Rogue2.
Wow, that's a fantastic job you have done.
For two reasons:
- First of all I understand it so I can validate it !
- Secondly it supplements perfectly my xl work and soc curves by giving probabilities of success ! Haven't programmed stuff for years I would not have been able to supplement my static XL with such macros or code to be able to run thousands of tries as you have done.
My remarks :
1) Intuitively, I would have put some 70% on Taïwan with end of trial by October. No surprise CVM gives 100% and is always above 10% in 2)
3) Agreed, I had similar finding.
4) Interesting
Now to go further on numbers :
- Have you tried to compute probabilities around a date for 298 ? Having in mind 133 events in 4 Feb 2017 and 208 events in 3 Feb 2018 will help.
- Have you read my interview of Dr Talor. There is a + and a -. The plus si that SoC OS is probably worse than litterature on oscc because most of patients have floor of the mouth cancer and likely more at STIV than STIII: on the low end of the survival curve (I would lower Taïwan by 3% to get 50% by Y5. Also UK soc is probably too good as it shows net survival and not overall survival. Dr Talor often talked about 50% at Y3, even lower than CVM's 55%). The - is that, by design, the study could have up to 15% of dropouts. I could not get the actual number of dropouts by Dr Talor, but I had the original study assumption from an old publication : 784 eligible patients are required to preserve the power of the study. That does not mean by any mean that it will be 15%, that only means that it could potentially hit that number. 5% is my hypothesis of work, but I could be totally wrong, the real max is 15%
Kindest regards and looking forward to your reply,
Fosco
