FROM BESPOKE MODELS TO SHARED INFRASTRUCTURE: NICE ECD18 AND THE FUTURE OF HTA MODELING
Author(s)
Gareth Wynne-McHardy, MBChB, Amanda Hansson Hedblom, MSc, Anna Louise McCormick, DPhil, Gerdi Strydom, MBA, PharmD.
Valid Insight, Macclesfield, United Kingdom.
Valid Insight, Macclesfield, United Kingdom.
OBJECTIVES: This study aims to describe the structural transition in NICE health economic (HE) modelling signalled by ECD18 and to assess the potential role of generative AI (GenAI) tools in supporting model development within emerging reference model frameworks.
METHODS: A narrative review of NICE methodological documentation published between 2016 and 2025 was conducted, with appraisals selected to represent distinct stages in the evolution towards standardised modelling, from multi-submission fragmentation (TA375) to formal pathway modelling (TA964). ECD18 and its cited literature were analysed alongside the ISPOR 2025 working group report on GenAI in health technology assessment (HTA), mapping AI capabilities across evidence synthesis, real-world evidence, and HE model development.
RESULTS: NICE's approach to HE modelling is evolving from bespoke to standardised. In 2022, the COVID-19 multiple technology appraisal introduced a shared, sponsor-independent framework enabling cross-product comparison. TA964 represents the most recent iteration: a reusable, sponsor-independent, disease-level pathway model designed to accommodate successive appraisals within a single validated structure. ECD18 formalises this trajectory by establishing the policy basis for disease-specific reference case extensions and independently developed reference models across therapeutic areas. The first reference case extension is expected for consultation in 2026, with metabolic dysfunction-associated steatohepatitis (MASH) anticipated to follow. As model architecture becomes pre-specified, differentiation opportunities shift towards evidence quality, the robustness of indirect treatment comparisons, and parameter justification. GenAI has shown early promise across all three: human-comparable performance in evidence screening and extraction, automated replication of network meta-analyses, and proof-of-concept reproduction of published economic models. However, ongoing concerns over reproducibility and the need for expert validation remain. AI-assisted tools may support alignment with pre-specified frameworks by expediting the generation of evidence, though their application in HTA contexts remains underdeveloped.
CONCLUSIONS: ECD18 marks a fundamental reorientation of HTA modelling. In doing so, it aligns future modelling with GenAI’s present strengths.
METHODS: A narrative review of NICE methodological documentation published between 2016 and 2025 was conducted, with appraisals selected to represent distinct stages in the evolution towards standardised modelling, from multi-submission fragmentation (TA375) to formal pathway modelling (TA964). ECD18 and its cited literature were analysed alongside the ISPOR 2025 working group report on GenAI in health technology assessment (HTA), mapping AI capabilities across evidence synthesis, real-world evidence, and HE model development.
RESULTS: NICE's approach to HE modelling is evolving from bespoke to standardised. In 2022, the COVID-19 multiple technology appraisal introduced a shared, sponsor-independent framework enabling cross-product comparison. TA964 represents the most recent iteration: a reusable, sponsor-independent, disease-level pathway model designed to accommodate successive appraisals within a single validated structure. ECD18 formalises this trajectory by establishing the policy basis for disease-specific reference case extensions and independently developed reference models across therapeutic areas. The first reference case extension is expected for consultation in 2026, with metabolic dysfunction-associated steatohepatitis (MASH) anticipated to follow. As model architecture becomes pre-specified, differentiation opportunities shift towards evidence quality, the robustness of indirect treatment comparisons, and parameter justification. GenAI has shown early promise across all three: human-comparable performance in evidence screening and extraction, automated replication of network meta-analyses, and proof-of-concept reproduction of published economic models. However, ongoing concerns over reproducibility and the need for expert validation remain. AI-assisted tools may support alignment with pre-specified frameworks by expediting the generation of evidence, though their application in HTA contexts remains underdeveloped.
CONCLUSIONS: ECD18 marks a fundamental reorientation of HTA modelling. In doing so, it aligns future modelling with GenAI’s present strengths.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
HTA386
Topic
Economic Evaluation, Health Technology Assessment, Organizational Practices
Topic Subcategory
Decision & Deliberative Processes, Systems & Structure
Disease
No Additional Disease & Conditions/Specialized Treatment Areas