Docstoc

PHELO EBM

Document Sample
PHELO EBM Powered By Docstoc
					Family Med EBM Conference
            Some History
• Often MI’s > abnormal beats (Ventricular
  Premature Depolarizations).
• Increase beats = increase risk of Sudden
  Death
• Therefore, Give meds & decrease extra
  beats = decrease risk & increase survival
• And in the 80’s, we did,…
              Question?
• “Are we really saving lives?”
• No research comparing encainide/flecainide
  vs placebo.
• So researcher attempted to prove it,..
• After 10 months they looked at the blinded
  data and saw,…
              I told you so,…
• A clear win for                    Group A Group B
  medicine,                          (730)   (725)
• Stopped trial &
  opened envelopes   Mortality       56          22
                                     (7.7%)      (3.0%)

                     arrhythmia 33               9
                     death or   (4.5%)           (1.2%)
                     cardiac
                     arrests

                     NEJM 1989; 321(6): 406-12
               That can’t be,…
• Actually, Treatment                 Treatment Placebo
  (encainide/flecainide)              (730)         (725)
  was harming people
• Relative risk of dying Mortality    56            22
  was 157% higher                     (7.7%)        (3.0%)
• The Number Needed
                         arrhythmia   33            9
  to Harm (kill) 1 extra
                         death or     (4.5%)        (1.2%)
  patient was only 21.
                         cardiac
                         arrests

                                 i. NEJM 1989; 321(6): 406-12
           It happens all the time,…
• Past teaching proven incorrect with evidence
  – Vitamins prevent oxidation
  – OTC cough suppressants for kids
  – Lubricant on speculum wrecks PAP tests
  – Morphine reduces Dx accuracy in Abdo pain
  – Low Carb Diets are the solution.
  – Topical NSAID is an inefficacious CAM Tx
  – Oral HRT for urinary incontinence
Lancet 2002; 360: 23-33 & Lancet 2003;361:2017-23. Arch Dis Child 2002; 86: 170-5.
Obstet Gynecol 2002; 100: 889-92. J Am Coll Surg 2003; 196: 18-31. JAMA 2003; 289
(14): 1837-50 & NEJM 2003; 348 (21): 2082-90. www.Bandolier.com March 05. JAMA
2005; 293(8): 935-48
                       If it’s not RCT,…


    • Case study & series lowest on the list but
        – Thalidomide danger 1st via case report
        – Torsades with Terfenadine & Itraconazole


    • Cohort: Smoking is Bad! (no RCT, ever*)


McBride Lancet 1961;ii, 1358. Pohjola-Sintonen et al. Eur J CLin Pharm 1993;45:191-3.
Doll & Peto. BMJ 1964;i;:1399-414,1460-7.
            In Summary?
• EBM is the best we’ve got?

• Understand the basics

• Always think and ask questions ?
So how did we get to this
     curriculum?
                Why This Curriculum:
                   Past Research
• Teaching critical appraisal in isolation doesn’t help
  doctors become EBM users.
• Doctors don’t do critical appraisal
• Doctors spend 2 minutes to look up questions
• Workshops increase Doctors Knowledge Base
• Integration into daily practice (with clinical questions &
  web-resources) changes practice.
• Biggest Barriers have been Time and Knowledge

Acad Med. 1999;74:686-94. CMAJ 1998;158:177-81. BMJ. 2002 7;325:1338-41. Acad
   Med. 2000;75:1212-4. BMJ 1999; 319: 358-61 Acad Med. 2002;77: 741-2. BMJ.
   2004;329: 1017. Acad Med. 2003 Mar;78:270-4.
       Why This Curriculum:
          Our Research
• Residents come from varied backgrounds
  and many report EBM training was limited
  in Quality and Quantity
• Residents think EBM results better patient
  care and can be practical.
• Biggest Barrier is time.
         Our EBM Curriculum.
• Give everyone some basic EBM knowledge
  – This Workshop.
• Give everyone easy access to EBM resources
  – The Desktop.
• An assignment integrated into clinical practice
  that reflects what is done in future practice.
  – The BEAR’s
• Give a journal club enhance rapid appraisal
  – Resident Journal Club
• Keep you up to date
  – The Quarterly Lit Review
BEAR Work-Sheet (Title): ____________________________________________

Name of Resident:                                          Date:

Question: ____________________________________________________________
_____________________________________________________________________

Search: (Check all that apply)
Pubmed/Ovid/Medline: □           Filtered Resources: □    Summary/Review Sites: □
   College/Society/Guidelines: □            Other: □ ( Describe: ___________________ )
Number of Resources Reviewed: ____

