Power for non-inferiority test of paired proportions

[humenghu]

Hi all,

I want to estimate the sample size for a
non-inferiority test of paired proportions. There is an excellent discussion on
this in PASS documentation: http://ncss.wpengine.netdna-cdn.com/wp-content/themes/ncss/pdf/Procedures/PASS/Non-Inferiority_Tests...

I know that in SAS there is PAIREDFREQ in PROC POWER, but  it seems that there is no straight forward solution for this non-inferiority problem.

Can anyone shed lights on how to solve it in SAS?

Thanks in advance! 

Hu

[Jack2012]

No clear about your meaning. You mean in PROC POWER you can't reach this aim ?

[lvm]

You can definitely do noninferiority power calculations. Here is a good description for some of this. Check the User's Guide for other applications.

48616 - Design and analysis of noninferiority studies

[CQ_Deng]

the link 'design and analysis of noninferiority studies' included only an example of non-inferiority test of two independent proportions. However, the original question is about the non-inferiority test of two paired proportions or two correlated proportions. For two correlated proportions, the proc power has a statement of pairedfreq. Not sure if proc power with pairedfreq statement can be used to calculate the sample size for non-inferiority test for two correlated proportions. 

[Rick_SAS]

I I think this is covered in Catello and Watts (2015). See Table 4 on p. 7. 

[CQ_Deng]

Table 4 on p. 7. In the paper by Catello and Watts are for non-inferiority test for lognormal data or ratio of two geometric means. We would like to know if SAS proc power can computer the power/sample size for non-interiority test of two correlated proportions.

A good example for this issue is illustrated in FDA guidance "Comparative Analysis and Related Comparative Use Human Factors Studies for a Drug-Device Combination Product Submitted in an ANDA" http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM536959.pdf



In the proposed study, each subject will test for two different medical devices (old device and new device). The error rates for using old device and using new device will then be calculated. The assumption is that if a patient is good at using the old device, he/she will be also good at using the new device. In other words, the error rate for using the old device is correlated with the error rate using the new device. If we are trying to design a non-inferiority study as outlined in FDA's guidance, will SAS proc power be used for calculating the power/sample size? The FDA guidance provided examples for sample sizes that were generated using the simulation.