Novo Nordisk obtained a serious enhance in March when the FDA authorized its GLP-1 diabetes and weight reduction drug Wegovy to additionally stop coronary heart assaults and strokes in sufferers with sure threat elements. The choice opened the market as much as greater than six million folks and broadened Medicare protection for the blockbuster drug, now on observe to turn into one of many best-sellers of 2024.
And that will solely be the tip of the iceberg, in keeping with a sweeping AI examine from Dandelion Well being, which recognized an extra 44 million lower-risk cardiovascular sufferers who may benefit from a GLP-1 remedy, hinting on the potential to drastically broaden insurance coverage protection.
Placing AI to work
Dandelion leaders ran the examine as a proof-of-concept take a look at of their real-world information platform, which makes use of business and tutorial AI packages to investigate affected person data obtained by way of revenue-sharing agreements with well being techniques, together with Sanford Well being, Sharp HealthCare and Texas Well being Sources. The platform might be a quicker and cheaper technique to ship outcomes than conventional medical trials. Research might take weeks, not years, and incorporate bigger and extra numerous affected person teams, in keeping with the corporate’s co-founder and CEO, Elliott Inexperienced.
“Nobody ever does preventative trials as a result of they’re simply too massive,” he mentioned. “However this AI proof-of-concept began to indicate that there is a path ahead to with the ability to obtain that.”
Mining real-world information
For the observational examine, Dandelion examined whether or not GLP-1 use may cut back the chance of main adversarial cardiac occasions reminiscent of coronary heart assault or stroke in obese and overweight sufferers. The examine used inclusion standards just like the Novo Nordisk-sponsored trial that helped persuade the FDA that Wegovy had coronary heart advantages. However, in contrast to SELECT, Dandelion’s affected person group didn’t have extreme preexisting heart problems.
Researchers utilized an AI algorithm from Pheiron educated to identify cardiovascular threat indicators on affected person electrocardiograms. The AI assigned sufferers a threat rating and in contrast predicted threat for coronary heart assaults and strokes in GLP-1 customers versus non-users. The outcomes expanded on the findings of Novo’s trial, which discovered Wegovy diminished the chance of main cardiac occasions within the examine group as a lot as 20%.
Dandelion’s examine, seven occasions the dimensions of Novo’s, decided that folks within the lower-risk group who took GLP-1 medication had 15% to twenty% decrease threat scores for main adversarial cardiovascular occasions after taking the drug for 3 years. Treating these further sufferers might head off 17,300 coronary heart assaults and 16,700 strokes a yr, in keeping with examine authors.
“With additional validation, this AI-driven method might probably unlock a brand new biomarker or surrogate endpoint that measures [major adverse cardiovascular event] threat, enabling shorter and smaller medical trials by predicting cardiovascular outcomes earlier with out ready for [adverse events] to happen,” they acknowledged.
Rushing analysis
Actual-world data-based AI might validate GLP-1s for broader functions, from neurological and liver ailments to habit and arthritis.
Inexperienced mentioned Dandelion already has one other GLP-1 examine within the works, utilizing AI to evaluate muscle modifications on affected person belly CT scans to find out whether or not GLP-1 medication additionally trigger muscle to wither throughout weight reduction. The platform might additionally assist pharma corporations discover and validate label growth alternatives and carry out different kinds of analysis. Dandelion is working with a number of corporations, Inexperienced mentioned.
“Scientific trials are very centered. You are available in with a really particular query, and also you get a really particular affected person inhabitants,” Inexperienced mentioned, noting {that a} broader method is commonly too difficult because of time and value constraints.
With AI, as an alternative of taking a look at a restricted space, just like what you may see underneath a single road mild, Inexperienced mentioned, you may illuminate the entire neighborhood.