Driving adoption and innovation of AI to benefit clinical development

AI holds immense possibility for transforming clinical trials – a transformation that has already begun.

From drug discovery and planning through operational execution to post-marketing, AI can accelerate processes and enhance efficiencies to bring life-saving therapies to market faster.

Realising the benefits of AI across key areas, however, requires a methodical approach to developing mature and responsive solutions.

AI and its sub-disciplines – including machine learning and deep learning, large language models, natural language processing and generative AI – have their roles in transforming clinical development.

Our approach to defining the role of AI centres on the eight ethical principles identified by our AI committee and places our people and our expertise first.

In so doing, we can leverage advancements and optimisations enabled by AI to augment our human capabilities to guide accountable, fair and transparent intelligent processes.

The resulting solutions amplify the benefits of AI with mitigated risk to ultimately make clinical trials more accessible, make investigators’ jobs easier and improve patient journeys.

The wider goal is to deploy AI where it will realise the most value for clinical trials, sponsors and patients. However, pursuing AI solutions is a multi-tiered process.

For instance, at my organisation we have a dedicated AI centre of excellence whose aim is to responsibly forward the integration of AI to benefit clinical development across the life cycle, supported by a governance framework that includes a cross-functional AI committee.

This approach requires a robust strategy, testing, development and deployment of solutions that deliver strategic goals in transforming how we execute clinical trials.

All solutions we pursue undergo many rounds of discovery through structured processes to hypothesise, observe and iterate to ultimately determine how AI could solve a business problem and how it could integrate into tech and business processes.
As we consider AI capabilities across industry, they must first align with four strategic pillars that guide responsible AI adoption:

- Increased execution efficiency
- Improved quality of execution
- Improved predictive analytics capabilities and reduction in complexities
- Accelerated clinical trial execution timelines.

As the solution proceeds through development, it is important to weigh the benefits and the risks of the implementation.

Given the ever-evolving nature of AI and digital technologies, we must also consider the ability to iterate, maintain and adapt these solutions and machines to the changing landscape, with special regard to regulatory and security considerations.

The goal is to maximise innovation and minimise risk at every step. As we develop AI solutions, we must also invest in a robust and scalable privacy strategy that mitigates the AI-specific challenges and aligns with current and upcoming global regulation, as well as ethical obligations to all stakeholders.

There are multiple advanced AI solutions. In our experience, effective solutions are integrated across three main areas of intelligence that span a wide range of business and scientific applications.

Intelligent automation – AI can enhance and enable task automation to accelerate timelines, reduce personnel burden and allow for more effective resource allocation.

This involves imbedding intelligent automation processes into areas such as pharmacovigilance case triage, site invoice processing, RFP/RFI processing and machine translation.

Resource intelligence – Combining generative AI and machine learning-assists in the planning and forecasting of resources across the life cycle of a clinical trial can more closely associate the requirements of all stages of study execution.

This insight enables improvements to patient enrolment, detailed and accurate resource forecasting and assignments, protocol amendments and risk management.

Docu intelligence – The power of generative AI and large language models can parse data, capture knowledge, learn from it and then make predictions and generate responses based on those learnings.

AI can be used to accelerate and drive quality improvements in documentation-heavy processes including early site activation and contract negotiations, health authority responses and legal processes. It also ignites the following areas:

Integrated solutions – Several AI-driven solutions have emerged over the past few years, which are now integral to clinical trial operations, delivering tangible benefits along the way.

These solutions have been developed through close collaboration between the ICON AI Centre of Excellence and the relevant subject matter experts for each function.

Site selection – This allows identification of the right clinical trial sites the first time by leveraging human-identified nuances in protocol and parsing massive amounts of real-world data to evaluate connections and rank results for best-fit sites.

Benefits include accelerating trial delivery, reducing non-enrolment, reducing costs and potentially extending the period between drug approval and patent expiration.

Post-marketing predictions – This facility precisely predicts post-marketing requirements for FDA or EMA for a new product by analysing clinical trial data, drug mechanisms of action data and data from regulatory authorities.

We continually retrain the AI model to reflect the evolving regulatory landscape so sponsors can assess and plan for post-marketing requirements earlier in development.

Identifying KOLs – Bringing together millions of publications into one view and visually illustrates their interconnectivity by indication, identifying key opinion leaders (KOLs) better positioned to support clinical efforts in rare disease – previously a highly manual and tedious effort.

AE/CM reconciliation – A clinical data science solution using natural language processing, greatly reduces human effort in data review processes for reconciliation of adverse events and concomitant medications.

Clinical outcomes – Selecting and managing clinical outcomes assessments to automatically curate data, improve data accessibility and provide deeper insights into COA effectiveness and trends across all indications to devise patient-centred endpoint strategies

Ultimately, a responsible approach to AI and dedication to digital transformation facilitates AI adoption and empowers next-generation thinking by promoting cross-organisational AI literacy and internal networks of AI and innovation champions.

A concerted approach to people-centric, ethical AI implementation that follows a stringent multi-tiered process can enhance clinical development in multiple ways.

This is exemplified by the successful results of our mature solutions, as well as the multiple initiatives on the near horizon that will support improved compliance with auto-filing, resource identification and forecasting, and AI-powered operational metrics for improved business intelligence.

AI and its sub-disciplines are significant tools for transformation within healthcare and clinical development.

The key to realising the full benefits of AI’s myriad applications, however, is a thorough development process backed by strong governance and supported by a culture of innovation.

Gerard Quinn is Vice president of IT innovation and informatics at ICON. Go to iconplc.com