Specification of Estimands for Complex Disease Processes Using Multistate Models and Utility Functions.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Complex Disease Modeling

Background:

• Complex diseases involve multiple semi-competing events and recurrent episodes, complicating clinical trial endpoint definition.
• Traditional composite endpoints may not fully capture the nuances of complex disease progression.
• Recent innovations like the win ratio offer ranking-based approaches to disease course analysis.

Purpose of the Study:

• To demonstrate the utility of multistate models for analyzing complex disease courses in clinical trials.
• To emphasize the interpretability and simplicity of using utilities to synthesize treatment effect evidence.
• To provide a robust statistical framework for primary analyses in clinical trials involving complex disease outcomes.

Main Methods:

• Application of multistate models to represent complex disease pathways.
• Development and use of utility functions to synthesize treatment effects across different disease aspects.
• Implementation of robust variance estimation using the infinitesimal jackknife for statistical validity.

Main Results:

• Multistate models provide a flexible framework for defining target estimands in complex diseases.
• Utilities offer a simple and interpretable method to combine evidence on various treatment effects.
• The infinitesimal jackknife ensures reliable variance estimation for primary trial analyses.

Conclusions:

• Multistate models and utilities are valuable tools for clinical trials in complex diseases.
• These methods enhance the ability to assess treatment effects on diverse disease outcomes, such as bleeding in cancer patients.
• The proposed approach supports robust and interpretable primary analyses in challenging clinical trial settings.