The individual-level surrogate threshold effect in a causal-inference setting with normally distributed endpoints.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Statistical Evaluation of Surrogate Endpoints

Background:

• Trial-level metrics like R-squared and surrogate threshold effect (STE) assess surrogacy but lack individual interpretability.
• Existing frameworks primarily focus on expected causal effects, not individual-level predictions.
• Alonso et al. proposed evaluating surrogacy based on individual causal treatment effects.

Purpose of the Study:

• Introduce the individual-level surrogate threshold effect (ISTE) for normally distributed surrogate (S) and true (T) endpoints.
• Provide a clinically interpretable metric for assessing individual-level surrogacy.
• Demonstrate the application and interpretation of ISTE in a case study.

Main Methods:

• Define ISTE as the minimum individual causal effect on S predicting a positive individual causal effect on T (lower prediction interval limit > 0).
• Develop methodology for normally distributed S and T variables.
• Apply the proposed ISTE methodology in a practical case study.

Main Results:

• The newly proposed ISTE metric demonstrates an appealing clinical interpretation.
• The case study illustrates the practical application and interpretability of ISTE.
• An R package 'surrogate' is available to implement the methodology.

Conclusions:

• ISTE provides a valuable, clinically interpretable metric for individual-level surrogate endpoint evaluation.
• The methodology extends existing surrogacy frameworks to individual predictions.
• ISTE enhances the assessment of surrogate endpoint validity in clinical research.