Ultra-Processed Foods and Health: Is Processing Really the Problem?

Few subjects in modern nutrition have become as polarised as ultra-processed food. If you ever want to get a group of healthcare professionals bickering like Tom and Jerry after 7 coffees, then just bring up this subject and watch the fireworks. It is a hotbed. Depending on which side of the debate you encounter, ultra-processed foods are either portrayed as one of the principal drivers of obesity and chronic disease or dismissed as an imprecisely defined category that tells us considerably less about the nutritional quality of food than its advocates claim.

The public conversation has consequently become remarkably simplistic. One position treats industrial processing almost as though it were inherently toxic, while the opposing position argues that humans have processed food for thousands of years and that nutrients, calories and overall dietary quality remain more important than the number of industrial processes involved in producing a food. There are legitimate arguments within both positions, but neither adequately captures where the science currently stands.

The evidence linking high consumption of ultra-processed foods with poorer health outcomes has become increasingly difficult to dismiss. Large prospective studies and meta-analyses repeatedly associate greater consumption with cardiovascular disease, type 2 diabetes, obesity and premature mortality. More importantly, controlled feeding experiments have demonstrated that certain diets dominated by ultra-processed foods can cause people to consume substantially more energy and gain weight, even when researchers attempt to match several conventional nutritional characteristics between diets.

What remains much less certain is whether industrial processing itself is the biological cause of these effects. A major scientific debate is now emerging around precisely this question. Is the health signal associated with ultra-processed foods produced by processing, additives and changes to the food matrix, or are we seeing the combined effects of energy density, rapid eating rate, hyperpalatability, refined carbohydrate, added fat, low fibre and the displacement of more nutritious foods? The answer matters because public-health advice needs to identify what we actually need to change. If processing itself is the principal problem, reducing ultra-processed foods as a category makes sense. If particular characteristics within that category drive most of the harm, a more targeted approach may ultimately prove considerably more useful.

The science is therefore beginning to move beyond asking whether ultra-processed foods are associated with poor health and towards the much more interesting question of why.

What exactly is an ultra-processed food?

The term ultra-processed food is most closely associated with the NOVA food-classification system developed by researchers at the University of São Paulo. NOVA classifies foods primarily according to the nature, extent and purpose of industrial processing rather than simply according to their nutrient composition.

The system contains four broad groups. Group one consists of unprocessed or minimally processed foods such as vegetables, fruits, grains, legumes, meat, fish, eggs and milk. Group two contains processed culinary ingredients including oils, butter, sugar and salt. Group three contains processed foods created by combining foods from the first group with ingredients from the second, such as many cheeses, canned vegetables and traditionally produced breads. Group four contains the foods NOVA defines as ultra-processed.

Ultra-processed foods are generally industrial formulations containing ingredients or processes that would not normally be used in domestic cooking. They may contain substances extracted or modified from foods, including protein isolates, modified starches and hydrogenated oils, alongside flavours, colours, emulsifiers, sweeteners and other additives designed to influence texture, appearance, shelf life or palatability. Typical examples include many soft drinks, confectionery products, packaged snacks, sweetened breakfast cereals, instant noodles, reconstituted meat products, some commercial breads, desserts and numerous ready meals.

The classification has been enormously influential because it introduced something conventional nutrient-based approaches can sometimes overlook. Foods are not merely collections of protein, carbohydrate, fat, vitamins and minerals. Their physical structure, method of manufacture, eating rate, palatability and place within the overall diet may also influence their physiological effects. However, that strength is also the source of much of the controversy surrounding NOVA.

Ultra-processed is not the same thing as processed

One of the first mistakes in the public discussion is treating all processing as though it were nutritionally undesirable. Humans have processed food for thousands of years. Cooking, fermentation, grinding, freezing, drying, canning and pasteurisation are all forms of food processing, and many of these processes improve food safety, digestibility, shelf life or nutrient availability.

Frozen vegetables remain highly nutritious. Canned beans provide an inexpensive and convenient source of protein and fibre. Pasteurisation dramatically reduces the risk of foodborne disease. Fermentation can produce foods with distinctive nutritional and microbial characteristics. Milling and cooking can make nutrients more accessible and foods considerably easier to digest.

