Research, published in JAMA Internal Medicine, reports that a higher consumption of ultraprocessed food may be linked to a higher risk of developing type 2 diabetes.
Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:
“This is an interesting and well conducted study which shows that people who consume ultra-processed foods have a slightly higher risk of type 2 diabetes. However, participants with higher ultra-processed food intake had also other risk factors for diabetes such as being less active, obese and were current smokers – but the authors have taken this into consideration.
“This study adds to growing evidence that shows an association between ultra-processed food intake and poor health outcomes – and this raises the question of a causal link. Food additives (‘E-numbers’) are often blamed, but their impact on health is regularly assessed – as is the case for contact materials. The main difficulty in establishing a causal link is that the definition of ultra-processed food includes a wide range of foods with vastly different nutrient profiles: according to the classification used, ultraprocessed foods include foods such as sausages, burgers, ice-cream, infant formula, cake mixes and all spirits. Some of these foods contain no additives (e.g. spirits) or only few additives (sausages, burgers, fish-fingers) – whereas others can contain a wide range of different ones. Some of these foods are important sources of nutrients (e.g. B12 and iron from meat products, omega-3 fatty acids from fish fingers), whereas others provide very little.
“The observed associations are strong enough to justify more research to understand the underlying reasons. But the current definition of ultra-processed food is too unspecific to derive specific dietary recommendations. Here is a link to the definition used: https://worldnutritionjournal.org/index.php/wn/article/view/5/4”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“This study compared the risk of being diagnosed with Type 2 diabetes in participants who ate different amounts of what are known as ultraprocessed foods, and found that the more of these foods people ate, the greater was their chance of becoming a new diabetes case in a given period of time. It’s a large study with over 100,000 participants, and it made pretty thorough use of appropriate statistical methods. So why am I sceptical about its findings? That’s because we can’t be sure whether eating ultraprocessed foods is what’s actually causing differences in diabetes risk; because, even if the association is one of cause and effect, it’s still unclear how that cause and effect works; and because the data come from a pretty untypical group of people.
“My first concern, which applies to any study like this, is that it’s observational. People weren’t allocated by the researchers to eat different diets – they ate what they chose to eat, and the researchers recorded what they ate and also recorded cases of type 2 diabetes. There are many other differences between people who ate different amounts of ultraprocessed foods, apart from their ultraprocessed food consumption. So some of these other differences might be the cause of differences in diabetes risk, and not the overall consumption of ultraprocessed foods at all. To some extent, the statistical results can be adjusted to try to allow for the effect of confounding factors like this, and indeed the researchers on this study looked at the effects of several such adjustments. These adjustments can never be perfect, though, and you can only adjust for factors on which you have data, so that the researchers are careful to point out that they may not have been able to take everything into account. There may be residual confounding, as it’s known. So the researchers rightly point out that this study on its own cannot show that eating a greater proportion of ultraprocessed food actually caused the participants’ diabetes risk to increase.
“The new study does not really throw much new light on how ultraprocessed food consumption might possibly be a cause of changes in diabetes risk. Better evidence on that might increase confidence that the association is one of cause and effect, to some extent at least. They do list various ways by which certain components or properties of some processed foods might have effects on the processes that lead to Type 2 diabetes. But there’s a problem in applying insights like that to the overall measures of ultraprocessed food consumption, because the list of ultraprocessed foods is very long and diverse. The possible causal pathways from the foods to the diabetes do not apply to every food on the ultraprocessed list, so it would need analysis of individual foods or groups of foods to throw clearer light on what might be going on, and despite the 100,000 participants, there isn’t really enough data for most such investigations. (Actually the 100,000 is a bit of a red herring – what really matters here is how many of the participants actually got diabetes during the study, and that was much smaller at 821.)
“A large proportion of the published research on ultraprocessed foods comes from just a very few cohorts of participants, and this particular group, the French NutriNet-Santé cohort, is the one that has by some margin been the most studied. It’s important to note that people volunteer to be in the cohort, which they are unlikely to do unless they have particular reasons to be interested in nutrition and health. As a result, the participants are not very typical of the general French population – they are on average younger, better educated, and have different dietary habits, and far more of them are female. (Four out of every five participants in this particular piece of research were women.) That might not be particularly important if the statistical adjustments can adequately allow for these idiosyncrasies, or if we actually knew how ultraprocessed foods might change diabetes risk, but we can’t be at all sure of these matters. What’s more, the participants were followed up for just 6 years on average (median), and since they are relatively young, that can’t tell us much about long-term effects in older people, for instance. And again, the great diversity of the list of ultraprocessed foods doesn’t help. Participants who eat different amounts of ultraprocessed foods may well eat very different types of these foods as well. Some of the statistical adjustments might have allowed for some of that, though we can’t be entirely sure. And the adjustments can’t allow for differences between the cohort of participants and the overall French population in terms of which ultraprocessed foods they tend to eat.
