drop out rate in proc seqdesign

[swannie]

Hi,

I need some help figuring out how to use the loss=exp(hazard=hazard rate) option in proc seqdesign correctly.  If the trial is expected to have 10% drop out total, but most is early (within 4 months of randomization), what's the proper (or best) way to calculate the hazard rate?  Any help vastly appreciated.

swannie

%let drop_out=0.0027;

ods graphics on;

proc seqdesign plots=boundary(hscale=samplesize) boundaryscale=stdz errspend ;

OneSidedOBrienFleming: design nstages=2 method=errfuncobf alpha=0.02 alt=upper beta=0.10 info=cum(0.50 1) stop=both(betaboundary=nonbinding);

samplesize model(ceiladjdesign=include)=twosamplesurvival (nullhazard=0.0608 hazard=0.0385 weight= 1 1 accrual=uniform acctime=&rectime. loss=exp(hazard=&droprate.) totaltime=&total_time. ceiling=n);

run;

ods graphics off;

[windfor]

The dropout rate were also converted to group loss hazards in the same way. Therefore, the 5% dropout rate for 70 months duration was corresponding to the group loss hazard of -ln(1-dropout rate)/StudyDuration =-ln(1-0.05)/70= 0.00073.