select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
Scientists respond to study published in BMJ Medicine* looking at energy density and levels of fat, sugar and salt in processed and ultra-processed food, and comparing existing and novel nutrient profiling models.
Prof Alberto Fiore, Professor in Food Chemistry and Technology, Abertay University, said:
“This is a classification exercise, not a health study – no people, no diets as eaten, and no health outcomes. It asks what proportion of the 1,898 packaged food and drink items in the UK’s National Diet and Nutrition Survey database (the study’s “processed” and “ultra-processed” categories) fails nutrient thresholds under different profiling systems. The study is carefully executed in its internals: it uses a nationally representative, up-to-date database; the NOVA coding was carried out with coders masked to the nutrient data and agreed between two coders including a registered dietitian; reporting follows STROBE; the statistical methods are appropriate; and the analysis code is publicly available. It also documents genuinely real inconsistencies between the UK’s existing systems for identifying unhealthy food.
“The headline “more than 90% unhealthy” figures, however, come from three new models constructed by the study’s authors themselves, in which a product is classed as unhealthy if it exceeds any single threshold for saturated fat, free sugar, sodium, total fat, or energy density, with those thresholds scaled from whole-diet recommendations onto individual foods. These models are new and untested – not previously published, not used in policy, and not validated against health outcomes or any external reference – so the finding that the HFSS and traffic light systems “miss” products is better described as a disagreement with the authors’ new yardstick than as an independent demonstration of failure. By the single-threshold rule, almost any packaged food fails – and, as the authors themselves note, even minimally processed foods such as unsalted nuts would be caught by the energy-density criterion if it were applied to them. The study also did not run its models over minimally processed foods, so there is no benchmark for how they perform across the whole food supply, and the FSA model was applied to all products rather than the 13 categories it actually regulates, so the comparison does not reflect how that system operates in the real world.
“In my view, and based on the evidence base as a whole rather than on this study, the nutrient-based systems the study criticises are the ones making the right calls: they penalise saturated fat, sugar and salt while crediting fibre, protein and wholegrain content, which is why a high-fibre bran cereal passes and a sausage does not. And in this study’s own data, the NOVA processing category does no discriminative work: under the authors’ most conservative model, 91.6% of “processed” (NOVA group 3) products and 91.5% of “ultra-processed” (NOVA group 4) products were flagged as unhealthy – virtually identical. If the level of processing tracked nutritional quality, those figures would diverge; they do not, because it is nutrient content that varies between products, not processing.
“In my view what the study shows is that everything of regulatory value in it – saturated fat, free sugar, sodium, energy density – is classic nutrient profiling; the NOVA category merely relabels it. In my view the vocabulary itself can be part of the problem: for example, “processed” and “ultraprocessed” have no consistent scientific definition, and the terms can group nutritionally balanced, industrially made foods – fortified wholegrain cereals, plant-based burgers, breaded fish – with confectionery and sugary drinks purely on the basis of how they were made rather than what is in them.Consumers might misunderstand and think very different foods are nutritionally equal. The debate about industrial food production remains a legitimate scientific discussion, but on this evidence the processing label is not adding information beyond what nutrient composition already tells us. Both the UCL and BMJ press releases keep this wording front and centre; journalists should treat it as the study’s framing, not its finding.”
Dr Rosiered Brownson-Smith, Postdoctoral Researcher in the Department of Nutritional Sciences, King’s College London, said:
“This is an interesting and timely study that highlights an important issue for UK food policy: the way we define and identify less healthy foods can vary considerably depending on the nutritional criteria we use.
“A key strength is that the researchers have applied different nutrient profiling models to the same dataset. This importantly demonstrates how the choice of model can substantially influence which products are classified as unhealthy. Their findings raise legitimate questions about whether existing approaches remain sufficiently aligned with current UK dietary recommendations, particularly given that the original FSA nutrient profiling model was developed around 20 years ago.
“However, it is important to interpret the findings carefully. The study assesses the nutritional composition of individual food products; it does not examine what people actually eat or whether consuming products classified as unhealthy leads to poorer health outcomes. The finding that almost 99% of ultra-processed foods were classified as unhealthy under the new model does not mean that 99% of these products have been demonstrated to cause harm.
“The proposed models also require further validation. In particular, translating dietary recommendations intended to apply to someone’s overall diet into thresholds for individual food products is not straightforward. Including energy density as a criterion is potentially valuable, but energy-dense foods are not necessarily unhealthy in isolation, and this needs to be considered across different food categories.
“Ultimately, this research provides useful evidence that the UK could benefit from a more consistent and up-to-date approach to nutrient profiling. However, the real test of any new system will be whether policies based on it improve the nutritional quality of the food supply and, importantly, people’s overall diets. Further research is needed to establish whether these proposed models achieve that.”
Dr Beverley O’Hara, Lecturer in Public Health Nutrition, Leeds Beckett University, said:
“The authors have tested various nutrient profile models to see if they are able to identify foods that have been categorised based on how processed they are. They report that assessing the amount of sugar, sodium or fat in foods is not a good way of identifying processed or highly processed foods.
The analysis considers 1,898 foods, which is a very small sample. In general, food databases contain upward of 500,000 food items. An important limitation is that the study is founded on the assumption that the NOVA classification system is an appropriate and robust framework for identifying less healthy foods, an assumption that remains subject to ongoing scientific debate. There is no independent or objective definition of what a ‘healthy’ food is to validate the analyses in this study.
“This paper illustrates that attempting to categorise foods based on healthiness is fraught with difficulty. In my opinion, the goal of producing an over-arching system for classifying all food products as ‘healthy’ or ‘unhealthy’ – particularly by incorporating the flawed NOVA system – has no scientific merit.”