The NOVA system does not classify all of these foods as ultra-processed simply because processing has occurred. Its concern is specifically with industrial formulations in which the original food may have been substantially transformed and combined with ingredients designed to produce particular sensory, structural and commercial characteristics.

Statements such as “processed food is bad for you” are therefore essentially meaningless. The health implications of processing depend on what has been done to the food, what the resulting product contains, how its physical structure has changed and how that food is consumed. The more difficult question is whether the narrower category of ultra-processing identifies something biologically important beyond conventional nutritional composition.

The observational evidence has become difficult to ignore

Over the last decade, a substantial body of prospective epidemiological research has associated greater consumption of ultra-processed food with adverse health outcomes. Systematic reviews and meta-analyses have reported associations with outcomes including obesity, type 2 diabetes, cardiovascular disease and all-cause mortality, although the strength and certainty of evidence vary considerably between individual outcomes.

An updated systematic review and dose-response meta-analysis published in 2025 examined 18 prospective cohort studies involving more than one million participants. Compared with those consuming the least ultra-processed food, people consuming the greatest amounts had a higher risk of all-cause mortality. Importantly, the researchers also identified evidence of a dose-response relationship, with mortality risk tending to increase as ultra-processed food consumption increased.

An umbrella review published in the BMJ similarly examined evidence across numerous meta-analyses and reported associations between greater exposure to ultra-processed foods and a broad range of adverse health outcomes. The certainty of those relationships was not uniform, and observational evidence inevitably contains limitations, but the sheer consistency with which a health signal has appeared across different populations means that it deserves serious attention.

Ultra-processed food has also entered healthy-ageing research. Large prospective studies examining dietary patterns over many years have associated greater consumption of ultra-processed foods with a lower probability of reaching older age free from major chronic disease while maintaining physical, cognitive and mental health. These findings are particularly interesting because they move the conversation beyond body weight and towards the broader question of whether long-term dietary patterns influence healthspan.

However, consistency of association does not automatically establish causality. People consuming large quantities of ultra-processed foods often differ from those consuming relatively little in numerous ways. Their diets may contain more energy, refined carbohydrate, added sugar, sodium and saturated fat while providing less fibre, fruit, vegetables and other minimally processed foods. They may also differ in smoking, physical activity, socioeconomic circumstances, sleep and other factors that influence long-term disease risk.

Researchers can statistically adjust for many of these variables, but statistical adjustment is never perfect. Some variables are measured imprecisely, others may not be measured at all, and dietary assessment itself introduces error. Observational studies can therefore establish a remarkably consistent relationship while still leaving open the question of what actually causes it.

The Hall study changed the UPF debate

In 2019, researchers led by Kevin Hall at the US National Institutes of Health conducted what remains one of the most influential experiments in this field. Twenty adults were admitted to a metabolic ward and randomly assigned to consume either an ultra-processed or an unprocessed diet for two weeks before switching to the alternative diet for another two weeks. Participants were allowed to eat as much or as little as they wanted.

The researchers attempted to match the diets as presented for several important nutritional variables, including total calories, energy density, macronutrients, sugar, sodium and fibre. This was important because it allowed the experiment to investigate whether differences in food processing could influence spontaneous food intake even when several conventional nutritional characteristics were controlled.

The result was striking. During the ultra-processed phase, participants consumed approximately 500 additional kilocalories per day and gained around 0.9 kilograms over the two-week period. During the unprocessed phase, participants lost approximately the same amount of weight. Because the same participants experienced both dietary conditions under tightly controlled circumstances, the experiment provided considerably stronger causal evidence than observational studies alone could provide.

The study demonstrated that something about the ultra-processed dietary condition encouraged substantially greater energy intake. What it could not establish was which characteristic of that diet produced the effect. Participants consumed the ultra-processed meals more rapidly, and there were differences in the proportions of carbohydrate and fat they selected. The foods also differed in physical structure, sensory characteristics and potentially in the speed with which nutrients could be consumed and absorbed. The experiment therefore showed that an ultra-processed dietary pattern can promote overeating under controlled conditions, but it did not prove that industrial processing itself was the singular mechanism responsible.

