The FDA positions credibility as an indispensable condition for utilising AI in pharmaceutical regulatory decision-making, within a risk-proportionate framework that demands traceability, transparency and validation throughout the entire model lifecycle. This approach has been synthesised and disseminated by Guillem López, CSV & QA Manager at OYTEC, author of the analysis “Credibilidad de la IA farmacéutica: Claves del borrador de la FDA” published in number 18th of FARMABIOTEC magazine.
Regulatory requirements
The draft guidance “Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products”, issued in January 2025, establishes that a credibility framework must be implemented for models, proportionate to the risk of this process, taking into account both the influence the model will have on decision-making and the consequences of making an incorrect decision (Model influence vs. Decision consequence).
The guidance proposes the establishment of a plan to determine credibility and a corresponding report demonstrating the model’s suitability for the context of use.
Key elements to be included within the plan must encompass the definition of the question of interest, the context of use, risk assurance and a technical description of how the model has been developed (including data, training, evaluation metrics, amongst others). Once established, the appropriateness of the results for the context of use must be properly executed and documented.
Like any GMP-critical system, once in the production phase, planning for maintenance and periodic performance review must not be omitted. However, in the case of AI models, this is particularly important due to the fact that they are systems which learn from our process, and this is constantly evolving (the Datadrift phenomenon). It is crucial to ensure that this suitability for the context of use is maintained through monitoring and retraining.
OYTEC’s and its author’s contribution
From OYTEC, Guillem López emphasises the need to precisely identify the Question of Interest (QoI) and context of use in order to establish a Credibility Assessment Plan and a Credibility Assessment Report that connect technical evidence with plant requirements, integrating GxP regulations and reference frameworks such as GAMP5 Ed.2 and ISPE to ensure reproducibility and change control.
How OYTEC implements IA
OYTEC implements AI solutions in GxP environments alongside technology partners such as Aizon, aligning credibility plans and reports with the FDA framework.
To explore further, please consult Guillem López’s article in FARMABIOTEC (n18, pages 66–71) and contact our experts.
Fast & Furious eBR Implementation
YES, that's right, a Fast & Furious eBR Implementation is possible. It is no longer a utopia to carry out an Electronic Batch Record implementation in a fast and massive way for any manufacturing plant. Regardless of the manufacturing area: bulk, weighing, or...
EMA and FDA: The 10 Principles of Pharmaceutical AI
The European Union and the United States published the Guiding Principles for Good AI Practices in Drug Development on January 14, 2026, as part of a strategic effort to harmonize regulatory criteria. The EMA-FDA collaborative work initiative has the mission of...
Kern Pharma Chooses OYTEC for Terrassa Site Digitalization
KERN PHARMA, one of the leading pharmaceutical groups in Spain and a market leader in generic drug manufacturing, has partnered with OYTEC to digitalize its industrial operations. As part of its Digital Transformation strategy, KERN PHARMA aims to optimize and enhance...
