Handle a Shared Comparator in a Multiarm Trial

Draw the arms before extracting effects

A trial with two active arms and one control can produce two contrasts, but those contrasts share the same control participants. Entering both into an ordinary pairwise meta-analysis as independent comparisons can give the trial excessive influence. Begin with an arm table containing randomized and analyzed counts, intervention definitions and eligible outcomes.

Cochrane’s chapter on trial variants describes strategies for trials with multiple intervention groups. Choose the strategy from the review question rather than from which arrangement produces a narrower interval.

Decide whether arms represent the same comparison

If both active arms meet the same intervention definition and combining them is scientifically sensible, create one active group before calculating the pairwise effect. Preserve the original arm data and the combination rule. Combining different doses may be inappropriate when dose is itself the question.

If the arms must remain separate, use a synthesis method that handles their dependence, or apply an explicitly justified approximation appropriate to the effect measure. Do not simply duplicate the full control count. Splitting a control group changes the representation and precision of the comparisons; it does not manufacture independent participants.

Check a hypothetical binary example

Imagine active arm A has 10 events among 50 participants, active arm B has 15 among 50, and control has 20 among 100. If A and B are eligible to combine, the active total is 25 events among 100 participants. The risk ratio against control is (25/100)/(20/100) = 1.25.

Analyzing A versus control and B versus control with all 100 controls entered twice would represent 300 participant entries even though only 200 people were randomized. That discrepancy is an audit signal, not a mathematical proof that every specialized multiarm model is wrong. A model can retain both contrasts correctly if it represents their covariance.

Preserve the scientific interpretation

For continuous outcomes, combining means requires group sizes and within- and between-group variability; averaging SDs is not the right calculation. Use the existing combined-summary guide when its independent-group assumptions fit.

Record excluded arms with the reason, selected combination, effect direction and resulting denominators. Check subgroup analyses for a reappearance of the same shared control problem. The final extraction should make it possible to reproduce both the original arm data and the chosen analysis representation without guessing which participants were reused.

Source verification: 6 November 2026. Examples are hypothetical.