Customizing ggcuminc graphs with ggsurvfit

[eloccodes]

Hi there -- thanks for such a lovely R package. Was wondering how to further customize ggcuminc graphs with ggplot2 language. See reprex and the example below. Thank you!

In this example, I am graphing the incidence of locoregional failure in a cancer population with death as a competing risk factor. In sum, I would like to be able to change the color and line type of each of my levels on the graph, as well as the titles, colors, etc. of the legend and axes.

Load libraries

library(tidyverse) # organizing and cleaning data
library(ggsurvfit) # graphing CI curves
library(tidycmprsk) # competing risk cumulative incidence

Create data frame with sample data

df <- data.frame(
ttprog = rexp(100),
status = factor(
sample(0:2, 100, replace=TRUE),
0:2,
c('censored', 'progression', 'death without progression')
),
variable = factor(sample(1:2, 100,replace=TRUE),
1:2,
c('TrtA','TrtB')
))

Create cuminc function

test <- cuminc(Surv(ttprog, status) ~ 1, data = df)

• Failure type "death without progression"
time n.risk estimate std.error 95% CI
1.00 39 0.150 0.039 0.084, 0.235
2.00 14 0.290 0.056 0.186, 0.403
3.00 7 0.290 0.056 0.186, 0.403

• Failure type "progression"
time n.risk estimate std.error 95% CI
1.00 39 0.300 0.050 0.206, 0.400
2.00 14 0.433 0.060 0.315, 0.545
3.00 7 0.479 0.063 0.352, 0.596

ggcuminc which produces a black-and-white gray curve

test %>%
ggcuminc(outcome = c("progression","death without progression")) +
add_confidence_interval() +
add_risktable() +
add_censor_mark() +
labs(title = "Incidence of LRF or Death", x = "Time (months)") +
theme_bw() +
theme(legend.position = "right")

adding ggplot2 language (e.g., to change fill values of lines) produces errors

however, color argument in the confidence interval works

test %>%
ggcuminc(outcome = c("progression","death without progression"), color = c("#54738E", "#82AC7C")) +
add_confidence_interval() +
add_risktable() +
add_censor_mark() +
labs(title = "Incidence of LRF or Death", x = "Time (months)") +
theme_bw() +
theme(legend.position = "right")
Error in ggplot2::geom_step():
! Problem while setting up geom aesthetics.
ℹ Error occurred in the 1st layer.
Caused by error in check_aesthetics():
! Aesthetics must be either length 1 or the same as the data (202)
✖ Fix the following mappings: colour
Run rlang::last_error() to see where the error occurred.

test %>%
ggcuminc(outcome = c("progression","death without progression")) +
add_confidence_interval(fill = "#508050") +
add_risktable() +
add_censor_mark() +
labs(title = "Incidence of LRF or Death", x = "Time (months)") +
theme_bw() +
theme(legend.position = "right")

[ddsjoberg]

Check out the first example in the figure gallery.
http://www.danieldsjoberg.com/ggsurvfit/articles/gallery.html

It provides an example of changing the colors.

gg_default <-
  [survfit2](http://www.danieldsjoberg.com/ggsurvfit/reference/survfit2.html)([Surv](https://rdrr.io/pkg/survival/man/Surv.html)(time, status) ~ surg, data = df_colon) [%>%](https://magrittr.tidyverse.org/reference/pipe.html)
  [ggsurvfit](http://www.danieldsjoberg.com/ggsurvfit/reference/ggsurvfit.html)() +
  [add_confidence_interval](http://www.danieldsjoberg.com/ggsurvfit/reference/add_confidence_interval.html)() +
  labs(title = "Default")

