Abstract
Abstract
RevMan is a difficult to use program with missing functionality and an old style command line interface. Comprehensive Meta-Analysis (CMA) in comparison costs money and also has an old interface. Both of these do not have advanced statistical flexibility. Finally command-line environments like R (metafor), require coding expertise but are still the gold standard. Pairwise Pro v2.2 is a new type of living html file with no installs or external dependencies. By comparing it against established tools such as R, we have shown Pairwise Pro enables advanced Bayesian and cross-disciplinary analyses—capabilities usually that can only be done in R—within a zero-dependency, browser-based interface. It is as good as standard GUIs by offering Decision-Driven Meta-Analysis (DDMA) probabilities.
Introduction: This is the first of a new wave of meta-analysis tools that can take data and analyse it without external dependencies quickly and to high degree of accuracy. They can be opened up 20 years later and will give the same result.
Methodological Comparison Standard GUIs use DerSimonian-Laird estimators and standard Wald confidence intervals but Pairwise Pro v2.2, validated against R’s metafor package, defaults to the more robust Restricted Maximum Likelihood (REML) estimator and Hartung-Knapp-Sidik-Jonkman (HKSJ) adjustments. This has been independently validated against Metafor. Many tools with large footprints lack native Bayesian capabilities. Pairwise Pro integrates a full MCMC sampler, allowing users to just using an html file generate posterior distributions and credible intervals. It will also give you R code to rerun this in R if you want.
We also have the Particle Data Group (PDG) scale factor from physics to handle variance inflation and Winner's Curse corrections from genomics. We use the binary "significant/non-significant" output of others with a DDMA framework, by giving the specific probability of clinical benefit and harm.
Conclusion: Pairwise Pro v2.2 offers the ability of R with the usability of RevMan. By allowing advanced metrics like heterogeneity variance decomposition and Bayesian inference, it represents a forward path for evidence synthesis. This will ensure that robust, decision-grade analysis is easy to use.
Articles in Gnosis are published under the Creative Commons Attribution 4.0 International (CC BY 4.0) licence. Authors retain copyright.