EVIDENCE BASED MEDICINE Dan O Connell MD Montefiore Family

EVIDENCE BASED MEDICINE Dan O’Connell, MD Montefiore Family Health Center

The Problems: We need evidence (about the accuracy of diagnostic tests, the power of prognostic markers, the comparative efficacy and safety of interventions, etc. ) about 5 times for every in-patient (and twice for every 3 out-patients). We get less than a third of it

Median minutes/week spent reading about my patients: Self-reports at 17 Grand Rounds: l Medical Students: 90 minutes l House Officers (PGY 1): 0 (up to 70%=none) l PGY 2 -4 20 (up to 15%=none) l Attendings: 45 (up to 40%=none) l Sr. Attendings 30 (up to 15%=none) l Specialists » Grad. Post 1975: 45 (up to 30%=none) » Grad. Pre 1975: 30 (up to 40%=none)

Three solutions Clinical performance can keep up to date: 1 by learning how to practice evidence-based medicine ourselves. 2 by seeking and applying evidence-based medical summaries generated by others. 3 by applying evidence-based strategies for changing our clinical behaviour.

What evidence-based medicine is: The practice of EBM is the integration of • individual clinical expertise with the • best available external clinical evidence from systematic research. and • patient’s values and expectations

I. Individual Clinical Expertise: • Clinical skills and clinical judgement • Vital for determining whether the evidence (or guideline) applies to the individual patient at all and, if so, how

II. Best External Evidence: • From real clinical research among patients. • Has a short doubling-time (10 years). • Replaces currently accepted diagnostic tests and treatments with new ones that are more powerful, more accurate, more efficacious, and safer.

III. Patients’ Values & Expectations • Have always played a central role in determining whether and which interventions take place • Patient adherence based on doctor-patient relationship and values and views of patients

What EBM is not: • EBM is not cook-book medicine – evidence needs extrapolation to my patient’s unique biology and values • EBM is not cost-cutting medicine – when efficacy for my patient is paramount, costs may rise, not fall • EBM is not a bludgeon to win an argument

Evidence-Based Medicine: The Practice When caring for patients creates the need for information: 1 Translation to an answerable question (PICO) (patient/intervention/comparison/outcome). 2 Efficient track-down of the best evidence – secondary (pre-appraised) sources e. g. , Cochrane; E-B Journals – primary literature

Evidence-Based Medicine: The Practice 3 Critical appraisal of the evidence for its validity and clinical applicability è generation of a 1 -page summary. 4 Integration of that critical appraisal with clinical expertise and the patient’s unique biology and beliefs è action. 5 Evaluation of one’s performance.

We needn’t always carry out all 5 steps to provide E-B Care m. Asking an answerable question. è Searching for the best evidence. Searching è Critically-appraising the evidence. appraising m. Integrating the evidence with our expertise and our patient’s unique biology and values mevaluating our performance

We’ve identified 3 different modes of practice è“Searching & appraising” – provides E-B care, but is expensive in time and resources è“Searching only” – much, quicker, and if carried out among E-B resources, can provide E-B care è“Replicating” the practice of experts – quickest, but may not distinguish evidence-based from ego-based recommendations

Even fully EB-trained clinicians work in all 3 modes 1. “Searching & appraising” mode for the problems I encounter daily. 2. “Searching only” mode among E-B resources for problems I encounter once a month. 3. “Replicating” the practice of experts mode for problems I encounter once a decade(and crossing my fingers!).

Three solutions Clinical performance can keep up to date: 1 by learning how to practice evidence-based medicine ourselves. 2 by seeking and applying evidence-based medical summaries generated by others. 3 by applying evidence-based strategies for changing our clinical behaviour.

Information required within seconds Systematic reviews, periodically updated, of randomised trials of the effects of health care (from all sources, and in all languages): The Cochrane Collaboration.

Cochrane Systematic Reviews (522; another 500 in preparation) Database of Abstracts of Reviews of Effectiveness (1895) Registry of Randomised Controlled Trials (218, 355)

EBM and E-B Guidelines • The best evidence comes from systematic reviews (such as Cochrane) and/or E-B journals of 2º publication: – Much more likely (than personal search and critical appraisal) to be true – Saves the clinician’s precious (scarce!) time • Avoids error and duplication of effort

SOURCES Systemic reviews: www. cochrane. org (Cochrane abstracts) Example: Cochrane Hepato. Biliary Group • Antibiotics for spontaneous bacterial peritonitis in cirrhotics (Cochrane Review)

• • Background: Spontaneous bacterial peritonitis is mainly a complication of cirrhotic ascites that occurs in the absence of any intra-abdominal, surgically treatable source of infection. Antibiotics have been recommended as the mainstay treatment for spontaneous bacterial peritonitis. However, this recommendation is not based on convincing evidence. It has been proposed that treatment should cover Gram-negative enteric bacteria and Gram-positive cocci, that are responsible for up to 90% of cases. Objectives: To evaluate the effectiveness and safety of different types and ways of antibiotic therapy for spontaneous bacterial peritonitis in cirrhotic patients. Search strategy: Electronic searches on the Cochrane Hepato-Biliary Group Trials Register (March 2000), the Cochrane Controlled Trials Register (Cochrane Library Issue 3, 2000), MEDLINE (1966 -2000), EMBASE (1980 -2000); scanning the references of all identified studies; contacting the first author of each included trial. Selection criteria: Randomised trials comparing different types of antibiotics for spontaneous bacterial peritonitis in cirrhotic patients. Data collection and analysis: Data were independently extracted by two reviewers. Relative risks or weighted mean differences, with their 95% confidence intervals were estimated using 'intention-to-treat' analyses. Main results: Nine trials dealing with 684 patients diagnosed with spontaneous bacterial peritonitis were included. No placebo-controlled trial was found. Each of the included trials compared different antibiotics, and no meta-analysis could be performed. We were unable to establish the optimal dose or duration of antibiotic therapy and found no convincing evidence that cefotaxime is more effective than ampicillin-tobramycin or that oral quinolones should be recommended for patients with less severe manifestations of the disease. Reviewers' conclusions: This review provides no clear evidence for the treatment of cirrhotic patients with spontaneous bacterial peritonitis. Until large, well-conducted, trials provide adequate evidence, treatment must be based on clinical experience. Citation: Soares-Weiser K, Brezis M, Leibovici L. Antibiotics for spontaneous bacterial peritonitis in cirrhotics (Cochrane Review). In:

