As I understand it and I've checked a couple of times with AI, the sponser of the trial selects the DSMB, often with help from the CRO. They are required to be independent of the sponser and have the required expertise.

That being true, I have no idea how that fiasco (I agree) happened with the 2020 CD12 DSMB... You'd think they would have chosen better, though I know we can be fooled at times re: people's intientions. God knows I've been.