Eating rate may be one of the missing pieces

One of the most interesting explanations involves something remarkably simple: many ultra-processed foods can be eaten very quickly. Food structure has a major influence on eating rate. Compare eating an intact apple with drinking apple juice, chewing whole nuts with eating a smooth nut butter, or eating a plate containing intact vegetables and protein with consuming a soft, highly refined ready meal. The ingredients may provide some overlapping nutrients, but the physical experience and rate of consumption can be completely different.

This matters because the physiological development of satiety is not instantaneous. Mechanical distension of the stomach, nutrient sensing within the intestine and hormonal signals including glucagon-like peptide-1, peptide YY and cholecystokinin contribute to the regulation of appetite during and after a meal. These processes unfold over time rather than responding immediately to the first mouthful of food.

Foods that require relatively little chewing and deliver large amounts of energy rapidly may therefore allow substantial energy consumption before the full constellation of satiation signals has developed. Research examining food texture and eating rate increasingly supports this possibility, with softer and easier-to-consume foods frequently allowing greater energy intake per minute.

Eating rate consequently provides a plausible mechanism linking some ultra-processed foods with excess energy intake without requiring industrial processing itself to be inherently harmful. Processing may still be important because it can create the physical properties that permit rapid consumption, but the immediate biological mechanism may be the rate at which energy enters the body rather than the mere fact that a food has undergone industrial manufacture.

Energy density and hyperpalatability may be equally important

Energy density provides another plausible part of the explanation. Foods containing combinations of refined carbohydrate and fat can deliver large amounts of energy in relatively small volumes, particularly when processing removes water or substantially disrupts the original food structure. This allows significant quantities of energy to be consumed before the physical volume of food becomes sufficient to generate strong satiation.

Palatability may add another layer. Foods containing carefully balanced combinations of fat, refined carbohydrate, sugar and sodium can be extremely easy and rewarding to eat. This does not require the more sensational claim that food manufacturers have discovered a mysterious way of overriding free will. Sensory properties have always influenced human eating behaviour, and foods that taste particularly good tend to encourage greater consumption.

The combination of high energy density, rapid eating rate and strong palatability could therefore create circumstances in which excess energy intake becomes remarkably easy. Importantly, none of these characteristics is unique to ultra-processed food. A homemade cake can be highly energy dense and extremely palatable, while a traditional pastry can combine refined starch and fat and be consumed very rapidly without necessarily meeting every definition of ultra-processing.

This is one of the central difficulties in treating NOVA as though it identifies a single biological exposure. Ultra-processed foods often possess characteristics that can plausibly promote overeating, but many of those characteristics can also occur in foods produced using relatively conventional culinary techniques.

Controlled feeding research is making the picture more complicated

Research following the original Hall experiment has increasingly attempted to identify which characteristics of ultra-processed diets influence energy intake. This work is beginning to reveal an important point: not all ultra-processed dietary patterns necessarily behave in the same way.

When researchers manipulate characteristics such as energy density, food texture, eating rate and hyperpalatability, differences in spontaneous energy intake can change substantially. This suggests that the physiological consequences attributed to ultra-processing may depend partly on the particular properties created by processing rather than simply on whether a product falls into NOVA group four.

A more recent randomised crossover trial in the UK provides an especially useful example. Researchers compared diets differing substantially in their degree of processing while ensuring that both complied with established healthy-eating guidelines. Participants lost weight during both dietary periods, although weight loss was greater during the minimally processed phase.

This does not overturn the evidence implicating high-UPF dietary patterns. Instead, it makes the question considerably more precise. A diet dominated by confectionery, sugary drinks, crisps and processed meat is nutritionally and structurally very different from one containing commercially produced wholegrain bread, fortified cereal and some packaged foods formulated to meet healthy-eating recommendations, even if products within both diets technically qualify as ultra-processed.

If different foods within the same classification produce substantially different effects, researchers need to understand which properties within the category are responsible for the health signal.

Is processing itself actually the problem?

This question has become the focus of an increasingly important scientific debate. Critics of the UPF concept do not necessarily argue that diets dominated by soft, energy-dense, highly palatable foods rich in fat, sugar and salt are harmless. Their criticism is more specific: current research has not conclusively demonstrated that ultra-processing itself is the independent causal factor responsible for the adverse health associations attributed to UPFs.

