Coding
Meta-analysis toolkit
Every systematic review asks a different question, and the right analysis depends on the outcome, the study designs and how the data were reported. This toolkit collects the methods I use across my published work, written as R code that can be reused in another review by swapping in its data.
- Matched to the question
- Explained line by line
- Ready to reuse
Pooling the outcomes
- 01Pairwise: binary outcomesRisk and odds ratios for events such as complications or mortality
- 02Pairwise: continuous outcomesMean and standardized mean differences for stay, operative time or pain
- 03Single-arm: binary outcomesPooled proportions, such as complication rates, without a comparison group
- 04Single-arm: continuous outcomesPooled means, such as operative time, without a comparison group
Testing how robust the result is
- 05Sensitivity analysisLeave-one-out and Baujat plots
- 06Subgroup analysisPre-specified subgroups and tests for subgroup differences
- 07Meta-regressionEffect sizes regressed on continuous study-level covariates, such as dose or year
- 08Publication biasFunnel plots and the Egger, Begg and Harbord tests for small-study effects
When the data call for a different design
- 09IPD from Kaplan-Meier curvesReconstructed patient-level survival data, for comparative and single-arm reviews
- 10IPD meta-analysisOne- and two-stage analysis of individual participant dataIn progress
- 11Dose-responseOutcome as a function of a continuous exposure, such as hospital volume
- 12Network meta-analysisSeveral treatments compared at once, combining direct and indirect evidenceIn progress
- 13Bayesian meta-analysisProbability statements and prior sensitivity when trials are fewIn progress