Beyond the Ingredient List: Ultra-Processed Food, Appetite and the Limits of Causal Evidence

Abstract

Background: Ultra-processed foods have become a prominent explanation for deteriorating population health. Public discussion, however, often conflates industrial processing, poor nutritional composition and the presence of additives. These exposures overlap but are not interchangeable.

Objective: To examine what selected experimental and observational evidence establishes about ultra-processed dietary patterns, distinguish causal findings from mechanistic hypotheses, and develop proportionate implications for everyday behaviour change.

Approach: A critical narrative synthesis of selected classification research, controlled feeding trials, an epidemiological umbrella review and methodological commentary. No systematic search, independent meta-analysis or formal certainty assessment was undertaken.

Synthesis: Controlled trials support causal effects of particular dietary interventions on energy intake and short-term weight change. They do not isolate a universal effect of processing independently of food composition and physical characteristics. Observational associations extend across multiple health outcomes, but their interpretation remains vulnerable to confounding, exposure misclassification and reverse causation. Nutritional quality and processing should therefore be assessed as related dimensions rather than competing explanations.

Conclusion: The evidence supports practical efforts to improve dietary patterns, including greater reliance on nutritionally adequate, minimally processed foods. It does not support blanket toxicity claims or the assumption that ingredient-list length predicts health risk. The most useful intervention identifies an appropriate replacement and makes that replacement sustainable.

The problem with “then versus now”

Images contrasting traditional recipes with contemporary packaged foods offer an intuitively persuasive narrative: familiar products have become more complex, and that complexity has made them harmful.

Three distinct propositions are embedded in this argument: a product’s formulation has changed; the change has altered its nutritional or physiological effects; and those altered effects have increased disease risk. Each requires different evidence.

Historical ingredient records can establish reformulation. Controlled comparisons can investigate physiological consequences. Longitudinal research can examine disease outcomes. A photograph of homemade vanilla ice cream beside a branded dessert establishes none of these propositions independently.

The scientific problem is therefore broader than whether an ingredient sounds unfamiliar. It concerns which measurable characteristics of foods influence consumption and health, under what conditions, and compared with which alternatives.

This review advances a central argument: dietary advice becomes more defensible when it specifies both the food being reduced and the food replacing it.

What does “ultra-processed” measure?

The NOVA framework groups foods according to the extent and purpose of processing. Its ultra-processed category generally encompasses industrial formulations involving food-derived substances and ingredients used to modify flavour, appearance or texture. It is not defined simply by a long ingredient list. [1]

A classification system can be useful without being a complete biological explanation. Three questions must remain separate: can researchers consistently assign foods to a category; does category membership help predict subsequent outcomes; and which properties of those foods produce the observed effects? Evidence for one does not automatically answer the others.

Braesco and colleagues examined the application of NOVA by food and nutrition specialists and documented inconsistencies in classification. Such findings warrant attention to coding procedures and the product information available to assessors. They do not, by themselves, invalidate every study using the framework. [2]

The methodological implication is straightforward: a category whose assignment varies introduces uncertainty into exposure measurement. Researchers should make that uncertainty visible rather than treat processing status as an error-free measurement.

The inpatient trial: a causal effect of the diets tested

Hall and colleagues randomised 20 adults to two consecutive two-week dietary conditions in an inpatient crossover trial. Meals were designed to match several presented nutritional characteristics, while participants could consume as much or as little as desired.

During the ultra-processed condition, mean energy intake was 508 kcal/day higher. Participants gained approximately 0.9 kg during that condition and lost approximately 0.9 kg during the unprocessed condition. [3]

Randomisation and controlled food provision strengthen the inference that the assigned diets caused the observed difference within this setting.

The scope of that inference matters. The intervention compared complete menus. It did not independently randomise processing, texture, additives and individual food types. Consequently, it establishes a causal effect of the dietary contrast, without identifying which component or combination produced that effect.

The short duration and small sample also limit extrapolation. The observed difference cannot responsibly be multiplied across a year to predict individual weight gain: that would assume an unchanged response, adherence and energy balance far beyond what was measured.

Guideline-concordant diets: a more demanding comparison

Dicken and colleagues compared two eight-week diets that both followed the UK Eatwell Guide. Fifty-five adults were randomised; 50 contributed to the intention-to-treat analysis.

Both conditions produced weight loss. Mean percentage weight change was −2.06% on the minimally processed diet and −1.05% on the ultra-processed diet. The estimated within-participant difference was −1.01 percentage points, with a 95% confidence interval of −1.87 to −0.14. [4]

This supports an additional benefit of the minimally processed dietary condition for the primary outcome. It does not show that the ultra-processed condition caused weight gain.

The confidence interval is also informative: the plausible magnitude ranges from a small difference to a more substantial one over eight weeks. Statistical significance alone does not establish long-term clinical importance.

