Proceedings of a Workshop
Challenges and Opportunities for Precision and Personalized Nutrition
Proceedings of a Workshop—in Brief
The Food Forum of the National Academies of Sciences, Engineering, and Medicine held a public workshop, Challenges and Opportunities for Precision and Personalized Nutrition, on August 10–12, 2021. It explored potential challenges and opportunities in applying precision and personalized nutrition approaches to optimize dietary guidance and improve nutritional status. The presentations discussed ways to define both approaches, described current research designs and methodologies in diverse populations, and examined limitations in design and data. They also reviewed innovative methodologies and technologies at the various scales of precision nutrition (including the genetic, physiologic/microbiome, individual, and social-ecologic scales) and discussed challenges and opportunities for implementing these approaches equitably, from the perspectives of academia, the federal government, and industry.
This Proceedings of a Workshop—in Brief highlights the workshop presentations and discussions and is not intended to provide a comprehensive summary of the information shared.1 The information summarized here reflects the knowledge and opinions of individual participants and should not be seen as a consensus of the participants, the Food Forum, or the National Academies.
THE CURRENT EVIDENCE BASE AND LIMITATIONS
The August 10 session featured six presentations that reviewed the current evidence base for precision and personalized nutrition, including potential definitions for these terms, research designs and methodologies, limitations in designs and data, and future challenges and opportunities for the field.
John Mathers, Newcastle University, discussed human variability and how it forms a basis for developing personalized and precision nutrition. People differ in visible ways, such as height and body shape, and also in less apparent ways, such as responses to specific foods and diets. As an example, Mathers referenced a study that found large interindividual variation in glycemic responses to several types of standardized meals (Zeevi et al., 2015).
According to Mathers, current public health approaches to changing diet are relatively ineffective because they typically consist of “one-size-fits-all” guidance, but precision or personalized nutrition approaches might lead to advice and support that are more likely to promote behavior change. Mathers proposed defining “personalized nutrition” as an approach that uses information on individual characteristics to develop targeted nutritional guidance, whereas “precision nutrition” suggests the possibility of obtaining a sufficient quantitative understanding about the complex relationships among an individual, their food consumption, and their phenotype to offer nutritional intervention that is individually beneficial. Precision nutrition is more ambitious, he clarified, as it demands much greater scientific certainty.
1 The workshop agenda, presentations, and other materials are available at https://www.nationalacademies.org/event/08-10-2021/challenges-and-opportunities-for-precision-and-personalized-nutrition-a-workshop (accessed August 19, 2021).
Mathers requested considering a biopsychosocial model when developing both approaches, which encompasses biological (e.g., genotype, epigenome, and gut microbiome), psychological, and sociological factors that contribute to interindividual variation in responses to food. He also suggested that precision and personalized nutrition focus more on individual health aspirations, food preferences, and barriers/facilitators to behavior change. The goal, he proposed, is to improve opportunity for everyone by developing wide-reaching, affordable, cost-effective approaches that address inequities in dietary intake and health outcomes.
Constance Hilliard, University of North Texas, explained that the genetic data on which precision approaches to health are based are often derived from people of European ancestry. This hinders their effectiveness to improve health in other populations, she maintained. She elaborated on the importance of precision when working with genetic populations by sharing a case study of high rates of salt-sensitive hypertension in African Americans of slave descent.
This population derived from a unique ecological niche, Hilliard explained, which provides clues about the etiology of its high rates of hypertension. She emphasized the importance of addressing three fallacies related to this population’s origins. First, race is not a scientific concept and, in her view, is not suitable for use in medicine except to address discrimination and past exclusion. Second, DNA ancestry provides a much sharper focus for medical research, and understanding ecological niche populations further enhances precision. Third, many ancestors of African Americans came from the deep interior of West Africa, and as Hilliard described, ended up on the coast only after being kidnapped and marched up to 1,000 miles away. This detail is critical, she explained, because coastal West Africans and Europeans were genetically accustomed to consuming 5,000 milligrams/day of sodium, whereas subsistence farmers in the interior—one of the most sodium-deficient regions in the world—had adapted to only 200 milligrams/day. Hilliard highlighted the discovery of gene variants that play a major role in sodium metabolism and are found almost exclusively in people descended from the West African interior. The presence of either variant has been associated with a 2- to 100-fold increased risk of developing kidney disease (NIH, 2017).
Hilliard illustrated that modern genomics and testing of DNA ancestry allows for greater precision in sodium guidelines and other nutrient guidance for all U.S. residents. The Dietary Guidelines advise those 14 years and older to take in less than 2,300 mg/day sodium, but this is substantially higher than that consumed by these West African ancestors. To improve the health of diverse genetic populations, Hilliard called for stratifying dietary guidance by DNA ancestry, which implies viewing U.S. residents as not only multiethnic or multiracial but also multigenomic.
Abigail Johnson, University of Minnesota, discussed potential approaches for integrating information about the microbiome with dietary intake data. She recounted her postdoctoral research study, which aimed to characterize day-to-day changes in the composition of the adult microbiome resulting from dietary intake. Study participants collected one microbiome (stool) sample daily for 17 days, and analysis revealed that participants had distinct microbiome conformations that were relatively unchanged from day to day.
