The growing popularity of glucagon-like peptide 1 (GLP-1) drugs opens up massive opportunities for biopharma companies, life sciences firms and researchers and represents a massive shift in how health systems approach obesity and diabetes treatment.
The industry is racing to figure out the effects of GLP-1 agonists on different conditions and populations and to test for secondary indications to treat cardiovascular diseases and liver disorders.
To unlock these insights, the industry needs large volumes of data and artificial intelligence technology for computational analysis.
Health tech startup Dandelion Health saw an opportunity to leverage its large, de-identified data set to help bring precision medicine to GLP-1s.
The company, a real-world data and clinical AI platform, has launched a GLP-1 data library as a multimodal real-world clinical data set built specifically to surface insights and opportunities related to the GLP-1 receptor agonist drug class.
Dandelion taped into its consortium of three major health system partners—Sharp HealthCare; Sanford Health in Sioux Falls, South Dakota; and Texas Health Resources—to collect structured and unstructured data for more than 10 million patients across a range of populations and longitudinal patient journeys.
The company's GLP-1 library incorporates all available clinical data for millions of patients, including 200,000 and counting who are taking a GLP-1, according to the company.
Elliott Green, co-founder and CEO of Dandelion Health, noted that nearly all of the unstructured clinical data that make up about 80% of medical data have been locked away in silos or are unusable without tools like AI algorithms that structure them for computational analysis.
These data include a range of clinician notes in electronic health records, imaging, such as MRIs and CT scans, in picture and archiving communications systems and echocardiogram (ECG) waveforms in vendor systems.
Dandelion Health has pulled together all these data, both structured and unstructured, into a curated GLP-1 data library so researchers, life sciences companies and healthcare organizations can advance precision medicine for cardiometabolic disease, Green noted, while also spurring potential new use cases for GLP-1s.
“While the past few years has produced great advancements in obesity care, there is still a wide gap between cardiometabolic care today and the high-innovation, high-investment, and deeply personalized care paradigms we see in immunology and oncology," Green said. “What got those markets to where they are today was data—data that revealed underlying mechanisms of disease, how individuals’ diseases look different, and consequently, how they might respond to therapy differently.”
Dandelion Health launched four years ago to help advance clinical AI. The company began amassing a de-identified data set for developers to build and test how AI algorithms perform.
While AI holds promise to improve healthcare delivery and outcomes, many AI products are built on narrow data sets that do not represent the diversity of the U.S. population and its care practices, according to the company.
Dandelion Health co-founders, including Chief Scientific Officer Ziad Obermeyer, M.D., wanted to fill the void.
Obermeyer, an associate professor and Blue Cross of California Distinguished Professor at the Berkeley School of Public Health, researches bias in healthcare. The company launched an initial pilot last year to test algorithms that use electrocardiograms to predict heart conditions, HealthcareITNews reported. The aim is to measure the performance and bias of AI across key racial, ethnic and geographic subgroups.
Dandelion Health is currently working with three health system partners—specifically chosen for their geographic, medical and demographic diversity as well as the longitudinal coverage of their patient data. Two more health systems will be added to the mix this year, Green said, to represent between 18 million and 20 million patients.
With its GLP-1 data library, researchers and life sciences companies also can use AI applications available on the Dandelion platform to get answers to open questions like GLP-1s' impact on body composition or their potential to reduce cardiovascular disease risks.
Research-driven healthcare organizations can use the GLP-1 data library to evaluate the quality of weight loss through endpoints found in body scans or compare the efficacy of treatments not studied head to head across a range of different real-world measures, according to Dandelion Health executives.
Organizations also can tap into the data library to demonstrate GLP-1’s therapeutic effects beyond current uses, including secondary benefits derived from exploratory use or quantify any side effects associated with GLP-1 use.
A precision medicine approach also could help match patients to the right treatments.
“Our goal was to curate a dataset that captures the full range of benefits that this new class of emerging treatments can offer—based on a larger and more representative patient population than would traditionally be seen in clinical trials,” said Shivaani Prakash, Ph.D., Dandelion’s head of data, in a statement.
“By making the rich, multi-dimensional data offered by unstructured modalities available and accessible—and connected to real-world treatment patterns and outcomes—people who use the library can answer key questions about the critical role that GLP-1 based treatments will play in clinical care.”
The company is working with AI developers and researchers on proofs of concept using its GLP-1 data library. Case in point, two researchers from a large academic medical center developed an algorithm that segments abdominal CTs to quantify fat loss and muscle and bone preservation.
Dandelion also is working on its own research project using the library’s structured EHR data to evaluate GLP-1s' efficacy across real-world patient cohorts and demographic groups and observable GLP-1 effects on other comorbidities that may suggest adjacent indications. Dandelion will publish these findings in a scientific preprint in the second quarter.