One of the difficulties is determining the appropriate comparison food. If an ultra-processed chocolate dessert is compared with a piece of fruit, almost every meaningful characteristic differs simultaneously. The foods differ in energy density, fibre, sugar, fat, physical structure, eating rate, micronutrients and phytochemicals as well as in their degree of processing. If different health effects subsequently occur, attributing them specifically to processing becomes extremely difficult.

To isolate processing experimentally, researchers would ideally need foods that are closely matched for nutrient composition, energy density, texture, eating rate, palatability and other relevant characteristics while differing meaningfully in the nature or degree of processing. Achieving this in real foods is exceptionally challenging because processing itself frequently creates changes in many of these characteristics.

The debate should therefore not be reduced to one group claiming UPFs are harmful while another claims they are harmless. The scientifically interesting disagreement concerns what the UPF health signal actually represents and whether NOVA identifies a causal exposure or a collection of food characteristics that frequently travel together.

The food matrix may help explain why processing matters

Conventional nutrition often describes foods according to their nutrient composition. Protein, carbohydrate, fat, fibre, vitamins and minerals are measured, and foods containing similar quantities can consequently appear nutritionally equivalent. Human digestion, however, does not encounter nutrients as numbers on a food label. It encounters those nutrients embedded within physical structures.

In plant foods, nutrients can remain enclosed within cellular walls. In intact grains, legumes and nuts, some nutrients are released relatively slowly because the original structure of the food remains partially preserved during digestion. Grinding, extrusion, refining and other forms of processing can disrupt these structures and increase the accessibility of starch, fat and protein to digestive enzymes.

This concept is known as the food matrix, and it helps explain why two foods containing similar nutrients can produce different physiological responses. Whole nuts and nut butter, for example, may contain essentially the same ingredients while differing in the proportion of fat that becomes available during digestion. Whole fruit and fruit juice similarly demonstrate how disruption of physical structure can alter eating rate, satiety and the delivery of carbohydrate.

Ultra-processing can profoundly modify this matrix. Depending on the food, this could influence chewing, eating rate, gastric emptying, nutrient absorption, glycaemic response, intestinal hormone signalling and the amount of undigested substrate reaching the colon and interacting with the gut microbiota.

The importance of food structure is one reason the UPF debate cannot be resolved simply by comparing nutrition labels. Two products may contain similar quantities of carbohydrate, fat and protein while behaving differently once consumed because those nutrients are presented to the gastrointestinal tract in very different physical forms.

Food additives remain an important but unresolved question

Food additives represent another major area of interest. Ultra-processed products frequently contain emulsifiers, stabilisers, sweeteners, colours and other compounds used to modify texture, shelf life, appearance or sensory properties. Experimental research has raised legitimate questions about whether some of these substances might influence the gut microbiota, intestinal mucus layer, glucose regulation or inflammatory signalling.

Emulsifiers have received particular attention because certain compounds have altered the gut microbiota and promoted inflammatory or metabolic changes in animal models. Small human experiments have also produced signals suggesting that particular emulsifiers may influence microbiota composition or metabolic responses in some individuals.

These findings are biologically interesting, but they do not justify treating all food additives as a single harmful category. Hundreds of permitted additives exist with very different chemical structures, biological properties and functions. Evidence concerning one emulsifier cannot automatically be extrapolated to another emulsifier, let alone to preservatives, colours, stabilisers or sweeteners.

Some additives may eventually prove physiologically important at commonly consumed doses, particularly when consumed repeatedly or in combinations. Others may have little meaningful effect on human health. Establishing this requires careful investigation of individual compounds and realistic exposure levels rather than assuming that an unfamiliar chemical name on an ingredient list is inherently harmful.

Additives may ultimately explain part of the relationship between certain UPFs and health, but current evidence does not support the conclusion that they account for the UPF health signal as a whole.

Ultra-processed foods can also displace nutritionally valuable foods

There is another explanation for part of the association that requires no novel molecular mechanism. If a large proportion of somebody's energy intake comes from soft drinks, confectionery, packaged snacks, desserts, processed meats and refined convenience foods, those foods inevitably replace something else within the diet.

They may displace vegetables, fruits, legumes, whole grains, nuts, seeds and other foods providing fibre, micronutrients and phytochemicals. A diet dominated by these products may also contain less intact food structure and provide less fermentable substrate for the gut microbiota.

