Articolo
The brave new world of predictive AI: Redefining the benefit-risk journey
-
Timm Pauli
-
Shilpa Patil, PhD
How predictive AI is helping to shake up benefit-risk assessments, rewriting past regulatory submission processes
Regulatory intelligence in the benefit-risk decision-making process is undergoing a digital revolution thanks to artificial intelligence (AI). Many regulatory submissions fail – or are delayed – not because of a lack of data, but because companies struggle to anticipate how regulators will interpret it. Predictive AI is changing that equation.
In the brave new world of AI, past practices to manage the regulatory submission process are being turned on their head. Instead of having regulatory teams hunt for the guideline, guess what is relevant, and try to find the gaps before regulators do, predictive AI does the heavy lifting. It seeks out the guideline, provides the relevant expectation, flags potential gaps before submission, and highlights health authority precedents.
When benefit-risk data, unstructured regulatory narratives, and precedent-based intelligence are integrated into a single framework, there is real potential to derisk the submission process.
In the brave new world of AI, past practices to manage the regulatory submission process are being turned on their head. Instead of having regulatory teams hunt for the guideline, guess what is relevant, and try to find the gaps before regulators do, predictive AI does the heavy lifting. It seeks out the guideline, provides the relevant expectation, flags potential gaps before submission, and highlights health authority precedents.
When benefit-risk data, unstructured regulatory narratives, and precedent-based intelligence are integrated into a single framework, there is real potential to derisk the submission process.
Shaking up benefit-risk assessments with AI
Central to any submission is the benefit-risk assessment. Regulators globally expect companies to balance the benefit of the product with the risk to the patient before it can be considered for approval or for continued use by patients.
The US Food and Drug Administration clearly laid out its benefit-risk expectations in its 2017 framework and issued guidance for new drugs and biologics in 2023.i Similarly, the European Medicines Agency has laid out its methods and attempts to quantify benefit-risk.ii
The challenge for industry has been to understand the benefit-risk factors and decision rationale of the various health authorities before submitting a new drug application (NDA), Biologics License Application (BLA), or, in Europe, a marketing authorization application (MAA). Companies have relied on good regulatory intelligence to understand how health authorities apply guidelines to benefit-risk decisions and how these decisions may differ globally depending on health authority priorities.
The US Food and Drug Administration clearly laid out its benefit-risk expectations in its 2017 framework and issued guidance for new drugs and biologics in 2023.i Similarly, the European Medicines Agency has laid out its methods and attempts to quantify benefit-risk.ii
The challenge for industry has been to understand the benefit-risk factors and decision rationale of the various health authorities before submitting a new drug application (NDA), Biologics License Application (BLA), or, in Europe, a marketing authorization application (MAA). Companies have relied on good regulatory intelligence to understand how health authorities apply guidelines to benefit-risk decisions and how these decisions may differ globally depending on health authority priorities.
Putting AI-enabled regulatory assessment to the test
As an example, Cencora’s pharmacovigilance, statistics and analytics, and digital innovators have been putting an AI-enabled benefit-risk assessment database to the test. The objective was to demonstrate the value of AI-supported gap identification and precedence analysis in the preparation of well-informed applications.
The digital and analytics team demonstrated that AI can be used to identify submission gaps and to systematically quantify and benchmark benefit-risk positions against past regulatory decisions to provide a deeper evidence-based foundation for approval strategies.
The digital and analytics team demonstrated that AI can be used to identify submission gaps and to systematically quantify and benchmark benefit-risk positions against past regulatory decisions to provide a deeper evidence-based foundation for approval strategies.
For example, for an immunotherapy application, comparing the benefit-risk factors – such as indication severity and prevalence, unmet need for the treatment, strength and consistency of the efficacy data, and risk profile and risk mitigation plans – from previous immunotherapy approvals in a similar therapeutic category can significantly contribute to establishing the relevance of the application.
A net benefit score – based on severity of condition, current availability of therapies, and efficacy data of the NDA/BLA, combined with adverse events and proposed risk mitigation plans – was used to support decision-making. The AI-assisted analysis is further supported by a quantified precedence database to assess gaps and deviations in a submission, which could help to predict the approval likelihood of future applications.
