Federated target trial emulation using distributed observational data for treatment effect estimation

Area of Science:

• Health Informatics
• Epidemiology
• Machine Learning

Background:

• Target trial emulation (TTE) uses real-world data to simulate clinical trials for treatment effect estimation.
• Distributed TTE enhances generalizability but faces privacy and data-sharing challenges.
• Existing methods struggle with cross-site analysis without compromising patient data.

Purpose of the Study:

• To introduce a novel Federated Learning-based Target Trial Emulation (FL-TTE) framework.
• To enable TTE across multiple sites without sharing patient-level data.
• To facilitate privacy-preserving, federated treatment effect estimation.

Main Methods:

• Developed FL-TTE incorporating federated protocol design.
• Implemented federated inverse probability of treatment weighting.
• Utilized a federated Cox proportional hazards model for time-to-event outcomes.

Main Results:

• Validated FL-TTE on Sepsis trials (eICU, MIMIC-IV) and Alzheimer's trials (INSIGHT Network).
• FL-TTE yielded less biased estimates compared to traditional meta-analysis.
• Demonstrated theoretical support for the federated approach.

Conclusions:

• FL-TTE successfully enables federated treatment effect estimation across distributed, heterogeneous data.
• The framework preserves data privacy, overcoming significant real-world data challenges.
• FL-TTE offers a robust solution for large-scale, multi-site clinical trial emulation.