Accelerating HTA Readiness with Generative AI

Bringing innovative therapies to patients has always required scientific rigor, clinical evidence, and thoughtful evaluation of safety and value. Today, the regulatory agencies making those evaluations include the National Institute for Health and Care Excellence (NICE) in the United Kingdom, Haute Autorité de Santé (HAS) in France, Federal Joint Committee (G-BA) and the Institute for Quality and Efficiency in Health Care in Germany, Canadian Agency for Drugs and Technologies in Health (CADTH) in Canada, FDA in the United States, and emerging bodies across Asia-Pacific (APAC) and Latin America (LATAM).

Now, with the European Union Health Technology Assessment (HTA) regulation coming into force as part of the required Joint Clinical Assessment (JCA) for oncology and advanced therapy medicinal products, the volume of evidence that must be aligned, organized, and well-structured across markets has increased exponentially (1).

Teams report that HTA preparation consumes an undue amount of bandwidth for clinical, medical, health economics outcomes research (HEOR), and market access professionals, individuals who are already stretched thin in their roles (2). Depending upon their markets, a given organization could be simultaneously conducting:

• a JCA for scientific evaluation in the EU
• a NICE Single Technology Appraisal for the UK
• a HAS Transparency Committee evaluation for reimbursement in France
• a G-BA dossier for Germany
• FDA evidence and labeling submissions
• pricing and reimbursement filings across LATAM and APAC.

For therapies in oncology, rare disease, and cell and gene therapy, these submissions are particularly demanding (3). Patient populations are small or heterogeneous. Clinical evidence may rely on single-arm trials or surrogate endpoints. Real-world evidence (RWE) is being required earlier than ever. Comparator landscapes can shift quickly, and payers request ongoing demonstration of value.

This is where generative artificial intelligence (GenAI) has begun to meaningfully reshape the process (Figure 1)—not by replacing scientific expertise, but by removing the friction between data, evidence, and narrative, and giving time back to the experts involved in advancing care (4–7).

What do the HTA workflows look like?

The workflow challenges of evidence synthesis, content generation, and regulatory reporting are daunting. It’s a laborious slog for busy, high-demand professions where their time is often better spent applying their expertise rather than collating, screening, and writing reports from scratch. Five practical HTA workflows from as-yet unpublished case studies are benefiting greatly from GenAI impact where it already is delivering measurable value, with several emerging frontier applications also poised to streamline the workflow.

Automating documentation creation without losing strategic voice. Every HTA submission involves substantial narrative documentation: clinical evidence summaries, unmet need justification, methodological rationale, comparator selection reasoning, adverse event interpretation, and more. Organizations are constantly required to develop reporting documentation. Traditionally, these documents are written and rewritten across multiple internal reviews, spanning weeks if not months.

With GenAI-supported document automation, organizations can:

• suggest population/patient, intervention, comparison, outcome (PICO) sets for scoping purposes
• generate first draft of JCA dossier and NICE-aligned documentation (among other HTA requirements) based on validated clinical trial data inputs
• produce consistently structured narratives for efficacy, safety, and patient relevance
• harmonize terminology across medical, regulatory, and HEOR contributions
• keep living documents updated automatically as new evidence emerges.

Example: Oncology JCA submission (solid tumors). An oncology manufacturer preparing for a JCA plus parallel NICE filing explored using GenAI capabilities to structure clinical narratives and outcome summaries directly from the protocol and clinical study report data. The AI platform would generate first drafts of evidence descriptions and comparator rationale sections in hours instead of weeks. The medical writing team would retain full scientific control, editing for nuance, but no longer must start from the ground up. The team has estimated that GenAI use would reduce its initial dossier drafting by about 40% timewise, with writers focusing on interpretation, not research compiling and formatting.

AI-assisted screening of abstracts reducing manual review burden. Systematic literature reviews (SLRs) underpin the majority of HTA submissions. Screening abstracts manually—especially in oncology, immunology, and metabolic disease—can involve thousands of citations.

GenAI can act as a priority ranking and second-reviewer mechanism by:

• identifying abstracts most likely to meet inclusion criteria
• suggesting probable exclusions with detailed auditable explanations
• supporting double-review workflows while preserving Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) compliance.

Example: Rare disease evidence landscape scan. A biotech company developing a therapy for a rare neuromuscular disorder faced a limited but globally dispersed literature base. The process the team typically used for SLR consumed months whenever comparator evidence was updated. With GenAI-enabled screening, the team cut its screening workload in half and moved faster toward the time-to-insights needed for payer engagement. In reducing screening labor by nearly 60%, the team achieved faster alignment on evidence gaps for value messaging.

Synthesizing published literature into clear, reviewer-ready evidence narratives. After screening comes synthesis, the most scientifically demanding phase of the process. Researchers must interpret several areas of dense information, including:

• study design differences
• population heterogeneity
• comparator selection
• endpoint variability
• clinical relevance.

With all screening decisions validated by the medical writer, a literature review can be enabled by GenAI without losing subject matter expert (SME) oversight, meeting HTA requirements while leveraging GenAI efficiency. GenAI helps generate evidence tables, comparison matrices, and narrative summaries that the SMEs will review, validate, and refine.

Example: First-in-class cell & gene therapy (CAR-T). For a chimeric antigen receptor T-cell (CAR-T) therapy with single-arm pivotal trial evidence, the manufacturer needed to contextualize outcomes using real-world comparator data. GenAI helped synthesize structured summaries of survival outcomes, toxicity management, and patient eligibility patterns across multiple observational sources.

Human HEOR professionals maintained full interpretive authority. They had the benefit of starting from clear, structured evidence, rather than raw literature. Improved clarity in payer narrative around clinical significance of durable response, as well as time-savings in producing results.