When the team integrated participants’ dietary intake data with their microbiome data, Johnson said it became clear that dietary composition was more variable than the corresponding microbiome composition from day to day. Dietary data analysis revealed that a person’s nutrient intake was more stable than individual food intake during the study period but also that nutrient intake did not pair with microbiome composition. At that point, the team clustered the foods into groups and generated a phenetic tree to statistically account for similarities in food group characteristics. The results indicated the distinct and highly variable nature of each participant’s dietary patterns, Johnson reported, and a successful pairing of microbiome composition and food intake. Both grain and fruit fiber diversity (a metric based on the diversity of fiber from grains and fruit, instead of the overall quantity of those food groups) also paired with microbiome composition using this method, which Johnson said was exciting.
Johnson reported that her team also applied computational tools to the study data to determine that certain dietary patterns appeared to be associated with the abundance of specific bacterial species and explained a large amount of the microbiome variation. Moreover, she continued, they found that most diet–microbe associations were personalized (many significant relationships existed between foods and bacterial species within a person, but very few of those relationships were repeated across people).
Christian Metallo, Salk Institute for Biological Studies, discussed applying metabolomics to studying nutrition and disease physiology, illustrated through a case study demonstrating how metabolic mechanisms affect the balance of nonessential amino acids and drive a disease state.
Macular telangiectasia (MacTel) is an eye disease characterized by vascular abnormalities resulting in central vision loss and difficulties in activities such as driving and reading at a relatively young age (around 40 years). Metallo reported that genome-wide association studies have identified several genetic variations that are associated with the biosynthesis pathway for serine, a nonessential amino acid, such that serine and glycine levels are lower in the plasma of MacTel patients compared with controls. Metallo’s team sought to understand serine’s role in the development of MacTel by examining how cells control its flux to different pathways. The team identified serine palmitoyl transferase
(SPT) as a key enzyme and was able to determine that SPT normally uses serine to catalyze biosynthesis of sphingolipids (structural components of cell membranes) but can also use alanine to generate deoxysphingolipids if serine levels are low or specific mutations are present in subunits of SPT. MacTel patients have higher levels of deoxysphingolipids than controls, Metallo noted, and plasma levels of the metabolite 1-deoxysphinganine are increased in patients with a type of hereditary neuropathy, some but not all of whom have MacTel. Metallo’s team next fed mice a serine-/glycine-free diet and observed that it could produce a similar neurological phenotype as is observed in MacTel patients, where the phenotype is driven by genetics.
Several lines of genetic evidence are now available to clarify that MacTel is a multigenic disease of dysregulated amino acid metabolism, Metallo confirmed, characterized by a phenotype with low serine and glycine levels and elevated alanine levels. Metallo’s team also found that a metabolic imbalance was triggered with a serine- and glycine-free, low-fat diet and a serine- and glycine-adequate, high-fat diet, but a serine- and glycine-free, Western-style, high-fat (60 percent of energy intake from fat) diet triggered the “metabolic catastrophe” that led to peripheral neuropathy. This is a severe and extreme diet, he admitted, which raised questions about how this metabolic pathway might function in individuals with a more typical diet. Metallo suggested that more sophisticated understanding of the body’s robust, genetically evolved biochemical engineering control mechanisms is needed to exploit them therapeutically through diet.
Susan Carnell, Johns Hopkins University School of Medicine and Appetite Lab, discussed how psychosocial and behavioral research on eating behavior could enhance precision nutrition by determining not only an optimal diet for an individual but also identifying the best strategies to support that individual in following the diet.
Carnell began with a discussion of “appetitive characteristics”—early, emerging, enduring dispositions toward food or eating styles that differ between individuals. Examples include food cue responsiveness (the degree of response to external food cues, such as sight or smell) and satiety responsiveness (the level of sensitivity to internal cues to stop eating, such as gut hormones or gastric distension). Carnell highlighted the Child Eating Behavior Questionnaire as a commonly used, parental report measure of a child’s traits associated with food approach (e.g., food responsiveness, enjoyment of food, and emotional overeating) or food avoidance (e.g., satiety responsiveness, slowness in eating, food fussiness, and emotional undereating). Carnell said that because appetitive behaviors seem to emerge early in life and are somewhat genetically determined, it may be better to work with them, as they may be hard to change.
Carnell reviewed evidence indicating that appetitive characteristics are associated with a number of outcomes in children, including weight/adiposity, diet, and responses to the food environment and to interventions designed to change eating behaviors. For example, food approach and avoidance traits have been positively and negatively associated with adiposity, respectively, and different food intake responses are exhibited to variable portion sizes based on levels of satiety responsiveness. Carnell conveyed that these findings emphasize the importance of considering appetite, food preferences, and other aspects of eating behaviors when developing personalized nutrition plans.
Michael Snyder, Stanford University, discussed using “big data” to support individualized profiling as a strategy to better manage health. It is increasingly possible to quantify factors that influence individual health, he said, but more importantly, their effects can also be quantified with detailed molecular and physiological measurements of an individual.
Snyder described his group’s longitudinal research on personal omics profiling, which involves in-depth measures (“omics measurements”) to gather information about participants’ genome, epigenome, transcriptome, proteome, cytokines, metabolome, lipidome, and microbiome, among others. Questionnaires, basic and advanced clinical tests, and wearable biosensor devices provide additional information, he explained, with the goal of better characterizing what a healthy profile looks like, how it changes over time and during phases of illness, and how it differs between people. Snyder highlighted the importance of the longitudinal nature of measurements in order to detect changes and better understand a health trajectory.
