NICE publishes methods framework for AI use in evidence generation and assessment

NICE has announced a new framework for using AI in evidence activities at ISPOR Europe 2026. Covering primary evidence generation, evidence synthesis, health economic evaluation and technical documentation, it sets out principles and best practice guidance for those developing, submitting and reviewing evidence used to inform NICE guidance.

The Baseline

• Announced at ISPOR Europe 2026, NICE’s AI methods framework supersedes its earlier statement of intent and position statement.
• Ten core principles address the foundations, governance, assurance and trustworthiness of AI use in evidence activities.
• Developed with stakeholders, the living framework will be updated regularly. It does not appraise individual AI tools or cover the assessment of AI-powered health technologies.

The National Institute for Health and Care Excellence (NICE) has announced the availability of its new AI methods for evidence framework at ISPOR Europe 2026. Published on November 9, 2026, the 90-page framework supersedes and builds on NICE’s statement of intent for AI (ECD12) and its position statement on the use of AI in evidence generation (ECD11).

Presenting the framework, Raphael Sonabend-Friend, Associate Director – AI Methods and Safety at NICE, highlighted the central role of evidence in cost-effectiveness evaluations and the development of recommendations for healthcare. He outlined the potential for AI to change these activities, while emphasizing that adoption must be accompanied by safeguards.

“AI has the potential to radically change that: the development, the synthesis, and the review. But it has to be done safely.”

The framework recognizes the potential of AI to support evidence activities across NICE’s programs, including health technology assessment (HTA) and guideline development. It focuses on methods that are beginning to influence these activities, or are likely to do so, combining principles applicable across uses with additional considerations for specific applications.

Ten core principles underpin the guidance, organized into four groups:

• Foundational principles: Data compliance and responsible use
• Governance principles: Accountability and human oversight
• Assurance principles: Validation, fairness and equity, reproducibility and AI lifecycle management
• Trustworthiness principles: Explainability and transparency

These principles apply irrespective of the particular AI tools or models used. NICE emphasizes that no AI system should be assumed to support the principles or use cases outlined in the framework. The publication does not review or appraise individual tools, and any references to models or providers are illustrative rather than endorsements.

Alongside these overarching principles, dedicated sections address AI methods for primary evidence generation, evidence synthesis, health economic evaluation and technical documentation. This structure brings together common expectations for AI use with considerations relevant to each evidence activity.

The framework is primarily intended for those who develop, submit, review or use evidence to inform NICE guidance. It may also support patients, healthcare professionals and organizations that generate or assess evidence for decision-making, including other HTA bodies and NICE’s system partners.

Speaking during the session, Mitch Higashi (ISPOR) described the framework as a starting point for navigating AI adoption and highlighted the importance of retaining flexibility as the field develops.

“Hopefully this guidance, this framework, will be flexible enough for us to keep moving up on this AI adoption curve.”

The publication draws a distinction between using AI to support evidence activities and evaluating health technology products that incorporate AI. Developing or assessing those products falls outside its scope, and developers should not interpret the framework as guidance for their evaluation. Equally, standards and requirements for AI-powered health technologies should not automatically be assumed to apply to the evidence activities addressed here.

Co-developed with a wide range of stakeholders, the framework seeks to establish a common standard for principles and guidance on AI use in HTA and guideline development. NICE describes it as a living framework, with regular updates planned to maintain its relevance as AI methods and applications evolve.