OBJECTIVES: Scrutinizing Health Technology Assessment (HTA) reports is a resource-intensive but critical task for a successful market access strategy. Our objective was to showcase Generative-AI's capabilities in navigating and summarizing HTA reports.

METHODS: A generative-AI technology tailored for the life sciences industry, ValueGen.AI, was developed to analyze HTA reports, synthesize evidence and respond to ad-hoc user queries. For this study, we considered hepatocellular carcinoma and included all relevant NICE reports from the UK (NICE_TA849 on Cabozantinib, NICE_TA666 on Atezolizumab plus Bevacizumab, and NICE_TA555 on Regorafenib) over the past 3 years. Data extraction and synthesis were conducted using GPT-4o and LangChain library in Python. For demonstration, the Generative-AI was tasked to: 1) provide a safety and tolerability profile summary, 2) identify primary and secondary endpoints 3) analyze key critiques raised in the reports.

RESULTS: The AI effectively navigated and summarized information from HTA reports. In terms of safety, Cabozantinib was associated with manageable adverse events like diarrhea, palmar-plantar erythrodysesthesia, and hypertension, with perceived higher toxicity than Regorafenib. Atezolizumab plus Bevacizumab was associated with common adverse events including diarrhea, fatigue, hypertension, and elevated liver enzymes. Regorafenib was associated with serious adverse events such as physical health deterioration, ascites, and hepatic failure. Primary and secondary endpoints varied: Cabozantinib focused on overall survival (OS) with PFS and ORR as secondary, while Atezolizumab plus Bevacizumab and Regorafenib emphasized OS and PFS with additional endpoints like ORR, TTP, DCR, and quality of life. Key critiques included concerns over trial population comparability, higher toxicity for Cabozantinib, uncertainties in NMA results and generalizability for Atezolizumab plus Bevacizumab, and efficacy uncertainties and potential biases in utility values for Regorafenib. All Generative-AI responses were validated by human-in-the-loop.

CONCLUSIONS: Generative-AI technology can provide valuable insights and comprehensive analysis of HTA reports and makes navigating the complex landscape of global health technology assessments easy.