Change risk table order

[saskia-c]

When plotting a cuminc() plot and adding a risk table, the order of the strata in the risk table is alphabetical. Is it possible for the order to follow the levels of a factor?

In the example below, I would like Drug B to be the top row in the risk table, and Drug A to come second.
It seems by default the risk table does not follow the levels of the factored stratum.

# data
trial_new <- trial %>%
  mutate(trt = factor(trt,
                      levels = c("Drug B", "Drug A"),
                      labels = c("Drug B", "Drug A")),
         cr_response = ifelse(response == 1, 1, death*2))

# plot
cuminc(Surv(ttdeath, factor(cr_response)) ~ trt, data = trial_new) %>%
  ggcuminc() +
  add_risktable(risktable_stats = "{n.risk} ({cum.event})") 

Not sure if this a bug or a missing feature, but any help on this would be greatly appreciated. Many thanks!

[ddsjoberg]

Dear @saskia-c , That seems to be a very reasonable expectation! Thanks for the post

[MichaelDismorrMD]

@saskia-c A temporary fix is to adjust the levels in plotobject$data$strata.
E.g. plot$data$strata <- forcats::fct_rev(plot$data$strata)
The order of the factor levels there seem to decide the order in the risk table

[lsIQ1]

@MichaelDismorrMD appreciate the workaround, I reordered the levels of my strata using this, however this unfortunately doesn't appear to apply to the risk table strata symbol colours, it appears to revert back to alphabetical order of the strata, eg.

# Reordering strata levels
cuminc_plot$data$strata <- forcats::fct_reorder(
  cuminc_plot$data$strata,
  "one",
  "two",
  "three"
)

# Adding strata symbols to risk table
cuminc_plot +
add_risktable_strata_symbol(symbol = "\U25CF", size = 10) 

results in a risk table that is correctly ordered one, two, three, however the strata symbols are back in the original alphabetical order (one, three, two). Any suggestions here?

[dchiu911]

A fix that works for reordering for both add_risktable() and add_risktable_strata_symbol():

library(purrr)
library(dplyr)

trial_new <- trial |>
  mutate(trt = factor(trt,
                      levels = c("Drug B", "Drug A"),
                      labels = c("Drug B", "Drug A")),
         cr_response = ifelse(response == 1, 1, death*2))

cuminc(Surv(ttdeath, factor(cr_response)) ~ trt, data = trial_new) |> 
  modify_in(list("tidy"), \(x) arrange(x, strata)) |>
  ggcuminc() +
  add_risktable(risktable_stats = "{n.risk} ({cum.event})") +
  add_risktable_strata_symbol(symbol = "\U25CF", size = 10)

It appears that although the levels are preserved in the "tidy" data the risk table functions use the original alphabetical ordering in the strata variable instead of its factor levels.

[ddsjoberg]

Hi @dchiu911 @lsIQ1 @MichaelDismorrMD @saskia-c , thank you for the posts!

Can you please install the latest version of tidycmprsk and try again? Thank you!