What is a Competing Event?

In survival analysis, a competing event is an event that prevents the occurrence of the primary event of interest. For example, if you're studying time to cancer relapse, death from another cause would be a competing event, as it precludes the possibility of relapse.

Standard Kaplan-Meier curves assume that no other event can interfere with the event of interest. This can lead to overestimated event probabilities in real-world scenarios. Using competing risks provides a more realistic view of patient outcomes.

Selecting competing events allows to use a cumulative incidence function (CIF), which provides more accurate probability estimates when multiple types of events can occur.

How to Use This Option?

You can enable competing risks in just a few steps:

1. Run a survival analysis on a primary event variable (e.g., "Cancer relapse").
2. Open the “Analysis options and censoring” panel in the configuration area.
3. Select one or more event-type variables under the “Competing events” section.
4. The analysis will automatically refresh and display the new results using the competing risks method.

Important Notes

This option is only available for:

• Univariate survival analyses
• Bivariate survival analyses (with a grouping variable)

It is not compatible with multivariate models such as the Cox proportional hazards model (for now).

Methodology

When one or more competing events are defined:

• The analysis uses the Aalen-Johansen method to estimate cumulative incidence.
• If a comparing variable is selected, Gray’s test is used to compare incidence curves across groups.

Why the Graphs Look Different?

When competing events are selected:

• The analysis uses the Aalen-Johansen estimator instead of the Kaplan-Meier method.
• The resulting graph shows Cumulative Incidence Functions (CIFs), which represent the actual probability of the event occurring over time, accounting for the risk of other competing events.
• This contrasts with Kaplan-Meier curves, which assume that censored cases could still experience the event later, even if they've had a competing event, which leads to overestimated probabilities.

In CIF plots:

• The curve typically rises more slowly than in KM plots.
• The y-axis represents the cumulative probability of experiencing the event of interest in the presence of competing events.