Dozens of Nutrient Profile Models Drive Front-of-Pack Food Labels, Review Finds

Walk down almost any supermarket aisle in Santiago, Paris, London or Sydney and you will encounter small, brightly coloured symbols on the front of food packages: black warning octagons, five-colour scores, traffic-light arrays or star ratings. These front-of-pack nutrition labelling systems, known as FOPNL, have become one of the most widely deployed policy tools in the global fight against diet-related disease. Yet behind every label sits a less visible piece of machinery: a nutrient profile model, or NPM, the algorithm that decides whether a product earns a green light, a warning sign or no label at all. A new systematic review published in Nature Food has, for the first time, mapped this hidden landscape in detail, identifying 42 distinct nutrient profile models developed or used for front-of-pack labelling worldwide, and the picture it reveals is one of remarkable diversity and limited consensus.

The review, conducted by Margarida Bica, Jessica Renzella, Asha Kaur and Mike Rayner of the University of Oxford, drew on both peer-reviewed literature and the often-overlooked grey literature of government regulations, technical manuals and policy documents. The researchers systematically searched for nutrient profile models specifically developed or applied to front-of-pack nutrition labelling systems, characterising each model according to the type of labelling scheme it supported, the nutrients and food components it restricted or encouraged, and the evidence of its scientific validation. The scope matters, the authors argue, because the sheer proliferation of NPMs now poses a practical challenge for policymakers: with dozens of options available, choosing an appropriate model for a national labelling policy has become a non-trivial scientific and political decision.

The taxonomy that emerged from the analysis mirrors the main families of front-of-pack labelling in use today. Fifteen of the 42 identified models were developed for warning-label schemes, the approach pioneered in Chile and now dominant across much of Latin America, in which products exceeding thresholds for nutrients of concern carry prominent black octagons declaring them high in sugar, sodium or saturated fat. Eleven models underpinned nutrient-specific systems, including the traffic-light labelling familiar to British shoppers, which displays colour-coded indicators for individual nutrients. Another 11 models supported health endorsement schemes, such as the Nordic Keyhole or Singapore’s Healthier Choice Symbol, which positively certify foods deemed healthier choices. Four models drove summary scores, exemplified by France’s Nutri-Score and Australia’s Health Star Rating, which condense a food’s overall nutritional quality into a single graded mark. One model served other system types.

Beneath this structural variety, the review found striking common ground on which nutrients matter most. Every one of the 42 models limited the content of sugar, reflecting the strength of the scientific evidence linking excess sugar intake to obesity, type 2 diabetes and dental disease. The vast majority also placed restrictions on sodium and saturated fat, the two other nutrients most consistently associated with non-communicable disease burden. This convergence suggests that, despite institutional and regional differences, the underlying nutritional science has produced a broadly shared view of which dietary components should trigger concern on the front of a package. Sweeteners, energy density, protein, fibre, fruit and vegetable content appeared in more variable combinations across models, reflecting differing philosophies about whether labels should simply flag harmful nutrients or also reward beneficial ones.

Where the consensus frays is in the details of implementation. Thresholds for what counts as high in sugar differ between models; some assess nutrients per 100 grams while others evaluate them per serving or as a proportion of total energy; category-specific rules treat staples such as cheese, oils and beverages differently across schemes. The review documents this heterogeneity in granular detail, showing that two products judged unhealthy under one national scheme may pass cleanly under another. For multinational food manufacturers, this patchwork creates compliance complexity and, critics argue, opportunities to formulate products to game particular algorithms. For policymakers in countries drafting their first labelling laws, the absence of a single validated standard complicates decisions that industry lobbyists are often eager to influence.

