select search filters
briefings
roundups & rapid reactions
before the headlines
Fiona fox's blog
A study published in the Lancet Planetary Health looks at Food processing and cancer risk in Europe from the prospective EPIC cohort study, and was press released by the International Agency for Research on Cancer (IARC).
Prof Tom Sanders, Professor emeritus of Nutrition and Dietetics, King’s College London, said:
“This a modelling study not a trial, and not a direct measurement of association of an individual’s food intake with risk of cancer. The major problem is that lifestyle factors other than food intake (especially smoking and alcohol intake) are strongly associated with risk of cancer. The increases in cancer reported to be associated with ultraprocessed food are also associated with alcohol intake (especially those of the mouth, pharynx and oesophagus which are especially high in smokers, and liver). The definition of ultraprocessed foods includes alcoholic beverages. When they adjusted for alcohol intake there were no statistically significant differences in risk of cancer of breast in post menopausal women, of rectum and of liver. Smoking had a very large effect on risk on cancer risk on cancers of the upper digestive tract. Consequently, it is difficult to rule out the effects of past smoking habits. The size of the effects reported to be associated with processed foods was small compared to the known lifestyle factors (e.g. smoking, alcohol, obesity).”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“The researchers have generally used appropriate statistical methods, but the data were not collected originally in order to investigate risks associated with eating ultra-processed foods, so there are some limitations associated with that. Also, this is an observational study. There are always some issues on sorting out cause and effect in observational studies, and this one is no exception. It’s important to note that the researchers, in the conclusions of their research paper, don’t go beyond saying that their findings suggest that substituting some processed or ultra-processed foods with unprocessed foods might reduce the risk for some cancers. They also ask for more research using more recent data on people’s diets, and to examine what might be going on in people’s bodies that might possible explain the associations they observed.
“The study is observational, in that the researchers did not try to change anyone’s diet or behaviour, but simply recorded what they ate, and some aspects of what they did, and followed them up and observed when any of the participants was diagnosed with a cancer. People who had different amounts of processed or ultra-processed foods in their diet also differed in many other factors too, on average. So any differences in cancer risk between the different diet groups could be caused, in whole or in part, by one or more of these other factors, and not just (or not at all) by the dietary differences.
“The researchers were, of course, aware of that issue, and they made statistical adjustments to allow for differences in these other factors, where they had relevant data. So they allowed for differences in age, sex, smoking, physical activity, some aspects of their diet apart from how much their food had been processed, and several more factors. They also allowed statistically for the fact that they had looked at associations between diet and several different cancer types – without that adjustment, the chance of finding a statistical association that is due to no more than chance would have been increased. But these adjustments, particularly the adjustment for the other potentially important factors, can never be perfect. So, as the researchers acknowledge, they can’t be totally sure about cause and effect in a study like this.
“The issue of cause and effect is made more relevant because this study couldn’t take account of changes in diet, or in other relevant factors, over the long follow-up period (14 years on average). The data on diet, and on most of the other factors that might affect cancer risks, were generally collected when participants first entered the EPIC study on which these finding are based. For many of them, that was more than 20 years ago. The availability of processed foods has changed over that time, as the researchers point out, and there may well also have been changes in things like people’s physical activity, and their use of tobacco and alcohol, which would not have been recorded and so could not be taken into account in the statistical adjustments. Indeed the recording of participants’ food consumption in the four NOVA categories is unlikely to have been perfect, since the original diet recording was not done to measure NOVA categories, which had not even been defined when many of the participants’ diets were recorded as they entered the EPIC study that provided the participants and the data for this work. For example, the researchers say that, over the past twenty years, there must have been changes in production of bread in the UK – they say that bread is now predominantly industrially produced (so counts as ultra-processed), whereas previously more of it was produced in bakeries that did not use industrial methods (which count as processed but not ultra-processed, NOVA group 3 rather than 4). The participants weren’t asked where they bought their bread when they were recruited, and weren’t asked about their diets at all during the years after recruitment, so there’s no data on this. The researchers did do additional statistical analyses to try to allow for these changes over time, but they certainly can’t account for everything, in my view.
“I’m also not entirely convinced by the ‘substitution analysis’ that the researchers performed, where they investigated the effect on cancer risk of moving 10% of the participants’ consumption from, first, the processed food category (NOVA 3) to the unprocessed category (NOVA 1), and then, separately, moving 10% from the ultra-processed category (NOVA 4) to NOVA 1. This was not done, and couldn’t have been done with the available data, by looking at how individuals’ cancer risks changed if they changed their diets in these ways. Instead, they are essentially using their statistical models to compare different people who reported different consumption levels in the relevant categories. That’s OK as far as it goes, but it certainly doesn’t tell us clearly what might happen if particular individuals did change their diets in these ways. Anyway, some changes between NOVA 3 or 4 and NOVA 1 could be quite straightforward, but replacing bread (NOVA 3 or 4 depending on its source) by entirely unprocessed foods (NOVA 1) may be more of a challenge.