Nutrition research repeatedly demonstrates that substitution matters. Removing one food or nutrient cannot be fully understood without considering what replaces it. Replacing saturated fat with polyunsaturated fat produces a different physiological effect from replacing it with refined carbohydrate. Replacing meat with legumes is nutritionally different from replacing it with refined starch. The same principle applies when reducing ultra-processed foods.

Replacing sugary drinks with water is highly likely to be beneficial in someone consuming substantial quantities. Replacing confectionery with fruit or nuts can increase fibre and micronutrient density while reducing energy density. By contrast, replacing a high-fibre commercial wholegrain bread with a homemade white loaf simply because the latter appears less processed may provide little nutritional advantage and could potentially make the diet worse.

This illustrates why processing classification can provide useful information about the overall dietary pattern while becoming less reliable when used as the sole method for judging individual foods.

Not all ultra-processed foods are nutritionally equivalent

One of the biggest problems with the way ultra-processed foods are discussed publicly is the tendency to speak as though every product within NOVA group four has the same biological significance. The category contains an enormous variety of foods, ranging from sugar-sweetened beverages, confectionery, packaged cakes and processed meats to some commercially produced wholegrain breads, yoghurts, breakfast cereals, plant-based products and fortified foods.

Their nutrient profiles, energy densities, fibre contents, protein contents, food matrices and likely effects on appetite can be completely different. It would therefore be surprising if their long-term health effects were identical.

Large observational studies examining individual UPF subgroups increasingly support this heterogeneity. Certain categories, particularly sugar-sweetened beverages and processed meat products, frequently show stronger associations with adverse cardiometabolic outcomes. Other categories demonstrate weaker, neutral or occasionally inverse associations depending on the population and outcome being examined.

These findings need cautious interpretation because people choosing particular categories of UPFs may differ in many other ways. Someone consuming commercially produced wholegrain bread and yoghurt may have a very different overall lifestyle from somebody consuming large quantities of processed meat, sugary drinks and confectionery. Residual confounding therefore remains possible.

Nevertheless, the subgroup evidence makes it increasingly difficult to justify treating every food classified as NOVA group four as biologically equivalent. The classification may successfully identify a broad dietary pattern associated with poor health while remaining too heterogeneous to function as a precise measure of the nutritional quality of every individual food.

Nutrient composition and processing answer different questions

Part of the UPF controversy arises because NOVA and conventional nutrient profiling are sometimes presented as competing systems, as though nutrition science must choose between them. In reality, they answer different questions and may be most informative when considered together.

Nutrient profiling asks what a food provides in terms of energy, protein, carbohydrate, fat, fibre, sodium, vitamins and minerals. NOVA asks how and why that food was manufactured. The first approach can identify nutritional characteristics associated with health, while the second may capture aspects of food structure, formulation and eating behaviour that conventional nutrient labels overlook.

A sugar-sweetened soft drink is nutritionally poor because it delivers free sugar rapidly, provides little fibre and contributes relatively little nutritional value beyond energy. Its ultra-processed classification adds information about its formulation, but we do not need NOVA to understand why consuming large quantities of sugary drinks may be undesirable.

Conversely, a commercially produced wholegrain bread may technically qualify as ultra-processed because it contains an emulsifier or another industrial ingredient while still providing useful amounts of fibre, micronutrients and wholegrain material. Rejecting its entire nutritional contribution because of its NOVA classification would make little sense.

Ignoring processing entirely may therefore miss meaningful aspects of food structure and eating behaviour, while ignoring nutritional composition because something is classified as ultra-processed can be equally misleading. A sophisticated assessment of food quality should be capable of considering both.

NOVA may be most useful at the level of the overall diet

The practical value of the ultra-processed concept becomes much clearer when it is applied to dietary patterns rather than used as a moral label for individual foods. If most of somebody's diet consists of vegetables, fruit, legumes, nuts, seeds, whole grains and appropriate sources of protein, the inclusion of some commercially produced foods is unlikely to transform an otherwise nutritious dietary pattern into an unhealthy one.

Conversely, if most daily energy comes from packaged snacks, confectionery, sugary drinks, refined convenience foods, desserts and highly processed meat products, the nutritional problems associated with that pattern are unlikely to disappear simply because individual products have been fortified with vitamins or reformulated to contain slightly less sugar.

