ECG data and AI will fuel pharma’s next drug development and commercialisation boom

It’s no secret that drug development is a huge gamble for pharmaceutical companies. The process involves betting billions of dollars and more than a decade of time that a compound will beat the 90% failure rate and generate value for both the creators and consumers of the therapy.

It’s a risk worth taking as the market continues to demand safer, more effective, and more accessible therapies for increasingly tailored applications. But, with so much on the line, it makes sense for pharma companies to do everything they can to make the process faster, less expensive, and more reliable — including exploring how to use artificial intelligence to enhance the utility of insight-rich data sources, such as waveform data, into the search for new therapies.

Waveform data includes electrocardiogram (ECG) tests — an affordable, non-invasive, and frequently ordered method of observing the heart’s electrical signals. In the clinical setting, the test can help identify arrhythmias and blood flow issues — and, by generating more than 3,000 data points over a 10-second period, a typical 12-lead ECG also contains a wealth of important information about the heart’s relationship to a wide variety of other organ systems and biological functions.

While at first glance ECGs may not seem like a top contender for revolutionising research, a deeper dive shows that, when AI algorithms are applied to it, this plentiful and highly predictive category of data can augment precision research in surprisingly powerful ways throughout the clinical trial process. Here’s how:

Screening and identifying patients for inclusion in clinical trials

ECGs are generally more accessible, faster, and about 95% less expensive than other tests, such as echocardiograms or cardiac MRIs, making them an attractive option for researchers looking to accurately screen potential enrollees for inclusion or exclusion criteria. By using this data to get the right patients enrolled at the beginning, researchers can avoid delays as well as reduce the high costs and administrative burdens of protocol amendments.

Additionally, they are increasingly being used to predict the likelihood of future conditions that may have an impact on the outcomes of clinical trials, so researchers can opportunistically screen individuals for participation.

AI and machine-learning models require access to rich, reliable, multimodal data and have been able to leverage ECGs for predicting or detecting a number of conditions with signals hidden in cardiac function data, including early Parkinson’s disease, pre-diabetes, and sleep apnoea, all of which may have an influence on patient selection and earlier diagnosis.

This could be especially useful for investigating therapies intended to prevent exacerbations of disease or new diagnoses. For example, identifying and enrolling the people who are most likely to have a second cardiac event is essential for maximising the investigative utility of a trial for a therapy designed to prevent these events from occurring.

Informing observational studies and gauging treatment impact over time

Because ECGs are simple and inexpensive, patients tend to have a lot more of them, and have them much more frequently, than other tests. Some wearable devices even allow users to take clinical-grade ECGs on demand, as often as they desire. That means researchers can look at many different consecutive readings over time, unlocking insights into the natural history of diseases and the impact of standard care protocols.

Series of ECGs have been used to look for signals of cirrhosis in non-alcoholic fatty liver disease before it’s visible on expensive, intermittent imaging tests, leading to an AI-powered severity score. It can also detect the likelihood of major adverse cardiovascular events, including sudden cardiac death, allowing researchers to examine the effectiveness of new therapies during clinical trials and even create predictive tools to prevent adverse outcomes.

Monitoring safety and efficacy during later-stage clinical research

The heart can speak loudly about what’s happening throughout the body, which can alert researchers to safety risks from investigational therapies. The AI and ECG combo can help uncover early signs of acute kidney injury, for example, or detect electrolyte abnormalities that may be caused by medications.

Since ECGs are easily obtainable and can provide varied information about multiple factors of interest to researchers, early safety signals gathered from waveform tests can inform how and when researchers look into potential issues with experimental treatments, such as providing vital data about if and when to discontinue a trial that is producing concerning results. While this is never a desirable outcome, knowing when to pivot, sooner rather than later, can safeguard patients and prevent wasted time and resources for the trial sponsor.

Expanding clinical indications to promote product success

In a time when pharmaceutical companies no longer have the economic luxury of decades-long development cycles, indication expansion can provide new treatment avenues for patients with few existing options while assisting with maximising a company’s financial investment in a specific product or therapy.

Indication expansion can be a challenging, costly, and time-consuming process, with an average lag of 1.7 years between the first and second indication. Longer times between expansions are lost opportunities for gaining market recognition. However, by reusing the vast amounts of ECG data collected in clinical trials for existing indications, companies can more quickly and efficiently identify opportunities for expansion, since ECGs can contain insights into how the drug may be affecting other biological functions, such as in the kidneys or liver.

Ultimately, using ECG data and other alternative datasets throughout key points in the clinical trial life cycle can help researchers achieve more of their goals with less cost, time, and effort. As AI models continue to mature into useful tools for analysing waveform data in conjunction with other real-world information, pharmaceutical companies have an intriguing opportunity to use these methods to accelerate research and develop safer, more informative, and more efficacious clinical trials.

About the authors

Elliott Green is the co-founder and CEO of Dandelion Health, a real-world data and clinical AI platform powering next-generation precision medicine and personalised care. His career has spanned finance and healthcare technology, culminating in executive roles within health tech start-ups focused on payers, providers, life sciences, and healthcare data. Green brings an in-depth understanding of the operational components of the US payer, provider, and life sciences ecosystems and an ability to establish and manage complex institutional partnerships.

Jamie Dermon, MD, is the director of clinical solutions and clinical informaticist at Dandelion Health, where she develops innovative solutions to improve patient safety and reduce risks in clinical trials. A board-certified emergency physician, with expertise in healthcare quality improvement, she leverages data-driven strategies to optimise patient care and enhance clinical research outcomes.