Presentation (CTI)

OBJECTIVES: Mandatory EU HTA Joint Clinical Assessments (JCAs) for oncology therapies necessitate scalable approaches to anticipate JCA PICOS requirements across EU Member States, enabling early identification of evidence gaps and development of evidence-generation strategies. We present a GenAI-assisted JCA-readiness case study in HLA-A*02:01-negative metastatic uveal melanoma, an indication characterised by a high unmet need and no approved systemic therapies.
METHODS: Anticipated JCA PICOS were derived from clinical management guidelines, prior HTA decisions, and development pipeline. An existing SLR (Chen 2026) was updated to a living SLR with downstream integration for PICOS simulations, feasibility assessment, and JCA readiness check. Evidence was mapped to assess feasibility, identify evidence gaps, and inform evidence-generation recommendations. AI-generated outputs were benchmarked against an independent investigator-led reference analysis using a pre-specified scoring framework.
RESULTS: Nine PICOs were simulated, with core comparators: immune checkpoint inhibitors (pembrolizumab, nivolumab plus ipilimumab), liver-directed therapies, and alkylating chemotherapies (dacarbazine, temozolomide, or fotemustine). Subgroups were defined by hepatic versus extrahepatic metastases and performance status. Comparative feasibility assessment indicated NMA was not feasible, anchored ITC was limited by HLA population heterogeneity and mixed investigator's-choice comparator, whereas MAIC/STC was conditionally feasible. Agreement with the reference analysis was achieved for 8/9 feasibility assessment. Evidence gaps included (i) an evolving targeted therapy landscape requiring periodic reassessment, (ii) absent HLA-stratified alkylating chemotherapy data warranting bridging analyses, and (iii) cross-cutting limitations including immature trial data and heterogeneous outcome definitions. The GenAI-assisted workflow, with integrated human oversight, was completed in 4-5 days, resulting in a 70-80% time reduction versus conventional manual workflow.
CONCLUSIONS: This case study demonstrates that multiple interdependent JCA-readiness activities can be integrated within a GenAI-assisted framework, producing outputs consistent with an independent expert reference analysis. Such approaches have the potential to shift HTA evidence planning upstream, enabling proactive evidence generation while complementing expert judgement. Prospective validation against published JCAs is warranted.