This perspective is consistent with nutrition science more broadly. Long-term health is rarely determined by one food eaten occasionally. It reflects the cumulative pattern of foods and behaviours repeated across months, years and decades. The proportion of the diet dominated by particular foods therefore matters considerably more than whether somebody occasionally consumes an individual product classified as ultra-processed.

Using NOVA in this way preserves one of its most useful insights. A food environment increasingly dominated by industrially formulated, rapidly consumed and energy-dense products may have health consequences that cannot be understood solely by examining isolated nutrients. What becomes less defensible is turning the classification into an absolute rule in which every NOVA 4 food is considered intrinsically harmful.

Complete elimination is neither necessary nor particularly useful

Modern food processing provides substantial practical benefits. It improves food safety, reduces spoilage, extends shelf life and makes nutritionally useful foods accessible to people who may not have the time, money, facilities or physical ability to prepare everything from basic ingredients. Any realistic discussion of UPFs therefore needs to acknowledge the role that convenience and affordability play in the food system.

A rigid anti-UPF approach can also produce peculiar nutritional decisions. Someone might reject a high-fibre commercial bread because it contains an emulsifier while consuming large quantities of butter, sugar and white flour in homemade baked goods because those products are technically less processed. The latter may satisfy a particular interpretation of NOVA while doing very little to improve the nutritional quality of the diet.

A more useful objective is to ensure that minimally processed and nutritionally dense foods provide the structural foundation of eating. Vegetables, fruits, legumes, nuts, seeds and other plant foods can provide fibre, micronutrients and diverse phytochemicals. Appropriate sources of protein support muscle and other physiological functions, while predominantly unsaturated sources of dietary fat can contribute to cardiovascular health.

Commercially produced foods can then be evaluated according to the role they actually play within that pattern. Their nutrient composition, fibre and protein content, energy density, physical structure, portion size and likely effect on eating behaviour all provide useful information. This moves the discussion away from fear of a packet or an ingredient list and towards a more meaningful assessment of nutritional quality.

Why the health signal should still be taken seriously

Uncertainty about the precise mechanism should not be mistaken for evidence that the overall relationship between UPF-rich diets and poor health is meaningless. Prospective epidemiological studies involving very large populations repeatedly identify adverse associations. Controlled experimental research demonstrates that at least some ultra-processed dietary patterns can substantially increase spontaneous energy intake. There are plausible mechanisms involving eating rate, energy density, palatability, food structure and potentially particular additives, while diets dominated by many common UPFs frequently displace foods with well-established nutritional benefits.

Taken together, this is sufficient reason to take high consumption of ultra-processed foods seriously. What the evidence does not yet establish is that ultra-processing represents one uniform biological mechanism explaining all of these observations.

This distinction has important implications for the future of nutrition policy and food reformulation. If eating rate proves to be a major driver, manufacturers could potentially change food texture and structure so that energy is consumed more slowly. If energy density and particular combinations of nutrients account for much of the effect, those become obvious targets for reformulation. If specific additives prove harmful at realistic exposure levels, regulation can focus on those compounds. If particular UPF categories account for most of the disease association, public-health guidance can become much more precise than simply advising people to avoid an enormous and heterogeneous category of foods.

Understanding the mechanism therefore does not weaken the argument for improving the food supply. It gives us the opportunity to identify exactly what needs improving.

Moving beyond the good-food-versus-bad-food argument

The evolution of UPF research is a good example of how nutrition science should progress. The first phase established an epidemiological signal in which people consuming greater quantities of ultra-processed foods repeatedly appeared to experience poorer health outcomes. The next phase produced controlled experiments demonstrating that certain ultra-processed dietary patterns could increase spontaneous energy intake and promote weight gain.

Research is now entering the more difficult and potentially more valuable phase of identifying which properties of those diets produce these effects. Processing may matter because it changes the physical structure of food. Texture may matter because it changes eating rate. Energy density may allow large quantities of energy to be consumed before satiation develops. Palatability may influence how much people choose to eat. Particular additives may affect intestinal physiology or the microbiota, while a diet dominated by nutritionally poor UPFs may simultaneously displace fibre-rich and minimally processed foods.

