Ultra-Processed Food Debate: Nova Researchers Defend the Evidence on Human Health

A fierce scientific dispute over the health effects of ultra-processed foods has erupted into print, with the Brazilian research group that created the Nova food classification system issuing a detailed rebuttal to critics who argue that the category is too broad, biochemically incoherent, and unreliable as a basis for public policy. The response, published as a Matters Arising commentary in BMC Medicine, answers a critique by Ibsen and colleagues of the main thesis advanced in a Lancet Series on ultra-processed foods and human health. At stake is not merely a definitional quarrel: the outcome will shape how governments regulate the industrial food supply at a moment when diet-related chronic disease is rising worldwide.

The authors of the response, led by researchers affiliated with the Center for Epidemiological Studies in Health and Nutrition at the University of São Paulo, include Carlos A. Monteiro, the epidemiologist who first proposed the Nova classification in the late 2000s. Their critics contend that ultra-processed foods, or UPFs, lump together harmful and benign products in a way that makes causal inference and policy unreliable. The São Paulo team counters that this criticism rests on a fundamental misunderstanding of what Nova measures and how the epidemiological evidence has been generated.

Nova, they explain, does not classify foods by nutrient composition, biochemical coherence, or expected biological effects. Instead, it sorts foods into four groups according to the extent and purpose of industrial processing: fresh and minimally processed foods, processed culinary ingredients, processed foods, and ultra-processed foods. The standard exposure metric used in most prospective studies is the dietary share of UPFs, meaning the proportion of total energy intake derived from Group 4 products. This metric, the authors argue, captures the health effects of the ultra-processed dietary pattern as a whole, reflecting the real-world displacement of the first three Nova groups, which form the basis of traditional healthy diets, by industrially formulated products.

The response draws an analogy with the Mediterranean diet to defuse the charge of heterogeneity. Foods within the Mediterranean pattern differ widely in nutrient profiles, texture, palatability, and potential health effects when considered in isolation, yet they are routinely grouped as a single dietary exposure in epidemiological studies and dietary guidelines. Variability within a category, the authors insist, does not invalidate the category. The same logic applies to ultra-processed foods: heterogeneity within the pattern does not undermine its relevance as a dietary exposure or as a target for policy.

Central to the São Paulo group’s argument is the claim that harmful attributes co-occur systematically in ultra-processed foods as a consequence of the manufacturing processes and ingredients needed to achieve their commercial purpose. These attributes include disrupted food matrices and excessive energy density that accelerate eating and increase energy intake; engineered hyper-palatable formulations that override satiety signals; additives such as emulsifiers and non-sugar sweeteners that have been linked to gut and metabolic dysfunction; and toxic industrial contaminants, including chemicals that leach from plastic packaging. Nutrient imbalances, they argue, are largely a downstream consequence of the expansion of this dietary pattern rather than its defining feature, which is why a nutrient-by-nutrient approach is insufficient to explain or reverse the global rise of diet-related chronic disease.

The authors also reject the criticism that the mechanisms linking ultra-processed foods to health outcomes are too non-specific to be meaningful. Mechanisms linking dietary patterns to health outcomes, they note, do not need to be specific: the beneficial effects of the Mediterranean diet are not discounted simply because fibre, unsaturated fats, and antioxidant compounds also appear in other dietary patterns. What matters is that harmful attributes are more prevalent and more systematically combined in ultra-processed diets than in traditional diets based on Nova Groups 1 through 3. Beyond driving nutrient imbalances and energy-dense eating, ultra-processed food intake is inversely associated with health-protective compounds such as flavonoids and phytoestrogens, and directly associated with processing-generated toxic compounds and potentially harmful additives. Soft textures, reduced chewing frequency, and faster eating rates further facilitate excess energy intake, and some researchers have argued that many ultra-processed products meet established criteria for addictive substances.

Crucially, the response points to experimental evidence that directly tests whether the health effects of ultra-processed diets can be explained away by poor nutrient profiles. Dozens of prospective studies have found associations with chronic disease even after adjusting for fibre, sodium, sugar, and saturated fat. Randomised crossover trials have shown that ultra-processed diets lead to higher energy intake and weight gain even when meals are matched for key nutrients. A recent UK trial found that non-ultra-processed diets adhering to national dietary guidelines produced better cardiometabolic outcomes than ultra-processed diets modified to meet the same criteria, a result the authors say reinforces the conclusion that reformulation does not neutralise harm.

On the charge that epidemiological studies rely on unrealistic comparisons and are confounded by other dietary components, the authors argue that the criticism inverts the logic of nutritional epidemiology. Dietary pattern analyses are not designed to isolate the effects of specific foods or nutrients; they estimate the health effects of real-world diets with high versus low ultra-processed food intake, implicitly capturing the displacement of unprocessed and minimally processed foods. Paradoxically, they contend, analyses of specific ultra-processed subgroups are more vulnerable to confounding, because they attempt to compare individual products against their non-ultra-processed counterparts within complex diets. Measurement error, meanwhile, is mitigated by validated protocols and trained raters, and its non-differential nature tends to underestimate true associations rather than fabricate false ones.

The authors concede a nuance on comparators: in countries such as the United Kingdom and the United States, where traditional diets based on Nova Groups 1 through 3 are no longer the default, comparisons between ready-to-consume processed and ultra-processed products may be relevant, and a more incremental policy approach starting with feasible targets may be warranted. But they stress that the Lancet Series was not designed for these settings alone. In most countries, cooked meals still predominate, and the evidence on the health effects of displacing traditional diets with ultra-processed products is particularly relevant. Early comprehensive policies targeting all ultra-processed foods, they argue, could help other nations avoid the dietary crises observed in the UK and US.

The policy implications are stark. The authors acknowledge that nutrient-focused interventions, such as trans-fat bans, sodium front-of-package warnings, and taxes on sugar-sweetened drinks, are valuable but insufficient, because they operate within the existing corporate food system and encourage reformulation without addressing the underlying production model. Removing trans fat does not eliminate other industrial contaminants; reducing salt does nothing to address energy density, soft texture, and engineered hyper-palatability; and cutting sugar often means increasing amounts of harmful additives. Several countries have already incorporated Nova into their dietary guidelines and implemented policies targeting ultra-processed foods, actions the authors say are justified by more than one hundred prospective studies, randomised trials, and mechanistic investigations. Scientific uncertainty, they conclude, should always be acknowledged, but it should not be conflated with complexity, nor should heterogeneity within a meaningful category justify inaction. In their view, the case for treating ultra-processed foods as a systemic driver of the pandemic of diet-related chronic disease is robust, and policy responses should reflect that reality.

Subject of Research: Scientific debate over the Nova classification of ultra-processed foods and the evidence linking them to human health outcomes

Article Title: Response to: Debate in relation to “Ultra‑processed foods and human health: the main thesis and the evidence”

Article References: Rezende, L. F. M., da Costa Louzada, M. L., Martinez-Steele, E., Cannon, G., Levy, R. B., Jaime, P. C., & Monteiro, C. A. (2026). Response to: Debate in relation to “Ultra‑processed foods and human health: the main thesis and the evidence”. BMC Medicine, 24(1), Article 525. https://doi.org/10.1186/s12916-026-05217-9

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05217-9

Keywords: ultra-processed foods, Nova classification, dietary patterns, nutritional epidemiology, public health policy, chronic disease, food industry, randomised trials, food additives, Mediterranean diet, BMC Medicine, dietary guidelines