Flexible parametric model for survival data subject to dependent censoring.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Standard survival analysis often assumes conditional independence between survival time (T) and censoring time (C), which is frequently unrealistic.
• Existing correction methods for dependent survival and censoring times often require prior knowledge or auxiliary data, limiting their applicability.
• This limitation hinders accurate analysis in many real-world scenarios, such as medical studies.

Purpose of the Study:

• To develop a flexible parametric model that can handle the association between survival time (T) and censoring time (C) without prior assumptions.
• To demonstrate the identifiability of the T-C association within the proposed model.
• To provide a robust method for survival data analysis when the independence assumption is violated.

Main Methods:

• Development of a flexible parametric model based on transformed linear models.
• Theoretical investigation of the identifiability of the association between survival and censoring times.
• Performance evaluation through asymptotic analysis and finite sample simulations.
• Introduction of a formal goodness-of-fit test for model assessment.

Main Results:

• The proposed flexible parametric model successfully identifies the association between survival time (T) and censoring time (C).
• The model's performance is validated through both theoretical analysis and simulation studies.
• A goodness-of-fit test is developed to ensure the reliability of the fitted models.

Conclusions:

• The developed flexible parametric model offers a viable solution for survival data with dependent censoring.
• The method overcomes limitations of existing approaches by not requiring prior knowledge or auxiliary data.
• The approach was successfully applied to liver transplant study data, demonstrating its practical utility.