Generative AI could help accelerate clinical development

By Andrew Bolt, principal, and Amy Cheung, principal, Deloitte Consulting LLP

In today’s challenging economic environment, leaders in life sciences are often faced with the difficulty of executing flexible and adaptable clinical trials that can serve the unique needs of various stakeholder groups in a cost-effective manner. This is not an easy task given that clinical trials are inherently labor intensive, complex, and regulated.

Novel digital technologies, automation tools, and patient-experience solutions have helped mitigate manual activities, and in turn, reduce overall cycle time and cost. However, these digitization tools have historically had only an incremental, not transformational, impact on clinical trials. Generative artificial intelligence (AI) might finally be the technology innovation that can be a transformative lever.

Notably, generative AI has the potential to slow rising costs and transform the landscape of clinical development by accelerating tasks across the clinical lifecycle. That could help bring a greater share of services back within the four walls of biopharma companies, while improving experiences for internal resources and patients alike, and ultimately contributing to more efficacious therapies.

Three processes that could be improved with generative AI

Generative AI is already transforming the way life sciences organizations decide which disease areas to invest in. It is also being used to identify targets, develop molecules, streamline and accelerate clinical trials, and submit findings for regulatory approval (see Can Life Sciences companies unlock the full value of GenAI?). Here is a look at the role generative AI might play in several aspects of a clinical trial:

1. Automate document generation activities to increase velocity: Generative AI could help biopharma organizations quickly draft and refine the documentation required to establish new test sites. The technology could use previous examples of clinical-trial protocols, site-contracting agreements, clinical report forms, and other key pieces of paperwork required to jumpstart clinical trials. Automating document generation could be a critical step in creating diversity within patient cohorts, allowing sites in under-served geographies to be established more quickly and with less effort.
2. Increase study retention by amplifying patient engagement: Attracting and retaining patients can be a major pain point for clinical trial sites. Generative AI can help facilitate patient recruitment by activating personalization at scale. It can also identify the doctors and ZIP codes of potential participants who could benefit from a new therapy and create customized outreach. The conversational nature of generative AI can function as the front-line in answering patient questions, sharing relevant information, and triaging the concerns that can cause patients to leave clinical trials early. Attrition can cost as much as $20,000 per patient. That expense gives organizations a significant incentive to explore all avenues for retaining clinical trial participants. 
3. Improve regulatory engagement with tailored submissions: To successfully deliver a submission to regulators, clinical teams typically must assemble immense dossiers that draw from all of their laboratory research and clinical development activities. Automated generative AI-enabled document tagging can make critical documents easier to locate. For example, intelligent or semantic search capabilities enabled by generative AI—wherein a search is based on meaning rather than individual key words—can enable much faster identification of relevant materials.

Three considerations to ensure the effective use of generative AI

Many of the life sciences leaders we talk to are excited by the potential of applying generative AI to clinical trials. They are also understandably concerned about managing risks to quality and employee experience. To that end, we provide the following considerations:

• Recognize the difference between a ‘task’ and a ‘job’: As generative AI use-cases grow, there are concerns that there will be less need for human workers. While this might be a valid concern, life sciences leaders should clearly differentiate between a task (i.e., an activity someone performs as part of their work) and a job, which typically has a far greater scope. Clinical development is made up of a variety of tasks that might be ripe for automation (e.g., repetitive, manual, and rule-based activities). The jobs that are attached to those tasks, however, remain critical.
• Anticipate a shift toward more specialized knowledge: As discrete tasks are increasingly shifted toward generative AI, organizations will likely need to stand-up the proper guardrails to ensure the integrity of the outputs. Generative AI is expected to increasingly drive content creation (e.g., outreach to trial participants, plain language summaries of clinical data) and, as a result, humans will likely need to validate those outputs.
• Be cautious of historical data: Clinical development has faced challenges related to creating diverse clinical trial cohorts. For that reason, leaders in R&D should be careful not to over-index on historical clinical data or they risk amplifying biases inherent in existing data sets. Moreover, ensuring that trustworthy AI frameworks and governance approaches are in place will mitigate potential for bias and unintended outcomes.

Combining GenAI with other digital platforms such as machine learning and predictive analytics could create an end-to-end business value stream—from clinical study startup through clinical study closeout.

Acknowledgments: Aditya Kudumala and Adam Israel