Extracting insights from real-world and unstructured clinical data. With RWE of increasing importance and HTA agencies beginning to expect RWE to support generalizability, teams need to analyze a variety of unstructured sources, like the following:

• electronic health record physician notes
• patient registry datasets
• chart review data
• patient-reported outcomes
• digital pathology and genomic markers.

GenAI models trained on biomedical language can be extremely beneficial, for example, by identifying cohorts, extracting survival and response outcomes, normalizing terminology, and mapping longitudinal patient journeys.

Example: Oncology plus genomic subtyping for HTA support. A precision oncology company needed to justify a biomarker-defined patient subgroup for reimbursement. GenAI extracted structured genomic and clinical outcomes across multiple registries, enabling the team to present a clear patient selection justification that aligned to HTA expectations. Payer reviewers gained confidence in population definition and real-world relevance.

Compiling and updating global value dossiers (GVDs). GVDs must remain living documents, adapting continually as evidence and care standards evolve. For GVDs, GenAI enables:

• automated GVD drafting from validated evidence models
• rapid updating when new trial or safety data emerges
• consistency across NICE, HAS, G-BA, JCA, and CADTH submissions
• reuse of approved core messaging across markets.

Example: Rare disease launch sequencing across Europe. A manufacturer preparing parallel submissions in the UK, France, Germany, and Spain considered using GenAI to maintain one centralized GVD that could automatically generate national dossier adaptations. Instead of rewriting country by country, the team would review and refine GenAI-generated country-specific variants. By launching countries in tighter sequence, the team would significantly reduce access delays.

What are some emerging frontier applications for GenAI?

Many organizations deploying GenAI in next-generation workflows are continuing to expand support for evidence synthesis and HTA reporting. Areas for expansion include:

• structuring omics and companion diagnostic performance data for HTAs
• identifying efficacy gaps for potential product repositioning or repurposing
• supporting sensitivity analyses for cost-effectiveness models (with human validation)
• maintaining living evidence ecosystems that alert teams when a new study impacts payer positioning
• generating early payer objection simulations based on precedent from historical HTA decisions.

Through all these projects, teams prioritized specific manual activities for GenAI support while ensuring their professional judgment was an essential part of the process. To remain compliant with HTA requirements, organizations typically rely on AI to synergize content and information using data sets and structures that enable transparency and traceability.

Following this strategy, GenAI does not eliminate the need for scientific reasoning, clinical judgment, or strategic narrative development. It accelerates the research while experts continue to validate the science. It frees experts to do their work at a greater scale.

The types of teams that succeed with GenAI in HTA workflows are those that observe the following steps:

• maintaining transparent data provenance as a top priority
• embedding structured expert, human oversight within their process as a best practice
• treating AI as a force multiplier, not a shortcut and certainly not a replacement for internal expertise
• aligning cross-functional teams on evidence interpretation, not document formatting, to optimize their time and resources in the most efficient way.

This hybrid model provides the best support for HTA deliverables when incorporating GenAI, enabling the preservation of scientific integrity while improving time-to-patient access.

What is a more efficient, more equitable path to access?

Once an industry player demonstrates a significant time- and efficiency-advantage in the regulatory path to patient access, the bar is raised for everyone else. Likewise, as GenAI becomes more accepted, HTAs will require faster evidence synthesis with ever more robust validation capabilities. More transparent justification of value will also be necessary as the technology is tested with projects of greater complexity.

As clinical landscapes shift and more innovative, unique therapies enter the market, AI solutions will need more agility to continue to evolve in such an environment. GenAI—implemented responsibly—creates the conditions that can enable such a possible future. Responsible deployment means that GenAI does not replace the experts but rather returns science to the center of their work. Most importantly, as HTAs consider GenAI as an enabler to research—again, when executed responsibly—it ultimately serves to shorten the path between medical innovation and the patients who need it.

References

1. European Commission. Joint Clinical Assessment for Medicinal Products (EC, January 2025). https://health.ec.europa.eu/document/download/ced91156-ffe1-472d-85eb-aa6a91dd707e_en?filename=hta_htar_factsheet-jca_en.pdf
2. Mitchell, M. and Mull, N. A Standardized HTA Report Summary for Rapidly Presenting Outside Findings, poster at Cochrane Colloquium virtual forum, Oct. 4-7, 2020. https://abstracts.cochrane.org/2020-abstracts/standardized-hta-report-summary-rapidly-presenting-outside-findings
3. Wafqui, F.; et al. The EU HTA Regulation: What Will It Mean for HTA Bodies, Industry and Patients?. The London School of Economics and Political Science, LSE Health research blog, https://www.lse.ac.uk/lse-health/research/mtrg-pages/EU-HTA-Regulation-Blog
4. Oliver, G. Streamlining JCA Response with GenAI. PharmaPhorum, April 16, 2025, https://pharmaphorum.com/market-access/streamlining-jca-response-genai
5. Oliver, G. Navigating Uncertainty with AI. Pharmaceutical Executive, July 29, 2025, https://www.pharmexec.com/view/navigating-uncertainty-ai-building-trust-way-forward
6. HTA. “Building Momentum and Moving Forward Together: The Path Ahead for the Digital Transformation of Health Technology Assessment,” presentation at the Health Technology Assessment international (HTAi) Global Policy Forum, April 30, 2025, https://htai.org/wp-content/uploads/2025/05/HTAiGPF2025_WhitePaper.pdf
7. Florence, R.; et al. Reporting Guidelines for the Use of Large Language Models in Health Economics and Outcomes Research: An ISPOR Working Group Report. Value in Health, 2025 28 (11) pp. 1611-1625, https://www.sciencedirect.com/science/article/pii/S1098301525024556

About the author

Angeline Dhas is an AI Product Specialist for MadeAi.