Methodology Brief

USDA FoodData Central: when to use Foundation Foods vs. Survey (FNDDS) vs. SR Legacy entries

A methodology brief

Sofia Rivera, MS, RD
Published December 4, 2024

Background

USDA FoodData Central (FDC), introduced in 2019, consolidates several historically separate food composition data types under a single interface. Users coming from the older Standard Reference (SR) world may not immediately appreciate that an identical food name (for example, “yogurt, plain, whole milk”) can return entries from Foundation Foods, FNDDS (Survey), SR Legacy, and branded entries, each with distinct provenance. Selecting the wrong data type can introduce systematic bias of several percent in energy and substantially more in micronutrients.

The intent of this brief is not to re-describe FDC documentation but to provide a practical decision rule for which data type to use in Initiative validation and secondary-analysis work.

The Method

Initiative selection rules, in order of preference for a given food:

1. Foundation Foods, when available, are the default for weighed-food validation reference analysis and for bench-top nutrient comparisons. Foundation entries carry analytic provenance, sample metadata, and uncertainty indicators, and are the USDA’s highest-curation data type.
2. FNDDS (Survey), which underlies NHANES dietary data, is the default when the analytic objective is comparability to NHANES or other survey-based consumption estimates. FNDDS entries are recipe-based and incorporate retention and yield factors relevant to as-consumed foods.
3. SR Legacy is used when neither Foundation nor FNDDS includes the food of interest, and when the analyst has reviewed whether a more recent Foundation entry has superseded the SR value. SR Legacy is no longer updated; the Initiative treats it as a secondary source.
4. Branded Foods are used for label-reconciliation studies and commercial-product work, but not as the primary reference in weighed-food validation because values are derived from manufacturer labels, which carry their own rounding and compliance allowances.

When two data types contain the same food and values disagree by more than a threshold (Initiative convention: 10% on energy, 20% on a micronutrient of interest), the discrepancy is recorded and the Foundation value is used. All FDC IDs used in a study are reported in the supplement.

Worked example

Consider “chicken, breast, cooked, roasted, meat only” for a weighed-food validation targeting protein and energy.

Under Initiative rules, the Foundation entry is selected for validation work; the FNDDS entry would be used if the same study were being harmonised against NHANES. The branded entry’s higher protein is consistent with label rounding and is not used as reference.

Common pitfalls

• Treating FDC as a single database and not recording which data type a value came from. Within-study switching between data types is a known source of non-reproducibility.
• Selecting FNDDS for a bench-top nutrient validation. FNDDS values may embed retention and yield factors inappropriate for a controlled kitchen preparation.
• Using branded entries for nutrient totals in validation work; label values round to the FDA’s rounding rules and do not reflect analytic measurement.
• Failing to document the FDC download date. FDC is updated on a schedule; reproducibility requires snapshot dates or the specific release version.
• Ignoring SR Legacy deprecation. Some SR entries have been superseded by Foundation values; analysts should re-check each food at study start.

Recommended reporting

• Report the FDC data type used for each food, not just the food name.
• Report the FDC release version or download date.
• If multiple data types were used within one study, justify the mixing and report a sensitivity analysis with a single-data-type alternative.
• Include FDC IDs in a supplementary table.
• Where a food was absent from FDC and a substitution was made, document the substitution rule.

References

1. Rivera M. A practical guide to FoodData Central for validation studies. J Nutr. 2023;153(10):2988-2995.
2. Ahlgren P, Santos J. Discrepancies between Foundation Foods and SR Legacy for selected produce items. Am J Clin Nutr. 2022;115(4):1102-1111.
3. Weiss R, Okafor N. Reproducibility of nutrient exposure estimates across FDC data types. Nutrients. 2023;15(18):3977.
4. Pelletier L, Hernandez A. FNDDS retention factors and their effect on cooked-food energy estimation. Public Health Nutr. 2021;24(12):3688-3697.
5. Branded-label rounding and its impact on commercial food composition datasets. Br J Nutr. 2020;123(11):1280-1289.
6. Rivera M, Patel R. Substitution rules for foods absent from national databases: a sensitivity analysis. Nutrients. 2024;16(3):412.
7. Holm T. Versioning and snapshotting of food composition data for reproducible nutrition research. Am J Clin Nutr. 2022;116(2):310-318.