Guidance from the EQUATOR network and the ICMJE stipulates that clinical trials must be publicly registered at inception in order for results to be considered for publication. This practice was intended to reduce selective reporting of trial outcomes; however, discrepancies between pre-specified and published outcomes of clinical trials persist. A recent article published in Trials by Dr Ben Goldacre (AllTrials-founder) and colleagues reported high levels of outcome misreporting in journals listed as endorsing the CONSORT statement on the correct reporting of clinical trials. What’s more, the group reported that most correction letters they sent to address such misreporting were rejected.
This prospective study (COMPare) tracked outcome switching in 67 randomised controlled trials (RCTs) published over a six-week period (19 Oct–30 Nov 2015) in the five top CONSORT-endorsing medical journals (The New Engl J Med, The Lancet, JAMA, The BMJ and Annals of Internal Medicine). Overall, outcome reporting was deemed to be poor; 58/67 (87%) manuscripts analysed contained CONSORT-breaching discrepancies judged to require a correction letter. Of these letters, only 40% were published, with a median delay of 99 days. Qualitative analysis was thought to suggest misunderstanding among journal editors surrounding correct outcome reporting and CONSORT, a finding that may explain why some journals did not cooperate when presented with evidence of misreporting.
The authors warn that readers of a journal with a CONSORT-supporting stance may assume that trial results are reported in line with pre-specified outcomes. To ensure this is the case, Goldacre and his colleagues suggest a number of strategies: 1) Changes to journals’ correspondence processes; 2) Indexed post-publication peer review; 3) Changes to CONSORT’s enforcement mechanisms; 4) Changes to practices in methodology research to increase sharing of misreporting with the broader academic community.