LIVING SYSTEMATIC REVIEWS (LSRS) AND ARTIFICIAL INTELLIGENCE (AI) IN THE EUROPEAN JOINT CLINICAL ASSESSMENT (JCA) ERA: ALIGNMENT AND GAPS ACROSS PRISMA, COCHRANE, ISPOR, AND HEALTH TECHNOLOGY ASSESSMENT (HTA) GUIDANCE
Author(s)
Stacy Grieve, PhD1, Mihaela Musat, PhD1, Arup Pramanik, MSc, MD, MBA2, Rozee Liu, MSc1, Anna Forsythe, MBA, MSc, PharmD1;
1Oncoscope-AI, Miami, FL, USA, 2Boehringer Ingelheim, Sharon, MA, USA
1Oncoscope-AI, Miami, FL, USA, 2Boehringer Ingelheim, Sharon, MA, USA
OBJECTIVES: LSRs aim to provide the most up-to-date information required in a highly dynamic landscape of pharmaceutical and clinical research. AI technologies can improve the speed and quality of LSRs; however, a clear methodological framework for their use is lacking. We evaluated whether existing methodological and HTA guidance address LSRs and AI-assisted LSRs conduct, reporting, and use in HTA and JCA.
METHODS: A structured review of publicly available guidance was conducted across six sources: PRISMA (PRISMA 2020, PRISMA-LSR extension, PRISMA-trAIce), Cochrane Handbook and LSR guidance, RAISE, ISPOR Good Practices and publications, NICE methods manuals, and JCA scoping guidance. Each document was reviewed for explicit guidance on: update frequency, versioning, study mapping, governance of changing inclusion/exclusion criteria, AI use (screening, extraction, synthesis), transparency/auditability, and PICO-stratified outputs. Findings were summarized quantitatively.
RESULTS: Across 14 distinct guidance documents, 50% explicitly defined or endorsed LSRs, but only 2 specified operational update mechanisms beyond periodic updates (for example continuous surveillance). None provide recommendations for daily/real-time updating. Guidance on AI use for systematic reviews was identified in 8/14 of documents; none presented clear, auditable requirements for model validation, performance variations, or traceability of AI decisions. Rules for managing multiple publications of the same study were partially addressed in 50% of documents, but without formal processes for grouping and documenting publications of the same study-family. Only one document discussed governance of evolving inclusion/exclusion criteria in response to new evidence. JCA guidance mandates PICO-specific reporting, yet there were no clear recommendations on dynamically re-synthesizing evidence by PICO as evidence changes.
CONCLUSIONS: Current guidance treats LSRs and AI-enabled systematic reviews as separate concepts, resulting in fragmented standards. Existing frameworks do not adequately support real-time LSRs aligned with JCA PICO requirements. An extension to PRISMA-LSR (“PRISMA-LSR+”) is needed to formalize real-time updating, study-family management, AI transparency, and dynamic PICO-based reporting.
METHODS: A structured review of publicly available guidance was conducted across six sources: PRISMA (PRISMA 2020, PRISMA-LSR extension, PRISMA-trAIce), Cochrane Handbook and LSR guidance, RAISE, ISPOR Good Practices and publications, NICE methods manuals, and JCA scoping guidance. Each document was reviewed for explicit guidance on: update frequency, versioning, study mapping, governance of changing inclusion/exclusion criteria, AI use (screening, extraction, synthesis), transparency/auditability, and PICO-stratified outputs. Findings were summarized quantitatively.
RESULTS: Across 14 distinct guidance documents, 50% explicitly defined or endorsed LSRs, but only 2 specified operational update mechanisms beyond periodic updates (for example continuous surveillance). None provide recommendations for daily/real-time updating. Guidance on AI use for systematic reviews was identified in 8/14 of documents; none presented clear, auditable requirements for model validation, performance variations, or traceability of AI decisions. Rules for managing multiple publications of the same study were partially addressed in 50% of documents, but without formal processes for grouping and documenting publications of the same study-family. Only one document discussed governance of evolving inclusion/exclusion criteria in response to new evidence. JCA guidance mandates PICO-specific reporting, yet there were no clear recommendations on dynamically re-synthesizing evidence by PICO as evidence changes.
CONCLUSIONS: Current guidance treats LSRs and AI-enabled systematic reviews as separate concepts, resulting in fragmented standards. Existing frameworks do not adequately support real-time LSRs aligned with JCA PICO requirements. An extension to PRISMA-LSR (“PRISMA-LSR+”) is needed to formalize real-time updating, study-family management, AI transparency, and dynamic PICO-based reporting.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
P3
Topic
Health Technology Assessment
Topic Subcategory
Systems & Structure
Disease
SDC: Oncology