OBJECTIVES: Our objective was to demonstrate the capabilities of generative AI technology in navigating, extracting, synthesizing, and summarizing common critiques raised by multiple HTA agencies, a time and labor-intensive task for humans.

METHODS: We developed a customized generative AI technology, called ValueGen.AI, to analyze and synthesize collective evidence in 13 HTA reports for Tofacitinib in the management of Ulcerative Colitis from NICE in the UK, HAS in France, and GBA in Germany. First, we extracted and synthesized data using GPT-4o and utilizing LangChain library in Python. Second, we converted processed data into Latex format using Jinja and generated summary PDF files using Miktex. The key functions of this generative AI technology included data extraction and synthesis, summarization of critiques, and insight generation.

RESULTS: The Generative-AI technology effectively synthesized information from the analyzed HTA reports, highlighting common critiques and insights. It revealed that Tofacitinib faced criticism across all HTA agencies for partial compliance with licensing and insufficient information on the comparator arm. Additionally, the trial design was scrutinized by all three countries for weakness in data collection and methodologies, affecting data reliability. NICE was the sole critic raising concerns on safety and analyzing adverse events whereas HAS was the only agency pointing out the impact on public health. Overall, NICE raised the highest number of criticisms among the three HTAs, while GBA raised the fewest. All Generative-AI findings were validated by human-in-the-loop.

CONCLUSIONS: ValueGen.AI, a customized Generative-AI technology, exemplifies a successful and specialized tool in HEOR and pharmaceutical market access. It has demonstrated its capability to provide valuable insights and conduct comprehensive analyses of HTA reports from multiple countries. This underscores Generative-AI’s potential to streamline and enhance the HTA process, offering in-depth insights, reducing manual effort and improving decision-making in healthcare evaluations.