multinma (version 0.8.0)

geom_km: Kaplan-Meier curves of survival data

Description

This helper function constructs a ggplot2 geom to plot Kaplan-Meier curves from a network containing survival or time-to-event outcomes. This is useful for overlaying the "raw" survival data on the estimated survival functions created with plotted with plot.surv_nma_summary(), but can also be used standalone to plot Kaplan-Meier curves before fitting a model.

Usage

geom_km(
  network,
  ...,
  transform = c("identity", "cloglog", "log", "cumhaz"),
  curve_args = list(),
  cens_args = list()
)

Value

A ggplot2 geom list that can be added to a ggplot2 plot object

Arguments

network

A nma_data network object containing survival outcomes

...

Additional arguments passed to survival::survfit()

transform

Character string giving the transformation to apply to the KM curves before plotting. The default is "identity" for no transformation; other options are "cloglog" for \(\log(-\log(S))\), "log" for \(\log(S)\), or "cumhaz" for the cumulative hazard \(-\log(S)\).

curve_args

Optional list of arguments to customise the curves plotted with ggplot2::geom_step()

cens_args

Optional list of arguments to customise the censoring marks plotted with ggplot2::geom_point()

Examples

# Set up newly-diagnosed multiple myeloma network

head(ndmm_ipd)
head(ndmm_agd)

ndmm_net <- combine_network(
  set_ipd(ndmm_ipd,
          study, trt,
          Surv = Surv(eventtime / 12, status)),
  set_agd_surv(ndmm_agd,
               study, trt,
               Surv = Surv(eventtime / 12, status),
               covariates = ndmm_agd_covs))
# Plot KM curves using ggplot2
library(ggplot2)

# We need to create an empty ggplot object to add the curves to
ggplot() + geom_km(ndmm_net)

# Adding plotting options, facets, axis labels, and a plot theme
ggplot() +
  geom_km(ndmm_net,
          curve_args = list(linewidth = 0.5),
          cens_args = list(size = 3, shape = 124)) +
  facet_wrap(vars(Study)) +
  labs(xlab = "Time", ylab = "Survival Probability") +
  theme_multinma()

# Using the transform argument to produce log-log plots (e.g. to assess the
# proportional hazards assumption)
ggplot() +
  geom_km(ndmm_net, transform = "cloglog") +
  facet_wrap(vars(Study)) +
  theme_multinma()

# Using the transform argument to produce cumulative hazard plots
ggplot() +
  geom_km(ndmm_net, transform = "cumhaz") +
  facet_wrap(vars(Study)) +
  theme_multinma()

# This function can also be used to add KM data to plots of estimated survival
# curves from a fitted model, in a similar manner
# \donttest{
# Run newly-diagnosed multiple myeloma example if not already available
if (!exists("ndmm_fit")) example("example_ndmm", run.donttest = TRUE)
# }
# Plot estimated survival curves, and overlay the KM data
# \donttest{
plot(predict(ndmm_fit, type = "survival")) + geom_km(ndmm_net)
# }