A human-AI collaboration
It is important to emphasize that while the system is automated and AI-supported, it is underpinned by human-in-the-loop validations. Given the complexity of benefit-risk assessment, any digitally led application analysis will require a combination of deep subject matter expertise – regulatory, safety, quality, etc. – and mastery of AI tools. The real value lies in human-AI collaboration – where AI scales pattern recognition and precedent analysis, while human experts apply judgment and context.
The point of predictive AI tools is to support well-informed regulatory applications by replacing laborious human-led relevant precedence research with an intelligent assistance system that can quickly and comprehensively review the information and predict potential gaps in the submissions. A tailored AI tool can not only provide insights through guidances, past reviewer queries and decisions, complete response letters (CRLs) and other public domain information, but also build a stronger submission by drawing on the company’s proprietary product information – restricted to ensure no cross-company access.
In future, this AI-human interchange is akin to having an AI reviewer sitting alongside subject matter experts as they prepare the dossier strategy and content.
By leading from an AI-generated regulatory intelligence perspective, companies can start to reduce risk of a CRL or non-approval of their submission. They can accelerate the approval process by reducing the likelihood of questions from the health authorities, and they can improve the important benefit-risk assessment process early on to help them make those crucial go/no-go decisions.
The point of predictive AI tools is to support well-informed regulatory applications by replacing laborious human-led relevant precedence research with an intelligent assistance system that can quickly and comprehensively review the information and predict potential gaps in the submissions. A tailored AI tool can not only provide insights through guidances, past reviewer queries and decisions, complete response letters (CRLs) and other public domain information, but also build a stronger submission by drawing on the company’s proprietary product information – restricted to ensure no cross-company access.
In future, this AI-human interchange is akin to having an AI reviewer sitting alongside subject matter experts as they prepare the dossier strategy and content.
By leading from an AI-generated regulatory intelligence perspective, companies can start to reduce risk of a CRL or non-approval of their submission. They can accelerate the approval process by reducing the likelihood of questions from the health authorities, and they can improve the important benefit-risk assessment process early on to help them make those crucial go/no-go decisions.
About the authors:
Timm Pauli is VP, Digital Solution Lead at Cencora, supporting the Regulatory, Compliance and Pharmacovigilance business group with the right digital tooling. He has around 25 years of experience in various functions within pharma R&D, including regulatory affairs and regulatory operations, pharmacovigilance, and clinical data management, and statistics.
Shilpa Patil, PhD, is a Senior Scientist in the digital innovation group at Cencora. She has more than eight years of experience in the field of translational research and healthcare, specializing in drug discovery, biomarker research, and AI-assisted personalized medicine in oncology. Shilpa and her team are working toward defining and building AI-based solutions that aim at adding value and efficiency to the regulatory applications and consultancy services offered.
Shilpa Patil, PhD, is a Senior Scientist in the digital innovation group at Cencora. She has more than eight years of experience in the field of translational research and healthcare, specializing in drug discovery, biomarker research, and AI-assisted personalized medicine in oncology. Shilpa and her team are working toward defining and building AI-based solutions that aim at adding value and efficiency to the regulatory applications and consultancy services offered.
Sources:
i Benefit-Risk Assessment for New Drug and Biological Products Guidance for Industry, FDA, 2023. https://www.fda.gov/media/152544/download
ii Benefit-risk methodology project, EMA, 2010. https://www.ema.europa.eu/en/documents/report/benefit-risk-methodology-project-work-package-2-report-applicability-current-tools-and-processes-regulatory-benefit-risk-assessment_en.pdf
i Benefit-Risk Assessment for New Drug and Biological Products Guidance for Industry, FDA, 2023. https://www.fda.gov/media/152544/download
ii Benefit-risk methodology project, EMA, 2010. https://www.ema.europa.eu/en/documents/report/benefit-risk-methodology-project-work-package-2-report-applicability-current-tools-and-processes-regulatory-benefit-risk-assessment_en.pdf
Dichiarazione di non responsabilità:
Le informazioni fornite in questo articolo non costituiscono una consulenza legale. Cencora, Inc. incoraggia vivamente i lettori a rivedere le informazioni disponibili relative agli argomenti discussi e a fare affidamento sulla propria esperienza e competenza nel prendere decisioni correlate.