He recounted nearly 50 occurrences of “major health discoveries” during the first few years of profiling participants, including of cardiovascular, metabolic, hematological, or oncological abnormalities that often precede disease. Every discovery was made before the person showed symptoms, he added, which enabled early action to treat underlying disease.
Snyder also pioneered the use of wearable sensors to collect daily data on biological indicators, such as heart rate and its variability, respiration, skin temperature, blood oxygen, and blood pressure. Smartwatches can collect other clinical biomarkers, such as hemoglobin levels, red blood cell counts, and fasting glucose, which he admitted are no substitute for diagnosis-grade measurements and health care consultation but can provide clues that something may be awry. For example, his research team developed an algorithm to detect infectious disease illness, such as Lyme and COVID-19, presymptomatically based on changes in resting heart rate data.
INNOVATIVE METHODOLOGIES AND TECHNOLOGIES
The August 11 session opened with an overview of the industry landscape followed by a review of a selection of innovative methodologies and technologies for personalized nutrition, which were grouped into four sets of scales—genetic, physiologic/microbiome, individual, and social-ecologic—each discussed by three speakers.
Industry Landscape in Personalized Nutrition
Mariëtte Abrahams, Qina, reviewed the past decade’s journey of the personalized nutrition industry. Abrahams described a gradual shift in focus from nutrigenetics in 2012 to the current more actionable, behavior-change-oriented solutions, highlighted by two key randomized controlled trials (RCTs) (Celis-Morales et al., 2017; Zeevi et al., 2015) that demonstrated the value of personalized nutrition and by investments in companies that pioneered and advanced the industry. By 2018, consumer interest was rising, she recalled, and the COVID-19 pandemic drove further demand for telehealth, smart eating apps, supplements, and content about how to improve personal health.
Abrahams estimated that nearly 400 companies are part of the 2021 personalized nutrition market, which she described as highly fragmented and difficult to track in terms of solutions offered and companies’ movement within segments. Companies are placing increasing emphasis on behavior-change techniques and adding services, such as connections with registered dietitians, to help people implement personalized advice. They are also using data to inform an optimal interface for interaction and the format and frequency of information delivery, which Abrahams said contribute to consumer success with behavior change.
Abrahams listed key market trends, including scientific advances driving new solutions, particularly around improved cardiometabolic health and microbiome composition; new entrants to the market, such as the pharmaceutical industry; wider applications of solutions, such as continuous glucose monitors for patients who do not have diabetes or elevated glucose but want to track their responses to foods; and demand for health care practitioner expertise in both understanding the science and technologies underlying personalized nutrition and translating the guidance into tailored individual behavior-change strategies.
In Abrahams’s view, the future of personalized nutrition will include emphasis on food as medicine/food for optimal health, with solutions that integrate grocery shopping lists and meal plans; retailers will play a larger role in helping consumers improve health, such as by providing in-store guidance with food selection; interest in sustainability will increase consumer demand for plant-based, consciously produced food options; and new regulations will be implemented to keep users’ data safe and promote appropriate controls on providing medical advice.
Overall, Abrahams said, this market is dynamic and growing, but she said that conducting more inclusive research and providing equitable access to such solutions are crucial.
The Genetic Scale
Denise Ney, University of Wisconsin–Madison, discussed population-wide newborn screening for inborn errors of metabolism (IEM) using whole exome sequencing (WES). Ney maintained that newborn screening for IEM provides an ideal model for evaluating the role of sequencing in population screening. WES consists of sequencing the protein-coding regions of genes, she explained, which are thought to be the location of most known pathogenic mutations. Ney reported results of a study that compared the sensitivity and specificity of WES in newborn screening for IEM with those of tandem mass spectrometry (MS/MS), the established technique for such programs (Adhikari et al., 2020). The study indicated that WES alone is insufficiently sensitive or specific to be the primary method for most newborn screening for IEM but that it offers advantages when used as a secondary test for infants with abnormal MS/MS screens. For example, its ability to identify false positives as a second-tier test can eliminate the need for further biochemical and clinical studies, which can be costly and stressful for families.
Jim Kaput, Vydiant, explained that single nucleotide polymorphisms (SNP) typically do not have effect sizes sufficient for explaining an individual’s response to a nutrient or susceptibility to disease. He and his collaborators sought to understand the genetic contribution to plasma vitamin B12 levels using combinations of SNPs. They assessed polygenic risk scores (PRSs), which are combinations of publicly available SNPs, to enhance understanding of these outcomes specific to meeting vitamin B12 needs. Researchers identified 36 SNPs within the genetic pathway of vitamin B12 metabolism that could be used for PRS analysis; Kaput reported that the PRSs of those SNPs explained 42 percent of phenotype (Fuzo et al., 2021).
As a proof of concept for the use of PRSs in personalized diets, Kaput explained that individuals could be stratified into terciles of PRSs that correlated with vitamin B12 levels. Those with a high PRS were associated with high baseline vitamin B12 levels, indicating a low risk for needing additional vitamin B12. Individuals in the middle tercile are
at mild risk. Those in the lowest tercile were at the highest risk, which Kaput said calls for screening, close monitoring, and regular assessment of B12 levels to develop specific dietary recommendations.
