Pseudo-observations for bivariate survival data.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• The pseudo-observations approach is popular for estimating covariate effects in censored survival data.
• Existing methods primarily focus on univariate failure-time data.
• There is a need to extend these methods for bivariate failure-time data.

Purpose of the Study:

• To generalize the pseudo-observations approach for bivariate failure-time data subject to right censoring.
• To enable estimation of covariate effects on joint survival functions and related quantities.
• To provide a statistically sound method for analyzing complex survival data.

Main Methods:

• Estimating the joint survival function using nonparametric methods (Lin and Ying, Dabrowska).
• Defining bivariate pseudo-observations based on the estimated joint survival function.
• Utilizing generalized linear models with pseudo-observations as responses.

Main Results:

• The proposed bivariate pseudo-observations approach yields consistent and asymptotically normal regression estimates.
• The method effectively estimates covariate effects on joint survival probabilities at specific or multiple time points.
• Demonstrated capability to estimate covariate-adjusted conditional survival probabilities.

Conclusions:

• The generalized pseudo-observations approach is a valid and powerful tool for bivariate censored survival data analysis.
• This method expands the applicability of pseudo-observations to more complex survival scenarios.
• The approach is robust, as shown by simulations and real-world data analyses.