department of Radiology, Duke University Medical Center, Box 3808, Durham, NC 27710. Address correspondence to N. M. Major.

AJR

2003;180:17–19

0361–803X/03/1801–17

© American Roentgen Ray Society

I

18


Major et al.

gical findings; surgery was performed by various or-thopedic surgeons. MR imaging data were thencategorized as true-positive, true-negative, false-pos-itive, and false-negative. From these data, sensitivityand specificity for the detection of meniscal and cru-ciate ligament tears in our study group of adoles-cents were computed and compared with the samevalues in adults for the same radiologists (obtainedfrom another study) over approximately the sametime period.

All the MR imaging examinations were per-formed on a 1.5-T magnet (Signa; General ElectricMedical Systems, Milwaukee, WI), and identicalprotocols were used for each of the examinations.Our standard knee protocol includes axial, sagittal,and coronal fast spin-echo T2-weighted imaging(TR/TE effective, 3500/65) with fat suppression andsagittal proton density imaging

(TR/TE, 2000/20)with fat suppression. The remaining parameters in-clude a matrix of 256

×

192, 2 excitations, a field ofview of 16

×

16 cm, and a slice thickness of 4 mm/0.4 mm.

Meniscus tears were identified if linear high sig-nal abutting the articular surface or abnormal mor-phology was seen. The anterior cruciate ligamentwas identified as torn if the fibers were disrupted andwere no longer parallel to the intercondylar notch.

Results

The sensitivity and specificity values of MRimaging for the detection of internal derange-ment of the knee in adolescents and adults areshown in Table 1.

In the group of adolescents with arthro-scopic correlation, arthroscopy showed 11 me-dial meniscus tears, 14 lateral meniscus tears,25 anterior cruciate ligament tears, and oneposterior cruciate ligament tear. Comparison

of the arthroscopic and MR imaging findingsyielded the following results. MR evaluationof the medial meniscus revealed 11 true-posi-tives, 41 true-negatives, six false-positives, andone false-negative; these values resulted in a92% sensitivity and 87% specificity. For thelateral meniscus, the MR interpretations con-sisted of 14 true-positives, 42 true-negatives,two false-positives, and one false-negative,which resulted in a 93% sensitivity and 95%specificity. MR findings for the anterior cruci-ate ligament yielded 26 true-positives and 33true-negatives with zero false-positives andzero false-negatives, which resulted in a 100%sensitivity and specificity. For the posteriorcruciate ligament, neither true-positives norfalse-positives were recorded for the MR im-aging findings; there were 58 true-negativesand one false-negative. These values yielded a0% sensitivity and 100% specificity.

The sensitivities and specificities of MRimaging for the detection of tears in the ado-lescent group were essentially the same asthose for the adult group, which included aseries of 203 patients (Table 1).

Of the 97 patients who did not undergo ar-throscopy, “normal” was assigned as the diag-nosis in 39 patients. Forty-six patients had noadditional follow-up. Other diagnoses encoun-tered were 10 bone contusions, seven patellardislocations (contusion pattern not counted inprevious group), four anterior cruciate liga-ment tears, two hematomas, two cases of Os-good-Schlatter disease, two cases of jumper’sknee, two osteochondral lesions, one posteriorcruciate ligament injury, one case of abnormal

signal in the Hoffa fat pad, one Wrisberg variantof discoid lateral meniscus, one medial collateralligament sprain, and one bucket-handle menis-cus tear. Of the four patients with anterior cruci-ate ligament tears, two refused surgery and twowere lost to follow-up. One patient had an ante-rior cruciate ligament tear and a bucket-handlemeniscus tear. Twenty-five patients had a finalimpression in the dictated report as “signal in ei-ther the meniscus or soft tissues not felt to be sig-nificant.” Therefore, the total number of casesthat were not diagnosed with pathology was 64.

Discussion

Our results indicate that the accuracy of MRimaging for the detection of internal derange-ment of the knee in adolescents is similar tothat in adults. The idea for this retrospectivedatabase study came from repeated discus-sions with various orthopedic surgeons at ourinstitution, during which we were told that MRimaging of the knee for detection of internalderangement is less useful in adolescents thanadults because of a report of decreased accu-racy. The surgeons referred us to an article byStanitski [4]. In this article, Stanitski comparedclinical examination findings, MR imaging re-sults, and arthroscopic findings in 28 childrenand adolescents (age range, 8–17 years) withknee injuries. Articular surface, anterior cruci-ate ligament, and meniscal injuries were re-viewed and the conclusions were as follows:“Overall, magnetic resonance imaging diag-noses added little guidance to patient manage-ment and at times provided spuriousinformation.” The data in Table 2 are fromStanitski’s article. Stanitski reported 75% totaldisagreement between clinical and MR imag-ing and 78.5% total disagreement between ar-throscopic findings and MR imaging results.In that study, total agreement was found be-tween clinical examination and arthroscopy in78.5% of the cases. Therefore, Stanitski as-serted that clinical examination and arthroscopyare superior to MR imaging of adolescents.

Theorthopedics literature includes a second articlewritten by McDermott et al. [5]; in that article,the authors state that the accuracy for MR im-aging of knee abnormalities in pediatric pa-tients is not well established. We believe thatour data refute the argument that MR imagingof the knee is less accurate in adolescents thanin adults.

A number of differences between our studyand that conducted by Stanitski [4] exist. First,the latter study included results from only 28patients, whereas our study included results

TABLE 1 Sensitivity and Specificity of MR Imaging of the Knee in Adolescents Versus Adults

LocationAdolescents (n = 59) Adults (n = 203)

Sensitivity (%) Specificity (%) Sensitivity (%) Specificity (%)

Medial meniscus 92 87 90 80Lateral meniscus 93 95 78 93Anterior cruciate ligament 100 100 100 97Posterior cruciate ligament 0 100 83 98

Note.—All data are from [4].

TABLE 2 Comparison of Findings from Clinical Examination, MR Imaging, and Arthroscopy of the Knee

Findings ComparedAgreement (%) Total Disagreement

(%)Total Partial

Clinical examination vs MR imaging 14.3 10.7 75Clinical examination vs arthroscopy 78.5 7.1 14.3MR imaging vs arthroscopy 7.1 14.3 78.5

MR Imaging of the Knee in Adolescents


19

from 59 patients. However, although the num-ber of patients in our study is more than doublethat in the other study, the total number is stillsmall. Therefore, a small sample is a potentialshortcoming of our study. Possible explana-tions for the small number of MR imagingstudies in adolescents include the reluctance oforthopedic surgeons to use MR imaging inthese patients because of the report by Sta-nitski and the possibility that adolescents areless likely to have internal derangement of theknee than adults.

Another difference between our study andthat of Stanitski [4] is that Stanitski usedgrade 2 meniscal signal abnormality as evi-dence for meniscus tear in an unspecifiednumber of patients. It is well known thatgrade 2 intrameniscal signal is evidence of in-trasubstance degeneration rather than a tear,because grade 2 intrameniscal signal does notdisrupt the articular surface. These cases wereerroneously diagnosed as tears in that study,therefore reducing the accuracy of MR imag-ing for revealing meniscal abnormalities. Inaddition, Stanitski did not provide the imagingparameters used to evaluate the meniscus. Ifsequences with a long TE were chosen to eval-uate the meniscus, tears could have been over-looked. Proper protocols will aid theradiologist (and surgeon) in accurately assess-ing the integrity of the meniscus.

In the Stanitski study [4], the accuracy ofthe radiologists’ interpretations of the MRimages of adults is not known. A potentialshortcoming in our study is that only muscu-loskeletal radiologists interpreted MR imagesrather than general radiologists. However, theradiology literature reports 95–100% accu-racy for anterior cruciate ligament tears, 90–

95% for medial meniscus tears, and 85–90%accuracy for lateral meniscus tears [1–3, 6,7], and there is no reason to believe that thesenumbers should not hold true for general ra-diologists. Stanitski asserted that sensitivityand specificity of MR imaging for detectinginternal derangements of the knee were infe-rior in adolescents compared with adults. Af-ter evaluating our data, we found that thesensitivity and specificity values for MR im-aging of adolescents and adults were essen-tially the same (Table 1).

Although our primary intention was todetermine the accuracy of MR imaging ofthe knee compared with arthroscopy in ado-lescents, we also assessed the outcomes forthe 97 patients who did not undergo arthros-copy. Forty-six patients did not undergo afollow-up examination. A lack of follow-upcould indicate that either the symptoms re-solved so clinical follow-up was not neededor the patient was seen elsewhere for addi-tional follow-up. Of the remaining patientswho did undergo follow-up, the visit con-sisted of one-time physical therapy or ortho-pedic follow-up without any additionalfollow-up or intervention. Four anterior cru-ciate ligament tears were identified, but thepatients did not undergo surgery at our insti-tution: two did not want surgery and theother two were lost to follow-up. The patientwith the bucket-handle meniscus tear wasamong these four patients. No additional“surgical lesions” were identified. None ofthe patella dislocations had associated carti-lage loss (our surgeons’ indication for oper-ating). The osteochondral lesions werestable by MR appearance. Surgery was notconsidered for these two patients.

In conclusion, we believe that MR imagingof the knee is just as useful as a clinical ad-junct in adolescents as in adults. Therefore,MR imaging of the knee in adolescents canassist in preventing unnecessary surgery suchas diagnostic arthroscopy. In circumstancesin which surgery is deemed necessary, MRimaging can aid in surgical planning, whichbenefits the orthopedic surgeon as well as thepatient because the information provided byMR imaging leads to decreased procedureand tourniquet time.

References

1. Mackenzie R, Palmer CR, Lomas DJ, Dixon AK.Magnetic resonance imaging of the knee: diag-nostic performance studies.

Clin Radiol

1996

;51:251–257

2. Mink J, Levy T, Crues JI. Tears of the anterior cru-ciate ligament and menisci of the knee: MR imag-ing evaluation.

Radiology

1988

;167:769–7743. De Smet AA, Graf BK. Meniscal tears missed on

MR imaging: relationship to meniscal tear pat-terns and anterior cruciate ligament tears.

AJR

1994

;162:905–9114. Stanitski CL. Correlation of arthroscopic and

clinical examinations with magnetic resonanceimaging findings of injured knees in children andadolescents.

Am J Sports Med

1998

;26:2–65. McDermott MJ, Bathgate B, Gillingham BL, Hen-

nrikus WL. Correlation of MRI and arthroscopicdiagnosis of knee pathology in children and adoles-cents.

J Pediatr Orthop

1998

;18:675–6786. Boeree NR, Watkinson AF, Ackroyd CE, Johnson

C. Magnetic resonance imaging of meniscal andcruciate injuries of the knee.

J Bone Joint Surg Br

1991

;73:452–4577. Lee JK, Yao L, Phelps CT, Wirth CR, Czajka J,

Lazman J. Anterior cruciate ligament tears: MRimaging compared with arthroscopy and clinicaltests.

Radiology

1988

;166:861–864

154

·

Januar y 20, 2000

The New England Journal of Medicine

VITAMIN E SUPPLEMENTATION AND CARDIOVASCULAR EVENTS IN HIGH-RISK PATIENTS

T

HE

H

EART

O

UTCOMES

P

REVENTION

E

VALUATION

S

TUDY

I

NVESTIGATORS

*

A

BSTRACT

Background

Observational and experimental stud-ies suggest that the amount of vitamin E ingested infood and in supplements is associated with a lowerrisk of coronary heart disease and atherosclerosis.

