Cel Sci Corporation (CVM)

[przgwxl]

@Bio: good response. I spent the morning doing an analysis. For me, the objective was to simplify matters. The product of the computation is the probability that the trial exceeds 10% OS improvement. The key independent variable is the 5 year survival probability for the control arm.

I get that if the 5 year survival probability for the control arm is less than 50%, then there is a > 90% probability that the trial is successful. This varies some with dropouts, I had the cumulative dropout percentage as 15.14%.

Everything moves around a bit with dropouts, but the key variable is the control arm survival at 5 years.

This is a much more interesting way of looking at things than computing the median impact on OS. For the numbers cited (15.14% dropouts, 50% OS survival in the control arm), the median effect is a 20.4% in OS. However, the probability that the trial fails is a non-negligible 17%. Put another way, the median effect is not representative of the distribution of outcomes as a whole, which appears to be fairly dispersed.

Having said all of this, I am liking the chances. I am thinking that for reasonable (conservative) OS numbers and dropout rates, the trial is likely to succeed with a 90% or better probability. If patients are really doing great with SOC, meaning >52% OS at 5 years, well, we are on a knife edge.