Presentation (CTI)

OBJECTIVES: Rapidly expanding clinical evidence can quickly reduce the currency of conventional systematic literature reviews (SLRs). Evolving health technology assessment (HTA) requirements, including the European Union Joint Clinical Assessment, require timely, transparent, and reproducible evidence syntheses within compressed timelines, necessitating living SLRs; for which responsible generative artificial intelligence (GenAI) offers a promising solution. We present a GenAI-enabled framework that generates de novo SLRs or transforms existing reviews into living SLRs, demonstrated through a case study in metastatic uveal melanoma (mUM).
METHODS: A GenAI-enabled living evidence framework was developed comprising: (1) planned quarterly/ milestone-based surveillance of bibliographic databases and trial registries; (2) incremental identification and ingestion of new evidence; (3) structured harmonisation of study characteristics, PICOs, and outcomes into a version-controlled repository; and (4) automated change detection-flagging clinically meaningful updates for validation. Responsible AI guardrails, including predefined decision rules and human-in-the-loop quality checkpoints, were embedded throughout. The framework was developed in alignment with Cochrane guidance for living SLRs and was demonstrated by updating a published mUM SLR (Chen, 2026).
RESULTS: Continuous surveillance over a 3-month period identified three newly eligible comparative studies-increasing the evidence base from 16 studies to 19, including an investigational combination targeted therapy assessed in HLA-A*02:01-negative mUM. Each new study fed the evidence repository, automatically refreshing structured evidence tables and PICO maps with full version control and traceability. Emerging evidence was autonomously flagged for reviewer assessment. The full pipeline, including human oversight at each stage was completed within one week, achieving a ~70-80% reduction in turnaround time versus a conventional human-only approach.
CONCLUSIONS: The proposed framework transforms SLR maintenance from a periodic, resource-intensive process into a continuous, structured workflow; extendable to downstream activities such as feasibility assessment, and ITC/NMA. Continuous surveillance with responsible GenAI-assisted updates enables scalable maintenance of living reviews while preserving transparency, traceability, reproducibility, with human oversight.