Methods

We enrolled a total of 2545 women and6996 men 55 years of age or older who were at highrisk for cardiovascular events because they had car-diovascular disease or diabetes in addition to one oth-er risk factor. These patients were randomly assignedaccording to a two-by-two factorial design to receiveeither 400 IU of vitamin E daily from natural sourcesor matching placebo and either an angiotensin-con-verting–enzyme inhibitor (ramipril) or matching pla-cebo for a mean of 4.5 years (the results of the com-parison of ramipril and placebo are reported in acompanion article). The primary outcome was a com-posite of myocardial infarction, stroke, and death fromcardiovascular causes. The secondary outcomes in-cluded unstable angina, congestive heart failure, re-vascularization or amputation, death from any cause,complications of diabetes, and cancer.

Results

A total of 772 of the 4761 patients as-signed to vitamin E (16.2 percent) and 739 of the 4780assigned to placebo (15.5 percent) had a primary out-come event (relative risk, 1.05; 95 percent confidenceinterval, 0.95 to 1.16; P=0.33). There were no signifi-cant differences in the numbers of deaths from car-diovascular causes (342 of those assigned to vita-min E vs. 328 of those assigned to placebo; relativerisk, 1.05; 95 percent confidence interval, 0.90 to 1.22),myocardial infarction (532 vs. 524; relative risk, 1.02;95 percent confidence interval, 0.90 to 1.15), or stroke(209 vs. 180; relative risk, 1.17; 95 percent confidenceinterval, 0.95 to 1.42). There were also no significantdifferences in the incidence of secondary cardiovascu-lar outcomes or in death from any cause. There wereno significant adverse effects of vitamin E.

Conclusions

In patients at high risk for cardiovas-cular events, treatment with vitamin E for a mean of4.5 years has no apparent effect on cardiovascularoutcomes. (N Engl J Med 2000;342:154-60.)

©2000, Massachusetts Medical Society.

Address reprint requests to Dr. Salim Yusuf at the Canadian CardiovascularCollaboration Project Office, Hamilton General Hospital, 237 Barton St.E., Hamilton, ON L8L 2X2, Canada, or at [email protected].

The writing group (Salim Yusuf, D.Phil., Gilles Dagenais, M.D., JanicePogue, M.Sc., Jackie Bosch, M.Sc., and Peter Sleight, D.M.) assumes re-sponsibility for the overall content and integrity of the manuscript.

*The investigators are listed in the Appendix of the Heart Outcomes Pre-vention Evaluation Study Investigators. Effects of an Angiotensin-Convert-ing–Enzyme Inhibitor, Ramipril, on Cardiovascular Events in High-RiskPatients. N Engl J Med 2000;342:145-53.

XIDATIVE modification of low-densitylipoprotein is an important step in the de-velopment and progression of atheroscle-rosis in experimental studies,

1,2

and antiox-idants such as vitamin E have been shown to slowatherosclerosis.

3-5

An inverse relation has been ob-served between coronary heart disease and the con-sumption of fruits, vegetables, and other foodscontaining vitamins, particularly vitamin E.

6-9

Obser-vational studies have indicated that persons who con-

O

sume more than 100 IU of vitamin E a day for morethan two years have lower rates of coronary events

10,11

and lower rates of progression of coronary artery le-sions.

12

However, observational studies cannot distin-guish whether the lower risk of coronary heart diseaseassociated with higher levels of vitamin E consump-tion is due to the vitamin or to other associated life-style factors such as increased exercise and otheraspects of diet. There have been four randomized,controlled trials of the relation between vitamin E andcoronary heart disease,

13-16

but their results are con-flicting, perhaps because of the low doses of vita-min E used in some studies,

13,14

the small numbersof events,

15

or the limited duration of treatment.

15,16

We evaluated a high dose (400 IU per day) of vi-tamin E from natural sources, which has high bio-availability, in a large, five-year, prospective study ofpatients at high risk for cardiovascular events. Theprimary outcome was a composite of myocardial in-farction, stroke, and death from cardiovascular causes.The secondary outcomes included death from anycause, hospitalization for unstable angina or conges-tive heart failure, revascularization or limb amputa-tion, complications of diabetes, and cancer. The trialwas also designed to evaluate the effects of an angio-tensin-converting–enzyme inhibitor, ramipril, on theincidence of cardiovascular events. After nearly 4.5years of follow-up, the collection of data on cardiovas-cular disease was stopped in April 1999 on the basisof a finding by the independent data and safety mon-itoring board that the trial had conclusively demon-strated the benefits of ramipril and a lack of effect ofvitamin E on cardiovascular events. This report pre-sents our findings relating to the effects of vitamin Eon the primary and secondary cardiovascular out-comes. The study has been continued in the majorityof centers to evaluate the effects of vitamin E on theincidence of cancer.

METHODS

Study Design

The Heart Outcomes Prevention Evaluation (HOPE) Study isa double-blind, randomized trial with a two-by-two factorial de-


Volume 342 Number 3

·

155

sign, conducted to evaluate the effects of ramipril and vitamin Ein 9541 patients at high risk for cardiovascular events. The resultsof the comparison of ramipril with placebo are reported in a com-panion article.

17

Details of the methods are given in that article

17

and in a previously published article.

18

Briefly, eligible patients athigh risk were randomly assigned to receive either 400 IU of vi-tamin E from natural sources or an equivalent placebo daily for4 to 6 years (mean, 4.5) and in addition to receive either 10 mg oframipril or an equivalent placebo daily. Patients were evaluatedevery six months for a variety of outcomes.

Outcomes

The primary outcome was a composite of myocardial infarction,stroke, and death from cardiovascular causes. Deaths classified asdue to cardiovascular causes were unexpected deaths presumed tobe due to ischemic cardiovascular disease and occurring within 24hours after the onset of symptoms without clinical or postmor-tem evidence of another cause; deaths from myocardial infarctionor stroke that occurred within seven days after the myocardial in-farction or stroke; and deaths from congestive heart failure, dys-rhythmia, pulmonary embolism, or ruptured abdominal aortic an-eurysm. Deaths for which the cause was uncertain were presumedto be due to cardiovascular disease. Myocardial infarction was diag-nosed when two of the following three criteria were met: typicalsymptoms, increased cardiac enzyme levels (at least twice the up-per limit of normal), and diagnostic electrocardiographic changes.Stroke was defined as a neurologic deficit lasting more than 24hours. A computed tomographic or magnetic resonance imagingexamination was recommended to define the type of stroke.

Secondary and other outcomes were death from any cause; un-stable angina, defined as worsening angina or angina at rest requir-ing hospitalization; hospitalization for heart failure with clinicaland radiologic signs of congestion; revascularization or limb am-putation; the development of overt nephropathy or the need fordialysis or laser therapy among patients with diabetes; and the de-velopment of heart failure or new or worsening angina regardlessof the need for hospitalization.

RESULTS

Characteristics of the Patients

The characteristics of the 9541 patients are shownin Table 1. The rate of compliance with the assignedregimen was high throughout the study. The percent-ages of patients who were taking vitamin E in thevitamin E and placebo groups, respectively, were 94.2percent and 1.0 percent at one year, 93.3 percent and1.7 percent at two years, 91.3 percent and 2.0 per-cent at three years, 90.2 percent and 2.7 percent atfour years, and 89.2 percent and 3.4 percent at thefinal visit.

Primary Cardiovascular Outcomes and Deaths from Any Cause

A total of 772 of the 4761 patients who were as-signed to receive vitamin E (16.2 percent) and 739of the 4780 who were assigned to placebo (15.5 per-cent) had a primary cardiovascular event (relative risk,1.05; 95 percent confidence interval, 0.95 to 1.16;P=0.33) (Table 2 and Fig. 1). There were no signif-icant differences between the groups in the numbersof deaths from cardiovascular causes (342 in the vi-tamin E group vs. 328 in the placebo group; relativerisk, 1.05), myocardial infarctions (532 vs. 524; rel-ative risk, 1.02), deaths from coronary heart disease

(287 vs. 277; relative risk, 1.06), or strokes (209 vs.180; relative risk, 1.17) (Fig. 2 and 3). The total num-bers of deaths were similar in the two groups (535 vs.537; relative risk, 1.00). Vitamin E had no signifi-cant effect on the primary outcome either amongpatients who were receiving ramipril (338 eventsamong those who were receiving vitamin E and 313events among those who were receiving placebo; rel-ative risk, 1.08) or among patients who were not re-ceiving ramipril (421 and 405 events, respectively;relative risk, 1.05).

Secondary Cardiovascular and Combined Outcomes

There were no differences between patients as-signed to vitamin E and those assigned to placebo inthe number of hospitalizations for unstable angina

*Plus–minus values are means ±SD.

†CABG denotes coronary-artery bypass grafting, PTCA percutaneoustransluminal coronary angioplasty, HDL high-density lipoprotein, andECG electrocardiogram.

‡The body-mass index is calculated as the weight in kilograms dividedby the square of the height in meters.

§Peripheral vascular disease included claudication, a history of peripheralarterial disease, or a ratio of blood pressure in the ankle to blood pressurein the arm of less than 0.90.

T

ABLE

1.

B

ASE

-L

INE

C

HARACTERISTICS

OF

THE

P

ATIENTS

.*

C

HARACTERISTIC

†V

ITAMIN

E G

ROUP

(N=4761)P

LACEBO

G

ROUP

(N=4780)

Age — yr 66±7 66±7

Blood pressure — mm Hg 139±20/79±11 139±20/79±11

Heart rate — beats/min 69±11 69±11

Body-mass index‡ 28±4 28±4

Female sex — no. (%) 1263 (26.5) 1282 (26.8)

History of coronary artery disease — no. (%)

3857 (81.0) 3832 (80.2)

Myocardial infarction 2499 (52.5) 2535 (53.0)Stable angina pectoris 2653 (55.7) 2668 (55.8)Unstable angina pectoris 1205 (25.3) 1246 (26.1)CABG 1229 (25.8) 1251 (26.2)PTCA 851 (17.9) 863 (18.1)

Stroke or transient ischemic attacks — no. (%)

530 (11.1) 500 (10.5)

Peripheral vascular disease — no. (%)§ 2109 (44.3) 2037 (42.6)

Hypertension — no. (%) 2219 (46.6) 2222 (46.5)

Diabetes — no. (%) 1838 (38.6) 1816 (38.0)

Known elevated total cholesterol — no. (%)

3109 (65.3) 3171 (66.3)

Known low HDL cholesterol — no. (%)

893 (18.8) 869 (18.2)

Current cigarette smoking — no. (%) 665 (14.0) 679 (14.2)

Medications — no. (%)Beta-blockers 1901 (39.9) 1870 (39.1)Aspirin or other antiplatelet agents 3665 (77.0) 3616 (75.6)Lipid-lowering agents 1352 (28.4) 1401 (29.3)Diuretics 728 (15.3) 717 (15.0)Calcium-channel blockers 2249 (47.2) 2236 (46.8)

Left ventricular hypertrophy on ECG— no. (%)

411 (8.6) 382 (8.0)

Microalbuminuria — no. (%) 1012 (21.3) 976 (20.4)

156

·

Januar y 20, 2000


(586 vs. 569; relative risk, 1.04), hospitalizations forheart failure (160 vs. 144; relative risk, 1.12), or re-vascularizations or limb amputations (848 vs. 787;relative risk, 1.09) (Table 3). There were no signifi-cant differences in the number of patients with anginaof new onset (278 vs. 245; relative risk, 1.15) or mi-crovascular complications of diabetes (340 vs. 325;relative risk, 1.06). A combined analysis of the pro-portion of patients who had any primary or second-ary event found a nonsignificantly higher rate amongthose assigned to vitamin E (1630 vs. 1576; relativerisk, 1.05; 95 percent confidence interval, 0.98 to1.13; P=0.14).

