Power calculation for sequential design with multiple endpoints (one binary superiority & one NI)

meganmoreton — Wed, 18 Mar 2020 12:30:51 GMT

Hi,

I am trying to perform a power calculation for a 2-arm randomised clinical study using a sequential design study with 2 outcomes - I having some difficulties.

The study has 2 primary outcomes which should be evaluated in a hierarchical manner (ie, outcome 2 is only evaluated if the null hypothesis for outcome 1 is rejected). The 1st outcome is a binary efficacy outcome for which superiority is necessary (expected 20% absolute difference - 2-sided alpha 0.05). The 2nd outcome is a safety binary outcome for which non-inferiority is necessary (10% margin - 1-sided alpha 0.025). The study has 3 equally spaced analysis timepoints & should be stopped for futility (ie, efficacy not established, null not rejected) or safety harm (ie, safety outcome not non-inferior, null rejected).

Does anyone have any advice on how to program this? I currently have something like this to cover outcome 1, but I am unsure of how to incorporate the non-inferiority outcome 2.

I really appreciate any help,

Megan

ods graphics on;
proc seqdesign altref=0.2
;
OneSidedOBrienFleming: design nstages=3
method=obf
alt=upper stop=accept
alpha=0.025 beta=0.20
;
samplesize model(ceiladjdesign=include)
=twosamplefreq(nullprop=0.3 test=prop);
ods output AdjustedBoundary=Bnd_Prop;
run;

topic Power calculation for sequential design with multiple endpoints (one binary superiority & one NI) in Statistical Procedures

Power calculation for sequential design with multiple endpoints (one binary superiority & one NI)