Open Access

Nearly all approaches to personalized nutrition (PN) use information such as the gene variants of individuals to deliver advice that is more beneficial than a generic one-size-fits-all recommendation. Despite great enthusiasm and the increased availability of commercial services, thus far, scientific studies have only revealed small to negligible effects on the efficacy and effectiveness of personalized dietary recommendations, even when using genetic or other individual information. In addition, from a public health perspective, scholars are critical of PN because it primarily targets socially privileged groups rather than the general population, thereby potentially widening health inequality. Therefore, in this perspective, we propose to extend current PN approaches by creating adaptive personalized nutrition advice systems (APNASs) that are tailored to the type and timing of personalized advice for individual needs, capacities, and receptivity in real-life food environments. These systems encompass a broadening of current PN goals (i.e., what should be achieved) to incorporate individual goal preferences beyond currently advocated biomedical targets (e.g., making sustainable food choices). Moreover, they cover the personalization processes of behavior change by providing in situ, just-in-time information in real-life environments (how and when to change), which accounts for individual capacities and constraints (e.g., economic resources). Finally, they are concerned with a participatory dialogue between individuals and experts (e.g., actual or virtual dieticians, nutritionists, and advisors), when setting goals and deriving measures of adaption. Within this framework, emerging digital nutrition ecosystems enable continuous, real-time monitoring, advice, and support in food environments from exposure to consumption. We present this vision of a novel PN framework along with scenarios and arguments that describe its potential to efficiently address individual and population needs and target groups that would benefit most from its implementation.

Background For a growing number of food-based dietary guidelines (FBDGs), diet optimization is the tool of choice to account for the complex demands of healthy and sustainable diets. However, decisions about such optimization models’ parameters are rarely reported nor systematically studied. Objectives The objectives were to develop a framework for (i) the formulation of decision variables based on a hierarchical food classification system; (ii) the mathematical form of the objective function; and (iii) approaches to incorporate nutrient goals. Methods To answer objective (i), food groups from FoodEx2 levels 3-7 were applied as decision variables in a model using acceptability constraints (5th and 95th percentile for food intakes of German adults (n = 10,419)) and minimizing the deviation from the average observed dietary intakes. Building upon, to answer objectives (ii) and (iii), twelve models were run using decision variables from FoodEx2 level 3 (n = 255), applying either a linear or squared and a relative or absolute way to deviate from observed dietary intakes, and three different lists of nutrient goals (allNUT-DRV, incorporating all nutrient goals; modNUT-DRV excluding nutrients with limited data quality; modNUT-AR using average requirements where applicable instead of recommended intakes). Results FoodEx2 food groups proved suitable as diet optimization decision variables. Regarding deviation, the largest differences were between the four different objective function types, e.g., in the linear-relative modNUT-DRV model, 46 food groups of the observed diet were changed to reach the model’s goal, in linear-absolute 78 food groups, squared-relative 167, and squared-absolute 248. The nutrient goals were fulfilled in all models, but the number of binding nutrient constraints was highest in the linear-relative models (e.g. allNUT-DRV: 11 vs. 7 in linear-absolute). Conclusion Considering the various possibilities to operationalize dietary aspects in an optimization model, this study offers valuable contributions to a framework for developing FBDGs via diet optimization.