10 - Review of Evidence Flashcards
What is biological plausibility?
- Part of the Bradford Hill’s criteria for a causal link
- However it is limited by current knowledge e.g Vit C and scurvy, and just because biologically plausible does not guarentee an association or effect
What is evidence based medicine?
- Process of turning clinical probs into qs then systematically locating, appraising and using research findings as basis to make decisions
What is the hierarcy of evidence?
A good RCT would not be higher than a poor systematic review
With reference to drug action, what does the statement ‘scientifically proven to work’ mean?
What is the difference between absolute risk and relative risk?
We use relative risk
Work out the relative risk and the number needed to harm using this scenario.
What is the purpose of a placebo?
Active treatment has element of placebo in it too
Why do you need to tell people in a clinical trial that they may recieve a placebo and what issues does this cause?
- Ethical
- Removes some of the placebo effect
What is a systematic review?
- Where you bring together a large group of studies which are all testing the same hypothesis
- All studies must answer the same question, be the same type of study (RCT), the same type of participants (age/gender), same outcome measures and same quality of studies
- Review is explicit, transparent and reproducible
A study has to show PICOS to be used in a systematic review? What is PICOS?
What is the difference between meta analysis and a systematic review?
Most systematic reviews include meta analysis
What are the four steps of evidence based healthcare?
What is the purpose of meta analysis?
- Can’t be done when there are concerns about the studies in a systematic review, e.g too different
- Used to collate study results so you can make statements about the information as it is based on a larger group
- Reduces problems in interpreting data
What is the methodology of a meta analysis?
- All studies must be standardised in the same way
- Take the odds ratio of all the studies and combine them to get one pooled estimate odds ratio
- Studies are then weighted according to their size and uncertainty in the odds ratio (based on how big their CI is - narrower means greater weighting)
Calculate the odds ratio of surviving when given aspirin following an MI compared to a placebo.
What is a Forest plot?
- Graphical display of estimated results from a number of scientific studies addressing the same question, along with the overall results
- Diamond is pooled estimate with the tips representing pooled 95% CI
- Solid line is null hypothesis OR
How would you interpret this forest plot for the study of taking aspirin after an MI?
What are the problems with meta analysis?
Differences in the heterogeneity can be overcome by using models for variation
What are the two different ways of modelling for variation?
Fixed effect model: (most common Forest Plot) assumes the studies are estimating exactly the same true effect size with no variation or variation due to random error
Random effect model: assumes the studies are estimating similar but not the same true effect size. this changes the weighting of studies on the Forrest plot to take into account variation. The odds ratio is often similar but the CI is often wider as theres more uncertainity. Takes into account variation but doesn’t explain what is causing it
What are the main differences in the Forrest plots of random effect and fixed effect models?
- Random has more evenly weighted boxes
- Random has wider CI
- Point estimate (odds ratio) is often similar
What does a test for heterogeneity in the meta analysis mean?
If it is a low value <0.05 this provides evidence of heterogeneity of intervention effects (variation in effect estimates beyond chance.
- If p-value of the test is low we can reject the hypothesis and heterogeneity is present. Use random effect model
If random effect model can accout for variation but not explain the variation, what method can be used to explain the variation?
Subgroup Analysis
- Look at study characteristics e.g follow up lengths, poor design, the study itself (e.g cohort)
- Look at patient profiles e.g sex, age, recruitment criteria
How can you assess the quality of the studies involved in a systematic review/meta analysis?
- Either define a basic quality standard (e.g good randomisation allocation methods, good blinding, ITT analysis, appropriate statistical analysis) and only include studies that satisfy this criteria
- Score each study for its quality and then incorporate the quality score into the weighting and use subgroup analysis to explore differences between high and low quality studies
Why is there publication bias when selecting studies to review and what are the consequences of this bias?
Publication bias: studies which are statistically signigicant are more likely to be published than those that aren’t. Meta analysis or systematic review can be flawed by this
How can we identify if publication bias is involved in a meta analysis?
- Check meta analysis protocol for method of identification of studies, e.g should include searching and identification of unpublished studies
- Funnel plot to plot results of identified studies against a measure of their size
- Use a stats test for publication bias - but these are often weak
When looking at a funnel plot, how can you tell there is publication bias?
- Likely to exist if there are few small studies with results indicating small or negative measure of effect
- Smaller studies can be expected to vary further from the central effect size
- ASYMMETRY
