Meta-analysis

Question 1

What is a Systematic Review?

Answer 1

Scientific Scientific strategy to

• assemble all relevant studies on a topic
• summarize summarize results results of identified studies
• appraise appraise quality quality of identified studies
• Take the results from each study and produce summary quantities produce summary quantities (answers), that are unbiased (as far as possible) and reproducible

Question 2

Inclusion criteria

Answer 2

Population

intervention/treatment

comparator

outcome measure

study design

Question 3

What is publication bias?

Answer 3

Tendency for negative studies not to be reported since:

• journals don’t want to publish
• researchers don’t want to write up
• Nb often smaller studies often smaller studies

Publication bias can be investigated by

• Searching trial registers
• Applying statistical methods e.g. The funnel plot

Question 4

Symmetrical funnel plot

have no publication bias

Answer 4

Each study is a dot

larger studies at the top

smaller studies have more spread out estimates

Question 5

Meta-analysis

Answer 5

Combination of results from studies using Meta-analysis analysis specific statistical methods

• Include comparable comparable studies (e.g. In terms of population, intervention and setting)
• Combine the results of all included studies into one summary statistic summary statistic
• Investigate variation of treatment effects among studies (known as “heterogeneity” = “lack of consistency”)

Question 6

Relative Risk

Answer 6

RR = Risk of outcome in the treatment group

/

Risk of outcome in the control group

• The summary statistic “odds ratio” may be used instead of RR in meta used instead of RR in meta-analysis analysis

Question 7

Addressing variation of treatment effects in meta effects in meta-analysis analysis

Answer 7

Null hypothesis: all studies estimate the same underlying population quantity (e.g. Relative Risk)

• Some random variation in study estimates is expected due to small random samples
• chi-squared test for heterogeneity indicates squared test for heterogeneity indicates whether the variation is significantly greater (p<0.05)  But is it is not significant (p>0.05, or “NS”) then there is no evidence to disbelieve that they are consistent.
• Consistent evidence adds confidence.
• Heterogeneous results need to examined further

Question 8

Evidence based medicine

Answer 8

Evidence-based medicine is the based medicine is the process of systematically finding and appraising research in order to make clinical decisions based on the best available evidence