Thank you for that. Clinical trial design is complex!Yes, you...

Thank you for that. Clinical trial design is complex!

Yes, you make some good points. Perhaps I could have expanded on the generalisability clause that I mentioned, which together with the analogy of fishing from the same fish pond, aimed to address.

Simply increasing the sample size comes with many issues!

On its own, simply increasing the number of recruits in the hope of achieving diversity will not always resolve issues of diversity. Some Covid vaccine trials had huge numbers, but the samples were never representative of the populations that were to receive those vaccines.

Thats why, in my view, fishing from a different fishpond has greater nenefits, than simply increasing the numbers that you pull out from that one fishpond, whose inhabitants already share a lot.

How do you even decide the number you need to recruit in order to achieve some measure of diversity? Will you dismiss a sample size that's been determined based on declared numbers and methods that everyone can see and use, and replace that with some large number that comes to mind? And then what: will you do subgroup analysis- and if so, will those smaller group analyses have the power you require to show a difference?
And if the results from the small group are opposite to the whole sample: will that invalidate your results?

May I add that, since clinical trials are designed with specific objectives, as per the starting hypothesis, in the frequentist world, you do a sample size calculation specifically to find the number that will help you address those requirements. You can not then simply recruit more (and expose them to drug X or Y) for an unspecified objective that has nothing to do with achieving the stated objective!

To me, doing this invalidates your hypothesis. By this, i mean: if your hypothesis leads you to needing 200, but you recruit 400, how do you analyse the data? What if the first 200 fail to achieve statistical significance, but adding the next 200 gets you across the line? Do you reject or accept the null hypothesis, and at which point?

Finally: on ethics! A human clinical trial is an experiment on humans, and as such, it will generally get stopped as soon as the objectives of the experiment have been achieved and the answer is obvious, or where it has become clear that the objectives will not be met on the planned sample size (futility). You can not simply power on for the sake of achieving diversity. Rather, you must add the factor (diversity) at the planning stage, and get the appropriate sample size at the very start.

It's complex, and debatable!