A few months ago, I posted a query regarding code for a sample size
estimate for a one arm survival trial. Below is some code I created to
calculate such an estimate - perhaps it may be of some use.
#cox.pow computes sample size for a one arm survival trial. med.0 is
the null median
#survival, med.a is the alternative median survival, a.time is the
accrual time, and
#f.time is the follow up (assumes constant accrual and two sided test).
Approach taken
#from Lawless 1982, p. 108
cox.pow<-function(med.0,med.a,a.time,f.time,alpha,beta) {
#initialize number of events to 0
r<-0
#find necessary number of events
while (qchisq((1-alpha/2),(2*r+1))*(med.0/med.a) >qchisq(beta,(2*r+1)) )
{r<-r+1}
#find proportion of patients that will experience event for given
median survivals,
#accrual and follow up times (formula from Cox 1983)
p <- 1 - (exp(-0.69*f.time/med.a)*(1 -
exp(-0.69*a.time/med.a))/(0.69*a.time/med.a))
#sample size is r/p
n<-round(r/p,0)
return(n)
}
#compute necessary sample size for alternative median 9.3 vs. null
median 6.3 with 20
#months accrual and 4 months follow up
cox.pow(6.3,9.3,20,4,.05,.2)
Regards,
-Cody
Cody Hamilton, Ph.D
Institute for Health Care Research and Improvement
Baylor Health Care System
This e-mail, facsimile, or letter and any files or attachmen...{{dropped}}