Validation emerged as another significant gap. The review assessed whether models had demonstrated convergent validity, meaning agreement with other established measures of nutritional quality, and criterion validity, meaning the ability to predict health-relevant outcomes. Convergent validity was demonstrated for nine of the 42 models, while criterion validity was shown for only three. In other words, the scientific evidence underpinning most front-of-pack labelling algorithms remains thin relative to their policy importance. The authors note that this imbalance reinforces the need for greater regional guidance and alignment, so that new labelling programmes can build on models whose performance has been rigorously tested rather than reinventing criteria from scratch under political pressure.

The policy stakes are considerable. Diet-related non-communicable diseases, including cardiovascular disease, diabetes and several cancers, are among the leading causes of death globally, and governments increasingly view front-of-pack labels as a fast, low-cost intervention that shifts both consumer choice and industry reformulation. Chile’s warning-label law, implemented in stages since 2016, has been associated with reductions in purchases of sugary drinks and prompted widespread product reformulation across the region. The Pan American Health Organization’s nutrient profile model has served as a regional benchmark, while the World Health Organization has issued guiding principles and framework manuals to support member states. The European Union is weighing a harmonised mandatory front-of-pack label, a debate in which the choice between Nutri-Score-style summary scores and other approaches has become intensely contested among member states, scientists and industry groups.

The review also situates nutrient profiling within a broader scientific conversation about what food labels should measure. A growing body of research argues that nutrient-based models, however well calibrated, fail to capture the degree of industrial processing, exemplified by the NOVA classification of ultra-processed foods, which has been independently associated with adverse health outcomes in large cohort studies. Recent work has explored modifying established algorithms, such as Australia’s Health Star Rating, to account for ultra-processing, and examining the complementarity between updated Nutri-Score criteria and NOVA categories. The Oxford review does not resolve this debate, but its systematic inventory provides the evidentiary scaffolding for it, documenting precisely which models exist, what they measure and where their validation evidence stands.

Methodologically, the review followed established systematic review practice, including structured searches across peer-reviewed databases and targeted collection of grey-literature policy documents from national ministries, regional health bodies and regulatory agencies across the Americas, Europe, Asia, Africa and the Eastern Mediterranean. The authors screened studies collaboratively, analysed model characteristics in detail and reported their findings with extensive supplementary tables. The inclusion of grey literature proved essential: many of the 42 models, particularly those embedded in national regulations from Chile and Mexico to Kenya, Samoa and the Gulf states, exist primarily in legal texts and technical manuals rather than scientific journals. Without such searches, the true global diversity of nutrient profiling for labelling would be substantially underestimated.

The review’s central message is ultimately one of opportunity tempered by fragmentation. Front-of-pack nutrition labelling has matured from a handful of voluntary schemes into a global policy movement with real evidence of impact, and the raw material for that movement, a library of 42 nutrient profile models, is now catalogued and characterised. But the heterogeneity the authors document suggests that the field has not yet converged on what constitutes the most appropriate model for labelling purposes, and that validation evidence lags behind deployment. Greater regional guidance and alignment, the authors conclude, would help countries adopt models with demonstrated performance, reduce industry gaming of inconsistent thresholds, and move the world’s food packages toward labels that are not only eye-catching but scientifically defensible. As more governments prepare to legislate, the choice of algorithm beneath the label may prove as consequential as the label itself.

Subject of Research: Nutrient profile models developed and used for front-of-pack nutrition labelling systems to classify foods and support policies against diet-related non-communicable diseases.

Article Title: Systematic review of nutrient profile models for front-of-pack nutrition labelling

Article References: Bica, M., Renzella, J., Kaur, A., & Rayner, M. (2026). Systematic review of nutrient profile models for front-of-pack nutrition labelling. Nature Food. https://doi.org/10.1038/s43016-026-01414-0

Image Credits: AI Generated

DOI: 10.1038/s43016-026-01414-0

Keywords: nutrient profile models, front-of-pack nutrition labelling, food policy, non-communicable diseases, sugar, sodium, saturated fat, warning labels, Nutri-Score, traffic-light labelling, systematic review, public health nutrition