Full Publication of Preprint Articles in Prevention Research: An Analysis of Publication Proportions and Results Consistency

Introduction There is concern that preprint articles will lead to an increase in the amount of scientifically invalid work available. The objectives of this study were 1) to determine the proportion of prevention preprints published within 12 months, 2) to assess the consistency of the effect estimates and conclusions between preprint and published articles, and 3) to explore the reasons for the nonpublication of preprints. Methods We developed a web crawler to search the preprint server medRxiv for prevention studies posted from January 1 to September 30, 2020. An update search was conducted 12 months later. We dually screened the results for prevention articles and developed a scheme to classify changes in effect sizes and conclusions. We modeled the effect of a set of predictors on the proportion of preprints published in peer-reviewed journals. We also developed a questionnaire for authors of unpublished preprints. Results Of the 329 prevention preprints that met our eligibility criteria, almost half (48.9%) were published in a peer-reviewed journal within 12 months of being posted, with the median time being 5.3 months (range 0.1–11.3 months). While 27 out of 161 (16.8%) published preprints showed some change in the magnitude of the primary outcome effect estimate, 4.4% were classified as having a major change. The style or wording of the conclusion changed in 42.2%, while the content of the conclusion changed in 3.1%. Preprints on chemoprevention, with a cross-sectional design, and with public and noncommercial funding had the highest probabilities of publication. The main reasons for the nonpublication of preprints were journal rejection or lack of time. Conclusion The reliability of preprint articles for evidence-based decision-making is questionable. Less than half of the preprint articles on prevention research are published in a peer-reviewed journal within 12 months, and significant changes in effect sizes and/or conclusions are still possible during the peer-review process.