Subgroup Analyses

There was no heterogeneity of results among sub-groups defined according to sex, age, previous car-diovascular disease, or use of other drugs with re-spect to the primary or secondary outcomes (data notshown). Specifically, there was no significant differ-ence in the incidence of the primary outcome amongpatients with diabetes (325 of those assigned to vita-min E vs. 313 of those assigned to placebo; relativerisk, 1.04) or among smokers (135 vs. 139; relativerisk, 1.02).

Adverse Effects

There was no significant difference between groupsin the incidence of adverse effects or in the numberof patients who stopped taking the study medication.There was no increase in hemorrhagic stroke associ-ated with vitamin E use (17 of those assigned to vi-tamin E had hemorrhagic stroke, as compared with13 of those assigned to placebo) or among those whowere also taking an antiplatelet agent (11 vs. 8).

DISCUSSION

In our study, vitamin E did not reduce the inci-dence of cardiovascular events, as compared with theincidence among patients assigned to placebo, dur-ing a follow-up period of four to six years. Given thelarge number of events and the consistent lack of dif-ference in all secondary cardiovascular outcomes, it isvery unlikely that vitamin E had any clinically worth-while beneficial effect on cardiovascular disease dur-ing four or five years of treatment.

Results have been reported from four randomizedtrials of the effects of vitamin E on cardiovascularevents. In a Chinese study, 29,584 adults from LinxianProvince, who did not have cardiovascular disease atentry, were randomly assigned to receive daily vita-

*CI denotes confidence interval.

†P values were calculated with use of the log-rank test.

‡The number of events among those receiving ramipril did not differ significantly between thoseassigned to receive vitamin E and those assigned to placebo (338 vs. 313). Similar results were ob-served among those who received matching placebo rather than ramipril (421 vs. 405).

§A patient may have had more than one event.

T

ABLE

2.

I

NCIDENCE

OF

THE

P

RIMARY

O

UTCOME

AND

OF

D

EATHS

FROM

A

NY

C

AUSE

.

O

UTCOME

V

ITAMIN

EG

ROUP

(N=4761)

P

LACEBO

G

ROUP

(N=4780)R

ELATIVE

R

ISK

(95% CI)* P V

ALUE

†

no. (%)

Myocardial infarction, stroke, or death from cardiovascular causes‡

Death from cardiovascular causes§

Myocardial infarction§

Stroke§

772 (16.2)

342 (7.2)

532 (11.2)

209 (4.4)

739 (15.5)

328 (6.9)

524 (11.0)

180 (3.8)

1.05 (0.95–1.16)

1.05 (0.90–1.22)

1.02 (0.90–1.15)

1.17 (0.95–1.42)

0.33

0.54

0.74

0.13

Death from any cause 535 (11.2) 537 (11.2) 1.00 (0.89–1.13) 0.99

Figure 1.

Kaplan–Meier Estimates of the Effect of Vitamin Eon the Composite Outcome of Nonfatal Myocardial Infarction,Stroke, or Death from Cardiovascular Causes.The relative risk of the composite outcome in the vitamin Egroup as compared with the placebo group was 1.05 (95 percentconfidence interval, 0.95 to 1.16; P=0.33).

0.00

0.20

0

0.05

0.10

0.15

500 1000 1500

Days of Follow-up

Primary Outcome

Vitamin EFPlacebo

Pro

po

rtio

n o

f P

atie

nts


Volume 342 Number 3

·

157

min E (30 mg), beta carotene, and selenium supple-ments or to receive placebo.

13

During the 5.2 years offollow-up, there was a 9 percent decrease in deathsfrom any cause without any significant reduction incardiovascular events. The dose of vitamin E in thisstudy was small, the nutritional status and cardiovas-cular risk of this population were very different fromthose of Western populations, and the beneficial ef-fects on overall mortality cannot be attributed only tovitamin E.

The second trial was the Alpha-Tocopherol, BetaCarotene Cancer Prevention Study, involving 29,133male smokers who were 50 to 69 years of age.

14

Dailytreatment with 50 mg of vitamin E for five to eightyears had no effect on the risk of death from coronaryheart disease. In a subgroup of 1862 men with a pre-

vious myocardial infarction at entry, there was a non-significant increase in the risk of death from coronaryheart disease (relative risk, 1.33; 95 percent confidenceinterval, 0.86 to 2.05; P=0.20). However, a reduc-tion in the risk of nonfatal myocardial infarction wasdocumented among men assigned to vitamin E only(40 vs. 55; relative risk, 0.62; 95 percent confidenceinterval, 0.41 to 0.96), but not among those receiv-ing the combination of vitamin E and beta carotene,in comparison with those receiving placebo only.

19

In this subgroup, the number of events was small. Inthe remaining patients in this study, there was no sig-nificant effect of vitamin E on nonfatal or fatal my-ocardial infarction, despite large numbers of events(1204 and 907, respectively).

20

Thus, in this well-conducted trial, vitamin E had no effect on coronaryheart disease. Although the trial used a low dose ofsynthetic vitamin E (50 mg per day), the median levelof alpha-tocopherol increased significantly, from 28.5µmol per liter at base line to 42.5 µmol per liter atthree months.

The third trial was the Cambridge Heart Antiox-idant Study, which randomly assigned 2002 patientswith coronary atherosclerosis to receive either vita-min E or placebo.

15

The mean alpha-tocopherol levelsincreased from 34.2 to 51.1 µmol per liter in pa-tients receiving 400 IU of vitamin E per day and to64.5 µmol per liter in patients receiving 800 IU perday. The majority of the patients received 400 IU perday. After a median follow-up of 1.4 years, a large re-duction in the number of patients with nonfatal my-ocardial infarction was observed (14 in the vitamin Egroup vs. 41 in the placebo group; relative risk, 0.53;95 percent confidence interval, 0.11 to 0.47; P=0.005), but there was no difference in deaths due to

Figure 2.

Kaplan–Meier Estimates of the Effect of Vitamin E onthe Incidence of Myocardial Infarction (Panel A) and Stroke(Panel B).The relative risk of myocardial infarction in the vitamin E groupas compared with the placebo group was 1.02 (95 percent confi-dence interval, 0.90 to 1.15; P=0.74), and the relative risk of strokewas 1.17 (95 percent confidence interval, 0.95 to 1.42; P=0.13).

0.00

0.15

0

0.05

0.10

500 1000 1500

Days of Follow-up

Myocardial Infarction

Vitamin EFPlacebo

Pro

po

rtio

n o

f P

atie

nts

0.00

0.06

0

0.02

0.04

500 1000 1500

Days of Follow-up

Stroke

Vitamin EFPlacebo

Pro

po

rtio

n o

f P

atie

nts

Figure 3.

Kaplan–Meier Estimates of the Effect of Vitamin E onthe Incidence of Death from Any Cause.The relative risk in the vitamin E group as compared with theplacebo group was 1.00 (95 percent confidence interval, 0.89 to1.13; P=0.99).

0.00

0.15

0

0.05

0.10

500 1000 1500

Days of Follow-up

Death from Any Cause

Vitamin EFPlacebo

Pro

po

rtio

n o

f P

atie

nts

A

B

158

·

Januar y 20, 2000


cardiovascular causes (27 vs. 23; relative risk, 1.18; 95percent confidence interval, 0.62 to 2.27; P=0.61).In this trial, the number of events was small and therewere imbalances in several base-line characteristics thatcall into question whether randomization resulted intruly comparable groups.

Furthermore, the very large reduction in nonfatalmyocardial infarction within a relatively short time(median, 1.4 years) is inconsistent with the results ofother interventions, such as lipid-lowering agents orantihypertensive medications, that reduce cardiovas-cular events. It is therefore likely that the results of theCambridge Heart Antioxidant Study may have beendue to chance. This possibility is supported by theresults of a recent Italian trial,

16

in which 11,000 pa-tients who had had myocardial infarctions were ran-domly assigned to receive 300 IU of vitamin E perday or placebo for a median of 3.5 years. The num-ber of patients with nonfatal myocardial infarctionwas slightly higher in the vitamin E group than the

placebo group (295 vs. 284; relative risk, 1.02; 95 per-cent confidence interval, 0.87 to 1.21), and the num-ber of deaths from coronary heart disease was slightlysmaller (227 vs. 249; relative risk, 0.92; 95 percentconfidence interval, 0.77 to 1.11). Neither differencewas statistically significant.

16

Our study used a high dose of vitamin E (400 IUper day), had high rates of compliance, and involvedhigh-risk patients. The study had a large number ofprimary outcomes and therefore had high statisticalpower (more than 90 percent power to detect a 13percent relative reduction in the risk of the primaryoutcome). Furthermore, a large number of secondaryoutcomes (e.g., revascularization or limb amputation,unstable angina, worsening angina, and heart failure)were examined. Such data are not available from mosttrials. Combining the data from all trials of vitamin Eindicates that such treatment has little effect on therisk of death or cardiovascular events (Table 4), at leastover a four-to-six-year period.

*CI denotes confidence interval.

†P values were calculated with use of the log-rank test.

‡Complications included nephropathy, dialysis, and laser therapy.

T

ABLE

3.

I

NCIDENCE

OF

S

ECONDARY

AND

O

THER

O

UTCOMES

.

O

UTCOME

V

ITAMIN

EG

ROUP

(N=4761)

P

LACEBO

G

ROUP

(N=4780)R

ELATIVE

R

ISK

(95% CI)*P

V

ALUE

†

no. (%)

Revascularization or limb amputation 848 (17.8) 787 (16.5) 1.09 (0.99–1.20) 0.07

Hospitalization for unstable angina 586 (12.3) 569 (11.9) 1.04 (0.93–1.17) 0.52

New-onset angina 278 (5.8) 245 (5.1) 1.15 (0.97–1.37) 0.11

Worsening angina 1215 (25.5) 1186 (24.8) 1.02 (0.94–1.11) 0.63

Claudication 762 (16.0) 753 (15.8) 1.02 (0.92–1.13) 0.70

Hospitalization for heart failure 160 (3.4) 144 (3.0) 1.12 (0.90–1.41) 0.32

Heart failure 530 (11.0) 457 (9.6) 1.17 (1.03–1.32) 0.02

Complications of diabetes‡ 340 (7.1) 325 (6.8) 1.06 (0.91–1.23) 0.47

*CI denotes confidence interval, ATBC Alpha-Tocopherol, Beta Carotene Cancer Prevention Study Group, CHAOSCambridge Heart Antioxidant Study, and GISSI Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico.

†Relative risks and confidence intervals were derived by the method of Yusuf et al.

21

; P=0.27.

T

ABLE

4.