Ahmed El-Sohemy, University of Toronto and Nutrigenomix Inc., reviewed several misconceptions that he said are often raised by people skeptical of the value of using genetics to personalize nutrition guidance. The first is that single SNPs are useless; El-Sohemy contended that in some cases, single SNPs can meaningfully affect an individual’s response to a nutrient. A second is that people will not change their behaviors. A third is that more RCT evidence is needed, he continued, but such evidence now available suggests that offering certain types of genetic information coupled with actionable recommendations motivates sustainable behavior change. A fourth misconception is that genetic test results are too complex for people to understand, which El-Sohemy agreed is likely true if only raw data are provided but suggested that a health care practitioner translating the results can help provide practical strategies for behavior changes. A fifth misconception is that family history is more informative than individual genetic makeup, but families also share common environments, he pointed out, and without an individual’s specific genetic information, it is not possible to know how each offspring has inherited specific genotypes that might affect the response to a dietary intervention. El-Sohemy shared his belief that the evidence base is more robust for genetics than other omics technologies, but because the availability and strength of evidence varies by genetic markers, it is important to distinguish where recommendations can and cannot be made.
The Physiologic/Microbiome Scale
Sarah Berry, King’s College London, described the Personalized Responses to Dietary Composition Trial (PREDICT) program’s use of big data and novel technologies to advance nutrition research and development of personalized nutrition programs.
Berry focused on the program’s first study, ZOE PREDICT 1, which used genetic, metabolomic, metagenomic, and meal context information to predict individuals’ postprandial responses to food. Participants completed a baseline clinic visit where they consumed test meals and underwent multiple assessments and biospecimen collections, then entered a 14-day home-based phase where they consumed both standardized and unstandardized meals, used digital devices to monitor continuous glucose levels and other parameters, provided biological specimens, and tracked dietary intake and satiety levels. The large volume of study data enabled investigators to explore the relative contribution of different exposures to a variety of outcomes, Berry explained, which led them to find that the importance of an exposure varied by outcome.
Berry said that ZOE, a start-up health care company that specializes in machine learning and artificial intelligence, used the PREDICT 1 results to create a machine learning model that uses an individual’s results from an at-home test to predict their responses to food and then delivers personalized dietary guidance. As for the future of ZOE’s PREDICT program, Berry said that the research team is “sitting on a goldmine of data” and has only just scratched the surface of its potential to improve personalized nutrition solutions.
Michal Rein, Weizmann Institute of Science and University of Haifa (Israel), described a unique approach to personalize nutrition guidance by predicting glycemic responses. She recounted a foundational study in which researchers measured 2-hour postprandial glucose responses (PPGR) in 1,000 healthy participants and found high interindividual variation to the same meal; responses were reproducible within participants. Researchers used these results and participants’ clinical and microbiome data to develop a machine learning algorithm that accurately predicted personalized PPGRs to any food combination. A short-term intervention tested the effects of personalized dietary intervention based on the algorithm, she recounted, and indicated that the diets lowered PPGR in people with prediabetes (Zeevi et al., 2015).
Rein described three subsequent interventions that assessed the clinical impact of a personalized post-prandial targeting (PPT) diet and supported its potential to improve glycemic control and metabolic health. Their results indicated that a PPT diet improved glycemic control and metabolic parameters, relative to a Mediterranean diet, in individuals with prediabetes (Ben-Yacov et al., 2021). In another unpublished pilot, a PPT diet induced diabetes remission in a majority of study participants (newly diagnosed type 2 diabetes patients not taking glucose-lowering medication); and in an ongoing intervention done by the company DayTwo, personalized diets improved glycemic control and reduced body weight in a majority of study participants (type 2 diabetes patients taking glucose-lowering medications).
Guru Banavar, Viome, highlighted two key technologies that Viome uses to create personalized food and supplement recommendations. The first, metatranscriptomics, is the measurement and detection of mRNA molecules across a range of organisms from microbes to humans. Viome has a scalable platform for extracting mRNA data from a stool sample, he said, which is introduced to its second key technology, artificial intelligence and machine learning, to generate personalized recommendations. Banavar contended that genetic expression as measured by mRNA molecules is superior to genetics for making dietary recommendations, based on Viome’s research that has used this data to pre-
dict differences in glycemic response between two foods for two individuals such that single foods can be classified as “enjoy” or “minimize” for each. Furthermore, he continued, Viome has also layered gut microbiome pathway analysis—a method to assess levels of gut metabolites that influence an individual’s response to foods—onto the glycemic response data to distinguish foods as “super,” “enjoy,” “minimize,” or “avoid.”
The Individual Scale
Andres Acosta, Mayo Clinic, discussed the use of pathophysiological and behavioral phenotypes for guiding obesity management to enhance weight loss. Existing “one-size-fits-all” approaches are not working, he contended, pointing out wide variation between patients in weight loss in response to a given treatment, whether it is lifestyle, drug, device, or surgery based. “Precision obesity” approaches could be more effective, he proposed, when using unique phenotypes to predict response to various treatments.
