Background: at half-year follow up times for 4y, patients may switch to a different medication group. To account for this, I've converted survival data into counting process form. I want to compare survival curves for medication groups A, B, and C. I am using an extended Cox model but want to do pairwise comparisons of each hazard function or do stratified log-rank tests.
pairwise_survdiff throws an error because of the form of my data, I think.
x<-data.frame(tstart=rep(seq(0,18,6),3),tstop=rep(seq(6,24,6),3), rx = rep(c("A","B","C"),4), death=c(rep(0,11),1))
survdiff in the
survdiff(Surv(tstart,tstop,death) ~ rx, data = x)
I get the error:
Error in survdiff(Surv(tstart, tstop, death) ~ rx, data = x) : Right censored data only
I think this stems from the counting process form, since I can't find an example online that compares survival curves for time-varying covariates.
Question: is there a quick fix to this problem? Or, is there an alternative package/function with the same versatility to compare survival curves, namely using different methods? How can I implement stratified log-rank tests using
survidff on counting process form data?
NOTE: this was marked as a known issue in the survminer package, see github issue here, but updating survminer did not solve my issue, and using one time interval, tstop-tstart wouldn't be correct, since that would leave, e.g., multiple entries at 6 months rather than out to the actual interval of risk.