M

ETA

-A

NALYSIS

OF

THE

E

FFECTS

OF

V

ITAMIN

E

ON

M

YOCARDIAL

I

NFARCTION

, S

TROKE

,

OR

D

EATH

FROM

C

ARDIOVASCULAR

C

AUSES

IN LARGE TRIALS.*

STUDY DAILY DOSE DURATION OF STUDY VITAMIN E PLACEBO RELATIVE RISK (95% CI)

mg yr no. with events/total no. (%)

ATBC14 50 5.0 1889/14,564 (13.0) 1970/14,569 (13.5) 0.96 (0.90–1.03)

CHAOS15 »400 1.3 41/1035 (4.0) 64/967 (6.6) 0.60 (0.40–0.89)

GISSI16 300 3.5 571/5660 (10.1) 584/5664 (10.3) 0.98 (0.87–1.10)

Current study 400 4.5 772/4761 (16.2) 739/4780 (15.5) 1.05 (0.95–1.16)

Total 3273/26,020 (12.6) 3357/25,980 (12.9) 0.97 (0.92–1.02)†


Volume 342 Number 3 · 159

Steinberg has hypothesized that unlike agents thatlower cholesterol or blood pressure, antioxidants mayhave to be used for more than five years to have a de-monstrable benefit, since the primary mechanism ofthese agents may be the prevention of new lesions.22

Therefore, in a population like the one we studied, itmay take longer than five years to detect an effect onclinical outcomes. However, the Physicians’ HealthStudy did not find a benefit of beta carotene (anotherantioxidant with a different action) after 12 years.23

Similar data are not available for vitamin E, but ob-servational studies that demonstrated a lower rate ofcoronary heart disease with vitamin E supplementa-tion suggested that a lower risk should be evident af-ter two years.10,11 In a nested substudy, we are examin-ing whether the thickness of the carotid intima andmedia (an indication of the risk of early atheroscle-rosis) can be favorably altered by vitamin E.24 If so,Steinberg’s hypothesis may be worth exploring withmore prolonged follow-up or treatment to assesswhether such changes in the development of athero-sclerosis would translate into a benefit in terms of clin-ical outcomes.

Although the moderate duration of vitamin E sup-plementation (four to six years) and the characteris-tics of the population may explain our finding of alack of benefit of vitamin E, another reason may beour use of vitamin E alone, without other antioxidants.In the epidemiologic studies that found an associa-tion between higher dietary intake of vitamin E andlower rates of coronary heart disease, higher vitamin Econsumption was also associated with higher intakeof a number of other antioxidants and micronutri-ents.6-9 It is possible that vitamin E supplementationrequires these cofactors to have a beneficial effect.25

Although the existence of interactions between vita-min E and other vitamins,10 beta carotene,6,14 orselenium13 is not supported by the findings of pro-spective observational studies or randomized trials,this hypothesis can be tested only in trials in whichcombinations of vitamins are given; some such trialsare now in progress.26-28

In conclusion, 400 IU of vitamin E administereddaily for four to six years had no beneficial effects oncardiovascular outcomes in a high-risk population ofpatients who were 55 years of age or older. Vita-min E was well tolerated, with no significant adverseevents as compared with placebo. This finding pro-vides some reassurance for the conduct of large,longer-term trials to address unanswered questionsregarding vitamin E, such as its possible effects inpreventing cancer.

Funded by the Medical Research Council of Canada, Natural Source Vi-tamin E Association, Negma, Hoechst–Marion Roussel, AstraZeneca, KingPharmaceuticals, and the Heart and Stroke Foundation of Ontario. Dr. Yusufwas supported by a Senior Scientist Award of the Medical Research Coun-cil of Canada and a Heart and Stroke Foundation of Ontario ResearchChair.

We are indebted to W. Whitehill, F. Schutze, N. Bender, B. Ran-goonwala, A. Ljunggren, G. Olsson, J.C. Dairon, J. Ghadiali, B.Carter, J.P. St. Pierre, W. Schulz, M. Jensen, L. Rios-Nogales, M.Bravo, J. Bourgouin, and C. Vint-Reed for support and to KarinDearness for secretarial help.

REFERENCES

1. Steinberg D, Parthasarathy S, Carew TE, Khoo JC, Witztum JL. Be-yond cholesterol: modifications of low-density lipoprotein that increase its atherogenicity. N Engl J Med 1989;320:915-24.2. Witztum JL. The oxidation hypothesis of atherosclerosis. Lancet 1994;344:793-5.3. Carew TE, Schwenke DC, Steinberg D. Antiatherogenic effect of probucol unrelated to its hypercholesterolemic effect: evidence that antiox-idants in vivo can selectively inhibit low density lipoprotein degradation in macrophage-rich fatty streaks and slow the progression of atherosclerosis in the Watanabe heritable hyperlipidemic rabbit. Proc Natl Acad Sci U S A 1987;84:7725-9.4. Verlangieri AJ, Bush MJ. Effects of d-a-tocopherol supplementation on experimentally induced primate atherosclerosis. J Am Coll Nutr 1992;11:131-8.5. Williams RJ, Motteram JM, Sharp CH, Gallagher PJ. Dietary vitamin E and the attenuation of early lesion development in modified Watanabe rabbits. Atherosclerosis 1992;94:153-9.6. Knekt P, Reunanen A, Jarvinen R, Seppanen R, Heliovaara M, Aromaa A. Antioxidant vitamin intake and coronary mortality in a longitudinal population study. Am J Epidemiol 1994;139:1180-9.7. Gaziano JM, Manson JE, Branch LG, Colditz GA, Willett WC, Buring JE. A prospective study of consumption of carotenoids in fruits and vege-tables and decreased cardiovascular mortality in the elderly. Ann Epidemiol 1995;5:255-60.8. Kushi LH, Folsom AR, Prineas RJ, Mink PJ, Wu Y, Bostick RM. Die-tary antioxidant vitamins and death from coronary heart disease in post-menopausal women. N Engl J Med 1996;334:1156-62.9. Klipstein-Grobusch K, Geleijnse JM, den Breeijen JH, et al. Dietary an-tioxidants and risk of myocardial infarction in the elderly: the Rotterdam Study. Am J Clin Nutr 1999;69:261-6.10. Stampfer MJ, Hennekens CH, Manson JE, Colditz GA, Rosner B, Willett WC. Vitamin E consumption and the risk of coronary disease in women. N Engl J Med 1993;328:1444-9.11. Rimm EB, Stampfer MJ, Ascherio A, Giovannucci E, Colditz GA, Willett WC. Vitamin E consumption and the risk of coronary heart disease in men. N Engl J Med 1993;328:1450-6.12. Hodis HN, Mack WJ, LaBree L, et al. Serial coronary angiographic evidence that antioxidant vitamin intake reduces progression of coronary artery atherosclerosis. JAMA 1995;273:1849-54.13. Blot WJ, Li JY, Taylor PR, et al. Nutrition intervention trials in Lin-xian, China: supplementation with specific vitamin/mineral combinations, cancer incidence, and disease-specific mortality in the general population. J Natl Cancer Inst 1993;85:1483-92.14. The Alpha-Tocopherol, Beta Carotene Cancer Prevention Study Group. The effect of vitamin E and beta carotene on the incidence of lung cancer and other cancers in male smokers. N Engl J Med 1994;330:1029-35.15. Stephens NG, Parsons A, Schofield PM, Kelly F, Cheeseman K, Mitchinson MJ. Randomised controlled trial of vitamin E in patients with coronary disease: Cambridge Heart Antioxidant Study. Lancet 1996;347:781-6.16. GISSI-Prevenzione Investigators (Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico). Dietary supplementation with n-3 polyunsaturated fatty acids and vitamin E after myocardial infarction: re-sults of the GISSI-Prevenzione trial. Lancet 1999;354:447-55.17. The Heart Outcomes Prevention Evaluation Study Investigators. Ef-fects of an angiotensin-converting–enzyme inhibitor, ramipril, on cardio-vascular events in high-risk patients. N Engl J Med 2000;342:145-53.18. The HOPE Study Investigators. The HOPE (Heart Outcomes Pre-vention Evaluation) Study: the design of a large, simple randomized trial of an angiotensin-converting enzyme inhibitor (ramipril) and vitamin E in patients at high risk of cardiovascular events. Can J Cardiol 1996;12:127-37.19. Rapola JM, Virtamo J, Ripatti S, et al. Randomised trial of a-tocoph-erol and b-carotene supplements on incidence of major coronary events in men with previous myocardial infarction. Lancet 1997;349:1715-20.20. Virtamo J, Rapola JM, Ripatti S, et al. Effect of vitamin E and beta carotene on the incidence of primary nonfatal myocardial infarction and fatal coronary heart disease. Arch Intern Med 1998;158:668-75.21. Yusuf S, Peto R, Lewis J, Collins R, Sleight P. Beta blockade during and after myocardial infarction: an overview of the randomized trials. Prog Cardiovasc Dis 1985;27:335-71.

160 · Januar y 20, 2000


22. Steinberg D. Clinical trials of antioxidants in atherosclerosis: are we doing the right thing? Lancet 1995;346:36-8.23. Hennekens CH, Buring JE, Manson JE, et al. Lack of effect of long-term supplementation with beta carotene on the incidence of malignant neoplasms and cardiovascular disease. N Engl J Med 1996;334:1145-9.24. Lonn EM, Yusuf S, Doris CI, et al. Study design and baseline charac-teristics of the Study to Evaluate Carotid Ultrasound Changes in Patients Treated with Ramipril and Vitamin E: SECURE. Am J Cardiol 1996;78:914-9.25. Upston JM, Terentis AC, Stocker R. Tocopherol-mediated peroxida-tion of lipoproteins: implications for vitamin E as a potential antiathero-genic supplement. FASEB J 1999;13:977-94.

26. Hercberg S, Galan P, Preziosi P, et al. Background and rationale be-hind the SU.VI.MAX Study, a prevention trial using nutritional doses of a combination of antioxidant and minerals to reduce cardiovascular diseases and cancers: SUpplementation en VItamines et Mineraux AntioXydants Study. Int J Vitam Nutr Res 1998;68:3-20.27. MRC/BHF Heart Protection Study of cholesterol-lowering therapy and of antioxidant vitamin supplementation in a wide range of patients at increased risk of coronary heart disease death: early safety and efficacy ex-perience. Eur Heart J 1999;20:725-41.28. Jha P, Flather M, Lonn E, Farkouh M, Yusuf S. The antioxidant vita-mins and cardiovascular disease: a critical review of epidemiologic and clin-ical trial data. Ann Intern Med 1995;123:860-72.

n engl j med

352;11

www.nejm.org march

17, 2005

1071

The

new england

journal

of

medicine

established in 1812

march

17

,

2005

vol. 352 no. 11

Cardiovascular Risk Associated with Celecoxib in a Clinical Trial for Colorectal Adenoma Prevention

Scott D. Solomon, M.D., John J.V. McMurray, M.D., Marc A. Pfeffer, M.D., Ph.D., Janet Wittes, Ph.D., Robert Fowler, M.S., Peter Finn, M.D., William F. Anderson, M.D., M.P.H., Ann Zauber, Ph.D.,

Ernest Hawk, M.D., M.P.H., and Monica Bertagnolli, M.D., for the Adenoma Prevention with Celecoxib (APC) Study Investigators*

abstract

From the Cardiovascular Division, Depart-ments of Medicine (S.D.S., M.A.P., P.F.)and Surgery (M.B.), Brigham and Women’sHospital, Harvard Medical School, Boston;Western Infirmary, University of Glasgow,Glasgow, Scotland (J.J.V.M.); Statistics Col-laborative, Washington, D.C. (J.W., R.F.);National Cancer Institute, Bethesda, Md.(W.F.A., E.H.); and Memorial Sloan-Ketter-ing Cancer Center, New York (A.Z.). Addressreprint requests to Dr. Solomon at the Car-diovascular Division, Brigham and Women’sHospital, 75 Francis St., Boston, MA 02115,or at [email protected].

*Participants in the APC study are listedin the Appendix.

This article was published at www.nejm.org on February 15, 2005.

N Engl J Med 2005;352:1071-80.

