Application of Computable Phenotyping to Real-World Data to Support the Active Safety Surveillance of CAR T-cell Therapies

External Institution: Johns Hopkins University  

External Collaborators: Taxiarchis Botsis, MSc, MPS, PhD; Tania Jain, MD; Syed Abbas Ali, MD; Julie Baker, MSN, DNP; Kory Kreimeyer, MS; Jonathan Spiker, BS

FDA Collaborators:  Jessica Zhou, MD; Hussein Ezzeldin, PhD; Wencel Valega-Mackenzie, PhD; Mary Rubin, MD; Katherine Barnett, MD; Asha Das, MD; Sundeep Agrawal, MD   

Date Started: 9/30/25 

Regulatory Science Challenge 

When cancer patients undergo CAR T-cell therapies, clinicians regularly monitor for side effects, known as adverse events (AEs). Currently, doctors must manually identify and report these AEs to the FDA, which is both time-consuming and prone to incomplete reporting. Important safety information often gets buried in complex electronic health records that contain extensive patient data from multiple treatments, making it difficult for the FDA to develop complete safety profiles for these therapies.

Objectives and Goals 

To address this problem, the objective of this project is to explore how AE information can be shared between hospitals and the FDA more efficiently. The project aims to create a computable platform within the Johns Hopkins Health System (JHHS) that can automatically detect CAR T-cell therapy AEs using real-world patient data. The platform will have three main components: 1) it will use advanced algorithms to identify AEs of varying complexity, with clinical confirmation by doctors; 2) it will automatically send information about confirmed AEs to the FDA through secure channels; and 3) it will implement technical standards to allow the FDA to access additional de-identified clinical information from JHHS electronic health records through existing health dat networks. This automated approach is designed to reduce the burden of manual reporting for healthcare providers while improving the FDA's ability to monitor the safety of CAR T-cell therapies.