Title: Critically Appraising Research A Consumers Perspective Gary S. Gronseth, M.D. Department of Neurolog
1Critically Appraising ResearchA Consumers
PerspectiveGary S. Gronseth, M.D.Department
of NeurologyUniversity of KansasOctober 12,
2006
2Clinical
Outcomes
Basic
What causes disease and defining approaches to
diagnosis and treatment?
What works and what does it cost in real-world
settings?
How can disease be prevented or treated under
controlled settings?
3(No Transcript)
4Steroids
5(No Transcript)
6- The use of steroids for Bells palsy has become
the standard of care in the community.
7- The consequences of disfiguring facial weakness
are so devastating that the use of steroids is
mandatory.
8- Fallacious
- Irrelevant
- Rhetoric
- Psychological appeal
- Emotion-Driven
- Persuasion
9Common fallacious arguments in medicine
- Popularity
- Begging the question
- Irrelevant Outcomes
10Irrelevant Outcomes
11- Fallacious
- Irrelevant
- Rhetoric
- Psychological appeal
- Emotion-Driven
- Persuasion
12Decison
13Decision
14Decision
15Deductive ArgumentPrinciples
- Steroids
- Reduce inflammation
- Reduce compression within temporal bone
Good for hypothesis generation Often not
Convincing When convincing, no controversy
16Decision
17Decision
18Decision
19Decision
20Anecdotal argument
- The last patient I saw with Bells palsy who
didnt get steroids never got better.
21ExperienceCollections of Anecdotes
22(No Transcript)
23Expert opinion
- From a practical view point, treatment with
steroids is appropriate for the management of
Bells Palsy. - Bells Palsy remains without a proven
efficacious treatment.
Stankiewicz 1987
May 1975
24Limitations of Experience
More likely to remember anecdotes that
are Recent Extreme Support our preconceptions
25EvidenceCollections of Recorded Anecdotes
26Decision
27(No Transcript)
28The results of this study demonstrate no
statistically significant beneficial effect of
steroid therapy upon recovery from Bells
Palsy.P 0.32
May et. al. 1975
- Relative Rate
- Complete Recovery
- 0.92 (0.60 to 1.4)
Insufficiently powered to exclude a benefit
29The control group included mainly those patients
who had refused to take prednisilone, because
they were afraid of developing complications, and
those with a relative contraindication to
steroids (e.g. moderate hypertension)
Shafshak et al 1994
30The Literature is full of
- Fallacious arguments
- Contradictory expert opinions
- Misinterpretations of errors of chance
- Studies with a high risk of bias
31Find and analyze all of the evidence in a
transparent and systematic fashion
or find someone who already has.
32AAN clinical practice guidelines Above the fray
Franklin, Gary M. MD, MPH Zahn, Catherine A. MD
Neurology Volume 59(7) 8 October 2002 pp 975-976
- Avoid fallacious arguments
- Involve experts but dont allow their conclusions
to go beyond the evidence - Find all relevant studies
- Systematically assess and grade the quality of
the studies to avoid spin
33Clinical
Outcomes
Basic
What causes disease and defining approaches to
diagnosis and treatment?
What works and what does it cost in real-world
settings?
How can disease be prevented or treated under
controlled settings?
Patient-based
Pre-clinical
Organizational- based
Patient Population-based
Translation 2 Humans to Bedside/Curbside
Translation 1 Bench to Humans
34Practice Guideline
- Systematic and Transparent
Many Specialty Societies
Cochrane collaboration
AAN
35Consensus vs. Evidence
?
36Consensus-Based Evidence-Based
?
37Evidence-based Argument
- Question
- Evidence
- Conclusion
- Recommendation
38Do steroids work for Bells palsy?
For patients with acute Bells palsy, do oral
steroids given within the first week improve
facial functional outcomes?
39Do steroids improve facial functional recovery in
patients with for Bells palsy?
Patients Acute Bells Palsy Intervention Ora
l Steroids Given within 1 week Outcomes F
acial Function
40Should we do this?
- For patients with (suspected) Bells palsy
- Therapeutic does early use of oral steroids
improve facial outcomes?
41Should we do this?
- For patients with (suspected) Bells palsy
- Therapeutic does early use of oral steroids
improve facial outcomes? - Prognostic does facial CMAP testing identify
patients at increased risk for poor
outcomes?
42Should we do this?
- For patients with (suspected) Bells palsy
- Therapeutic does early use of oral steroids
improve facial outcomes? - Prognostic does facial CMAP testing identify
patients at increased risk for poor
outcomes? - Screening does head MRI identify patients
with treatable etiologies?
43The Answer
- Qualitative
- Yes, (No, Maybe, Cant tell) for patients with
condition A, intervention B is effective
(useful, superior) to improve outcome C. - QuantitativeHow much does the intervention
improve outcomes - CertaintyHow confident are we in the answer
44Evidence-based Argument
- Question
- Evidence
- Conclusion
- Recommendation
45(No Transcript)
46Evidence
- Find all relevant studies
- Assess
- Risk of Bias
- Results
- Random Error
- Generalizability
47Find the EvidenceThe Literature
48Literature Search
Apriori inclusion criteria.
49Literature SearchSteroids for Bells Palsy
230
Studies of outcomes in Bells Palsy patients some
getting steroids, some not getting steroids
8
50Steroid Studies
- Yr Author
- 94 Shafshak
- 93 Austin
- 87 Abraham
- 82 Brown
- 78 Wolf
- 76 May
- 72 Adour
- 54 Taverner
51How good is the Evidence?
