Surrogate markers and joint models for longitudinal and survival data

Area of Science:

• Biostatistics
• Clinical Trial Design
• Epidemiology

Background:

• Surrogate endpoints offer faster, cheaper clinical trials but require rigorous validation.
• Prentice's framework defines conditions for valid surrogate endpoints, focusing on joint distributions.
• Quantifying the proportion of treatment explained (PTE) is an intuitive measure for surrogate marker utility.

Purpose of the Study:

• To investigate the consistency of survival models used for PTE calculation.
• To evaluate the estimation of PTE and other surrogacy measures using a joint model.
• To emphasize the importance of understanding marker-clinical endpoint joint distributions for surrogate endpoint validation.

Main Methods:

• Developed a joint model for longitudinal markers and clinical endpoints.
• Analyzed the consistency requirements for separate survival models used in PTE calculation.
• Conducted simulation studies to compare PTE statistics and surrogacy measures within the joint model framework.

Main Results:

• Demonstrated that consistency between two survival models necessitates a specific joint model assumption.
• Showed that the Freedman et al. PTE statistic can be estimated using the joint model.
• Validated the joint model approach for estimating surrogacy measures derived from Prentice's framework.

Conclusions:

• Accurate surrogate endpoint evaluation hinges on a thorough understanding of the joint marker-clinical endpoint distribution.
• The proposed joint modeling approach provides a consistent framework for assessing surrogate marker validity.
• This work highlights the necessity of joint modeling for reliable surrogate endpoint validation in clinical research.