A nonparametric approach to a survival study with surrogate endpoints

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Estimating joint distributions of survival and early surrogate responses is crucial in clinical research.
• Existing methods may not fully capture the complexities of time-to-event data with early markers.
• The Kaplan Meier estimator is a standard for univariate survival analysis.

Purpose of the Study:

• To develop a novel nonparametric estimator for the joint distribution of survival time and surrogate response time.
• To demonstrate the estimator's reduction to the established Kaplan Meier estimator when surrogate data is absent.
• To explore the extension of this methodology to multiple state processes.

Main Methods:

• A nonparametric estimation approach is presented for the joint distribution.
• The method utilizes an exchangeable process, specifically reinforced random walks, to model individual observations.
• The estimator is derived in a novel manner, distinct from existing techniques.

Main Results:

• A nonparametric estimator for the joint distribution of survival time and surrogate response time is successfully derived.
• The estimator is shown to converge to the Kaplan Meier estimator in the absence of surrogate response variables.
• The methodology demonstrates adaptability for modeling multiple state processes.

Conclusions:

• The proposed nonparametric estimator provides a robust method for analyzing joint survival and surrogate response data.
• The novel use of reinforced random walks offers a new perspective in survival analysis modeling.
• The framework's extensibility to multiple states suggests broad applicability in various research settings.