Methodological study
A review of the methodology and reporting of multi-arm, parallel group, randomised clinical trials
Ed Juszczak, Doug Altman, An-Wen Chan
Publications on randomised clinical trials (RCTs) largely focus on two-arm parallel group trials. However, trials with more than two arms raise complex issues in design, conduct, analysis and interpretation, mainly arising from the multiplicity of comparisons.
These issues include (i) justifying the multi-arm design and explicitly defining the objectives; (ii) pre-defining a clear analysis strategy, consistent across outcomes; (iii) pre-specifying all planned comparisons of intervention groups, and (iv) addressing the issue of adjustment for multiple testing.
The multiplicity of possible comparisons also leads to complexities in the sample size calculation per se, trial monitoring/interim analysis and in interpretation.
In order to clarify these issues we are performing a review of current recommendations and practice, which will offer guidance to researchers, and inform the planned extension of the CONSORT statement to such trials.
Within a cohort of 519 Medline-indexed RCTs published in December 2000. 97 (25%) of the 388 parallel group RCTs had more than two arms. Primary reports of a random sample of 60 trials are being reviewed, which will allow a good description of current practice. Each article is being assessed in duplicate using a purpose-made pre-tested form addressing issues relating to the design, conduct, analysis and interpretation specific to multi-arm trials.
We will also examine textbooks on RCTs and carry out a Medline search to seek recommendations specific to the design and analysis of multi-arm parallel group trials, over and above generic information that applies to all RCTs.
We will present a summary of findings from the review of published trials and textbooks, and make recommendations for the reporting of such trials.
