Understanding concepts of Evidenced Based Medicine Frank J. Domino, M. D. Professor University of Massachusetts Medical School
Learning Objectives by the end of the session, you will n n Appreciate the basic statistical concepts involved in EBM Contrast absolute risk reduction (ARR) with relative risk reduction (RRR) and Calculate the number needed treat (NNT); and Understand how sometimes statistical interpretation can leads us astray….
Outline of Talk 1. Types of Papers in the Medical Literature 1. 2. 3. Review vs. Systematic Review RCT, Cohort, Case-Control Studies Bias 2. Statistics: AR, RRR, NNT 3. Patient vs Disease Oriented Evidence
What is Evidence-Based Medicine? n Integration of • Best research evidence (from Systematic Reviews) Patient Values • Clinical expertise • Patient values n Patient Oriented Outcomes [POE] vs Disease/Intermetdiate Outcomes [DOE] 4 Sacket et al “How to Teach and Practice EBM”, Churchill Livingston, 2000 Len
Types of Research Information Review Articles Vs Systematic Reviews
Traditional Review Article - Summary of the literature Often valuable reviews for medical practice Problems: Do not necessarily include all relevant evidence n Author bias mixed with the evidence n Publisher’s motives in ? n
A Systematic Review Article – Utilizes quality standards to judge the literature n n n Clear objective for evaluation of study Identify studies (Randomized Controlled Trials) meeting review criteria The results of acceptable studies combined Outcome of those studies published If Quantitative, combined -> Meta Analysis
Systematic Review Meta Analysis Systematic Review
Antibiotics and Acute Sinusitis A Cochrane Systematic Review www. cochrane. org 1. Objectives: ‘To determine whether antibiotics are indicated for acute sinusitis, and if so, which antibiotic classes are most effective. ’ 2. Search strategy: Studies identified via searches of MEDLINE & EMABASE, contacts w/ pharmaceutical companies and bibliographies of included studies.
3. Selection criteria n n n Randomized controlled trials n > 30 adults. Compare antibiotic to control or other Abx. DX confirmed by radiograph or aspiration. Outcomes: cure or symptom improvement. Of 2058 potentially relevant studies, only 49 studies (13, 660 pts) met review criteria! > 2000 of questionable significance……. .
4. Data Combined Reviewers' conclusions: • For acute maxillary sinusitis, current evidence is limited but supports penicillin or amoxicillin for 7 to 14 days. • Clinicians should weigh the moderate benefits of antibiotic treatment against the potential for adverse effects. The Cochrane Library, Oxford
The Forest Plot Estimates with 95% confidence intervals Line of no effect Kennedy 1997 Locke 1952 A Estimate and confidence interval for each study Lopes 1997 Reynolds 1998 Estimate and confidence for the meta-analysis Seiberth 1994 0. 2 1. 0 Risk ratio Favours LR 5 Favours control Scale (effect measure) Direction of effect Forest Plot: If <> to your Left, Intervention was Effective at Lowering Risk of Outcome. 12 Len
Interpreting a Meta Analysis using a Forest Plot Len 13
Study Designs The Systematic Reviews use: 1. Randomized Controlled Trials Other Study types: 2. Cohort Studies 3. Case Control Studies
Randomized Controlled Clinical Trials n n A population is chosen, then randomly assigned to an intervention or not An intervention is given to the study population. The Non-Intervention group receives a placebo or some standard of care. Differences in pre identified outcomes are measured and produce ABSOLUTE RISK of those outcomes
Randomized Controlled Trials Bias in the Medical Literature Attrition bias – how were “drop outs” accounted for Publication bias – only positive results get published Comparator bias - new Tx compared to placebo rather than current standard Commercial Bias – who funded the study and what motivated the researchers
Attrition Bias n n n What happened to everyone who was randomized? Was the study “Intention to Treat”? May need to review the data to determine if they “add up” Bob 17
