In GSK-Biostatistics/tlang: Applies Display Metadata to Analysis Results Datasets

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(tfrmt)
library(dplyr)
library(tidyr)
library(patchwork)
library(survival)
library(ggfortify)

Print to ggplot outputs a ggplot table. This table can then be added to an existing plot to provide additional information. For this we recommend the use of the patchwork package.

Survival analysis example

A common use case of this function could be in the production of survival plots. The following code sets up a Kaplan Meier plot using the colon data from the survival package.

# Set up survival data
fit <- survfit(Surv(time,status)~rx, data=colon)

# Plot kaplan meier between times 0-3000
km_plot <- autoplot(fit) +
  xlim(c(0,3000))

As with print_to_gt, print_to_ggplot requires an input table with label, value,param and column variables. The following code sets up our mock input table.

risk <- tibble(time=c(rep(c(0,1000,2000,3000),3)),
               label=c(rep("Obs",4),rep("Lev",4),rep("Lev+5FU",4)),
               value=c(630,372,256,11,620,360,266,8,608,425,328,14),
               param=rep("n",12))

A tfrmt object is required to specify the formatting of the ggplot table. This can then be piped out to print_to_ggplot as seen below.

table <- tfrmt(
  # specify columns in the data
  label = label ,
  column = time,
  param = param,
  value = value,
  body_plan = body_plan(
    frmt_structure(group_val = ".default", label_val = ".default",
                   frmt("X"))
  )) %>%
  print_to_ggplot(risk)

table

Now using the patchwork package we can combine the original plot, and our ggplot table. Since we want the table below the plot, we use /. For more information on using patchwork, refer to the documentation here.

km_plot/table

Because we don't have to duplicate the time points, we can just remove the x-axis labels using the theme

table2 <- table +
  theme(axis.text.x = NULL)

km_plot/table2

Using grouping

You can also apply groups to your ggplot table. The code below adds groupings to the risk table above for a mock example.

riska <- risk %>%
  mutate(group="A")

riskb <- risk %>%
  mutate(group="B",
         value=value+10)

risk_group <- riska %>%
  rbind(riskb)

Now we need to add group to our tfrmt specification and patch together:

group_table <- tfrmt(
  # specify columns in the data
  group = group,
  label = label ,
  column = time,
  param = param,
  value = value,
   body_plan = body_plan(
    frmt_structure(group_val = ".default", label_val = ".default",
                   frmt("X"))
  )) %>% 
  print_to_ggplot(risk_group) +
  theme(axis.text.x = NULL)

km_plot/group_table

Forest Plots

The same logic used in the survival plots can also be used to create forest plots, but instead of stacking the plots we will just put them side by side. First we can make the table using print_to_ggplot.

aes <- factor(c("Fever", "Malaise", "Local Allergic Reaction"),
              levels = c("Fever", "Malaise", "Local Allergic Reaction"))

tbl_dat <- tibble(
  grp = "1 to 7 days after treatment",
  ae = rep(aes, each = 4),
  trt = rep(rep(c("TRT A", "Placebo"), each = 2), 3),
  param = rep(c("n", "pct"), 6),
  value = c(60, 7.7, 26, 3.3, 183, 23.5, 89, 11.4, 94, 12, 55, 7)
)

tbl_p <- tfrmt_n_pct() %>% 
  tfrmt(label = ae,
        group = grp,
        column = trt,
        param = param, 
        value = "value") %>% 
  print_to_ggplot(tbl_dat)

tbl_p

Next we need to plot the rate ratios. This plot requires a bit more refining because we need to remove the y-axis label so it can be combined with the table plot. We also need to add a row in the data for the group value to make the plots match up correctly.

plot_dat <- tibble(
  ae = aes,
  mean = c(3, 2.3, 2),
  lower = c(1.95, 1.9, 1.2),
  upper = c(4, 3.5, 3.4)
  ) %>% 
  bind_rows(c(ae = "1 to 7 days after treatment"))

plot_p <- ggplot(data=plot_dat, aes(x=ae, y=mean, ymin=lower, ymax=upper)) +
  geom_pointrange() + 
  geom_hline(yintercept=1, lty=2) +  # add a dotted line at x=1 after flip
  coord_flip() +
  xlab("") +
  ylab("Adjusted Rate Ratio (95% CI)")+
  scale_x_discrete(limits=rev)+
  scale_y_discrete(position = "right") + 
  theme_minimal() +
  theme(axis.text.y=element_blank(), # remove x axis labels
        axis.ticks.y=element_blank())
plot_p

Now thanks to patchwork combining the two plots is relatively easy.

tbl_p + plot_p

Table Styling

A final note about the ggplot tables - most styling can be adjusted with theme once the table is create. But, the table body needs to be adjusted in the print_to_ggplot call, by supplying the requirements to the .... So if you need to change the size table boy you just add that to the print_to_ggplot.

tfrmt_n_pct() %>% 
  tfrmt(label = ae,
        group = grp,
        column = trt,
        param = param, 
        value = "value") %>% 
  print_to_ggplot(tbl_dat, size = 8)