introduction to meta-analysishlm.tzuchi.com.tw/epi-stat/images/class/2016/2016_class4.pdf• karl...
TRANSCRIPT
Introduction to Meta-analysis
What does a statistical significant
or non-significant result mean?
Type I & II error
Statistical power
• Many (most) clinical trials do not have enough
statistical power to detect clinically important
effects.
Research synthesis
• Karl Pearson synthesized data on efficacy of typhoid
vaccination (1904)
– “Many of the groups… are far too small to allow of any
definite opinion being formed at all, having regard to the
size of the probable error involved.”
• "It is surely a great criticism of our profession that we have
not organised a critical summary, by specialty or subspecialty,
adapted periodically, of all relevant randomized controlled
trials”-- Archibald Cochrane
Research synthesis
• Meta-analysis is a statistical technique for
combining the findings from independent
studies.
• Meta-analysis is most often used to assess the
clinical effectiveness of healthcare
interventions; it does this by combining data
from two or more randomised control trials.
Hierarchy of evidence
• Systematic reviews and meta-analyses of randomized controlled trials
• Randomized controlled trials
• Non-randomized intervention studies
• Cohort studies
• Case control studies
• Case series
• Case report
• Expert opinion
Meta-analyses indexed in PubMed
Search: meta-analysis(Publication Type)
Types of reviews
• Traditional, narrative review
• Overview
• Systematic review
• Meta-analysis
• Pooled analysis
Issues with traditional, narrative reviews
• Subjective, prone to bias and error
• Inadequate literature search strategy
• Unclear inclusion/exclusion criteria
• No description of the methods used by the review
• Lack of consistent methodology to draw consistent qualitative or
quantitative conclusions
– Replication or validation of review conclusions issues
– Conflicting conclusions based on the same evidence
Mammography Screening
• Gotzsche & Olsen [Nordic Cochrane Centre] concluded that
“screening for breast cancer with mammography is unjustified”
(Gotzsche & Olsen. Lancet 2000;355:129 – 34. & Lancet
2001;358:1340-42) RR =1.00 (95% CI 0.96, 1.05)
• The US Preventive Services Task Force reviewed the same set of trials (2001):
– The USPSTF recommends screening mammography, with or without clinical breast examination, every 1-2 years for women aged 40 and older.” RR=0.77 (95 percent CI, 0.67-0.89)
Meta-analyses Steps
• Formulate question & write a protocol
• Retrieve study articles
• Assess study quality: inclusion/exclusion
• Extract data
• Analyze data
• Conclusions and reports
P - Patient population
I - Intervention or exposure
C –Comparison group
O - Outcome compared
Counsell C. Formulating questions and locating primary studies for inclusion in systematic reviews. Ann Intern Med 1997;127:380-7
Formulate question
Issues with meta-analyses
• Meta-analysis should be part of a systematic review
• Poor quality studies should be excluded
• Heterogeneity of studies should be addressed
• Be aware of biases
– Publication bias
– Time lag bias
– Duplicate publication bias
– Language bias
– Outcome reporting bias
Heterogeneity
• Heterogeneity index
–Cochrane's Q and I2 statistic
• Non-homogeneous studies should not be
pooled and analyzed together.
Meta-analyses Evaluation
• The quality of meta-analyses are highly
variable
• Cochrane reviews, on average, may be better
reported than journal reviews
• The quality of meta-analyses cannot be taken
for granted
Appraising Meta-analyses
Examples sources of references
• http://media.wix.com/ugd/dded87_a02ff2e3445f495
2992d5a96ca562576.pdf
• Users’ Guides to the Medical Literature 3rd ed.
(2015)
Effect Size
• The effect size
– the “dependent variable” of meta-analyses
– standardizes findings across studies
• Different meta-analyses may use different effect size
indices
– Central Tendency Research
– Pre-Post Contrasts
– Group Contrasts
– Association Between Variables
– Regression analyses
Summarizing studies
• A weighted average according to the amount of information that each trail contains is more reasonable
• Weight by:
– Size (sample size (n))
– Event rate
– Homogeneity (inverse of the variance)
– Quality
– Other factors…
A two-stage process
• A summary statistic for each study
• A summary (pooled) treatment effect
estimate as a weighted average of the
treatment effects estimated in the individual
studies.
What to look for in a systematic review?
• A clearly defined question
• Comprehensive literature search strategy
• Explicit, reproducible inclusion/exclusion criteria
• Quality assessment of primary studies
• Reproducible data extraction procedure
• Data analysis and reporting – Exploration of heterogeneity, publication bias, etc.
• Limitations of evidence
• Data Interpretation
• Implications for patient care and future research
Conclusions
• Rosiglitazone was associated with a significant
increase in the risk of myocardial infarction
and with an increase in the risk of death from
cardiovascular causes that had borderline
significance.
The End