Feature request: additional arguments for `theme_risktable_default()`

[jpnolte]

First of all, thank you so much for the amazing work. Your package is so much better than the survminer package I used before. I've done a few projects now with your package and some details in adapting the risk table would be a great addition.

Lets take this a small reprex:

p <- survfit2(Surv(time, status) ~ sex, data = df_lung) |>
  ggsurvfit() +
  add_risktable(risktable_stats = "n.risk") + 
  scale_ggsurvfit()

(1) When dealing with larger patient numbers (e.g., everything above 1000) it will usually lead to the numbers at timepoint 0 overlap with the axis text on the y axis of the risk table. In the add_risktable() documentation you can see it in example 3. So my solution up until now is to either increase the expand argument in scale_x_continouus() (e.g., scale_x_continuous(expand = c(0.015, 0))) or play with the width argument when saving with ggsave() until the text does not overlap anymore.

But what if you would include an argument called margin within theme_risktable_default() which would work like the following but for the risktable and not the plot.

p + theme(axis.text.y = element_text(margin = margin(r = 10)))

(2) Left align the axis text of the y-axis of the risk table. This is more of a subjective option, but if you are already thinking about creating the first (important) one, it might be fast to also integrate this as an additional argument (hjust or align) to theme_risktable_default(). Here is how to do it for the plot. Would be nice to have it for the risk table as well.

p + theme(axis.text.y = element_text(hjust = 0))

(3) Positioning the stats_label above the axis text of the y-axis in the risk table instead of on top of the table itself. I didn't have time yet to look up your way of defining the stats_label (e.g., At risk), but it looks like you use ggtitle(). If so, this might also be quite easy to integrate and useful since some other programs seem to support it.

p +
  ggtitle("Testing this") + 
  theme(
    plot.caption = element_text(hjust = 0),
    plot.title.position = "plot"
  )

(4) Right align text (the numbers) from geom_text() in the risk table. I know this is only a matter of appearence, but it gets way easier to read and compare larger numbers with smaller ones. This one I did not figure out yet, since providing geom_text() through add_risktable() with the hjust argument for some reason puts a small offset there additionally.

survfit2(Surv(time, status) ~ sex, data = df_lung) |>
  ggsurvfit() +
  add_risktable(risktable_stats = "n.risk", hjust = "right") 

I would really appreaciate if you could add the first suggestion, since this is rather annoying. Also it would allow me to set expand to c(0, 0) so the axes would start at the same spot. The other suggestions would be the icing on the cake.

[ddsjoberg]

Thanks for the post @jpnolte !

Find responses below. Please note that the defaults were set to meet many needs. But we know that it does not meet everyone's need, which is why we try to make it possible modify the defaults.

1. You're calling the scale_ggsurvfit() function which reduces the padding/spacing. If you don't make this call, or you modify the defaults padding values, you can make any display fit.
2. Can you pass theme = list(theme_risktable_default(), theme(axis.text.y = element_text(hjust = 0))) to get what you want?
3. I think these options can also be passed to the theme argument, right?
4. Any arguments passed in add_risktable(...) are directly passed to geom_text(...). Will that work for you?

[jpnolte]

Thank you so much for the quick response. You are absolutely right, (2) and (3) can be achieved by providing the arguments through theme() in the list, nice! So this way you can change the position of the risk table title and also the alignment of the axis text. Maybe adding an example with the theme() in the list to the documentation is also helpful to others?

survfit2(Surv(time, status) ~ sex, data = df_lung) |>
  ggsurvfit() +
  scale_x_continuous(expand = c(0.005, 0), 
                     breaks = seq(0, 34, 2),
                     limits = c(0, 34)) + 
  scale_y_continuous(expand = c(0, 0.05), 
                     breaks = seq(0, 1, 0.25),
                     labels = scales::label_percent()) + 
  add_risktable(
    risktable_stats = "n.risk",
    theme = list(
      theme_risktable_default(), 
      theme(
        axis.text.y = element_text(hjust = 0, margin = margin(r = 10)),
        plot.caption = element_text(hjust = 0),
        plot.title.position = "plot"
      )
    )
  ) 

ad 1) Yes, this is exactly my point. I think, and I know others that agree, it looks better the less empty space there is between the actual start of the y and x axis and the first break. In the example above I reduced the empty space causing the overlap. I was thinking that the margin argument can change that, but of course the risk table ends at the same position as the plot, which is the actual problem of the numbers disappearing. I think, this might be a limitation of ggplot? It is not a big problem when you only have 138 patients, since the empty space I have to add to avoid overlapping is quite small. But the bigger the numbers, the more empty space is necessary.

ad 4) I already tried to provide the hjust argument directly to geom_text(), but apparently ggplot will not center and right align them, but instead move the numbers slightly to the left. Than it is of course slightly off in comparison to the axis text of the x axis of the plot.

survfit2(Surv(time, status) ~ sex, data = df_lung) |>
  ggsurvfit() +
  scale_x_continuous(expand = c(0.04, 0), 
                     breaks = seq(0, 34, 2),
                     limits = c(0, 34)) + 
  scale_y_continuous(expand = c(0, 0.05), 
                     breaks = seq(0, 1, 0.25),
                     labels = scales::label_percent()) + 
  add_risktable(
    risktable_stats = "n.risk",
    theme = list(
      theme_risktable_default(), 
      theme(
        axis.text.y = element_text(hjust = 0, margin = margin(r = 5)),
        plot.caption = element_text(hjust = 0),
        plot.title.position = "plot"
      )
    ),
    hjust = 1
  ) 

If you would then also right align the axis text on the x axis you can see how ggplot handles the alignment, moving both away from the ticks.

survfit2(Surv(time, status) ~ sex, data = df_lung) |>
  ggsurvfit() +
  scale_x_continuous(expand = c(0.04, 0), 
                     breaks = seq(0, 34, 2),
                     limits = c(0, 34)) + 
  scale_y_continuous(expand = c(0, 0.05), 
                     breaks = seq(0, 1, 0.25),
                     labels = scales::label_percent()) + 
  add_risktable(
    risktable_stats = "n.risk",
    theme = list(
      theme_risktable_default(), 
      theme(
        axis.text.y = element_text(hjust = 0, margin = margin(r = 5)),
        plot.caption = element_text(hjust = 0),
        plot.title.position = "plot"
      )
    ),
    hjust = 1
  ) + 
  theme(axis.text.x = element_text(hjust = 1))

So I think this is also a limitation of ggplot?