The Mayo Clinic classifies patients with obesity based on their results to a series of tests to measure domains that influence energy intake and expenditure and identified four phenotypes: hungry brain, characterized by abnormal satiation; hungry gut, characterized by abnormal postprandial satiety; emotional hunger, characterized by eating in response to positive and/or negative emotions; and slow burn, characterized by an abnormal metabolism. In a sample of 450 patients with obesity, around one-quarter had more than one phenotype and about 15 percent had none (Acosta et al., 2021).
Acosta reported that the Mayo Clinic has multiple proof of concept, placebo-controlled clinical trials to demonstrate that the phenotypes can enhance weight loss in people with obesity. For example, phenotype-guided pharmacotherapy in clinical practice can almost double weight loss at 12 months from baseline compared with standard pharmacotherapy (Acosta et al., 2021).
Samantha Kleinberg, Stevens Institute of Technology, reviewed efforts to develop passive, technology-based methods for tracking individual dietary intake. She relayed that devices to track meal timing include audio sensors (e.g., ear buds that track chewing sounds), motion sensors (e.g., wearables that measure head motion and arm- and wrist-to-mouth motions), and multimodal sensors that combine audio and motion. By combining data from lab and free-living environments, Kleinberg said that her lab was able to correctly identify each bite of food by type (e.g., steak, potato, or salad) with more than 80 percent accuracy.
Kleinberg acknowledged that major gaps remain; for example, these methods do not provide information about how people make food decisions, their levels of hunger and fullness, or the social and emotional context in which eating occasions occur. Another significant limitation is generalizability of the information, she added; although passive methods can reliably classify intake within 40–50 different types of food, the same types may vary considerably in composition between cultures and countries.
Diana Thomas, U.S. Military Academy at West Point, emphasized the richness of data collected in open-ended, free-form survey questions and analyzed with natural language processing. Thomas illustrated the value of such data with an example of a 26-question survey on weight bias that included only one question with a free-form text response: “in your opinion, what does the American public think about people with obesity?” Sentiment assessment is one of the first approaches that statisticians take to analyze free-form text, Thomas explained, with support from software packages with built-in dictionaries that assign various levels of positive or negative sentiment levels to different words and classify words into emotions. Mostly negative sentiment was detected, Thomas reported, and when the sentiment data were concatenated the survey’s Likert-style question in which respondents were asked to classify their self-perceived weight status, sentiment became more negative as that status increased. Thomas reiterated that this combination of question types provided depth of insight beyond either type alone.
The Social-Ecological Scale
Kayla de la Haye, University of Southern California, discussed using new sources of social data and network science to more precisely understand how social exposures and social networks influence eating habits.
Longitudinal social network research with network analytic methods has shown that nutritional health and people’s social networks are interdependent, de la Haye explained, because people tend to select and form social ties with those who are similar to themselves. She also described strong evidence that these networks influence people’s health behaviors, including eating, through social influence mechanisms, such as mimicry, normative influence, and social support. These dynamics of social selection, social influence, and corresponding shared exposures influence diet and health, she summarized: people with the greatest nutritional health risks are more likely to find themselves in networks that also confer health risks. These social exposures reinforce and influence nutritional health behaviors and can be barriers to behavior change or targeted in network interventions.
de la Haye proposed that another opportunity to improve eating habits is to use innovations in data, data science, and behavioral and social science to gain more precise insights into the social and built environments where people acquire and eat food. She suggested that this type of data could be leveraged to intervene in real time to help people make healthy choices in certain food environments to which they are exposed, for example.
Sean Duffy, Omada Health, shared examples of the company’s insights that illustrate how virtual care can improve health behaviors. He explained that when new members are asked what barriers and challenges they expect as they try to change their health behaviors, the temporal nature of their responses appears to be associated with the success they achieve in the program. For example, responses in the present tense, such as “I hate to cook,” are correlated with less weight loss, whereas responses in the past tense, such as “I used to eat extra portions of sweets,” are correlated with more. Omada termed the latter framing a “growth mindset,” Duffy explained, in which intelligence and talent are viewed as more flexible and changeable.
A second insight, Duffy continued, is that combining technology with human support appears to influence subtle changes in behavior. For example, the Omada team observed that when its coaches engaged directly with members after noticing an unconscious behavioral pattern, members modified their health behavior and their rapport with the coach improved. Duffy attributed the improvement to the human touchpoint and outreach of care and support. Based on that insight, Omada developed a successful strategy in which coaches identify patterns of member behavior and respond with personalized suggestions for improvements.
Michael Howell, Google, gave a presentation on the Google-developed datasets that epidemiologists and public health teams are using to study how people interact with their communities. He reviewed three categories of publicly available, anonymized, and aggregated datasets. The first category is aggregated, anonymized search queries, which he explained can provide insights into people’s information needs and how they change over time. Public health–related examples of search query resources are Google Trends, the COVID-19 Search Symptoms Dataset (geographic trends in searches for specific health symptoms), and the COVID-19 Vaccine Search Insights (trends from searches for vaccine topics, such as safety and side effects). A second category is maps for understanding distance and how it relates to travel time, which Howell explained can help provide information about built environments. Google has also created a COVID-19 Vaccination Access Dataset, which Boston Children’s Hospital and Ariadne Labs have used to identify “vaccine deserts” and calculate travel time to vaccine locations from various starting points. Howell’s third category of anonymized, aggregated datasets is community mobility, which he said have been used to assess trends in community mobility to various destinations during the COVID-19 pandemic.