Publication Bias Factors that prevent publication 40% of All RCT not published n RCT of using 400 IU/day of Vitamin E to prevent Coronary Artery Disease. Intervention MI 4. 2 3. 9 Placebo p=0. 7 No benefit to supplementing diet with Vitamin E in the prevention of CAD 18 Bob
Comparator Bias Comparing new treatment to no treatment, rather than the current standard of care n n “Azithromycin is superior to placebo in the treatment of Acute Sinusitis” That is nice, but is it superior to Amoxicillin? Don’t know; they didn’t do (or won’t publish) that study. Bob 19
Commercial Bias n n Look for Disclosures of the Authors Who funded the study Bottom of The Front page and just before Reference Section Ex. JUPITER Study BIAS: Assume it is Present $$ Bob 20
P < 0. 001 Statisticall y significant! AR MI placebo - 0. 76% AR MI Crestor - 0. 35%
Cohort Studies n n Prospective, Observational studies with conclusions Produces a Relative Risk (RR) RR=Incidence of disease in Exposed divided by Incidence of disease in Unexposed population Ex: Framingham Heart study
Relative Risk (RR) RR = Number of times more or less than 1 an event will happen in one group when compared to another If < 1. 0, risk is REDUCED If > 1. 0, risk is INCREASED
“The RR of death if involved in an MVA without a seat belt = 3. 5” “Tea drinkers have a 0. 6 RR of dying from CAD”
Does tea drinking prevent CAD? ? ? Correlation does not prove cause and effect!
Vitamin E & CAD 3 cohort studies mid-90’s concluded vitamin E use correlated with a lower risk for CAD. 1. Antioxidant Vitamins and Coronary Heart Disease 2. Vitamin E Consumption and Risk of Coronary Disease in Men 3. Vitamin E Consumption and Risk of CAD in Women
The Heart Outcomes Prevention Evaluation (HOPE) Study 2000: Controlled Trial of Vitamin E for 5 Yr: rates of MI, CVA & CV death did not differ significantly from placebo (16. 2% v 15. 5 %). • HOPE II (JAMA 2005) Long-term use of vit E in CHD or DM patients does not prevent cancer or major cardiovascular events and may increase the risk for HF
Case Control Studies n n Identify a Potential trend in disease Collect exposure history of “CASES” Identify similar people age/gender like Cases but without disease Compare Cases to Controls to determine if exposure increased ODDS of disease. Ex: Cigarette smoking
Outline of Talk 1. Types of Papers in the Medical Literature 1. 2. 3. Review vs. Systematic Review RCT, Cohort, Case-Control Studies Bias 2. Statistics: AR, RRR, NNT 1. Patient vs Disease Oriented Evidence
Which drug would you take? n n n Drug A can reduce your MI risk by 1/3 Drug B can reduce your MI risk by 9 % Drug C every 11 patients who take Drug C, one MI will be prevented A B C
The Scandinavian Simvastatin Survival Study (4 S) n n 4444 pts with angina or previous MI and ↑ cholesterol Randomized to simvastatin or placebo. After 5 yrs, simvastatin reduced TC 25%, LDL 35% and increased HDL 8%. [DOE] MI or death: 622(28%) control group (placebo) 431 pts (19%) simvastatin group [POE]
Disease Oriented vs. Patient Oriented Evidence DOE Lidocaine ↓ V. Tach POE Prophylactic Lido ↑ CV death HRT will ↓ LDL; ↑ HDL Increased risk of stroke & Mortality Fluoride ↑ bone density Does not prevent fractures
Absolute Risk Reduction (ARR) 4 S Trial: The difference between the incidence of outcome in the control group (28%) and the incidence in the treatment group (19%). ARR = Incid Control – Incid Treatment 28%- 19% = 9% ARR
Relative Risk Reduction Absolute Risk Reduction (ARR) divided by Incidence in the control group RRR = (28 -19)/28 = 33% RRR is not the same as AR or RR
Relative Risk Reduction A way to describe (and often over inflate) the relative impact of a treatment on an outcome 35 Bob
Number Needed to Treat (NNT) n n NNT - the number of people needed to receive an intervention before one person gets the expected outcome! NNT = 100 divided by Absolute Risk Reduction. NNT = 100/(28 -19) = 11 11 pts w/ CAD need to be treated with simvastatin to prevent 1 from having a subsequent MI or death.
