One of the many promising applications of artificial intelligence (AI) in health care is assistance with personalized nutrition strategies. Such plans are developed by gathering disparate cognitive, metabolic, microbiome, and physiologic data and integrating it with genomics, medical history, environmental, behavioral, and cultural demographic data to tailor an individual’s diet for optimal health outcomes.

However, as with many rapid AI advancements in health care, some question whether the technology is ready for utilization in a real-world environment. Two experts debated the merits of AI and its use with current personalized nutrition strategies during the Sunday, June 23, session The Role of Artificial Intelligence in Transforming Nutrition and Diabetes Care—Ready for Prime Time or Not Quite?

The session can be viewed on-demand by registered meeting participants on the virtual meeting platform. If you haven’t registered for the 84^th Scientific Sessions, register today to access the valuable meeting content through Aug. 26.

Samantha Kleinberg, PhD, Associate Professor at Stevens Institute of Technology, presented the case that AI and personalized nutrition strategies provide essential insights for patients and providers, emphasizing that such plans can be tailored to individuals’ unique glycemic responses to different foods.

“Responses to diet are extremely individual,” Dr. Kleinberg said. “People can have very different responses to the exact same food.”

Researchers can leverage physiological data to create personalized diet plans that target optimal glycemic response. Dr. Kleinberg presented research that showed personalized nutrition plans led to superior outcomes compared to the gold-standard Mediterranean diet.

One advantage of personalized nutrition strategies at the community level is the ability to account for a group’s food environment, meaning the spatial, financial, and cultural access to certain foods in a defined area, as well as relevant psychological cues that can shape and reinforce eating habits. Tracking these environmental characteristics is essential because they can shape disparities or reveal unique needs for an individual or community.

While investigators have researched and analyzed food environments, many have focused on static locations around an individual’s or population’s home, work, or school environments, according to Dr. Kleinberg, neglecting to account for movement outside of these static areas throughout the day.

“With AI and other types of sensing like GPS, we can actually track this [mobility data] in-depth. We don’t have to just focus on where someone lives or really crude measures like their ZIP code,” Dr. Kleinberg explained.

A study published in March that gathered data from more than 1 million people showed that individuals in larger urban settings or those who commute longer distances are often exposed to food environments throughout the day that aren’t near their homes. The data showed that 19 percent of visits to food outlets were for fast food, and among those visits, only 7 percent occurred within individuals’ home census tracts.

With mobility data, appropriate environmental interventions can be made to impact communal health.

“This is not a far-fetched, future thing,” Dr. Kleinberg said. “This is something various cities have done, either banning fast food outlets in an area or creating programs and incentives for healthier food outlets to open.”

Holly Nicastro, PhD, Coordinator of Nutrition for Precision Health through the All of Us Research Program, Office of Nutrition Research, National Institutes of Health, also advocated for personalized nutrition planning but contends that more time and research are needed to enact these strategies safely and effectively.

“We need a stronger evidence base that includes data from diverse participants,” Dr. Nicastro said.

She cited personalized nutrition studies, including PREDICT, and showed that the data could not be stratified to analyze factors like race, income, education, or geographic location (rural versus urban settings). Dr. Nicastro also emphasized that investigators must prioritize factors such as diversity in how populations interact with the health care system or comparing outcomes between healthy individuals versus those with comorbidities.

“Forging ahead [with AI for nutrition care in diabetes] has the potential to widen health disparities if we find that our models are working better for those overrepresented people or the well-studied people or the privileged groups,” Dr. Nicastro said.

She also argued that additional validation studies were required before officially sharing recommendations that can affect human health.

“Our methodologies, or the information we’re collecting on people, might not yet be precise enough to really show what’s going on,” Dr. Nicastro said.

With the Nutrition for Precision Health study, Dr. Nicastro and her team are working to achieve this aim by developing algorithms to predict individual responses to foods and dietary patterns using data that captures individuals’ holistic health profiles with a large and diverse population of participants often underrepresented in medical research. At the end of each module, participants will undergo a standardized meal test to assess metabolic phenotypes and measure health factors like insulin sensitivity, beta-cell function, and lipid metabolism.