Skip to content
HabitSeek
Back to results

Base most meals on minimally processed foods

Building most meals and snacks from unprocessed or minimally processed foods and ingredients rather than from industrial formulations such as packaged snacks, reconstituted meat products, instant noodles and sweetened packaged baked goods.

At a glance

  • Evidence: One inpatient randomised crossover trial found about 500 kcal per day higher energy intake and roughly 0.9 kg more weight gained on an ultra-processed diet than on a matched unprocessed one.Ad libitum energy intake and body weight (randomised feeding trial) · Limited certainty · Randomized intervention
  • What’s uncertain: The trial is very small, with 20 participants, and each diet period lasted only two weeks, so it cannot show what happens to weight over months or years.
  • First step: Anchor the change on meals rather than on labels: cook or assemble the main part of most meals from single ingredients, and treat packaged ready-to-eat products as the exception rather than the base.
  • Before you try it: Replacing ultra-processed food with the wrong substitute can be worse than leaving it. All precautions
Read the full takeaway

Randomized intake evidence

Shifting most meals towards minimally processed food is a reasonable default, and the single inpatient randomised trial supports a real effect on how much people eat when the two diets are matched on presented calories, energy density, macronutrients, sugar, sodium and fibre.

Cohort findings

Everything beyond that is observational. Pooled cohort analyses associate higher ultra-processed food intake with type 2 diabetes, cardiovascular death and all-cause mortality, and the umbrella reviews grade some of those associations as convincing or highly suggestive, but almost all of the pooled analyses were themselves rated low or very low quality by GRADE.

Classification and confounding

The classification is the weak point rather than the statistics: as the critical review of NOVA by Louie 2025 notes, NOVA puts fortified breakfast cereal, some dairy foods and mass-produced wholegrain bread in the same group as processed meat and soft drinks, and the cohort evidence shows the associations are carried by particular subgroups rather than by every product the classification calls ultra-processed. An E-value analysis (Robinson 2024) found that unmeasured confounding of only small-to-moderate strength would be enough to explain the prospective association with weight gain.

Overall interpretation

The honest reading is that reducing ultra-processed food is defensible and low-risk, that the weight and energy-intake signal has randomised support, and that the disease-risk numbers are more fragile than they are usually presented.

Fresh chopped vegetables ready for a saladDiscussed by3 peopleView statements ↓
NutritionLong-term healthVaries · Moderate · Moderate
Reviewed Sep 23, 2026
Save

Scientific evidence

What the research shows

Outcome assessments based on research, including its uncertainties and limitations.

Ad libitum energy intake and body weight (randomised feeding trial)

Finding: Benefit supportedCertainty: Limitedstudies found a benefit from doing this

Evidence basis: randomised trials

One inpatient randomised crossover trial found about 500 kcal per day higher energy intake and roughly 0.9 kg more weight gained on an ultra-processed diet than on a matched unprocessed one.

Population: 20 weight-stable adults with mean age 31.2 years and mean body mass index 27 kg/m2, admitted as inpatients and fed ad libitum in a randomised crossover design with two-week diet periods.

Research details

Summary

In the only randomised controlled trial to test the question directly, 20 weight-stable adults were admitted to a clinical research centre and assigned either an ultra-processed or an unprocessed diet for two weeks and then crossed over to the other for two weeks. Meals were designed to be matched for presented calories, energy density, macronutrients, sugar, sodium and fibre, and participants were told to eat as much or as little as they wanted. Energy intake was greater during the ultra-processed diet (508 +/- 106 kcal/day; p = 0.0001), with increased consumption of carbohydrate (280 +/- 54 kcal/day) and fat (230 +/- 53 kcal/day) but not protein (-2 +/- 12 kcal/day). Weight changes were highly correlated with energy intake (r = 0.8), with participants gaining 0.9 +/- 0.3 kg on the ultra-processed diet and losing 0.9 +/- 0.3 kg on the unprocessed diet. A later review of ultra-processed food and obesity in adults noted that this was the only randomised controlled trial testing causality of the association and that mechanisms proposed include increased sugar intake, decreased fibre and lower protein density.

Why this certainty

Rated Limited rather than Moderate. The randomised crossover design, controlled inpatient setting and matching of the two diets are substantial methodological strengths, and the direction of the effect is unambiguous within the trial. Certainty stays at Limited because there is a single very small trial, the exposure periods were only two weeks, blinding was impossible, and no independent replication in a randomised setting exists. The outcome tested was energy intake and short-term weight, not disease.

