IDENTIFYING VALUE DRIVERS IN HTA SUBMISSIONS USING GENERATIVE AI: A CASE STUDY OF LONG-ACTING INJECTABLE HIV THERAPY
Author(s)
Turgay Ayer, PhD1, Sumeyye Samur, PhD1, Ismail F. Yildirim, MSc1, Mine Tekman, PhD1, Jag Chhatwal, PhD2;
1Value Analytics Labs, Boston, MA, USA, 2Massachusetts General Hospital/Harvard Medical School, Boston, MA, USA
1Value Analytics Labs, Boston, MA, USA, 2Massachusetts General Hospital/Harvard Medical School, Boston, MA, USA
OBJECTIVES: Understanding the drivers of historical health technology assessment (HTA) decisions can inform future evidence generation and support timely, decision-relevant research. This study evaluated whether GenAI can systematically identify, structure, and synthesize key value drivers across European HTA agencies, using long-acting injectable HIV therapy as a case study.
METHODS: An agentic GenAI architecture was developed using the LangGraph framework to orchestrate more than 20 specialized sub-agents designed to decompose complex queries and coordinate evidence extraction across multiple sources. The platform, ValueGen.AI, was applied to long-acting injectable antiretroviral therapy for HIV, using cabotegravir plus rilpivirine (CAB/RPV) as an exemplar. The system automatically retrieved, synthesized, and organized publicly available evidence cited in European HTA assessments, including clinical trial publications (ATLAS, FLAIR, ATLAS-2M), patient-reported outcome studies, real-world and implementation literature, and national HTA decisions from NICE (England and Wales), SMC (Scotland), HAS (France), IQWiG/G-BA (Germany), and AIFA (Italy). Extracted evidence was categorized into clinical effectiveness, patient preference and health-related quality of life (HRQoL), adherence and persistence, resistance risk and eligibility, healthcare resource use, and cost-effectiveness drivers.
RESULTS: Across 5 HTA bodies, ValueGen.AI identified common core clinical value drivers including non-inferior maintenance of virologic suppression versus daily oral antiretroviral therapy, and non-inferior every-two-month dosing compared with monthly administration. Patient preference, treatment satisfaction, and reduced pill burden were frequently cited as important contributors to perceived value, although their quantitative impact was applied conservatively in economic models. Common risk considerations included strict eligibility criteria, resistance management related to the long pharmacokinetic “tail,” and clinic resource requirements for injection delivery. Cost-effectiveness conclusions varied by jurisdiction and were strongly influenced by pricing arrangements, dosing frequency, clinic costs, and assumptions regarding adherence effects.
CONCLUSIONS: Generative AI can systematically deconstruct and organize HTA decision drivers across jurisdictions, offering a scalable approach to inform evidence planning, value messaging, and economic modeling for future submissions.
METHODS: An agentic GenAI architecture was developed using the LangGraph framework to orchestrate more than 20 specialized sub-agents designed to decompose complex queries and coordinate evidence extraction across multiple sources. The platform, ValueGen.AI, was applied to long-acting injectable antiretroviral therapy for HIV, using cabotegravir plus rilpivirine (CAB/RPV) as an exemplar. The system automatically retrieved, synthesized, and organized publicly available evidence cited in European HTA assessments, including clinical trial publications (ATLAS, FLAIR, ATLAS-2M), patient-reported outcome studies, real-world and implementation literature, and national HTA decisions from NICE (England and Wales), SMC (Scotland), HAS (France), IQWiG/G-BA (Germany), and AIFA (Italy). Extracted evidence was categorized into clinical effectiveness, patient preference and health-related quality of life (HRQoL), adherence and persistence, resistance risk and eligibility, healthcare resource use, and cost-effectiveness drivers.
RESULTS: Across 5 HTA bodies, ValueGen.AI identified common core clinical value drivers including non-inferior maintenance of virologic suppression versus daily oral antiretroviral therapy, and non-inferior every-two-month dosing compared with monthly administration. Patient preference, treatment satisfaction, and reduced pill burden were frequently cited as important contributors to perceived value, although their quantitative impact was applied conservatively in economic models. Common risk considerations included strict eligibility criteria, resistance management related to the long pharmacokinetic “tail,” and clinic resource requirements for injection delivery. Cost-effectiveness conclusions varied by jurisdiction and were strongly influenced by pricing arrangements, dosing frequency, clinic costs, and assumptions regarding adherence effects.
CONCLUSIONS: Generative AI can systematically deconstruct and organize HTA decision drivers across jurisdictions, offering a scalable approach to inform evidence planning, value messaging, and economic modeling for future submissions.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
HTA21
Topic
Health Technology Assessment
Topic Subcategory
Decision & Deliberative Processes, Value Frameworks & Dossier Format
Disease
SDC: Infectious Disease (non-vaccine)