BMJ Umbrella Review on Ultra-Processed Foods: What the NOVA Evidence Means for Reformulation

Introduction

In February 2024 The BMJ published the largest overview yet of the evidence on ultra-processed foods (UPFs) and health: an umbrella review by Lane and colleagues that pooled earlier meta-analyses and graded how credible each association was.^[2] It is now the study most often quoted in debates on UPFs, front-of-pack rules and product reformulation. This article sets out what the review found, how strong the evidence is, where the NOVA classification behind it is criticised, and what a brand can sensibly do with the findings when it reformulates.

What Is the NOVA Classification?

The four NOVA groups

NOVA sorts foods into four groups by the extent and purpose of industrial processing: (1) unprocessed or minimally processed foods, (2) processed culinary ingredients, (3) processed foods and (4) ultra-processed foods.^[1] It is the most commonly used food-processing classification worldwide and has been recognised in reports by the Food and Agriculture Organization of the United Nations and the Pan American Health Organization.^[2] In research, exposure is usually measured as the share of the diet that comes from UPFs, by energy or by weight.^[1] For the processing steps behind each group, see our guide to processing methods behind each NOVA group and the four categories of processed foods.

How NOVA decides that a food is ultra-processed

The Lancet Series paper by NOVA’s developers and collaborators describes UPFs as branded, commercial formulations made from cheap ingredients extracted or derived from whole foods and combined with additives; most contain little to no whole food.^[1] The deciding test is the ingredient list, not a threshold for sugar, salt or fat. A product is ultra-processed when it contains any of NOVA’s markers:^[1]

• Food substances of exclusive or almost exclusive industrial use, such as plant protein isolates, mechanically separated meat, and modified starches and oils.
• Sensory or “cosmetic” additive classes, such as colours, flavours, flavour enhancers, non-sugar sweeteners and emulsifiers.

Preservatives and other shelf-life additives are also found in group 3 processed foods, so on their own they do not mark a product as ultra-processed.^[1] Two products with the same nutrient profile can therefore sit in different NOVA groups, and that is the source of both the research signal and the reformulation challenge.

What the 2024 BMJ Umbrella Review Found

Scope and strength of the evidence

Lane and colleagues searched for meta-analyses published from 2009 to June 2023 and included 14 meta-analyses containing 45 unique pooled analyses: 13 dose-response and 32 non-dose-response. The pooled analyses had a combined 9,888,373 participants (people can appear in more than one analysis). Higher UPF exposure was directly associated with a worse outcome in 32 of the 45 analyses (71%), spanning mortality, cancer, and mental, respiratory, cardiovascular, gastrointestinal and metabolic health.^[2]

The review graded each association twice. Evidence class (I convincing to V no evidence) tests the statistics: sample size, P value, heterogeneity, prediction intervals and signs of bias. GRADE certainty rates how far the result can be trusted. On GRADE, 4 pooled analyses were moderate, 22 low and 19 very low. Using AMSTAR 2, the authors also rated their confidence in 7 of the 14 meta-analyses as low, mainly because it was unclear whether review methods had been set in advance.^[2]

Which health outcomes were linked to UPFs

Four associations met the class I (convincing) criteria and seven more met class II (highly suggestive).^[2]

Table 1. Class I and II associations in the 2024 BMJ umbrella review

Source: Lane et al., 2024.^[2] RR, risk ratio; OR, odds ratio; HR, hazard ratio. “Prevalent” outcomes come largely from cross-sectional data, which cannot show whether UPF intake came before the outcome.

Evidence class is not certainty. The cardiovascular mortality association was convincing on the statistics but very low on GRADE, while the dose-response diabetes association was the only class I result with moderate certainty. That diabetes estimate is small (about 12% higher risk for each 10% increase in the UPF share of the diet) but tightly estimated.^[2]

Criticisms and Limits of NOVA and the Evidence

Association, not causation

Every study in the review was observational. The authors state that they did not examine how each study adjusted for confounders, that different dietary assessment tools introduce measurement bias, and that residual confounding is a limitation of this kind of evidence.^[2] They also found excess significance bias in 9 of the 28 pooled analyses with three or more studies, including those for all-cause mortality, obesity and type 2 diabetes.^[2] An umbrella review cannot correct biases in the studies it pools.