Limitations

The trial is very small, with 20 participants, and each diet period lasted only two weeks, so it cannot show what happens to weight over months or years. Participants could not be blinded because the diets were visibly different. It was deliberately designed to isolate processing from nutrient content by matching the diets on energy density, macronutrients, sugar, sodium and fibre, which means the result cannot be attributed to any of those matched factors and the responsible mechanism remains unidentified. It enrolled weight-stable adults with overweight rather than a general population, and the food was provided and prepared by the research kitchen, so it says nothing about adherence in ordinary life.

Effect estimate

Energy intake 508 +/- 106 kcal/day higher on the ultra-processed diet (p = 0.0001), with weight change +0.9 +/- 0.3 kg on the ultra-processed diet versus -0.9 +/- 0.3 kg on the unprocessed diet, in 20 adults over two-week periods.

Sources for this outcome:

Type 2 diabetes, cardiovascular mortality and all-cause mortality (observational)

Finding: Benefit supportedCertainty: Moderatestudies found a benefit from doing this

Evidence basis: observational studies (association, not proof of cause)

Umbrella reviews of pooled cohorts associate higher ultra-processed food intake with type 2 diabetes and with cardiovascular and all-cause mortality, but grade most of that pooled evidence as low or very low quality.

Population: Adults in prospective cohort studies pooled across umbrella and dose-response meta-analyses: 9,888,373 participants in the largest umbrella review, 5,750,133 individuals across 40 prospective cohorts in the pooled all-cause mortality meta-analysis of Taneri 2022, and 114,064 US health professionals followed for a median of 31 to 34 years in the largest single cohort.

Research details

Summary

An umbrella review of 45 pooled analyses covering 9,888,373 participants classified evidence as convincing for a direct association between greater ultra-processed food exposure and incident type 2 diabetes (dose-response risk ratio 1.12, 95% CI 1.11 to 1.13; GRADE moderate) and cardiovascular disease related mortality (risk ratio 1.50, 95% CI 1.37 to 1.63; GRADE very low), and as highly suggestive for all-cause mortality (risk ratio 1.21, 95% CI 1.15 to 1.27; GRADE low) and heart disease related mortality (hazard ratio 1.66, 95% CI 1.51 to 1.84; GRADE low). A separate umbrella review that updated meta-analyses with 122 additional articles across 49 outcomes found 25 health outcomes associated with ultra-processed food consumption, with convincing evidence only for renal function decline and wheezing in children and adolescents, and highly suggestive evidence for diabetes mellitus, overweight, obesity, depression and common mental disorders. Pooled cohort estimates for type 2 diabetes are consistent in direction with the umbrella findings, with the largest single cohort analysis reporting a 4% higher all-cause mortality in the highest quarter of ultra-processed food intake (hazard ratio 1.04, 95% CI 1.01 to 1.07) and no association with cancer or cardiovascular mortality.

Why this certainty

Rated Moderate. The direction replicates across independent umbrella reviews with very large participant numbers, the dose-response gradients are statistically robust, and the classification of convincing or highly suggestive evidence was made using pre-specified criteria with GRADE. Certainty is held below Strong because there is no randomised trial of disease outcomes, the great majority of the pooled analyses were rated low or very low quality — in the largest umbrella review, 22 pooled analyses were low quality, 19 very low and only four moderate — the exposure classification is contested and heterogeneous, and a formal confounding analysis found the association with weight gain is small enough to be explained by unmeasured confounding. It is held at Moderate rather than Limited because the type 2 diabetes dose-response association survived stringent credibility criteria and was graded moderate quality by the same review that graded the other 41 pooled analyses low or very low.

Limitations

The evidence is entirely observational, and the umbrella reviews themselves grade it down: in the largest one, 22 pooled analyses were rated low quality, 19 very low quality and only four moderate quality. The exposure is contested, because NOVA classifies fortified breakfast cereals and some dairy foods alongside processed meat and soft drinks, and one meta-analysis of 40 prospective cohorts found breakfast cereals associated with lower all-cause mortality (risk ratio 0.85, 95% CI 0.79 to 0.92) while sugar-sweetened beverages (1.11, 1.04 to 1.18), artificially sweetened beverages (1.14, 1.05 to 1.22) and processed meat or red meat (1.15, 1.10 to 1.21) were associated with higher mortality, and the highest category of ultra-processed food overall with 1.29 (1.17 to 1.42). Associations within the largest cohort were not consistent across quarters of overall diet quality, and they varied widely by ultra-processed food subgroup (hazard ratios 1.06 to 1.43). An E-value analysis of the prospective association with weight gain (weight-gain RR 1.14, E value 1.55) estimated that unmeasured confounding of only small-to-moderate strength could explain that association. Dietary exposure is self-reported and measured at intervals, and reverse causation is possible because early disease changes both what people eat and their mortality risk.

