Healthcare decisions should be based on all relevant evidence.1 Usually, this is provided by randomised controlled trials (RCTs) comparing two or more interventions for a condition affecting a target population of interest, although other forms of evidence can be considered.1 2 When more than one study is available, meta-analysis can be used to combine multiple treatment effects and obtain an overall estimate of the effect in the target population. To assess clinical effectiveness, evidence from RCTs is typically used and relative treatment effects estimated in individual trials are pooled using methods that preserve within-trial randomisation. However, for the majority of health conditions, there are more than two interventions of interest. In such cases, performing multiple pairwise meta-analyses (comparing interventions two at a time) or lumping every active intervention to be compared with a ‘control’ is of limited use for decision-making and does not allow for coherent and transparent decisions. Decisions involving 20 interventions are not uncommon.3–6 The number of pairwise comparisons required to make a decision between 3 interventions is 3, with 5 interventions it is 10, with 10 interventions it is 45 and with 41 interventions4 it is 820. Clearly, not all comparisons will have been carried out in RCTs but looking at multiple separate pairwise analyses carried out using different sets of trials makes it impossible to decide which intervention is best.
展开▼