Posting: # 20932
I am usually not afraid of thinking in untraditional ways. Here's an concept that just popped up in my mind while I am waiting for a flight at MCO.
When we calculate sample size some of the usual clauses relating to ethics and science are ICH E6 § 6.9.2, 2.2, 2.3, or E9 § 3.5. All good. they indeed support the usual concept of first defining a minimum level of power then going ahead with a sample size calculation using some assumptions.
How about doing it differently, for ethical reasons? We could apply e.g. E6 §2.2 and 2.3 in a different way. Remembering that Hötzi a few times argued that the loss of a few subjects does not necessarily mean the loss of a lot of power, I am thinking that we could look at the benefit associated with the 'last subject(s)'.
In stead of aiming for a minimum power per se, or in addition to doing so, we could in principle (also) say that we are only going to add subjects as long as they contribute by some ethical margin X to the power of the trial. If we have 82.1% power with 36 subjects, and 82.2% with 38 subjects then one might say the addition of the last two would not be justified, because adding 0.1% points of power does not by some criterion or other make much difference and should therefore not take place
What do you think about this? [kindly note: I am not asking if you think this would be acceptable to a regulator, what the guideline specifically say, or what a reasonable value of X is. I am just asking what you think, from an ethical perspective, about approaching sample size this way.] One could of course combine minimum power and minimum added power per subject in one and the same calculation, if necessary. Or have other flavours of a similar theme.
Pros and cons, anyone?
Le tits now.