Effect estimate

Type 2 diabetes dose-response risk ratio 1.12 (95% CI 1.11 to 1.13) per increment in ultra-processed food exposure, and all-cause mortality risk ratio 1.21 (95% CI 1.15 to 1.27) for highest versus lowest intake, from umbrella review of pooled cohort analyses.

Sources for this outcome:

What people say

Who discusses this?

Public statements, with positions specific to each source and topic. These views do not determine the scientific evidence assessment.

What the research shows ↑

Practical steps and precautions

How to try it

  1. Anchor the change on meals rather than on labels: cook or assemble the main part of most meals from single ingredients, and treat packaged ready-to-eat products as the exception rather than the base.
  2. Use the practical NOVA distinction rather than a strict ban. The trial's benefit came from an unprocessed diet built from whole foods, and the evidence implicates particular subgroups rather than every product the classification calls ultra-processed: the critical review notes that sugar-sweetened beverages are consistently adverse while fortified cereals and certain dairy products are neutral or protective, and the pooled mortality analysis found breakfast cereals associated with lower all-cause mortality (risk ratio 0.85, 95% CI 0.79 to 0.92).
  3. Prioritise the subgroups the data actually implicate: processed and reconstituted meat products, sugar-sweetened and artificially sweetened drinks, packaged desserts, and instant or ready-to-eat snack foods.
  4. Do not strip out fortified wholegrain breads and cereals on processing grounds alone. Analyses of national survey data found that excluding ultra-processed grain foods would lower intakes of key nutrients for groups such as women of childbearing age.

Precautions

  • Replacing ultra-processed food with the wrong substitute can be worse than leaving it. If the replacement is a home-cooked diet based on red and processed meat, the pooled cohort evidence implicates those subgroups most strongly of all (processed meat and red meat risk ratio 1.15, 95% CI 1.10 to 1.21 for all-cause mortality).
  • Wholesale avoidance can reduce intake of fortified nutrients. Fortification analyses found exclusion of ultra-processed grain foods would lower intakes of nutrients of concern for at-risk groups, and that realistic non-ultra-processed substitutes may not be affordable or available for everyone.
  • Cost and access are real constraints. The critical review notes that ultra-processed products often supply cheap calories and nutrients for disadvantaged households, so treat this as a shift in the balance of the diet rather than a categorical exclusion.
  • Do not present this as a treatment for diagnosed type 2 diabetes or heart disease. The randomised evidence tested energy intake and body weight; the disease outcomes come from observational cohorts only.

References

Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain: An Inpatient Randomized Controlled Trial of Ad Libitum Food Intake.

Hall KD, Ayuketah A, Brychta R, Cai H, Cassimatis T, Chen KY, Chung ST, Costa E, Courville A, Darcey V, Fletcher LA, Forde CG, Gharib AM, Guo J, Howard R, Joseph PV, McGehee S, Ouwerkerk R, Raisinger K, Rozga I, Stagliano M, Walter M, Walter PJ, Yang S, Zhou M. · Cell metabolism · 2019

Inpatient randomised controlled crossover feeding trialVerified from abstract
Open original source

Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses.

Lane MM, Gamage E, Du S, Ashtree DN, McGuinness AJ, Gauci S, Baker P, Lawrence M, Rebholz CM, Srour B, Touvier M, Jacka FN, O'Neil A, Segasby T, Marx W. · BMJ (Clinical research ed.) · 2024

Umbrella review of meta-analyses of cohort, case-control and cross-sectional studiesVerified from full text
Open original source

Ultra-processed foods and human health: An umbrella review and updated meta-analyses of observational evidence.

Dai S, Wellens J, Yang N, Li D, Wang J, Wang L, Yuan S, He Y, Song P, Munger R, Kent MP, MacFarlane AJ, Mullie P, Duthie S, Little J, Theodoratou E, Li X. · Clinical nutrition (Edinburgh, Scotland) · 2024

Umbrella review with updated meta-analyses of observational studiesVerified from abstract
Open original source

Association of ultra-processed food consumption with all cause and cause specific mortality: population based cohort study.

Fang Z, Rossato SL, Hang D, Khandpur N, Wang K, Lo CH, Willett WC, Giovannucci EL, Song M. · BMJ (Clinical research ed.) · 2024

Population based prospective cohort studyVerified from full text
Open original source

Association Between Ultra-Processed Food Intake and All-Cause Mortality: A Systematic Review and Meta-Analysis.

Taneri PE, Wehrli F, Roa-Díaz ZM, Itodo OA, Salvador D, Raeisi-Dehkordi H, Bally L, Minder B, Kiefte-de Jong JC, Laine JE, Bano A, Glisic M, Muka T · American journal of epidemiology · 2022

Systematic review and meta-analysis of prospective cohort studiesVerified from abstract
Open original source

Causality or confounding? Applying E values to examine associations between ultra-processed food consumption and risk of weight gain.

