Thread by @Ayjchan

I've been sitting on a major topic that I think the non-scientist public needs a primer on, with particular significance to COVID-19 research.

That topic: Research Misconduct
ori.hhs.gov/definition-mis…

And what to do about papers that are found to have engaged in misconduct.

One of the most notable instances of misconduct was the Surgisphere HCQ papers. @TheLancet eventually decided to retract the paper & commentary because they would be too misleading in their original form. They adopted a "retract and replace" approach... retractionwatch.com/2020/07/10/a-m…

... because the editorial had been written by innocent parties who were not aware of the data issues, @TheLancet published a new editorial to explain what had transpired - in order to rightfully preserve the reputations of scientists who had been misled. retractionwatch.com/2018/03/29/a-n…

This is a good read. Particularly, D. Goldstein's quote: “Given the amount of data that was in the [Surgisphere] database, it’s just hard to believe someone would [fabricate] something like this.” - Many of us underestimate the scope of misconduct. sciencemag.org/news/2020/06/w…

According to the @HHS_ORI, research misconduct comes in 3 forms (1) fabrication=made-up s**t, (2) falsification=manipulating, changing, omitting data or results > inaccurate representation in the research record, (3) plagiarism=taking someone's ideas/results without attribution.

The first two are especially pernicious because they can mislead the entire scientific community, wasting millions of taxpayer $$$ and (prime) years of research by brilliant and/or hardworking scientists. I didn't get an education in this until I moved to Boston.

But the last one can also be dangerous, especially in the form of deliberate self-plagiarism. For example, publishing the same data again and again as new and separate data. This misleads the entire field into thinking that there are many many instances of the same phenomenon.

Example: what if there is one phase 3 vaccine subject who suffered adverse effects, but leaders of the vaccine project kept on publishing details on that patient's adverse events across different publications, different journals, without informing the reader that this was 1 case.

Due to data falsification+self-plagiarism, readers of these papers would have the impression that several phase 3 subjects are suffering adverse effects from the vaccine.

In such a scenario, I don't see how these papers could be corrected instead of straight-up retracted.

What do journals do when they obtain evidence of misconduct?

Each journal has its own protocol. However, some journals can take years to investigate. A typical research misconduct inquiry+investigation can take 2 years. By that time, dozens of scientists would've been misled.

What is the incentive for a journal or university/institute to resolve allegations of research misconduct in a timely manner?

Maybe only reputational damage. And that will not even stop the hoards of people trying to get into top journals and schools.

This year, I've been shocked, repeatedly, that established scientists would try to salvage data from publications that have signs of misconduct, mainly falsification. There's a desperation to believe in data+conclusions of some papers, even after misconduct has been revealed.

It's like someone who realizes that their food is horribly rotten but tries to pick out the parts that are at least visibly not moldy or covered in maggots.

This is not the behavior of top journals, top institutes, top scientists.

This may be cheeky, but @NIH it's not the trainees who should be compelled to attend Responsible Conduct of Research meetings. It's the PIs. It's the journal editors. If they don't take the lead on research integrity, you can't expect trainees to!

In one of the RCR courses I took, the scenario: what should you do if you're a grad student who noticed that a postdoc falsified data for your PI who happily presented it at a conference.

What the hell is a grad student or even another postdoc supposed to do in this scenario?

I suspect the course would be 10x more useful if you taught trainees how to document research misconduct via emails, data, lab meeting slides. And also walked them through the >2-years of horror of trying to report misconduct, where they become known primarily as a whistleblower.

Whether you’re a journal, an institute, or a national program, if you demonstrate an inability/reluctance to identify research misconduct, and a propensity for quietly covering it up, you’re signalling to bad actors that they can continue to count on this protection in future.