select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
A study published in the Lancet Regional Health – Europe looks at ultra-processed foods and the risk of cancer and cardiometabolic diseases.
Dr Duane Mellor, Registered Dietitian and Senior Lecturer, Aston Medical School, Aston University, said:
“This is a large study which collected data between 1992 and 2000 from over half a million people in ten different European counties including the UK, although not all this data could be analysed as in countries like France, Norway and Greece they had not recorded heart disease in the same way. This paper took data from a number of different ways of measuring food intake over the last 12 months using a food questionnaire at the start of the study. This data was then used to estimate ultra processed food intake, these people were then followed up 11.2 years later to see if people had developed heart disease, cancer and type 2 diabetes where individuals have more than one of these conditions together – something called multi-morbidity.
“This study had to try and use the NOVA classification which describes food in four categories, with ultra processed foods being ones which typically cannot be made in the home and use processed ingredients. To estimate it researchers had to break down foods into different ingredients to try and work out if it is ultra-processed or not. This approach, especially as the food data is up to 30 years old, could make this type of interpretation of historical data using a modern definition open to error. Additionally not all the food questionnaires were designed to fully quantify food intake, which potentially could increase risk of error. This is before considering the risk of bias, relying on individuals having to remember what they ate over the last year, and then assuming this is constant during the 11.2 year follow up.
“To try and reduce the risk of other factors influencing the risk of these diseases, researchers tried to control for factors like age and education, which are known to increase risk of poor health. But they did not consider other key factors – this may be an issue especially with factors like income or level of deprivation, as it is known that they impact health massively, particularly the risk of developing multi-morbidity which was the focus of this study.
“It is also noted in this study that when groups of ultra-processed foods were investigated only sweetened drinks (which in this study included both sugar sweetened and artificially sweetened versions) and processed meats were significantly associated with increased risk of multi-morbidity. This perhaps suggests that the concept of ultra-processed foods is too broad, and this study suggested that there was no significant association of ultra-processed cereals (which include most types of bread) and risk of multi-morbidity. This also aligns with current government recommendations to reduce intake of sugar and sweetened drinks along with processed meats as part of a healthy diet based on vegetables, fruit, nuts, seeds, wholegrain foods and legumes with moderate amounts of dairy products and non-processed meats. So, although this study might suggest the problem is ultra-processed foods, it could actually be that foods already widely accepted as increasing risk of multimorbidity cancer, heart disease and diabetes could be the ones responsible for the associations seen in this study.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, Open University, said:
“This is an interesting study, and in my view a good one in statistical terms. Like all studies of this type, its interpretation is made complicated because it’s an observational study. But, unlike most (but not all) studies of associations between consumption of ultra-processed foods (UPFs) and health outcomes, this one looked separately at subgroups of ultra-processed foods. That’s useful, because the overall category of ultra-processed foods is pretty wide.
“This study’s conclusions about subgroups still have limitations, for example because they also come from observational data. But I’d say that they throw a bit of light on the whole notion of treating all ultra-processed foods as a single grouping. Comments in the press release emphasise this.
“I think it does make sense to look, as this study does, at combinations of illnesses (that is, multimorbidity). Dietary studies of many kinds, over many years, not just on UPFs, have indicated links between diet and many different non-communicable diseases. It’s hard enough, for the patient and for health professionals, dealing with just one serious non-communicable disease, and having more than one makes it harder, so if there’s some association between diet and combinations of these diseases, it’s important to know about the details, as far as possible.
“The issue about this being an observational study is the usual problem of what’s known as confounding. In a study like this one, people aren’t told what to eat – the researchers simply recorded what they said they ate when they joined the study, and then followed them up for many years (11 on average) and recorded if and when they had a diagnosis of a cancer, type 2 diabetes, or a cardiovascular disease (heart attacks, strokes, and so on). They found that, in some ways, the participants’ risks of having these health outcomes was higher if they had reported more consumption of UPFs all those years before. But that doesn’t mean that the risk differences were caused by the participants’ diets.
“There were many other differences, on other factors than diet, between the people who reported eating different amounts of UPFs. One or more of these other factors, known as confounders, might be what was actually causing the differences in health risks, in whole or in part, rather than the aspects of diet that the researchers recorded. Obviously the researchers were aware of this possibility. It is possible to make statistical adjustments to allow for confounders, and the researchers did that for a wide range of potential confounders. But you still can’t be sure that all confounding has been taken into account – there may be confounders on which the researchers just had no data, or maybe there is a confounder they did not consider. So a study like this simply can’t show with any degree of certainty that the differences in health risks are caused by the differences in diet. The researchers acknowledge that explicitly.
