Merging survival curves on a KM fit model

[qalid7]

First of all, many thanks for this absolutely wonderful package! I use it pretty much on a daily basis!

I've seen the two related issues (#97, #226), and have also tried the useful pairwise_survdiff.R function.

However, my suggestion/request is simpler. How can I merge survival curves within the same fit?

For example, merge the three groups (red, purple and green) in the below fit (into one group/curve) and see how the separation would look like against the cyan group - does that make sense?

Basically, how to produce the survival curves presented as shown in (#97)?

Working on TCGA Lung, I can tell this would be a formidable tool especially when we combine 2+ variables on a KM fit.

PS. I realize sex shouldn't be used in surv_cutpoint, this is just a quick reproducible example.

data(lung)
cut <- surv_cutpoint(lung, time = "time", event = "status", minprop = 0.2,
                             variables = c("ph.ecog", "ph.karno", "pat.karno", "sex"))
cat <- surv_categorize(cut)
fit <- survfit(Surv(time, status) ~ ph.ecog+sex, data = cat)
ggsurvplot(fit, data = cat, risk.table = TRUE, pval=TRUE)

Thanks a lot!

[kassambara]

Hi,

The easiest way to go is to create a new grouping variable into the cat data frame.

For example:

library(dplyr)

data(lung)
cut <- surv_cutpoint(lung, time = "time", event = "status", minprop = 0.2,
                     variables = c("ph.ecog", "ph.karno", "pat.karno", "sex"))

cat <- surv_categorize(cut)
cat <- cat %>%
  mutate(custom.grps = ifelse(ph.ecog == "low" & sex == "high", 1, 0))

fit <- survfit(Surv(time, status) ~ custom.grps, data = cat)
ggsurvplot(fit, data = cat, risk.table = TRUE, pval=TRUE)

[qalid7]

Wonderful! Thanks.