Hostname: page-component-7c8c6479df-7qhmt Total loading time: 0 Render date: 2024-03-28T20:42:40.425Z Has data issue: false hasContentIssue false

Ultra-processed foods in public health nutrition: the unanswered questions

Published online by Cambridge University Press:  14 December 2022

Michael J. Gibney*
Affiliation:
Institute of Food and Health, University College Dublin, Belfield, Dublin D04 V1W8, Ireland
Rights & Permissions [Opens in a new window]

Abstract

There is a growing interest in the study of the degree of food processing and both health and nutritional outcomes. To that end, several definitions of the degree of processing have been proposed. However, when each of these is used on a common database of nutritional, clinical and anthropometric variables, the observed effect of high intakes of highly processed food, varies considerably.. Moreover, assigning a given food  by nutritional experts, to its appropriate level of processing, has been shown to be variable. Thus, the subjective definitions of the degree of food processing and the coding of foods according to these classifications is prone to error  is  prone to error. Another issue that need resolution is the relative importance of the degree of food processing and the formulation of a processed food. Although correlational studies linking processed food and obesity abound, there is a need for more investigative studies.

Type
Invited Commentary
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2022. Published by Cambridge University Press on behalf of The Nutrition Society

Ultra-processed foods and public health nutrition: the unanswered questions

Since the dawn of nutritional science, foods with a common origin have been categorised into food groups such as milk and milk products, spreadable fats, meat and meat products or vegetables. These categories are usually used to report patterns of food consumption, given that typical food composition tables might contain several thousands of individual foods. There is no record of any ambiguity as to the meaning of such food groups and, using raw data from dietary surveys, the foods within any category can be changed at will, to suit the research question in mind. In recent years, a new categorisation of foods, based on their degree of processing, has been proposed and is a rapidly increasing source of published literature. The present paper seeks to examine the strength of evidence that the degree of food processing is useful in the science of public health nutrition.

Defining highly processed foods

The most widely used definition of highly processed foods comes from the University of Sao Paulo where the NOVA classification of foods has been developed(Reference Monteiro, Cannon and Lawrence1). Within that system, four levels of processing are defined: minimally processed (MP) foods, processed culinary ingredients, processed foods and ultra-processed foods (UPF). The latter are defined as foods which contain ‘substances never or rarely used in kitchens, or classes of additives whose function is to make the final product palatable or more appealing’. The University of North Carolina (UNC) classification system builds on the NOVA definition of UPF and the European Prospective Investigation in Cancer (EPIC) also developed a definition of highly processed foods(Reference Poti, Mendez and Ng2,Reference Eicher-Miller, Fulgoni and Keast3) . Finally, the International Food Information Council (IFIC) developed a categorisation of foods based on the degree of processing and two of these categories (ready-to-eat processed foods and prepared meals or foods) are combined to provide a definition of highly processed foods(Reference Slimani, Deharveng and Southgate4). The extent to which these four systems of food processing classification agree on the impact of UPF consumption on biochemical, clinical and anthropometric outcomes was examined using a large Spanish database (PREDIMED-Plus Cohort). Food intake data were recorded for the degree of food processing according to the four approaches (NOVA, UNC, EPIC and IFIC)(Reference Martinez-Perez, San-Cristobel and Guallar-Castillon5). The results showed that, using a fully adjusted linear model, the interpretation of the relationship between the level of consumption of UPF on the parameters studied varied across the four definitions. The NOVA classification found a positive association between UPF intake and BMI (kg/m2) whereas no such association was observed with any other definition of UPF. In contrast, the UNC classification showed an effect of UPF intake on both systolic and diastolic blood pressure but none of the other three definition found this effect. Whereas none of the four classification systems found any link between the level of UPF consumption and LDL-cholesterol levels, three (IFIC, UNC, EPIC) found a positive association between UPF intake and HDL-cholesterol while NOVA alone did not show this association. These contrasting findings, as to the link between UPF intake and measures of health, are due to the subjective approach to defining UPF. There is no scientific basis for choosing any one of these four definitions over the others; however, popular any individual definition might be with authors of papers on highly processed foods and health. Even within a given food processing classification system, there exists subjectivity in assigning individual foods to particular degrees of processing. A French study examined the ability of food and nutrition experts to correctly assign generic or marketed foods to one of the four levels of processing in the NOVA system(Reference Braesco, Souchon and Sauvant6). Irrespective of whether or not the full ingredient data were provided, the authors found a high level of discordance across the evaluators. Clearly, this variability between classification systems coupled with the poor inter-individual assignment of foods to specific processing categories highlights a major problem with the use of food processing classification systems in public health nutrition.

