I have to use a group sequential tests that use an O’Brien-Fleming group sequential design methodology ---to access sample size and boundaries values for a clinical trials that randomized subjects to two treatment groups A or B, within centers. The end point is the proportion of subjects fully cured after receiving the treatment after day 5. The study is powered for 90% with a 2-sided 5% alpha to detect a 20% difference in the primary endpoint from 60% on treatment A to 80% on treatment B. We know, centers are crossed with treatment, in other words subjects in each center has subjects who receive treatment A or B. We do not want to ignore the within center correlation in the outcomes because of homogeneity of subjects and institutional practice within centers. Thus, we cannot use the usual chi-square test to compare these two treatment groups but specifically use GEE model with centers treated as clusters -via proc genmod to help account for the correlation in data. Now the questions Proc Seqdesign does not allow me to use/incorporate gee model under the "samplesize model= ...." see below, so I am not able to adjust for the correlation in computing the SAMPLE SIZE. Can someone advise on how to go around this to get boundary values to use
ODS graphics on;
proc seqdesign altref=0.10
boundaryscale=mle;
OBrienFleming: design method=obf
nstages=4
stop=both
alpha=0.05;
samplesize model=twosamplefreq(nullprop=0.2 test=prop);
ODS output Boundary=Bnd_Count;
run;
ods graphics off;
