Understanding dietary patterns is crucial for preventing and treating chronic diseases like obesity, type 2 diabetes, and cardiovascular disease. Numerous studies have shown the benefits of specific diets, such as the Mediterranean diet and the DASH diet. However, current dietary guidelines largely focus on individual nutrients and food groups, overlooking the complex relationships within the human diet. This study addresses this gap by exploring the concept of nutritional redundancy. The research question focuses on whether the stability of overall nutrient intake, despite individual food intake variability, is a significant factor in health outcomes. The study's purpose is to define and quantify nutritional redundancy, investigate its relationship with established healthy diet scores, and assess its predictive power for healthy aging and chronic disease risk. The importance of this study lies in its potential to offer a novel perspective on dietary assessment and its application in personalized nutrition and disease prevention.

Existing literature extensively documents the benefits of specific dietary patterns on health outcomes. Randomized controlled trials have demonstrated the positive effects of the Mediterranean diet on cardiovascular health and the DASH diet on blood pressure. Nutrient profiling, a key area of research, uses databases like the USDA's Food and Nutrient Database for Dietary Studies (FNDDS) to understand nutrient composition in foods. However, the existing literature largely focuses on specific nutrients or food groups, overlooking the potential for redundancy in nutrient intake. This study builds upon this foundation by examining the broader network of food-nutrient relationships to quantify and understand the phenomenon of nutritional redundancy.

This study analyzed five datasets with varying dietary data collection methods and time scales: 1. Diet-microbiome association study (DMAS): Daily dietary intake data from 30 healthy individuals over 17 days using ASA24. 2. Nurse's Health Study (NHS): Eight time points of FFQ data from 35,256 female nurses. 3. Health Professionals Follow-up Study (HPFS): Seven time points of FFQ data from 17,529 male health professionals. 4. Women's Lifestyle Validation Study (WLVS): Four ASA24 records within 1 year from 216 women. 5. Men's Lifestyle Validation Study (MLVS): Four ASA24 records within 1 year from 451 men. The authors calculated the relative abundance of foods and nutrients for each individual in each dataset. To quantify the variability in food and nutrient profiles, they used beta diversity measures (Bray-Curtis dissimilarity, root Jensen-Shannon divergence, Yue-Clayton distance, and negative Spearman correlation). A food-nutrient network (FNN) was constructed, represented as a bipartite graph connecting foods and nutrients. Nutritional redundancy (NR) was defined as the difference between food diversity (FD, using the Gini-Simpson index) and nutrient diversity (ND, using Rao's quadratic entropy). The authors then assessed the correlation between NR and existing healthy diet scores (HEI-2005, AHEI-2010, AMED, DASH). They also examined the association between NR and various host factors, and used a random forest classifier to predict healthy aging status based on NR and other factors. Finally, Cox proportional hazard models were used to assess the association between NR and risks of type 2 diabetes and cardiovascular disease in NHS and HPFS.

The study revealed that food profiles were highly dynamic and personalized, while nutrient profiles were remarkably stable across individuals and time. This phenomenon, termed nutritional redundancy (NR), was quantified using a novel measure derived from the food-nutrient network. NR values were around 0.3 in all five cohorts, indicating comparable nutritional redundancy and diversity. NR did not strongly correlate with classical healthy diet scores, suggesting they capture different aspects of dietary intake. However, NR showed comparable performance in predicting healthy aging to existing healthy diet scores. After adjusting for age, higher NR was associated with a lower risk of cardiovascular disease and type 2 diabetes in both NHS and HPFS. This association persisted even after adjusting for several confounding factors in HPFS. Analysis of food consumption patterns showed that individuals with high NR consumed more fruits, vegetables, and dairy products, while consuming less beverages. Randomization tests indicated that the observed NR was not simply due to random food choices but reflected inherent structural features of the food-nutrient network, particularly its nestedness.

The findings address the research question by demonstrating that nutritional redundancy, despite variability in food intake, is a significant factor in predicting healthy aging and reducing the risk of chronic diseases. The significant association between NR and lower disease risk suggests that the diversity of nutrient sources may be more important for health than the diversity of food choices. The comparable predictive power of NR and existing healthy diet scores highlights the potential of NR as a complementary metric in dietary assessment. This research challenges the traditional focus on specific food groups and nutrients, emphasizing the importance of considering the overall nutrient profile and the structure of the food-nutrient network in dietary guidance and disease prevention.

This study introduces the concept of nutritional redundancy (NR) and demonstrates its potential as a novel metric in nutritional epidemiology. NR offers a new perspective on dietary assessment, considering the complex relationship between food choices and nutrient intake. Future research should explore NR's application in dietary interventions and investigate the mechanisms underlying its association with health outcomes. Further research is needed to refine NR calculation methods, incorporate more complete nutrient data, and explore its application across diverse populations.

The study relies on self-reported dietary intake data (ASA24 and FFQ), which are prone to measurement error. The nutritional components considered in the FNN are limited by the completeness of existing databases, potentially omitting bioactive molecules and variations in nutrient absorption. While several null models were tested, exploring additional network characteristics could offer further insights into NR. The food distance measure used might introduce biases due to correlated nutrients. Future studies using objective measures of nutrient intake (biomarkers) and more comprehensive food composition data are recommended to validate and refine these findings.