Excess cumulative incidence estimation for matched cohort survival studies

Area of Science:

• Epidemiology
• Biostatistics
• Survival Analysis

Background:

• Estimating excess cumulative incidence function (CIF) in competing risk settings with matched data is challenging.
• Existing methods often require estimating separate CIFs for exposed and unexposed groups, complicating analysis.
• Handling multiple time scales (e.g., age, time since exposure) and left truncation adds complexity.

Purpose of the Study:

• To propose a novel regression approach for estimating excess cumulative incidence function (CIF) using matched data.
• To define and estimate excess risk as the difference between exposed and unexposed CIFs in a competing risk framework.
• To develop a method that naturally incorporates matched data structures and multiple time scales.

Main Methods:

• Utilizing an extended binomial regression model that leverages the matched data structure.
• Defining excess risk as CIF_exposed - CIF_unexposed (background CIF).
• The model inherently handles age and time since exposure, and left truncation on age.

Main Results:

• The proposed regression approach effectively estimates excess risk without needing separate CIF estimations.
• Simulations demonstrate the model's practical utility and accuracy.
• The method allows for straightforward prediction of individual excess risk scenarios and covariate effect interpretation.

Conclusions:

• The developed regression model provides an efficient and interpretable method for estimating excess cumulative incidence in matched, competing risk data.
• This approach simplifies complex survival analyses, particularly in fields like cancer survivorship research.
• The model was successfully applied to investigate late event risks in childhood cancer survivors from the ALiCCS study.