Thread by @MicrobiomDigest

Here is an example of a paper with some very serious problems, where it appears the journal @CellStemCell took a very long time to retract: more than 7 years.
HT: @SmutClyde
sciencedirect.com/science/articl…

Issues with figures from this and three other papers from the same first author have been reported on @PubPeer in November 2015.
pubpeer.com/publications/A…
and e.g.: pubpeer.com/publications/3…

The retraction notice states that 'The Regenerative Research Foundation conducted an investigation following its policies and the NIH Office of Research Integrity guidelines and determined that there had been image manipulation in these figures'.

But there is a 7-year old (March 2016) New York Post article that already mentions that the first author 'was placed on administrative leave last month while a foundation that oversees the Neuro Stem Cell Institute investigates'
nypost.com/2016/03/07/sci…

It is not clear whether the foundation's investigation took 7 years to complete, or if the journal took so long to retract. The retraction notice appears to be deliberately vague about the time line.

The retraction notice also states it is the *authors* requesting retraction. Why didn't the journal make that decision?
Wouldn't that be similar to only giving a fine to people who admitted they were driving under the influence?
If the evidence is there, don't wait for authors.

Here are things that scientific publishers could do better:
1. If evidence is clear, put an immediate Expression of Concern on the paper
2. Retract within a year after first problems are noted
...
#SSP2023

3. More transparency in retraction notices about time interval between first allegations, conclusion of institutional investigation, to retraction.
4. Do not always wait for authors or institutions to agree. If evidence is clear, retract.
#SSP2023

I'd like to point out (again) this Editorial by @hholdenthorp:
Problems with papers can be addressed quickly by e.g. retraction. Institutional investigations can take years. Often, these can be two different parallel processes, not sequential.
#SSP2023
science.org/doi/10.1126/sc…