Several of these mechanisms could operate at the same time, and their relative importance may differ between products. A sugar-sweetened drink may influence health predominantly through rapidly consumed free sugar and excess energy intake, while processed meat raises a different collection of nutritional questions. A soft, energy-dense snack may promote passive overconsumption through yet another combination of mechanisms.

Expecting one explanation to account for every food placed within NOVA group four may therefore be asking the classification to do something it was never capable of doing. The eventual answer is unlikely to be that ultra-processing is either completely irrelevant or uniquely responsible for modern chronic disease. Human nutrition rarely provides explanations that simple.

What should we actually do with this information?

For practical purposes, the current evidence supports a dietary approach that is considerably less dramatic than much of the public discussion. Whole and minimally processed foods should provide the nutritional architecture of the diet because this naturally increases consumption of foods supplying fibre, micronutrients, protein, unsaturated fats and intact food structure while reducing dependence on products that make large quantities of energy extremely easy to consume.

Within that framework, there is little scientific justification for fearing every food classified as ultra-processed. A packaged food can still make a useful nutritional contribution, particularly when it provides meaningful amounts of fibre, protein or essential nutrients and makes a healthy dietary pattern more affordable, convenient or sustainable.

Equally, the imperfections of NOVA should not be used as an excuse to ignore the nutritional problems associated with diets dominated by sugary drinks, confectionery, packaged snacks, refined convenience foods and processed meats. The epidemiological and experimental evidence gives us good reason to reduce reliance on these foods, particularly where they contribute substantial amounts of energy while displacing foods with greater nutritional value.

A more sophisticated assessment therefore asks several questions simultaneously. We can consider how processed a food is, but also what it contains, how energy dense it is, how quickly it can be consumed, how much of its original physical structure remains, whether it provides fibre and protein, what role it plays in appetite regulation and what it replaces elsewhere in the diet. None of these questions needs to exclude the others.

Ultra-processed food remains an important concept because it forced nutrition science to look beyond isolated nutrients and consider how the modern food supply has changed the physical and sensory characteristics of what we eat. That was a valuable contribution and continues to generate important research.

The next stage is to understand the category with greater precision. The evidence increasingly suggests that dietary patterns dominated by ultra-processed foods deserve concern, but the biological explanation is unlikely to be captured by the word “processing” alone. Identifying the combination of food structure, eating rate, energy density, nutrient composition, palatability, displacement and potentially specific additives responsible for the observed health effects will allow nutrition science to move from classification towards mechanism.

That is ultimately a far more useful question than whether every food carrying the label “ultra-processed” should simply be considered good or bad.

References

Lane, M.M., Gamage, E., Du, S. et al. (2024) ‘Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses’, BMJ, 384, e077310. doi: 10.1136/bmj-2023-077310.

Liang, S., Zhou, Y., Zhang, Q., Yu, S. and Wu, S. (2025) ‘Ultra-processed foods and risk of all-cause mortality: an updated systematic review and dose-response meta-analysis of prospective cohort studies’, Systematic Reviews, 14, 53. doi: 10.1186/s13643-025-02800-8.

Barbaresko, J. et al. (2025) ‘Ultra-processed food consumption and human health: an umbrella review of systematic reviews with meta-analyses’, Critical Reviews in Food Science and Nutrition, 65(11), pp. 1999–2007. doi: 10.1080/10408398.2024.2317877.

Hall, K.D., Ayuketah, A., Brychta, R. et al. (2019) ‘Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake’, Cell Metabolism, 30(1), pp. 67–77.e3. doi: 10.1016/j.cmet.2019.05.008.

Monteiro, C.A., Cannon, G., Levy, R.B. et al. (2019) ‘Ultra-processed foods: what they are and how to identify them’, Public Health Nutrition, 22(5), pp. 936–941. doi: 10.1017/S1368980018003762.

Tessier, A.J. et al. (2025) ‘Optimal dietary patterns for healthy aging’, Nature Medicine, 31, pp. 1644–1652. doi: 10.1038/s41591-025-03570-5.

Dicken, S.J. et al. (2025) ‘Ultra-processed or minimally processed diets following healthy dietary guidelines on weight and cardiometabolic health: a randomised crossover trial’, Nature Medicine.