Indirect Comparison

A method estimating the relative effect of two treatments never studied against each other directly, using their separate effects versus a shared comparator.

Source: Bucher et al. 1997

An indirect comparison estimates the relative effect of two treatments that were never trialled against each other by using their results against a shared comparator. If one treatment beat placebo by a certain margin and another beat the same placebo by a different margin, the gap between those margins gives an estimate of how the two would compare directly. This works only if the trials are similar enough that the shared comparator behaves the same way in each. Bridging through a common reference is its logic. Dias and colleagues (2013) describe this method.

Source: Dias et al. 2013

An indirect comparison works by using the results of trials that share a common comparator: the effect of treatment A versus the comparator and the effect of treatment B versus the comparator are combined, preserving the randomisation within each trial, to estimate the relative effect of A versus B. The Bucher method does this for a single common comparator. The estimate relies on the trials being similar enough that the comparison is valid. So an indirect comparison combines the effects of two treatments against a shared comparator to infer their relative effect, using the common comparator as the link between the otherwise unconnected treatments.

Source: Bucher et al. 1997

An indirect comparison is used when direct head-to-head evidence between two treatments is unavailable but each has been compared with a common comparator, which is common because treatments are often tested against placebo or standard care rather than against each other. It allows their relative effectiveness to be estimated to inform decisions between them. Indirect comparison also underlies network meta-analysis. So an indirect comparison is applied to fill the frequent gap left by the absence of head-to-head trials, providing an estimate of the relative effect of two treatments through their shared comparator when direct evidence is lacking.

Source: Dias et al. 2013

An indirect comparison requires the assumption of similarity, or transitivity: that the trials linked through the common comparator are sufficiently alike in their populations, methods, and the behaviour of the comparator, so that the indirect comparison validly reflects the true relative effect. If the trials differ in factors that modify the treatment effect, the comparison can be biased. So the validity of an indirect comparison depends on the linked trials being comparable, particularly in effect-modifying characteristics, and this assumption must be considered when interpreting the results, since violations of it can distort the estimated relative effect derived through the shared comparator.

Source: Bucher et al. 1997

The limitations of an indirect comparison arise because it depends on the assumption that the linked trials are comparable, and differences between them in effect-modifying factors can bias the estimate; it provides weaker, less certain evidence than a direct head-to-head trial; and errors can arise if the similarity assumption fails. Population-adjustment methods can address some differences but rely on further assumptions. So an indirect comparison is interpreted cautiously, with attention to the comparability of the trials and the greater uncertainty inherent in comparing treatments without direct evidence, and direct evidence is preferred where available.

Source: Bucher et al. 1997