Anyone ?

Hi,

The most appropriate method to have a head-to-head comparison of the three drugs is to conduct a network meta-analysis (NMA).

NMA uses evidence from direct head-to-head trials and also information from indirect comparisons (A vs. placebo and B vs. placebo to get an idea about A vs. B) to identify the best of the various drugs used for a given condition.

The Cochrane library has an excellent webinar on NMA:

http://training.cochrane.org/resource/introduction-network-meta-analysis-nma

Short of that, you can only speculate the 'best' based on the point estimates of the relative risks of each of the comparisons of A vs. placebo, B vs. placebo, and C vs. placebo.

Hope that helps.

Jeeva

The simplest example of a multiple-treatments meta-analysis is the indirect comparison.

Indirect comparisons are made between interventions in the absence of head-to-head randomized studies. For example, suppose that some trials have compared the effectiveness of ‘dietician versus doctor’ in providing dietary advice, and others have compared the effectiveness of ‘dietician versus nurse’, but no trials have compared the effectiveness of ‘doctor versus nurse’. We might then wish to learn about the relative effectiveness of ‘doctor versus nurse’ by making indirect comparisons. In fact, doctors and nurses can be compared indirectly by contrasting trials of ‘dietician versus doctor’ with trials of ‘dietician versus nurse’.

One approach that should never be used is the direct comparison of the relevant single arms of the trials. For example, patients receiving advice from a nurse (in the ‘dietician versus nurse’ trials) should not be compared directly with patients receiving advice from a doctor (in the ‘dietician versus doctor’ trials). This comparison ignores the potential benefits of randomization and suffers from the same (usually extreme) biases as a comparison of independent cohort studies.

More appropriate methods for indirect comparisons are available, but the assumptions underlying the methods need to be considered carefully. A relatively simple method is to perform subgroup analyses, the different subgroups being defined by the different comparisons being made. For the particular case of two subgroups (two comparisons; three interventions) the difference between the subgroups can be estimated, and the statistical significance determined, using a simple procedure described by Bucher (Bucher 1997). In the previous example, one subgroup would be the ‘dietician versus doctor’ trials, and the other subgroup the ‘dietician versus nurse’ trials. The difference between the summary effects in the two subgroups will provide an estimate of the desired comparison, ‘doctor versus nurse’. The test can be performed using the test for differences between subgroups, as implemented in RevMan (see Chapter 9, Section 9.6.3.1). The validity of an indirect comparison relies on the different subgroups of trials being similar, on average, in all other factors that may affect outcome. More extensive discussions of indirect comparisons are available (Song 2003, Glenny 2005)