The reason I believe success for DCVAX Phase-3 is all but guaranteed.
There have been a lot of posts lately stating that 1000's of events would be required to have statistically compelling evidence that DCVAX is effective so I went back and decided to take at very simple look at the trial. We know 2/3rd of that patients received DCVAX from the getgo and 90% of the 1/3rd placebo patients received DCVAX after progression so for simplicity lets assume we have just one group of 331 patients most of which received DCVAX at some point as a treatment for GBM after SOC and ask is there compelling statistical evidence that DCVAX is effective?
So what else do we know about this population? Median enrollment was May 2014
Final enrollment was November 2015
100 of 331 patients were still alive June 2017
So what do we know about GBM Overall Survival?
From what I can gather and what has been posted the GBM patient survival with SOC is something like this: 1 year --> 50%
2 year --> 30%
3 year --> 15%
4 year --> 8%
5 year --> 5%
So how do the 331 DCVAX patients compare to SOC statistically?
Lets first look at the 100 patients still alive June 2017
For simplicity lets break this into pieces. Lets first look at the 166 patients enrolled prior to the median enrollment in May 2014. We received a very detailed patient enrollment timeline from NWBO CTO Marnix Bosch August 31, 2017.
Looking closely at the we see that prior to October 2011 we had about 7 patients enrolled. From October 2011 to the midpoint in May 2014 enrollment is pretty linear at about 5 to 6 a week which represents a roughly even distribution of patients ranging from 38 to 68 weeks. Knowing that SOC has about 5% survival at 5 years, 8% at 4 years and 15% at 3 years it is a simple matter to create a spread sheet and determine that if the population of 166 had this type of survival rate then by June 2017 we would expect to have only 12 of the 166 patients to still be alive.
Now if we look at enrollment from the median enrollment May 2014 through the final enrollment it too was fairly linear but a little steeper at 10 or 11 patients entering the trial each month up to final enrollment October 2015. So now comparing that 100 patients alive June 2017 to the 60 patients that would have been expected to be alive using the GBM typical SOC survival rates we have almost 1.7X (~40%) more patients still alive with DCVAX vs expected SOC with a P value of 0.0003.
The P value as most know is the probability of rejecting the null hypothesis when it is in fact true is called a Type I error. In this case the hypothesis would be that DCVAX overall survival is better than SOC . . . So even in June 2017 this is starting to look pretty compelling already even with this simplistic highly conservative (pessimistic) view.
Try it yourself. There is an easy to use, easy to understand statistical tool you can access with this URL if you want to play around with the data yourself https://www.medcalc.org/calc/relative_risk.php
The equations used are described in detail and it give you some other interesting statistics. For example it provides the Number Needed to Treat (NNT) metric which is the estimated number of patients who need to be treated with the new treatment rather than the standard treatment (or no treatment) for one additional patient to benefit (Altman 1998). In this case it is 8.275. It also provide the 95% Confidence Interval for NNT which is 5.393 to 17.770 patients. Here also is the URL for the 95% confidence interval is calculated according to Daly (1998) and is reported as suggested by Altman (1998). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1114210/
I was curious so found it and read through it.
Even though I have been trying to be the most conservative and pessimistic possible with this analysis if there is general agreement with the normal GBM survival rates I have listed then it would not be a stretch to increase the population size of this group since we certainly have a much much large number of patients that have received SOC.
So for example if we do that and compare: DCVAX
Patients with bad outcomes ==> 231 (331-100)
Patients with good outcomes ==> 100
Patients with bad outcomes ==> 271 (331-60) instead make it 100X ==> 27100
Patients with good outcomes ==> 60 instead make it 100X ==> 6000
This improved the P value to 0.00001.
The 95% Confidence interval for NNT narrow to 6.148 to 12.650 patients
So given this why would Linda want to keep going with the trial since this already in June 2017 looks fairly compelling?
One reason might be that there may be some debate about the 1, 2, 3, 4, 5 year OS for SOC that I have listed here. There are certainly those that don't want DCVAX to succeed so why not try to raise doubt and stir the pot. Also it might be argued that FDA has a stringent statistical approach that must be followed and this is not it and this is not the 2 patient population design established for the Phase-3. That is all well and good but also contrary to what the new and improved FDA is looking to do so we will see.
Perhaps more important is that if we are focused on the long tail and we wait until the end of 2018 then all the patients will have been in the trial for just about 5 years or more. The numbers assuming 2 patients a month die between June 2017 and then will look like this now. DCVAX
Patients with bad outcomes ==> 267
Patients with good outcomes ==> 64
Patients with bad outcomes ==> 304
Patients with good outcomes ==> 27
So it is almost 2.4X better survival with 60% more patient surviving . . . WOW
The P value is still the same 0.00001. (hard to get any better than this)
The 95% Confidence interval for NNT narrow to 6.148 to 12.650 patients (this is with the larger sample size)
Do we have to wait that long? What does it look like now?
If we look right now Feb 2018 assuming 2 patients a month continued to die since June 2017 it will look like this now. DCVAX
Patients with bad outcomes ==> 243
Patients with good outcomes ==> 88
Patients with bad outcomes ==> 286
Patients with good outcomes ==> 45
So it is almost 2X better survival with about 50% more patient surviving right now.
The P value is still the same 0.00001. (again hard to get any better than this)
The 95% Confidence interval for NNT narrows to 5.183 to 10.790 patients (this is with the larger sample size). Hmmm . . . so that is interesting. If we were to cut and run now we could claim 2X improvement in overall survival with DCVAX with a extraordinary P value of 0.0001. If we wait until year end and still continue at 2 deaths per month we will have a 2.4X better overall survival with DCVAX vs SO with the same extraordinary P value of 0.0001, but the confidence interval gets a bit worse. Anyway one looks at this trial data DCVAX gets approved.
It is just a question of when and my take is now is as good if not better than year end. Perhaps the expert statistician that looks at the trial data each spring will reach a similar conclusion. Also keep in mind this is a very very conservative pessimistic analysis assuming we simply look at the whole 331 group of patients compared to SOC. In reality there are many other ways to slice this that will be even more compelling. Many of us have already discussed that DCVAX may be effectively a cure for the worse performing mesenchymal subgroup so it would be very interesting to see that data broken out separately. It will be interesting to see if the paper we are all sitting on the edge of our seats waiting for reaches similar conclusions. They certainly have even more detailed data than I even with them still blinded.
Keep me honest. If you see any mistakes or errors in my analysis let me know. This is hot off the press.