“You might wonder why researchers bother with prospective cohorts studies like this, and other observational studies, if each one of them can’s establish cause and effect. There exist other types of study, such as randomised clinical trials (RCTs), that can say much more about cause and effect. But imagine what an RCT to study links between multimorbidity and diet would look like. The diseases involved would typically take years, maybe even decades, to occur. So researchers would need to allocate participants at random to eat, for many years, different diets containing different amounts of UPFs, and participants would have to agree to do this. That’s completely impracticable. (To my knowledge, there has been only one published RCT about UPFs, but it was a small trial with 20 participants, looking only at short-term effects, over two weeks.) However, observational studies can add to the evidence base, and can point to new research directions that could be followed up with other observational studies, or with studies in cell cultures or, perhaps, experimental animals to throw light on the bodily mechanisms that could be involved. The number of observational studies that have found associations of various kinds between levels of UPF consumption and various health outcomes is now pretty large, though few have looked at multimorbidity as this new study does.
“Does all this begin to add up to persuasive evidence that eating more UPFs, in general, is bad for you? I’d say no, and this new study throws light on why not. My concern is that the category of UPFs is very wide and not always very clearly defined. (The researchers write in their paper, “We acknowledge that the Nova group 4 (i.e., UPFs) consists of very heterogeneous foods representing virtually all major food groups.” Also, though this is not entirely in my expertise as a statistician, not enough seems to be known about exactly how UPFs could cause ill health, though there’s been a lot of speculation. Is it because many UPFs contain high levels of fat, or sugar, or not enough fibre – that is, because of factors that have been known about and studied before the notion of UPFs even arose? Is it because many UPFs contain emulsifiers, even though after looking at detailed evidence on 28 different emulsifiers, a study1 published in September this year found evidence of association between emulsifier consumption and CVD risk for only 5 of them (and that was one observational study, so couldn’t establish cause and effect on its own)? Answering this sort of question is important because it can be difficult for some people to avoid UPFs entirely, perhaps on the grounds of cost, or time, or both, and if it’s possible to change the ingredients and processing to reduce any health risks, that would help.
“It’s worth mentioning that the absolute level of risk of disease in the participants in this study was not large anyway, and wasn’t very much higher in people who reported eating more UPFs. About 17 in a group of 1,000 people like those the study would have a diagnosis of at least two of the three conditions involved (cancer, type 2 diabetes, CVD) during the average 11-year follow-up period. In a group of 1,000 people who were similar to the first group, but who each had each eaten an extra amount of UPFs equal to the fairly large increment in UPF consumption considered in the researchers’ model (one standard deviation of consumption), if the association really was one of cause and effect, that 17 in 1,000 figure would go up to about 18 in 1,000. So we’re not talking about a large increase in risk.
“This new study looks not only at associations between overall reported UPF consumption and the multimorbidities that they examined, but also at associations for subgroups of the UPF category. They did this in a sensible statistical way, looking for the additional effect on health risk of changing consumption of UPFs in one subgroup if the consumption of other UPF subgroups was unchanged. And the subgroups differed in what was found.
“They looked at nine subgroups, and for only two of them was there reasonably persuasive statistical evidence that higher consumption was associated with increased risk. I must repeat that we can’t conclude these associations are caused by the UPF consumption, because the study is observational. Those two groups were animal-based products, and artificially and sugar-sweetened beverages. Both of those groups involve food or drink for which there’s already evidence of links with health. (UPF meat-based products, for example, often contain processed red meat, which is linked to bowel cancer risk, though the increase in risk from eating reasonable amounts is not immense.)
“But also there was one group, ultra-processed breads and cereals, for which people who ate more of them tended to have a lower risk of multimorbidity, not higher, though again we can’t know whether that association is one of cause and effect, and the statistical evidence for this association is pretty marginal. The researchers suggest that, if this association does really exist and is one of cause and effect, it could be because of fibre in those bakery products, which is known from previous work to be protective against some diseases. The other six UPF subgroups that were considered showed no clear evidence of association with multimorbidity risk.
