Fraudstein complains that FEV data was "per protocol" and perhaps based on a restricted and cherry-picked poplulation. But he didn't even notice that the number of patients in the FEV primary endpoint analysis exceeds those in the so-called mITT analyses. In other words, a broader sample was in the primary endpoint
analysis. I am sure the statisticians used every data point that they possibly could. As has been pointed out, no human trials are perfect and you're going to lose some data, whether due to hangnail or heart attack. Fraudstein, though, immediately points to company fraud, naturally.
You make an interesting point, but looking at Fig 4 it actually seems like something was at best mislabeled. If you look week to week there are earlier weeks with fewer patients than later weeks, suggesting they did not impute missing values and/or allowed subjects who missed a rating to continue on. That would suggest possible inclusion of possible protocol villagers, precisely why a PP analysis is rarely used. In any event, per other posts, what AF missed, including the above, is that if a PP analysis was specified there is no problem. If it was not, there is no dispute from the poster that different populations were used for the primary versus other endpoints, and that could be an issue. Finally, you are mistaken in assuming that the 128 patients in a PP population would have to overlap with the 128 patients in the mITT populations.
Did you see Figure 2? PRIMARY ENDPOINT: Relative Change in FEV1 at Week 24 using the Per Protocol Population with a Noninferiority Margin of (minus) 5%. The measured difference was minus 1.31% with a lower 95% confidence interval that did not encompass the minus 5%, hence the statistical finding of noninferiority. I would imagine the slight dip in numbers during the middle part of the trial could be explained by the fact that patients were not available for those particular readings, but having continued with the protocol treatment, were included in the final analysis. That is just my best reasonable assumption.