The participant that I know of who withdrew did it after she had finished the trial. That's when she found out they had links to disability insurers. So they already had all her data. That's why they claim it took months to recover from the removal. I fail to understand how, with computers, removing one person's data would take a team on a big trial "months" of work, but that's what the claim is, I guess. I'm not sure who the second person referenced in that paragraph is. The 2011 paper itself only shows one person's data removed after completion.