Homogeneity of ultra-processed food

The NOVA classification of UPF is based on twelve or so food categories (it varies over time) and almost all studies use this broad classification to study the link between intake of UPF and health outcomes(Reference Gibney7). Foods defined as UPF generally account for 60 % of total energy intake and, with such a wide coverage of the food chain, it is not surprising that dietary sub-groups can be identified among UPF consumers. The lifestyle prospective study examined the relationship between UPF and the development of type 2 diabetes in over 70 000 adults followed for 41 months(Reference Duan, Vinke and Navis8). Four patterns of UPF consumers were identified using principal component analysis: two involved snacks (one hot, one cold), one represented the traditional Dutch diet and one was high in sweets and pastries. The two snack clusters showed a positive relationship with the onset of diabetes while the sweet and pastry cluster showed a negative relationship. The traditional Dutch diet cluster showed no relationship. Another study found that whereas the totality of UPF categories was associated with all-cause mortality in renal transplant patients, only two of the twelve categories of food within the NOVA definition of UPF (sugar-sweetened beverages and processed meats) showed a significant association with all-cause mortality(Reference Osté, Duan and Gomes-Neto9). Both of these studies indicate that the gross classification of twelve food categories into one large category of UPF may yield results which do not correctly drive options within public health nutrition.

Nutrients or additives

Three studies examined the relationship between the intake of NOVA-defined UPF and chronic disease and have shown that the effect of UPF on chronic disease remained even when the nutritional quality of the diets of individuals was included in the logistic regression analysis(Reference Du, Kim and Rebholz10Reference Solans, Fernández-Barrés and Romaguera12). However, one other study failed to find such an effect(Reference Griffin, Albaloul and Kopytek13). This raises the question of whether the true causative agent in UPF is not the nutrient profile but rather, the food additive content. The approval of a food additive for use in the human food chain is subject to very extensive toxicological evaluation in cell lines and with animal models. Carcinogenic and mutagenic properties are included within the toxicological profiles used for approval of a food additive. Subsequent post-approval data from human epidemiological studies may indicate a possible association between an additive and some chronic disease. In that case, the approval for the use of the food additive for human consumption will be re-assessed and either withdrawn, approved for use with altered conditions or the original approval and conditions of use upheld. It is therefore difficult to envisage how a putative significant association between the intake of a food additive and a given chronic disease could be missed in the lengthy and extensive toxicological evaluation of the additive in question. Assessing the occurrence and usage levels of food additive is extremely challenging, putting exposure estimates beyond the abilities of most research groups(Reference Gibney and Lambe14). If sufficient data are available from local or national public analysts’ laboratories on the additive content of commercially produced foods, it is possible to compute exposure data(Reference Chazelas, Druesne-Pecollo and Esseddik15). However, given the decades-long consumer concern about food additives, a significant number of food additives currently in use are naturally occurring and estimates of exposure to a given food additive will be confounded by intakes of the chemical in question from natural sources(Reference Kim, Lee and Kwon16). In effect, the only reliable approach is to conduct total diet studies, specifically designed to target food additives. Total diet studies draw on data from food consumption studies but involve the subsequent purchasing and cooking of foods or food groups prior to analysis for a specific food additive content(Reference Gibney and Forde17). Most large food retailers list the full ingredients of processed foods and it would be valuable as a starting point in studying food additives as causative agents of UPF, to document the occurrence of different additives among high and low consumers of UPF.

Ultra-processed food and energy balance

Whereas most studies of NOVA-defined UPF intake and obesity have shown positive associations, several key studies have failed to confirm such a link(Reference Gibney7). This may be due to what has been previously mentioned, the high level of discordance in coding foods according to the UPF classification system. One randomly controlled study in a metabolic ward setting found that increased consumption of UPF was associated with a rise in body weight while the control arm, fed a MP diet, showed no such trend(Reference Hall, Ayuketah and Brychta18). Because this trial involved ad libitum intakes of food, the range of foods offered to either arm of the study at each meal exceeded the predicted energy intake and the two arms of foods offered had identical mean energy densities. However, from the foods offered, those chosen by the subjects in the UPF arm had a higher energy density than the foods chosen on the MP diet. Energy density is a known driver of energy intake and thus should be controlled for in studies of UPF intake and obesity. Sadly, it has never been retained as a variable in any of the studies of UPF intake and obesity. A Dutch study has shown that whereas the mean energy density of UPF is higher that less processed foods, the within-category variability is also very high(Reference Forde, Mars and de Graaf19). Thus, MP foods can have a high energy density (avocados, peanuts, butter) while foods designated as highly processed can have a low energy density (breads, breakfast cereals, flavoured low-fat yogurts).

