Regression analysis of a graphical proportional hazards model for informatively left-truncated current status data

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Real-world survival data often present challenges like censoring, truncation, and complex covariate network structures.
• Existing methods may not adequately address the interplay between censoring, failure times, and intricate covariate relationships.

Purpose of the Study:

• To develop a robust variable selection method for left-truncated current status data with complex covariate networks.
• To model the dependence between censoring and failure times using a copula-based approach.
• To enhance the flexibility of the Proportional Hazards (PH) model for covariate correlation structures.

Main Methods:

• Employed a copula-based framework to link censoring and failure time.
• Integrated Markov Random Fields (MRF) with the Proportional Hazards (PH) model to capture covariate network structures.
• Utilized a penalized optimization method and spline functions for parameter estimation and baseline hazard function estimation.

Main Results:

• The proposed model demonstrated robustness in handling complex disease data.
• Numerical simulations and clinical trial data case studies confirmed the model's effectiveness and performance.
• The parameter inference strategy showed high precision and reliability.

Conclusions:

• The copula-MRF-PH model offers a powerful tool for variable selection in complex survival data.
• The method accurately handles left-truncated current status data and covariate dependencies.
• Validated effectiveness and reliability for real-world applications, particularly in clinical research.