GRADE POINTS FOR CLINICAL GUIDELINE PANELISTS - PUBLICATION BIAS

All studies should be reported. This is the premise behind the All Trials campaign launched in 2013 which uses the slogan "All trials registered - All results reported". Also in 2013, Doug Altman and David Moher published a declaration of transparency where they suggest that "authors should sign a publication transparency declaration as part of every journal submission for each research article". In that editorial they also stated, "Failure to publish the findings of all studies, especially randomised trials, seriously distorts the evidence base for clinical decision making." 

Without all the data, we cannot accurately assess the evidence for many clinical questions.

GRADE accounts for this in the rating of evidence from RCTs through the fifth and final domain for downgrading evidence, which they name publication bias.

Publication bias occurs when the decision whether or not to publish a study is influenced by the results of the study.

GRADE summarizes some of the reasons for differential publication leading to what we know of as publication bias, especially that negative studies are more likely to go unpublished than are studies with positive results. The reasons are many, including authors failing to invest time to write up or submit study results that are negative and editors or peer reviewers finding negative studies less interesting. 

GRADE recommends downgrading for publication bias when the studies are small and especially if most are industry sponsored. If there are sufficient studies, use of funnel plots to look for asymmetric distribution is offered, but GRADE warns that "visual assessment of funnel plots is distressingly prone to error."

GRADE acknowledges the challenges in determining whether or not there are missing studies. As a result, GRADE suggests different terminology and method for rating down evidence for this domain. When suspicion is low, they suggest using the label "undetected". GRADE suggests rating down a maximum of one level for publication bias and using the label "strongly suspected".

GRADE cautions that since early negative trials may face delayed publication, this makes early, positive trials, especially small ones, suspect.

As always, for the definitive resource on all things GRADE, consult the GRADE HANDBOOK

TheEvidencedoc November 27, 2017