The food system’s environment impact has become a pressing concern due to its contributions to greenhouse gas emissions (GHGe), land use and biodiversity loss(Reference Springmann, Clark and Mason-D’Croz1). Intensive agricultural practices, especially monocultures such as maize, wheat and soya, degrade ecosystems and narrow crop diversity. Ultra-processed foods (UPF), composed largely of ingredients produced from these high-yield crops and livestock, have been indicated to have a negative impact on the environment due to their contribution to limited crop diversity and increased vulnerability to environmental pressures(Reference Leite, Khandpur and Andrade2). In addition, many UPF are characterised by hyperpalatability, low satiety potential and heavy marketing that can encourage overconsumption, leading to excessive food production and associated environmental pressures, while also contributing to significant public health challenges(Reference Anastasiou, Baker and Hadjikakou3,Reference Seferidi, Scrinis and Huybrechts4) .
UPF have been linked to negative health outcomes such as obesity, CVD, depressive symptoms and certain cancers(Reference Lane, Gamage and Du5). Consequently, countries like Mexico have incorporated recommendations to limit UPF consumption in dietary guidelines(6). However, the environmental impacts of UPF have received less attention, and the potential implications of substituting UPF with unprocessed or minimally processed foods remain underexplored. With diets shifting towards greater UPF consumption globally, understanding their impact on the environment is critical, particularly in terms of GHGe, land use and preservation of food biodiversity(Reference Popkin and Ng7–9). This convergence suggests that UPF-driven overconsumption represents a shared pathophysiological mechanism underlying both human and environmental health. The same hyperpalatable formulations, low satiety signals and marketing strategies that promote excessive energy intake could simultaneously drive increased food demand and production, amplifying environmental pressures. This dual pathway through overconsumption represents a novel framework for understanding how food processing impacts both human and planetary health through a common mechanism.
This study examined the relationship between dietary intake across food processing levels and environmental outcomes – specifically GHGe, land use and food biodiversity – and evaluates the potential impact of substituting minimally processed foods for UPF among adults in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort.
Methods
The European Prospective Investigation into Cancer and Nutrition cohort
The EPIC cohort is a large multicentre cohort examining the links between metabolic, lifestyleand environmental factors of cancer and chronic diseases. Between 1991 and 2000, over 500 000 participants aged 25–70 years were recruited across twenty-three centres in ten European countries: Denmark, France, Germany, Greece, Italy, the Netherlands, Norway, Spain, Sweden and the United Kingdom. Dietary intake at enrolment was assessed using validated, country-specific questionnaires capturing habitual consumption over the past 12 months(Reference Riboli, Hunt and Slimani10). In order to study associations in a disease-free population, participants with missing dietary data, extreme energy intake-to-requirement ratios, lack of follow-up or prevalent diseases at baseline were excluded. Due to administrative constraints, cohorts from Greece, Norway and Sweden were excluded, resulting in 368 733 participants (see online supplementary material, Supplemental Figure 1).
Dietary assessment
In the 1990s, participants’ usual food intake over the previous 12 months was assessed at baseline with country-specific dietary questionnaires. Depending on the study centre, quantitative dietary questionnaires, semi-quantitative FFQ or a combination of semi-quantitative FFQ and 7-day food records were used. Data on frequencies, portion sizes or intakes in grams per day were stored in a central International Agency for Research on Cancer (IARC) database(Reference Riboli, Hunt and Slimani10). Post-harmonisation of dietary data was conducted, following standardised procedures (e.g. disaggregating recipes into ingredients), to obtain a standardised food list for which the level of detail is comparable between countries. The EPIC food composition database comprises more than 11 000 food and beverage items reflecting the specificities of each country.
Exposure - Nova classification
Standardised EPIC food items were categorised by processing level using the Nova classification: Nova 1 (unprocessed or minimally processed foods, e.g. fruits, vegetables), Nova 2 (processed culinary ingredients, e.g. oils, sugar), Nova 3 (processed foods, e.g. cheese, bread) and Nova 4 (UPF, e.g. soft drinks, flavoured yoghurts). Since the Nova classification system was developed after the EPIC dietary data collection, there was some uncertainty in classifying certain food items according to their level of food processing. Therefore, three classification scenarios were developed to address this uncertainty, a lower, middle and upper-bound scenario. This study used the most probable, middle-bound scenario(Reference Huybrechts, Rauber and Couto11). Dietary contribution from each Nova class was expressed in both grams and kcal per day, as grams reflect absolute consumption, while kcal accounts for energy density, providing complementary insights into environmental impacts.