Séamus Higgins, Associate Professor in Food Process Engineering, University of Nottingham, said:
“This is a well-designed comparison of how different nutrition rules classify the same foods. The researchers applied each system to nearly 1,900 food and drink items from the UK National Diet and Nutrition Survey database. Among ultra-processed items, the traffic-light criteria flagged 48%, and the marketing-restriction scoring system flagged 58%. The three new models based on the Eatwell Guide flagged 92% to 99%. Comparing the same items across all five systems shows how much scoring rules and thresholds matter.
“The (UCL) press release captures that gap, but its headline is too broad. The Eatwell Guide offers advice on the diet as a whole. The higher figures stem from the researchers’ models for translating that advice into rules for individual products. The researchers used the NOVA classification to group foods by degree of processing, then applied the scoring systems to the processed and ultra-processed groups. The study excluded minimally processed foods and did not test whether using any of these models would change what people buy or eat to improve health outcomes.
“The findings highlight a wider challenge for the food industry. Unless we can agree on consistent criteria for defining healthier foods, manufacturers will struggle to reformulate products with confidence, regulators will struggle to provide clear guidance, and consumers will struggle to make informed choices. Ultimately, greater consistency is essential if we are to achieve better health outcomes. This study makes a case for aligning dietary guidance, food labels and marketing rules more closely. More consistent regulatory criteria would also level the playing field for manufacturers and give shoppers greater clarity. Manufacturers would know which standards to work towards, and shoppers could more easily identify healthier food options.
“The bigger question is whether the Government has the political will to agree those rules and assess whether they help people across the UK source and afford healthier food.”
Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:
“This is a very interesting study which highlights the difficulties in providing useful dietary information to the general public. It highlights apparent discrepancies between dietary information provided by the government, although these discrepancies are partly due to the nature of the advice given. While the Eatwell guide provides broad dietary recommendations for the general public, nutrient profiling models are used to categorise individual foods into categories, to regulate marketing and provide information to consumers.
“The results of the study support the assumption that the often observed adverse health effect of so-called “ultra-processed foods” is mainly due to their nutrient composition. They find that a large proportion of these foods fall into the “unhealthy” category with high amounts of fat, sugar and salt. It is therefore likely that many of the effects attributed to ultra-processing are explained by nutrient. A diet meeting dietary recommendations – whether ultra-processed or not – is therefore unlikely to have an adverse effect on health. This has been shown by some of the authors in a study conducted last year: in a study comparing a minimally processed and an ultra-processed diet that both followed the Eatwell guide did not show significant adverse health effects for either diet (https://www.nature.com/articles/s41591-025-03842-0). That suggests that nutrient composition is more important for health effects than processing.
“The study is well conducted and analyses foods commonly consumed in the UK. The results are therefore very relevant to the UK consumer.
“The current study did exclude minimally processed foods, although these results would have been very useful for context. Many foods in these categories would fall into the “unhealthy” category of the models used, e.g. milk, eggs, nuts and avocados due to their fat or saturated fat content. It would be more complicated for fruits, as the sugar naturally occurring in fruits is treated differently from the same sugar in fruit juice.
“A key difficulty of all food labelling systems is that many foods are not uniformly “healthy” or “unhealthy” – and this is the reason why a balanced, varied diet is important. Foods might be energy dense or high in fat, but at the same time provide large amounts of fibre or micro-nutrients. It is difficult to capture this in a single “healthy” or “unhealthy” category.
“The results of this study demonstrate these difficulties well: The 2018 FSA nutrient profiling model – which uses a more holistic approach to determine the “healthiness” of foods – considers only about 70% of ultra-processed, and 55% of processed foods as unhealthy. Models that do not consider the composition of the entire food beyond the small group of nutrients consider more than 90% of foods as unhealthy.
“Such a finding is not only not helpful to a consumer (who might wish to choose the least unhealthy option), but also does not reflect the fact that diets are complex and based on many different foods.”
* ‘Evaluation of energy density and levels of fat, sugar, and salt in processed and ultra-processed food with existing and novel nutrient profiling models to identify unhealthy food and drink: cross sectional analysis of a nationally representative database of food and drink items’ by Samuel J Dicken et al. was published in BMJ Medicine on Tuesday 10 November 2026
DOI: 10.1136/ bmjmed-2025-001750
Declared interests
Prof Alberto Fiore: “Alberto Fiore is Professor of Food Chemistry and Food Technology at Abertay University, Dundee, Scotland. He has no financial or personal relationships with the authors of this study or their funders, no involvement in the study, and no financial interest with food industries.”
Dr Rosiered Brownson-Smith: “No COIs.”
Dr Beverley O’Hara: “I have no conflicts of interest to declare.”
Seamus Higgins: “Previously held technical and leadership roles within the food industry, including with Nestlé and Premier Foods. Since joining the University of Nottingham approximately ten years ago, he has had no employment, consultancy, advisory, financial, or other commercial relationships with these companies and reports no competing interests in relation to this article.”
Prof Gunter Kuhnle: “I am a former member of the UK Committee on Toxicity of Chemicals in Food, Consumer Products, and the Environment; a current member of the Advisory Committee on Novel Foods and Processes; the Director of the Chemical Analysis Facility at the University of Reading, which provides analytical services to academic and commercial clients; have received research funding (2010–20) from Mars for work on flavanols; and have received consultancy payments from RSM UK and EQT, paid to the University of Reading. As a member of the EFSA ANS panel (2018-2019) and the UK’s COT (2019-2025) I was involved in the evaluation of food additives.”