Acknowledgments: Thank you to members of the Clinical Feasibility Consortium for their contributions to this article.
Clinical trials continue to be fraught with delays and challenges that ultimately impede our ability to deliver critical medical breakthroughs. A significant contributing factor is slower-than-expected patient enrollment, typically taking 1.8 times longer than planned and affecting approximately 85% of clinical trials. This prolonged duration corresponds to a daily loss in sales ranging from $600,000 to $8 million for each day the trial is postponed. The predominant factors leading to these adverse trends primarily stem from difficulties in clinical trial planning and execution. Hurdles in patient enrollment arise from complex trial designs that are challenging to enroll and inefficient country and site selection, while challenges in trial operations result from operational inefficiencies, burgeoning trial complexity and difficulty maintaining patient retention and adherence.
Artificial intelligence (AI) presents significant opportunities for improving the planning and execution of clinical trials, promising to address key clinical feasibility challenges, including trial design, protocol optimization and country and site selection. Various forms of AI, including multimodal, generative and predictive, are being explored for their potential to enhance clinical trial processes.
While the potential benefits are immense, the integration of AI into clinical trials also comes with a set of unique challenges. These include data access and quality issues; the availability of requisite skill sets; the scalability of AI models; the need for robust change management to drive adoption and institutionalize new ways of working; and the persistent trust deficit and skepticism toward AI platforms and applications.
Earlier this year, the Clinical Feasibility Consortium assembled a team of 12 feasibility leaders from premier pharmaceutical organizations. Their mission was to discuss the practical aspects, hurdles and prospective solutions for incorporating AI into the processes of clinical trial planning that will enable successful downstream execution. Through these discussions, several key factors to enable widespread adoption of AI in the pharmaceutical sector were identified:
- External data access and quality: Healthcare data, a vital input for AI systems, often presents significant challenges. It can be hard to access, is frequently unstructured and contains a large proportion of missing data points. Both access and improving quality of this data will be key for successful use of AI.
- Internal data maturity: Internal data has significant potential for AI models as it is more readily available and accessible. However, progress is needed on data readiness, visibility and quality challenges.
- Data integration: The harmonization of data will greatly support adoption of AI, especially in use cases with high-data volume. This can be accomplished with IT support to seamlessly integrate multiple data sources and establish standardized ontologies across data sources.
- Skill set: The demand for professionals adept in data science and with an intricate understanding of the clinical development life cycle far outweighs the current supply, creating a significant skill gap. Current roles in business will need to evolve to incorporate the operation of AI models as well as the validation of AI outputs.
- Scalability: AI models often prove successful at the proof-of-concept stage but encounter difficulties when scaling them to address real-world requirements, which is a critical step for their widespread use. Scalability will need to be considered from the start.
- Change management: High AI adoption requires a cultural shift within organizations, from a model dominated by experience-based and leader-driven decision-making to one which prioritizes data-driven decision-making.
In addition to these factors, there are critical considerations related to the regulatory landscape governing the use of AI in clinical trials. The regulatory framework for AI in healthcare is currently underdeveloped and subject to change, and this uncertainty can pose a challenge to the integration of AI in clinical trial processes. There’s a pressing need for collaboration between pharma companies, AI experts and regulatory bodies to establish robust guidelines that balance innovation and patient safety.
As the pharma industry continues on this journey, it can glean insights from other sectors that have successfully integrated AI into their operations. The tech sector, for example, has extensive experience in handling unstructured and missing data and could provide guidance on how to manage similar issues within clinical trials. Additionally, experts from the finance and e-commerce sectors, where data science is deeply ingrained, could help address the current skill set shortage in pharma.
Another area where other industries might offer lessons is in managing the transition to a more data-driven culture. Behavioral scientists, change management consultants and experts from sectors that have undergone similar transitions can prove instrumental in guiding this shift within pharma organizations.
The successful integration of AI into clinical trial processes will require concerted efforts addressing challenges of data quality, skill sets, scalability and change management. Harnessing AI for streamlined clinical trial planning and operations has the potential to revolutionize the way clinical research is conducted, making it more efficient, cost-effective and patient-centric. AI technology can streamline and contextualize key processes and decisions, transforming our scope of work from generation to validation of insights. The Clinical Feasibility Consortium intends to further explore how AI can be used to enhance several aspects in the clinical feasibility space, including optimization of trial design and site selection, as well as consideration of practical ways to approach the challenges of trust and scalability.
Add insights to your inbox
We’ll send you content you’ll want to read – and put to use.