According to a recent study published in Psychological Medicine, reporting and citation biases are eliminating negative study results from the scientific literature. The authors of the research assembled information on 105 trials of antidepressants that had been registered with the FDA and identified the cumulative effect of four specific publishing biases:
- Study publication bias: The results of half of the trials examined were considered positive by the FDA and 98% of these were published, while the other half were considered negative or questionable and only 48% of these were published.
- Outcome reporting bias: The authors considered that 10 of the published negative trials were reported as ‘positive’ within the publication, by switching the status of the primary and secondary outcomes or failing to include less favourable data.
- Spin: Eleven of the remaining published negative trials were written using language the authors felt made the negative results appear positive, such as ‘a trend for efficacy’.
- Citation bias: Positive trials and those including spin in the abstract were cited more frequently than negative trials.
Interestingly, the authors noted that all of the negative trials that remained unpublished were completed before 2004. Therefore, the more recent requirement for trials to be prospectively registered and to report results following completion may be helping to prevent study publication bias. However, a review conducted by Aaron Carroll in The New York Times delved further into the issue, highlighting the serious effects of various cumulative biases, including how they can skew the results of meta-analyses, a tool critical for evidence-based decision-making. Carrol calls for the scientific community to encourage journals to publish negative results and asks us to ”celebrate and elevate negative results, in both our arguments and reporting, as we do positive ones”.