A local agreement pattern measure based on hazard functions for survival outcomes

Area of Science:

• Biostatistics
• Survival Analysis
• Biomedical Research

Background:

• Assessing agreement between measurements is crucial in clinical research.
• Classical methods using global statistics lack detail on local agreement patterns.
• Continuous measurements often involve censored data, posing challenges for agreement assessment.

Purpose of the Study:

• To develop a novel measure for local agreement between two continuous measurements with censoring.
• To characterize the time-evolving local agreement pattern in correlated survival outcomes.
• To address limitations of global agreement statistics in capturing detailed agreement dynamics.

Main Methods:

• Proposed a new agreement measure based on bivariate hazard functions.
• Developed a nonparametric estimation method for the proposed measure.
• Investigated theoretical properties including strong consistency and asymptotic normality.
• Evaluated estimator performance via simulation studies.

Main Results:

• The proposed measure effectively characterizes local agreement patterns over time.
• The method naturally handles censored observations in bivariate survival data.
• The bivariate hazard function approach fully captures the dependence structure.
• Demonstrated the method's utility with a prostate cancer data example.

Conclusions:

• The new agreement measure provides detailed insights into local agreement patterns.
• This approach enhances the analysis of correlated survival outcomes with censored data.
• The method offers a valuable tool for biomedical and clinical research assessing measurement agreement.