Outcomes - Environmental impacts and food biodiversity
Environmental outcomes were assessed using the SHARP indicators database, which estimates GHGe and land use from life cycle assessment data encompassing production, packaging, transport and preparation(Reference Mertens, Kuijsten and Dofková12). Food items were matched between the EPIC and SHARP databases using EFSA’s FoodEx2 base-term codes(13). Diet-related GHGe and land use were computed for each individual by summing the amounts for all foods consumed; GHGe was expressed as kg CO2-equivalents per day and land use as m² per day(Reference Mertens, Kuijsten and Dofková12,13) . Food biodiversity was quantified using dietary species richness (DSR), defined as the count of unique biological species consumed across foods, beverages and mixed dishes(Reference Hanley-Cook, Huybrechts and Biessy14). Composite foods were decomposed into ingredients using standard recipes and foods consumed ‘never or less than once per month’ were not considered in the DSR computation(Reference Hanley-Cook, Huybrechts and Biessy14).
Study covariates
Sociodemographic and anthropometric covariates included in the models were: age at recruitment, BMI height, sex, educational level, smoking status at baseline, physical activity using the Cambridge index and alcohol intake.
Statistical analysis
Consumption of the Nova classes (g/d or kcal/d) was modelled as continuous variables. Multivariable mixed linear models with random intercepts for study centres and adjustment for sociodemographic and anthropometric variables were fitted to assess associations between Nova class consumption, GHGe, land use and DSR. Additive models assessed associations of the additional consumption of a Nova class. For this, weight- and energy-based all-component models were constructed, mutually adjusting for each Nova class, to account for the total weight or energy intake(Reference Tomova, Arnold and Gilthorpe15). Additionally, substitution analyses were performed, using the ‘leave-one-out’ method estimated associations of replacing a specific amount of Nova 4 with Nova 1, by keeping total intake constant(Reference Tomova, Gilthorpe and Tennant16). For instance, the substitution of Nova 4 by Nova 1 in GHGe can be parameterised as:
\begin{align*}{\rm{GHGe}} = & \, {\rm{\;}}\alpha + \;{\gamma _1}Nova\;1 + \;{\gamma _2}Nova\;2 + \;{\gamma _3}Nova\;3 \\& + \;{\gamma _4}total\;intake + \;covariates + \left( {1Study\;Centre} \right) + \;\varepsilon \end{align*}
Here, γ 1 represents the relative estimate for replacing a quantity of Nova 4 with an equivalent amount of Nova 1, keeping the total intake constant.
Estimates were expressed as: (1) a 1-sd increment in consumption of a Nova class, or (2) a 10 % increase from the mean absolute total dietary intake. To interpret the results as percentage differences, these estimates were divided by the mean value of the respective outcome measure.
Sensitivity analyses included baseline models only mutually adjusted for each Nova class and main models further adjusted for the Mediterranean diet score (0–18 points)(Reference Couto, Boffetta and Lagiou17). Statistical analyses were performed in RStudio (v4.0.4.1) with two-sided testing, and P values < 0·05 were considered statistically significant.
Results
Sample characteristics
This study included 368 733 participants from the EPIC cohort, of whom 259 268 (70·3 %) were females. The mean (sd) age at recruitment was 51·3 (9·9) years, and the average BMI was 25·4 (4·3) kg/m² at baseline. On average, participants consumed 364 g (278) or 672·9 kcal (412·0) of UPF daily, representing 12·9 % (8·5) of total intake by weight and 30·5 % (15·3) by energy. Mean dietary GHGe and land use were 5·3 (1·82) kg CO2-equivalents per day and 6·9 (2·6) m² per day, respectively. Average DSR was 68·2 (15·2) species per year (Table 1).
