A new review shows a majority of COVID-19 randomized, controlled trial preprints have been published with changes to the components of the original findings.
The time between online preprint posting and journal publication for randomized clinical trials on COVID-19 subjects can range from 2 weeks to more than a year, according to new research.
Data from a team of Canadian investigators showed preprint research articles on COVID-19 take a median time of more than 6 months to actually publish in a journal. The significant lag in publication—as well as varied outcomes of publication for preprint articles based on their trial sample sizes and risk of biases—speak to the lack of uniform strategy nor prioritization of professional review of the latest research into the pandemic virus.
Investigators led by Anthony D. Bai, MD, of the Division of Infectious Diseases in the Department of Medicine at Queen’s University, sought to interpret time to journal publication of COVID-19 randomized clinical trial (RCT) preprints, as well as differences in research published both as a preprint and eventually as a journal article.
Bai and colleagues stressed the “impetus” to provide RCT data surrounding COVID-19 as soon as possible in order to aid clinical care strategies during the pandemic. They cited preprint successes, such as the RECOVERY trial’s outcomes surrounding dexamethasone as a beneficial, standard-of-care treatment for COVID-19.
“However, preprints may not meet the required standards for scientific publication without peer review, they may be subject to change because the manuscript is not yet finalized, and they may contain errors or false information,” they wrote. “As a result, preprints can mislead the public.”
Few previous trials have observed differences in COVID-19 RCT preprint and eventual journal-published article pairs; the investigators’ systematic review and meta-epidemiologic analysis sought to answer how exactly journal articles differ in time to publication, outcomes, analyses, results and conclusions based on preprint on journal-published status.
“We hypothesized that there would be a substantial time lag from preprint posting to journal publication and that there would be notable differences between preprints and corresponding published journal articles,” they wrote.
The team conducted a literature search of the World Health Organization (WHO) COVID-19 database and Embase to identify preprint RCT articles published between January 1 and December 31, 2021. Eligible trials included human participants and research questions pertaining to the treatment or prevention of COVID-19.
For observed preprints, investigators then conducted a literature search to locate its corresponding journal article. Outcomes including risk of bias and time to publication were defined by independent reviewers and the Cox proportional hazards regression model, respectively. Bai and colleagues independently described RCT trial pair differences by outcomes, analyses, results and conclusions.
The final assessment included 152 preprints, of which, 119 (78.3%) had been published in journals by October 1, 2022. Of the published RCTs, 65 (54.6%) were published with changes in outcomes, analyses, results or conclusions. Median time to publication was 186 days (range, 17 – 407). Just 2 (1.7%) studies were published with conclusions that contradicted the conclusion included in the preprint version.
The team identified an association between larger RCT sample sizes and lower risks for bias with journal publications. RCT patient or sample sizes of 201 – 1000 were 23% more likely to be published versus a reference of RCTs with ≤200 samples (hazard ratio [HR], 1.23; 95% CI, 0.80 – 1.91). Among RCTs with >1000 samples, likelihood of publication was more than two-fold (HR, 2.19; 95% CI, 1.36 – 3.53).
Using an RCT with a high risk of bias as reference, investigators observed that medium-risk RCTs with some concerns for bias were 77% more likely to be published (HR, 1.77; 95% CI, 1.02 – 3.09); low-risk RCTs were three-fold more likely (HR, 3.01; 95% CI, 1.71 – 5.30).
Bai and colleagues concluded the time lag from preprint to publication in COVID-19 RCTs was “substantial,” as were the differences in outcomes, analyses, sample size and results from preprint to publication in a majority of the RCTs. Among the most implicative findings they noted were the need for more context relevant to preprint data including in the journal-published RCTs.
“However, if investigators include preprints, they should use caution in synthesizing and interpreting knowledge and should acknowledge that unpublished preprints are likely to have a higher risk of bias,” they wrote. “In this study, it was reassuring that the primary outcome estimate and conclusion typically remained consistent in preprints and corresponding journal articles, so these updates likely would not change the overall conclusion.”
The study, “Comparison of Preprint Postings of Randomized Clinical Trials on COVID-19 and Corresponding Published Journal Articles,” was published online in JAMA Network Open.