Preprints and peer-reviewed versions: what changes

ICMJE recommends clearly identifying preprints as work that has not undergone peer review and linking to the subsequent publication. For verifiable reading, record the identifier, date and version as well as the title. A file downloaded months earlier may be outdated.

It also matters whether two documents describe the same sample. Counting them as independent studies would give the same data additional weight. First establish that correspondence, even if the title or author order changed.

A 2020 observational study compared reporting quality in biomedical preprints and peer-reviewed articles, including paired versions of the same work. It found a small average difference favoring the peer-reviewed publications.

The variable examined was the information communicated, such as how easily relevant elements could be located. That helps assess transparency but is not equivalent to repeating experiments or checking all original data. Clearer writing may reveal a limitation that was previously poorly explained.

Another investigation followed epidemiological estimates from COVID-19 preprints to their peer-reviewed publications. Estimates remained closely related between versions, and confidence intervals narrowed on average. The results apply to that sample of publications.

For a specific comparison, the lesson is to separate the estimated value from its uncertainty. A small change in one may accompany an important change in the other. Checking only whether the verbal conclusion remains positive leaves out that information.

When comparing documents, record changes in included participants, exclusions, outcomes, statistical adjustments or supplementary data. This is an editorial comparison criterion: each modification should be connected with the figure or claim being summarized.

If an estimate no longer appears, do not present it as a current result of the final publication. If it remains, check that its definition and units are unchanged. Matching numbers do not guarantee that precisely the same quantity is being measured.

Where a relevant peer-reviewed publication exists, ICMJE recommends citing it rather than the preprint when appropriate. If the analysis concerns a change between versions, identify both and explain the difference.

The cited empirical comparisons do not justify saying all preprints are weak or all peer-reviewed publications correct. They support more precise questions about reporting, estimates and the version supporting each statement.