Skip to main content
NUTRITION 2026 Main banner

(OR13-03-26) A Harmonized Additive Detection Framework for Screening Ultra-Processed Foods in Eight Countries

Sunday, July 26, 2026
2:54 PM - 3:06 PM ET
Location: National Harbor 12-13
Emerging Leaders in Nutrition Science Poster Competition

    Presenting Author(s)

  • Muyao Li, MSPH

    MSPH Graduate
    Johns Hopkins Bloomberg School of Public Health
    Baltimore, Maryland, United States

    Disclosure(s): No relevant financial relationship(s) with ineligible companies to disclose.

    • Co-Author(s)

  • FD

    Fabio Da Silva Gomes

    Pan American Health Organization, United States

  • LV

    Lesly Vejar Renteria

    Pan American Health Organization, United States

  • MM

    Mika Matsuzaki

    Johns Hopkins Bloomberg School of Public Health, United States

  • VG

    Vanessa Garcia-Larsen

    Pan American Health Organization, United States

Disclosure(s):
Muyao Li, MSPH: No relevant financial relationship(s) with ineligible companies to disclose.
Objectives: To develop a harmonized additive reference linking E-numbers with multilingual additive names and to evaluate ingredient-based rules for identifying ultra-processed foods (UPF) compared with Open Food Facts (OFF) NOVA classifications.

Methods: Packaged food products from eight Pan American Health Organization (PAHO) countries with front-of-pack (FOP) labeling policies (Argentina, Brazil, Chile, Colombia, Ecuador, Mexico, Peru, and Uruguay) were analyzed using ingredient lists from the OFF database. An additive reference linking E-numbers with additive functions and names in the three primary working languages of PAHO (English, Spanish, and Portuguese) was constructed using Codex GSFA/INS, the EU E-number system, FDA GRAS notices, and OFF additive tags.

Ingredient tokens were matched against the additive reference to detect cosmetic additives. A second set of markers included manually identified ultra-processed ingredients (UPI). Two screening rules were evaluated: cosmetic additives only, and a combined algorithm including both cosmetic additives and UPI markers. Products were classified as UPF if at least one marker was detected. Prevalence and performance metrics (sensitivity, specificity, positive predictive value, and accuracy) were calculated using OFF NOVA classification as the reference standard.

Results: Among 20,400 packaged foods across the eight countries, 70.8% were classified as UPF according to the NOVA classification in OFF. Cosmetic additive markers alone identified 65.6% of products as UPF, while the combined algorithm incorporating both cosmetic additives and UPI markers identified 67.2%. Across countries, the combined algorithm showed high sensitivity for detecting NOVA 4 products (range: 0.886–0.995) with moderate specificity (range: 0.602–0.806). Positive predictive values ranged from 0.860 to 0.931 and overall accuracy from 0.856 to 0.931. These findings indicate that ingredient-based markers can identify most ultra-processed foods while maintaining moderate specificity across different national food supplies.

Conclusions: This framework provides a reproducible method for detecting food additives across multilingual ingredient lists and may support scalable monitoring of ultra-processed foods where NOVA classification.

Funding Sources: Pan American Health Organization