AUTOMATING BRIEFING BOOK DEVELOPMENT FOR MAJOR EUROPEAN HEALTH TECHNOLOGY ASSESSMENT SUBMISSIONS USING GENERATIVE AI
Author(s)
Namita Tundia, MS, PhD1, Timon Schicht, MSc2, Schiffon L. Wong, MPH3, Sukriti Sharma, MSc4, Ankita Sood, PharmD4, Barinder Singh, RPh4.
1EMD Serono, Billerica, MA, USA, 2Merck Healthcare KGaA, Darmstadt, Germany, 3Schiffon Wong Strategic Advisory, Greater Boston, MA, USA, 4Pharmacoevidence, Mohali, India.
1EMD Serono, Billerica, MA, USA, 2Merck Healthcare KGaA, Darmstadt, Germany, 3Schiffon Wong Strategic Advisory, Greater Boston, MA, USA, 4Pharmacoevidence, Mohali, India.
OBJECTIVES: Briefing books (BB) are strategic documents that facilitate early scientific dialogue with health technology assessment (HTA) bodies, informing trial design and strengthening reimbursement potential. The development process is resource-intensive, demanding significant time and specialist expertise. This research explored the potential of integrating generative artificial intelligence (GenAI) within a Retrieval Augmented Generation (RAG) framework to automate development of traceable, high-quality BB content.
METHODS: Evaior® (Pharmacoevidence AI tool) was utilized to develop BB content for a stroke use case. An automated, end-to-end workflow, with human oversight at each stage, was developed to streamline BB generation for European HTA agencies (NICE, G-BA, HAS). Source documents were indexed within a RAG pipeline. A multi-agentic approach developed specific sections (e.g., disease background, epidemiology, disease burden and management) to produce accurate, traceable outputs with human oversight. Historical submissions for approved stroke interventions were used to shape and contextualize advisory questions tailored to each HTA body.
RESULTS: For each BB, a 10-15-page document was generated. Subject matter experts (SMEs) validated GenAI-generated content for relevance, accuracy, and traceability. Outputs were generated as text and tables and tailored to market requirements. Human input was required for ~5% of disease background and disease management, and 2% of unmet needs sections. GenAI generated questions from historical submissions, those on study design and economic model were incorporated through human review. Validated content was translated into German for G-BA submission and reviewed by a native speaker for contextual and linguistic accuracy. This AI-human hybrid approach generated BB sections within a day vs 5-6 days manually (~75% time savings).
CONCLUSIONS: This research establishes the feasibility of GenAI-generated BB documents for HTA submissions, substantially reducing time and effort while maintaining relevance, accuracy and traceability. The scalability of this approach across multiple HTA agencies positions it as a high-value, operationally efficient solution for manufacturers navigating complex, multi-market reimbursement landscapes.
METHODS: Evaior® (Pharmacoevidence AI tool) was utilized to develop BB content for a stroke use case. An automated, end-to-end workflow, with human oversight at each stage, was developed to streamline BB generation for European HTA agencies (NICE, G-BA, HAS). Source documents were indexed within a RAG pipeline. A multi-agentic approach developed specific sections (e.g., disease background, epidemiology, disease burden and management) to produce accurate, traceable outputs with human oversight. Historical submissions for approved stroke interventions were used to shape and contextualize advisory questions tailored to each HTA body.
RESULTS: For each BB, a 10-15-page document was generated. Subject matter experts (SMEs) validated GenAI-generated content for relevance, accuracy, and traceability. Outputs were generated as text and tables and tailored to market requirements. Human input was required for ~5% of disease background and disease management, and 2% of unmet needs sections. GenAI generated questions from historical submissions, those on study design and economic model were incorporated through human review. Validated content was translated into German for G-BA submission and reviewed by a native speaker for contextual and linguistic accuracy. This AI-human hybrid approach generated BB sections within a day vs 5-6 days manually (~75% time savings).
CONCLUSIONS: This research establishes the feasibility of GenAI-generated BB documents for HTA submissions, substantially reducing time and effort while maintaining relevance, accuracy and traceability. The scalability of this approach across multiple HTA agencies positions it as a high-value, operationally efficient solution for manufacturers navigating complex, multi-market reimbursement landscapes.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR18
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
Cardiovascular Disorders (including MI, Stroke, Circulatory), Neurological Disorders