The subjective interpretation of negative trial results during oral plenary presentations

Abstract

<h3>Objectives:</h3> Oral presentations of phase 3 randomized controlled trials (RCTs) at oncology meetings often do not undergo peer review; this may lead to conclusions that do not reflect the primary results of the study. For example, the presentation may include a positive conclusion despite a negative trial result. Presentations at medical society meetings may have significant impact on the oncology community. The purpose of this study is to quantify and categorize not-negative conclusions made in oral plenary presentations of phase 3 RCTs for gynecologic malignancies. <h3>Methods:</h3> Abstracts related to oral presentations of phase 3 RCTs at the Society of Gynecologic Oncology's Annual Meetings on Women's Cancer between 2005-2020 were reviewed. Studies with a primary endpoint of overall survival (OS) or progression free survival (PFS) and with a formally negative primary endpoint were included. Abstract conclusion sentences were classified as negative or not-negative. Trials with formally negative results were categorized based on the type of not-negative conclusions: 1) positive subgroup emphasis, 2) positive secondary endpoint emphasis, 3) emphasis on better numerical outcome despite nonsignificant p-value, 4) noninferiority interpretation of negative superiority trial. Studies with negative results and not-negative conclusions were compared to respective published manuscripts if available. The results and conclusion from the manuscript were compared to quantify and categorize not-negative conclusions. <h3>Results:</h3> Oral presentations of 61 phase 3 RCTs met inclusion criteria. Of these, 22 had a formally negative primary PFS or OS endpoint, of which 6/22 (27%) presented a not-negative conclusion. There was a higher proportion of not-negative conclusions among negative trials in more recent years, with 50% (5/10) of abstracts from 2015-2020 including not-negative conclusions, vs just 8.3% (1/12) in studies from the preceding decade 2005-2014 (p=0.03; Figure 1). Authors emphasized a positive subgroup in 4/6 studies and a positive secondary endpoint in 1/6 studies. A numerically better outcome in the experimental arm was highlighted in 2/6 studies despite a nonsignificant p-value, and 1/6 studies made a non-inferiority interpretation of a negative superiority trial. Of 21 studies with formally negative results, 56% (5/9) for-profit studies had not-negative conclusions, whereas 8.3% (1/12) non-profit studies had not negative conclusions (p=0.02). Published manuscripts were available for 3/6 not-negative studies, all similarly incorporating not-negative conclusions despite negative results. <h3>Conclusions:</h3> Since 2005, 27% of RCTs presented at SGO made not-negative conclusions despite formally negative results, with a majority emphasizing a positive subgroup and funded by for-profit organizations. These results further emphasize the importance of presenters' accurate portrayal of results and attendees' attention to bias during presentations.