The authors give the counter-argument too. In their view, the consistency of findings across most analyses suggests residual confounding does not fully explain them, and a recent meta-analysis found that adjusting for diet quality did not remove the associations. They note that long-term trials on hard outcomes such as cardiovascular disease would be unethical, so only short-term trials on intermediate outcomes are feasible.^[2]

The UK Scientific Advisory Committee on Nutrition (SACN) called the associations concerning but advised caution. It noted inconsistent adjustment for covariables between studies and said it is unclear whether the associations are explained by the high energy density, salt, saturated fat or free sugars typical of many UPFs.^[3]

What trials add

Controlled trials are few. The review authors point to a 2019 inpatient trial in which an ultra-processed diet led to about 500 kcal a day more energy intake and 0.9 kg weight gain compared with a minimally processed diet.^[2] In the 2025 UPDATE crossover trial, 55 adults with overweight or obesity were randomised to two 8-week diets, both following the UK Eatwell Guide, with 50 included in the analysis. Weight fell on both diets, but more on the minimally processed diet (−2.06% vs −1.05%; difference −1.01%, 95% CI −1.87 to −0.14). LDL cholesterol fell only on the UPF diet.^[4] The UPF diet included reformulated products such as breakfast cereals, ready meals and plant-based alternatives. Limits: 90.9% of those randomised were women, the trial was short, and there was a significant diet-order effect.^[4]

Where NOVA is hard to apply

SACN found NOVA’s categories broad, grouping foods with different nutritional attributes. It also noted varying inter-assessor agreement and found some NOVA classifications discordant with nutritional classifications. Because the literature is dominated by NOVA, any limitation in the system may repeat across studies.^[3] Within the review, a type 2 diabetes meta-analysis found some UPF subgroups (including dark or wholegrain bread, yoghurt and dairy-based desserts) inversely associated with risk, even though total UPF intake was linked to higher risk.^[2] The Lancet Series, written by NOVA’s developers and collaborators, argues that the totality of the evidence already justifies policy action.^[1] For a formulator, NOVA group 4 is best treated as a flag for review, not a verdict on a single product.

What It Means for Reformulation

Fix the nutrient profile first

The evidence cannot yet separate processing from nutrient profile,^[3] so a reformulation plan has to address both. Start with what is well established: energy density, salt, saturated fat and free sugars.^[3] Then rebuild what high-UPF diets tend to lack. Diets with a higher UPF share contain less fibre, protein, potassium, zinc, magnesium and several vitamins.^[1] A cleaner label that loses fibre or protein trades one problem for another.

Audit the NOVA markers, one function at a time

Because any single marker places a product in group 4,^[1] removing one additive rarely changes its NOVA group. List every marker on the ingredient declaration, record the job it does in the product, and decide whether that job can be done another way. The table below is FRL formulation practice, not a finding of the cited studies.

Table 2. NOVA markers, their functions and reformulation routes (FRL view)

Each swap moves a risk somewhere else. Emulsifier or modified-starch removal is where shelf life most often fails, so plan a shelf-life study alongside the bench work; our article on how nutrient-led reformulation can break emulsion stability and shelf life covers the same trade-off for nutrient targets. Reformulation alone also may not close the gap: the UPF diet in the UPDATE trial met dietary guidelines and still produced less weight loss than the minimally processed diet.^[4]

Keep claims inside the evidence

The review shows associations, not proof of cause.^[2] Avoid disease-risk or “healthier because less processed” claims that the evidence cannot support, and describe what actually changed: fewer additives, more fibre, less salt.

Watch the US definition

In the US (see our summary of CDC data on ultra-processed foods in US diets), the FDA and USDA issued a request for information on 24 July 2025 on what a uniform federal definition of UPFs should include; the comment period closed on 23 October 2025.^[5] On 10 August 2026 HHS and USDA announced that a proposed definition had been sent to the Office of Management and Budget for review. The agencies describe it as a white paper and research tool rather than a regulation, and as of 13 October 2026 its text had not been published.^[6] Map each SKU against NOVA now, but expect the federal criteria to weigh processing, ingredients and nutrients in their own way.

