| Sumario: | Real-world data derived from electronic medical records are increasingly used in medical research, but data sharing among institutions still faces privacy and operational challenges. To address these challenges, we developed the Full-Cloud Personal Health Train (FC-PHT), a cloud-based federated analysis system that allows secure, end-to-end data extraction and analysis without transferring raw data. The FC-PHT architecture, implemented using Google Cloud services, consists of three major environments: Analysis Environment, Handler Environment, and Data Provider Environment, which work seamlessly to perform federated analysis. As a pilot evaluation, the FC-PHT was implemented across three Clinical Research Core Hospitals in Japan, and a preliminary analysis of heart failure biomarker related to angiotensin receptor–neprilysin inhibitor (ARNI) therapy was conducted using realworld data databases in these hospitals to evaluate the system’s functionality and feasibility. In the ARNI-related biomarker analysis, BNP showed no notable change after therapy, whereas NT-proBNP tended to decrease, consistent with expected pharmacological effects. The pilot evaluation demonstrated the feasibility of performing cloud-based federated analysis across institutions using the FC-PHT without physical data transfer. These findings suggest that the FC-PHT can securely and efficiently enable multi-institutional real-world data analysis while maintaining privacy and reducing operational burdens on researchers and data management staff.
|
|---|