select search filters
briefings
roundups & rapid reactions
factsheets & briefing notes
before the headlines
Fiona fox's blog

expert reaction to study looking at ultraprocessed food consumption and death

Research published in JAMA Internal Medicine demonstrates that an increase in ultraprocessed food consumption appears to be associated with an overall higher mortality risk.

Prof Julian Cooper, Chair of Scientific Committee, Institute of Food Science and Technology (IFST), said:

“The study refers to ‘ultra-processed foods’, a very unclear and confusing terminology, and is deemed to require further research due to significant limitations.

“Foods, such as snacks, desserts and some meats, prepared by food manufacturers, enable consumers to have convenient, safe and shelf stable choices, which are packaged for protection during distribution.  Approved food additives perform a wide variety of functions in foods, for example preservatives, acidity regulators and antioxidants which keep food in a good condition, thus reducing waste, whilst preserving the nutritional quality.  Colours, sweeteners, emulsifiers and stabilisers contribute to the product quality and improve the eating experience.”

Catherine Collins RD FBDA, Registered dietitian, NHS dietitian, said:

“This is yet another paper from the research group running the French online health survey NutriNet-Santé1, inferring an association between a higher percentage calorie intake from ‘ultra-processed’ foods in the overall diet and higher all-cause mortality.

“Theoretically this should be of significant interest to us in the UK, given half our daily calorie intake is deemed to come from ‘ultra-processed’ foods using the same definitions as used in this study2.

“In this subgroup of 44,500 NutriNet-Santé participants, 602 (1.4%) participants died over a 7 year period.  The authors infer that a 10% increase in the proportion of ‘ultra-processed’ food in the diet was associated with a 14% higher risk of ‘all cause’ mortality.  Just under half the deaths were attributed to cancer (n=219) or cardiovascular disease (n=34), both conditions known to be influenced by diet and lifestyle.  Over half the causes of death weren’t defined and so their relationship to diet could prove irrelevant (e.g. death from major trauma).

“The highest intake of arbitrarily defined ‘ultra-processed’ foods3 was reported by those on the lowest income, living alone, male, a current smoker, obese, or reporting low levels of physical activity.  All these factors are independent predictors of mortality potentially unrelated to dietary intake.

“A higher ‘ultra-processed’ food intake was associated with a proportionally increased total and saturated fat, sugars, salt and energy intake, and a lower protein, vitamin and mineral intake.  However, with no absolute nutrient values provided in the main text, no assumptions regarding dietary inadequacy or nutritional deficiencies can be inferred.  However, when pre-existing health issues prior to recruitment, or death within the first year of the study were both excluded, the association between ‘ultra-processed’ foods and mortality no longer proved statistically significant.

“Professionally, I’ve three concerns about the nutritional aspects of NutriNet-Santé.  First, NutriNet-Santé is an on-going, self-recruiting study with some 270,000 participants currently registered.  Initial public interest promoted registration as an ‘exemplary citizen act’.  This selection bias has potential problems when it comes to health and diet assessment and the accuracy of documentation.

“In relation to diet, ‘Nutrinauts’ (as participants are called) complete 5 online diet and health questionnaires on registration and repeat the process yearly.  The diet survey requires 3 ‘self-selected’ days to be recorded online.  Portion sizes are estimated using online food photos from which researchers calculate nutritional intake and assign a degree of ‘ultra-processing’.  This is prone to description error.

“It’s incredibly difficult to confirm that diet diaries of this 44,500 subgroup taken at yearly intervals were fully representative of usual dietary intake (given participants could choose when they wanted to record information), nor be a valid representation of their usual diet.

“Secondly, NOVA categorises foods according to the extent and purpose of food processing, rather than in terms of the nutrients found in those foods – according to the group who devised this classification3.  The group define a food as ‘ultra-processed’ if it contains five or more ingredients, and/or EU permitted food additives, or contains food extracts such as whey protein.  However, we can’t infer a food is nutritionally inferior simply on the basis of recipe formulation or food processing techniques.*  For example, bread or biscuits baked at home would not be considered ultra-processed, whereby shop bought versions would, despite identical ingredients. This classification is not grounded in nutritional science.

“Third, the authors mention things like acrylamide, plasticisers, additives, artificial sweeteners and/or potential endocrine disruptors from packaging materials in terms of processing – however the International Agency for Research on Cancer (IARC) and European Food Safety Agency (EFSA) finding ‘inadequate evidence in humans for the carcinogenicity of acrylamide’ and that artificial sweeteners have been approved by EFSA as safe.

“Finally, it’s not helpful to group ‘ultra-processed’ foods together.  Even the general public understand that a 100g bar of milk chocolate provides less protein, micronutrients and fibre than a nutritionally-balanced microwave dinner, despite both providing around 530kcals.  It’s the role of Dietitians and Registered Nutritionists to promote healthful diets which are based on the food matrix, not the degree of processing.  However, this study confirms global findings from other studies that a higher intake of ‘ultra-processed’ foods increases overall energy intake and risk of obesity.”

