Advances in Nutrition · 4 authors, 2 centres
This summary was generated by AI from a single paper. It has not been reviewed by a clinician and is not clinical advice. Verify against the source before acting on it.
This review evaluates the quality and FAIRness (findable, accessible, interoperable, reusable) of nutrient databases, using the USDA SR Legacy as a gold standard. It finds that SR Legacy is incomplete for both Nutrition Facts Panel (mean 84.9%) and NASEM essential nutrients (mean 70.3%), and phytonutrient coverage is minimal. The authors call for incorporating data science principles to improve database completeness and FAIRness, which is critical for advancing precision nutrition.
**Background:** Nutrient databases are foundational for nutrition science and precision nutrition (PN), which aims to personalize dietary recommendations. However, the quality and FAIRness (findable, accessible, interoperable, reusable) of these databases have not been systematically evaluated. This review assesses the completeness and FAIRness of current food composition data, using the USDA SR Legacy as a surrogate gold standard, and proposes improvements.
**Methods:** SR Legacy was downloaded from FoodData Central on 1 December 2021. Completeness was defined as having values for all 15 Nutrition Facts Panel (NFP) measures and all 40 NASEM essential nutrients. Phytonutrient data were collected from the 4 USDA Special Interest Databases (flavonoids, proanthocyanidins, isoflavones) on 4 October 2022, covering 38 phytonutrient measures. For FAIRness evaluation, 175 food and nutrient data sources were identified from a search of the top 5 nutrition journals (2020 publications) and additional sources. Data sources were characterized for findability, accessibility, interoperability, and reusability.
**Key Results:** - **Completeness:** SR Legacy had a mean availability of 84.9% for NFP nutrients (range: 53.6% for transfat to 100% for calories and total fat) and 70.3% for NASEM essential nutrients (range: 66.5% for vitamin D to 94.8% for vitamin A). By food group, mean NASEM nutrient availability ranged from 66.3% (fruit) to 79.2% (meats). Phytonutrient coverage was limited: e.g., total isoflavone was provided for 544 foods, quercetin for 3161 foods, out of 38 measures across the 4 databases. - **FAIRness:** Of 175 data sources, 88.0% were findable; 68.4% of URLs worked at collection. Of 154 findable sources, 78.6% were accessible without payment; 43.8% were exportable in Excel/CSV, 11.6% in Access/SAS, 47.1% as PDFs, and 9.7% were viewable but not exportable. Interoperability was limited by lack of documentation on linking identifiers. For reusability, 48.9% of food composition data citations included version or retrieval date; 51.1% did not. - **Data source characterization:** USDA databases were incorporated into 57.7% of sources (84% if including likely USDA data). Sources originated from 48 countries; 43.0% from North America (37.1% US, 6.9% Canada). Food composition data was the most common type (56.6%).
**Clinical Implications:** The incompleteness and poor FAIRness of current nutrient databases undermine the reliability of dietary assessments and precision nutrition research. Improving data completeness (including all essential nutrients and phytonutrients) and FAIRness (persistent identifiers, accessible formats, clear documentation, citation standards) is essential for robust PN. The authors recommend collaboration between nutrition and data scientists, adoption of FAIR principles, and setting high-quality standards in large studies like the NIH Nutrition for Precision Health study.