CMOR Lunch’n’Learn
6 July 2023
Ross Wilson
From Cochrane:
A systematic review is a study that
“attempts to identify, appraise, and synthesize all the empirical evidence that meets pre-specified eligibility criteria to answer a specific research question”
Meta-analysis is the statistical combination of results from two or more separate studies
Do your systematic review well:
Define the review aims & scope (PICO)
Ensure your search strategy, screening, etc. is effective
Identify & summarise study characteristics
Decide whether studies are ‘similar enough’ to be grouped & synthesized
Extract (or calculate) a summary statistic for each study
Calculate (weighted) average effect across all studies
Calculate the standard error of the summary effect (to derive confidence intervals and/or p-values)
Estimate heterogeneity between studies
Summarise findings graphically (usually with a forest plot)
\[\text{Weighted average} = \frac{\sum{Y_i W_i}}{\sum{W_i}}\]
where \(W_i = (1 / SE_i)^2\)
\[\text{Standard error} = \frac{1}{\sqrt{\sum{W_i}}}\]
Heterogeneity is measured by
\(Q = \sum{W_i (Y_i - \overline{Y})^2}\)
and
\(I^2 = \text{max}\left\{\frac{Q - (k - 1)}{Q}, 0\right\}\)
Q is a test statistic following a chi-squared distribution, and I2 is interpreted as the proportion of total variance in study estimates due to heterogeneity rather than sampling error
Meta-analyses are usually illustrated using a forest plot
A forest plot displays effect estimates for both individual studies and the overall meta-analysis result
Each study is represented by a square (usually) at the point estimate, and horizontal lines extending to bounds of the confidence interval
The summary result is presented as a diamond at the bottom, centred on the meta-analysis point estimate and with width showing the confidence interval
metafor
is a comprehensive package with functions for converting published results to common summary statistics, fitting inverse-variance weighted and related models, forest and funnel plots, heterogeneity measures, and diagnostic statistics (among others)Cochrane Handbook for Systematic Reviews of Interventions
https://training.cochrane.org/handbook/current
Egger M, Smith GD, Altman DG. Systematic Reviews in Health Care: Meta-Analysis in Context, 2nd Ed. London, UK: BMJ Publishing Group. 2001.
https://doi.org/10.1002/9780470693926
Jin Z-C, Zhou X-H, He J. Statistical methods for dealing with publication bias in meta-analysis. Statistics in Medicine 2015;34(2):343-360
https://doi.org/10.1002/sim.6342
metafor
R package https://wviechtb.github.io/metafor/