A 2024 BMJ umbrella review linked greater exposure to ultra-processed food with 32 adverse health outcomes across 45 pooled analyses. The review covered nearly 9.89 million participants, but most of its evidence came from observational research and cannot by itself establish that ultra-processing caused each outcome.

NOVA offers a way to describe how food is made, not a simple division between natural and harmful products. Frozen vegetables, pasteurized milk and canned beans can be minimally processed or processed foods, while soft drinks, packaged snacks and reconstituted meat products often fall into the fourth group.

When symptoms need urgent care

Ultra-processed food intake is generally a long-term dietary issue, not an acute emergency, and a NOVA category cannot explain a new symptom. A person with diabetes who develops intense thirst and frequent urination together with vomiting, deep breathing, fruity-smelling breath, marked drowsiness or confusion needs urgent medical assessment because these can be signs of diabetic ketoacidosis.

The United Kingdom's National Health Service describes diabetic ketoacidosis as life-threatening and says it requires urgent hospital treatment. Children, pregnant or breastfeeding people, older adults at risk of malnutrition, and people managing diabetes or kidney disease should discuss broad food restrictions with a clinician or registered dietitian rather than removing foods solely because of a NOVA label.

Packaged cookies, potato chips and instant noodles displayed on store shelves (illustrative image)

NOVA classifies the purpose and extent of processing

The researchers who developed NOVA describe four groups based on the extent and purpose of industrial processing. Group one covers unprocessed or minimally processed foods, including fresh or frozen produce, grains, eggs and pasteurized milk when salt, sugar, oil or other food substances have not been added.

Group two contains processed culinary ingredients such as oil, sugar and salt. Group three covers foods made mainly by combining group-one foods with group-two ingredients, including many canned vegetables, canned fish, cheeses and freshly made breads.

Group four contains ultra-processed formulations assembled largely from substances derived from food and additives, often with little intact group-one food. Examples include carbonated soft drinks, sweet or savory packaged snacks, instant noodles, reconstituted meat products and many ready-to-heat dishes.

The method matters more than whether a package, freezer or can was involved. Freezing, pasteurizing, drying and canning without added sugar or excess sodium can preserve foods without turning them into group-four products.

Ingredient markers are more useful than package claims

NOVA's developers identify ingredients rarely used in home kitchens as markers of ultra-processing. These include high-fructose corn syrup, maltodextrin, hydrogenated or interesterified oils, hydrolyzed proteins and protein isolates, along with additives used mainly to change sensory qualities, such as flavors, colors, sweeteners, emulsifiers and thickeners.

A long ingredient list can prompt a closer look, but length alone is not a formal test. A product may contain several ordinary culinary ingredients, while a shorter formulation may still use an industrial ingredient or additive that places it in group four. Claims such as natural, organic or plant-based also do not determine the NOVA group.

Classification is not always consistent. In a study involving 159 to 177 French food and nutrition specialists, agreement on NOVA assignments was low, with Fleiss' kappa values of 0.32 and 0.34 for marketed products and generic foods. About one-quarter of the foods in each list drew particularly mixed assignments, even when evaluators received ingredient information.

The largest evidence review found associations of varying certainty

The 2024 BMJ umbrella review identified 45 pooled analyses covering 9,888,373 participants and reported direct associations between greater ultra-processed food exposure and 32 of 45 health outcomes. The outcomes spanned mortality, cardiovascular disease, metabolic health, mental health and other areas.

The strongest statistical category included a 50 percent higher relative risk of cardiovascular-disease mortality in higher-exposure groups and a 12 percent higher relative risk of type 2 diabetes for each increment used in the underlying dose-response analysis. Those figures are not interchangeable: one compares exposure groups, while the other measures a study-defined increase in intake.

The review's own certainty assessments also require restraint. It rated the cardiovascular-mortality evidence as very low certainty under the Grading of Recommendations Assessment, Development and Evaluation approach, while type 2 diabetes received moderate certainty in the dose-response analysis. Residual confounding, inconsistent exposure measures and the possibility that high consumers differed in other behaviors prevent the review from proving each causal link.

A controlled trial found higher intake over two weeks

A small inpatient experiment provides stronger evidence for short-term food intake and weight change. In 2019, US National Institutes of Health researchers reported that 20 adults ate about 500 more calories per day on an ultra-processed diet than on a minimally processed diet, despite menus matched for presented calories and macronutrients.

Participants spent two weeks on each diet in random order. They gained an average of 0.9 kilograms during the ultra-processed phase and lost 0.9 kilograms during the minimally processed phase. The inpatient design helped control the food offered, but the sample was small and each phase was brief, so it cannot establish years-long effects in people choosing their own meals.

Processed meat is a separate cancer finding

The cancer classification for processed meat should not be extended to every NOVA group-four product. The World Health Organization says the International Agency for Research on Cancer classified processed meat as Group 1, carcinogenic to humans, based on sufficient evidence that it causes colorectal cancer.

Group 1 describes the strength of evidence that an exposure can cause cancer, not the size of the risk. WHO says an analysis of 10 studies estimated that eating 50 grams of processed meat daily was associated with about an 18 percent increase in colorectal-cancer risk. The finding applies to processed meats such as sausages, ham and bacon, not to soft drinks, packaged snacks or every product classified as ultra-processed.

Food choices still require a nutrient check

NOVA does not replace nutrition information. Two group-four products can have different amounts of sodium, free sugar, saturated fat, fiber and micronutrients, and some group-three products can still contain enough salt or sugar to warrant comparison.

The World Health Organization's 2026 healthy-diet guidance says a variety of minimally processed and unprocessed foods low in unhealthy fats, free sugars and sodium should form the foundation of a healthy diet. It also says frozen and canned fruit and vegetables can be good choices when they do not contain added sugars or excess sodium.

A practical review of a packaged product can therefore use two questions. First, does the ingredient list contain markers of an industrial formulation; second, what does the nutrition panel report for the portion actually eaten? Replacing a frequently consumed sugary drink or salty snack with water, unsweetened tea, fruit, nuts without added salt or a less processed meal can reduce the relevant exposure without treating every processed food as the same.

A home meal with cooked vegetables, brown rice and steamed fish on a dining table (illustrative image)

Frequently asked questions

Are all frozen or canned foods ultra-processed?
No. Frozen produce can remain in NOVA group one, and vegetables or fish preserved with salt or oil may fall in group three. The ingredients and purpose of processing matter more than the container.

Does a long ingredient list prove that a food is ultra-processed?
No. It is a reason to inspect the ingredients, but NOVA focuses on industrial substances and additives used to create the formulation. List length by itself is not the classification rule.

Does the BMJ review prove that ultra-processed food causes disease?
No. It found many associations, but most underlying evidence was observational and vulnerable to residual confounding. The short NIH trial supports a causal effect on calorie intake and weight over two weeks, not every long-term health outcome.

Does IARC Group 1 mean processed meat is as dangerous as smoking?
No. Group 1 describes how strong the evidence of carcinogenicity is, not the magnitude of risk. It also applies to processed meat rather than all ultra-processed products.

What is the most useful first check?
Read the ingredient list for markers of ultra-processing, then compare sugar, sodium, saturated fat and fiber for the portion consumed. NOVA adds information about formulation but does not replace the nutrition panel or individual dietary needs.