“Confounding can hide an association that really exists, as well as making one appear that doesn’t really exist, and it can even change the direction of an association from higher consumption increasing risk to higher consumption decreasing risk, or vice versa. So we really can’t be sure of anything here, in terms of causation.
“What the study does indicate, though, is that the associations they found could, possibly, be explained by known mechanisms rather than by some newly discovered properties of some or all UPFs. That’s why the research paper talks about possibly advising people to change consumption of some UPFs, and, as the researchers suggest in their conclusion, that “more nuanced subgroup analyses of UPFs are warranted”.”
Prof Christine Williams, Emeritus Professor Human Nutrition, University of Reading, said:
The study:
“This Prospective Cohort Study is an observational study of the association between consumption of Ultra Processed Foods (UPFs) and risk of multimorbidity. Multimorbidity was defined as co-current diagnosis of at least two of three chronic diseases (cancer, cardiovascular disease and Type 2 diabetes). The study is large with recruitment of over a quarter of a million volunteers from seven European countries and uses a well-established food frequency questionnaire to assess habitual baseline intakes of foods over a one-year period (1992-2000). Volunteers were subsequently followed up to assess diagnoses of the three common chronic diseases in order to determine any relationships between intake of UPFs measured 20-30 years previously and subsequent diagnosis of at least two of these diseases.
Main findings:
“An important aspect of the study was that as well as measuring total UPF levels, seven UPF sub-groups were measured. They found two of the sub-groups, ‘Animal-based products’ and ‘Artificially and sugar-sweetened beverages’, were significantly associated with increased risk of multimorbidity. However another group- ‘Ultra-processed breads and cereals’ were significantly associated with lower levels of multimorbidity. The latter is an important finding because many informed commentators have identified these foods as important for intakes of fibre, micronutrients and complex carbohydrates. These foods are commonly eaten by children and those on limited incomes and should not be demonised by categorising them as UPFs. The reduced risk of multimorbidity with breads and cereals suggests the high UPF clasification may be misleading for some groups of foods. It should be noted this is an observational study and cannot conclude a cause and effect relationship between diet and diseases outcome – only the direction of an association; increase or decrease in risk.
Missing analyses:
“Another key question for UPFs are whether the reported associations are due to processing per se, or whether these reflect the fact that high UPF diets are high in fat, sugar and salt, high in processed meats but low in fruits and vegetables. It may be these dietary characteristics which are the real culprits rather than UPFs per se – adverse associations of some UPF sub groups with health outcomes might be because of their high levels of fat, sugar and salt. A statistical method can account for the known effects of each of these nutrients/foods on disease risk; if the risk association with UPFs disappears when these adjustments are made, then this suggest it is the diet in general rather than UPFs that are the problem. If the risk remains then it suggests other, as yet unknown, factors in UPFs that may be contributing to risk. It is disappointing that the investigators did not show the levels of these nutrients and foods in the diets, or adjust for them, especially as these analyses will be readily available to them. By not recording and accounting for these, the study has not resolved a critical issue which keeps rumbling on.
Criticism:
“During 1992-2000 when diets were assessed, the food frequency questionnaires will not have recorded information on the degree of processing. It is likely there will have been insufficient detail to allow the NOVA system (used from 2016 onwards) to be used to estimate dietary UPF levels accurately. To overcome the criticism the investigators attempted to adjust the classification of UPFs to take account of the customary levels of UPFs in individual countries over time. It is not clear whether those adjustments have been subjected to rigorous validation.”
Prof Tim Chico, Professor of Cardiovascular Medicine and Honorary Consultant Cardiologist, University of Sheffield, said:
“Over a quarter of all adults in the UK have two or more medical conditions (called multi-morbidity). These include diabetes, heart disease and cancer. We know already that many of these diseases can be caused by the same factors, like smoking, being overweight, not doing enough physical activity, and diet.
“This study has used well-established scientific approaches to understand the relationships between how much UPF someone eats and their risk of later cancer, heart disease and diabetes. Because it is very hard to continually measure what someone eats, the study asked people to complete a questionnaire about their diet in the previous year.
“The study found that people who reported eating more UPF had higher rates of diseases in later years, which is in keeping with a lot of other evidence about the relationships between poor diet and future health. It cannot prove conclusively that the UPF is the direct cause of the diseases, but taken with all the other scientific evidence it is very likely that some types of UPF do increase the risk of later disease, either because they are directly harmful or because they replace healthier foods such as vegetables, fruit, nuts, seeds, olive oils, etc.
