Systematic Reviews, Meta-Analysis And Heterogeneity Flashcards
What are the key aspects of a systematic review?
- EXPLICIT statements are made in regards to study design and outcomes
- TRANSPARENT methodology
- REPRODUCIBLE by anyone
Define ‘systematic review’
An overview of primary studies that uses explicit, transparent and reproducible methods
Define ‘meta analysis’
A quantitative synthesis of the results of two or more primary studies that addressed the same hypothesis in the same way
What is the purpose of a meta analysis?
- Quantify effect sizes (OR and 95% CI) of a large number of studies as a POOLED ESTIMATE
- Systematically collate and facilitate the synthesis of a large number of study results
- Reduce problems in interpretation due to variations in sampling
What model is used to display a meta analysis?
Forest plot
Explain how a pooled estimate is calculated in a meta analysis
- Odds ratio and 95% CI calculated for each study in meta analysis
- Combined to give a pooled OR and 95% CI using statistical computer
- Studies are WEIGHTED according to their size and uncertainty in OR (error factor)
What is the effect of error factor on the weighting of studies in a meta analysis?
Smaller error factor means LARGER WEIGHTING GIVEN
Describe what the following mean in a forest plot: 1 - square 2 - dotted line 3 - diamond 4 - solid line
1 - shows each individual odds ratio with their 95% CI (larger square means greater weighting is given)
2 - pooled odds ratio for all studies
3 - pooled 95% CI (points on diamond represent lower and upper limits)
4 - null value
What is heterogeneity?
- % measure of how similar the studies in the meta analysis are
- Assesses whether the difference in the individual trials was due to random variation (within studies)
How could you tell if the differences between studies in a meta analysis was due to random variation with the studies themselves?
- If p
How could you model for variation between study designs in a meta analysis?
- Fixed effect model vs Random effects model
- Often a WEAK test for heterogeneity
What is the difference between a fixed effect model and a random effects model?
- Fixed effect model assumes studies are ESTIMATING THE EXACT SAME EFFECT SIZE
- Random effects model assumes the studies are ESTIMATING SIMILAR, BUT NOT THE SAME EFFECT SIZE
How does the weighting of studies differ in a random effects model compared to a fixed effect model?
- Weighting is MORE EQUAL
- However there is a WIDER 95% CI
Describe a way in which you could test for heterogeneity without using a fixed/random effects model approach
- SUB GROUP ANALYSIS
- Stratification of studies by study characteristics and participant profile
Describe how the quality of studies can vary
- Some studies are more prone to confounding and bias
- Poor study design and protocol
- Poor protocol implementation
