Publication Bias in Abstracts Presented at the American Diabetes Association Scientific Sessions: A Retrospective Cohort Study
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Objective
To determine the publication rate of abstracts presented at the American Diabetes Association Scientific Sessions and to evaluate the association between statistical significance of study results and subsequent publication.
Research Design and Methods
We conducted a retrospective cohort study of abstracts presented at the 2018 American Diabetes Association Scientific Sessions. The primary exposure was study result category (statistically significant vs. non-statistically significant findings), and the primary outcome was publication in an indexed journal within 5 years after conference presentation. Publication status was determined through PubMed/MEDLINE and Scopus searches. Adjusted relative risks (RRs) and 95% CIs were estimated using generalized linear models with Poisson distribution and robust variance.
Results
Among 541 included abstracts, 321 (59.3%) were subsequently published in indexed journals. Abstracts reporting statistically significant findings had a higher publication rate than those reporting non-statistically significant findings (61.9% vs. 42.3%; p=0.002). In the adjusted analysis, abstracts with non-statistically significant findings had a lower likelihood of publication compared with those reporting statistically significant findings (adjusted RR 0.71 [95% CI 0.54–0.93]; p=0.012).
Conclusions
Approximately four in ten abstracts presented at the ADA Scientific Sessions were not published within 5 years. Abstracts reporting non-statistically significant findings had a lower likelihood of subsequent publication, suggesting persistent publication bias in diabetology research. Future initiatives promoting the interpretation of effect estimates, confidence intervals and clinical relevance, rather than statistical significance alone, may help reduce selective dissemination of evidence
