A Semi-parametric Transformation Frailty Model for Semi-competing Risks Survival Data.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Semi-competing risks data analysis commonly uses multi-state models with shared frailty terms.
• Parametric frailty distributions (e.g., Gamma) are often assumed for identifiability, but misspecification can lead to bias.
• Existing methods struggle with consistent estimation when frailty distributions are misspecified.

Purpose of the Study:

• To propose a novel class of transformation models for semi-competing risks analysis.
• To enable non-parametric specification of the frailty distribution, overcoming limitations of parametric assumptions.
• To develop robust estimation procedures for multivariate time-to-event data.

Main Methods:

• Introduced a transformation model class allowing non-parametric frailty distribution specification.
• Ensured identifiability through parametric transformation and flexible error distribution specifications.
• Derived semi-parametric efficient scores and proposed a non-parametric score imputation for right censoring.

Main Results:

• Established consistency and asymptotic normality of the proposed estimators.
• Evaluated small-sample operating characteristics through simulation studies.
• Demonstrated applicability to real-world data, such as hospital readmission in pancreatic cancer patients.

Conclusions:

• The proposed semi-parametric transformation model offers a flexible and robust approach to semi-competing risks analysis.
• The non-parametric score imputation method effectively handles right censoring in multivariate time-to-event data.
• This methodology is broadly applicable to correlated time-to-event outcomes beyond semi-competing risks.