Applying the Nova food classification to food product databases using discriminative ingredients: a methodological...

Area of Science:

• Nutrition Science
• Food Science
• Public Health

Background:

• Growing evidence links ultra-processed foods (UPFs) to adverse health outcomes.
• Stakeholder interest in the Nova food classification system is increasing.
• Accurate UPF identification is vital for research and regulatory efforts.

Purpose of the Study:

• To develop and validate replicable methods for identifying UPFs.
• To test the sensitivity and specificity of proposed UPF identification methods.
• To utilize the 2017 Brazilian Food Labels Database for method validation.

Main Methods:

• Five scenarios were created to identify UPFs based on rare culinary substances and food additives.
• These scenarios were compared against the Nova classification's product name and category method.
• Diagnostic tests, ROC curves, and sensitivity analyses were employed to evaluate performance.

Main Results:

• UPF prevalence varied from 47% to 72% across scenarios, versus 70% with the classic method.
• Scenario 3 (cosmetic additives and rare culinary substances) identified 65% UPFs with good sensitivity and specificity.
• The addition of vitamins and minerals did not significantly alter UPF identification.

Conclusions:

• Ingredient-based criteria, specifically cosmetic additives and rare culinary substances, offer a reliable method for UPF identification.
• The proposed methods demonstrate reproducibility, supporting their application in research.
• These findings aid policy development and regulatory actions concerning UPFs.