This last week has given me plenty of time to reflect on the announcement from last Friday. Having considered the timing of the announcement , the "finding" by Disco (although I knew the difference between the “enrollment” dates and OS calculations, I did not think about this difference in timing when doing my model since this is a model I have used on several trials. I think/hope the thought that the delay in the enrollment is ~equal to the DSMB delay is a valid idea.), and settling on which OS curve to use - I ran my model again and I’m ready to make my “predictions” for OS. First off I’m using the UCLA 20 month OS curve for the control arm. Although I know this causes heartburn with several of you – you must not underestimate the determination of the patients that willing to enroll in an experimental trial. The failure of many trials is the control arm out-living historical norms – if the control arm does not hit the 20 month OS then we are even better off… With this control OS and an early June announcement – I’m getting an 80% probability of just under 6 months additional OS in the treatment arm. (Disclaimer – we all should remember that there are many assumptions that are built in models and models can only predict the range of possible outcomes for a given set of OS curves – not what will actually occur!) I believe I can say with certainty that the trial will not end in 2013!
My outstanding concerns: the OS and shape of the control arm curve, the change in the manufacturer (I have read a few posts on the internet about some patients in the trial who experienced issues with getting doses – this concerns me greatly and I would assume this is why things where changed mid-stream which is pretty odd and usually does not happen), and the actual delay between the 32nd event and the announcement (I would like for us to get management to speak about this…).
Thanks phishes3. Can you tell me what your median prediction is (50% probability)?
To me, it's less important what month we use, and more important to understand why the results in Phase II are lower than in Phase I. I predicted that the we would have the results sooner because I thought that we would have less than 75% fully resected patients. To me, that precludes the possibility of a 20 month OS. It isn't just the patients have access to a variety of experimental treatments, its the type of patients that you select into your trial. If every patient had a frontal lobe single hemisphere tumor that was fully resected, I would expect that you could get 20 months OS.
But the problem is explaining why ICT-107 appeared to be less effective than in Phase I. My theory is that it isn't less effective, fewer of the patients are fully resected. Because after all, if you say that the patients could live 20 months+ OS in this trial, how long do you propose that the patients would have lived in Phase I if they were placebo patients? 27 months? That is to say that you believe that Phase I is anything other than an outlier. Typically, Phase I results are better than Phase II results, which are better than Phase III results because at each step, the trials typically include a higher percentage of the population.
The problem with me using an OS that high for the placebo group, is that it requires me to use a predictive OS even higher for a would-be placebo group in Phase I which I cannot rationalize or substantiate. How could a placebo group live even longer in Phase I? Furthermore, even if we could produce those type of results with a 75% fully resected population, it still begs the question why we would be using a 75% fully resected population in a double-blind placebo controlled trial anyway?
All that I need to make sense of this is that I can confirm that the fully resected population in this trial is less than 75%.
There's another reason why the numbers drop from phase I to phase III on average. It is because with small sample sizes you can more easily get good results easier just by luck instead of because of real efficacy. Of course it can go the other way also, but only the trials with the good results will advance to the next phase. This means if the drug makes it to the next phase there should be a bias in the estimate of efficacy. It would be an interesting study to see if the drop off is normally more than expected due to this bias.
50% probability with the 20 month control was ~8.5 months (I think the number was like 8.4 but I don't have my model with me right now) ...just think of the 20 months as added insurance... I have modeled a trial that the cumulative OS looked good but failed because the control OS was longer than historical norms.... It's a waiting game now - and from the looks of it, a long one...
(I have read a few posts on the internet about some patients in the trial who experienced issues with getting doses – this concerns me greatly and I would assume this is why things where changed mid-stream which is pretty odd and usually does not happen
Basically what Smith said is correct, that becuase of this there is no chance of FDA approval on this P2, a long expensive P3 will be needed.
Also it seems with this "enrollment-treatment" fudge the OS look far inferior to NWBO. A good time to sell the rest of my IMUC and move on over to NWBO? I am soooooooooo glad I split my investment here long time ago. Like some other posters here I see $1.80-$1.40 this summer.
Shaping up to be ther CLSN disaster as another poster already pointed out.
FWIW, I don't understand how you could possibly feel safer about a company (NWBO), that produced good Phase I results over 10 years ago, and still has not filled enrollment in their trial. Nor could I imagine rationalizing to myself that without them telling me where they where in their enrollment, how could I rationalize that I even know for sure that the trial will fill, let alone be confident that I know it would work. Nor could I possibly imagine trusting someone (Smith) who understands so little, that he actually believes that final results for IMUC will come in Q4 2013.
What was the advantage for Provenge? Less than two mos.. That advantage worked out to be worth over a billion. With the control @ 20 mo., this is a 30% increase in life expectancy. Thousands of % better than the advantage for Provenge. I can live with 6mos. (pun intended).
Hippo - I think you need to remember that there are as many if not more risks with NWBO... When I looked into NWBO I was not impressed with the way it has been managed and made my choice to invest in IMUC - not that I wont ever invest in NWBO in the future but they need to get their act straightened out first... I do think their technology is ok but that wont be any good unless things are run well.
phishes .... I'm starting to wonder if you are still invested here........ (lol)
"just under 6 months in the treatment arm" - now we know why everybody sold. When 2nd quarter holdings are reported, we'll get confirmation of institutional sentiment. Up or down.
off topic for a moment ...............
I don't care for posters who feel it necessary to pat themselves on the back, reminding us of an earlier post. As an example here's my post dated 5/30. : ) : )
interim announcement will happen "shortly" -
I have no doubt whatsoever, company knows the 32nd event has occurred. Cedars-Sinai is one of 25 sites, is it possible the CRO can lock the database on those patients and the doctors have no idea ? I suppose. . .. . .. really, they don't have to guess, being paid $5 million to run the trial by Imuc, the cro called Gengos, told him.
DSMC recommendation soon, 1 or 2 weeks
Thanks for your input. I was waiting to see your take on interim timing.
Question : If I say 17mOS in control, do I add 3 months to tx arm and get 9mo. advantage ? or bio-stat does not work like that ?
ps- hopefully the stat guys will reply to your post and push mine to the bottom.
" "just under 6 months in the treatment arm" - now we know why everybody sold. When 2nd quarter holdings are reported, we'll get confirmation of institutional sentiment. Up or down."
I don't know if I understand... I think 6 months is quite nice, and I see a 50% probability of ~8.5 months OS (50% isn't enough for me to hang my hat on....)
....and YES I did get a low probability of no benefit in the treatment arm but it was less than 2%
" Question : If I say 17mOS in control, do I add 3 months to tx arm and get 9mo. advantage ? or bio-stat does not work like that ?" Most trials that I have modeled are 50/50 enrolled in the control/treatment arms and in these cases, it does work like that (normally I just give a cumulative OS prediction) but in this case it does not work this way because the mix is 33/67 - I'm sure there is a formula that can be used to approximate this conversion but I have not spent time thinking about it yet.