I agree with you. Actually I had done it that way too. Remember that I had to have the “39%” before I could break it down to month by month percentages. But if presented it that way was because: - It was more illustrative for the process doing it month by month, and I already had got the “month to month” percentages from my previous projections for both Arms mixed
- This way I got fewer number of events. I was being conservative. If I was already impressed by the numbers in the “month by month” way, just imagine doing it as you mention... It gave me like 5 to 6 more events. (Of course, this method would be “scraping up” events that the other one would miss)
I don’t know about that. You may be right, but I took my time on this one. I have had this question in mind from the very beginning, but you see, I also have month 0 , which is the month that patients are enrolled. So month #1 means that a whole month passed by, and I consider that enrolling a patient means starting the treatment right then.
Yes. Doing it month by month I had to use rounding of the numbers to the closest integer, otherwise I would have got half the numbers and that was way too conservative. Had it done as you mention I only had to round the numbers once, instead of the many roundings as months had that group of patients I did
I guess we both agree that we would need something around 150 events for the 1st look-in, regardless everything else and if you say that “my numbers” were not completely off the wall… I suppose it is not completely off the wall to believe that the number of events for the Bavi Arm might be half the numbers than the Placebo Arm.
Thanks so much for your kind reply and for paying attention to my first cumbersome post.