Solutions for surrogacy validation with longitudinal outcomes for a gene therapy

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Causal Inference

Background:

• Surrogate endpoints (S) are crucial for assessing treatment efficacy when true outcomes (T) are delayed or difficult to measure.
• Validating surrogate endpoints ensures reliable and efficient clinical trial outcomes.
• Existing validation methods may not fully capture the complexities of longitudinal data and treatment designs.

Purpose of the Study:

• To propose and validate a causal inference approach for surrogate endpoints using longitudinal true outcome measurements.
• To define principal surrogacy criteria applicable across different study periods and treatment designs.
• To address identifiability challenges using Bayesian methods and sensitivity analyses.

Main Methods:

• Utilized a causal inference framework incorporating longitudinal data of true outcomes.
• Employed mixed modeling and principal surrogacy criteria for validation.
• Applied a Bayesian approach for estimation, incorporating informative priors and sensitivity analyses for identifiability.

Main Results:

• Developed a method to validate surrogate endpoints considering surrogate-dependent treatment efficacy curves.
• Extended the approach to delayed-start treatment designs.
• Evaluated the frequentist properties of the proposed Bayesian imputation method.

Conclusions:

• The proposed causal inference approach provides a robust framework for validating surrogate endpoints with longitudinal data.
• This method is applicable to complex trial designs, including delayed-start treatments.
• The Bayesian approach with sensitivity analysis improves parameter identifiability and inference reliability in surrogate endpoint validation.