Posting: # 18293
For starters, I may easily have chosen the wrong category, but I think this is a discussion on how to assess and handle study results, so it may fit.
This is from a project some time ago (brought up by a similar but not identical situation in a current study) and I would be interested an other peoples point of view (and maybe the issue is interesting for somebody else, too). As disclaimer, I was not able to find anything close to this issue by means of the search function, but if I should have, please feel free to post the link.
As further disclaimer: the project is buried and dead (in that order), is not connected with my current employer (I am not that prepared to take risks ) and actually was quite successful, but only one part of a larger package.
Lets assume a simple, open parallel group study for standard BE.
For one subject someone managed to apply the wrong treatment (for example by transposed numbers) and this was only observed later. So the question came up how to handle this subject in the evaluation.
Formally, this would have been easy: wrong treatment results in exclusion from the study and the per-protocol (perfect) set. Easy decision, no discussion.
But on the other hand, it was too late, deviation committed, treatment administered, blood taken - would it then be correct (ethically) to exclude perfectly valid data from an analysis?
As an argument: we would not exclude or even be allowed to exclude the data in other clinical trials, where an evaluation in the full analysis set was done, just because the "wrong" treatment was administered.
I am not thinking about checking the data first and decide based on the fact whether inclusion will be favourable or not. I am thinking of a decision prior to bioanalysis or whatever, when the decision is not data-driven.
Back then we decided to make this very clear in the report as a deviation, but ultimately to keep the data in the evaluation (was irrelevant anyway in the end, but that is not an argument).
As an amusing side-note: the ITT argument came up back then and as a consequence I was also asked whether we maybe should keep the subject but have to affiliate it with the intended treatment.
So if such a situation came up, what would be your arguments?
Would be interested.