what could go wrong?, page-113

An interesting series of posts here. Neo, you explained your position very clearly as regards risk/reward. However, I regard the probability of success (i.e. > 8+ week median PFS experimental over control) as 80% plus with the upper limit to the SP being primarily controlled by the degree of median PFS in excess of 8 weeks – say 80c to $1.50. This of course changes the risk/reward equation substantially and makes ACL a great bet according to my opinion.

My 80% figure is of course pure assumption, as is your 50% assumption. My confidence is for the reasons as follows (all my opinion only).

1. The biggest unknown in the trial is the effect of HA+FOLFIRI cf FOLFIRI alone on PFS. It was designed by professionals to measure this.

2. As part of the design they would have assumed a ‘typical’ control arm PFS interval based on available historical data.

3. Based on this ‘typical’ PFS assumption the trial is indicating a median PFS far greater than 8 weeks – some estimate up to 20 weeks atm)

4. A major confounding influence on the trial duration could be the ‘typical’ PFS assumption. If, in fact, it is erroneous (for this large cohort) its ‘atypicality’ (wrt initial assumptions) should be distributed roughly equally between the two arms (i.e. not reflected more in one than another).

5. At the current trial duration there is quite a bit of padding (perhaps up to 12 weeks) to allow for this ‘atypicality’ (if it, in fact, exists at all).

Despite my opinion, we are heading for a binary outcome in this instance, so good luck to all those who are brave (foolish?) enough to hold through to the result. [I am still haunted by my effort in turning a substantial profit in PBT to a minor loss, by holding through an unexpected negative result – so don’t listen to me - but I am going to plough on undeterred.]