IMPLEMENTATION OF PRECISION AND PERSONALIZED NUTRITION
The August 12 session discussed the implementation of precision and personalized nutrition and included academic, regulatory, and industry perspectives on opportunities and challenges.
Christina Roberto, University of Pennsylvania, discussed challenges to precision nutrition’s potential to achieve greater health equity. The discussions tend to focus on individual-level factors that affect health, she observed, but broader thinking is important because those interact with many environmental- and systems-level forces that influence health behaviors.
Roberto pointed out that, according to data published in the Dietary Guidelines for Americans, 2020–2025, many people do not adhere to dietary recommendations, and she expressed her doubt that targeted nutrition advice will solve the problem. People have limited memories and information processing requires effort, she explained, but most food decisions are made quickly and automatically, using associations and emotions. Simple nutrition guidelines have tremendous recall advantages, Roberto maintained, relaying an example in which researchers found that personalized guidelines about the number of daily servings for each food group were less memorable and actionable than a generalized guideline to fill half the plate with fruits and vegetables. Technologies that remind people of advice reduce recall burden, Roberto acknowledged, but she expressed concern about increasing disparities by relying on them, given that mobile devices and health apps are most likely to be used by younger, educated women.
Roberto also raised concern about precision nutrition being exploited for profit. Online shopping could be a positive or negative source of personalized marketing, she suggested, noting that some retailers already organize products by dietary preferences and use customers’ browsing and purchasing data to recommend products. This could be beneficial if the products are healthy foods, she pointed out, but could exacerbate poor eating habits if not.
Roberto ended with a reference to a framework for achieving health equity, which she proposed incorporating into conversations about implementing precision nutrition in a way that does not widen disparities. The framework emphasizes the importance of developing interventions that aim to increase healthy options and/or reduce deterrents to unhealthy behaviors, she explained, while also incorporating considerations related to social disadvantage and social determinants of health (Kumanyika, 2019).
Peter Lurie, Center for Science in the Public Interest, cautioned that exciting science alone is not enough to improve public health. He proposed that current federal regulatory structures are inadequate for supporting the translation of personalized nutrition science into meaningful public health gains.
Lurie supported this hypothesis by describing what he called “an almost perfect storm of poor regulation,” characterized by high demand for personalized nutrition products (particularly among higher-income, “worried well” consumers); a largely unregulated market for laboratory-developed tests (LDTs), a subset of diagnostics; a policy for “general wellness” products that provides “safe harbor” for many devices; and claims on food products that could lead consumers to believe that their health-promoting potential exceeds what evidence suggests.
Lurie elaborated on his concerns with the U.S. Food and Drug Administration (FDA) oversight of LDTs, maintaining that it has infrequently exercised its enforcement authority. He listed resulting deficiencies in oversight, such as inadequate adverse event reporting and product labeling, unsupported manufacturer claims, and no premarket review of performance data.
Lurie described four types of FDA-permitted claims on food as another element of regulation relevant to precision nutrition. Different types of claims require different levels of supporting scientific evidence, he explained, but claim language sometimes blurs the line between the types and may be misconstrued by consumers. For example, health claims convey a relationship between an ingredient and a disease and require “significant scientific agreement” and preapproval, whereas structure/function claims describe the effect on normal body structure or function and require only “competent and reliable” evidence and no preapproval.
Lurie ended by highlighting that a challenge for personalized nutrition is to determine when it makes more sense than a population-based approach; he maintained that the choice between the two should be guided by which will have the most public health impact.
Robert Califf, Verily and Google Health, provided a regulatory perspective on challenges and opportunities for precision nutrition. FDA’s mission is to preserve and protect the public health, he began, and although it is science based, Califf emphasized that it is tasked with making decisions even when more research would be helpful to clarify an optimal choice.
Califf highlighted that the current environment offers real-time access to massive amounts of data and the ability to rapidly analyze data to gain insights and guidance. This environment enables integrating comprehensive health data, which he said is valuable for both assessing the entire human being and interacting components and generating evidence to inform public health decisions.
To build on his point about the value of comprehensive health data, Califf explained that individual biomarkers are unlikely to predict a food’s effect on health except in cases of specific nutritional deficiencies and that substantial evidence must exist for a biomarker of one measure to serve as a surrogate (i.e., a substitute for a clinical end point). The vast majority of biomarkers are not valid surrogates, he pointed out, because they exist in a milieu of biological complexity. The effect of food on health is a prime example of this complexity, he added, because many nutrients and substances are simultaneously interacting, each affecting multiple pathways and each with a potential impact on health. Biomarker tests need to be reliable, reproducible across multiple laboratories and clinical settings, and of adequate sensitivity and specificity before data based on them can be used for evaluation. Otherwise, Califf cautioned, personalized recommendations could cause harm that may not be apparent until many years later.
Josh Anthony, Nlumn, shared lessons learned from his experience in the personalized nutrition industry. As providers rush to understand how they can fit into the rapidly growing market’s ecosystem, he began, they often make the mistake of putting themselves and their brand at the center instead of starting with understanding consumer needs, behaviors, and values.
Consumers seeking personalized nutrition approaches hold widely different views about what health means, Anthony explained, with many embracing scientific controversies and seeking self-affirming information or groups aligned with their beliefs and a smaller segment seeking objective, data-driven personalization. Consumer expectations of personal nutrition’s delivery of solutions are broad and inconsistent, he continued, ranging from managing chronic disease risk to maintaining sustained energy and reducing stress levels.
