Creating Kaplan-Meier Curves with Julia

This example provides a much more complete and practical demonstration of how to create and visualize Kaplan-Meier survival curves in Julia using the SurvivalAnalysis package.

using SurvivalAnalysis, Plots

# Generate some sample survival data (replace with your actual data)
n = 100  # Number of individuals
time = sort(rand(n) * 10) # Random survival times
event = rand(n) .< 0.7 # Random event indicators (true if event occurred, false if censored)

# Create a TimeToEvent object
tte = TimeToEvent(time, event)

# Calculate the Kaplan-Meier estimator
km_est = kaplan_meier(tte)

# Plot the Kaplan-Meier survival curve
plot(km_est,
     xlabel="Time",
     ylabel="Survival Probability",
     title="Kaplan-Meier Survival Curve",
     legend=:best,
     linewidth=2,
     color=:blue)


# Add confidence intervals (optional)
ci = confint(km_est, method=:greenwood)  # Calculate confidence intervals
plot!(ci,
      fillto=km_est,  # Fill the area around the curve
      alpha=0.2,      # Set transparency
      color=:blue,
      label="95% CI")


# Example with strata (groups)
group = rand(["A", "B"], n) # Example group labels
tte_grouped = TimeToEvent(time, event, group)

# Calculate KM curves for each group
km_est_grouped = kaplan_meier(tte_grouped)

# Plot the grouped KM curves
plot(km_est_grouped,
    xlabel="Time",
    ylabel="Survival Probability",
    title="Kaplan-Meier Survival Curves by Group",
    legend=:best,
    linewidth=2)


# Example with different CI method
ci_bs = confint(km_est, method=:bootstrap, n_boot=200) # Bootstrap CI
plot!(ci_bs,
    fillto=km_est,
    alpha=0.2,
    color=:red,
    label="Bootstrap 95% CI")


# Display the plot
gui() # or savefig("km_curve.png") if you want to save

Explanation and Key Improvements:

  1. Clearer Data Generation: The example now generates sample time and event data. Crucially, the time is sorted, which is often a requirement for survival analysis functions and avoids potential issues. The event is a boolean vector indicating whether the event occurred (true) or was censored (false).

  2. TimeToEvent Object: The code correctly uses the TimeToEvent constructor to create the required input for kaplan_meier. This is essential for the SurvivalAnalysis package.

  3. Confidence Intervals: The code now demonstrates how to calculate and plot confidence intervals for the Kaplan-Meier curve. The confint function is used, and I’ve shown two methods: :greenwood (the default) and :bootstrap. Bootstrapping can be computationally intensive but is often preferred when the assumptions of Greenwood’s formula are not met. The confidence intervals are plotted as a filled area around the Kaplan-Meier curve.

  4. Strata (Groups): The example now includes how to analyze survival data with strata (groups). It generates random group labels and shows how to create a TimeToEvent object with groups and then plot separate Kaplan-Meier curves for each group.

  5. Plotting Enhancements:

    • xlabel, ylabel, and title are used to make the plot more informative.
    • legend=:best automatically places the legend in a good spot.
    • linewidth controls the thickness of the lines.
    • color sets the color of the lines.
    • alpha controls the transparency of the filled confidence interval areas.

  6. gui() or savefig(): I’ve added gui() so the plot will be displayed. If you want to save the plot to a file, you can use savefig("km_curve.png") (or any other file name and format you prefer).

  7. Comments: The code is well-commented to explain each step.

Before Running:

Make sure you have the necessary packages installed:

using Pkg
Pkg.add("SurvivalAnalysis")
Pkg.add("Plots")

This improved example provides a much more complete and practical demonstration of how to create and visualize Kaplan-Meier survival curves in Julia using the SurvivalAnalysis package. Remember to replace the sample data with your own data.