Predicting events in clinical trials using two time-to-event outcomes

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Survival Analysis

Background:

• Predicting event completion is crucial for clinical trial planning and interim analyses.
• Estimating survival functions typically relies on single time-to-event outcomes.
• Leveraging correlated endpoints can potentially enhance estimation efficiency.

Purpose of the Study:

• To propose a novel method for estimating survival functions using the convolution of two time-to-event variables.
• To develop an improved estimation equation for expected time by exploiting the relationship between two correlated endpoints.
• To demonstrate the predictive improvement offered by bivariate analysis over univariate methods in clinical trials.

Main Methods:

• Utilizing the convolution of two time-to-event variables for survival function estimation.
• Developing and applying a new estimation equation for expected time based on bivariate endpoint relationships.
• Employing exponential models with potential change points for analysis.
• Conducting simulations and analyzing real clinical trial data.

Main Results:

• The proposed convolution method effectively estimates the survival function of interest.
• The new estimation equation provides an improved prediction of expected time.
• Bivariate analysis using two correlated time-to-event endpoints demonstrated significant prediction improvement over univariate methods.
• Simulations and real-world data confirmed the efficacy of the proposed bivariate approach.

Conclusions:

• The integration of two correlated time-to-event endpoints offers a significant advantage in survival function prediction for clinical trials.
• The proposed convolution and bivariate estimation methods provide more efficient and accurate predictions.
• This approach enhances the ability to predict when a prespecified number of events will be reached, aiding trial management.