Table 1. Baseline characteristics of 368 733 middle-aged adults enrolled in the European Prospective Investigation into Cancer and Nutrition (EPIC) study
DSR, dietary species richness. Nova 1, unprocessed or minimally processed foods. Nova 2, processed culinary ingredients. Nova 3, processed foods. Nova 4, ultra-processed foods.
Associations between Nova class consumptions and greenhouse gas emissions, land use and food biodiversity
Figure 1 illustrates the percentage difference relative to the mean of GHGe, land use and DSR associated with higher consumption of each Nova class. A 1-sd increment in consumption of each Nova class, either in gram or kcal per day, was associated with significantly higher GHGe, land use and DSR, with Nova 4 consumption being more strongly associated with GHGe and land use compared with Nova 1. To illustrate, a 1-sd increment of Nova 4 consumption in kcal per day was related to 15·8 % (95 % CI: 15·8, 16·0) higher GHGe, 16·9 % (16·9, 17·1) higher land use and 1·0 % (0·9, 1·1) higher DSR, while for Nova 1 this was 13·8 % (13·8, 14·0) for GHGe, 12·8 % (12·7, 14·0) for land use. Similar findings were reported for consumption of the different Nova classes in g/d. Strengths of associations differed within Nova 4 subgroups, with animal-based products showing the strongest positive associations with GHGe and land use, while plant-based alternatives and savoury snacks showed the weakest associations (see online supplementary material, Supplemental Table 1).
Figure 1. Linear association between the consumption of each Nova class and dietary greenhouse gas emissions (GHGe), land use and DSR. The left panel shows the additive estimates for 1 sd or 10 % of mean total absolute intake increase in consumption (95 % confidence intervals) across Nova classes, while the right panel presents substitution estimates for 1-sd or 10 % of mean total absolute intake substitution of Nova 4 for Nova 1 among 368 733 adults enrolled in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Nova 1: unprocessed or minimally processed foods, Nova 2: processed culinary ingredients, Nova 3: processed foods and Nova 4: ultra-processed foods. Additive models were mutually adjusted for each Nova class. Substitution models were adjusted for Nova 1, 2, 3 and total intake. Both models were also adjusted for socio-demographic and anthropometrics covariates including: age at recruitment (years), BMI (kg/m2), height (cm), sex (male, female), educational level (none, primary school, secondary school/technical school, higher education, unknown), smoking status at baseline (never, former, current and unknown), physical activity (Cambridge index; inactive, moderately inactive, moderately active, active and unknown) and alcohol intake (non-drinker, > 0–6, > 6–12, > 12–24 and > 24 gram per day), and centre was included as a random intercept. For consumption in kcal per day, the sds are 271·6 for Nova 1, 145·2 for Nova 2, 336·0 for Nova 3 and 394·11 for Nova 4. For consumption in gram per day, the sds are 833·3 for Nova 1, 23·5 for Nova 2, 308·1 for Nova 3 and 264·3 for Nova 4. The 10 % of the mean total absolute intake in kcal per day was 218·8 and 281·9 for total absolute intake in gram per day. Substitution models substituted 1-sd of Nova 4 with an equivalent amount of Nova 1. All P values < 0·001. To facilitate direct comparison, the same y-axis scale was used for both 1sd and 10 % increment estimates. This may reduce visual contrast for smaller effect sizes but improves interpretability across models.
Substitution of ultra-processed with unprocessed or minimally processed foods
10 % of the mean total intake in grams per day substitution of Nova 4 substitution with Nova 1 was associated with 8·9 % (95 % CI: –9·0, –8·9) lower GHGe and 9·3 % (–9·5, –9·2) lower land use (Figure 1). However, such a substitution was related to marginally lower DSR (–0·1 %; –0·2, –0·1). Conversely, a Nova 4 substitution with Nova 1, 10 % of the mean total intake in kcal per day, was associated with higher GHGe (2·6 %; 2·5, 2·6), land use (1·2 %; 1·0, 1·3) and DSR (1·0 %; 1·0, 1·1) (Figure 1).
Sensitivity analysis confirmed our main findings (data not shown).
