Avid Bioservices Inc (CDMO)

[tradero]

Thank you Golfho for your detailed explanations.

I find interesting what you said

That implies that there is, in my view a 28% probability that the trial could be stopped at the first look-in.

Now let me take look at this from another perspective. We are getting closer to the end line and thus we can more and more base our projections on facts.

Number of enrolled patients in Sunrise:
- Beginning of Dec., SK said we had achieved more than 90% of enrollment goal. 582*0.9 > 524 (fact)
- We know that Sunrise Centers started the transition to the Durva Combo trial in January… So this means that PPHM thought that the Sunrise trial was already good enough. No more time and energy unnecessarily wasted and let’s start the new trial a.s.a.p. So let’s round up that 524 number with a few more enrolments in December/January… up to 552 (final number). It may not be far from reality.

Also, we know that 1st & 2nd look-in happens at 33 & 50 % of ??… NOT necessarily ?? means the total enrolled patients. Because besides making the look-ins unreasonably long and not fair for the trial (bear in mind “droppedouts”)… Projections would send first look-in into the summer, while SK expected them to be 2016Q1… Mind you, though projections may not be very accurate… but they can give you an order of magnitude…. And the more they are based on actual data the closer they get to reality.

So, for me it makes sense that look-ins be a percentage of still “active” patients in the trial (alive or evented). This is: total enrolled patients minus censored patients. So let me speculate with this assumption.

From the Opdivo trial (see chart below , thanks Mojojojo) we know:
- there were around a 17% of censored patients
- Censoring really started after 15 months of enrollment. Before that 15th month, “dropped outs” were insignificant (no need to take this into account in our calculations)
- There were more than double censored patients in the Opdivo Arm than the Placebo Arm

Now, Sunrise is in its mid-life, with about only 50 patients who have been enrolled for 15 months or longer (see table below), so the number of censored patients needs to be an estimated number. We have the 17% figure from the Opdivo trial but I believe that this number will be bigger in Sunrise. Why? As a continuation (transition) of Sunrise we have the Durvalumab Combo trial at the same sites and it seems to me that many Sunrise patients will be enrolled in that new Durva Combo trial as well… and thus Sunrise IMHO will be heavily censored. Doctors’ number one priority should be their patients’ health. And Durva + Bavi may be another step in their treatment.

So, I wouldn’t be surprised if they estimate that the number of censored patients is 20% or more. Therefore “active” patients might be 552 * 0.8 => for convenience in calculations let’s round it up to 444 patients
1st look-in: 444 *0.66 = 148 events ; 2nd look-in: 444 * 0.5 = 222

Now let’s take a look at March 2016 as a pivotal month in the trial.
The enrolment numbers are based on Golfho’s first model but updated to the present circumstances. And since the model is based on the factual dates when Centers signed in for Sunrise… data might be “fairly” accurate



Let’s take a look at the situation for the Placebo Arm in February and March 2016. Let’s estimate how many events there might have been just for the Placebo Arm.

These calculations will be based on:
- The enrolment model Table (see above)
- Assuming that the Placebo Arm will behave like the Placebo Arm at Herbst’s trial, assumption that can be reasonably accurate since patient population in both trials are of a similar type.

The results are:
- On March 1st, the Placebo Arm may have AROUND 119 events
- On April 1st, the Placebo Arm may have AROUND 130 events (see below table on how these numbers came about)

If these assumptions are right, these estimations and the fact that 1st look-in has not happened yet tells me that the number of events from the Bavi Arm (the unknown in the equation) is rather low (considerably lower than the Placebo Arm), which is good news.

Now, let´s remember that 148 was the estimated number of events needed for the 1st Look-in. Now, 139 is another magic number for the Placebo Arm (and the Bavi Arm for that matter). It is more than half of the Patients Arm group. If when 1st look-in is triggered off there were 139 (or more) patients from the Placebo Arm, that would mean that the Median of the Placebo Arm would have already been established (in our sample we estimated it a 9.9 MOS). And that would have given hopes for a Trial Halt and not just a “futility” check. But I don’t have much hopes it will.

However for 2nd look-in… Let’s take a look at the month of June:
On July 1st there would have been AROUND 171 events for the Placebo Arm (171 > 139). So the Median for that Arm would have been established and there would be around 222 – 171 = 51 events for the Bavi Arm (51 as a maximum, less if 2nd look-in comes later).
The disproportion would be such that … it is easy to guess that the Bavi MOS would beat the Placebo Arm by far… So according to these assumptions I would expect the trial to be stopped by 2nd look-in.

The table below shows the method used to estimate the number of events for the Placebo Arm in March 2016.



Censoring Chart - Opdivo