Anthony described a model of personalized nutrition that supports creating and sustaining engagement with a wide variety of potential users. The model calls for first selecting the optimal health outcome of interest for the consumer, then collecting an objective, validated measure of health or function. The consumer’s result for that measure leads to personalized recommendations to improve health and lifestyle, which are intended to drive behavior change. The cyclical, continuous nature of the process supports reassessment and new advice, he said, as the consumer sets new outcome goals and/or health measures data change.
Regarding the state of the science of personalization, Anthony suggested that companies can differentiate themselves by committing to evidence-based practices, although he observed that most do not have published data to support
their specific program and its purported outputs. He said personalized nutrition is most successful when the program delivers health and functional benefits that exceed benefits from “one-size-fits-all” approaches and are sustained over time. In his view, most of the programs that are more rooted in science are starting to show some health and functional benefits to users but have yet to demonstrate that their outcomes are sustainable and superior to those from broader approaches.
Martin Hahn, Hogan Lovells, discussed communicating personalized nutrition information to the public while maintaining compliance with the current regulatory framework. Hahn explained that FDA regulates food products based on intended use. That determines whether a product is regulated as a food or a dietary supplement, for example, and what type of claim it is permitted to carry. Hahn opined that tension exists between identifying claim language that effectively conveys the intended message while staying within regulatory guardrails.
Hahn underscored that messages to communicate the benefits of personalized and precision nutrition must be truthful and not misleading, supported by competent, reliable scientific evidence for both explicit and implied messages conveyed by claims. Hahn pointed out a paradox of structure/function claims, which is that many imply disease prevention. For instance, he recognized that the approved claims “calcium builds strong bones” and “whole oats support heart health” imply reduced risk for osteoporosis and coronary heart disease, respectively, potentially conveying a message that may not have been intended.
Hahn contended that the current regulatory system is not well suited to allow personalized nutrition technologies to communicate their benefits without risking allegations of misleading consumers or being regulated in a way that is inconsistent with their intended use. As an example, Hahn explained that a genetic analysis used to identify a diet that helps reduce the risk of developing diabetes could be communicated through an FDA-approved health claim. But if it is used to identify a diet that reduces insulin dependence for an individual with diagnosed diabetes, then it is communicating a treatment effect and warrants a disease claim, which would trigger filing as a drug. The new drug approval process is lengthy and expensive, he said, and could stifle innovation because it is not a viable option for many of the personalized nutrition technologies under way. The product could more safely make a structure/function claim, he added (e.g., that it supports blood glucose levels within normal ranges), but this would not communicate its full benefits and likely would not attract as much consumer interest.
Hahn espoused the value of a stakeholder dialogue to discuss a strategy for supporting a regulatory structure that allows personalized nutrition technologies to advance and, when they have a sufficient evidence base, to clearly and effectively communicate their benefits to target populations without risking unintended regulatory consequences. This could involve considering new legislation to give FDA the authority to develop a new regulatory framework for personalized nutrition solutions, he suggested, or adapting existing regulatory frameworks, such as those governing foods for special dietary uses or medical foods.
Patsy Brannon, Cornell University, discussed challenges and opportunities for incorporating precision nutrition into both public health guidelines and individualized guidelines. For context, she contrasted food- and nutrient-based guidelines for public health with precision nutrition–based recommendations for individuals in terms of the types of actions they support, the way they are implemented, and the strength of evidence required.
Brannon reviewed the expanded Dietary Reference Intakes (DRIs) model and highlighted opportunities for precision nutrition to help build stronger reference intake values. She listed 12 nutrients for which the distribution of need is not yet understood, primarily because no adequacy outcome has been identified. Precision nutrition could help build stronger reference intake values for these nutrients, she proposed, as research proceeds to identify biomarkers and their variance. When a DRI for chronic disease is warranted, precision nutrition could help identify surrogate outcomes for chronic disease.
Another opportunity, Brannon continued, is to determine population subgroups with specific nutrient needs, such as subgroups that have distinctly different distributions of requirements for a nutrient and responses to intake of that nutrient and subgroups of responders and nonresponders to a particular nutrient. Precision nutrition could also fill the research gap of defining a “healthy population” when the prevalence of chronic disease is high, which would promote better informed public health guidelines for nutrients and dietary patterns, Brannon explained, and enhance understanding of variance within a healthy population.
Brannon raised the issue of whether individual algorithms can be linked to public health guidelines so that a person’s specific nutrient needs could be used to tailor DRIs and Dietary Guidelines for them. The strength of evidence required for such tailoring remains to be determined, she noted, and individual guidelines based on precision nutrition also raise the issue of efficacy versus effectiveness. Brannon suggested that the effectiveness of some precision nutrition approaches may be related to participants’ self-selection bias. She encouraged the consideration of behavior-change theories, as they could provide insight into participant characteristics and the subgroup of participants who are successful. Last, she raised the questions of how to ensure equitable access for people at the highest risk of poor health and that practitioners receive an appropriate scope and depth of training to implement individual guidelines. ◆◆◆
Acosta, A., M. Camilleri, B. Abu Dayyeh, G. Calderon, D. Gonzalez, A. McRae, W. Rossini, S. Singh, D. Burton, and M. M. Clark. 2021. Selection of antiobesity medications based on phenotypes enhances weight loss: A pragmatic trial in an obesity clinic. Obesity (Silver Spring) 29(4):662–671.
