A computationally simple bivariate survival estimator for efficacy and safety

Area of Science:

• Clinical trial methodology
• Biostatistics
• Survival analysis

Background:

• Treatment efficacy and safety are primary clinical trial endpoints.
• Efficacy often uses continuous time-to-event data, while safety uses discrete scales.
• Existing methods struggle with joint analysis of mixed-type, censored data.

Purpose of the Study:

• To develop a non-parametric estimator for bivariate survival functions with mixed continuous and discrete time-to-event data.
• To enable robust joint analysis of treatment efficacy and safety in clinical trials.
• To provide a computationally simple method for handling right-censored data.

Main Methods:

• Proposed a non-parametric estimator for the bivariate survival function.
• Assumed independent censoring times, potentially dependent on each other.
• Derived closed-form covariance estimators for the survivor function and its temporal evolution.

Main Results:

• The novel estimator effectively analyzes joint efficacy and safety data with mixed continuous and discrete outcomes.
• The derived covariance estimators facilitate statistical inference and sequential analysis.
• The method is computationally simple and applicable to right-censored data.

Conclusions:

• The proposed method offers a significant advancement in the joint analysis of efficacy and safety endpoints in clinical trials.
• This approach enhances the statistical rigor for evaluating complex clinical trial data, particularly in oncology.
• The derived estimators support flexible inference and sequential monitoring in clinical studies.