Transformation model based regression with dependently truncated and independently censored data

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Truncated survival data present challenges for conventional methods like Kaplan-Meier and Cox models.
• These conventional methods often assume quasi-independence between truncation and event times, which may not hold in real-world scenarios.

Purpose of the Study:

• To extend the structural transformation model for estimating survival functions under dependent truncation to a regression setting.
• To introduce novel, flexible, and computationally efficient methods for analyzing dependent truncated survival data.
• To provide an accessible R package (tranSurv) for implementing these advanced statistical techniques.

Main Methods:

• Extension of the structural transformation model to incorporate regression.
• Development of three new regression methods based on the transformation model.
• Introduction of a piecewise transformation model for enhanced flexibility.
• Validation through simulation studies and application to Alzheimer's disease cognitive decline data.

Main Results:

• The proposed transformation model-based methods demonstrate effective analysis of survival data with dependent truncation.
• Simulation studies confirm the performance of the developed statistical models.
• The methods were successfully applied to real-world data from the National Alzheimer's Coordinating Center.

Conclusions:

• The extended transformation model provides a robust framework for survival analysis with dependent truncation.
• The developed methods and the tranSurv R package offer valuable tools for researchers dealing with complex survival data.
• This approach enhances the ability to study event time distributions and regression coefficients in the presence of truncation.