Which drug would you take? n Drug A can reduce your MI risk by 1/3 rd • Relative Risk Reduction (RRR = 33%) n Drug B can reduce your MI risk by 9% • Absolute Risk Reduction (ARR = 9 %) n Drug C prevent an MI for every 11 patients who take it regularly • Number Needed to Treat (NNT = 11)
Another example of NNT: Does alendronate prevent hip fractures in postmenopausal women ? n n The published study: Low bone mass density One or more fractures at baseline Alendronate 5 mg/d x 24 mo, then 10 mg/d Outcome - subsequent hip fractures
Outcomes P < 0. 044 Statistically significant! Statistical Significant does NOT equal Clinical Significant
Statistical Significance n n n An indication that the findings are not due to chance! P < 0. 05 means there is < 5% chance the difference between the placebo and treated group is a chance occurrence. It is not an implication of clinical meaning (significance).
A Closer Look… Hip fracture rate in treated group 1% Hip fracture rate in control group - 2. 2%
Absolute Risk Reduction Placebo incidence – Treated incidence 2. 2% - 1% = 1. 2% A high risk patient can reduce her risk of a hip fracture by 1. 2% from alendronate x 4 yrs
What happens to the other 82 women? n They receive the medication, incurring the cost of treatment along with the exposure to the potential side effects, but…. . obtained no identified benefit!
What is this 56% ? ? ? Relative Risk Reduction 2. 2 -1(ARR)/2. 2 = 56%
Number Needed to Treat 100/ARR = 100/(2. 2 -1) = 83 83 high risk women would have to be treated with alendronate for 3 years to prevent one additional hip fracture n Evidence-Based Medicine
What’s a “good” NNT ? n n n Depends on risks of intervention vs outcome The best NNT would be 1 - every treated patient benefited, but no placebo benefit NNTs < 5 indicate very effective treatments
Good NNT n n NNT is a VALUE BASED decision NNTs of 50 or 100 useful for interventions to reduce death after heart attack. NNS in the 1000’s as screened population includes those with and without disease. Decisions should also consider costs and risk of the intervention (NNH)
Some NNTs Prevent 1 Drug NNT MI/Death ASA X 1 year 500 healthy ♂ 25 unstable angina 700 mild HTN 15 severe HTN Stroke/MI/Death Anti-HTN X 1 yr Death ACEi 18 CHF/post-MI Acute Rheumatic fever Hip Fx Penicillin 3500 strep throats Ca/Vit D X 3 yr 30 ambulatory♀
Number Needed to Harm? Drug Prevent 1 Hip Fx Cause 1 Stroke Cause 1 Breast Ca NNT 333 ♀ HRT X 5 yrs 250 ♀ 200 ♀ Cause 1 DVT OM pain relief Cause Vomiting Cause Diarrhea Cause a Rash NNH 100 ♀ 15 Antibiotic X 10 days 12 12 12
Crestor for Low Risk Patients AR MI with placebo - 0. 76% AR MI with Crestor - 0. 35% ARR = 0. 76 -0. 35 =0. 41% RRR= 0. 41/0. 76 = 54% NNT = 100/0. 41 = 244 low risk patients would have to be treated with rosuvastatin for 2 years to prevent one MI
What happens to the other 243 patients? n They receive the medication, incurring the cost of treatment along with the exposure to the potential side effects, but…. . obtained no recognized benefit!
