The ggsurvfit package eases the creation of time-to-event (aka survival) summary figures with ggplot2. The concise and modular code creates images that are ready for publication or sharing. Competing risks cumulative incidence is also supported via
Use ggplot2 functions: Each ggsurvfit add-on function (e.g.
add_risktable(), etc.) is written as a proper ggplot2 ‘geom’, meaning the package functions can be woven with ggplot2 functions seamlessly. You don’t need to learn how to style the plot within the ggsurvfit functions: rather, rely on the suite of ggplot2 functions you already know.
Publishable Legends: Raw variable names do not appear in the figure legend, e.g.
Limitless Customization: You can modify the x-axis scales or any other plot feature and the risk table will still align with the plot.
Install ggsurvfit from CRAN with:
You can install the development version from GitHub with:
# install.packages("devtools") devtools::install_github("ddsjoberg/ggsurvfit")
Review the figure gallery for many more examples.
library(ggsurvfit) library(ggplot2) survfit2(Surv(time, status) ~ surg, data = df_colon) |> # build Kaplan-Meier plot ---------------------------------------------------- ggsurvfit(size = 1) + add_confidence_interval() + add_risktable() + add_quantile(y_value = 0.6, color = "gray50", size = 0.75) + # use ggplot2 functions to style the plot and update the labels -------------- # limit plot to show 8 years and less coord_cartesian(xlim = c(0, 8)) + # update figure labels/titles labs( y = "Percentage Survival", title = "Recurrence by Time From Surgery to Randomization", ) + # reduce padding on edges of figure and format axes scale_y_continuous(label = scales::percent, breaks = seq(0, 1, by = 0.2), expand = c(0.015, 0)) + scale_x_continuous(breaks = 0:10, expand = c(0.02, 0))
- We can reliably remove the raw variable names from the figure legend, e.g.
- P-values can be calculated with
survfit_p()and added to figures.
- The items above are often possible using
survfit(). However, by utilizing the calling environment we are assured the correct elements are found, rather than crossing our fingers that the search path contains the needed elements.
The package also includes a gem for those using the CDISC ADaM ADTTE data model. The event indicator in ADTTE data sets is named
"CNSR" and is coded in the opposite way the survival package expects outcomes—
1 = 'censored' and
0 = 'event'. This difference creates an opportunity for errors to be introduced in an analysis. The package exports a function called
Surv_CNSR() to resolve this concern.
The function creates a survival object (e.g.
survival::Surv()) that uses CDISC ADaM ADTTE coding conventions and converts the arguments to the status/event variable convention used in the survival package for both the event indicator and the time component—
"AVAL". The function can be used in ggsurvfit as well as any other package that uses