Copyright © 2005 Massachusetts Medical Society.

background

Selective cyclooxygenase-2 (COX-2) inhibitors have come under scrutiny because of re-ports suggesting an increased cardiovascular risk associated with their use. Experimen-tal research suggesting that these drugs may contribute to a prothrombotic state pro-vides support for this concern.

methods

We reviewed all potentially serious cardiovascular events among 2035 patients with ahistory of colorectal neoplasia who were enrolled in a trial comparing two doses ofcelecoxib (200 mg or 400 mg twice daily) with placebo for the prevention of colorectaladenomas. All deaths were categorized as cardiovascular or noncardiovascular, andnonfatal cardiovascular events were categorized in a blinded fashion according to aprespecified scheme.

results

For all patients except those who died, 2.8 to 3.1 years of follow-up data were available.A composite cardiovascular end point of death from cardiovascular causes, myocardialinfarction, stroke, or heart failure was reached in 7 of 679 patients in the placebo group(1.0 percent), as compared with 16 of 685 patients receiving 200 mg of celecoxib twicedaily (2.3 percent; hazard ratio, 2.3; 95 percent confidence interval, 0.9 to 5.5) and with23 of 671 patients receiving 400 mg of celecoxib twice daily (3.4 percent; hazard ratio,3.4; 95 percent confidence interval, 1.4 to 7.8). Similar trends were observed for othercomposite end points. On the basis of these observations, the data and safety monitor-ing board recommended early discontinuation of the study drug.

conclusions

Celecoxib use was associated with a dose-related increase in the composite end pointof death from cardiovascular causes, myocardial infarction, stroke, or heart failure. Inlight of recent reports of cardiovascular harm associated with treatment with otheragents in this class, these data provide further evidence that the use of COX-2 inhibitorsmay increase the risk of serious cardiovascular events.

Copyright © 2005 Massachusetts Medical Society. All rights reserved. Downloaded from www.nejm.org on March 18, 2005 . This article is being provided free of charge for use in Philippines.

n engl j med

352;11

www.nejm.org march

17

,

2005

The

new england journal

of

medicine

1072

he promise of a lower incidence

of gastrointestinal side effects with the useof selective cyclooxygenase-2 (COX-2) in-

hibitors than with the use of nonselective nonste-roidal antiinflammatory drugs (NSAIDs) or aspirinhas led to a marked increase in prescriptions forCOX-2 inhibitors, despite the fact that they offersimilar degrees of pain relief.

1-3

In addition, theidentification of COX-2 as a promoter of intestinaltumorigenesis suggested that inhibiting this en-zyme could prevent the formation of premalignantcolorectal adenomas.

4-8

Recently, however, this classof drugs has come under scrutiny because of clini-cal reports that they were associated with an in-creased risk of serious cardiovascular harm.

9-11

Themechanism of this effect is suggested in part by ev-idence that selective inhibition of COX-2 can blockthe production of prostacyclin without affectingthe synthesis of thromboxane A

2

,

10

thereby poten-tially creating a prothrombotic state.

The observation of an increased incidence ofdeath from cardiovascular causes, myocardial in-farction, or stroke among patients receiving rofe-coxib in the Adenomatous Polyp Prevention onVioxx (APPROVe) trial and the associated voluntarywithdrawal of this drug from the market promptedthe data and safety monitoring board and steeringcommittee of a similar ongoing trial of celecoxib torequest a focused reassessment of data on cardio-vascular safety by an independent committee, withthe results presented at their scheduled meeting onDecember 10, 2004. The study was a prospective,randomized, double-blind, multicenter trial assess-ing the efficacy of celecoxib for the prevention ofadenomatous polyps in patients who had under-gone endoscopic polypectomy. Because neitherprior clinical trials nor observational studies hadreported a clearly increased risk of cardiovascularevents with celecoxib use,

2,5,12-16

this longer-term,placebo-controlled trial provided an important op-portunity to evaluate the potential association. Thisreport describes the findings of the independentcardiovascular safety committee.

patients

The Adenoma Prevention with Celecoxib (APC)study compared the efficacy and safety of 200 mg ofcelecoxib twice daily, 400 mg of celecoxib twicedaily, and placebo in reducing the occurrence of ad-

enomatous polyps in the colon and rectum one yearand three years after endoscopic polypectomy. Thetrial was led by the Strang Cancer Prevention Cen-ter (New York) and cosponsored by the NationalCancer Institute and Pfizer. Ninety-one sites par-ticipated (72 in the United States, 1 in the UnitedKingdom, 8 in Australia, and 10 in Canada). Partici-pants ranged from 32 to 88 years of age and wereconsidered to have a clinically significant risk ofcolorectal adenoma on the basis of a history of ei-ther multiple adenomas or a single adenoma thatwas at least 0.5 cm in diameter. All known adeno-mas were removed colonoscopically before drugtreatment began.

A detailed medical history, including baseline as-sessment of cardiovascular disease status and riskfactors for cardiovascular disease, was obtained foreach patient. The protocol was reviewed and ap-proved by the appropriate institutional reviewboards, and all patients provided written informedconsent before enrollment. Patients were randomlyassigned to treatment with the use of a computer-generated randomization schedule. At the time ofdata review, 2035 patients had undergone random-ization in a double-blind manner at a 1:1:1 ratio,after stratification according to the use or nonuse ofaspirin for cardiovascular prophylaxis and the en-rolling center. Enrollment began in November 1999and concluded in March 2002. Compliance was as-sessed by means of both pill counts and standardmonitoring of medical records every 6 to 12 weeks.

review of cardiovascular safety

The cardiovascular safety committee developed end-point definitions as guidelines for adjudication. Thecommittee classified and adjudicated the end pointsby defining a hierarchy of composite end points(based on clinical importance and the prior findingswith rofecoxib). These guidelines were designedspecifically to assess cardiovascular safety (listedin the Supplementary Appendix, available with thefull text of this article at www.nejm.org). An initialreview identified all deaths and potential nonfatalcardiovascular adverse events. Two experienced in-dependent assessors reviewed these events usingmedical records and narratives supplied by site in-vestigators. Myocardial infarction was defined onthe basis of either a clinical presentation character-ized by typical symptoms, signs, or electrocardio-graphic changes associated with an elevation inthe level of a cardiac marker or angiographic evi-

t

methods


n engl j med

352;11

www.nejm.org march

17, 2005

celecoxib and cardiovascular risk

1073

dence of coronary thrombosis. Stroke was definedas a persistent focal neurologic event whose onsetwas sudden and was not due to trauma or a tumor.Other cardiovascular events were categorized ac-cording to a preplanned schema. When this initialdocumentation was insufficient for adjudication,additional information was obtained from the in-vestigative sites.

The entire cardiovascular safety committee wasunaware of the patients’ treatment assignmentsthroughout the review process. For the purposes ofthis analysis, we evaluated a hierarchy of compos-ite end points, including death from cardiovascularcauses, myocardial infarction, stroke, heart failure,unstable angina, and the need for a cardiovascularprocedure.

statistical analysis

Randomization codes were provided to StatisticsCollaborative (Washington, D.C.). All analyses wereperformed according to the intention-to-treat prin-ciple, with data on each patient analyzed accordingto the original randomized treatment assignment.Log-rank tests were used to compare the time to acardiovascular event in the three groups for eachcomposite end point of interest. Cox models, withthe treatment group as the only covariate, wereused to estimate hazard ratios for the two celecoxibgroups as compared with the placebo group. Al-though the randomization was stratified accordingto the baseline use or nonuse of aspirin and the cen-ter, the Cox models did not include these stratifyingvariables. Censoring was defined by assuming thata patient was followed for 37 months, until death,or until January 6, 2005 (the date defined for thisanalysis as the common close-out date) — which-ever came first. At the time of this review, we hadfollow-up information for more than 97 percent ofthe patient-years at risk. Incidence rates were cal-culated for individual and composite cardiovascu-lar events by dividing the number of patients withevents by the number of patient-years at risk.

Important subgroups based on baseline charac-teristics were prespecified. To examine whether theeffect of celecoxib varied between subgroups, weconstructed Cox models with terms for treatment,subgroup, and the interaction between subgroupand treatment and evaluated the interaction termsfor statistical significance.

Recommendations to the study’s data and safetymonitoring board were made on the basis of data

available at the time of the original analysis. Thisanalysis contains data on three additional cardio-vascular events that were not included in the origi-nal report.

At the time of the analysis, 77 percent of the 2035patients had completed the study, and all of the re-maining surviving patients had completed at least2.8 years of follow-up (range, 2.8 to 3.1). The base-line characteristics were similar among the threegroups (Table 1). The incidence of the prespecifiedcomposite cardiovascular end points, analyzed ac-cording to the time to the first event, and the asso-ciated hazard ratios are shown in Table 2. As com-pared with the placebo group, the group given 200mg of celecoxib twice daily had a hazard ratio fordeath from cardiovascular causes, myocardial in-farction, stroke, or heart failure of 2.3 (95 percentconfidence interval, 0.9 to 5.5), and the group re-ceiving 400 mg of celecoxib twice daily had a hazardratio of 3.4 (95 percent confidence interval, 1.4 to7.8). The results for the individual components ofthe composite end point are shown in Table 3.

There were six deaths in the placebo group, six inthe group given 200 mg of celecoxib twice daily, andnine in the group given 400 mg twice daily, and one,three, and six of the deaths, respectively, were dueto cardiovascular causes. The Kaplan–Meier curvesfor the combined end point of death from cardio-vascular causes, myocardial infarction, stroke, orheart failure in the three groups are shown in Fig-ure 1. The annualized incidence of death from car-diovascular causes, stroke, myocardial infarction, orheart failure was 3.4 events per 1000 patient-yearsin the placebo group, 7.8 events per 1000 patient-years in the group given 200 mg of celecoxib twicedaily, and 11.4 events per 1000 patient-years in thegroup given 400 mg twice daily.

In addition to the increased risk of the prespec-ified composite end point of cardiovascular events,the point estimate of the number of venous throm-boembolic events was also increased (though notsignificantly) among patients receiving celecoxib:four in the group given 400 mg of celecoxib twicedaily and three in the group given 200 mg twice dai-ly, as compared with one in the placebo group (haz-ard ratio for the two celecoxib groups combined,3.5; 95 percent confidence interval, 0.4 to 28.5).There was no apparent increase in the risk of un-

results


n engl j med

352;11

www.nejm.org march

17

,

2005

The

new england journal

of

medicine

1074

stable angina, arrhythmia, or the need for a cardio-vascular procedure. The hazard ratios associatedwith celecoxib use decreased when a broader classof cardiovascular events, including unstable angi-na and the need for a cardiovascular procedure,was added to the composite end point. The hazardratio associated with celecoxib was not significant-ly affected by any of the baseline characteristics ex-amined, including aspirin use at baseline (Table 4).

On December 16, 2004, on the basis of thesefindings, the advice of the cardiovascular safetycommittee, and previous findings with drugs in thesame class, the data and safety monitoring boardconcluded that continued exposure to celecoxibplaced patients at increased risk for serious cardio-vascular events. On the basis of this recommenda-tion, the steering committee stopped the use ofstudy medication among the patients remaining inthe trial. The trial remained blinded, and follow-upfor the end point of adenoma continued. Threeevents that were documented after the study wasstopped are included in the present analysis; theirinclusion does not alter the overall conclusions ofthe report issued on December 16, 2004.