“Research like this is very important. However, we already know fairly well what we should be eating (and doing) to reduce our risk of disease. We also know not enough of us achieve this. We are facing a growing epidemic of poor health. The most important question is what should we do about it?”
Dr Ian Johnson, Nutrition researcher and Emeritus Fellow, Quadram Institute, said:
“This new paper by Cordova et al draws upon data from over 266,000 European citizens enrolled in the longstanding EPIC prospective cohort study. The aim of this particular investigation was to explore the relationship between consumption of ultra-processed food (UPF) and the onset of multimorbidity, defined as the co-occurrence of at least two chronic diseases, including cancer, cardiovascular disease or type-2 diabetes in the same individual. After more than 11 years of follow-up, higher consumption of UPF was shown to be associated with a modest but statistically significant increase in risk of developing multimorbidity. The authors recognised however that the definition of UPF covers a very broad and diverse range of foods. They therefore broke the classification of UPF down into subdivisions and explored the contributions of the various different food types to the risk of developing multimorbidity. The ultra-processed foods most strongly associated with risk were found to be artificially sweetened and sugar-sweetened beverages, processed animal products, and sauces and condiments. However, a very wide range of other products, including ready-to-eat dishes, savoury snacks, and sweets and deserts were not shown to be associated with increased risk. Importantly, ultra-processed bread and cereal products showed an association with a reduction in risk. These observations do suggest a role for some UPF in the onset of multiple chronic disease, but they also show that the common assumption that all UPF foods are linked to adverse health effects is probably wrong. Furthermore, ultra-processed cereal products may be beneficial to health, perhaps because some provide convenient and palatable sources of dietary fibre.”
Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:
“This is a very detailed and thorough analysis of the association between ultra-processed food intake and health in a large European cohort. As this is an observational study, the authors relied on dietary data that was collected during the recruitment in the 1990s. This introduces some limitations – most importantly: the dietary assessment methods were not designed to determine ultra-processed food intake and the authors had to use a combination of different methods. The authors have attempted to address this in various ways in order to be able to estimate ultra-process food intake. However, differences in the methods and classification of foods (e.g. bread in the UK is usually considered ‘ultra-processed’, but not in all other countries) are likely to affect results.
“The authors acknowledge that “ultra-processed foods” is a very heterogeneous group of foods that include many different food groups. Although they are often referred to as “junk foods”, not all ultra-processed foods are high in fat or sugar, are energy dense or lead to increased food consumption. The authors have therefore investigated the effect of different food groups: like previous studies, they found that the consumption of soft drinks (sugar and artificially sweetened beverages) was strongly associated with an increased risk of mutli-morbidity. A strong negative association between the consumption of these drinks and heart diseases is well known and is one likely driver of the observed associations between ultra-processed food intake and disease risk.
“The authors investigated this further and found that adjusting for soft drink consumption attenuated the association between ultra-processed food intake and multi-morbidity, i.e. it explains the association to some extent. The authors also found that the consumption of animal based ultra-processed foods was associated with an increased disease risk. For most other food groups within the ultra-processed category, there was no association with any change in disease risk – the exceptions being bread and cereals (which showed a borderline reduced risk) and sauces and condiments (which showed a borderline increased risk).
“These findings are important, because they show that not all UPFs are created equal and only some are associated with an increased risk to health and any research and recommendations should focus on these food groups. Notably, these are food groups where it was already known that they are associated with an adverse effect on health.”
Prof Tom Sanders, Professor emeritus of Nutrition and Dietetics, King’s College London, said:
“This paper reports a 9% increase in risk of multimorbidity to be associated with higher intake of ultraprocessed food (UPF). However, among the various categories of UPF only animal products and sweetened beverages were associated with risk. There was no evidence that ultraprocessed bread and cereal products or other categories of UPF were related to risk. Consequently, the finding cannot be generalised to all UPF.
“Estimates of UPF intake were made from food frequency questionnaires completed 23-28 years ago among participants of European Prospective Investigation in Cancer (EPIC). The participants were followed up and new cases of morbidity from cancer, cardiovascular disease and type 2 diabetes recorded. Analyses were then done to see how many patients developed multimorbidity.
