Ranking treatments that were never compared head-to-head
9 October 2026 · Cohira team
Guidelines rarely ask "is A better than placebo?" They ask "which of the five options should we use?" Most of those five have never met in a trial.
Indirect evidence
If A beat placebo and B beat placebo, the two trials imply a comparison of A with B. A network meta-analysis formalises that logic across every trial at once, keeping the randomisation inside each trial intact.
The assumptions that matter
Transitivity: the trials must be similar enough to connect. Consistency: direct and indirect evidence should agree, and node-splitting tests whether they do. A network that fails these checks produces rankings that mislead.
Reading a ranking
A P-score or SUCRA value is a probability, not a verdict. Rankings belong next to their intervals and the certainty of the evidence, which is why CINeMA assessment is part of the standard package.