Competing risks models with two time scales

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Competing risks models often use a single time scale, limiting their application in complex scenarios like cancer mortality.
• Jointly considering multiple time scales (e.g., age and time since diagnosis) is crucial for accurately assessing cause-specific hazards.
• Existing methods for multiple time scales in competing risks are limited, necessitating novel approaches.

Purpose of the Study:

• To propose and implement a flexible statistical model for competing risks analysis incorporating two time scales.
• To estimate cause-specific hazards that vary smoothly over two dimensions using penalized splines.
• To address challenges with coarsely grouped data in real-world datasets like the SEER program.

Main Methods:

• Developed a novel competing risks model utilizing two-dimensional P-splines for hazard smoothing.
• Leveraged the equivalence between hazard smoothing and Poisson regression for estimation.
• Employed generalized linear array models for computational efficiency and a penalized composite link model for data ungrouping.
• Implemented the model in the R-package TwoTimeScales.

Main Results:

• The proposed model effectively estimates cause-specific hazards across two time scales.
• The method successfully handles coarsely grouped data, demonstrated using SEER breast cancer mortality data.
• The R-package TwoTimeScales provides a practical tool for applying this advanced statistical methodology.

Conclusions:

• The novel two-time scale competing risks model offers a significant advancement for analyzing complex survival data.
• This approach enhances the understanding of mortality patterns in diseases like breast cancer by considering age and time since diagnosis.
• The developed methodology and software facilitate more accurate risk assessment and epidemiological research.