Statistical method combining direct and indirect trial comparisons across multiple interventions
A network meta-analysis pools evidence from studies that compare different pairs of interventions (e.g. cluster sets vs traditional sets, rest-redistribution vs traditional sets) to estimate both direct and indirect head-to-head effects within a single model. A Bayesian framework generates probability-based rankings and credible intervals rather than the frequentist p-values and confidence intervals used in pairwise meta-analysis. Rankings can look precise even when few studies directly compared the interventions in question, so sparse direct evidence should temper confidence in the ordering.
This guide was auto-drafted and is pending editorial review.