Marginal regression of multivariate event times based on linear transformation models

Area of Science:

• Biostatistics
• Medical Statistics
• Survival Analysis

Background:

• Multivariate event time data are prevalent in medical research, encompassing multiple event types, recurrent events, or clustered event times.
• Existing methods often require specifying dependence structures, which can be challenging for complex event data.

Purpose of the Study:

• To develop a robust statistical method for analyzing multivariate and recurrent event time data.
• To simultaneously estimate regression parameters and transformation functions without prespecifying dependence structures.

Main Methods:

• Utilized semiparametric linear transformation models for marginal distributions.
• Developed novel estimating equations for simultaneous parameter and transformation function estimation.
• Employed the plug-in method for consistent estimation of the variance-covariance matrix.

Main Results:

• Regression estimators were shown to be asymptotically normal.
• The variance-covariance matrix has a closed form and is consistently estimable.
• Simulation studies confirmed the practical applicability and appropriateness of the proposed method.

Conclusions:

• The proposed method offers a flexible and effective approach for analyzing complex event time data in medical studies.
• The methodology is illustrated with a real-world application to bladder cancer tumor recurrence data.