Understanding Competing Risks in Survival Analysis: A Simple Guide

Introduction

In survival analysis, we are often interested in a single event, such as death from a specific disease. However, in the real world, patients can experience other events that “compete” with the event of interest. For example, a patient with cancer could die from heart disease before they die from the cancer. This is what we call a competing risk.

When competing risks are present, standard survival analysis methods (like the Kaplan-Meier method) can be misleading. The jSurvival module provides a specific tool for analyzing data with competing risks.

Learning Objectives:

By the end of this guide, you will be able to:

Understand what a competing risk is and why it’s important.
Perform a competing risks analysis using the jSurvival module in jamovi.
Interpret a cumulative incidence plot.

Why Do We Need a Special Analysis for Competing Risks?

Let’s imagine we are studying deaths from breast cancer. If we only use the standard Kaplan-Meier method, we would treat deaths from other causes (like heart disease) as “censored.” This assumes that these patients are still at risk of dying from breast cancer, which is not true. This can lead to an overestimation of the survival probability from breast cancer.

Competing risks analysis solves this problem by modeling the probability of each event type separately.

Performing a Competing Risks Analysis in jamovi

Let’s walk through an example. We have a dataset of patients with prostate cancer, and we are interested in the time to death from prostate cancer. However, many of the patients in our study are elderly and may die from other causes.

Step 1: Your Data

For a competing risks analysis, your outcome variable needs to have at least three levels:

“No Event” (i.e., the patient is still alive).
The event of interest (e.g., “Death from Prostate Cancer”).
The competing event(s) (e.g., “Death from Other Causes”).

Step 2: Set up the Analysis

Go to Analyses > jSurvival > Competing Risks Survival Analysis.
Assign your Time Elapsed variable.
Assign your multi-level Outcome variable.
You will need to tell jamovi which level of your outcome variable corresponds to which event type.

Competing risks setup in jSurvival

Step 3: Interpret the Results

The main output of a competing risks analysis is the cumulative incidence function (CIF) plot.

Cumulative incidence (CIF) plot example

How to Read the CIF Plot:

The y-axis shows the cumulative incidence, which is the probability that an event has occurred by a certain time.
The x-axis shows the time.
You will see a separate curve for each event type.

Interpretation (example wording)

By 24 months, disease‑specific death reaches 18%, while death from other causes is 10%.
This indicates a meaningful burden from competing events; report both CIFs rather than a single KM estimate.

For example, the curve for “Death from Prostate Cancer” will show you the probability of a patient dying from prostate cancer by a certain time, taking into account that they could have died from other causes.

Key Differences from a Kaplan-Meier Plot:

In a Kaplan-Meier plot, the curve goes down, showing the probability of survival.
In a CIF plot, the curves go up, showing the probability of an event.

Conclusion

Competing risks are a common issue in clinical research, especially when studying elderly populations or diseases with multiple causes of death. The jSurvival module in jamovi provides a user-friendly way to perform a competing risks analysis and get a more accurate understanding of the probability of your event of interest.

Key Takeaways:

Use a competing risks analysis when patients can experience events that prevent the event of interest from occurring.
The cumulative incidence function (CIF) plot is the main tool for visualizing the results of a competing risks analysis.
A CIF plot shows the probability of an event occurring over time, accounting for competing risks.

Common Pitfalls

Treating competing events as censored in KM analyses overestimates disease‑specific survival.
Outcome labelling: Ensure outcome levels are clearly mapped (no event, event of interest, competing event[s]).
Interpretation: CIF curves go up (probability of event), unlike KM curves which go down (probability of survival).

ClinicoPath Development Team