Is there a way to use ggplot2 to plot survival curves from a svycoxph object?

I used svycoxph from the survey package to create an object:

object <- svycoxph(Surv(time, status) ~ group, design=design)

And then I used predict() as documented here on page 88 to predict curves:

prediction <- predict(object, type="curve", newdata = data.frame(group=c("A", "B")) )

Is there a way to use ggplot2 functionality to plot the survival curves from the prediction? I tried using the base plotting features as documented on the same page 88, but I can't tell which line is A and which line is B, and I can't make it match plot templates I have that were made with ggsurvfit().

One other thing I tried was extracting the data from the prediction object and using ggplot(), but I can't tell if I labeled the A and B lines correctly:

prediction.data <- data.frame(time=prediction[["1"]][["time"]], 
                              A_surv=prediction[["1"]][["surv"]], 
                              B_surv=prediction[["2"]][["surv"]])

ggplot(prediction.data, aes(x=time)) + 
  geom_line(aes(y=A_surv), color="red") + 
  geom_line(aes(y=B_surv), color="blue")

You're on the right track. Essentially you need to cajole the predictions into a data frame and use this for the plot.

Here's a full reprex. First we will load the package, create some reproducible sample data with the same names as in your example, and specify a design:

library(survey)

set.seed(1)

df <- data.frame(time = round(c(rexp(50, 0.01), rexp(50, 0.015))),
                 group  = rep(c("A", "B"), each = 50))
df$status <- ifelse(df$time > 60, 0, 1)
df$time[df$time > 60] <- 60

design <- svydesign(id = ~ group, weights = 1, data = df)

Now we create our model

object <- svycoxph(Surv(time, status) ~ group, data = df, design = design)

Now let's get prediction curves for our two levels of group

preds <- predict(object, newdata = data.frame(group = c("A", "B")), 
                 type = "curve", se = TRUE)

The trickiest part is extracting the curves into a data frame

plot_df <- do.call(rbind, 
        Map(function(x, y) cbind(as.data.frame(unclass(x)), group = y),
            preds, c("A", "B")))

plot_df$lower <- exp(log(plot_df$surv) - 1.96 * sqrt(plot_df$varlog))
plot_df$upper <- exp(log(plot_df$surv) + 1.96 * sqrt(plot_df$varlog))

Now we are ready to plot:

library(ggplot2)

ggplot(plot_df, aes(time, surv, color = group)) +
  geom_step(linewidth = 1) +
  geom_step(aes(y = lower), linetype = 2, alpha = 0.5) +
  geom_step(aes(y = upper), linetype = 2, alpha = 0.5) +
  scale_color_manual(values = c("orangered", "deepskyblue4")) +
  theme_minimal(base_size = 16) +
  scale_y_continuous("Survival", limits = 0:1, labels = scales::percent) 

^{Created on 2023-06-15 with reprex v2.0.2}