Fitting competing risks with an assumed copula

Area of Science:

• Biostatistics
• Survival Analysis
• Medical Statistics

Background:

• Competing risks data analysis is crucial in many fields, including medicine.
• Traditional methods often assume independence between failure types, which may not hold true.
• Understanding dependence structures is key for accurate risk assessment.

Purpose of the Study:

• To propose a fully parametric model for analyzing competing risks data with potentially dependent failure types.
• To model the dependence between cause-specific survival times using copula functions.
• To assess the impact of concomitant variables on marginal survival and the relationship between causes of death.

Main Methods:

• Development of a fully parametric competing risks model incorporating copula functions.
• Ensuring identifiability and flexibility in choosing marginal survival functions.
• Adjustment for concomitant variables and a dependence parameter.

Main Results:

• The proposed copula model effectively captures dependence structures between competing risks.
• The model allows for adjustment of concomitant variables and assessment of their effects.
• Application to prostate cancer data yielded more accurate inferences compared to independent models.

Conclusions:

• The copula-based parametric model offers a flexible and accurate approach for analyzing dependent competing risks.
• This method improves upon simpler models by explicitly addressing inter-risk dependencies.
• The findings have implications for risk prediction and understanding disease progression in complex scenarios.