In the Randomised Controlled Trial that found an association between the intake of UPF and obesity, a higher eating rate was also found on the UPF arm of the trial. A high eating rate is strongly associated with weight gain(Reference Ohkuma, Hirakawa and Nakamura20). To differentiate between the effects of degree of processing and eating rate, a recent study compared the eating rates of hard and soft lunches representative of both minimally and UPF(Reference Teo, Lim and Goh21). The hard, MP and hard-ultra-processed meals were consumed at an equally slower rate than their soft counterparts (UPF and MP), reducing weight of food intake by 21 % and energy intake by 26 %, irrespective of the degree of processing. Texture alone significantly influenced eating rate. Thus, more studies on the physical nature of processed foods are needed to fully understand any putative effect of UPF on food choice

Conclusions

In any scientific discipline, definitions must be objectively derived. The difficulty with the present approach is that all definitions are subjective, simply reflecting the personal opinions of those deriving the definition. The NOVA definition of UPF refers to…. ‘additives whose function is to make the final product palatable or more appealing’. Palatability is not just a function of a given food but is primarily determined by genetic, phenotypic and environmental factors(Reference Johnson and Wardle22). Palatability is therefore a subjective term as in ‘chacun à son gout’ (each to their own taste). The subjective nature of the definition continues with reference to the ability of food additives to make a food ‘more appealing’. If a preservative is used in a bread, does that make it ‘more appealing’ to a consumer than an identical bread with no additives present? If the degree and nature of processing of foods are to be considered as an important driver of public health nutrition, then some level of objectivity in the definition of highly processed foods is needed. Moreover, the selection of candidate foods for consideration as highly processed must first be examined to understand their population impact on nutrient intakes. Breads and breakfast cereals are classed as NOVA UPF foods and yet they make a considerable positive impact on population nutrient intake. Further gains in nutrient intake with bread and breakfast cereals can be achieved by promoting whole grain varieties and through reformulation. However, the NOVA recommendation that all UPF foods be avoided, including industrially prepared breads and breakfast cereals, does not make sense for public health nutrition policy. Moreover, NOVA opposes the concept of reformulation of foods on the grounds that one cannot make an unhealthy food (subjectively defined) healthy(Reference Monteiro and Cannon23,Reference Scrinis and Monteiro24) . This ignores extensive efforts by governments to encourage reformulation of foods to lower salt, added sugars and fats. By the same token, foods that are considered treats, such as chocolate, and which make a modest contribution to energy intake, might be excluded from consideration as UPF. Presently chocolate is deemed to be UPF and thus to be avoided. If the degree and nature of food processing are to be considered within the strategies of public health nutrition policies, a robust, objective, evidence-based definition must be devised and the criteria for considering a food as highly processed must first take account of that food’s impact on population nutrient intake.

Declaration of Interest

The author receives an honorarium as chair of the International Breakfast Research Initiative sponsored by Cereal Partners Worldwide and as chair of a European research consortium on confectionery portion sizes sponsored by CAOBISCO. The author has acted as a paid and non-paid consultant to leading companies with an interest in food.

Footnotes

This article has been co-published with permission in British Journal of Nutrition and in Public Health Nutrition.

