Ultra-Processed Food Research: What's Been Measured

This content is educational. It describes the research on ultra-processed foods. It is not dietary advice.

Why this matters

Ultra-processed food has become one of the most-discussed topics in consumer nutrition over the past decade. Headlines confidently claim that "ultra-processed food causes obesity," "UPFs drive cancer," or "modern food is killing us." The research is genuinely interesting but considerably more nuanced than the headlines.

This page describes what the published research has actually measured about ultra-processed foods — how the category is defined, what observational studies have shown, what the single major RCT (Kevin Hall's 2019 metabolic-ward trial) found, and where the methodological limits sit.

It is not a guide to what to eat. It is a guide to reading the rapidly-growing UPF research literature with appropriate calibration.

What ultra-processed food actually means

The most-used definition comes from the NOVA classification system, developed by Carlos Monteiro and colleagues at the University of São Paulo. NOVA categorises foods by degree of industrial processing, not nutrient content [Monteiro et al. 2019]:

Group 4 (ultra-processed) is the contested category. Examples include sugar-sweetened soft drinks, mass-produced packaged breads, breakfast cereals with added flavours and colours, packaged snacks, instant noodles, ready meals, energy bars, plant-based meat substitutes, many protein powders.

The classification is based on processing complexity and typical ingredient lists, not nutrient profile. A breakfast cereal can be ultra-processed and nutritionally fortified. A home-baked cake using flour, butter, eggs, and sugar would be processed but not ultra-processed.

This is part of what makes the NOVA system contested — it treats two foods with similar nutrient profiles differently if one is industrially produced.

What observational studies have measured

Observational research on UPF intake and health outcomes has grown rapidly. Several large cohorts have reported consistent patterns:

Cardiovascular outcomes

The NutriNet-Santé cohort in France (n>100,000) reported that each 10% increase in UPF share of diet was associated with a 12% higher risk of cardiovascular disease over follow-up [Srour et al. 2019].

The SUN cohort in Spain reported similar magnitude associations between UPF intake and cardiovascular mortality [Rico-Campà et al. 2019].

A 2024 meta-analysis pooling cohort data on UPF intake and mortality concluded that higher UPF consumption was associated with elevated all-cause mortality risk across cohorts [Lane et al. 2024].

Cancer

Several cohort studies have reported associations between UPF intake and certain cancers — particularly breast and colorectal cancer. Effect sizes are typically modest (relative risks 1.10-1.25 for high vs low UPF tertiles) [Fiolet et al. 2018].

Metabolic outcomes

UPF intake has been associated with type 2 diabetes risk, weight gain, and metabolic syndrome markers across multiple cohorts [Chen et al. 2023].

Mental health

A growing literature has reported associations between high UPF intake and depression incidence, though confounding by diet quality more broadly is hard to rule out [Lane et al. 2022].

Why the observational data has limitations

The observational pattern is real and consistent. Several methodological problems make it difficult to attribute the associations to ultra-processing specifically:

Diet pattern confounding. People who eat more UPF tend to eat fewer fruits, vegetables, whole grains, and fish. Statistically separating "UPF effect" from "overall diet quality" is hard.

Lifestyle confounding. UPF intake is associated with lower socioeconomic position, less physical activity, higher smoking rates, and reduced healthcare access — all of which independently affect health outcomes.

Reverse causation. People with developing chronic disease may shift toward more convenient (often ultra-processed) food rather than UPF causing their disease.

Self-reported intake. All cohort studies rely on food frequency questionnaires with their well-documented accuracy limitations (covered in the nutrition research methodology piece).

Measurement of UPF specifically. NOVA classification of cohort dietary data is often done retrospectively and approximately, with substantial inter-coder variability [Braesco et al. 2022].

The observational literature establishes a real association. It does not establish that ultra-processing causes the observed outcomes.

The one major RCT (Kevin Hall, 2019)

A single trial has directly tested the causal hypothesis under controlled conditions. It is methodologically important and worth understanding in detail.

In 2019, Kevin Hall and colleagues at the NIH conducted a metabolic-ward study with 20 inpatient adults. Each participant spent 2 weeks eating an ultra-processed diet (NOVA Group 4) and 2 weeks eating an unprocessed diet, in random order. Crucially, the diets were matched for total calories, macronutrients, sugar, sodium, and fibre. Participants were told to eat as much or as little as they wanted [Hall et al. 2019].