In a large, randomized, placebo-controlled, double-blind, multicenter trial, we found that twice-daily

treatment with 200 or 400 mg of celecoxib to pre-vent colorectal adenomas led to a dose-related in-crease in the risk of serious cardiovascular events,including death from cardiovascular causes, myo-cardial infarction, stroke, and heart failure. Theseresults were consistent among the individual com-ponents of the composite end point. Because theuse of other selective COX-2 inhibitors, includingrofecoxib, valdecoxib, and parecoxib, has also beenassociated with an increased rate of cardiovascularevents,

17,18

our results heighten concern that thisclass of drug may be associated with increased car-diovascular risk. The cardiovascular safety commit-tee also completed a preliminary review of cardio-vascular safety in another study, the Prevention ofSpontaneous Adenomatous Polyps (PreSAP) trial,which randomly assigned patients with a history ofcolorectal adenomas to receive either 400 mg of cele-coxib once a day or placebo. The preliminary analy-sis did not show an increase in risk at this dose.

The reason for the apparent increase in cardio-vascular risk associated with the use of COX-2 in-hibitors is uncertain. One prominent hypothesisinvolves the effects of COX-2 inhibitors on twokey prostanoids, prostacyclin and thromboxaneA

2

, which have a crucial role in vascular homeo-stasis.

9,19,20

These prostanoids are generated bythe action of the cyclooxygenase-1 (COX-1) andCOX-2 isoenzymes on arachidonic acid.

21

Throm-

discussion

* Plus–minus values are means ±SD. There were no significant differences among the groups.

† Data were missing for one patient in the placebo group.

Table 1. Baseline Characteristics of the Patients.*

CharacteristicPlacebo(N=679)

Celecoxib, 200 mgTwice Daily (N=685)

Celecoxib, 400 mg Twice Daily (N=671)

Age — yr 59.7±9.7 59.7±9.4 59.9±9.4

Male sex — no. (%) 473 (69.7) 460 (67.2) 454 (67.7)

History of cardiovascular events — no. (%) 321 (47.3) 335 (48.9) 307 (45.8)

Myocardial infarction 29 (4.3) 22 (3.2) 31 (4.6)

Cerebrovascular disease 14 (2.1) 20 (2.9) 13 (1.9)

Congestive heart failure 14 (2.1) 6 (0.9) 11 (1.6)

Angina 51 (7.5) 50 (7.3) 42 (6.3)

Hypertension 277 (40.8) 287 (41.9) 260 (38.7)

Diabetes — no. (%)† 61 (9.0) 66 (9.6) 64 (9.5)

Current smoker — no. (%) 122 (18.0) 119 (17.4) 96 (14.3)

Aspirin use — no. (%) 213 (31.4) 201 (29.3) 200 (29.8)

Use of lipid-lowering drug — no. (%) 184 (27.1) 188 (27.4) 191 (28.5)


n engl j med

352;11

www.nejm.org march

17, 2005


1075

*M

I den

otes

myo

card

ial i

nfar

ctio

n.

Tabl

e 2.

Inci

denc

e of

and

Haz

ard

Rat

ios

for

the

Com

posi

te E

nd P

oint

s in

the

Cel

ecox

ib G

roup

s R

elat

ive

to th

e Pl

aceb

o G

roup

.

End

Poin

t*Pl

aceb

o (N

=679

)

Cel

ecox

ib,

200

mg

Twic

e D

aily

(N

=68

5)

Cel

ecox

ib,

400

mg

Twic

e D

aily

(N

=67

1)

Bot

h C

elec

oxib

G

roup

s (N

=13

56)

Plac

ebo

(N=

679)

Cel

ecox

ib,

200

mg

Twic

e D

aily

(N

=68

5)

Cel

ecox

ib,

400

mg

Twic

e D

aily

(N

=67

1)

Bot

h C

elec

oxib

G

roup

s (N

=13

56)

Cel

ecox

ib,

200

mg

Twic

e D

aily

(N

=68

5)

Cel

ecox

ib,

400

mg

Twic

e D

aily

(N

=67

1)

Bot

h C

elec

oxib

G

roup

s (N

=13

56)

num

ber o

f pat

ient

s (p

erce

nt)

rate

/100

0 pa

tient

-yea

rsha

zard

ratio

(95

% c

onfid

ence

inte

rval

)

Dea

th fr

om c

ardi

ovas

-cu

lar

caus

es1

(0.1

)3

(0.4

)6

(0.9

)9

(0.7

)0.

51.

42.

92.

23.

0 (0

.3–2

8.6)

6.1

(0.7

–50.

3)4.

5 (0

.6–3

5.5)

Dea

th fr

om c

ardi

ovas

-cu

lar c

ause

s or

non

-fa

tal M

I

4 (0

.6)

12 (

1.8)

15 (

2.2)

27 (

2.0)

1.9

5.8

7.4

6.6

3.0

(1.0

–9.3

)3.

8 (1

.3–1

1.5)

3.4

(1.2

–9.7

)

Dea

th fr

om c

ardi

ovas

-cu

lar

caus

es, n

on-

fata

l MI,

or s

trok

e

6 (0

.9)

15 (

2.2)

20 (

3.0)

35 (

2.6)

2.9

7.3

9.9

8.6

2.5

(1.0

–6.4

)3.

4 (1

.4–8

.5)

2.9

(1.2

–7.0

)

Dea

th fr

om c

ardi

ovas

-cu

lar

caus

es, n

on-

fata

l MI,

stro

ke,

or h

eart

failu

re

7 (1

.0)

16 (

2.3)

23 (

3.4)

39 (

2.9)

3.4

7.8

11.4

9.6

2.3

(0.9

–5.5

)3.

4 (1

.4–7

.8)

2.8

(1.3

–6.3

)

Dea

th fr

om c

ardi

ovas

-cu

lar

caus

es, n

on-

fata

l MI,

stro

ke, h

eart

fa

ilure

, or

angi

na

11 (

1.6)

18 (

2.6)

25 (

3.7)

43 (

3.2)

5.4

8.7

12.5

10. 6

1.6

(0.8

–3.4

)2.

3 (1

.1–4

.7)

2.0

(1.0

–3.8

)

Dea

th fr

om c

ardi

ovas

-cu

lar

caus

es, n

on-

fata

l MI,

stro

ke, h

eart

fa

ilure

, or

angi

na

or n

eed

for

a ca

rdio

-va

scul

ar p

roce

dure

17 (

2.5)

26 (

3.8)

31 (

4.6)

57 (

4.2)

8.4

12.7

15.5

14.1

1.5

(0.8

–2.8

)1.

9 (1

.0–3

.3)

1.7

(1.0

–2.9

)


n engl j med

352;11

www.nejm.org march

17

,

2005

The

new england journal

of

medicine

1076

boxane A

2

, which promotes platelet aggregation,vasoconstriction, and smooth-muscle proliferation,is synthesized primarily in platelets, which expressonly COX-1. Conversely, prostacyclin, which has an-tiaggregative, antiproliferative, and vasodilatory ac-tions, is the main prostanoid product of endothelialcells, synthesized as a result of the action of COX-2.

22

Whereas nonselective NSAIDs inhibit both

COX-1 and COX-2, selective COX-2 inhibitors actprimarily on COX-2.

9

The selective COX-2 inhibi-tors may therefore suppress vascular productionof prostacyclin without affecting the synthesis ofplatelet-derived thromboxane A

2

. This imbalancemay promote thrombosis and increase the risk ofcardiovascular events.

10

Nonaspirin, nonselectiveNSAIDs may also not sufficiently reduce throm-boxane A

2

synthesis long enough to prevent plate-let aggregation and atherosclerotic events.

10

Otherpotentially detrimental effects of COX-2 inhibi-tors have been suggested, including elevated bloodpressure, though some reports have indicated thatthese drugs may have beneficial effects on vascu-lar health.

23

In contrast to our findings, most of the earlierclinical trials of selective COX-2 inhibitors in pa-tients with arthritis did not appear to show an in-crease in cardiovascular risk.

2,5,14,24

These trials,

however, were generally short-term studies designedto assess the use of this class of drug for pain reliefand to evaluate associated adverse gastrointestinalevents. They included a relatively small proportionof patients at high risk for cardiovascular events orexcluded such patients, despite the fact that manypatients who are taking these drugs or who are con-sidered candidates for this therapy are at high car-diovascular risk.

25

Consequently, the studies lackedadequate statistical power to confirm or refute a car-diovascular hazard related to the use of COX-2 in-hibitors.

11

The use of active rather than placebo con-trols in many of these studies also made the findingsdifficult to interpret.

The results of the Vioxx Gastrointestinal Out-comes

Research (VIGOR) trial

3

and a subsequentstudy, APPROVe,

26

raised questions about the safetyof rofecoxib. The VIGOR trial, which compared anine-month course of 50 mg of rofecoxib per day(a larger dose than that usually recommended forthe long-term treatment of arthritis) with naproxenin patients with rheumatoid arthritis, reported ahigher risk of myocardial infarction among the pa-tients receiving rofecoxib.

27

Some have attributedthese findings to the potentially cardioprotective ef-fects of naproxen,

28,29

although this interpretationhas been a source of contention.

18,20

Table 3. Incidence of Individual Cardiovascular and Fatal Events.

End PointPlacebo (N=679)



Both Celecoxib Groups (N=1356)

number of patients (percent)

Death from any cause 6 (0.9) 6 (0.9) 9 (1.3) 15 (1.1)

Death from cardiovascular causes 1 (0.1) 3 (0.4) 6 (0.9) 9 (0.7)

Death from noncardiovascular causes 5 (0.7) 3 (0.4) 3 (0.4) 6 (0.4)

Nonfatal cardiovascular events

Myocardial infarction 3 (0.4) 9 (1.3) 9 (1.3) 18 (1.3)

Stroke 3 (0.4) 3 (0.4) 5 (0.7) 8 (0.6)

Heart failure 2 (0.3) 1 (0.1) 4 (0.6) 5 (0.4)

Thromboembolic event 1 (0.1) 3 (0.4) 4 (0.6) 7 (0.5)

Resuscitation after sudden cardiac arrest

0 0 1 (0.1) 1 (0.1)

Hospitalization for unstable angina

5 (0.7) 4 (0.6) 2 (0.3) 6 (0.4)

Arrhythmia 9 (1.3) 4 (0.6) 7 (1.0) 11 (0.8)

Cardiovascular procedure 7 (1.0) 9 (1.3) 6 (0.9) 15 (1.1)

Other 9 (1.3) 11 (1.6) 14 (2.1) 25 (1.8)


n engl j med

352;11

www.nejm.org march

17, 2005


1077

More recently, the APPROVe trial, a randomized,placebo-controlled trial designed to evaluate the ef-ficacy of rofecoxib for preventing colorectal polypsin patients with a history of colorectal adenomas,was terminated early because of an increased riskof cardiovascular events.

10,26

These results prompt-ed voluntary withdrawal of rofecoxib from the mar-ket. Topol reported that another controlled trial alsoshowed an increased risk of cardiovascular eventswith treatment with 12.5 mg of rofecoxib per day,as compared with nabumetone or placebo.

30

The results of other studies have aroused con-cern about the safety of selective COX-2 inhibitors.In a placebo-controlled trial of pain relief after cor-onary-artery bypass surgery, the use of the paren-teral COX-2 inhibitor parecoxib followed by oraltreatment with its active metabolite valdecoxib, ortreatment with placebo followed by valdecoxib, wasassociated with a significantly increased risk of car-diovascular thromboembolic events.

31

In this issueof the

Journal,

Nussmeier et al. report a second trialshowing a significant increase in cardiovascularevents when parecoxib and valdecoxib were used inthe immediate postoperative period after coronary-artery bypass surgery.