1 Hercberg S et al. The Nutrinet-Santé Study: a web-based prospective study on the relationship between nutrition and health and determinants of dietary patterns and nutritional status. BMC Public Health. 2010 May ;10:242. PubMed PMID: 20459807

2 Monteiro CA et al. Household availability of ultra-processed foods and obesity in nineteen European countries. Public Health Nutr. 2018; 21:18-26. PMID: 28714422

3 NOVA classification. https://archive.wphna.org/wp-content/uploads/2016/01/WN-2016-7-1-3-28-38-Monteiro-Cannon-Levy-et-al-NOVA.pdf

*Edited at authors request from “However, this does not necessarily infer a food because of these criteria for their classification model.” to “However, we can’t infer a food is nutritionally inferior simply on the basis of recipe formulation or food processing techniques” – 16:57 11/02/2019

Prof Nita Forouhi, MRC Epidemiology Unit, University of Cambridge, said:

“The case against highly processed foods is mounting up, with this study adding importantly to a growing body of evidence on the health harms of ultra-processed foods.  Though this research is observational, which has well-known limitations, the authors have made demonstrable attempts to reduce these.  But some important limitations remain.

“Is this research definitive? – no, more evidence is still needed to confirm these findings, we need greater clarity on type of food processing and extent of any misclassification of foods, and mechanisms of the link need further study, yet we would ignore these findings at public health’s peril.

“A vital take-away message is that consumption of highly processed foods reflects social inequalities – they are consumed disproportionately more by individuals with lower incomes or education levels, or those living alone.  Such foods are attractive because they tend to be cheaper, are highly palatable due to high sugar, salt and saturated fat content, are widely available, highly marketed, ready-to-eat, and their use-by-dates are lengthy, so they last longer.  More needs to be done to address these inequalities.

“Public health action is crucial. Solutions rest within the three Ps – people, places and products. People need more information and health education; places that sell foods need to provide easy access to and promote healthier options; and products need to be available that are not ultra-processed, and are affordable.  In the UK, Public Health England has already challenged the food industry to cut 20% of calories from processed foods by 2024, using strategies to reformulate food products and/or reduce portion size and promote healthier options.

“The authors appropriately discuss many of the relevant limitations and strengths of their research.  As mentioned, this research is observational so cannot prove cause and effect.  But enthusiasm for randomised controlled trials on this will be low among both funders and participants, given there is already other research pointing to the likely health harms of ultra-processed foods.  We may argue whether it would even be ethical to put people into the ultra-processed food group.  And even if that was possible, classic double blind placebo controlled trials will not be possible, for obvious reasons.  So we need to be smart, and use this type of observational evidence together with further improved methods, and with replication of findings in different populations.

“Other limitations should be further considered, and improved upon in future research.  The numbers of deaths over an average of 7 years of follow-up in the study are not large enough to allow analysis by cause of death.  The overall effect size was small, and not significant when deaths in the first two years of follow up were excluded, or people with existing cancer or cardiovascular disease were excluded.  Their decision to present results by ultra-processed foods as a proportion of the total weight of food is limiting, and it would have been helpful to see the results in the main article for these as a proportion of total energy intake too (supplementary files were not available at time of this comment).  As three quarters of the study cohort was made up of women, and it is known from other research that generally women tend to have healthier diets than men, therefore the results might have been different if more men were included.”

Dr Ian Johnson, Nutrition researcher and Emeritus Fellow, Quadram Institute Bioscience, said:

“This is a large, carefully conducted prospective study of healthy middle-aged and older French people, in which a statistically significant association between death from any cause and a relatively higher consumption of ‘ultraprocessed foods’ has been identified.  To put things into perspective, although the risk of dying over the seven-year period of investigation was about 15% higher among those consuming more of these foods, the background risk across the whole group was low.  Happily, 98.6% of the participants were still alive at the end of the study.

“It is very difficult to draw any firm conclusions from this investigation.  The consumption of ‘ultraprocessed foods’, which includes an enormous range of products, from soft drinks to snacks and ready-made meals, was linked, often adversely, to many measures of general nutritional quality, as well as to other lifestyle factors such as smoking, education and physical activity.  The authors have made great efforts to correct for such confounders, but many others probably remain unknown, and as they themselves state, “no causality can be established for the observed associations”.  The authors do feel justified in discussing the possible adverse effects of food additives, but as they have no data on the actual levels of consumption of these in their study group, this is pure speculation.

“I do agree however with the authors’ overall conclusion that further studies are needed to disentangle the various mechanisms by which ultraprocessed foods may affect health.”

Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“This is a large, generally competent piece of research.  If its findings applied in the UK as well as in France where it was carried out, its findings might possibly be considerably more important here than there.  A 2017 study1 reported that the UK households bought the highest proportion of ultraprocessed foods out of the 19 European countries studied, with these foods providing just over half of the calories bought.  France was close to being the lowest on this comparison, with only about 14% of calories coming from these foods.  But in fact it’s not necessarily true that its results can be applied to the UK, as I’ll explain.  Furthermore, the nature of this new study means that it really can’t take us much further forward in our understanding of how food affects health.  It doesn’t really tell us what causes what, it gets us no further forward on exactly what could be harmful in ultraprocessed foods, and the size of the association that it found was pretty small anyway.  All that the researchers propose as actions to be taken in the light of their findings is that more precise research is carried out to try to disentangle some of this.  That might well be helpful, but I’d agree with them that their results don’t warrant any new changes in policy or advice to the general public.

“The study can’t establish that eating more ultraprocessed food actually causes an increase in the short-term risk of death, because it’s an observational study.  That is, people weren’t allocated randomly to eat particular diets.  They just ate what they were going to eat anyway, and the researchers recorded that, and recorded whether and when they died.  The research report points out that the amount of ultraprocessed food they ate varied according to age, gender, marital status, income, where they lived, whether they smoked, and a lot of other things.  Perhaps any differences in death rates are actually caused by one or more of these things, or by some other differences that were not investigated, and not by their consumption of processed food at all.  The researchers did carry out statistical adjustments to take some account of these other differences between people as far as they could, but that can only be done for differences on which they have data.  The pattern of what causes what could be very complicated.  The researchers point out that they can’t even rule out reverse causation – that is, some people get ill with a chronic disease, which increases their risk of death, and as a consequence they change their diet, so that in a sense the increased risk of death causes dietary differences, rather than the other way round.

“Even if it is the case that the increase in death rates is caused by differences in consumption of ultraprocessed food, this study can’t tell us what it is about the ultraprocessed food that might increase the risk.  That’s because the list of ultraprocessed foods is long and varied, and different people eat different combinations of these foods.  It’s perfectly possible that the observed increase in risk in people who get more of their nutrition from ultraprocessed foods is because of differences in the exact type of ultraprocessed foods they eat, rather than the overall amount.  It’s possible that many foods on the list aren’t related to death risk at all, or even that some of them lower the short-term risk of death.  It’s also possible that the observed increases in death rates were due to some aspect of the foods that aren’t classified as ultraprocessed.  Maybe those who ate more ultraprocessed foods also ate different types of other foods, and it’s that difference in other foods that is actually related to mortality.  We just can’t tell from this study.

“It’s important to understand that the differences in death rates in this study were not very large.  On average, participants were followed up for an average of about 7 years, which isn’t very long in terms of picking up deaths from the sort of chronic diseases most likely to be affected by diet.  Out of every 1000 participants (all aged at least 45), about 14 died during the follow-up period.  Suppose, for the sake of argument, that it is indeed ultraprocessed food consumption that causes the increased death rates.  Now imagine that every one of the participants had increased the proportion of ultraprocessed foods in their diet by 10 percentage points, so for instance, of their proportion of ultraporcessed foods was 15%, they increased it to 25%.  That’s a very substantial increase and (if it applied across the country) would move France a long way up the European league table of ultraprocessed food consumption.  Then there would be about 2 extra deaths during the follow-up period out of 1000 participants – not very many. *If the effects are causal and the proportional size of the effect on death rates doesn’t change during adult life – and neither of these assumptions is actually confirmed by this study – the decrease in life expectancy from this change in consumption of ultraprocessed foods would be about a year and a half.*

“Despite these issues, does this new study tell us anything about associations between ultraprocessed food consumption in other countries than France, in the UK for instance?  Not necessarily, I’d say.  People’s diets in the UK are, on average, very different from those in France.  In the UK we (on average) eat far more in the way of foods classified as ultraprocessed, but we also probably eat different types of ultraprocessed foods, and indeed different types of foods that aren’t ultraprocessed.  Since this study doesn’t given any data on what aspects of the foods might cause problems, it could be that its findings don’t carry over to the UK at all – we just can’t tell.  Indeed, as the researchers point out, the participants in the study weren’t even very typical of the general French population – they were all volunteers in a long-term nutrition study, are more health-conscious than the general population, and so may well eat differently and have differences in other health-related behaviours.  So perhaps the results aren’t even typical of the overall position in France, let alone elsewhere.”

1 Monteiro, CA et al. (2017) ‘Household availability of ultra-processed foods and obesity in nineteen European countries’, Public Health Nutrition: 21(1), 18–26, doi:10.1017/S1368980017001379

*Quote amended at authors request. 09:23 12/02/2019

‘Association between ultraprocessed food consumption and risk of mortality among middle-aged adults in France’ by Laure Schnabel et al. was published in JAMA Internal Medicine at 16:00 UK time on Monday 11 February 2019.

Declared interests

Catherine Collins: “Recent chair (paid) of a DairyUK meeting on dairy foods, health and sustainability.”

Dr Ian Johnson: “Ian Johnson participated in the SACN Working Group on carbohydrates and health as an external expert from 2009 to 2015, and has served in the past as an advisor to the food industry.”

Prof Nita Forouhi: “None.”

Prof Kevin McConway: “Kevin McConway is a Trustee of the SMC.”

in this section

filter RoundUps by year

search by tag