Assessing treatment effects with surrogate survival outcomes using an internal validation subsample.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Survival Analysis

Background:

• Studies often have surrogate outcomes for all subjects but true outcomes for only a subset.
• Assessing treatment effects requires methods that can integrate both types of endpoints.
• Existing methods may not fully leverage available data when true outcomes are partially missing.

Purpose of the Study:

• To develop a novel semiparametric estimated likelihood method for proportional hazards models.
• To enable real-time validation of surrogate outcomes and handle flexible censoring.
• To accurately estimate treatment effects using both surrogate and true outcome data.

Main Methods:

• Developed a semiparametric estimated likelihood approach for discrete time data.
• Incorporated a binary covariate of interest within a proportional hazards framework.
• Allowed for flexible censoring and real-time validation of surrogate endpoints.

Main Results:

• The proposed estimator is consistent and asymptotically normal.
• Demonstrated unbiased estimation of covariate effects compared to using only surrogate endpoints.
• Showed improved efficiency over complete-case analysis when dealing with missing true outcomes.
• Illustrated application using Alzheimer's Disease Neuroimaging Initiative data.

Conclusions:

• The method effectively accounts for surrogate outcome uncertainty using a validation subsample.
• The proposed estimator surpasses standard semiparametric survival analysis methods.
• This approach can reduce trial costs and enhance the power to detect treatment effects.