“The position is even less clear if we consider how far these results might apply to other countries, such as the UK, where eating habits do still tend to be quite a lot different from France. Because of that, and because fundamentally we can’t be sure about what causes what, I’m not even going to attempt to provide numbers that translate the results of this research to the UK context.
“None of these reasons for scepticism mean that there’s definitely no effect of eating ultraprocessed foods (or some of them at least) on the risk of type 2 diabetes. I’m just saying that we still can’t be at all sure what’s actually going on.”
Dr Duane Mellor, Registered Dietitian and Senior Teaching Fellow, Aston Medical School, Aston University, said:
“This is a large study of over 100,000 French adults which were followed up for an average (median) of 6 years. This study confirms messages in current recommendations such as those from Diabetes UK and American Diabetes Association with respect to the management of type 2 diabetes. These both aimed to make recommendations based on food, not just nutrients and focus on suggesting people try to follow a healthy overall dietary pattern, such as a Mediterranean type diet. In this study of a French adults, it appears that high intakes of ultraprocessed foods, which are the foods which fall outside the main plate model of the Eatwell guide and therefore are recommended to be limited, are linked to higher risk of developing type 2 diabetes. The problem being that we know from countries like Australia which has similar dietary habits as the UK and US (and increasingly France who are moving to this Westernised diet) that around a third of our daily calories comes from these types of foods. So, reducing how much of these ultraprocessed foods, which is already a core part of current dietary guidelines is a simple step that could improve our health individually and as a society, so this study largely tells us what we already know.
“The study is clear that it does not show that this link is causal, it is only an association. There are some confusing aspects in that the statistics try to take out the effect of the diet quality measured using the UK Food Standard’s Agency model for foods high in fat, sugar and salt, which is the one used in the advertising ban on the Transport of London estate. This is problematic as the authors suggest one reason that ultraprocessed foods may increase risk of type 2 diabetes is because of the fat, sugar and salt. This should not be shown in the statistics as it seems that it should have been accounted for in the statistical modelling, meaning to suggest this study suggests what we already know to eat less fat, sugar and salt was not clear, as before looking at the effect of ultra-processed foods they corrected for the effect of these nutrients.
“Perhaps as a scientific community we need to be more responsible as the data from this exact same population has been used to look at the effect of ultraprocessed foods on risk of cardiovascular disease which was only published in May this year. Whilst it’s not uncommon to reuse a data set with different outcomes, it doesn’t not always help the public. Although the message that a diet based on more fresh vegetables, fruits, nuts and seeds with some dairy, fish and meat is healthy is not new. The problem with ultraprocessed foods is that not all foods are easily classified, and it has been questioned whether this approach is any better than messages to eat more fresh foods. It is of note, that, unlike with the author’s previous papers looking at cardiovascular disease risk in the Nutrient-Sante population which showed a link between the UK Food Standard’s Agency nutrient profiling and heart disease, which was also linked to ultra processed foods, with respect to type 2 diabetes and this new study, no analysis of the effect of nutrient profiling has been published and it is only used as a controlling factor in the analysis of this paper. This would be interesting as it would help guide if our current labelling approach is effective or if a moved to labelling foods according to level of profiling is better.
“It is also potentially concerning that the authors suggest effects of packaging which they even admit were banned part way through the study in the form of BPA (bisphenol -A) which were not assessed in the study and have been stated as being at safe levels in the UK diet according to European Food Safety Agency assessments. So, this was a speculative theory only. So, although this data supports current recommendations it makes a number of speculative suggestions for mechanisms which are not fully grounded in the current evidence.”
‘Ultraprocessed food consumption and risk of type 2 diabetes among participants of the NutriNet-Santé prospective cohort’ by Bernard Srour et al. was published in JAMA Internal Medicine at 16:00 UK time on Monday 16 December 2019.
Prof Gunter Kuhnle: “No CoIs.”
Prof Kevin McConway: “Prof McConway is a member of the SMC Advisory Committee, but his quote above is in his capacity as a professional statistician.”
Dr Duane Mellor: “I was funded by an EU 2020 Inclusilver grant to look at nutrient profiling within an online recipe application aimed to support people living with diabetes. Also I am working on a project with the out of home advertising industry to explore the effect of advertising on obesity, including how advertising can be used as a positive way to improve public health.”