52Grading the Evidence
The quality of evidence is always judged as it
pertains to the clinical question.
53AANs Evidence Classification Schemes
- Therapeutic
- Diagnostic/Prognostic Accuracy
- Screening
54AANClassification of evidence
Risk of Bias Low Moderately Low Moderately
High High
55Study
56AAN Grading the evidence
Controlled
57No Comparison Group
Poor Good
St
58No Comparison Group
- Natural History
- Placebo Response
- Regression to the Mean
59Confounding
Potential known and unknown confounding
differences between treated untreated group
St
-St
60ControlledMatching
Potential unknown confounding differences between
treated untreated group
St
M
-St
61ControlledRandomized Trial
No confounding differences between treated
untreated group
St
R
-St
...unless you are unlucky.
Or
62ControlledConcealed Allocation
Is the study truly randomized
St
R
-St
63Cross OversLost Randomization
St
R
-St
64Cross OversIntent-to-treat
St
R
-St
65Losses to Follow UpLost Randomization
St
R
-St
66Losses to Follow UpIntent to Treat
St
R
-St
67AAN Grading the evidence
Controlled randomized matched comparative unc
ontrolled
68AAN Grading the evidence
Controlled randomized matched comparative unc
ontrolled
Masking
69Masking
St
-St
70Masking
- Patient
- Treating Providers
- Outcome assessors
71AAN Grading the evidence
Controlled randomized matched comparative unc
ontrolled
Masking masked outcome masked
outcome independent non-independent
72Steroid Studies
- Author Random. Comp FU Masked
- May R 100 Y
- Taverner R 100 Y
- Brown R 100 Y
- Austin R 71 Y
- Shafshak NR 100 Y
- Wolf R 100 N
- Adour NR 85 N
- Abraham NR 100 N
- Class
- I
- I
- II
- II
- III
- IV
- IV
- IV
73What does the evidence show?
- The studys results
- (Not the authors conclusions)
74Effect of SteroidsRate Difference
Outcome
Treatment
Poor
Good
Steroids
15
85
No Steroids
33
67
75Categorical Outcomes are the most clinically
relevant
76Measures of Effect
- Therapeutic questions
- Rate Difference
- Relative Rates
- Odds Ratios
- Diagnostic questions
- Sensitivity/Specificity
- Likelihood Ratios
- Screening questions
- Yields
77Rate Difference Good RecoverySteroids vs No
Steroids
60
I
II
III
IV
40
20
Rate Difference
0
-20
-40
May
Taverner
Brown
Austin
Shafshak
Wolf
Adour
78Evidence-based Argument
- Question
- Evidence
- Conclusion
- Recommendation
79Synthesis
60
I
II
III
IV
40
20
Rate Difference
0
-20
-40
May
Taverner
Brown
Austin
Shafshak
Wolf
Adour
80Decrease Random Error (Include all studies)
Decrease Risk of Bias (Exclude studies with a
higher risk of Bias)
81Cumulative Meta-analysis
- Do not Consider Class IV evidence
- Iteratively combine results of studies with the
least risk of bias until you arrive at a
sufficiently precise measure of effect or you run
out of studies - Grade the strength of the conclusion based on the
most bias prone study used to determine the effect
82Combine Class I evidence
60
I
40
20
Rate Difference
0
-20
-40
Pooled
May
Taverner
83Hypothetical
60
I
40
20
Rate Difference
0
For patients with Bells palsy steroids are
established to increase the likelihood of good
facial functional recovery by about 12 percent.
-20
-40
Pooled
May
Taverner
84Combine Class I and Class II evidence
60
I
II
40
20
Rate Difference
0
For patients with Bells palsy steroids probably
increase the likelihood of good facial functional
recovery by about 12 percent.
-20
-40
Pooled
May
Taverner
Brown
Austin
85Hypothetical
60
I
II
III
40
20
Rate Difference
0
For patients with Bells palsy there is
insufficient evidence to conclude that steroids
increase or decrease the probability of good
facial functional recovery.
-20
-40
Pooled
May
Taverner
Brown
Austin
Shafshak
86Conclusions
- For patients with Bells palsy...
- Based on consistent results from two class I and
two class II studies, steroids probably increase
the likelihood of good facial functional recovery
by about 12 percent.
87Evidence-based Argument
- Question
- Evidence
- Conclusion
- Recommendation
88Relative Value
Benefits
Risks
89Relative Value
Patient Preferences Consensus
Value of Steroids is Large
Good Facial Function Permanent
Steroid Side Effects Temporary
90Strength of Recommendation
Strength of Evidence 2 Class I 2 Class II 2
Class III Value of Benefit/Risk Large Moderate Small Too
close to call
Level of recommend. A. Should be offered B.
Should be considered C. May be considered U. No
recommendation
91Strength of Recommendation
Strength of Evidence 2 Class I 2 Class II 2
Class III Value of Benefit/Risk Large Moderate Small Too
close to call
Level of recommend. A. Should be offered B.
Should be considered C. May be considered U. No
recommendation
92Recommendation
- For patients with Bells palsy...
- oral steroids should be considered to increase
the probability of good facial functional
recovery. (Level B)
93Level A Recommendation
94Level C Recommendation
95(No Transcript)