Findings:

• On the UPF diet, participants consumed approximately 500 more calories per day than on the unprocessed diet — despite the diets being nutritionally matched
• Participants gained an average of 0.9 kg over 2 weeks on the UPF diet
• Participants lost an average of 0.9 kg over 2 weeks on the unprocessed diet
• The participants didn't report eating more — the increased intake was unconscious
• The UPF was eaten faster (more bites per minute) than the unprocessed food

This is the strongest experimental evidence that ultra-processed food drives increased caloric intake even when nutritionally matched to less-processed alternatives. The mechanism remains debated — possibilities include energy density, eating rate, palatability, hedonic reward, or food matrix effects.

Limitations of the trial: short duration, small sample, inpatient setting (not real-world), single research group (not yet replicated). But the design is methodologically strong and the finding is widely cited.

What's actively being researched

The UPF literature is moving fast. Active research questions include:

• Mechanism. Is it specific additives (emulsifiers, sweeteners, colours), food matrix structure (how nutrients are bound), energy density, eating rate, or hedonic reward that drives the observed effects?
• Heterogeneity within UPF. All NOVA Group 4 foods are not equivalent. Yoghurt with added flavourings is ultra-processed but very different from an energy drink. Some research is trying to identify which specific UPF subcategories drive the strongest health associations.
• Dose-response. Is there a threshold below which UPF intake has minimal effect, or is the effect linear from zero?
• Reformulation. Can ultra-processed foods be reformulated to remove the problematic features while keeping the convenience? Industry research is heavily focused here.
• Specific ingredient research. Emulsifiers, artificial sweeteners, and certain colours are being studied for individual contributions to metabolic and gut-microbiome outcomes.

What can be said with confidence in 2026

Calibrated summary of the research as of mid-2026:

Established with reasonable confidence:

• UPF intake is associated with adverse health outcomes across multiple large cohorts
• The association is consistent across countries and dietary patterns
• Under controlled conditions, UPF diets produce overconsumption and weight gain compared with nutrient-matched unprocessed diets
• Effect sizes for individual outcomes are modest (10-25% relative risk increases per intake category)

Less established:

• The specific mechanism by which UPF drives the observed outcomes
• Whether the entire NOVA Group 4 category behaves homogeneously
• Whether reformulation can neutralise the effects while retaining convenience
• The contribution of ultra-processing per se versus correlated lifestyle factors

Where marketing has gone beyond evidence:

• "UPF causes [specific disease]" — usually overstated; observational associations are not causal proofs
• "Avoiding all UPF will [specific outcome]" — no trials have measured this
• Specific anti-UPF intervention claims (apps, programs, supplements) — mostly unvalidated

How to read UPF claims in consumer content

When evaluating UPF claims, the methodology checks are the same as for other nutrition research (see our nutrition methodology piece):

1. Is the claim based on observational or experimental data?
2. What was the effect size? "Significantly higher risk" without numbers is uninformative
3. Was the comparison group eating similar amounts of food, or just different food?
4. Has the finding replicated across cohorts in different populations?
5. Is the claim about ultra-processed food in general or about specific subcategories?

Confident "UPF causes X" claims usually elide the methodological complexity. More careful claims acknowledge the observational nature of most data and the single RCT's findings about caloric overconsumption.

Related Proco pages

• Why nutrition research is uniquely hard
• How to read a clinical trial
• The wellness economy in 2026

Sources

Proco provides educational, research-based information. This page describes what the research has measured about ultra-processed food. It is not dietary advice. Decisions about your own diet belong with a qualified healthcare professional or registered dietitian, particularly if you manage a chronic condition, are pregnant or breastfeeding, or have a history of disordered eating.

Schema (for implementation)

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Ultra-Processed Food Research: What's Been Measured",
  "description": "Observational research consistently associates ultra-processed food intake with adverse health outcomes. The single controlled trial (Hall 2019) demonstrated calorie overconsumption on UPF diets. This page describes what each has measured.",
  "datePublished": "2026-06-01",
  "dateModified": "2026-05-31",
  "author": {"@type": "Organization", "name": "Proco"},
  "publisher": {"@type": "Organization", "name": "Proco", "url": "https://procohq.com"},
  "about": {"@type": "Thing", "name": "Ultra-processed food research"}
}