Trials

, 14:O123 | Cite as

Design, analysis and reporting of active-control randomised trials: a systematic review

  • Lang'o Odondi
  • Chris Metcalfe
  • Jonathan Sterne
Open Access
Oral presentation
  • 567 Downloads

Keywords

Secondary Analysis Primary Outcome Measure Primary Analysis Average Treatment Widespread Application 

Background

Active-control randomised trials are typically conducted to evaluate the benefit of a novel intervention relative to an established treatment. The appropriate design, analysis and conclusions depend upon whether the novel intervention is anticipated to be superior or non-inferior on the primary outcome measure. But despite their widespread application, active-control studies are often poorly understood, improperly applied, and incorrectly interpreted.

Methods

We systematically review recently published active-control trials in a high-impact journal, to appraise their design, analysis and the conclusions drawn. All trials published in 2010 were identified and data extracted in duplicate using a standard proforma. Meta-analyses examined differences in the average treatment effect between superiority and non-inferiority studies.

Results

Thirty-seven studies met our inclusion criteria; 12 designed as non-inferiority studies. Non-inferiority studies did not have larger sample sizes (median=702) compared to superiority studies (median=725). Margins of non-inferiority were explicitly defined for the non-inferiority studies. All studies employed an intention-to-treat as primary analysis, with one non-inferiority study including per-protocol secondary analysis. On average, superiority studies favoured the novel treatment (OR=0.75), which was not the case for non-inferiority studies (OR=1.31). Few studies swapped between the superiority and non-inferiority approaches as the study progressed, with no evidence that those swaps were in response to the results obtained.

Conclusion

These active-control studies were found to have appropriate sample size targets, with no evidence of non-inferiority studies being larger. Only one non-inferiority study used per-protocol analysis as suggested in the CONSORT guidelines. There was no evidence of reporting bias due to switches between superiority and non-inferiority approaches.

Copyright information

© Odondi et al; licensee BioMed Central Ltd. 2013

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Authors and Affiliations

  • Lang'o Odondi
    • 1
  • Chris Metcalfe
    • 1
  • Jonathan Sterne
    • 1
  1. 1.University of BristolBristolUK

Personalised recommendations