Discussion
This study found that higher UPF consumption was more strongly associated with increased dietary GHGe and land use compared with unprocessed or minimally processed foods. For DSR, associations were shown to be marginal. Energy-based substitution of UPF with unprocessed or minimally processed foods were associated with higher environmental impacts, whereas weight-based substitutions were associated with lower environmental impacts.
These discrepancies likely stem from the higher energy density of UPF. Energy-based substitutions require larger quantities of unprocessed or minimally processed foods to achieve isocaloric substitutions, potentially increasing environmental impacts(Reference Gupta, Hawk and Aggarwal18). Research suggests that individuals consuming diets high in unprocessed or minimally processed foods tend to have lower energy intake compared with those with UPF-rich diets, meaning isocaloric substitution may not fully capture these differences(Reference Hall, Ayuketah and Brychta19). In contrast, weight-based substitutions, which emphasise food weight rather than caloric equivalence, show environmental benefits that align with UPF’ well-documented tendency to promote overconsumption through their hyper palatability, low satiety, softer textures requiring less chewing, widespread availability and lower cost per calorie, which could lead to excessive energy intake, contributing to rising obesity rates(Reference Anastasiou, Baker and Hadjikakou3,Reference Seferidi, Scrinis and Huybrechts4,Reference Hall, Ayuketah and Brychta19) . Such overconsumption drives higher demand for foods, amplifying environmental impacts further. Additionally, while low-impact plant-based UPF have lower environmental footprints, animal-based UPF remain highly impactful, underlining the importance of considering UPF subgroups(Reference Clark, Domingo and Colgan20,Reference Cordova, Viallon and Fontvieille21) . These findings support the hypothesis that overconsumption serves as a critical link between UPF consumption and environmental harm, paralleling established mechanisms for UPF-associated health risks.
Additionally, negligible DSR differences were observed when substituting UPF with unprocessed or minimal foods, diverging from findings in Brazilian diets where UPF involved fewer species(Reference Leite, Khandpur and Andrade8). This discrepancy may reflect methodological differences: the Brazilian study examined species diversity within UPF products at the food system level, while our analysis assessed how individual dietary patterns relate to overall species consumption.
Our findings suggest that food biodiversity operates independently from processing level in individual diets. Substituting UPF with unprocessed foods may not increase species diversity if individuals simply consume larger quantities of the same limited set of species they already consume. Therefore, reducing UPF consumption alone may be insufficient to improve dietary biodiversity without concurrent efforts to promote species diversification. Alternatively, UPF-driven overconsumption may increase total food intake, maintaining DSR through higher consumption volumes rather than dietary diversification.
Limited observational evidence on UPF’ environmental impacts exists, with most insights coming from life cycle assessments(Reference Jones, Hoey and Blesh22). In a French cohort study, it was found that UPF accounted for 19 % of energy intake in the diet and contributed to 24 % of GHGe, 23 % of land use and 26 % of energy demand. These highlight the significant environmental burden associated with diets rich in UPF, with higher contributions from post-farm stages, in particular processing regarding energy demand(Reference Kesse-Guyot, Allès and Brunin23). A longitudinal study showed reducing UPF consumption lowered GHGe and energy demand, but increased water use(Reference García, Bouzas and Mateos24). Our study is unique in its large, diverse European cohort, allowing a comprehensive assessment of food processing levels and substitution effects.
