For a particular multi-arm study, the intervention groups of relevance to a systematic review are all those that could be included in a pair-wise comparison of intervention groups that, if investigated alone, would meet the criteria for including studies in the review. For example, a review addressing only a comparison of ‘nicotine replacement therapy versus placebo’ for smoking cessation might identify a study comparing ‘nicotine gum versus behavioural therapy versus placebo gum’. Of the three possible pair-wise comparisons of interventions, only one (‘nicotine gum versus placebo gum’) addresses the review objective, and no comparison involving behavioural therapy does. Thus, the behavioural therapy group is not relevant to the review. However, if the study had compared ‘nicotine gum plus behavioural therapy versus behavioural therapy plus placebo gum versus placebo gum alone’, then a comparison of the first two interventions might be considered relevant and the placebo gum group not.
As an example of multiple control groups, a review addressing the comparison ‘acupuncture versus no acupuncture’ might identify a study comparing ‘acupuncture versus sham acupuncture versus no intervention’. The review authors would ask whether, on the one hand, a study of ‘acupuncture versus sham acupuncture’ would be included in the review and, on the other hand, a study of ‘acupuncture versus no intervention’ would be included. If both of them would, then all three intervention groups of the study are relevant to the review.
As a general rule, and to avoid any confusion for the reader over the identity and nature of each study, it is recommended that all intervention groups of a multi-intervention study be mentioned in the table of ‘Characteristics of included studies’, either in the ‘Interventions’ cell or the ‘Notes’ cell. However, it is necessary to provide detailed descriptions of only the intervention groups relevant to the review, and only these groups should be used in analyses.
The same considerations of relevance apply when determining which intervention groups of a study should be included in a particular meta-analysis. Each meta-analysis addresses only a single pair-wise comparison, so review authors should consider whether a study of each possible pair-wise comparison of interventions in the study would be eligible for the meta-analysis. To draw the distinction between the review-level decision and the meta-analysis-level decision consider a review of ‘nicotine therapy versus placebo or other comparators’. All intervention groups of a study of ‘nicotine gum versus behavioural therapy versus placebo gum’ might be relevant to the review. However, the presence of multiple interventions may not pose any problem for meta-analyses, since it is likely that ‘nicotine gum versus placebo gum’, and ‘nicotine gum versus behavioural therapy’ would be addressed in different meta-analyses. Conversely, all groups of the study of ‘acupuncture versus sham acupuncture versus no intervention’ might be considered eligible for the same meta-analysis, if the meta-analysis would include a study of ‘acupuncture versus sham acupuncture’ and a study of ‘acupuncture versus no intervention’. We describe methods for dealing with the latter situation in Section 16.5.4.