Response to Lighrock's Insights for P3 Calculator Variables
Lightrock,
Responding to your comments on Cancer Stage: I believe that we don't know each enrollee's cancer type; however, the "guardrails" are Stage III SOC at best and StageIII/IVa at worst. This is why the calculator examines two different control arms. One assumes StageIII SOC adjusted by the Clinical Trial Effect variable and the other assumes the Stage III/IVa SOC adjusted by the Clinical Trial Effect number. In the Control Arm Survival Inputs tab one can redraw the survival curve by changing the five numbers for each control arm scenario. For example, one could study an average of the two "guardrails" and look at Overall Survival in that scenario. That new scenario is subsequently modified by the Clinical Trial Effect number to establish the survival curve for that scenario. I don't have a dog in this hunt. I really don't know what to expect and if anyone has clear input for these curves I would appreciate hearing about it.
Regarding the "...% responding to injection assumptions" you mention below, I'm not familiar with this? Is this an assumption that the Phase III trial will mirror in some way the Phase II trial in that 2 of the 19 enrollees had their tumor completely disappear in three weeks before the SOC regimen could begin? I'm unclear. Would appreciate more information on this concept. I can certainly add more variables to this calculator.
I too considered analyzing this stochastically and also using my Bayesian analysis which I developed in my day-job. What I learned doing both on my day job with life settlement portfolios is that with around 900 participants there is little new information to be learned. This is one reason why the 298 is deemed to be "statistically significant." The right-handed barbell appearance of the enrollee census complicates "gut reaction" or "shirtsleeve guesses." Excel guesses hundreds of times to identify a single survival function that when applied to the enrollment census results in the exact number of survivors who didn't die in the control group (for that scenario). I SUSPECT that the "separation" that you "observed between Q2 and Q4 last year" was indeed related more to the odd enrollment rates. I don't know for sure.
I apologize to everyone for using fractional people! It would have been a lot more work to integerize this (if such a work even exists!) It think you end up at about the same place anyway.
Cheers from a wet Oregon. Freezing level dropping tomorrow!!
Jim
News 
Market Data 
Discover 
