This problem has several constraints which limit the possible ways in which the problem could have occurred.
1) How could the original Kaplan-Meier survival curve be so good if there was a mix up between the control arm and the 1 mg/kg arm?
2) The fact that they are making a new control arm combining the placebo arm and 1 mg/kg arm
tells me that there are too many patients in both arms affected to just throw them out. However, how can this be reconciled with the first statement above?
If the mix up had a big effect then the curves for the placebo and 1 mg/kg arms should be closer together.
3) That leaves the possibility that the mix up only caused a small effect, but it occurred with many patients.
My theory here is that many, maybe even all, of the patients in the placebo and 1 mg/kg arms received one incorrect dose. So all the placebo patients got 5 doses of placebo and one dose of
bavi at 1 mg/kg. I think this would cause hardly any change. Likewise, all patients in the 1 mg/kg arm got 5 doses of bavi and one dose of placebo. Again, this may have caused hardly any change. This might be especially true if the mix up occurred at the beginning or at the end of the 6 treatments. So the result is all the patients in these two arms were affected, but the mixed up doses didn't make any noticeable difference in the survival curves. You can see that correcting the data is not possible since there would be no one left. So the only thing to do is to combine them and make a new "control" arm. Alternatively, you can just recognize that the effect was small and not do anything. Luckily, the 3 mg/kg arm was not involved and so the question is what do you compare it with. I maintain that the comparison with the original placebo arm is still good enough if this happened the way I believe it did.
That is my theory. When Peregrine meets the FDA they will have all the data and know exactly what happened and when. If they can show the FDA the numbers I think it would be hard to not accept the fact that bavituximab works very well. There does not have to be statistical significance to be given the go ahead for a phase 3 trial, the FDA just has to believe bavi is safe and has a good chance of being effective for second-line NSCLC.
Hopefully, in the near future Peregrine will release a much more detailed explanation of the nature of the mix up.