The minimally processed diet had lower energy density, and the biomarker findings were mixed; LDL cholesterol decreased more under the ultra-processed condition. Weight results should therefore not be presented as evidence of superiority across every measured outcome. [4]

A disagreement about the causal question

Methodological commentary by Louie argues that interpretations of feeding trials sometimes overstate processing-specific causation and underemphasise neutral or favourable findings in ultra-processed conditions. This is a critical perspective, rather than a new experiment. [5]

The underlying disagreement can be clarified by distinguishing two questions. The practical question asks what happens when people receive these different dietary patterns. The mechanistic question asks what happens when processing changes while other relevant properties remain constant. A trial can answer the first convincingly while leaving the second unresolved.

There is a further complication. If processing changes texture or energy density, those characteristics may be mediators of its effects rather than unrelated confounders. Matching them away could remove part of the effect being investigated.

This is a methodological interpretation: researchers must define whether they seek the total effect of a dietary strategy or an effect independent of particular pathways. Neither question should silently substitute for the other.

Observational evidence: broad concern, uneven certainty

Lane and colleagues’ umbrella review evaluated 45 pooled analyses and reported associations between greater ultra-processed food exposure and 32 health parameters. Under GRADE, four analyses were rated moderate quality, 22 low and 19 very low. [6]

These findings justify concern, but the number of associated outcomes is not equivalent to the number of independently proven causal effects.

Residual confounding: food choices occur within social and behavioural circumstances. Statistical adjustment cannot guarantee that all relevant differences have been measured adequately.

Reverse causation: existing illness, distress or functional limitations may change purchasing and eating behaviour.

Exposure misclassification: dietary assessment may not capture the manufacturing details needed for precise classification.

Comparator ambiguity: a higher proportion of energy from one food category necessarily leaves less energy from others. The association therefore partly depends on what is displaced.

Large datasets improve statistical precision. They do not automatically correct systematic measurement error or establish causality.

Relative risks also require care. A relative increase cannot be translated into an individual’s absolute probability without an appropriate baseline risk, exposure definition and time horizon.

Why the replacement matters

“Reduce ultra-processed food” is an incomplete intervention description.

Replacing a sweetened drink with water, replacing a convenient meal with an inadequate snack, and replacing a packaged staple with a nutritionally similar alternative are different interventions. Their consequences should not be assumed identical.

The following practical framework is an authorial synthesis, not a validated assessment instrument. Its purpose is to move advice from category avoidance towards a specified, assessable change.

• Nutritional contribution: what does the current food provide, and what is missing?
• Consumption pattern: how much is eaten, how frequently, and in what context?
• Replacement: what will take its place?
• Feasibility: can that alternative be afforded, obtained, stored and prepared?
• Evaluation: what outcome would indicate that the change is useful?

Implications for behavioural practice

From a psychotherapeutic and coaching perspective, the useful contribution is to help people understand the circumstances surrounding repeated choices and develop feasible alternatives.

That work should remain distinct from diagnosing nutritional deficiencies, treating metabolic disease or making claims that psychotherapy reverses food-related physiological harm.

A practical conversation can begin with one recurring situation: what happens on the evenings when you arrive home exhausted? What food is available? What could make a satisfactory meal easier?

This shifts the proposed intervention towards the setting in which the behaviour occurs.

The outcome should also match the purpose. If the aim is greater dietary variety, variety should be assessed. If the aim is making meals manageable, preparation burden and consistency matter. Weight should not become the default measure of success for every person.

These are practice proposals derived from the analysis. The cited dietary trials did not test this coaching framework.

A research agenda

Stronger evidence would address several complementary questions. Mechanistic trials and pragmatic trials serve different purposes. The former can isolate pathways; the latter can establish whether an intervention works under ordinary constraints. A useful evidence base needs both.

• Which ultra-processed subcategories account for observed associations?
• How do texture and energy density contribute to effects under controlled conditions?
• Do short-term differences persist when people purchase and prepare their own food?
• How do cost, adherence and preparation burden alter real-world effectiveness?
• Which specified replacements improve meaningful health outcomes?

Limitations

This review is selective rather than systematic. It presents central studies and methodological issues without claiming comprehensive coverage of all evidence available in October 2026.

No new participant data were collected, no pooled effect was calculated, and no independent risk-of-bias assessment was performed. The proposed behavioural framework requires separate evaluation.

These limitations define what the article can support: a critical interpretation and practical perspective, rather than a definitive clinical guideline.

Conclusion

The experimental evidence warrants taking ultra-processed dietary patterns seriously. Particular dietary contrasts have produced measurable differences in spontaneous intake and short-term weight outcomes.

The same evidence does not establish that every ultra-processed product is harmful, that additives explain the observed effects, or that ingredient-list length is a reliable measure of risk.

For practical guidance, the strongest position combines attention to processing with nutritional adequacy, the replacement food and the person’s circumstances.

A useful dietary recommendation should explain what to change, what to choose instead, and how to make that choice sustainable.

About the author

Benjamin Brixley is an NLP psychotherapist and founder of ORIGIN, a company developing wellbeing products and services. His work focuses on personal development and sustainable behaviour change.