Dandelion Health, the maker of a real-world data and clinical AI platform, has launched the first multimodal real-world clinical dataset focused on uncovering insights and opportunities related to GLP-1 receptor agonist drugs. The company describes the GLP-1 data library as “the first multimodal real-world clinical dataset” related to GLP-1 receptor agonists.
The library is sourced from Dandelion’s consortium of non-academic medical center health system partners spanning diverse geographic regions, including Sanford Health (a rural Midwest system), Sharp HealthCare (an urban California system) and Texas Health Resources (operating in cities and suburbs across the Southwest).
The multimodal aspect of the data sets it apart from many traditional healthcare datasets. Elliott Green, CEO and cofounder of the firm, pointed out, “80-90% of all healthcare data generated is imaging and waveforms — 12-lead ECGs, CTs, MRIs, X-rays, inpatient telemetry.” Historically, such information has not been deeply mined for insights. “Dandelion’s approach involves obtaining raw biological data that is typically only available through clinical trials.”
Amassing a research library for GLP-1 receptor agonists
Elliott Green
While GLP-1 receptor agonists remain a relatively new class of drugs, interest in them has reached a fevered pitch in recent years given their potential to help manage obesity. This interest surge has led to a demand for more comprehensive data on the long-term effects of these medications, as well as their potential for other chronic diseases, given the core influence metabolism has on an array of diseases.
Uniting millions of patient journeys in data
With GLP-1s, Dandelion has curated data including millions of longitudinal patient journeys. Consequently, healthcare organizations armed with that data “can effectively answer many questions about how the drugs have potentially affected a specific condition, population, etc.,” Green said.
One example of Dandelion’s capabilities is using ECG data to predict cardiac event risk pre- and post-GLP-1 treatment in specific patient subgroups. Dandelion’s data has subsets of patients taking a GLP-1 that captures the before and after effects. This includes patients who had ECGs before GLP-1 therapy and then data from patients who had undergone the therapy.
Green noted that the available biological data allows for comparisons, even in the absence of lab tests or documented effects. “An ECG is 30,000 data points in 10 seconds,” Green said. Take that and incorporate data on co-morbidities and other patient info, and researchers can start to understand whether the medication class appears to reduce cardiac event risk for males, say, aged 30-45. Or whether it works better in one demographic versus another.
Iteratively redefining scale and depth
The scale of Dandelion’s dataset is already substantial, but it is growing rapidly. “By the end of this year with five health systems, we’ll have around 18–20 million patients,” Green said. “The larger sample size and variety of sub-cohorts in our dataset could provide deeper dosing insights compared to traditional clinical trials.” This represents an advantage over typical clinical trials, which might enroll only 1,500 to 5,000 people in a Phase 3 study. In contrast, millions of patients could be on a drug in the real world, depending on its indications and popularity. “It’s crazy we’re not doing more with that real-world data once the drug is out in the market,” Green said.
Real-world data has the potential to surface new treatment indications faster than traditional clinical research, Green noted. “Historically, you’d run 10, 20, 30, 40, 50 clinical trials to prove out those secondary impacts,” he said. But the data is constantly emerging.
Potentially broadening the scope of GLP-1 indications
Growing data availability and rising sophistication of healthcare organizations points to potentially faster, more efficient treatment indication discovery and tailored patient applications. “We can see it with the right data, and that data lives in health systems, it’s just not been accessible,” Green said.
GLP-1 indications beyond diabetes and obesity could increase in the coming years in areas such as neurodegenerative diseases and cardiology. “If you’d have said five years ago you have a drug that will help diabetes, massive weight loss, and potentially Parkinson’s or Alzheimer’s, people would think that’s crazy,” Green said. “But that’s what we’re dealing with now.” An approach to identifying these new indications lies in combining GLP-1 data with comprehensive patient data, especially from before and after patients began GLP-1 therapy. By analyzing this data, researchers can identify relationships between GLP-1 use and a spectrum of health markers, such as changes in brain plaques on MRI or head CT scans for neurological issues, or changes in ECG readings and cardiac events. This approach allows researchers to pinpoint promising areas for further investigation and drug discovery. “You start to see the correlation across so many things,” Green said. “You get a sense of where to hunt.”
Filed Under: Drug Discovery, machine learning and AI, Metabolic disease/endicrinology
