Tested, tried and true: Why digital biomarkers are the AI & ML strategy pharma should prioritise

Rather than chasing generative AI, pharma executives can focus their AI & machine learning (ML) strategies for drug development on investing in and deploying connected sensors to reduce costs & improve patient outcomes.

The topic of generative AI has taken the world, pharma included, by storm — however, the field is relatively new and untested in the realm of pharmaceutical research, drug discovery, and drug development. Despite this, many pharma companies are leaning into these unknowns across the R&D lifecycle instead of doubling down on existing AI & ML-powered innovations that have proven effective and reliable repeatedly in drug development: digital biomarkers.

In August 2023, the European Medicines Agency formally qualified the first DHT-derived digital primary endpoint with a new stride-velocity measure for drug development in Duchenne Muscular Dystrophy as an alternative to traditional walking tests. The benefits of the digital endpoint — continuous monitoring over long periods of time, reduced reliance on patient motivation, and more — drive increased likelihood of approval beyond the capabilities of the 6-minute walk test. There is over a decade of validated research on deployment of digital biomarkers in life sciences, with proven return on investment to reduce trial cost and de-risk trial success, and these efforts are getting more attention and investment than ever before.

We’re seeing support for this innovation from regulatory bodies within the U.S. as well. This past summer, the FDA published a paper on AI & ML for Drug Development, with case studies from discovery through to development, commercial, and manufacturing. While generative AI systems like ChatGPT have recently gotten the lion’s share of attention, the FDA chose to focus its paper largely on biomarkers measured by digital health technologies (DHTs), such as connected sensors, which nearly all already use AI & ML to analyze and interpret data.

Regulatory support goes beyond drug discovery — the new FDA paper outlines six opportunity areas for leveraging AI & ML for drug development, several of which rely heavily on DHTs. These technologies can help improve clinical trial quality by supporting sensor-derived data collection and analysis, as well as better detection of drug safety and efficacy by monitoring patient symptoms in real-time.

The use of DHTs has the potential to improve trial diversity, an especially important consideration in the wake of the new regulatory mandates and legislation from the DEPICT Act and FDORA which are requiring diversity action plans to drive down trial costs from reduced sample size, reduced trial duration, and increased patient retention and improved sustainability.

At the JP Morgan Healthcare conference in January 2023, Roche CFO Alan Hippe cited a 70% reduction in sample size and half the required trial duration in their PASADENA study by using a novel digital endpoint.
In June 2023, the FDA endorsed a digital outcome measure of physical activity as a primary endpoint in Bellerophon’s pivotal trial using wearable data, showing that data from the sensor could be used to reduce the trial’s sample size by more than 50%.
Astrazeneca’s CRESCENDO trial is currently leveraging a smart spirometer and a cloud-based AI solution for quality control in data collection for both primary and secondary endpoints, predicting to cut treatment duration in half and reduce overall trial duration by up to 15%.

Compared to digital biomarkers, the field of generative AI is new and unproven for pharma, an industry where millions of lives and billions of dollars depend on scientific evidence for proof of efficacy and safety. Despite this, we’re seeing pharma lean into the more novel, unknowns of AI, instead of doubling down on the existing biomarker innovation that has proven effective and reliable.

In June, Sanofi announced that they are going “all in” on artificial intelligence and data science, and called out a number of partnerships in areas such as drug discovery. Beyond drug discovery, there is an extended opportunity to expand investment strategies into drug development. Objectives of trial sponsors are to de-risk trial success and FDA approval, while also showing influence in the industry by investing in innovation and new technology to accelerate their operations. While the two may seem contradictory, investment in evidence-based methods that leverage AI & ML is an easy win — DHTs and data models have been using AI & ML for more than ten years, with thousands of trials conducted to safeguard their investment.

With data curated by the Atlas platform for precision measurement, we know that more than 130 pharma and biotech companies have invested in DHTs. Metrics range from walking speed and blood pressure to sleep quality and cognition, with over 1,300 AI-powered digital biomarkers deployed in clinical trials in the past 15 years.

Source: Using Artificial Intelligence & Machine Learning in the Development of Drug & Biological Products; Atlas

As pharma executives craft their AI & ML strategies, they could explore new tools like generative AI, or they could choose to invest in digital biomarkers that have been consistently validated and are well-known for safe and effective clinical trials. Prioritizing digital biomarkers is no longer a nice-to-have, but rather a tried and true investment that already leverages innovative solutions within AI & ML to improve patient outcomes, faster.

Rather than chasing generative AI, pharma executives can focus their AI & machine learning (ML) strategies for drug development on investing in and deploying connected sensors to reduce costs & improve patient outcomes.

Blog authored by Abby DeVito, Sr. Director, Go-to-Market, ICON. Originally written for HumanFirst, a company acquired by Icon in january 2024.

In this section

Connect with us