“A final point is that the researchers certainly did not find an association between the risk of every cancer that they considered, and the consumption of processed food. There was an association with the risk of ‘all cancers’, but the ‘all cancers’ data are found by adding together diagnoses of every type of cancer. As their other analyses showed, there seem not to be associations at all with rather a lot of individual types of cancer. If only one cancer site was associated with diet (and these researcher did in fact find more associated cancer sites than just one), that would mean that the total number of cancers, ‘all cancers’, could well be associated with diet, because that total number would include the one cancer site that was associated with diet, even if all the other types of cancer had no association with diet. To understand more of this, it would be necessary to know more about how processed and ultra-processed food are associated with cancer, in terms of what’s going on in people’s bodies. A study like this one simply can’t look at these so-called mechanisms behind the associations, as the researchers point out by asking for more research on that aspect.”
Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:
“The EPIC cohort is large and includes people from all over Europe, so can provide very good data on the association between diet and health. In the past, the EPIC cohort has provided many ground-breaking results. It is therefore not surprising that they have investigated the association between ‘food processing’ and cancer risk. In this study, they have used a so-called ‘substitution analysis’ to find out what would happen if people swapped different food items – this is very useful to understand the potential impact of dietary changes.
“The main shortcoming of the study is the reliance on the ‘NOVA’ classification, which is a very arbitrary system of classifying food according to a perceived state of their processing. In this study, the authors put a considerable amount of effort into the classification of foods as ‘NOVA’ can be very ambiguous: for example, bread is considered “ultra-processed” when purchased in a supermarket but only “processed” when purchased from a bakery – irrespective of the ingredients used. As the baseline of EPIC – and thus the dietary assessment – happened in the 90s, the authors had to make assumptions about the likely sources and classifications of foods.
“The authors found that reducing the amount of processed and ultra-processed food intake was associated with a reduced risk for some cancer – and this is not surprising as other studies have shown similar results. The question is however whether this is due to ‘food processing’ or whether there are other factors. The group of ‘ultra-processed’ foods is incredibly heterogenous and includes breads, baked beans and biscuits – but also tinned fruits and high-fibre breakfast cereals. The group of ‘processed foods’ is even more heterogeneous and these foods have very little in common. Previous studies have showed that some – but not all – ‘ultra-processed’ foods are hyperpalatable and therefore increase the risk of obesity and associated diseases. Soft-drinks – classed as ‘ultra processed foods’ are known to increase the risk of a range of diseases when consumed in large amounts.
“It is unfortunate that the press release refers to ‘processing’ and claims that ‘processing’ introduces harmful compounds. Food processing is crucial for the food safety and security we enjoy, and it is very likely that our health would be considerably worse without food processing. The formation of acrylamide – a known carcinogen – is usually much better controlled in an industrial process than the domestic kitchen. Food processing is also crucial to reduce food waste and food cost by extending shelf life – preservation methods are crucial to ensure a secure food supply.
“We definitely need to understand the underlying reasons for these observations, but they are definitely much more complex than a four-category classification system that currently enjoys popularity.”
Dr Duane Mellor, Registered Dietitian and Senior Lecturer, Aston Medical School, Aston University, said:
“Although this study looked at over 450000 people and their risk of a range of cancers, it is important to note that it used food frequency questionnaires to collect information about dietary habits, except in the UK when a 7-day diet diary was used and Sweden where it recorded hot meals. So, the quality of the food intake data is mixed, making it more challenging to measure and compare intake of all foods let alone processed foods.
“The data on food intake was collected between 20-30 years ago, so the researchers had to model the data to look at how processed food intake might have changed over time. This means that although the paper claims a reduction in risk of all cancers when 10% of intake of processed foods are substituted, this is not quite correct – it compares people it estimates to eat processed food with those who eat 10% less processed foods (and therefore 10% more minimally processed foods).
“Although it claims to show that minimally processed foods are linked to lower risk of cancer, this is not really any different to current dietary recommendations to the general public as well as those aimed at reducing risk of developing cancers – that we should try to eat a wider variety and increase the amount vegetables, fruit, pulses, nuts, wholegrain and seeds in our diet and eat less processed meats and foods high in fat, salt and sugar such as pies, cakes, biscuits, sweets and other confectionary.”
‘Food processing and cancer risk in Europe: results from the prospective EPIC cohort study’ by Nathalie Kliemann et al. was published in the Lancet Planetary Health.
DOI: 10.1016/S2542-5196(23)00021-9
https://www.thelancet.com/journals/lanplh/article/PIIS2542-5196(23)00021-9/fulltext
IARC press release: https://www.iarc.who.int/wp-content/uploads/2023/03/pr330_E.pdf
Declared interests
Prof Tom Sanders: “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.”
Prof Gunter Kuhnle: “No COI.”
Dr Duane Mellor: “No conflicts of interest.”