Several limitations must be acknowledged. The EPIC cohort may differ substantially from current European populations. UPF intake has risen dramatically – from approximately one-third of energy intake in our cohort to over half in recent studies, due to changes in food environments and consumption patterns(Reference Monteiro, Moubarac and Levy25). Although educational attainment has increased across EU member states, this has not corresponded with expected reductions in UPF consumption, suggesting altered socioeconomic determinants of dietary choices(Reference Dicken, Qamar and Batterham26). The shift toward sedentary lifestyles correlates with increased convenience food reliance, while younger populations exhibit greater price sensitivity toward UPF products(Reference Costa, Del-Ponte and Assunção27,Reference Machado, Steele and Levy28) . These transitions suggest our cohort likely underestimates the environmental impacts of contemporary European diets. Misclassification within the Nova system and reliance on SHARP database estimates, which lack country specificity and farming method variations, may introduce error. Additionally, many UPF-specific ingredients (e.g. additives) lack environmental impact assessments, and UPF typically rely on more intensively produced commodity ingredients than non-UPF, differences our analysis cannot fully capture. These errors might underestimate the true associations due to differential measurement error. Variations in dietary assessment methods and the number of items included between centres could also affect DSR, and taxonomic limitations hinder further analysis of food biodiversity. The questionnaires did not distinguish between homemade and industrianlly processed foods, which could overlook ingredient differences leading to varying environmental impacts. Lastly, this study compared individuals rather than actual substitutions, and context-specific factors such as preparation time, cost and food safety may influence dietary shifts and willingness to make substitutions(Reference Gupta, Hawk and Aggarwal18,Reference Wolfson, Martinez-Steele and Tucker29) . For instance, while unprocessed or minimally processed foods are often more nutrient-dense, UPF offer greater accessibility and food safety(Reference Amorim, Silva and do Amaral Sobral30).
In conclusion, UPF consumption was more strongly associated with GHGe and land use as compared with unprocessed or minimally processed food consumption, while associations with food biodiversity were marginal. Substituting UPF with unprocessed or minimally processed foods resulted in differing directions of associations with environmental impacts, depending on whether substitutions were weight or calorie based.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980025101067
Data availability
EPIC data and biospecimens are available to investigators in the context of research projects that are consistent with the legal and ethical standard practices of IARC/WHO and the EPIC Centres. The use of a random sample of anonymised data from the EPIC study can be requested by contacting epic@iarc.fr. For information on the EPIC data access policy and on how to submit an application for gaining access to EPIC data and/or biospecimens, please follow the instructions at iarc.who.int
Financial support
The coordination of EPIC is financially supported by the IARC and the Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London which has additional infrastructure support provided by the NIHR Imperial Biomedical Research Centre (BRC). The national cohorts are supported by: Danish Cancer Society (Denmark); Ligue Contre le Cancer, Institut Gustave Roussy, Mutuelle Générale de l’Education Nationale, Institut National de la Santé et de la Recherche Médicale (INSERM) (France); German Cancer Aid, German Cancer Research Center (DKFZ), German Institute of Human Nutrition Potsdam-Rehbruecke (DIfE), Federal Ministry of Education and Research (BMBF) (Germany); Associazione Italiana per la Ricerca sul Cancro-AIRC-Italy, Compagnia di SanPaolo and National Research Council (Italy); Dutch Ministry of Public Health, Welfare and Sports (VWS), Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF) (The Netherlands); Health Research Fund (FIS) – Instituto de Salud Carlos III (ISCIII), Regional Governments of Andalucía, Asturias, Basque Country, Murcia and Navarra and the Catalan Institute of Oncology – ICO (Spain); Swedish Cancer Society, Swedish Research Council and County Councils of Skåne and Västerbotten (Sweden); Cancer Research UK (14136 to EPIC-Norfolk; C8221/A29017 to EPIC-Oxford), Medical Research Council (1000143 to EPIC-Norfolk; MR/M012190/1 to EPIC-Oxford). (United Kingdom), UiT The Arctic University of Norway. EKC was supported by the Ministry of National Education, Türkiye, through the YLSY International Graduate Education Scholarship Programme.
Funding for grant number IIG_FULL_2020_034 was obtained from Wereld Kanker Onderzoek Fonds (WKOF), as part of the World Cancer Research Fund International grant programme. Funding for grant IIG_FULL_2020_033 was obtained from World Cancer Research Fund (WCRF UK), as part of the World Cancer Research Fund International grant programme. Researchers were independent from the funders. Funders had no role in the collection, analysis and interpretation of data, the writing of the manuscript and the decision to submit the article for publication. PV and CJM are supported by the NIHR GHRC on NCD and Environmental Change (NIHR203247) using UK aid from the UK Government to support global health research. Researchers were independent from the funders. Funders had no role in the collection, analysis and interpretation of data, the writing of the manuscript and the decision to submit the article for publication.