gg_styled <-
  gg_default +
  coord_cartesian(xlim = [c](https://rdrr.io/r/base/c.html)(0, 8)) +
  scale_y_continuous(
    limits = [c](https://rdrr.io/r/base/c.html)(0, 1),
    labels = scales::[percent](https://scales.r-lib.org/reference/percent_format.html), 
    expand = [c](https://rdrr.io/r/base/c.html)(0.01, 0)
  ) +
  scale_x_continuous(breaks = 0:9, expand = [c](https://rdrr.io/r/base/c.html)(0.02, 0)) +
  scale_color_manual(values = [c](https://rdrr.io/r/base/c.html)('#54738E', '#82AC7C')) +
  scale_fill_manual(values = [c](https://rdrr.io/r/base/c.html)('#54738E', '#82AC7C')) +
  theme_minimal() +
  theme(legend.position = "bottom") +
  guides(color = guide_legend(ncol = 1)) +
  labs(
    title = "Modified",
    y = "Percentage Survival"
  )

gg_default + gg_styled

[eloccodes]

Thank you so much for the quick reply! I have definitely pursued the gallery and couldn't get the code to work for me specifically with the ggcuminc function. For example for the exact code from the gallery example:

test %>% ggcuminc(outcome = c("progression","death without progression")) + add_confidence_interval() + add_risktable() + add_censor_mark() + labs(title = "Incidence of LRF or Death", x = "Time (months)") + coord_cartesian(xlim = c(0, 4)) + scale_y_continuous( limits = c(0, 1), labels = scales::percent, expand = c(0.01, 0) ) + scale_x_continuous(breaks = 0:4, expand = c(0.02, 0)) + scale_color_manual(values = c('#54738E', '#82AC7C')) + scale_fill_manual(values = c('#54738E', '#82AC7C')) + theme_minimal() + theme(legend.position = "bottom") + guides(color = guide_legend(ncol = 1))

It still comes out as black and white. However, the scale modifications and other code work, wonderfully!
I have tried other ggplot language without avail, as well. Sorry for such a bother--really appreciate the help and great package!

[ddsjoberg]

The code from the gallery works for me:

library(ggsurvfit)
#> Loading required package: ggplot2
library(tidycmprsk)

cuminc(Surv(ttdeath, death_cr) ~ trt, trial) %>%
  ggcuminc(outcome = "death from cancer") +
  add_confidence_interval() +
  add_censor_mark() +
  add_risktable() +
  theme_bw() +
  scale_ggsurvfit(x = list(breaks = seq(0, 24, by = 6))) +
  scale_color_manual(values = c('#54738E', '#82AC7C')) +
  scale_fill_manual(values = c('#54738E', '#82AC7C'))

^{Created on 2022-12-28 with reprex v2.0.2}

[ddsjoberg]

If you're still having issues, run your code piece by piece to identify where the error is arising.

If you would like to share more code, please review the reprex package and make a proper reprex to copy and paste into the discussion. Also be sure to keep your code as minimal as possible, aka only include the most minimal code needed to illustrate an issue.

[saskia-c]

I have the same question on customizing lines in the ggcuminc() function.

The gallery code works to change the line colours if there are two groups (e.g. Drug A and Drug B). However, if there is only one group, but you would like to plot two (or more) outcomes, then it is not clear how to change the line colours.

Below examples using 'trial' dataset

library(tidycmprsk)

# Was expecting this to work, but it returns a black and white plot

cuminc(Surv(ttdeath, death_cr) ~ 1, trial) %>%
  ggcuminc(outcome = c("death from cancer", "death other causes")) +
  add_confidence_interval() +
  scale_color_manual(values = c('#54738E', '#82AC7C')) +
  scale_fill_manual(values = c('#54738E', '#82AC7C'))


# Alternatively thought this might work, but it returns same error as described by eloccodes

cuminc(Surv(ttdeath, death_cr) ~ 1, trial) %>%
  ggcuminc(outcome = c("death from cancer", "death other causes"),
                   color = c('#54738E', '#82AC7C')) +
  add_confidence_interval() 

Many thanks in advance!

[ddsjoberg]

Hmmm, I am not sure it's possible. The aes is constructed with aes(x = time, y = estimate, linetype = outcome). The colour aes is added if the model is stratified.

I am not sure if colors can be modified if there is no call to a color aes?

[saskia-c]

Oke, thank you for looking into it!