32

The Therapeutic ArthritisResearch and Gastrointestinal Event Trial (TARGET)also showed a nonsignificant increase in the risk ofcardiovascular events with lumiracoxib therapy,

10

ascompared with naproxen or ibuprofen therapy, butonly among patients who were not taking aspirin.

In contrast, to our knowledge, neither phar-macoepidemiologic studies nor randomized, con-trolled trials have reported clear evidence of anincreased cardiovascular risk associated with cele-coxib. The failure of pharmacoepidemiologic stud-ies to show an increased risk may be due in part tothe lower doses and shorter duration of use in thesestudies than in clinical trials and in part to the po-tential for selection bias in nonrandomized studies.Nevertheless, the Celecoxib in Long-term ArthritisSafety Study (CLASS),

2

which used the same doseof celecoxib (400 mg twice daily) that was given toone group in the APC study and compared celecoxibwith two nonselective NSAIDs, did not show an in-creased rate of cardiovascular events.

2

CLASS dif-fered from the VIGOR study in several importantways. A short-term study not designed for system-atic and formal assessment of cardiovascularevents, CLASS enrolled relatively low-risk patientsand allowed the use of aspirin for cardiovascularprotection. In addition, FitzGerald has suggestedthat CLASS did not completely refute evidence of

an increased cardiac risk associated with celecoxibuse in non–aspirin users, as compared with thosetaking ibuprofen (but not diclofenac).

20

Moreover,the results of a randomized, controlled clinical trialof celecoxib in patients with Alzheimer’s disease,reported to the Food and Drug Administration,demonstrated an increase in cardiovascular eventsamong patients receiving celecoxib.

33

Although we found that patients with an in-creased cardiovascular risk at baseline appeared tohave a higher absolute rate of events than those withno increase in cardiovascular risk at baseline, for-mal statistical tests of interaction showed no dif-ferential effect of celecoxib with respect to baselinecardiovascular risk. One prespecified subgroup in-cluded users of cardioprotective aspirin at base-line. Although the overall absolute risk appeared tobe higher among such patients, analysis of the dataon aspirin users in this study shows that they had ahigher frequency of cardiovascular risk factors atbaseline than did nonusers.

The cardiovascular findings with regard to cele-coxib use in the APC study are consistent with thoseidentified for rofecoxib use in the APPROVe trial. Incontrast, preliminary analyses from the PreSAP trial,which involved a daily dose of 400 mg of celecoxib,showed no apparent increase in cardiovascular risk.The differences in the dosing regimens betweenthese two trials — twice daily in the APC study, as

Figure 1. Kaplan–Meier Estimates of the Risk of the Composite End Point of Death from Cardiovascular Causes, Myocardial Infarction, Stroke, or Heart Failure among Patients Who Received Celecoxib (200 mg Twice Daily or 400 mg Twice Daily) or Placebo.

The log-rank statistic of 8.73, which has two degrees of freedom, was used to determine the P value.

Estim

ated

Pro

babi

lity

of C

ompo

site

End

Poi

nt

0.0500.0450.040

0.0300.025

0.0100.015

0.005

0.035

0.020

0.0000 6 12 18 24 30 36

Months after First Dose

No. at RiskCelecoxib, 400 mg Celecoxib, 200 mg Placebo

648 670 667

576 595 585

651 673 668

665 676 675

655 675 672

669 681 677

671685679

P=0.01

Placebo

Celecoxib, 400 mg

Celecoxib, 200 mg


n engl j med

352;11

www.nejm.org march

17

,

2005

The

new england journal

of

medicine

1078

compared with once daily in the PreSAP study —support the hypothesis that sustained inhibition ofprostacyclin may contribute to the increase in car-diovascular risk. Other potential differences in thetrials, including geographic differences, differencesin the patient population, and differences in use ofconcomitant medications, may have contributed tothe disparity in the preliminary findings.

The increased cardiovascular risk in the APC trialwas based on a small number of events in a trial thatwas not designed or statistically powered to evalu-ate cardiovascular risk. Although we believe we haveidentified all adverse cardiovascular events, we can-not rule out the possibility that some events re-mained unreported. Our results must therefore beinterpreted with caution.

Still, in the context of the results of the other tri-als reviewed involving agents in the same class,these data suggest that there may be a real increasein cardiovascular risk associated with the use of cele-coxib in particular and the class of selective COX-2inhibitors in general. If correct, this interpretation

has substantial implications for public health,

11,34

patient education,

35

and drug regulation.

36,37

Giventhe experience with COX-2 inhibitors, we supportthe call for regulatory agencies to consider request-ing a formal evaluation of long-term cardiovascu-lar outcomes of any new drug with a mechanism ofaction that could augment the risk of cardiac andvascular events, especially if many patients who arelikely to use the new agent are prone to cardiovas-cular disease.

25

This category may include nonse-lective NSAIDs (other than aspirin), as discussedearlier. More broadly, this experience underscoresboth the need for long-term, placebo-controlled tri-als to assess safety as well as efficacy and the needto improve methods for assessing potential adversecardiovascular outcomes in studies with noncar-diovascular primary end points.

In summary, a blinded review of cardiovascularevents in a large, randomized, controlled study oftwo doses of celecoxib for the prevention of colorec-tal adenomas showed a dose-related risk of suchevents, including death from cardiovascular causes,

* CI denotes confidence interval.

Table 4. Incidence of Death from Cardiovascular Causes, Myocardial Infarction, Stroke, or Heart Failure According to Baseline Characteristics.

SubgroupNo. of

Patients PlaceboBoth Celecoxib

GroupsHazard Ratio

(95% CI)*P Value for Interaction

no./total no. (%)

Age 0.61

<60 yr 1078 4/372 (1.1) 14/706 (2.0) 1.8 (0.6–5.6)

≥60 yr 957 3/307 (1.0) 25/650 (3.8) 4.0 (1.2–13.2)

Sex 0.55

Female 648 2/206 (1.0) 8/442 (1.8) 1.9 (0.4–8.7)

Male 1387 5/473 (1.1) 31/914 (3.4) 3.2 (1.3–8.3)

Baseline cardiovascular risk factors 0.44

Yes 963 4/321 (1.2) 28/642 (4.4) 3.5 (1.2–10.1)

No 1072 3/358 (0.8) 11/714 (1.5) 1.8 (0.5–6.6)

Diabetes 0.86

Yes 191 1/61 (1.6) 5/130 (3.8) 2.3 (0.3–19.9)

No 1843 6/617 (1.0) 34/1226 (2.8) 2.9 (1.2–6.8)

Aspirin use 0.63

Yes 614 2/213 (0.9) 14/401 (3.5) 3.8 (0.9–16.6)

No 1421 5/466 (1.1) 25/955 (2.6) 2.4 (0.9–6.4)

Use of lipid-lowering drug 0.79

Yes 563 3/184 (1.6) 15/379 (4.0) 2.4 (0.7–8.4)

No 1472 4/495 (0.8) 24/977 (2.5) 3.1 (1.1–8.8)


n engl j med

352;11

www.nejm.org march

17, 2005


1079

myocardial infarction, stroke, and heart failure. Inlight of other recent reports of the adverse cardio-vascular effects of other agents in this class, thesedata provide further evidence that long-term use ofCOX-2 inhibitors may increase the risk of seriouscardiovascular events. These risks will need to beweighed against any potential benefits of celecox-

ib in preventing colorectal neoplasia and in reliev-ing pain.

The APC was sponsored by the National Cancer Institute and co-sponsored by Pfizer. This cardiovascular review was funded solely bythe National Cancer Institute.

Drs. McMurray, Pfeffer, and Zauber report having received con-sulting fees from Pfizer. Drs. Solomon, McMurray, and Pfeffer re-port having received lecture fees from Pfizer. Dr. Wittes reports hav-ing received consulting fees from Merck within the past two years.

appendix

The following persons participated in the APC Study:

Steering Committee:

M.M. Bertagnolli, E. Hawk, C. Eagle;

Statistical Team:

A. Zaub-er, K.M. Kim, D. Corle, R. Rosenstein, J. Tang, T. Hess, A. Wilton;

Medical Monitors:

W. Anderson, L. Doody;

Central Pathology Review:

M.Redston;

Project Directors

: M. Woloj, D. Bagheri, A. Crawford, M. Schietrum, V. Ladouceur;

Data and Safety Monitoring Board

: S. Rosen(chair), L. Friedman, R. Makuch, R. Phillips, P. Taylor;

Principal Investigators: United States:

S. Auerbach (California Professional Re-search, Newport Beach), C.F. Barish (Wake Research Associates, Raleigh, N.C.), T. Barringer (Carolinas Medical Center, Charlotte, N.C.),R.W. Bennetts (Northwest Gastroenterology Clinic, Portland, Oreg.), M. Blitstein (Associates in Gastroenterology and Liver Disease, LakeForest, Ill.), J. Bruggen (Wake Forest University Baptist Medical Center, Winston-Salem, N.C.), P. Carricaburu (Veterans Affairs Hospital,Sheridan, Wyo.), D. Chung (Massachusetts General Hospital, Boston), F. Colizzo (Pentucket Medical Associates, Haverhill, Mass.), R. Cur-tis (Newton–Wellesley Hospital, Newton, Mass.), T. Dewar (Harris Methodist Hospital Fort Worth, Ft. Worth, Tex.), R. DuBois (VanderbiltUniversity Medical Center, Nashville), T. Feinstat (Gastroenterology Consultants of Sacramento, Roseville, Calif.), T.R. Foley (RegionalGastroenterology Associates of Lancaster, Lancaster, Pa.), D. Gabbaizadeh (Huntington Research Group, Huntington Station, N.Y.), J.Geenen (Wisconsin Center for Advanced Research, Milwaukee), F. Giardiello (Johns Hopkins Hospital, Baltimore), A. Goetsch (nTouch Re-search, Huntsville, Ala.), M. Goldberg (Regional Gastroenterology Associates of Lancaster, Evanston, Ill.), J.L. Goldstein (University of Illi-nois at Chicago, Chicago), W. Harlan, III (Asheville Gastroenterology Associates, Asheville, N.C.), R. Hogan (Gastrointestinal Associates,Jackson, Miss.), M. Kamionkowski (Gastroenterology Associates of Cleveland, Mayfield Heights, Ohio), M. Kelfer (Fallon Clinic, WestBoylston, Mass.), B. Kerzner (Health Trends Research, Baltimore), K. Kim (University of Chicago Medical Center, Chicago), I. Klimberg(Gastroenterology Associates of Ocala, Ocala, Fla.), G. Koval (West Hills Gastroenterology Associates, Portland, Oreg.), C. Krone (Ad-vanced Clinical Therapeutics, Tucson, Ariz.), S. Krumholz (Waterside Clinical Research, West Palm Beach, Fla.), M.W. Layton (South PugetSound Clinical Research Center, Olympia, Wash.), C. Lightdale (Columbia-Presbyterian Medical Center, New York), P.J. Limburg (MayoClinic, Rochester, Minn.), C. Lind (Vanderbilt University Medical Center, Nashville), D. Lipkis (Institute for Health Care Assessment, SanDiego, Calif.), M. Lloyd (Idaho Gastroenterology, Meridian), D. Maccini (Spokane Digestive Disease Center, Spokane, Wash.), F. MacMilan,Sr. (Pentucket Medical Associates, Haverhill, Mass.), R. Madoff (University of Minnesota, Minneapolis), A. Malik (Advanced Clinical Re-search, North Providence, R.I.), A. Markowitz (Memorial Sloan-Kettering Cancer Center, New York), R. Marks (Alabama Digestive Re-search Center, Alabaster), C.J. McDougall (Manhattan Associates, New York), P. Miner (Oklahoma Foundation for Digestive Research,Oklahoma City), M. Murphy (Southeastern Digestive and Liver Disease Institute, Savannah, Ga.), A. Namias (Gastrointestinal Physicians,Salem, Mass.), N. Nickl (University of Kentucky Medical Center, Lexington), M. Pochapin (Jay Monahan Center for Gastrointestinal Health,New York), R.E. Pruitt (Nashville Medical Research Institute, Nashville), J. Puolos (Cumberland Research Associates, Fayetteville, N.C.),D.S. Riff (AGMG Clinical Research, Anaheim, Calif.), R. Roman (South Denver Gastroenterology, Englewood, Colo.), L. Rubin (New JerseyPhysicians, Passaic), D. Ruff (Healthcare Discoveries, San Antonio, Tex.), M. Safdi (Consultants for Clinical Research, Cincinnati), J. Saltz-man (Brigham and Women’s Hospital, Boston), B. Salzberg (Atlanta Gastroenterology Associates, Atlanta), J.A. Sattler (Western ClinicalResearch, Torrance, Calif.), P. Schleinitz (Americas Doctors Research, Medford, Oreg.), J. Schwartz (Northwest Gastroenterologists, Ar-lington Heights, Ill.), M. Schwartz (Jupiter Research Association, Jupiter, Fla.), M. Silpa (Gastroenterology Associates of The East Bay Med-ical Group, Berkeley, Calif.), D. Silvers (Drug Research Services, Metairie, La.), D. Smoot (Howard University Cancer Center, Washington,D.C.), S. Sontag (Veterans Affairs Medical Center, Hines, Ill.), R.J. Sorrell (Gastroenterology Specialties, Lincoln, Nebr.), D. Stanton (Com-munity Clinical Trials, Orange, Calif.), J. Sturgeon (Americas Doctors Research, Shawnee Mission, Kans.), J.P. Tracey (Hawthorne MedicalAssociates, North Dartmouth, Mass.), T. Werth (Charlotte Gastroenterology and Hepatology, Charlotte, N.C.), C.M. Wilcox (University ofAlabama at Birmingham, Birmingham), R. Wohlman (Northwest Gastroenterology Associates, Bellevue, Wash.), S. Woods (Gastroenter-ology Associates of Fairfield County, Bridgeport, Conn.);