Conflict of interest
The authors have declared that no competing interests exist. IARC disclaimer: Where authors are identified as personnel of IARC/WHO, the authors alone are responsible for the views expressed in this article, and they do not necessarily represent the decisions, policy or views of IARC/WHO. FAO disclaimer: The views expressed in this publication are those of the authors and do not necessarily reflect the views or policies of the FAO of the UN.
Authorship
Conceptualisation: J. Berden, C. Lachat, I. Huybrechts. Data curation: I. Huybrechts, G. Nicolas, P. Ferrari. Formal analysis: J. Berden, I. Huybrechts, G. Nicolas. Funding acquisition: G. T. Hanley-Cook, I. Huybrechts, C. Lachat, M. Touvier. Investigation: J. Berden, G. T. Hanley-Cook, B. Chimera, D. Aune, M. G. M. Pinho, G. Nicolas, B. Srour, C. J. Millett, E. K. Cakmak, E. Kesse-Guyot, E. M. González-Gil, E. P. Vamos, J. B. Lopez, J. Baudry, J. Berlivet, K. Chang, M. Touvier, C. L. Cornet, C. Marques, C. C. Dahm, D. B. Ibsen, F. Jannasch, G. Skeie, M-J. Sanchez, M. B. Schulze, S. Grioni, Y. T. van der Schouw, A. M. J. Zabala, A. Winkvist, A. Tjønneland, C. Sacerdote, C. Kyrø, E. Weiderpas, M. Guevara, P. Frenoy, R. Tumino, S. Panico, V. Katzke, X. Ren, P. Vineis, P. Ferrari, C. Lachat, I. Huybrechts. Methodology: J. Berden, C. Lachat, I. Huybrechts. Project administration: J. Berden, C. Lachat, I. Huybrechts. Software: J. Berden. Resources: J. Berden, G. T. Hanley-Cook, B. Chimera, D. Aune, M. G. M. Pinho, G. Nicolas, B. Srour, C. J. Millett, E. K. Cakmak, E. Kesse-Guyot, E. M. González-Gil, E. P. Vamos, J. B. Lopez, J. Baudry, J. Berlivet, K. Chang, M. Touvier, C. L. Cornet, C. Marques, C. C. Dahm, D. B. Ibsen, F. Jannasch, G. Skeie, M-J. Sanchez, M. B. Schulze, S. Grioni, Y. T. van der Schouw, A. M. J. Zabala, A. Winkvist, A. Tjønneland, C. Sacerdote, C. Kyrø, E. Weiderpas, M. Guevara, P. Frenoy, R. Tumino, S. Panico, V. Katzke, X. Ren, P. Vineis, P. Ferrari, C. Lachat, I. Huybrechts. Supervision: C. Lachat, I. Huybrechts. Visualisation: J. Berden. Validation: G. Nicolas, C. Lachat, I. Huybrechts. Writing original draft: J. Berden. Writing paper draft: J. Berden, G. T. Hanley-Cook, B. Chimera, D. Aune, M. G. M. Pinho, G. Nicolas, B. Srour, C. J. Millett, E. K. Cakmak, E. Kesse-Guyot, E. M. González-Gil, E. P. Vamos, J. B. Lopez, J. Baudry, J. Berlivet, K. Chang, M. Touvier, C. L. Cornet, C. Marques, C. C. Dahm, D. B. Ibsen, F. Jannasch, G. Skeie, M-J. Sanchez, M. B. Schulze, S. Grioni, Y. T. van der Schouw, A. M. J. Zabala, A. Winkvist, A. Tjønneland, C. Sacerdote, C. Kyrø, E. Weiderpas, M. Guevara, P. Frenoy, R. Tumino, S. Panico, V. Katzke, X. Ren, P. Vineis, P. Ferrari, C. Lachat, I. Huybrechts.
Ethics of human subject participation
The present study is relevant to guide local agrifood systems transformation and nutrition research. Local research teams were invited to contribute to the manuscript and were included as co-authors in accordance with ICJME criteria.
This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research study participants were approved by local ethics committees and by the Internal Review Board of the International Agency for Research on Cancer. Written informed consent was obtained from all subjects. For participants who were illiterate or otherwise unable to provide written consent, informed consent was obtained from a legal guardian or an appropriate representative on their behalf.
Open access
