NEW YORK – A study underway at the US National Institutes of Health's All of Us Research Program is laying the groundwork to develop machine learning algorithms that predict how patients will respond to certain diets.
Participants in the Nutrition for Precision Health study, which started enrollment in May, will provide a variety of data, including information on their daily diet, health, genetics, and gut microbiome. Investigators hope to use this information to identify biological differences that influence how individuals respond to different foods and, one day, develop tailored dietary recommendations.
Today, dietary guidelines tend to take a one-size-fits-all approach, said Holly Nicastro, coordinator for the Nutrition for Precision Health study.
"Even if everybody followed the healthy eating guidelines, which is hard enough to do, we still think that we may not be able to achieve optimal health," she said. "We want to understand the differences [and] the factors behind them, so we can predict how someone responds, and then ultimately, tailor dietary interventions to the individual or to subgroups of individuals."
Armed with such guidance, clinicians could potentially prescribe a diet tailored to a patient's specific needs as part of care plans to improve health or reduce their risk of chronic diseases like hypertension, diabetes, and cancer.
Investigators plan to gather data from 10,000 participants for the first phase, or module 1, of the Nutrition for Precision Health study. In an attempt to ensure findings are generalizable, there are limited exclusion criteria for module 1, though participants must be enrolled in the wider All of Us program, be at least 18 years of age, and live near one of the clinical centers, since participants must visit a study site twice to record vital signs and provide biological samples.
"Unless we study diverse groups and diverse populations, we're not going to be able to figure out what those key factors are that lead to differences in metabolism and differences in response," Nicastro said.
At the initial study site visit, participants will be given an accelerometer for tracking physical activity and a continuous glucose monitor for tracking blood glucose and will receive instructions for collecting a stool sample at home that researchers will use to analyze their gut microbiome. Over the next 10 days, participants will record their daily diets through surveys and test new methods to track food consumption, such as an AI tool that analyzes photos of meals.
During the second study site visit at the end of the research period, researchers will measure participants' height, weight, body circumference, respiratory rate, and blood pressure, as well as obtain blood and urine samples for lab tests to gauge metabolic biomarkers. All of this data is linked to data that participants provide to the broader All of Us program, such as genetic results and health records.
A key component at the end of module 1 is a "meal challenge," in which participants drink a liquid meal smoothie, so that researchers can assess biological and physiological changes before and after consumption.
Typically, nutrition studies compare one diet against another but miss the opportunity to understand why people respond to diets differently, said Leanne Redman, director of the reproductive endocrinology and women's health laboratory at the Pennington Biomedical Research Center at Louisiana State University, who leads the Nutrition for Precision Health clinical site there. In her prior work in which patients were admitted to the research center and overfed, participants varied in terms of weight gain, with some not gaining any weight at all. She hopes data collected as part of the Nutrition for Precision Health study will start to provide clues as to why.
A subset of participants from module 1 will continue to modules 2 or 3, in which the study provides participants with meals that align with three specific diets varying in fruits and vegetables, whole grains, refined grains, dairy, and sugar-sweetened drinks and desserts, among other food groups. Each of the diets is designed for weight maintenance, rather than weight loss or gain, Redman said.
As part of module 2, about 1,500 participants will be asked to adhere to each of the three diets for two weeks, with at least a two-week washout period in between. In module 3, about 500 participants will stay at a clinical research center while on those diets, during which they'll receive meals at standard times and follow a schedule for sleeping and physical activity. Module 3 also involves more intensive data collection, with participants weighed in person and a subset providing daily stool samples for microbiome analysis.
Participants can receive up to $300 in compensation for module 1, up to $1,500 for module 2, and up to $6,200 for module 3, an All of Us spokesperson said in an email.
The diets remain unnamed to avoid unintended bias, Nicastro said. Since participants in modules 2 and 3 are limited to specific dietary patterns, the exclusion criteria is stricter than for module 1. Participants who have received treatment for bulimia or anorexia nervosa in the past three years, for example, aren't eligible to participate in modules 2 or 3, nor are diabetics taking insulin or those with uncontrolled hypertension.
Investigators expect to launch modules 2 and 3 in late summer, according to Nicastro.
Building on All of Us' focus on returning results directly to participants, those enrolled in the nutrition study will receive some interpreted information from the data and samples they provide, such as their body fat percentage, microbiome makeup, blood glucose, and metabolism, Nicastro said.
Nutrition for Precision Health is a discovery science study, investigating many potential predictors, she noted. Data will be made available through All of Us' data platform, called the Researcher Workbench, where researchers can access it for their own studies. Meanwhile, investigators involved in Nutrition for Precision Health will use AI to analyze the trove of data gleaned from participants and build predictive algorithms from the most relevant factors.
"We don't yet know what those most important predictive factors will be, which is kind of exciting," Nicastro said.
Nutrition for Precision Health is a five-year study supported by the NIH Common Fund. If there is more money for a second phase of the study, Nicastro said she'd like to validate the algorithms they develop, use them to create customized diets for participants, and gauge if they respond to these tailored diets as expected.