Resources (Top 3)
#1 Resource: Abstract □      Paper □            Filtered Article □   Summary □
Review/Meta-Analysis □       College/Society/Guideline Paper □       Other Research □
    –   Abbreviated Citation: __________________________________________
    –   Strengths:___________________________________________________
    –   Weaknesses:_________________________________________________
Take-Home Message:__________________________________________________

#2,…

Bottom-line: ____________________________________________________________
________________________________________________________________________
_____________________________________________________________________

Practice (These findings had a):
Large Change □      Small Change □               Reassured □                  No Help □
    Therapy Articles
Much Thanks to: Rob Hayward & Tanya Voth, CCHE
    Objectives: Practical EBM
• Generally:
  – Recognize the primary features which answer
     • Are the Results Valid
     • Number needed to treat.
• Specifically,
     • Learn the Language
     • Learn where to look to rapidly get answers.
Users Guides
    Basics of the Randomized
       Control Trial (RCT)
• RCT is an experiment
• We divide people in to 2 or more groups
  (“Randomly”)
• We subject the groups to different
  interventions (a Treatment and a “Control”)
• Then compare them
        Are the Results Valid
1) Did groups begin the study with a
  similar prognosis?
  a) Were patients randomized?
    • Without:
      – Prognostic factors = determinants of outcome
      – Sometimes we know Prognostic factors,
        sometimes we don’t
      – Outcomes affected by many factors other than the
        experimental intervention
       Are the Results Valid?
1) Did groups begin the study with a
  similar prognosis?
  a) Were patients randomized?
    • Remedy:
      – Effect of randomization is to evenly distribute prognostic
        factors between experimental and control groups
      – Balancing both known
        and unknown prognostic factors
      – eliminating selection & confounding biases
         Are the Results Valid?

1) Did groups begin the study with a
  similar prognosis?
  b) Was randomization concealed?
   (“Allocation Concealment”)
    • Threat:
       – Inconvenience (difficult) and influence (clinicians
         want power to allocate pts)  inconsistent use
         of randomization
       – Randomization can fail
                Are the Results Valid?

  1) Did groups begin the study with a
    similar prognosis?
      b) Was randomization concealed?
          • Remedy:
              – Remote randomization: allocation made by persons
                not working with patients
              – Concealed randomization: e.g., sealed, opaque
                envelopes
              – Only done in 7-55% of trials1 and it can bias trials
                33-41%2
1. BMJ 2004 3;328:22-4. ACP J Club. 2000 Mar-Apr;132:A11 2. JAMA 1995; 273: 408-
412. Lancet. 1998;352:609-13.
        Are the Results Valid?
1) Did groups begin the study with a similar
  prognosis?
   c) Were patients analyzed in the groups
    to which they were randomized?
    (“Intention to Treat Analysis”)?
    • Threat:
       – Compliance correlates with outcome
       – Rates of cross-over or loss to follow-up correlate with
         outcome
              Are the Results Valid?
1) Did groups begin the study with a similar
  prognosis?
    c) Were patients analyzed in the groups to
       which they were randomized? (“Intention
       to Treat Analysis”)?
        • Remedy:
             – Intention to treat analysis  balance effect of
               compliance losses in both groups
             – Only 7-48% of articles do ITT (& less properly)
             – It can bias trials up to 51%

BMJ. 1999; 319: 670-4. J Fam Pract. 2002; 51: 969-71. BMJ 2003;326: 117-5. J Clin
           Are the Results Valid?
1) Did groups begin the study with a similar
  prognosis?
   d) Were the patients in treatment and
    control groups similar with respect to
    known prognostic factors?
    (Baseline Characteristics)
     • Threat:
        – Randomization does not always work, or may not be possible
           imbalance of prognostic factors
        Are the Results Valid?
1) Did groups begin the study with a similar
  prognosis?
   d) Were the patients in treatment and
    control groups similar with respect to
    known prognostic factors?
    (Baseline Characteristics)
     • Remedy:
       – Magnitude of imbalance?
       – Strength of poorly distributed factor(s)?
       – Are adjusted and unadjusted analyses the same?
        Are the Results Valid?
2) Did groups retain a similar prognosis
  after the study started?
  a) Were patients clinicians and analyzers
   aware of the group allocation? (Blinded)
    • Threat:
       – Placebo effects, expectation bias, Co-interventions,
    • Remedy:
       – Mask (blind) patients to exposure type
       – Inadequate blinding favours treatment by about
         17%

