(Total Views: 411)
Posted On: 02/25/2021 10:17:38 AM
Post# of 145871
Respert24,
You are right, however, remember that nobody knows in advance how many "events" (in our case percentages of deaths/survivals) will take place.
(One just can assume lots of things when designing a trial.)
((Or just "play it safe" as BP does (lots of money and time). With enough patients one does not have to worry about this (power).))
(((When one is "poor" or time is of the essence, the trial is designed to meet its objectives in some amount of time with some limited (if you will) amount of patients.)))
((((And, then, you take a "peek" midway (or 75%) and see that your assumptions were not quite correct and try to adjust the parameters (i.e. patients, end-points, frequencies) to meet them at the end.))))
(((((And, of course FDA has also to decide what is "approvable" or not (level of power good enough).)))))
(((((((So, with initial design everything is good. But life is not always what we think. It is mathematically quite possible to meet the false positive but not the false negative (even though both are dependent of trial parameters like the number of patients and the number of evens measured). This depends on how "tight" you design the trial with ones assumptions in hand. For example we could have said: let's do 780 patients ... sure, better statistics but ... you get the point.))))))
(((((((
)))))))
You are right, however, remember that nobody knows in advance how many "events" (in our case percentages of deaths/survivals) will take place.
(One just can assume lots of things when designing a trial.)
((Or just "play it safe" as BP does (lots of money and time). With enough patients one does not have to worry about this (power).))
(((When one is "poor" or time is of the essence, the trial is designed to meet its objectives in some amount of time with some limited (if you will) amount of patients.)))
((((And, then, you take a "peek" midway (or 75%) and see that your assumptions were not quite correct and try to adjust the parameters (i.e. patients, end-points, frequencies) to meet them at the end.))))
(((((And, of course FDA has also to decide what is "approvable" or not (level of power good enough).)))))
(((((((So, with initial design everything is good. But life is not always what we think. It is mathematically quite possible to meet the false positive but not the false negative (even though both are dependent of trial parameters like the number of patients and the number of evens measured). This depends on how "tight" you design the trial with ones assumptions in hand. For example we could have said: let's do 780 patients ... sure, better statistics but ... you get the point.))))))
(((((((
![](/m/images/icons/icon_smile.gif)
![](/m/images/icons/icon_smile.gif)
![](/m/images/thumb-up.png)
![](/m/images/thumb-down.png)