Federated target trial emulation for time-to-event outcomes via POLARIS: Pooled-equivalent One-shot Likelihood...

Area of Science:

• Health Informatics
• Pharmacovigilance
• Clinical Trial Methodology

Background:

• Precision medicine requires detecting heterogeneous treatment effects (HTE), but real-world studies often lack scale and diversity.
• Data-sharing restrictions hinder multi-site analyses, limiting the ability to study HTE across different institutions.
• Existing methods struggle to enable collaborative analysis of sensitive patient data while preserving privacy.

Purpose of the Study:

• To introduce POLARIS, a federated framework for time-to-event target trial emulation.
• To enable scalable, distributed assessment of HTE across diverse health systems without sharing patient-level data.
• To investigate gastrointestinal outcomes associated with GLP-1 receptor agonists (GLP-1RAs) compared to SGLT2is and DPP4is.

Main Methods:

• POLARIS utilizes a federated approach, converting site-specific weighted Cox risk functions into compact tensors.
• These tensors are shared with a coordinating center, allowing lossless reproduction of pooled estimates.
• The framework was applied across five U.S. health systems for target trial emulation.

Main Results:

• GLP-1 receptor agonists (GLP-1RAs) were associated with increased risks of nausea and vomiting.
• HTE were observed, with higher risks noted in men, patients with HbA1c ≥ 8.5%, and those on lipid therapy.
• The study successfully demonstrated the feasibility of distributed HTE assessment.

Conclusions:

• POLARIS offers a scalable and privacy-preserving solution for federated target trial emulation.
• The framework facilitates fine-grained assessment of HTE across diverse patient populations and healthcare settings.
• Findings highlight specific patient subgroups at higher risk for gastrointestinal adverse events with GLP-1RAs.