                   JAMA. 1995;273:408-12.
     Are the Results Valid?
2) Did groups retain a similar prognosis
  after the study started?
  d) Was follow-up complete?
    • Threat:
       – Losses are all persons whose status is not
         known at the end of the study.
    • Remedy:
       – Equivalent follow-up for the two groups?
       – Sensitivity analysis to losses
             Users Guides
• Are the results valid?
  – Prognosis similar at the beginning?
  – Prognosis similar at the end?
• What are the results?
  – Size of effect?
  – Precision of effect?
• How can I apply the results?
  – Patients like yours?
  – All important outcomes?
          What are the results?
    • How large was the treatment effect?
       Control and Experimental Event Rates?
       (often Placebo and Drug Event rates = CER & EER)

•                 Outcome
•                 +    -
•                              Risk of Outcome
•   Experimental a      b      EER = a / total in Exp
•   Control      c      d      CER = c / total in Control
        Hypothetical Post-MI Trial
            of 200 Patients
                      Mortality     Total Enrolled/
                                        Group
   Placebo               20               100

   Treatment             10              100

CER = # with outcome in Control / # who got Control
CER = 20 / 100 = 20%

EER = # with outcome in Experiment / # who got Exp
EER = 10 / 100 = 10%
     What are the results?
• How large was the treatment effect?
 What is the absolute risk reduction?


         •       The absolute risk reduction is the
     •       difference in risk between the control
             •       group and the treated group:
 •

                 •     ARR = CER-EER
  How can I apply the results?
• Are the likely benefits worth the harms and
  costs?
  What is the number needed to treat?


    •   The number needed to treat (NNT) is the
           •  number of patients requiring
    •   treatment for one outcome to be avoided:

        •     NNT = 1/ARR = 1/(CER-EER)
         RRR, ARR and NNT
e.g. Treating Hypertension for MI Risk
              (Framingham data)

  CER: 60 y.o. M, HDL 1.0, BP>160 = 20%

  EER: 60 y.o. M, HDL 1.0, BP<130 = 10%



 • ARR = CER – EER       ARR= 20% - 10% = 10%
 • RRR= ARR / CER        RRR= 10% / 20% = 50%
 • NNT = 1 / AR          NNT= 1 / 0.1 = 10
     RRR when event unlikely &
       RRR vs ARR (NNT)
  Common Diseases     Uncommon Diseases
• RRR = 50%         • RRR = 50%

•   CER = 20%       •   CER = 1%
•   EER = 10%       •   EER = 0.5%
•   ARR = 10%       •   ARR = 0.5%
•   NNT = 10        •   NNT = 200
    The Media: Women’s Health Initiative
• TV Evening News            • Actual Risk per 10,000
  (September 24, 2003)         (JAMA 288(3):321-33)


•   41% increase in stroke   •   8 strokes in 10,000 p. y.
•   29% increase in MI       •   7 MI’s in 10,000 p. y.
•   Double rate of VTE       •   18 VTE in 10,000 p. y.
•   26% in Breast Ca         •   8 Breast Ca in 10,000
•   No benefits mentioned    •   6 Colorectal Ca & 5 Hip
                                 Fractures Less/10,000
              Bottom Line
• Absolute Excess Adverse Events in the
  Global Index are 19 in 10,000 patient years
  OR
• 100 women on HRT for 5 years = 1 additional
  adverse event
        What are the results?
• How precise was the treatment effect?
    What is confidence interval on the results?


•    95% CI:
     the range of treatment effect estimates that includes
     the true treatment effect 95% of the time.
•    95% CI reflects:
     precision (reliability) of the result; statistical
     significance.
             Users Guides
• Are the results valid?
  – Prognosis similar at the beginning?
  – Prognosis similar at the end?
• What are the results?
  – Size of effect?
  – Precision of effect?
• How can I apply the results?
  – Patients like yours?
  – All important outcomes Considered?
  How can I apply the results?
• Were the study patients similar to your
  patients?
  – Do study inclusion criteria fit your patients’
    prognostic factors?
  – Compelling reasons for non-generalizability?
  – Dependent on subgroup analysis?
  How can I apply the results?
• Were all clinically important outcomes
  considered?
  – What were the primary and secondary
    endpoints?
  – Were surrogate outcomes used?
  – Were adverse outcomes considered?
 How can I apply the results?
• Are the likely benefits worth the
  harms and costs?
  Is the reduction in illness worth the cost and
   risk of harm?
     • NNT vs NNH?
     • Patients’ preferences and values?
     • Costs?
               Summary
• The goal of a RCT is to make exposed and
  unexposed similar in all respects other than
  the intervention... and to maintain that
  balance throughout.
• The size and precision of treatment effects
  determines the importance of the results.
• Who was enrolled and what was measured
  (outcomes) are the most important
  determinants of applicability
               The End




Much Thanks to: Rob Hayward & Tanya Voth, CCHE

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:5
posted:12/29/2011
language:English
pages:43