Predictive cluster level surrogacy in the presence of interference

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trial Design

Background:

• Evaluating surrogate outcomes is challenging, particularly with interference in cluster-level studies.
• Existing causal inference frameworks provide a basis for understanding treatment effects but require extension for complex scenarios.

Purpose of the Study:

• To propose a novel procedure for evaluating surrogate outcomes in the presence of interference at the cluster level.
• To extend causal inference methods to assess direct, indirect, and total effects for surrogate evaluation.
• To identify predictive cluster-level surrogates for interventions like transmission-blocking malaria vaccines.

Main Methods:

• Leveraging the causal-association paradigm for defining surrogacy.
• Extending Hudgens and Halloran's estimators for causal inference with interference.
• Applying proposed criteria and procedures to simulated data from a malaria vaccine trial.

Main Results:

• The proposed procedure facilitates the evaluation of surrogates for indirect and spill-over clinical effects.
• Existing estimators can be adapted to evaluate biomarkers under the new definition of surrogacy.
• Demonstrated utility in identifying predictive cluster-level surrogates using simulated malaria vaccine trial data.

Conclusions:

• The developed framework enables robust surrogate evaluation in the complex setting of cluster-level interference.
• This approach is vital for developing interventions like transmission-blocking vaccines where direct effects may be minimal.
• The study provides a practical procedure for identifying predictive biomarkers in public health interventions.