Analysis

Harms of outcome switching in reports of randomised trials: CONSORT perspective

BMJ 2017; 356 doi: https://doi.org/10.1136/bmj.j396 (Published 14 February 2017) Cite this as: BMJ 2017;356:j396
  1. Douglas G Altman, professor1,
  2. David Moher, senior scientist2,
  3. Kenneth F Schulz, distinguished scientist3 4
  1. 2Ottawa Hospital Research Institute and Canadian EQUATOR Centre, University of Ottawa, Ottawa, Canada
  2. 3FHI 360, Durham, NC, USA
  3. 4University of North Carolina School of Medicine, Chapel Hill, NC, USA
  1. Accepted 13 January 2017
  1. 1Centre for Statistics in Medicine and UK EQUATOR Centre, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford, UK
  2. Correspondence to: D G Altmandoug.altman{at}csm.ox.ac.uk

The outcomes reported for trials often differ from those specified in the protocol. Douglas Altman, David Moher, and Kenneth Schulz call for journals to do more to ensure that authors follow guidelines

Rigorous and transparent conduct of clinical trials is essential for confidence in their results. Randomised trials must have a protocol that gives details of the planned study design and outlines the intended methods of analysis, reflecting substantial discussions among the investigators. Key elements of the study design should also be included in an entry on a clinical trials registry. The core principles, as set out in the SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) statement,1 are that the protocol should give full details about objectives and planned conduct; the protocol should be adhered to; and any changes from what was initially planned should be reported and, ideally, explained. The SPIRIT guidance for protocols intentionally overlaps considerably with CONSORT recommendations for what information should be included in the report of a randomised trial.2

A particular concern is that researchers should specify the primary and secondary outcomes that will be assessed, including the method and timing of measurement. When the trial’s findings are published, the main reported results should be for those predeclared outcomes and any differences should be highlighted and explained. Adherence to a prespecified plan avoids bias, gives credibility to the results, and avoids confusion for the reader. Conversely, when authors substitute other outcomes after the trial has started there must be concern that such changes were done with knowledge of the data. That casts doubt on the reliability and integrity of the results.

Switching outcomes

Despite this guidance, Chan and colleagues identified frequent discrepancies between the outcomes specified in trial protocols and those reported in subsequent journal articles.3 In particular, in over half …

View Full Text

Sign in

Log in through your institution

Subscribe