Originally published by our sister publication Gastroenterology & Endoscopy News
By Marcus A. Banks
WASHINGTON—Machine learning tools can sift through genetic profiles to predict which patients with obesity will best respond to glucagon-like peptide-1 agonists such as liraglutide (Saxenda/Victoza, Novo Nordisk) or semaglutide (Ozempic/Wegovy, Novo Nordisk), according to a presentation by Mayo Clinic researchers at DDW 2024.
Mayo researchers previously had found that patients with obesity and the “hungry gut” phenotype—meaning they felt hungry shortly after eating a full meal—were more likely to respond to liraglutide than those without the phenotype whose satiety lasted longer (Obesity 2021;29[4]:662-671).
Subsequently, they developed a machine learning tool that analyzed the blood and saliva samples of people with obesity, and then predicted which patients had the hungry gut phenotype on the basis of their proportions of 40 gene variants present in the samples. At DDW, researchers reported that people with the phenotype responded better to semaglutide, just as they had with liraglutide (presentation 638).
“This opens up the possibility that we can use machine learning to predict which patients will benefit from GLP-1 agonists, rather than just prescribing them to everyone with obesity,” said study investigator Andres Acosta, MD, PhD, a gastroenterologist who studies food intake and obesity at Mayo Clinic in Rochester, Minn. Dr. Acosta is also the co-founder of Phenomix Sciences, which licenses obesity phenotyping data and built the machine learning tool presented at DDW.
The study included 84 participants (age, 47.6±10.9 years; body mass index, 38.8±6.9 kg/m2); 51 of the patients had the hungry gut phenotype based on the machine learning analysis, and 33 did not. All the patients were taking semaglutide, in doses ranging from 0.25 to 2.4 mg once per week, regardless of their gut phenotype. Almost all (94%) participants were white.
By nine months, people in the hungry gut group had greater levels of total body weight loss (–14.4% ± 6.6% vs. –10.3% ± 7.0%; P=0.045) than those without the phenotype, a difference that had strengthened at 12 months (–19.5% ± 11.4% vs. –10.0% ± 9.3%; P=0.01). Of note, the difference was not seen earlier in the study at three and six months. The machine learning algorithm was 95% accurate at predicting which patients would do best with semaglutide, the researchers reported.
“There’s a great need for similar studies with more diverse populations, but we think the work to date is encouraging,” Dr. Acosta said. The study did not control for people’s diets, so it is possible that improved dietary choices influenced the results. But as doses of semaglutide were comparable between groups, and diet was not tracked in either group, Dr. Acosta still finds that having the hungry gut phenotype predicts GLP-1 benefit.
“We have other medications for controlling obesity, not just GLP-1s,” such as phentermine-topiramate extended release (Qsymia, Vivus) and naltrexone-bupropion extended release (Contrave, Currax), Dr. Acosta noted. If machine learning tools can pinpoint who will benefit from a GLP-1—and who won’t—then clinicians can make rational prescribing decisions that better help patients while controlling drug costs, Dr. Acosta said. “This is a great example of precision medicine,” he added.
Caution Against Overuse of GLP-1s
“From my perspective, it’s just as important to figure out which would be at risk for taking them and who should avoid them as it is who would respond well,” said Christopher Thompson, MD, the director of endoscopy and co-director of the Center for Weight Management and Wellness at Brigham and Women’s Hospital, in Boston.
Physicians who treat people with osteopenia or sarcopenia should think twice before prescribing a GLP-1 agonist, Dr. Thompson advised, because the medications are likely to reduce lean body mass far more than surgery or an endoscopic procedure. Before writing a GLP-1 prescription, the clinician should order a bone density scan, Dr. Thompson said, and only proceed if in the clinician’s view the patient has sufficient bone density to absorb a GLP-1 agonist.
“I get a little anxious about everyone going on these,” Dr. Thompson noted, with some patients perhaps receiving prescriptions via online vendors that require a brief consultation rather than a full evaluation.
When Dr. Thompson prescribes a GLP-1 agonist, he said it is always in conjunction with a resistance training program that helps people stave off the inevitable loss in lean body mass that will accompany the GLP-1 agonist. In addition, he said he always alerts patients that the medication is lifelong, with weight likely to return soon after they stop taking it.
Dr. Acosta is the co-founder of Phenomix Sciences. Dr. Thompson reported no relevant financial disclosures.