The NICE AI Methods for Evidence Framework: What It Means for HTA and Evidence Generation


November 17, 2026

Artificial intelligence (AI) is rapidly transforming how evidence is generated, synthesized, and communicated across the healthcare ecosystem. From automating literature screening and extracting study data to supporting health economic modelling and drafting technical reports, AI technologies are increasingly being embedded across the HTA value chain.

Recognizing both the opportunities and risks associated with these technologies, NICE has published its new AI Methods for Evidence Framework, providing a comprehensive approach to the responsible use of AI in evidence activities that may inform NICE guidance. The framework aims to improve transparency, quality, consistency, and trust in AI-enabled evidence generation while supporting continued innovation.

 

Why the framework matters

Over the past few years, AI adoption has accelerated dramatically across pharmaceutical companies, consultancies, contract research organizations, and HTA agencies. Large language models (LLMs), machine learning algorithms, and generative AI tools are increasingly being used to support systematic reviews, real-world evidence studies, evidence synthesis, economic evaluations, and technical writing.

While AI offers significant potential to improve efficiency and reduce resource requirements, it also introduces important concerns regarding transparency, reproducibility, bias, accountability, and scientific validity. Decision-makers need confidence that AI-assisted evidence remains robust and trustworthy.

The NICE framework seeks to address these concerns by providing practical guidance on how AI should be used, governed, validated, and reported. Importantly, the framework does not establish minimum evidence standards or determine whether evidence is acceptable for decision-making. Instead, it focuses on ensuring that AI-supported methods are applied and documented appropriately. It is intended to be a “living” framework to be updated when needed. For example, “Agentic AI” has not been covered in this framework but will be in a future update.

 

A principles-based framework

Rather than prescribing specific technologies, software platforms, or models, NICE adopts a principles-based approach that can evolve alongside rapidly changing AI capabilities.

The framework is built around four categories of principles that together support responsible AI adoption.

 

Foundational principles: Using AI responsibly

The foundational principles establish that responsibility for evidence remains with the organization generating or submitting it, regardless of whether AI has been used.

Key expectations include:

  • Maintaining human accountability for scientific decisions.
  • Considering potential bias and fairness issues.
  • Ensuring transparency regarding AI use.
  • Protecting confidentiality and data security.
  • Using AI in ways that support trustworthy evidence generation.

These principles reinforce that AI should augment, rather than replace, expert judgement.

 

Governance principles: Using AI appropriately

The framework emphasizes the need for proportionate governance based on the intended use and risk level of the AI application.

Organizations are expected to:

  • Assess whether AI is appropriate for a specific task.
  • Understand the limitations of the chosen technology.
  • Implement appropriate oversight and controls.
  • Establish clear accountability processes.

This reflects a risk-based approach in which higher-impact applications require greater scrutiny and governance.

 

Assurance principles: Using AI safely

Assurance focuses on demonstrating that AI-supported activities are reliable, accurate, and fit for purpose.

Users should be able to:

  • Verify outputs generated by AI systems.
  • Evaluate performance and accuracy.
  • Assess uncertainty and limitations.
  • Demonstrate that appropriate quality assurance processes have been applied.

For HTA practitioners, these principles are particularly important when AI is used in evidence synthesis, comparative effectiveness analyses, or economic modelling activities that may directly influence reimbursement decisions.

 

Trustworthiness principles: Ensuring transparency

Transparency is a central theme throughout the framework.

NICE expects users to clearly report:

  • Which AI tools were used.
  • Where AI was used within the workflow.
  • What outputs were generated.
  • What validation or quality checks were performed.
  • How human oversight was applied.

The objective is to enable reviewers and decision-makers to understand the influence of AI on the final evidence package and judge its appropriateness accordingly. An AI Declaration (AID) Checklist is also provided in the framework.

 

Recognizing different levels of AI involvement

A notable strength of the framework is its recognition that AI use exists on a spectrum rather than as a simple yes-or-no proposition.

Some applications involve AI providing limited assistance while humans retain full control over decisions. Others involve semi-automated workflows in which AI performs substantial portions of the work under human supervision.

By acknowledging these different levels of automation, NICE encourages a proportionate approach to governance and reporting based on the degree of AI influence and the potential impact on evidence quality.

 

AI across the evidence lifecycle

The framework applies across the major evidence activities that support HTA.

 

Primary Evidence Generation

AI may be used during study planning, participant identification, data collection, data preparation, statistical analysis, and interpretation of findings.

