Title: METAANALYSES: APPROPRIATE GROWTH OR MALIGNANT TUMOR
1META-ANALYSES APPROPRIATE GROWTH OR MALIGNANT
TUMOR?
Ian Shrier MD, PhD, Dip Sport Med (FACSM)
- Associate Professor, Dept of Fam Med, McGill
University - Centre for Clinical Epidemiology and Community
Studies, SMBD-Jewish General Hospital and McGill
University - Past-president, Canadian Academy of Sport Medicine
2GROWTH VS QUALITY?
- Cochrane Collaboration 10 points
- collaboration
- building on the enthusiasm of individuals
- avoiding duplication
- minimizing bias
- keeping up to date
- striving for relevance
- promoting access
- ensuring quality
- continuity
- enabling wide participation
3GROWTH VS QUALITY?
- Cochrane Collaboration 10 points
- collaboration
- building on the enthusiasm of individuals
- avoiding duplication
- minimizing bias
- keeping up to date
- striving for relevance
- promoting access
- ensuring quality
- continuity
- enabling wide participation
Growth
4DATA ENTRY
5OBJECTIVE?
A Meta-analysis is an Observational Study,
Subject to Selection Bias, Confounding, and
Measurement Error Like Any Other!
Computer programs and standardized scoring make
it very easy to make mistakes that are difficult
for the reviewer to pick up.
Platt, personal communication
6OVERVIEW
- Inclusion Criteria / Data abstraction
- Analyses
- Interpretation of Evidence
7INCLUSION CRITERIA - SEARCH
- Search Strategy
- Does Vitamin C prevent death due to cancer in
the general population
8INCLUSION CRITERIA RCTs?
9INCLUSION CRITERIA RCTs?
Furlan, 2006
10DATA ABSTRACTION
- Peer-review for stretching article
- The RR 0.79 was based on the article in Pope
2000.
- Actual article
- The X article Cox regression differing
exposure times for different subjects.
univariate hazard ratio of 0.95 (CI 0.77,1.18)
- multivariate analysis HR1.04 (0.82,1.33)
11OVERVIEW
- Inclusion Criteria / Data abstraction
- Analyses
- Interpretation of Evidence
12ANALYSES
- Effect Modifiers
- Study-level confounders
- Subject-level confounders
- Publication Bias
- Papers with significant results are more likely
to be published (i.e. selected) - Reporting Bias
- Outcomes that are statistically significant are
more likely to be reported in publications (i.e.
selected)
13ANALYSES
14ANALYSES
15OVERVIEW
- Inclusion Criteria / Data abstraction
- Analyses
- Interpretation of Evidence
16GRADE SCALE DECISIONS
Recommendation
Balance Benefit vs Harm
Rater
1
2
- Reasons
- Lack of information in summaries
- Missing baseline risk
- What is sparse information?
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Atkins. BMC Health Services Research 2005
17INTERPRETATIONS
I believe magnesium has now been shown to be
beneficial for patients during the post-MI period
(SD-SA)
1-23
1-20
1-10
1-5
1-3
1
RCTs
N
Fixed OR
Rand. OR
I2
Rev 1
Rev 2
Rev 3
Rev 4
Rev 5
Rev 6
Rev 7
Rev 8
18CONCLUSION
- The explosion of meta-analyses has occurred
because of apparently simple recipes - However, a meta-analysis is an observational
study with potential errors at many levels - Authors of meta-analyses need to be appropriately
trained - Readers need to be appropriately skeptical
- The benefits of meta-analyses may be lost if
current growth remains unrestricted and becomes a
malignancy
19(No Transcript)