AI/ML IN HTA EVIDENCE SYNTHESIS: WHERE DO WE STAND TODAY?
Author(s)
Raju Gautam, PhD1, Saeed Anwar, MSPharm2, Ratna Pandey, MSc2, Khushbu Baranwal, MSc2, Tushar Srivastava, MSc1;
1ConnectHEOR, London, United Kingdom, 2ConnectHEOR, Delhi, India
1ConnectHEOR, London, United Kingdom, 2ConnectHEOR, Delhi, India
OBJECTIVES: National Institute for Health and Care Excellence (NICE) and Canada Drug Agency (CDA) published an AI-position statement - a guiding principle for the use of artificial intelligence (AI) in the generation and submission of evidence. AI/machine learning (AI/ML) for systematic literature review (SLR) has increased significantly, yet their documented use in formal health technology assessment (HTA) submissions remain unclear. This study aims to identify HTA submissions that reported the use of AI/ML for SLRs.
METHODS: We conducted comprehensive research that followed a two-step process. First, HTA guidance documents from major agencies (NICE, HAS, IQWiG, NCPE, SMC, TLV, CDA, and PBAC) were systematically reviewed to identify guidance on the use of AI/ML in HTA-compliant SLRs. Second, for HTA agencies that refer to AI/ML in the context of SLR, a comprehensive review of HTA submissions from past three years was conducted. The HTA submissions included were reviewed to assess the scope and context of AI use in these SLRs.
RESULTS: Only NICE, CDA and IQWiG recommended the use of AI/ML for SLRs. A total of 1,391 HTA documents (NICE [468], CDA [264], and IQWiG [659]) were screened. Across the reviewed documents, AI/ML methods were reported in only three HTA documents, all from NICE. One technology appraisal (TA11540) reported the use of AI for SLR study screening. Two additional technology appraisals reported limited application of AI, with TA1071 using an AI-assisted ad-hoc literature search to help identify potential treatment effect modifiers, and TA962 applying a natural language processing approach to update the SLR and assist with search strategy development through identification of keywords within publications.
CONCLUSIONS: Current HTA position papers acknowledge the potential of AI in evidence generation, however, its implementation within HTA evidence reviews remains limited. Bridging this gap could enable more efficient and robust evidence synthesis processes, ultimately facilitating more informed and accelerated decision-making.
METHODS: We conducted comprehensive research that followed a two-step process. First, HTA guidance documents from major agencies (NICE, HAS, IQWiG, NCPE, SMC, TLV, CDA, and PBAC) were systematically reviewed to identify guidance on the use of AI/ML in HTA-compliant SLRs. Second, for HTA agencies that refer to AI/ML in the context of SLR, a comprehensive review of HTA submissions from past three years was conducted. The HTA submissions included were reviewed to assess the scope and context of AI use in these SLRs.
RESULTS: Only NICE, CDA and IQWiG recommended the use of AI/ML for SLRs. A total of 1,391 HTA documents (NICE [468], CDA [264], and IQWiG [659]) were screened. Across the reviewed documents, AI/ML methods were reported in only three HTA documents, all from NICE. One technology appraisal (TA11540) reported the use of AI for SLR study screening. Two additional technology appraisals reported limited application of AI, with TA1071 using an AI-assisted ad-hoc literature search to help identify potential treatment effect modifiers, and TA962 applying a natural language processing approach to update the SLR and assist with search strategy development through identification of keywords within publications.
CONCLUSIONS: Current HTA position papers acknowledge the potential of AI in evidence generation, however, its implementation within HTA evidence reviews remains limited. Bridging this gap could enable more efficient and robust evidence synthesis processes, ultimately facilitating more informed and accelerated decision-making.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR140
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas