other·nutrition, public health, epidemiology·PMC10002453
Comparison of Nutri-Score and Health Star Rating Nutrient Profiling Models Using Large Branded Foods Composition Database and Sales Data
International Journal of Environmental Research and Public Health · 2 authors, 3 centres
AI SUMMARY
FIDELITY 100%
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 study compared two front-of-package nutrition labelling schemes, Nutri-Score (NS) and Health Star Rating (HSR), using a large database of 17,226 Slovenian branded foods and 12-month sales data. The main finding was strong overall agreement (70%, κ=0.62) and very strong correlation (rho=0.87) between the models, though notable disagreements occurred in subcategories like cheese and cooking oils. A key limitation is that the data is from a single country's food supply, and the models were developed for different markets, which may affect generalizability.
Full summary
1,620 CHARS
BACKGROUND
Front-of-package nutrition labelling (FOPNL) grading schemes like Nutri-Score (NS) and Health Star Rating (HSR) aim to guide healthier food choices. This study compared their alignment using a large, real-world food supply dataset. Methods: Researchers applied NS and HSR algorithms to 17,226 pre-packed foods from the 2020 Slovenian branded foods database. They evaluated agreement (percentage, Cohen's Kappa) and correlation (Spearman rho) between the models. A 12-month nationwide sales dataset was used for sale-weighting to account for market-share differences. Key Results: Both models showed good discriminatory ability. NS graded 22% of products as healthy (A/B), while HSR graded 33% as healthy (≥3.5 stars). Overall agreement was strong (70%, κ=0.62) with a very strong correlation (rho=0.87). Alignment was highest in Beverages and Bread and bakery products. Notable disagreements were found in Cheese and processed cheese (8% agreement, κ=0.01, rho=0.38) and Cooking oils (27% agreement, κ=0.11, rho=0.40). Sale-weighting increased overall agreement to 81%. Limitations: The study used a single country's food supply, and some nutritional data (e.g., fibre, % FVNL) were estimated. Sales data were not available for all products. The models were developed for different markets with different public health priorities. Implications: The findings highlight that while NS and HSR are largely compliant, differences in specific subcategories reflect challenges in nutrient profiling. International harmonization could support the development of more universally acceptable grading schemes for FOPNL.
