Avid Bioservices Inc (CDMO)

[freethemice]

I want to look at Mojojojo's result for pancreatic cancer in a different way. I am using pancreatic cancer because
the MOS values from the control arms of the phase 3 trials in my tables are fairly uniform. That is, there is not
too much variation in the results. This is probably because it is such a deadly disease that kills patients quickly
and the standard chemotherapy, gemcitabine, isn't very effective.

For each of the 13 clinical trails there is a group of patients which has certain characteristics. These include
the two most important, PS score, and disease extent, as well as others, such as CA 19-9 baseline values,
and other things such as age, gender, etc. There are also the unknowns, such as the individual's response to
gemcitabine, immune system condition, etc., things we think might influence the patient outcomes, but for which
we have no data. There are also the unknown unknowns. However, there are also constraints placed on what
patients are eligible, which are encoded in the inclusion, and exclusion, criteria for the trial. These criteria
are somewhat similar between the trials, so the patient groups have considerable overlap in
their characteristics, but also have their differences.

If we could take just one group of patients and run the same trial on them over, and over, we would get a
distribution of control arm MOS values about some mean. Obviously, that is not possible. But we do
have these 13 control arms from phase 3 trials. Each of them has a different input, that is group of patients,
which are somewhat different. Then they go through their trials and eventually the MOS values for the control
arms are computed. These values should also be distributed about some mean. We can think of the differences
in these control arm MOS values as due to noise which arises from the different inputs (patient groups).
See the figure below.

Let us now assume that the MOS for the treatment arm of the Peregrine pancreatic trial does come
out at 10.1 months, as Mojojojo has estimated. Let us make the hypothesis that Bavi is nothing but
a placebo, so the treatment arm can then be considered as a control arm receiving gemcitabine plus
placebo. If that is true, then what is the probability of a control arm producing a MOS value of 10.1 months?
I have tried to calculate that using the data I have and excel functions. I am no statistician, but I think I worked it out.

The value of 10.1 months is about 5 standard deviation units from the mean of the 13 control arms, as you
can see in the table below. That is an incredible difference. I am assuming that the control arm of the
Peregrine trial will turn out to be fairly similar to the mean of the 13 control arms I am using here. Since the
standard deviation of the those 13 MOS values is fairly small, it is probably a good assumption.
Of course, the only thing that is important to the FDA is the difference between the treatment arm and the
control arm of the same trial. I am just trying to get a measure of how different the bavi treatment arm might
be from the historic control arms. As an aside, the threshold of detection used by the physicists at CERN to
detect the Higgs boson was a 5-sigma event.


Thanks for reading,
FTM