Digital Twins Bring Precision Medicine Closer to the Masses

The phrase precision medicine has been a headline term for almost a decade, ever since President Barack Obama launched the Precision Medicine Initiative in 2015. According to clinicians at the cutting edge of the field, we are getting close to the point where mass adoption is practical and very effective.

And digital medicine, rather than pinpointed pharmaceuticals, will be a big part of that adoption, according to Jeffrey Mechanick, MD, medical director of the Marie-Josee and Henry R. Kravis Center for Cardiovascular Health at Mount Sinai Fuster Heart Hospital in New York, NY.

“We are at an inflection point now,” Mechanick told MD+DI. “We are building a culture of medical centers and academic centers incorporating institutes for health computing and research departments for [artificial intelligence] AI. I foresee a lot of biosensors and computational medicine. This isn’t going to be an interpretive science — not a soft science or word of mouth. This is hard science at the same level of rigor you would see for big pharma.”

Mechanick, while serving as lead author, recently reinforced this sentiment through a study published in the Journal of the American College of Cardiology. The study demonstrated digital twin technology developed by Mountain View, CA-based Twin Health, including AI, resulted in significantly improved blood pressure control in patients with type 2 diabetes (T2D) — Mechanick is a member of the company’s medical advisory board.

Related:IBD Patients Gain a Holistic Edge with SonarMD’s New Partnerships

The study, a secondary analysis of a one-year randomized controlled trial that looked at remission rates of T2D patients who used the technology, included 59 patients who were also taking medications to treat high blood pressure. After one year, patients with high blood pressure who used the digital twin AI platform, and its associated coaching by healthcare professionals, achieved higher rates of normalized blood pressure — 40.9% vs 6.7% who received standard care.

Twin Health is not alone in demonstrating the utility of digital twin/AI technology. Singapore-based Mesh Bio recently received the nation’s regulatory approval for its HealthVector Diabetes product, a metabolic digital twin-based software-as-a-medical-device platform.

Developed using computational systems biology and AI, it is intended to predict a three-year onset risk of chronic kidney disease in patients with T2D. In a recent study demonstrating the platform’s abilities, the model performed strongly, achieving an area under the curve between 0.80 and 0.82. In prediction, generalized metabolic fluxes computed with complete model parameters attained high performance with an AUC of 0.86, and with incomplete parameters it achieved an AUC of 0.75.

Related:HMI: The Future of Patient Care Is Here!

Images courtesy of Jeffrey Mechanick and Lisa Shah

Sensors and sensibilities

Digital twins — the detailed computational simulations of complex systems — are not new but the proliferation of Internet of Things technology and ultra-fast AI platforms are making them more prolific in areas as disparate as self-driving cars and power grid resilience.

Lisa Shah, MD, Twin Health’s chief medical officer, said the human metabolic system certainly qualifies as a complex system, and by integrating sensor data and self-reported data with lab results, Twin Health can deliver personalized recommendations via its mobile app and team of coaches. According to company material, the platform analyzes more than 3,000 data points daily.

“People think about things that are biometric when [discussing] metabolic disease,” Shah said, “You can get a continuous glucose meter read on everyone every five minutes. But we also get preferences and behavior data. We get obvious biometric data but also activity and sleep data. [We] care a lot more about how that data intersects with things like glucose, weight, and muscle mass. We want to understand [a patient’s] environment. We get everything from how much you can afford to spend on food a week, do you live in a safe neighborhood, or work a night shift? Can you walk outside at night? Are you a single parent? How many mouths do you have to feed? Even more deeply, we get likes and preferences so a member can give a thumbs up or down to recommendations from the platform.”

Related:5 Private Medical Device Companies Analysts Are Talking About

One of the digital twin’s appeals, Shah said, is its immediacy. Rather than base recommendations on data such as a lab-drawn A1C that may be a 90-day average, Shah said the platform lets people see their own data in the moment. “To be able to eat a pizza and see that spike in front of you and then be able to eat a salad and see the difference is massively impactful on someone’s behavior change,” she said.

Opportunities and challenges

In a recent review published in Trends in Endocrinology & Metabolism, Oregon Health & Science University researchers Peter G. Jacobs and Clara Mosquera-Lopez outlined benefits to digital twin/AI platforms similar to Shah’s comments, and also outlined several challenges to its widespread adoption.

“Three primary challenges have been identified to translating digital twin research into clinical practice: the fusion of multimodal data for the development of digital twin models, clinical implementation, and data governance and product oversight,” they wrote in the review.

Another challenge that often arises in evaluating AI algorithm effectiveness is its often proprietary black-box intellectual property status. Without reproducibility, the argument goes, adequate validation of an AI platform’s claims are often impossible.

Shah, however, said Twin Health’s revenue model, in which the company assumes full risk for outcomes, validates the platform’s utility to its customers. “Our value-based approach is about outcomes and improving health to the point there is a significant ROI for them. The per-patient, per-month model of the past is really becoming less preferred — it’s not high-value. Customers want to move the needle on expensive healthcare and drugs like GLP-1’s.”

She also said the company’s business model is customizable as to what point a customer wishes to begin enrollment. Launched commercially in the US in 2021, she said Twin Health now has 60 commercial partners.

“We agree with our partners on entry eligibility criteria based on what they choose to cover,” Shah said. “There are three ‘front doors’ — a diabetes diagnosis, a pre-diabetes diagnosis, and elevated BMI. Partners can decide to cover a person at various BMI levels and we work with them to screen entry criteria when members come in. They are looking for something that is fully metabolic health and fully behavioral. The entry points are diabetes, pre-diabetes, and weight, but you can impact other things such as fatty liver disease, high blood pressure, hypercholesterolemia, dyslipdemia, and Framingham risk score. Greater metabolic improvement means one solution for a lot more of your population.”

Mesh Bio executives did not respond to MD+DI’s request for comment.