Robinson E, Jones A · International Journal of Obesity · 2024

Meta-analysis of prospective studies with E-value sensitivity analysisVerified from abstract
Open original source

Fortification of grain foods and NOVA: the potential for altered nutrient intakes while avoiding ultra-processed foods.

Estell ML, Barrett EM, Kissock KR, Grafenauer SJ, Jones JM, Beck EJ · European Journal of Nutrition · 2022

Quantitative analysis of a national nutrition survey with dietary modellingVerified from abstract
Open original source

Are all ultra-processed foods bad? A critical review of the NOVA classification system.

Louie JCY · The Proceedings of the Nutrition Society · 2025

Critical reviewVerified from abstract
Open original source

Community

Discussion

Have you tried this? What helped or made it difficult?

Reader experiences are separate from the scientific evidence. Explore all discussions

No discussions yet.

Be the first to ask a question or share your experience.

Sign in to vote or join the discussion.

Related habits

Explore other habits in a shared category. Each has its own evidence and precautions.

Editorial history and review method

Revision 3 · Approved for publication · Codex (publication approval explicitly authorized by site owner; category migration)

Prepared and checked with AI. Publication approval does not imply independent expert review. Request a correction.

Audit correction pass (group 7). Sources the entry's claims depended on but did not cite are now cited, and one previously unidentified source was retrieved. (1) The population field's '5,750,133 individuals across 40 prospective cohorts' and the limitations' 'fortified breakfast cereals ... risk ratio 0.85, 95% CI 0.79 to 0.92' were unsourced: they come from Taneri 2022 (DOI 10.1093/aje/kwac039, PMID 35231930), a meta-analysis of 40 prospective cohorts in 5,750,133 individuals, which reports breakfast cereals RR 0.85 (0.79 to 0.92), sugar-sweetened beverages 1.11 (1.04 to 1.18), artificially sweetened beverages 1.14 (1.05 to 1.22), processed meat/red meat 1.15 (1.10 to 1.21) and highest-versus-lowest ultra-processed food 1.29 (1.17 to 1.42). That record was retrieved from Europe PMC and added to content/sources.json as taneri-2022-upf-mortality; the claim is therefore sourced rather than deleted. (2) robinson-2024-upf-evalue, estell-2022-fortification-nova and louie-2025-nova-critical-review were added to the outcome's sourceIds, which substantiates the E-value sentence in verdict, limitations and certaintyRationale, the fortification sentences in implementation 4 and precaution 2, and the NOVA critique in verdict and precaution 3. (3) The unsupported subgroup sentence in the verdict and implementation 2 was rewritten to the subgroups the cited sources actually report (processed meat and sweetened drinks from Fang 2024; breakfast cereals from Taneri 2022, and with the direction stated correctly as lower mortality); the 'dairy desserts' attribution to the strongest associations was dropped, since Fang 2024 reports dairy-based desserts only for all-cause mortality among many subgroups. (4) Card summary corrected: the E-value result applies to the prospective association with weight gain, not to diabetes or death, and the card now says so. (5) certaintyRationale now states the quality asymmetry explicitly (22 low, 19 very low, 4 moderate in the largest umbrella review) while keeping Moderate on the strength of the GRADE-moderate type 2 diabetes dose-response. A second reviewer must re-check: (1) that the Taneri 2022 estimates are transcribed exactly as its abstract prints them, including the mixed punctuation around the confidence intervals; (2) that the type 2 diabetes and mortality outcome is not double-counting pooled analyses across lane-2024-upf-umbrella, dai-2024-upf-umbrella and taneri-2022-upf-mortality, which was an open item before this pass and is now more pressing with a third synthesis cited; (3) whether the certainty should drop to Limited instead, which the audit offered as the alternative option; (4) that the entry's stated source count ('seven sources') now matches the record; and (5) that no sentence implies a randomised trial of disease outcomes, since none exists. Publication release 2026-10-04: site owner authorized a bounded AI audit and release of this exact revision. See docs/release-audit-2026-10-04.md for source-check findings and limits. No independent health-professional or peer review is claimed. Category migration 2026-10-04: removed the separate goals field; categories: Nutrition, Long-term health → Nutrition, Long-term health. Research assessments, sources, guidance, precautions, and evidence review dates are unchanged. This revision awaits publication approval; no new evidence review is claimed. Publication approval 2026-10-04: site owner explicitly accepted all current category-migration revisions for deployment and requested that remaining limitations be retained in docs/content-review-backlog.md. Approval applies to this exact content revision. No new evidence review, independent expert review, or clinical validation is claimed; existing precautions and disclosures remain in effect.