Meta-analysis Flashcards
Combining the outcomes from different studies
Fixed / random effects model
Fixed effects model
Assumes all the studies share the same common treatment effect (homogeneous)
Can only be applied when heterogeneity can safely be excluded by testing for it
Fixed effect statistics
Mantel-Haenszel ratio: useful even when wide differences exist between individual studies in ratios of the size of two groups
Peto ratio: mostly restricted to reviewing RCTs as it can produce biased results in unequal groups
Random effects analysis
Assumes each study shows a different effect which are normally distributed around true mean
Gives proportionally greater weight to smaller studies
Susceptible to publication bias and results in wider less precise confidence intervals
Significant heterogeneity
May be judged graphically or measured statistically
Heterogeneity graphs
L’Abbe plot - modified scatter plot. CER is plotted against EER.
Galbraith plot - alternative to a forest plot. Precision (1/SE) on x axis, effect size / z score / standard normal deviate (odds ratio / SE) plotted on y axis
Forest plots (blobbograms)
Chi-square tests of heterogeneity
Can test for significance in heterogeneity but cannot measure it
Q test or I2 test
Cochran’s Q
Quantify heterogeneity
Weighted sum of squared differences between individual study effects and the pooled effect across studies
I2 statistic
The percentage of variation across studies that are due to heterogeneity rather than chance
Meta-regression and sub-group analysis
Can throw light on the causes of heterogeneity
Funnel plot
Commonest method to detect publication bias
Failsafe N (file-drawer number)
The number of zero-effect studies that would be required to nullify the mean effect seen in a meta-analysis
Effect sizes for dichotomous variables
RR or OR
RR is generally used in cochrane reviews
Effect sizes in continuous variables
Simple difference between the mean values
Standardised difference between the mean values - calculated using cohen’s d or hedges’ g (avoiding a bias in cohen’s d seen with small number of subjects)
Standard format for meta-analysis
QUORUM statement
Standard format for RCTs
CONSORT statement
Evaluating meta-analyses
CER = n/N (for control group)
EER = n/N (for other arm)