Adhikari, S., E. C. Nice, E. W. Deutsch, L. Lane, G. S. Omenn, S. R. Pennington, Y-K Paik, C. M. Overall, F. J. Corrales, I. M. Cristea, J. E. Van Eyk, M. Uhlén, C. Lindskog, D. W. Chan, A. Bairoch, J. C. Waddington, J. L. Justice, J. LaBaer, H. Rodriguez, F. He, et al. 2020. A high-stringency blueprint of the human proteome. Nature Communications 11(1):5301.
Ben-Yacov, O., A. Godneva, M. Rein, S. Shilo, D. Kolobkov, N. Koren, N. Cohen Dolev, T. Travinsky Shmul, B. C. Wolf, N. Kosower, K. Sagiv, M. Lotan-Pompan, N. Zmora, A. Weinberger, E. Elinav, and E. Segal. 2021. Personalized postprandial glucose response-targeting diet versus Mediterranean diet for glycemic control in prediabetes. Diabetes Care 44(9):1980–1991.
Celis-Morales, C., K. M. Livingstone, C. F. Marsaux, A. L. Macready, R. Fallaize, C. B. O’Donovan, C. Woolhead, H. Forster, M. C. Walsh, S. Navas-Carretero, R. San-Cristobal, L. Tsirigoti, C. P. Lambrinou, C. Mavrogianni, G. Moschonis, S. Kolossa, J. Hallmann, M. Godlewska, A. Surwillo, I. Traczyk, et al. 2017. Effect of personalized nutrition on health-related behaviour change: Evidence from the Food4Me European randomized controlled trial. International Journal of Epidemiology 46(2):578–588.
Fuzo, C. A., U. da Veiga, S. Moco, O. Cominetti, S. Métairon, S. Pruvost, A. Charpagne, J. Carayol, R. Torrieri, W. A. Silva, Jr., P. Descombes, J. Kaput, and J. P. Monteiro. 2021. Contribution of genetic ancestry and polygenic risk score in meeting vitamin B12 needs in healthy Brazilian children and adolescents. Scientific Reports 11(1):11992.
Kumanyika, S. K. 2019. A framework for increasing equity impact in obesity prevention. American Journal of Public Health 109:1350–1357.
NIH (National Institutes of Health). 2017. Story of discovery—APOL1 gene variants: Unraveling the genetic basis of elevated risk for kidney disease in African Americans. Washington, DC: National Institute of Diabetes and Digestive and Kidney Diseases. https://www.niddk.nih.gov/news/archive/2017/story-variants-unraveling-genetic-basis-elevated-risk-kidney-disease-african-americans (accessed September 27, 2021).
Zeevi, D., T. Korem, N. Zmora, D. Israeli, D. Rothschild, A. Weinberger, O. Ben-Yacov, D. Lador, T. Avnit-Sagi, M. Lotan-Pompan, J. Suez, J. A. Mahdi, E. Matot, G. Malka, N. Kosower, M. Rein, G. Zilberman-Schapira, L. Dohnalová, M. Pevsner-Fischer, R. Bikovsky, et al. 2015. Personalized nutrition by prediction of glycemic responses. Cell 163(5):P1079–P1094.
DISCLAIMER: This Proceedings of a Workshop—in Brief was prepared by Emily A. Callahan as a factual summary of what occurred at the workshop. The statements made are those of the rapporteur or individual workshop participants and do not necessarily represent the views of all workshop participants, the planning committee, or the National Academies of Sciences, Engineering, and Medicine.
*The National Academies of Sciences, Engineering, and Medicine’s planning committees are solely responsible for organizing the workshop, identifying topics, and choosing speakers. The responsibility for this published Proceedings of a Workshop—in Brief rests with the rapporteur and the institution.
REVIEWERS: To ensure that it meets institutional standards for quality and objectivity, this Proceedings of a Workshop—in Brief was reviewed by Josiemer Mattei, Harvard T.H. Chan School of Public Health, and Robin McKinnon, U.S. Food and Drug Administration. Leslie Sim, National Academies of Sciences, Engineering, and Medicine, served as the review coordinator.
SPONSORS: This workshop was partially supported by the American Heart Association; American Institute for Cancer Research; American Society for Nutrition; Cargill, Inc.; Coca-Cola Company; Conagra Brands; Center for Science in the Public Interest; Danone North America; General Mills, Inc.; Keurig Dr Pepper; Mars, Inc.; Mondelēz International; National Institutes of Health; Ocean Spray Cranberries, Inc.; Unilever; U.S. Department of Agriculture; and U.S. Food and Drug Administration, with additional support from the Academy of Nutrition and Dietetics.
For additional information regarding the meeting, visit https://www.nationalacademies.org/our-work/food-forum.
Suggested citation: National Academies of Sciences, Engineering, and Medicine. 2021. Challenges and opportunities for precision and personalized nutrition: Proceedings of a workshop—in brief. Washington, DC: The National Academies Press. https://doi.org/10.17226/26407.
Health and Medicine Division
Copyright 2021 by the National Academy of Sciences. All rights reserved.