select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
A study published in eClinicalMedicine looks at consumption of ultra-processed food, risk of cancer and cancer mortality.
Prof Tom Sanders, Professor emeritus of Nutrition and Dietetics, King’s College London, said:
“This study is an exploratory analysis searching for possible associations between the consumption of ultraprocessed foods and risk of cancer at multiple sites. The major weakness of this approach is that when a very large number of comparisons are made some will appear statistically significant by play of chance. It can be useful for picking some new risk factors for further investigation. However, the definition of ultraprocessed food is so vague which makes establishing any cause-effect relationship problematic.
“The participants of the study were from the U.K. Biobank. Unfortunately, there were a very large number of statistically significant differences between those consuming the highest vs the lowest level of intake with regard smoking habit, obesity, socioeconomic status, physical activity, and educational attainment. This makes it statistically almost impossible to correct for these correlated behaviours. The study found those with the highest intake of ultraprocessed foods had a 25% higher risk of lung cancer (given that 90% of lung cancer is smoking related). As cancer of the lung accounted for most of the excess cancer risk, it questions the validity of reporting an increase in the overall cancer risk being attributed to a high intake of ultraprocessed foods.
“The primary statistical analysis of study also found significant trends for a greater risk of head and neck cancer, a lower risk for cancer of the oral cavity, and a borderline statistically significant trend (P=0.03) for ovarian cancer in those with the highest reported intake of ultra-processed foods. There were no associations with gastrointestinal cancers which are known to be associated with dietary intake. After attempting to adjust for potential confounding factors, the association with ovarian cancer was increased to a 19% increased risk of ovarian cancer.
“The claim that ultraprocessed foods are associated with an increased risk of all cancer is unsound because smoking related cancers such as those of the head and neck, and lung, were greater in those with the highest intake. The authors recognise that the association with head and neck tumours might be confounded by alcohol and tobacco use. However, risk of ovarian cancer does not appear to be related to cigarette smoking. The Word Cancer Research Fund rolling review found strong evidence for increased adult height (reflecting growth in early life and childhood) to be associated with ovarian cancer and probably an effect of obesity but no other dietary factors (e.g., yogurt and egg intake) have been consistently related. The association with ultraprocessed food and risk of ovarian cancer in this study is novel but given the relatively small number of cases (291) of ovarian cancer, the finding needs replication in other prospective cohorts before taking the claim that ultraprocessed foods increase the risk of cancer seriously.”
Dr Simon Steenson, Nutrition Scientist, British Nutrition Foundation, said:
“One of the main strengths of this study is the large data set from the widely used UK Biobank, meaning that the findings are relevant to UK adults. The authors also included 34 different cancer types in their analysis, giving much more information than in previous studies, especially on less common cancer types. However, a significant association between ultra-processed foods (UPFs) and risk of death from specific cancers was only found for ovarian and breast cancers, and not for others that were analysed.
“The findings add to previous studies showing an association between a greater proportion of ultra-processed foods (UPFs) in the diet and a higher risk of obesity, heart attacks, stroke, and type 2 diabetes. However, an important limitation of these previous studies and the new analysis published today is that the findings are observational and so do not provide evidence of a clear causal link between UPFs and cancer, or the risk of other diseases.
“It is possible that a higher proportion of UPFs in the diet is a marker of an overall poorer diet, which is higher in energy, saturated fat, salt, and free sugars, and lower in fruit, vegetables, fibre and essential nutrients – dietary factors that are known to negatively affect health.
“A few different explanations have been put forward for why eating more UPFs is associated with a higher risk of chronic diseases. For example, it has been suggested that the negative association might be linked to potential effects of food additives, compounds produced during processing, or chemicals found in food packaging. However, at the moment, we don’t have enough data from human studies to know if any (or a combination) of these of factors might be involved in the negative link between UPFs and health found in observational studies. Food processing encompasses a wide range of different methods and some studies have indicated, for example, that changes to food structure may impact nutrient absorption and eating rate. Research into the effects of processing on health is therefore ongoing.
“Many foods and drinks that would be classified as UPF are high fat, salt and sugar (e.g. cakes, biscuits, fried snacks, confectionery) and it is already recommended that we consume this type of foods less often and in smaller amounts (if at all). However, an issue with the concept of UPFs is that this category can also contain commonly consumed foods that provide important nutrients, such as packaged wholemeal bread, which contains fibre and essential vitamins and minerals, or high fibre, lower sugar breakfast cereals that are commonly fortified with key nutrients. These and other healthier foods that would be classified as UPF, such as reduced sugar and salt baked beans or vegetable-based pasta sauces, can form an important part of a healthy, balanced diet and provide affordable and widely available options that can form the basis of nutritious meals.”
Dr Duane Mellor, Registered Dietitian and Senior Lecturer, Aston Medical School, Aston University, said:
“This is the latest in a number of studies looking at the association between ultra-processed foods and health risk – in this case, the risk of developing or dying from cancer.
