prana example; endpoints don't mean much?, page-6

thanks Pete

this was interesting:

"Ten patients seem good candidates to be in your study, sign informed consent and after screening (for drug abuse, for non-gulliver problems that might confound the study) are cleared to be in the study. All start taking pill X. The protocol calls for them to return in one week for follow-up.

A week later, only nine patients show up. You call the tenth patient, but her phone has been disconnected. And no one answers when you call her emergency contact. Of those remaining nine patients, seven are doing OK. But one of the nine remaining CG patients has been vomiting like mad, and he thinks drug X is to blame. He didn’t call, however, and is still taking X. Another, you find, has a shockingly low white blood cell count. The vomiting patient doesn’t want to stay in the study. And you do not feel that the leukopenic patient should stay on X. You tell both to stop pill X, tell them that they are out of the trial, but ask them to come back in a week for follow-up to be sure they are OK. You now have, altogether, three patients who are not successful study outcomes.

Seven CG patients keep taking X. But prior to the next study visit, one of them calls. Actually it is his wife that calls. The patient has had a seizure and is in the hospital. The doctors feel he has gotten West Nile virus and seized because of the brain infection. You have to tell the patient that this is a serious adverse event, and that even if the study drug did not evidently cause it, he needs to stop taking X. After all, X is investigational. Not everything is yet known about it.

If it seems I am making this convoluted, I don’t mean to be. This is real-world clinical study drama, folks....".

...

"The FDA requires a different sort of more stringent data breakdown, called intention-to-treat analysis (ITA). ITA requires that you look at ALL patients you screened and cleared to be in the study, and count them in reckoning the study outcome. You intended to treat them, and so whether they completed treatment or not, you have to count them in the study outcome. This is a way of catching not only adverse effects from drugs that are known, but also adverse effects not recognized."