topic Re: Sample Size for Non-inferiority with interim analyses (group sequential) study in Statistical Procedures

Sample Size for Non-inferiority with interim analyses (group sequential) study

AndreyMyslivets — Mon, 02 Nov 2015 20:49:52 GMT

Hello everybody!

I would be grateful if you could help me with calculations of sample size if I have following data&colon;

1. Outcome is binary (Yes/No), main hypothesis is "non-inferiority";

2. There are two planned analyses (60% - interim, 100% - final);

3. Non-inferiority margin = -20%;

4. True planed difference between reference and test drugs = -5% (reference effect=56% - Yes, test effect=51% - Yes);

5. Test one-tailed, alpha=5%, betta=80%.

I'm looking forward to hearing any suggestions or recommendations for solving this problem.

Best Regards,

Anrey

Re: Sample Size for Non-inferiority with interim analyses (group sequential) study

lvm — Thu, 05 Nov 2015 01:44:17 GMT

I recommend that you check out PROC POWER. It should have what you need. Also, read:

http://support.sas.com/kb/48/616.html

Re: Sample Size for Non-inferiority with interim analyses (group sequential) study

AndreyMyslivets — Tue, 10 Nov 2015 07:46:35 GMT

Thank you very much for your answer!

I think that I understand some key points at the moment and want to share with you (hope it is interesting) .

First of all there are 3 procedures in the SAS for SS calculation:

1. power procedure

2. glmpower procedure

3. seqdesign procedure

It is necessary to make some alpha (or/and beta correction for the analysis). The seqdesign procedure can present the alpha and beta spending functions for correction the SS in any sequential design (There are many types of it, I use Lan-DeMets O'brien-Fleming approximation spending function of alpha in my example).

When I have information about correction, I can use any of this three procedures for calculation using corrected alpha.

Best Regards,

Andrey

Re: Sample Size for Non-inferiority with interim analyses (group sequential) study

sarijaan — Fri, 19 Nov 2021 08:36:02 GMT

Did you have an example of how you made it ?