I have basically been keeping a hidden factor of safety in all of my numbers for quite some time. I have modeled and calculated the time of 32 events based on enrollment, not diagnosis. The delay between diagnosis and enrollment will vary from person to person, but it is about one month. You can look at a few blogs to verify this. So when IMUC tells you that they modeled 32 events at the end of Q1 with a 9 month survival advantage 80% of the time, that is calculated based on diagnosis, not enrollment. Everyone who has been modeling based on enrollment data and has piggy-backed off of my numbers will have the same "hidden" factor of safety. So when I tell you that 50% of the time, if the median survival was 18.8 months that 32 events would have happened mid January if no benefit, there was some padding there because the calculation was based on enrollment, not diagnosis. It really would have been early December. Just thought you would like to know so you can be prepared.
32 events is coming very soon, whether they company was told or not. We really shouldn't go much longer, even if OS is 38 months, if we only have 14.6 months OS for the control group. Yes it is longer with 38 months OS and 18.8 OS, but 18.8 OS was for 100% fully resected, and 38 months was for 75% fully resected. Even though we were told 75% fully resected two CCs ago, it should really be less than that, if the population in this trial is in anyway representative of the GBM population.
I think, and this is my opinion and only my opinion, what we are missing is that amongst all the variables in a complex drug study we have add the variables that the FDA is a federal agency with all its inefficiencies, ineptness, and regulations beyond belief. This all adds time and cost to any process. Any time you have to depend on our federal government in a process inefficiencies and time are inserted into the mix. Statistics being a pure science can’t account for these variables. I am strong and long on IMUC knowing that as bad and slow as our federal management is it will at the end of the day make the correct decision.
“So when IMUC tells you that they modeled 32 events at the end of Q1 with a 9 month survival advantage 80% of the time, that is calculated based on diagnosis, not enrollment.”
That would be really odd. What makes you believe this?
Well if I take the 18.8 curve and the 38.4 curve off of the IMUC presentation and put them in my model - on average I see the 32nd event happening at the end of Sept (I don't have my computer right now so I don't remember the range on my 250 runs) and my model starts at diagnosis.... so I'm not sure why we all get numbers that are so different - I know the shape of the curve has a lot to do with it... try the curves off of the presentation and tell me what y'all get. bdoglo - try those also and tell me your outcome.
Our numbers are not different. For the end of July, for the 14.6 months OS group, I have 30.5118 deaths. And for the 18.8 group, I have only 25.2238 deaths. So the fact that you modeled the end of September is not a surprise at all for the median. You probably also modeled early August for 14.6 months.
The first key issue is that our prediction is based on ENROLLMENT, not diagnosis. That's why your numbers are so different from what the company predicts. I doubt that you concluded that in 80% of your runs, you would have a 9 month survival advantage by the end of March. The reason they did is because they have diagnosis numbers. If you had diagnosis numbers, you would have 32 events occurring in mid August with the same curve based on 18.8 numbers.
The second key issue is that I really doubt that we have a 38.4 month OS because even though Dr. Yu said that he assumes that we have 75% fully resected in this trial, I really doubt that.
If you use something like 17 months OS, and the ICT-107 group having a 32-34 months OS, then you figure that there is a delay of about a month or a month and half between enrollment and diagnosis, you probably get 32 happening, not being reported, this month.
I can go as far as 18.8 / 43.8 months and still hit 32 in June. The point for me is not thinking that this is somehow going to be an accurate spot on number, rather, to recognize just how far over the top this is likely to be when it all shakes out. This is going to be recognized "out there" soon enough.
Keep up the good work Disco!
I think we will hear about the 32 events in Q3.
One question. After the 32 events, will we see the interim at the end of 2013, q1 2014? or will this depend on hitting 64 events?
Could you send me (by yahoo-mail) your spreadsheet/enrollment data?
Sentiment: Strong Buy
We won't see 64 this year. No one has modeled 64 events that early. Gengos evened softened his stance to maybe 64 events before the end of this year. Maybe Q2 2014 or later. Trial stops at the end of Oct 2014 whether we have 64 or not. But the first patient enrolled in this trial in Feb 2011 and we still don't have 32 deaths. How are we going to have 64 deaths by the end of this year when there are only 41 control patients (approximately) - rhetorical question?