Competing risks predictions with different time scales under the additive risk model.

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Traditional competing risks analysis often uses a single time scale for all event types.
• The proportional hazards model, commonly used, may not always be appropriate due to its proportionality assumption.
• Different event types may have distinct underlying time scales.

Purpose of the Study:

• To propose and evaluate predictions of cumulative incidence functions using cause-specific additive risk models with distinct time scales for different event types.
• To provide a flexible alternative to the proportional hazards model in competing risks analysis.
• To establish theoretical properties and practical utility of the proposed methods.

Main Methods:

• Developed cause-specific additive risk models allowing for different time scales per event type.
• Derived predictions for cumulative incidence functions.
• Established consistency and asymptotic normality of predictions using empirical processes.
• Developed consistent variance estimators.

Main Results:

• The proposed prediction methods for cumulative incidence functions demonstrate good performance in simulation studies.
• Theoretical properties (consistency, asymptotic normality) of the predictions were established.
• The methods were successfully illustrated using real-world data.

Conclusions:

• The cause-specific additive risk model with differing time scales offers a valuable approach for competing risks data analysis.
• The proposed prediction methods are statistically sound and practically applicable.
• This methodology enhances the analysis of complex event data, as shown with breast cancer data from the SEER program.