NICE emphasizes that researchers must understand how AI contributes to study conduct and ensure that scientific validity is maintained throughout the research process.

 

Evidence Synthesis

Evidence synthesis is one of the most active areas for AI adoption.

Potential applications include:

  • Literature searching
  • Study screening
  • Data extraction
  • Evidence summarization
  • Risk-of-bias support activities
  • Report drafting

The framework recognizes significant efficiency gains but stresses the importance of human verification and transparent documentation. For organizations conducting systematic literature reviews or network meta-analyses, this section is likely to have immediate practical relevance.

 

Health Economic Evaluation

The framework also considers AI-supported health economic modelling.

Potential applications include:

  • Identification of model inputs
  • Data preparation and analysis
  • Programming support
  • Scenario generation
  • Technical documentation

Given the importance of economic models in HTA decision making, NICE highlights the need for transparency, auditability, and clear understanding of how AI contributes to modelling outputs.

 

Technical Documentation

Generative AI tools are increasingly being used to draft reports, evidence summaries, and other technical documents.

While NICE acknowledges the productivity benefits of these tools, it makes clear that human review remains essential and that organizations retain responsibility for all final content.

 

Reporting will become increasingly important

Perhaps the most significant practical implication of the framework is the growing expectation that AI use should be actively documented and disclosed.

Historically, many evidence-generation activities have incorporated automation without systematically reporting it. The NICE framework signals a shift towards greater transparency.

For manufacturers and evidence-generation teams, this means documenting:

  • The purpose of AI use.
  • The specific tools employed.
  • The tasks completed using AI.
  • Validation procedures.
  • Human review processes.
  • Known limitations and mitigation strategies.

Robust AI reporting is likely to become a routine component of evidence submissions, much like reporting statistical methods or evidence synthesis procedures today.

 

What this means for the future of HTA

The publication of the NICE framework represents an important milestone in the evolution of AI governance within HTA. It signals that AI is no longer viewed as an experimental technology operating at the margins of evidence generation. Instead, it is becoming an established methodological tool that requires clear standards for transparency, oversight, and accountability.

For pharmaceutical companies, HEOR teams, evidence synthesis groups, and HTA consultancies, the message is clear: organizations should begin developing formal AI governance frameworks, validation processes, and reporting practices now.

Ultimately, the framework does not position AI as a replacement for scientific expertise. Rather, it promotes a model in which AI augments human capability while maintaining the rigor, transparency, and reproducibility essential for evidence-informed healthcare decision-making. As AI adoption continues to accelerate, the NICE framework provides an important foundation for ensuring that innovation is accompanied by trust.

Contact Us!
Subscribe to our newsletter

Dalia Dawoud

Research Principal, HTA Policy and Strategy

Dalia Dawoud is Research Principal, HTA Policy and Strategy at Cytel. Prof. Dawoud holds an MSc in economic evaluation in healthcare (City University London) and a PhD in pharmaceutical policy and economics (King’s College London) and has practiced as health economist and researcher for over 20 years. Her work is largely focused on the application of health economics and outcomes research (HEOR) in HTA and clinical guideline development. Prior to joining Cytel Inc., she worked at leading organizations including NICE, where was the founding Associate Director of the newly established NICE HTA Innovation Laboratory (HTA Lab) with projects in the areas of RWE, HTA methods, and health economics, focusing on managed access, evaluating emerging therapies, such as dementia treatments and multi-indication diagnostics, and the use of AI in economic modelling. She also led a portfolio of HORIZON Europe projects such as HTx, SUSTAIN HTA, and EDiHTA, with combined funding of over 5 million euros. Dalia also worked at the Royal College of Physicians – London and King’s College London among other academic institutions.

She is widely published in the area of HEOR, HTA, and pharmacy policy and serves as Associate Editor of the ISPOR journal Value in Health, was recently elected as Director of the HTAi Board (2026–2029), and served as Director on ISPOR Board of Directors (2023–2026). She is also a member of ISPOR AI Working Group, Living HTA Working Group, and ISPOR Task Force on using GenAI in systematic reviews. She also holds Professor position at the Faculty of Pharmacy, Cairo University.

Read full employee bio

Claim your free 30-minute strategy session

Book a free, no-obligation strategy session with a Cytel expert to get advice on how to improve your drug’s probability of success and plot a clearer route to market.

glow-ring
glow-ring-second