“This study looked at a large number of individuals who were part of the UK Biobank study. It did look to try to control for many of factors which are strongly linked to risk of cancer including household income (living in poverty), smoking and age which may link to intake of processed foods. They did report a significant association between consumption of ultra-processed foods and risk of all types of cancers (when all cases are combined together) and specifically the risk of ovarian cancers (but not risk of specific cancers at other sites) and the risk of death from all types of cancers, as well as specifically ovarian cancer and breast cancer.
“When looking at this type of study it is important to recognise in our modern often urban world, some processing of food has improved health as diseases linked to aflatoxins from fungi in food, which include some forms of liver cancer, have been virtually eradicated in countries like the UK.
“When looking at food intake of people reported to be consuming more ultra-processed foods, they also tended to drink more fizzy drinks and less tea and coffee, as well as less vegetables and other foods associated with a healthy dietary pattern (e.g. pulses, nuts, seeds, fruit etc.). This could mean that it may not be an effect specifically of the ultra-processed foods themselves, but instead reflect the impact of a lower intake of healthier food. There is also a potential issue by what is meant by ultra-processed foods, as the definition used can mean that ultra-processed foods can include many shop bought item such as a biscuit or potentially even a loaf of bread, are deemed worse than the near identical food made at home. The difference being time and cost, resulting in potential stigma, which is even more damaging at a time of a cost of living crisis, making people feel guilty about the food that they can afford and know their families will eat.
“So, although this study showed an association between ultra-processed food intake and some cancers, which the authors acknowledge does not show a causal effect, it does not really change the messages from the study’s funder, World Cancer Research Fund, about encouraging a healthy dietary pattern rich in vegetables, pulses, nuts, seeds and fruit, with wholegrains and less processed meat, red meat and alcohol with the aim of staying physically active and maintaining a healthy body weight. Hopefully, the attention that the currently popular theme of ultra-processed food is getting does not distract from these basic messages which ultimately have not changed, and that they do not add to the stigma associated with food choice, by simplistically labelling then as good or bad, or in this case minimally processed or ultra-processed.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“This is yet another study that uses data from a large cohort of people, drawn from an existing long-term study, to investigate possible associations between consumption of ultra-processed foods (including drinks) and a health issue – in this case, cancers. In statistical terms, the study uses appropriate methods and the statistical analysis is thorough.
“The researchers found evidence of increased numbers of diagnoses of all cancers taken together, and of increased numbers of deaths from all cancers taken together, in people who diets included greater proportions of ultra-processed food. These increases in cancer diagnoses and deaths were arguably not all that large, proportionally, though because cancers are such an important cause of ill-health and death, even a relatively small increase could matter.
“However the researchers also looked separately at cancer diagnoses and deaths for separate sites in the body. They found evidence of stronger associations between ultra-processed food consumption and ovarian cancer diagnoses and deaths. There was also a hint of an association between ultra-processed food consumption and breast cancer deaths, though the evidence for that is not strong statistically. I use the word ‘association’, as do the researchers, because these are patterns of correlation, and research of this type simply cannot establish that the higher levels of cancer diagnoses or deaths were actually caused by consumption of ultra-processed foods.
“This issue, of not being able to conclude what causes what, applies to all the research of this general type on ultra-processed food consumption. Indeed the new research has several other important limitations that it shares with pretty well all the previous studies of this type.
“The problem about cause and effect arises because the study is observational. The researchers did not seek to influence what food and drink the participants consumed. They just recorded what they consumed, and followed them up for about ten years on average, and recorded new cancer diagnoses and cancer deaths. So participants who consumed different amounts of ultra-processed food happened to differ in terms of many other factors too. (That can be seen in Table 1 of the research paper, which shows differences between ultra-processed food consumption groups in terms of all 21 other factors recorded in the table.) So the differences that the researchers found in cancer diagnosis and death rates between the consumption groups could be caused, in part or in while, by some of these other factors, rather than the ultra-processed food consumption.
“Of course the researchers understood this, and they made statistical adjustments to allow for a number of factors that could have influenced cancer rates. The snag is that one can never be sure whether everything relevant has been adjusted for, or that the adjustments have been done appropriately. The researchers clearly acknowledge this in their conclusion, writing that “causality may not be implied owing to the observational nature of the study”. But this does not mean that high ultra-processed food consumption definitely does not cause ovarian cancer (for instance). (Sorry, double negative.) It might or it might not – this research just can’t tell us which.
“It’s worth mentioning that these issues of other so-called confounding factors in observational research can apply the other way. The researchers in this study did not find evidence, for most of the cancer sites that they looked at, of an association between ultra-processed food consumption and cancer diagnoses or deaths from cancers in those sites. In many cases that might well be because in fact there is no link to find evidence of. In other cases it could be simply because there wasn’t enough evidence, even from the almost 200,000 participants they had, to separate an association from the noise in the data. But in yet others, it’s possible that there is an association, that could indeed be causal, but the pattern of other factors is hiding that association. Generally the research results can’t clearly tell us which of these possibilities might be the case, for those other cancer sites.