References

Monteiro, CA, Cannon, G, Lawrence, M, et al. (2019) Ultra-Processed Foods, Diet Quality, and Health using the NOVA Classification System. Rome: FAO.Google Scholar
Poti, JM, Mendez, MA, Ng, SW, et al. (2015) Is the degree of food processing and convenience linked with the nutritional quality of foods purchased by US households? Am J Clin Nutr 101, 12511262.10.3945/ajcn.114.100925CrossRefGoogle ScholarPubMed
Eicher-Miller, HA, Fulgoni, VL III, Keast, DR, et al. (2012) Contributions of processed foods to dietary intake in the US from 2003–2008: a report of the Food and Nutrition Science Solutions Joint Task Force of the Academy of Nutrition and Dietetics, American Society for Nutrition, Institute of Food Technologists, and International Food Information Council. J Nutr 142, 2065S2072S.10.3945/jn.112.164442CrossRefGoogle Scholar
Slimani, N, Deharveng, G, Southgate, DAT, et al. (2009) Contribution of highly industrially processed foods to the nutrient intakes and patterns of middle-aged populations in the European Prospective Investigation into Cancer and Nutrition study. Eur J Clin Nutr 63, S206S225.CrossRefGoogle Scholar
Martinez-Perez, C, San-Cristobel, R, Guallar-Castillon, P, et al. (2021) Use of different food classification systems to assess the association between ultra-processed food consumption and cardiometabolic health in an elderly population with metabolic syndrome (PREDIMED-Plus Cohort). Nutrients 13, 24712489.CrossRefGoogle Scholar
Braesco, V, Souchon, I, Sauvant, P, et al. (2022) Ultra-processed foods: how functional is the NOVA system? Eur J Clin Nutr (Epublication ahead of print version 21 March 2022).10.1038/s41430-022-01099-1CrossRefGoogle ScholarPubMed
Gibney, MJ (2018) Ultra-processed foods: definitions and policy issues. Curr Dev Nutr 3, nzy077.CrossRefGoogle ScholarPubMed
Duan, M-J, Vinke, PC, Navis, G, et al. (2022) Ultra-processed food and incident type 2 diabetes: studying the underlying consumption patterns to unravel the health effects of this heterogeneous food category in the prospective Lifelines cohort. Published Online: BMC Med 20, 7.Google ScholarPubMed
Osté, MCJ, Duan, M-J, Gomes-Neto, AW, et al. (2022) Ultra-processed foods and risk of all-cause mortality in renal transplant recipients Am J Clin Nutr 115, 16461657 10.1093/ajcn/nqac053CrossRefGoogle ScholarPubMed
Du, S, Kim, H, Rebholz, CM, et al. (2021) Higher ultra-processed food consumption is associated with increased risk of incident coronary artery disease in the atherosclerosis risk in communities study. J Nutr 151(12), 37463754.10.1093/jn/nxab285CrossRefGoogle ScholarPubMed
Fiolet, T, Srour, B, Sellem, L, et al. (2021) Consumption of ultra-processed foods and cancer risk: results from NutriNet-Santé prospective cohort (2018). Br Med J 360, k322.Google Scholar
Solans, M, Fernández-Barrés, S, Romaguera, D, et al. (2021) Consumption of ultra-processed food and drinks and chronic lymphocytic Leukemia in the MCC-Spain study. Int J Environ Res Public Health 18, 112.10.3390/ijerph18105457CrossRefGoogle ScholarPubMed
Griffin, J, Albaloul, A, Kopytek, A, et al. (2021) Effect of ultra-processed food intake on cardiometabolic risk is mediated by diet quality: a cross-sectional study. BMJ Nutr Prev Health 4, 174180.CrossRefGoogle Scholar
Gibney, MJ & Lambe, J (1996) Estimation of food additive intake: methodology overview. Food Addit Contam 13, 405410.10.1080/02652039609374424CrossRefGoogle ScholarPubMed
Chazelas, E, Druesne-Pecollo, N, Esseddik, Y, et al. (2021) Exposure to food additive mixtures in 106,000 French adults from the NutriNet-Santé cohort. Sci Rep 11, 19680.10.1038/s41598-021-98496-6CrossRefGoogle ScholarPubMed
Kim, CI, Lee, J, Kwon, S, et al. (2015) Total diet study: for a closer-to-real estimate of dietary exposure to chemical substances. Toxicol Res 31, 227240.10.5487/TR.2015.31.3.227CrossRefGoogle ScholarPubMed
Gibney, MJ & Forde, CG (2022) Nutrition research challenges for processed food and health. Nat Food 3, 104109.10.1038/s43016-021-00457-9CrossRefGoogle ScholarPubMed
Hall, KD, Ayuketah, A, Brychta, R, et al. (2020) Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake. Cell Metab 32, 6777.CrossRefGoogle ScholarPubMed
Forde, CG, Mars, M & de Graaf, K (2020) Ultra-processing or oral processing? A role for energy density and eating rate in moderating energy intake from processed foods. Curr Dev Nutr 4, nzaa019.CrossRefGoogle ScholarPubMed
Ohkuma, T, Hirakawa, Y, Nakamura, U, et al. (2015) Association between eating rate and obesity: a systematic review and meta-analysis. Int J Obes 39, 15891596.10.1038/ijo.2015.96CrossRefGoogle ScholarPubMed
Teo, PS, Lim, AJ, Goh, AT, et al. (2022) Texture-based differences in eating rate influence energy intake for minimally-processed and ultra-processed meals. Am J Clin Nutr 116, 244254.10.1093/ajcn/nqac068CrossRefGoogle ScholarPubMed
Johnson, F & Wardle, J (2014) Variety, palatability, and obesity. Adv Nutr 5, 851859.10.3945/an.114.007120CrossRefGoogle ScholarPubMed
Monteiro, CA & Cannon, G (2012) Product reformulation will not improve public health. World Nutr 3, 406434.Google Scholar
Scrinis, G & Monteiro, CA (2018) Ultra-processed foods and the limits of product reformulation. Public Health Nutr 21, 247252.10.1017/S1368980017001392CrossRefGoogle ScholarPubMed