Practical Reformulation Priorities for Brands

1. Classify each SKU. Record its NOVA group and the marker ingredients that put it there, because assessors do not always agree.^[3]
2. Profile the nutrients. Check energy density, salt, saturated fat, free sugars, fibre and protein.^[1][3]
3. Map each marker to its function. Replace a marker only where the function can be matched or designed out.
4. Re-test after every swap. Confirm stability, texture, sensory quality and shelf life before the next change.
5. Keep claims to the evidence. The review shows association, so avoid causal or disease claims.^[2]
6. Track the definitions. Watch for publication of the US white paper and any UK or EU follow-up.^[6]

Conclusion

The 2024 BMJ umbrella review found consistent associations between higher UPF intake and worse health, with the most credible signals for cardiovascular mortality, type 2 diabetes and common mental disorders.^[2] Most of the evidence is low or very low certainty, it is observational, and NOVA itself has known weaknesses.^[2][3] For brands, the practical response is neither to ignore the findings nor to chase a single additive. Improve the nutrient profile, remove NOVA markers where their function can be replaced, and prove the product still works.

For Brands & Formulators

If you are reviewing a range against NOVA or preparing for a UPF definition, FRL’s new product development service can turn the evidence into a reformulation plan for each product.

• Classify: NOVA group, marker ingredients and nutrient profile for each SKU.
• Plan: marker-by-marker options, ranked by technical risk and label impact.
• Prove: bench trials with sensory, stability and shelf-life testing of the reformulated product.

Share your product and ingredient list, or contact the team to scope the work.

1. Monteiro, C. A., Louzada, M. L. C., Steele-Martinez, E., Cannon, G., Andrade, G. C., Baker, P., Bes-Rastrollo, M., Bonaccio, M., Gearhardt, A. N., Khandpur, N., Kolby, M., Levy, R. B., Machado, P. P., Moubarac, J.-C., Rezende, L. F. M., Rivera, J. A., Scrinis, G., Srour, B., Swinburn, B., & Touvier, M. (2025). Ultra-processed foods and human health: The main thesis and the evidence. The Lancet, 406(10520), 2667–2684. https://doi.org/10.1016/S0140-6736(25)01565-X Cited for the NOVA groups and UPF definition, NOVA marker ingredients, how UPF exposure is measured, nutrient profiles of high-UPF diets and the Series’ policy position.
2. Lane, M. M., Gamage, E., Du, S., Ashtree, D. N., McGuinness, A. J., Gauci, S., Baker, P., Lawrence, M., Rebholz, C. M., Srour, B., Touvier, M., Jacka, F. N., O’Neil, A., Segasby, T., & Marx, W. (2024). Ultra-processed food exposure and adverse health outcomes: Umbrella review of epidemiological meta-analyses. BMJ, 384, Article e077310. https://doi.org/10.1136/bmj-2023-077310 Cited for all umbrella-review figures (scope, effect sizes, evidence classes, GRADE, AMSTAR 2, bias tests), the authors’ limitations and counter-arguments, and the 2019 inpatient trial they cite.
3. Scientific Advisory Committee on Nutrition. (2023, July 11). SACN statement on processed foods and health: Summary report. GOV.UK. https://www.gov.uk/government/publications/sacn-statement-on-processed-foods-and-health/sacn-statement-on-processed-foods-and-health-summary-report Cited for SACN’s assessment of NOVA and of the evidence on processed foods and health.
4. Dicken, S. J., Jassil, F. C., Brown, A., Kalis, M., Stanley, C., Ranson, C., Ruwona, T., Qamar, S., Buck, C., Mallik, R., Hamid, N., Bird, J. M., Brown, A., Norton, B., Gandini Wheeler-Kingshott, C. A. M., Hamer, M., van Tulleken, C., Hall, K. D., Fisher, A., Makaronidis, J., & Batterham, R. L. (2025). Ultraprocessed or minimally processed diets following healthy dietary guidelines on weight and cardiometabolic health: A randomized, crossover trial. Nature Medicine, 31(10), 3297–3308. https://doi.org/10.1038/s41591-025-03842-0 Cited for the UPDATE trial design, weight and LDL results and limitations.
5. U.S. Food and Drug Administration. (2025, September 18). Ultra-processed foods. https://www.fda.gov/food/nutrition-food-labeling-and-critical-foods/ultra-processed-foods Cited for the FDA–USDA request for information (24 July 2025) and its extended comment deadline (23 October 2025).
6. Greenberg Traurig. (2026, October 13). More than research? Why the coming federal definition of ultra-processed foods matters [GT Alert]. https://www.gtlaw.com/en/insights/2026/10/more-than-research-why-the-coming-federal-definition-of-ultra-processed-foods-matters Cited for the August 2026 submission of the proposed US UPF definition (white paper) to OMB and its unpublished status as of 13 October 2026.