Solving Questions on the Fly: Using Best Evidence Frank J. Domino, M. D. Professor Dept. Family Medicine & Community Health University of Massachusetts Medical School Worcester, Massachusetts
Objectives: 1. 2. 3. Understand what characteristics denote “Best” medical evidence Appreciate spectrum of resources available to use at bedside or with E. M. R. Solve your questions in real time 55
Why this session? n n n President Obama January 2009: Provision of $40, 000 in incentives (beginning in 2011) for physicians to use an EHR Funds to coordinate interoperable EHRs Education programs to train in EHR use Creation of HIT grant and loan programs Acceleration of the construction of the National Health Information Network (NHIN) 56
Who has an E. H. R. ? http: //www. aafp. org/fpm/2009/1100/10. html 57
Now that Patient Information is at the bedside, why not solve questions there too? 58
1. Screening: 56 year old male wants to know why there is a controversy about Prostate Cancer screening; “if you find it early, won’t I stand a better chance of living longer? ” For Questions about SCREENING, use: www. ahrq. gov 59
Prevention: www. ahrq. gov 60
62
64
2. Basics: You have a MS III in your office who sees a patient w/ Irritable Bowel Syndrome & wants to learn about this Diagnosis n www. emedicine. com free www. epocrates. com free/$ www. 5 mcc. com $ n Not: Google or Wikipedia n n 66
OR, go to: www. emedicine. com 67
68
69
Epocrates: www. epocrates. com 70
Continuing Med. Ed. “Listen to Lecture” +: “hit PLAY” -: Industry Funded/Biased, Boring Point of Care (tracks usage) +: Fills YOUR Knowledge Gaps, rewards your desire to improve Up. To. Date, 5 Minute CC, Dynamed 71
3. Deeper Dive 79 y/o becomes acutely ill after his most recent intra-vesicle BCG instillation for Transitional Cell Ca of Bladder; what should you do? www. uptodate. com $ www. accessmedicine. com (Harrison’s) $ n 72
73
150 Hits 74
4. Best Evidence You are frustrated about the controversy re: serum homocysteine levels in patients at risk for CHD and you want the BEST EVIDENCE. n Systematic Review Databases • Cochrane Database of Systematic Reviews • Others from PUBMED. GOV 75
What is Evidence Based Practice? n Integration of • Best research evidence • Clinical expertise • Patient values n Most Explicit: Use a Systematic Review to guide Patient Care Sacket et al “How to Teach and Practice EBM”, Churchill Livingston, 2000 76
www. pubmed. gov
78
www. pubmed. gov n n n Use Clinical Queries Filter Search suffixes * adds any suffix to the end of the word (ie. asthm*) [ti] = any term in the title (ie. Asthm*[ti]) [ab] = any term in abstract (ie. Asthm*[ab]
Online & Electronic Tools Cost/Yr Pros Cons SCREENING: www. ahrq. gov $0. 00 -Free -Unbiased -Standard of Care - Too many “I” ratings BASICS: www. emedicine. com $0. 00 Free -Current -Lots of ads, now owned by Medscape -May become biased based upon funders -PDA version weak BASICS: : www. epocrates. com $195. 00 -Rx, Basics, DDX, Lab, calculators, Pt Ed -Not Biased -PDA & Web -Web not done yet BASICS: : www. 5 mcc. com $99. 00 -Rx, Basics, DDX, Lab, Video, Calculators, Patient Education -One fee Book, PDA, Web -no CME yet Deeper Dive www. uptodate. com $500 -1 st Yr $400 -Comprehensive Medicine content Subspecialist Author -can be overwhelming -PDA version HUGE -PDA, CME Deeper Dive www. accessmedicine. c om/harrisons $200/yr Adult Medicine -have to buy subscriptions to other resources, which adds up quickly BEST EVIDENCE www. pubmed. gov Free Use Clinical Queries filter Not all reviews listed are 82 systematic reviews, read carefully
Solving Questions on the Fly: Using Best Evidence Frank J. Domino, M. D. Professor Dept. Family Medicine & Community Health University of Massachusetts Medical School Worcester, Massachusetts Frank. domino@umassmemorial. org
Let’s look for answers to your questions 84
Скачать презентацию Understanding concepts of Evidenced Based Medicine Frank J
a2b190937a867361a776e54242dace46.ppt