0

It has been a long time since I last performed a survival analysis, and I have a basic question about it. I have three cohorts, each with a different median follow-up time value that I have calculated using the reverse Kaplan-Meier estimator. The R code I used is:

install.packages("prodlim")
library(prodlim)

quantile(prodlim(Hist(FOLLOW_UP_TIME, event=EXITUS)~1, data= COHORT1, reverse=T))

quantile(prodlim(Hist(FOLLOW_UP_TIME, event=EXITUS)~1, data= COHORT2, reverse=T))

quantile(prodlim(Hist(FOLLOW_UP_TIME, event=EXITUS)~1, data= COHORT3, reverse=T))

The median follow-up times are 100 (70-115), 60 (45-80), and 30 (15-45), respectively. My question is, which maximum time should I use in the plots, and at which time should I give the several survival and cumulative incidence probabilities?

Can anyone help me?

Thank you very much.

EI_Stats

EI_Stats
  • 21
  • 2
  • First off, you should include needed `library` calls. Apparently you are using the `prodlim` package. Second, why do you think you should impose a limit on the plots? (This is probably a better question for stats.stackexchange.com , since it has no data and is woefully incomplete as a coding question.) – IRTFM May 07 '23 at 00:45
  • @IRTFM First of all, thank you very much for your response. I have already added the library calls. Secondly, I have already asked the same question on stats.stackexchange.com. You are right that perhaps there is no need to cut the graph at any point, but my question is more about whether data from groups with longer follow-up times should be censored to avoid bias, and at what time it would be optimal to give survival probabilities. I imagine that, for example, if you establish giving probabilities at the follow-up time of the group with a lower median, you lose some information. – EI_Stats May 08 '23 at 07:56
  • When you ask about "bias" you need to say what sorts of effect measures you are talking about and why you think that they might be distorted by changing the times at which they are calculated. The plot of all the data should allow you to read off comparisons at particular follow-up intervals without difficulty. The survival fractions are always going to be conditional on the data available to a particular follow-up interval. – IRTFM May 08 '23 at 15:24

0 Answers0