Does sample size calculators for non-inferiority take prevalence into account?

[Sara Hansen]

When using sample size calculators such as ssi or artbin for non-inferiority study designs, does it take the prevalence into account?
For example, I want to calculate the sample size for a study investigating if methodB is non-inferior to methodA (with a 5% non-inferiority margin). The outcome is sensitivity, so I will test if the proportions are non-inferior.
I use the following code:

ssi 0.87 0.05, a(0.025) power(0.9) non

I get the result that I need 951 in each group. My question is: since sensitivity is calculated on true positives, do I need 951 true positives in each group? Or do I just need 951 in total in each group?

I hope it makes sense, and thank you so much for your time in advance.

[Ian White]

Sara, this is probably too late for you, but let me reply anyway.

You can get the same sample size using artbin, pr(.87 .87) margin(.05) power(.9) wald. This takes the prevalence of the outcome (expected to be 0.87) into account.

But I think you are asking a different question, about true positives vs all-comers. If your trial is to compare the sensitivities, then you will only analyse true positives, and the sample size (1902) given is for true positives. If you want to calculate how many all-comers you need to recruit in order to get 1902 true positives, then that's a non-standard calculation for a clinical trial, and you will have to work it out yourself using the prevalence.

I'd also point out that comparing two methods is usually done by applying both methods to each individual, in which case you don't have independent samples & shouldn't use artbin: instead look at power pairedproportions.