Transforming the clinical trials landscape with the brave application of AI

Conversations around AI can often either seem very futuristic, or centre around whether tools such as ChatGPT can be trusted to write an email to your boss.

What is less acknowledged is the way in which it has already begun to transform the drug development process.

From drug design to biomarker identification, many companies are leveraging the power of AI to accelerate medical breakthroughs and bring life-saving drugs and therapies to patients faster.

One area where AI has particular potential to catalyse significant progress is in clinical trials. Challenges such as patient identification and recruitment, trial design and patient diversity could be tackled with AI solutions at a much faster rate.

But with this new era of AI-driven technologies comes questions around how to deploy it most effectively, and how it should be regulated to keep patients, and their data, safe.

AI has the potential to transform the entire clinical trial life cycle, particularly in site selection and performance, study planning and design, and optimising patient outcomes.

Advanced AI models can be instrumental in site selection, from forecasting trial enrolment at different sites across therapeutic areas to tracking how sites are likely to perform over time.

This means that companies can select sites in a way that maximises their success, accelerating the time that it takes to get a trial up and running.

Today we can also predict, using AI tools, which patients are more likely to enrol in trials, suffer adverse events or expect a positive treatment response. This in turn can lead to better patient selection, leading to fewer drop-outs.

When it comes to study planning and design, AI makes it easier to run trials as efficiently as possible by allowing pharma companies to test various scenarios in a simulation before confirming a trial design.

Generative AI can be used to model alternative trial scenarios and designs, create optimal protocols and predict obstacles before they arise, resulting in faster, more successful clinical trial programmes.

From the past 15 years of AI innovation in the industry, we have been able to develop new solutions with compelling use cases, leaving no doubt around their potential impact.

Examples of areas where AI has supported the drug development process in just the last year include accelerating trial enrolment for phase 3 trials by more than six weeks through better prediction of site performance and data-driven site selection; providing clear evidence that a new medication improves survival in pancreatic cancer using synthetic control arms (where historical trial data is used to create a placebo study arm); and early completion of a trial designed through predictive modelling due to the ‘overwhelming benefit’ of the study drug.

In simple terms, AI systems work by combining an algorithm – a series of instructions – with large amounts of data, that allows the system to learn from the data and recognise patterns in it.

However, the complexity of many AI systems makes explaining the process of AI decision-making very difficult.

This can pose a significant barrier to the adoption of these solutions – especially in healthcare & life science – as users are rightfully less willing to trust a decision-making system whose inner workings are not understandable.

As a result, a lot of the industry is in ‘watch and wait’ mode – companies want to see the impact of using AI in the clinical trial process before they commit to using it themselves.

Only by understanding how AI tools work and how they can support us, can we maximise their benefit.

Consequently, designing models that are explainable and approachable is a critical step in developing AI solutions that are trustworthy, particularly in an industry such as healthcare and pharma.

It is therefore more important than ever that we are kept ‘in the loop’ when it comes to how AI-driven technology operates. To trust these systems, we can’t become naïve to how they are generating the solutions they are providing.

An example of everyday use of an AI model is a navigation system such as Google Maps. We trust these systems to generate a route, ETA and turn-by-turn instructions for all of our journeys, even those we are familiar with.

If the system starts to take you on a different, unfamiliar route, it will inform you that this is due to road works, or an accident, or heavy traffic.

This is an example of how designing AI tools in a way that keeps us ‘in the loop’ in the decision-making process ensures we are able to make an informed decision on whether to accept and act on the recommendation or solution AI has provided.

In the life sciences industry similar design principles need to be adopted when it comes to AI features. Companies should always be transparent about where AI is being used, what data it is using to make inferences and how that data has been processed by their algorithms to suggests a trial site or scientific conclusion.

Developing and using AI in healthcare is fundamentally about the use of data. One concern that has emerged with the implementation of AI in clinical trials is data privacy and the use of patient data.

Patients’ involvement in the decisions of how their data may be used is key to responsible AI use. It is vital that patients have autonomy over their data and can make an informed choice about how their data is used.

Promisingly, many patients appear to recognise the value of their data in supporting the development of new medicines.

When surveyed, 93% of clinical trial participants believed their data should be shared with university scientists to accelerate medical breakthroughs.

Regulations that ensure data security without stifling innovation are crucial as AI capabilities expand. To date, the industry has worked closely with regulators to provide assurances to all stakeholders including, perhaps most importantly, patients, to demonstrate our ability to guarantee data privacy.

Trust is a vital part of patient engagement, and being open and transparent with patients ensures that we maintain their confidence.

AI is already playing a transformative role in the clinical trial process, but its impact is far from over. There is no doubt that all of the innovation that AI has spawned can have a significant impact on healthcare and drug development – and most of that impact is still to come.

For example, we could use AI to go beyond simply predicting which patients will enrol in a trial to proactively identify patients most likely to benefit from therapies.

Combining AI with historic clinical trial data and patient medical records could reveal the gap between patients being enrolled in trials and those who are receiving treatment.

This would allow us to directly engage with target patient populations rather than relying on patients or clinicians to be aware of upcoming trials.

Some of the biggest challenges in clinical trials have the potential to be solved through AI technology.

From diversifying clinical trials, to efficiently recruiting patients most likely to benefit from treatment and predicting those likely to experience adverse effects, AI is set to deliver safer, more efficient clinical trials. By innovating today, we can meet the needs of patients today and in the future.

Jacob Aptekar, Vice President of Platform AI & Data Science at Medidata. Go to medidata.com