Generative AI in Clinical Development

Artificial intelligence (AI) has become quite the buzzword over the past couple of years, and for good reason. We’ve seen significant advances both in how compelling AI can be—from writing a bedtime story to composing a tweet—and the impact it can deliver in almost every sector, from driving cars to providing medical advice. While the rate of AI development and adoption today is unprecedented, it’s important to remember that AI has been a vital and reliable technology in the life sciences for some time.

Data is the lifeblood of drug discovery and clinical development, but understanding that data and applying the insights it provides to deliver better, safer medicines to patients sooner, has been the domain of advanced analytics and AI for years.

AI is adept at detecting patterns and anomalies in data—like the subtle or complex patterns in data that may represent a genetic predisposition that could be helped or hurt by a drug, or anomalies in how data is being collected that may be a sign of a malfunctioning system or human error. AI in the form of statistical learning can also leverage real patient data from past trials to create synthetic control arms to be utilized for trials where including contemporaneous control groups would be difficult, essentially ensuring that all participants are receiving the trial drug.

But this kind of analytic or simulative AI isn’t what’s been grabbing all the headlines, the real buzz is around generative AI and the far more interactive and human-like solutions it can deliver. It’s therefore unsurprising that the implementation of gen-AI solutions in clinical development is an attractive proposition.

A recent report from Everest Group—The Promise of Generative AI in Clinical Development—states that 30% of pharma enterprises consider gen-AI to be a core focus and the primary area of technology investment for the next 12 months, with 34% already implementing proof of concept solutions.

The Benefits of Gen-AI?

So, what can gen-AI bring to clinical development over and above what traditional AI solutions are already providing? According to Everest Group , the potential benefits are many, from cost efficiencies and faster time-to-market, to enhanced stakeholder experience and satisfaction—gen-AI can help improve the clinical trials process across the board, while also enhancing patient experience.

One of the core benefits of gen-AI is its ability to explain potentially complex documents in a manner that is understandable to almost any audience—essentially providing each patient with a personalized AI concierge to help them fully understand the trial protocol before enrolling. Everest Group suggests that gen-AI can 'improve the success rates of clinical trials with well designed, patient-centric protocols,' and patient centricity should be a core focus for any clinical trial.

A well-implemented gen-AI solution can streamline the clinical development process, from candidate recruitment, through to final data analysis, but its ability to monitor and evolve workflows and processes throughout a trial could be invaluable. Whether that be uncovering patterns and correlations to drive informed decisions, or identifying and automating repetitive and labor-intensive tasks, gen-AI can constantly learn, extract insight, and drive efficiency.

The Data Dilemma

While there are myriad potential benefits to utilizing gen-AI within the clinical development process, there are also many challenges to consider and overcome before the impact of gen-AI can truly be felt.

One of the major challenges for generative AI is data quality, especially the data on which models are trained. Every effort should be made to avoid influences like political agendas, personal bias or simple inaccuracy—in life sciences, even when implementing pre-trained LLMs, checks and safeguards should always be implemented to account for and mitigate any such issues.?

However, as Everest Group points out, even with accurate data, unintentional bias can still be an issue. For example, if there’s a lack of patient diversity in the data pool, an AI model could make decisions affecting minority demographics that aren’t based on that group’s needs. It’s therefore vital to ensure that any gen-AI platform is trained on data that is trusted, accurate, and representative for the task at hand. This is even more important for clinical development focusing on rare diseases, or precision medicine.

“We have made significant investments to ensure that the quality of our data is exceptional. And by maintaining this high standard, the data’s value is greatly enhanced.” — Jia Chen, PhD , Sr. Director, Medidata AI

Regulatory Considerations

Another challenge is regulatory compliance, with some regulatory bodies still unsure of how to regulate the use of AI technology in areas such as drug discovery and clinical development. We are, however, seeing an appetite among major regulators to address and expedite the need for regulatory frameworks around AI, with the Food and Drug Administration ( FDA ) currently considering such a framework. Jacob Aptekar , VP, Trial Design Solutions, Medidata AI, said that the FDA’s recognition of AI as a potential alternative evidential source “could significantly accelerate the clinical development process, by focusing clinical development in areas of true scientific uncertainty, while reducing the need for confirmatory data on treatments or patient groups that have been oversampled in the recent past.”

Data privacy is a key concern around gen-AI in the wild, where public cloud infrastructure and often free usage models can mean that any data gathered during use, is then open to reuse. Clearly that’s not viable when it comes to personal medical data.

“I think many patients are genuinely willing to share their data in the hope of accelerating the discovery of cures,” says Chen. “But they are concerned about privacy.” This means that any gen-AI solution that touches patient data must come with provable data security and privacy guarantees in place. And this, in turn, raises another concern: cost.

The Future of AI in Clinical Development

To guarantee data security and privacy, any enterprise in the drug discovery and clinical development space looking to leverage gen-AI will need to either develop significant expertise in AI-specific information security processes or work with vendors who have invested in developing AI systems that comply with regional regulations as well as emerging industrial best practices.

Medidata has built a robust reputation for creating and advancing AI platforms that integrate seamlessly into clinical development ecosystems. The Medidata platform is already analyzing real-time data and comparing results and insights with a vast pool of historical data, pulled from over 33,000 successful clinical trials. While predictive modeling can simulate a trial, build a synthetic control arm, or develop virtual twins that can accelerate precision medicine research.?

Generative AI is the logical next step for Medidata, continuing a pattern of cutting-edge technological innovation in clinical trials, but according to Chen it’s also not a step the company is taking lightly:?

“Despite our extensive experience with AI, we’re adopting a prudent approach, when it comes to gen-AI, to ensure that privacy, security, and trustworthiness are upheld.”

“We have made substantial investments in AI governance, ensuring that our workflows incorporate a 'human in the loop' framework,” continues Chen. “Additionally, transparency is paramount. The critical question we continuously address is: How can we responsibly utilize this technology?”?

By innovating responsibly and keeping humans in the loop, organizations can harness the strengths of AI while mitigating its potential risks, ensuring a balanced, effective, and sustainable approach to technology deployments.?

While there may be challenges to overcome, the potential benefits that generative AI can bring to the clinical development sector are huge, and consequently, the potential benefits to patients could be even greater.

Only Medidata combines innovative technology, AI, advanced analytics, and the industry’s largest historical patient-level clinical trial data set to transform how clinical research is performed. Learn more about how Medidata is harnessing the power of AI in clinical trials.

This article was originally published on the Medidata Clinical Minds blog.