“There are other limitations stemming from the details of the study, and the researchers acknowledge these too. The cohort of people that they studies was not particularly typical of the UK population, including in terms of ethnicity, deprivation, obesity and indeed the level of consumption of ultra-processed foods. So it’s possible that patterns observed in this cohort are not entirely typical of the patterns in the general population, in ways that might relate to the association between ultra-processed foods and health. And there are issues about how their consumption was recorded. The researchers relied on questionnaires recording what people ate over 24-hour periods. This happened five times between 2009 and 2012, though in fact, for almost two in every five participants they had data from only one of those 24-hour periods. So for many participants, there could be no record of changes in diet, and no changes between 2012 and when the study ended in 2021 could have been recorded for anyone. Also, although the questionnaire was well validated, it was not designed specifically to categorise foods as ultra-processed. The researchers had to make that classification depending on the categories of food that were originally used, and in some cases it isn’t clear whether a food counts as ultra-processed or not. (For instance, bread may or may not count as ultra-processed depending on how and where it is made.)
“A further issue is that research like this cannot examine what it is about eating and drinking ultra-processed foods that could lead to cancer or other ill health. The research report does suggest a number of possible mechanisms that could occur within the body and lead to cancers in people who consume a lot of ultra-processed food, but evidence from this particular research can’t itself throw light on which, if any, of them is important. The category of ultra-processed foods is wide, simply because there are many ways of processing foods, and the processing methods and added ingredients can differ a lot between different ultra-processed foods. Maybe some types of processing and some ingredients are particularly damaging to health, and others actually aren’t? This type of research can’t examine that. In this new study, the differences in consumption of ultra-processed foods seem to have been largely in just a few categories – particularly fizzy drinks, fruit-based drinks, ready meals and industrially made bread (Figure 1 in the paper). Is it those that are causally linked to cancers, and might some of the others not matter so much? This research just can’t tell us.
“There are other associations and correlations between possible risk factors and health outcomes where a lot of the evidence has come from observational studies, but over time and across many studies, it becomes generally accepted that the risk factor is indeed causing differences in health. There have now been many observational studies that link high consumption of ultra-processed foods (and drinks) to poor health outcomes. Can they be put together to say that there is now evidence of cause and effect? I’d say not, and that’s because the observational studies pretty well all have the same issues as this one does. This means that important doubts about cause and effect arise in the same ways in all of them, so the existence of more and more such studies doesn’t really rule out those doubts.
“Yes, the studies are observational, but they have more in common than that. They are carried out in existing long-term studies, generally some of the recording of consumption was only made at the start of the study, or only infrequently in many participants, and most cannot really link consumption of specific types of ultra-processed foods to the health outcomes they looked at. Really this is inevitable to some extent. You need a lot of participants to pick up the associations, you need a lot of data on them to make the appropriate statistical adjustments, and you need to follow them up for a long time to be able to pick up diseases like heart disease and cancer that can take many years to develop. So you really need a large group who have volunteered to take part in a long-term study of health, possibly related to nutrition, possible more general (as with UK Biobank, used in this new research). People who volunteer for something like that won’t be typical of the general population. Setting up a large enough study cohort specifically to look at ultra-processed food would be very expensive, compared to a study like UK Biobank that shares its costs among many different investigations. The food consumption recording was generally started some time ago, to allow for the necessary length of follow-up, so generally it is not tailored to measuring ultra-processed food consumption. The concept of ultra-processed food wasn’t published until about 2009, and the current four-category food classification used in this paper came out about 2016, so couldn’t possibly have been used to design the food questionnaires used here and in most of the other studies.
“But what really helps in establishing cause and effect, where much of the research has to be observational, is good research on possible mechanisms by which the risk factor can cause the ill health. This research paper reports that some of this is going on, though not in this study or studies like it.”
‘Ultra-processed food consumption, cancer risk and cancer mortality: a large-scale prospective analysis within the UK Biobank’ by Kiara Chang et al. was published in eClinicalMedicine at 22:00 UK time on Tuesday 31 January 2023.
DOI: 10.1016/j.eclinm.2023.101840
Declared interests
Prof Tom Sanders: “No conflict of interest in the past 3 years.”
Dr Simon Steenson: “Funding to support the British Nutrition Foundation’s charitable aims and objectives comes from a range of sources including membership, donations and project grants from food producers and manufacturers, retailers and food service companies, contracts with government departments; conferences, publications and training; overseas projects; funding from grant providing bodies, trusts and other charities. Further information about the British Nutrition Foundation’s activities and funding can be found at http://www.nutrition.org.uk/aboutbnf/.”
Dr Duane Mellor: “No conflicts of interest.”
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.”