TRIPDATABASE. com www. tripdatabase. com Previously free, now subscribed Evidence based articles Clinical Guidelines Query-answering services E Textbooks Clinical Queries on Pub. Med Limitation: frequently links only to abstract Pub. Med. com - “clinical query” separates wheat from chaff

Mixed public and subscription • www. uptodate. com Extensive “Topic Reviews”, article links to subscribers • Organized somewhat like skimmable textbook chapters in evolution, with topics to skip to by headings on left side

By subscription • www. Info. POEMS. com • http: //www. acponline. org/journals/ecp (access to current issue) • www. ovid. com - links to EBM databases

AECOM LIBRARY REMOTE ACCESS • For full articles, not just abstract: • www. library. aecom. yu. edu • “remote access” • 10 digit library bar code, password • AECOM library 430 -3104 • MMC library 920 -4666 (sheila smalling)

FOR PDAs • www. library. uchc. edu - PDA resources – medical databases, medical link resources • www. handango. com • Atp III calculator, OB wheel, Med Calc, etc

www. google. com

www. google. com • • No evidence Internet is dominated by commerce Health is dominated by “alternative health” LOOK for educational web sites “. edu” as these frequently are academic medicine • Great for news “lipitor cancer june 2003” • Great for non-digested info about obscure topics – drugs from other countries, unusual treatments

SOURCES • • COCHRANE. ORG TRIPDATABASE. COM UPTODATE. COM LIBRARY. AECOM. . YU. EDU PUBMED. COM - CLINICAL QUERIES CEBM. NET DARE: www. nhscrd. york. ac. uk/darehp. htm WWW. EBMNY. ORG (connect to DARE, other web sites

PICO – generating questions and search terms • Patient/Population • Intervention • Comparison • Outcome

PICO • 3 y. o. , fever x 1 day and red ear. ? Antibiotics? • 82 y o CHF on Beta Blocker, ACE, diuretic, increased symptoms. ? Add digoxin? • 52 yo perimenopausal hot flashes. ? risks, benefits of short term HRT?

RRR, ARR, NNTT: LIPOPRILOLOL • STUDY A: Lipiprilolol lowers Heart attacks and strokes by 33% • STUDY B: 2% fewer patients Lipiprilolol treated patients had heart attacks and strokes • STUDY C: Treating 50 patients with Lipiprilolol resulted in 1 fewer heart attack and stroke

RRR, ARR, NNTT • Fictitious study : LIPIPRILOLOL lipid/ACE/Beta blocker vs placebo give to healthy 50 year olds • • 100 given placebo, 100 given drug, CAD/CVA events measured Placebo - 6 events Drug – 4 events What is RRR? What is ARR? What is NNTT?

RRR RELATIVE RISK REDUCTION Proportionate decrease between the treatment and placebo groups (Treatment% – placebo%) DIVIDED BY / placebo % d E. G. if 4% in treatment group and 6% in placebo group: ( 6%-4%) / 6% = 2/6 = 33%

ARR Absolute Risk Reduction • Difference between the percent of events in the placebo group and the percent of events in the treatment group • (% Placebo) MINUS (%Treatment) E. G. if 6% in placebo group, 4% in treatment group ARR = 6% - 4% = 2% (or 0. 06 – 0. 04 = 0. 02)

NNTT Number Needed to Treat • Number of people to be treated to achieve one outcome • Inverse of ARR Or • 100% / (divided by) (%Events Placebo/Total Placebo MINUS %Events Treatment/Total Treatment) 6% with treatment 4% with placebo 100% / (6%-4%) = 100/2 = 50 (or 1. 0/ 0. 02) = 50

RRR, ARR, NNTT Example 2 : Enuresis • 200 7 year olds treated with alarm blankets and behavioral techniques (less fluid after bed etc) • 200 no alarm blanket • After 3 months, 10 of placebo with dry weeks • 60 of alarm blanket group with dry week • What is RRR? • What is ARR? • What is NNTT?

• 60/ 200 = 30% • 10/200 = 5% • RRR: = (30% - 5%) / 30% =25/30 =83% • ARR = (60/200) – (10/200)=50/200= 25% or 30% - 5% = 25% • NNT = 100 / (30% - 5%) = 100/25 = 4

SOURCES • • COCHRANE. ORG TRIPDATABASE. COM UPTODATE. COM LIBRARY. AECOM. . YU. EDU PUBMED. COM - CLINICAL QUERIES CEBM. NET DARE: www. nhscrd. york. ac. uk/darehp. htm WWW. EBMNY. ORG (connect to DARE, other web sites