Trustworthiness of medical publications can depend on either good faith or verifiable data. In recent years, it has increasingly become apparent that good faith alone is not enough, with estimates of fabricated research being 30% of published randomised clinical trials. This seminar discusses the role of authors and their institutes, peer review pre-and post publication and the role of publishers in this problem.

Comments from chat:
This is an excellent talk on making people aware of the problem that many of papers published in medical journals are flawed by fraud. Methodological errors is also a common problem. But a main obstacle to improving the scientific rigor in medical journal papers is the unwillingness of most medical journal editors to publish replication papers, unlike economic journals. Economic journals also require that authors make their de-identified data and codes available to independent authors for replication purposes. Several medical journal editors told me at the 2022 Congress of Peer Review that they do not publish replication papers because it believe it will worsen their impact factor. One editor told me they don’t publish replication papers because their audience aren’t interested in replication papers. Question to audience: What can we do to change medical journals’ culture from one of “deny and defend” to transparency?
But why would most people work to identify fraud if their replication paper that identified fraud are not published?

If people want to have a look at my problem paper and refusal of authors and the editor to even bother to answer   / 1648587873359200261  

I'm not an expert in the field, but I recently had the chance to read a review on possible applications of synthetic data in médicine, i.e. data produced by an AI by "amplifying" a smaller dataset, with the primary aim of AI training, but with several possible different implications. Do you think that misuse of synthetic data might worsen the issue of fraud? Or could the transformations brought about by AIs be a fruitful chance to take into account the problem of fraud as well, and bring greater awareness on the issue of data reliability?

Replying to "If people want to ..."

In terms of transparency, this is now a journalist who's asking the questions given the EiC isn't acting

WRT to frauds being more sophistocated, I think the hope is that criminals get bolder and more reckless with time (think Malcolm Pearce & BJOG papers, or even our medical murderer Harold Shipman). When finally caught because becoming lax, the lookback finds they've been miscreants for years.

I have to go now but thank you Ben, this was an extremely illuminating talk.

Parker, Lisa and Boughton, Stephanie and Lawrence, Rosa and Bero, Lisa, 'Somebody Might Actually Make Up Data': An Empirical Qualitative Study on How to Screen for Fake Research (2022). Available at SSRN: https://ssrn.com/abstract=4071671 or http://dx.doi.org/10.2139/ssrn.4071671

It beats me why more people don't say straight out that AI, or at least LLMs, are a spectacular failure

I observe that most of the fraud studies were clinical trials and not other types of studies. Is that true ? and why would that be the case , any thoughts?