Artificial intelligence (AI) approaches may help improve identification and monitoring of immune-related adverse events associated with immune checkpoint inhibitors, according to a scoping review published in JCO Clinical Cancer Informatics.
Immune checkpoint inhibitors have expanded treatment options across multiple tumour types, but their immune-activating mechanism can also lead to immune-related adverse events (irAEs) affecting organs such as the lungs, liver, gastrointestinal tract, endocrine glands, and heart. Detecting these toxicities can be challenging, particularly because they may be documented in narrative clinical notes rather than structured fields within electronic health records.
CLINICAL SUMMARY
What was examined
A scoping review assessed studies investigating the use of artificial intelligence methods—including machine learning and natural language processing—to detect and predict immune-related adverse events associated with immune checkpoint inhibitors.
Key findings
-
AI tools can analyse electronic health records and clinical notes to identify immune-related adverse events.
-
Natural language processing may help detect toxicities documented in narrative clinical documentation but not captured in structured coding systems.
-
Machine-learning models have also been explored to estimate patient risk of immunotherapy toxicities.
Clinical implications
-
AI approaches may support improved monitoring of immunotherapy-related adverse events.
-
Further prospective validation is required before routine use in clinical practice.
To explore how computational tools might help address this challenge, researchers reviewed studies evaluating AI approaches—including machine learning, deep learning, and natural language processing—for identifying, predicting, and monitoring immunotherapy-associated toxicities.
Across the literature, natural language processing was commonly used to analyse unstructured clinical documentation and identify adverse events that may not be captured through traditional diagnostic coding. Some studies suggest these approaches may support more comprehensive detection of irAEs compared with conventional coding-based surveillance.
Featured Event
Several studies also explored machine-learning models designed to estimate a patient’s risk of developing immune-related toxicities, although these predictive approaches remain under investigation.
However, most studies identified in the review were retrospective and based on data from single institutions, highlighting the need for prospective validation and evaluation across different healthcare settings.
The authors note that further research will be required to determine how AI tools could be integrated into clinical workflows. As oncology datasets continue to expand, computational approaches may help support more systematic monitoring of immunotherapy-associated toxicities.
Paper: Chin Hang Yiu et al. Leveraging Artificial Intelligence for Immune Checkpoint Inhibitor Safety: A Scoping Review of Current Applications. JCO Clin Cancer Inform 10, e2500323(2026). DOI:10.1200/CCI-25-00323. Access online here.
