We are at the brink of a fundamental shift to transform clinical research and development. This is due to the emergence of two trends that together create conditions ripe for scientific discovery:
- Real-world data (RWD): The availability of large, multimodal clinical datasets
- AI: The proliferation of clinical AI tools to extract insights from these datasets
To assess the impact of this combination we asked the question below, with the help of our friends at Pheiron and the incredible work of the Dandelion team (thank you, all of you).
Can GLP-1s reduce the risk of heart attacks and strokes among those who do not yet suffer from severe cardiovascular disease?
The goal was to go beyond what the Novo Nordisk SELECT trial had previously done when they studied people that had recently had cardiac events.
Here are our 4 key findings:
1. Expanded population: GLP-1s may serve as primary prevention for obese patients with mild or moderate CVD – almost 40 million additional patients in the U.S. alone (this only reflects patients with ECGs so the number could be nearer 100 million).
2. Impact: Our study found a statistically significant 15-20% reduction of Major Adverse Cardiac Events (MACE) risk using AI. This is in line with MACE incidence results observed in a similar past clinical trial (17,604 participants and 5 years).
3. Accelerated Signal: Using AI, we detected decreased risk just 1.7 years after GLP-1 initiation (as opposed to the 3.3 year study period in the Novo Nordisk SELECT trial). AI enables faster the signal by predicting cardiovascular risk, rather than analyzing actual adverse events.
4. Reduced Trial Size: AI may decrease trial duration and required recruitment size by more accurately identifying appropriate trial participants.
This was all covered by Fierce Healthcare and CNN. Thank you to Deidre McPhillips and Emma Beavins
If you wish to read our white paper that these findings come from, please follow the link in the comments to download it.
We genuinely hope this is the start of opening up the floodgates to more people working out how to utilise AI in order to reduce the time and expense of drug development, whilst increasing the accuracy and effectiveness of our trial process.
With many thanks to:
Shivaani Prakash, MSc, PhD, Ana-Vanessa “AV” Ploumpis, Gonzalo Hernandez, Jamie Dermon, MD, CPHQ, Ross Bierbryer, Isaac Moshe, Luke Sagers, Nick Burke, Kiko Wemmer, Maxence Frenette, Owen Kasozi, Manqing L., Nicholas Gossen, Richard Y., Matthew Stewart, PhD, Mathilde Williams, Balazs Lengyel, MD, Nick Znajkowski, Mara Alexeev, Sofia (Becerra) Van Dusen, Nicholas Shea, Seema Sood, Jason Ma, Armend Cobovic, Thore Buergel, Jakob Steinfeldt, MD, Dave Clarke, Steven Gipstein M.D., Sharp HealthCare, Texas Health Resources Sanford Health and of course, Ziad Obermeyer, Sendhil Mullainathan and Niyum Gandhi.