United Kingdom:

J. Burn (South Cleveland Hospital, Middlesbrough);

Australia:

H. Ee (Sir Charles Gairdner Hospital, Nedlands, W.A.), M. Korman (Monash Medical Centre, Clayton, Victoria), A. Lee (Concord Repatria-tion and General Hospital, Concord, N.S.W.), B. Leggett (Royal Brisbane Hospital, Herston, Queensland), F. Macrae (Royal MelbourneHospital, Melbourne, Victoria), L. Mollison (Freemantle Hospital, Freemantle, W.A.), N. Yeomans (Western Hospital, Footscray, Victo-ria), G. Young (Flinders Medical Centre, Bedford, S.A.);

Canada:

G. Aumais (Hospital Maisonneuve-Rosemont, Montreal), R. Bailey (HysMedical Centre, Edmonton, Alta.), C. Bernstein (Winnipeg Health Sciences Centre, Winnipeg, Man.), L. Cohen (Sunnybrook and Wom-en’s Hospital, Toronto), C. Dallaire, R. Dube (Centre Hospitalier Universitaire de Quebec, Quebec, Que.), D. Morgan (McMaster University,Hamilton, Ont.), T. Sylwestrowicz (St. Paul’s Hospital, Saskatoon, Sask.), G. Van Rosendaal (University of Calgary Health Sciences Centre,Calgary, Alta.), S.J. Van Zantan (Queen Elizabeth II Health Sciences Centre, Halifax, N.S.).

references

1.

Marnett LJ, Rowlinson SW, GoodwinDC, Kalgutkar AS, Lanzo CA. Arachidonicacid oxygenation by COX-1 and COX-2:mechanisms of catalysis and inhibition.J Biol Chem 1999;274:22903-6.

2.

Silverstein FE, Faich G, Goldstein JL, etal. Gastrointestinal toxicity with celecoxib vsnonsteroidal anti-inflammatory drugs for

osteoarthritis and rheumatoid arthritis: theCLASS study: a randomized controlled trial.JAMA 2000;284:1247-55.

3.

Bombardier C, Laine L, Reicin A, et al.Comparison of upper gastrointestinal toxic-ity of rofecoxib and naproxen in patientswith rheumatoid arthritis. N Engl J Med2000;343:1520-8.

4.

Marnett LJ, Kalgutkar AS. Cyclooxygen-ase 2 inhibitors: discovery, selectivity andthe future. Trends Pharmacol Sci 1999;20:465-9.

5.

Steinbach G, Lynch PM, Phillips RKS, etal. The effect of celecoxib, a cyclooxygenase-2 inhibitor, in familial adenomatous poly-posis. N Engl J Med 2000;342:1946-52.


n engl j med

352;11

www.nejm.org march

17

,

2005

1080


6.

Giardiello FM, Yang VW, Hylind LM, etal. Primary chemoprevention of familial ad-enomatous polyposis with sulindac. N EnglJ Med 2002;346:1054-9.

7.

Baron JA. Epidemiology of non-steroi-dal anti-inflammatory drugs and cancer.Prog Exp Tumor Res 2003;37:1-24.

8.

Hawk ET, Viner J, Richmond E, Umar A.Non-steroidal anti-inflammatory drugs(NSAIDs) for colorectal cancer prevention.Cancer Chemother Biol Response Modif2003;21:759-89.

9.

FitzGerald GA, Patrono C. The coxibs,selective inhibitors of cyclooxygenase-2.N Engl J Med 2001;345:433-42.

10.

FitzGerald GA. Coxibs and cardiovascu-lar disease. N Engl J Med 2004;351:1709-11.

11.

Topol EJ, Falk GW. A coxib a day won’tkeep the doctor away. Lancet 2004;364:639-40.

12.

White WB, Faich G, Borer JS, MakuchRW. Cardiovascular thrombotic events in ar-thritis trials of the cyclooxygenase-2 inhibi-tor celecoxib. Am J Cardiol 2003;92:411-8.

13.

Mamdani M, Juurlink DN, Lee DS, et al.Cyclo-oxygenase-2 inhibitors versus non-selective non-steroidal anti-inflammatorydrugs and congestive heart failure outcomesin elderly patients: a population-based co-hort study. Lancet 2004;363:1751-6.

14.

Solomon DH, Schneeweiss S, Glynn RJ,et al. Relationship between selective cyclo-oxygenase-2 inhibitors and acute myocar-dial infarction in older adults. Circulation2004;109:2068-73.

15.

Ray WA, Stein CM, Daugherty JR, HallK, Arbogast PG, Griffin MR. COX-2 selectivenon-steroidal anti-inflammatory drugs andrisk of serious coronary heart disease. Lan-cet 2002;360:1071-3.

16.

Mamdani M, Rochon P, Juurlink DN, etal. Effect of selective cyclooxygenase 2 in-hibitors and naproxen on short-term risk ofacute myocardial infarction in the elderly.Arch Intern Med 2003;163:481-6.

17.

Furberg CD, Psaty BM, FitzGerald GA.

Parecoxib, valdecoxib, and cardiovascularrisk. Circulation 2005;111:249.

18.

Juni P, Nartey L, Reichenbach S, SterchiR, Dieppe PA, Egger M. Risk of cardiovascu-lar events and rofecoxib: cumulative meta-analysis. Lancet 2004;364:2021-9.

19.

Crofford LJ, Lipsky PE, Brooks P,Abramson SB, Simon LS, van de Putte LB.Basic biology and clinical application of spe-cific cyclooxygenase-2 inhibitors. ArthritisRheum 2000;43:4-13.

20.

FitzGerald GA. COX-2 and beyond: ap-proaches to prostaglandin inhibition in hu-man disease. Nat Rev Drug Discov 2003;2:879-90.

21.

McAdam BF, Catella-Lawson F, MardiniIA, Kapoor S, Lawson JA, FitzGerald GA.Systemic biosynthesis of prostacyclin by cy-clooxygenase (COX)-2: the human pharma-cology of a selective inhibitor of COX-2. ProcNatl Acad Sci U S A 1999;96:272-7. [Erra-tum, Proc Natl Acad Sci U S A 1999;96:5890.]

22.

Brock TG, McNish RW, Peters-GoldenM. Arachidonic acid is preferentially metab-olized by cyclooxygenase-2 to prostacyclinand prostaglandin E2. J Biol Chem 1999;274:11660-6.

23.

Chenevard R, Hurlimann D, Bechir M,et al. Selective COX-2 inhibition improvesendothelial function in coronary artery dis-ease. Circulation 2003;107:405-9.

24.

Farkouh ME, Kirshner H, HarringtonRA, et al. Comparison of lumiracoxib withnaproxen and ibuprofen in the TherapeuticArthritis Research and GastrointestinalEvent Trial (TARGET), cardiovascular out-comes: randomised controlled trial. Lancet2004;364:675-84.

25.

Zhao SZ, Burke TA, Whelton A, von All-men H, Henderson SC. Comparison of thebaseline cardiovascular risk profile amonghypertensive patients prescribed COX-2-specific inhibitors or nonspecific NSAIDs:data from real-life practice. Am J ManagCare 2002;8:Suppl:S392-S400.

26.

Bresalier RS, Sandler RS, Quan H, et al.Cardiovascular events associated with rofe-coxib in a colorectal adenoma chemopreven-tion trial. N Engl J Med 2005;352:1092-102.

27.

Mukherjee D, Nissen SE, Topol EJ. Riskof cardiovascular events associated with se-lective COX-2 inhibitors. JAMA 2001;286:954-9.

28. Weir MR, Sperling RS, Reicin A, GertzBJ. Selective COX-2 inhibition and cardio-vascular effects: a review of the rofecoxib de-velopment program. Am Heart J 2003;146:591-604.29. Konstam MA, Weir MR, Reicin A, et al.Cardiovascular thrombotic events in con-trolled, clinical trials of rofecoxib. Circula-tion 2001;104:2280-8.30. Topol EJ. Rofecoxib, Merck, and theFDA. N Engl J Med 2004;351:2877-8.31. Ott E, Nussmeier NA, Duke PC, et al. Ef-ficacy and safety of the cyclooxygenase 2 in-hibitors parecoxib and valdecoxib in pa-tients undergoing coronary artery bypasssurgery. J Thorac Cardiovasc Surg 2003;125:1481-92.32. Nussmeier NA, Whelton AA, BrownMT, et al. Complications of the COX-2 inhib-itors parecoxib and valdecoxib after cardiacsurgery. N Engl J Med 2005;352:1081-91.33. Celecoxib. PhRMA IQ5-97-02-001. (Ac-cessed February 18, 2005, at http://www.clinicalstudyresults.org/documents/company-study_76_0.pdf.)34. Topol EJ. Failing the public health —rofecoxib, Merck, and the FDA. N Engl J Med2004;351:1707-9.35. Dieppe PA, Ebrahim S, Martin RM, JuniP. Lessons from the withdrawal of rofecox-ib. BMJ 2004;329:867-8.36. Josefson D. FDA warns Merck over itspromotion of rofecoxib. BMJ 2001;323:767.37. Mukherjee D, Topol EJ. Pharmaceuticaladvertising versus research spending: areprofits more important than patients? AmHeart J 2003;146:563-4.Copyright © 2005 Massachusetts Medical Society.


ebm self instructional manual 2008[1]

Documents