“Increases in risk of developing type 2 diabetes (7%) and CVD (5%) and higher total UPF intake were found. These findings confirm what has previous been reported for processed meats and sweetened beverages.
“In my opinion the use of multimorbidity is a bit of a red herring because type 2 diabetes is almost inevitably linked to multimorbidity. Indeed, between a quarter and a third of hospital patients have type 2 diabetes compared with 6-8% in the general population.
“Patients with CVD are also at greater risk of developing type 2 diabetes as well as renal failure. In this study most of the multimorbidity was associated with CVD (30%) and diabetes (29%) compared to cancer (16%).
“The strengths of the study are the long period of follow-up and relatively robust statistical analysis which has excluded the contribution of alcoholic drinks, which are included in the NOVA definition of ultraprocessed foods. The weakness is that food intake was measured by a questionnaire on one occasion a long-time ago. This is important as dietary patterns have changed quite markedly in the past twenty five years with more food eaten outside the home and more ready prepared food being purchased.”
Prof Keith Frayn, Emeritus Professor of Human Metabolism, University of Oxford, said:
“Many studies have shown adverse health outcomes associated with ultra-processed food consumption, and we now need to understand better which components of these foods are the culprits. The advantage of this study is that individual components of those foods have been assessed, and point towards animal-based products and sweetened beverages as those that cause harm. As with any observational study, there may be many confounding factors. This study is also limited because the data on foods eaten was collected at baseline, and may well have changed during the average 11 years of follow-up. But accepting the results as they are, it is particularly notable that adjustment of the results for BMI makes, in general, little difference, so these adverse outcomes seem not to be related just to increased body weight. The results point once again to a need to be vigilant about ultra-processed food consumption, but also for further research.”
‘Consumption of ultra-processed foods and risk of multimorbidity of cancer and cardiometabolic diseases: a multinational cohort study’ by Reynalda Cordova et al. was published in Lancet Regional Health – Europe at 23:30 UK time on Monday 13 November 2023.
DOI: 10.1016/j.lanepe.2023.100771
Declared interests
Dr Duane Mellor: “I have no commercial interest in these comments, however I am speaking to a number of groups, including food companies about what is meant by UPF and their potential to adversely impact health.”
Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee. My quote above is in my capacity as an independent professional statistician.”
Prof Christine Williams: “2020-23: Trustee Academy of Nutrition Sciences; Chair, 2022-2023.
Nominations Committee, BNF.
2020-23: Chair, World Cancer Research Fund Grant Panel; Chair, Transforming Food Systems Grant Panel (UKRI); Trustee, Scottish Rural University College.”
Prof Tim Chico: “No conflicts.”
Dr Ian Johnson: “No conflict of interest.”
Prof Gunter Kuhnle: “Employment: University of Reading (since 2010) and member of the USS.
Research funding: current funding from BBSRC TUKFS (Co-I, since 2020); past research funding from Mars, Inc (2014-2018); EU Horizon Programme (2012-2016).
Committee membership: Committee on Toxicity (COT) and various working groups, past member of the EFSA ANS panel and the American Academy of Nutrition and Dietetics working group on flavanols.
Scientific organisations: British Mass Spectrometry Society, Nutrition Society, Deutsche Gesellschaft für Epidemiologie and Deutsche Gesellschaft für Ernährung; Trinity Hall Alumni Organisation; Registered Nutritionist.
Member of the organising committee of the International Conference on Polyphenols and Health.
Trustee of a Parent Teacher Association.
Family Vineyard.”
Prof Tom Sanders: “Member of the Science Committee British Nutrition Foundation. Honorary Nutritional Director HEART UK.
Before my retirement from King’s College London in 2014, I acted as a consultant to many companies and organisations involved in the manufacture of what are now designated ultraprocessed foods.”
Prof Keith Frayn: “I have no conflict of evidence to declare other than that I am the author of books on nutrition and metabolism.
Author of:
Lipids: Biochemistry, Biotechnology and Health 6th edn (Wiley) published 2016– with Mike Gurr and others;
Human Metabolism: a Regulatory Perspective 4th edition (Wiley) published 2019 with Rhys Evans;
Understanding Human Metabolism (CUP) published 2022;
Forthcoming:: The Calorie Equation (Piatkus) 2025.”
For all other experts, no reply to our request for DOIs was received.