METAANALYSES: APPROPRIATE GROWTH OR MALIGNANT TUMOR

1
META-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

2
GROWTH 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

3
GROWTH 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
4
DATA ENTRY
5
OBJECTIVE?
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
6
OVERVIEW

• Inclusion Criteria / Data abstraction
• Analyses
• Interpretation of Evidence

7
INCLUSION CRITERIA - SEARCH

• Search Strategy
• Does Vitamin C prevent death due to cancer in
  the general population

8
INCLUSION CRITERIA RCTs?
9
INCLUSION CRITERIA RCTs?
Furlan, 2006
10
DATA 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)

11
OVERVIEW

• Inclusion Criteria / Data abstraction
• Analyses
• Interpretation of Evidence

12
ANALYSES

• 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)

13
ANALYSES
14
ANALYSES
15
OVERVIEW

• Inclusion Criteria / Data abstraction
• Analyses
• Interpretation of Evidence

16
GRADE 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
17
INTERPRETATIONS
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
18
CONCLUSION

• 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)