Date created: | Last Updated:

Category: Project

Description: Objectives: To describe and summarise the reporting of "non-significant" results in clinical trials, and to estimate how commonly clinical trial reports make an erroneous claim of no treatment difference based on a non-statistically significant result. Design: Retrospective survey. Setting: Four high impact factor general medical journals, published between June 2016 and June 2017. Participants: Reports of randomised controlled trials that did not find a difference between the interventions they compared. Interventions: No intervention. Primary and secondary outcome measures: We used a 10-category classification for the text describing results for the primary outcome or outcomes, in the Results and Conclusions sections of the Abstract of each paper. Proportion of papers making claims that were not justified by the results. Results: Eighty-five trial reports were included, reporting 111 treatment comparisons. The majority of papers (55%) concluded that there was no treatment benefit. The other common approaches were to state that there was no significant benefit (12.6%) or no significant difference (11.7%). Conclusions: Despite decades of warnings, the error of concluding a lack of treatment benefit from a non-statistically significant result remains common.