select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
Scientist comment on a study, published in Diabetes, on UPF consumption, physical activity, genetic susceptibility and type 2 diabetes, which is being presented at the European Association for the Study of Diabetes (EASD) conference in Milan.
Dr Marie Spreckley, Weight Management Researcher, Prevention of Diabetes and Related Metabolic Disorders in High Risk Groups, Institute of Metabolic Science, University of Cambridge, said:
“In this study, researchers followed 56,964 UK Biobank participants for an average of 7.8 years, during which 926 developed type 2 diabetes. They used 24-hour dietary recalls, wearable devices to measure activity and linked health records to identify new cases. They also accounted for several other risk factors, and the findings were similar when they took overall diet quality into account. Those are strengths, although diet relied on participants’ recall and activity was measured for just one week.
“The headline is too strong, though. People whose diets had a smaller share of ultra-processed foods and who were more active were less likely to develop type 2 diabetes, including those at higher genetic risk. But we cannot say that changing those behaviours would cause the reductions reported here. Other differences between participants may have played a part.
“The ‘up to 70%’ figure compares two groups with intermediate genetic risk: people with a lower share of ultra-processed foods in their diets who met the activity target, and people with a higher share who did not. It does not mean an individual could cut their risk by 70%. The researchers found no clear evidence that the associations for the individual lifestyle factors differed by genetic risk. Their genetic analyses were also limited to people of European ancestry, making ‘regardless of genetic make-up’ too broad a claim.”
Dr Rosiered Brownson-Smith, Nutritional Sciences Researcher, King’s College London, said:
“This is an interesting and timely analysis of UK Biobank data, particularly because it considers ultra-processed food intake alongside objectively measured physical activity and genetic susceptibility to type 2 diabetes. The large sample size, prospective design and use of wearable-measured physical activity are strengths, and the study adds to a growing body of evidence linking higher UPF intake with increased risk of type 2 diabetes.
“Importantly, however, this is an observational study, so the findings show associations rather than cause and effect. We cannot conclude from these data that reducing ultra-processed food intake or increasing physical activity will directly reduce an individual’s risk of developing type 2 diabetes. This is particularly important when interpreting the headline claim that reducing UPF intake and being physically active can “cut” diabetes risk. I would therefore prefer language such as “was associated with a lower risk” rather than “can cut risk”. Observational studies are nevertheless extremely valuable for identifying patterns in large populations, which can then be investigated further using other types of research, including controlled intervention studies.
“The findings are broadly consistent with the existing epidemiological literature: higher UPF consumption has repeatedly been associated with a higher risk of type 2 diabetes, while greater physical activity is associated with lower risk. What is interesting here is the examination of these factors alongside genetic susceptibility. The study found that the associations between lower UPF intake and physical activity and lower diabetes risk were also observed among people with higher genetic susceptibility. This adds to the evidence that genetic predisposition does not completely determine an individual’s risk, and that potentially modifiable behaviours remain relevant across different levels of genetic risk.
“There are, however, some important complexities around the interpretation of ultra-processed food itself. UPF is a broad and heterogeneous category, and the study uses the NOVA classification, which categorises foods according to the extent and purpose of processing. Foods within this category can differ substantially in their nutritional composition, energy density, fibre content, additives, food structure and other characteristics. Therefore, an association between UPF intake and diabetes risk does not necessarily tell us that processing itself is responsible for the observed association.
“There are several possible explanations. The association could reflect characteristics of some UPFs themselves, particular ingredients or additives, overall dietary quality, the displacement of less processed and more nutrient-dense foods, or the tendency for some UPFs to be energy dense and relatively easy to consume quickly and in larger quantities. At present, we cannot confidently disentangle these possibilities from observational studies alone.
“The way UPF intake is measured is another consideration. Although this study uses repeated 24-hour dietary records, the underlying UK Biobank dietary data do not provide comprehensive brand-level information. This matters because products that appear similar can have quite different ingredients, nutrient profiles and formulations, and the same type of food can sometimes fall into different NOVA categories depending on its formulation. Consequently, it remains difficult to isolate the effect of food processing itself from the characteristics of the foods being consumed.
“It is also worth noting that the researchers did make efforts to account for potential confounding. Their fully adjusted analyses included demographic, socioeconomic, behavioural and clinical factors, and the findings remained broadly consistent after additional adjustment for overall diet quality, night-shift work, medication use and family history of diabetes. This strengthens confidence that the associations are not simply explained by a single obvious confounding factor, although residual confounding can never be completely excluded in an observational study.
“Overall, I would describe this as a useful addition to the evidence base rather than definitive evidence that changing UPF intake or physical activity will prevent type 2 diabetes. The most interesting finding is perhaps that associations with healthier lifestyle behaviours were observed across genetic risk groups. This highlights that having a higher inherited susceptibility does not mean that potentially modifiable lifestyle factors are irrelevant. However, the study should not be interpreted as showing that lifestyle changes can “overcome” genetic risk, nor that they eliminate the risk associated with a high UPF intake.”
Dr Beverley O’Hara, Lecturer in Public Health Nutrition, Leeds Beckett University, said:
“This study is based on observational data so it cannot demonstrate cause and effect. The researchers have investigated the relationship between levels of genetic risk, physical exercise, sleep and intake of ‘ultra processed’ food (UPF).
“The statement ‘higher UPF intake was linearly associated with incident type 2 diabetes risk’ is misleading. Firstly, it is specifically ‘ultraprocessed’ drinks that appear to be driving the observed association between UPF and incidence of diabetes (no other categories of UPF were associated with diabetes risk). Secondly, the size of the reported association is small and difficult to interpret within the context of peoples’ diets. The 7% is a relative risk increase, not an absolute one. Starting from a baseline diabetes risk of around 7% in the UK, each 10% increase in intake of ‘ultraprocessed’ drinks would raise that risk to roughly 7.5% — a small absolute change, even though “7% increase” sounds larger.
“The results of this study also show that UPF intake level was not associated with risk of diabetes across all genetic risk strata for individuals who meet physical activity guidelines, yet this finding is somewhat overlooked.
“The authors do not present detailed information about the methodology, so it is difficult to judge the overall quality of the science. The authors report that the dietary assessment methodology was based on 24-hr recall data. This dietary methodology was not designed to capture intake of UPF and is therefore at hight risk of misclassifying foods. In addition, food composition databases are not designed for processing-level food categorisation. These issues regarding classification and quantification of UPFs seriously limit the ability to draw robust conclusions from these data.
“The take-home message from this research has less to do with UPF and should in fact be about physical activity and sleep, which were measured more reliably than diet and showed greater effect sizes. An important finding is that physical activity appears to be substantially associated with a lower risk of developing diabetes; of significance is that the greatest difference in risk appears to be comparing those individuals who are inactive to those who are moderately active.
“The press release places undue emphasis on the association with ultra-processed foods, an area of considerable current interest, rather than on the findings most directly supported by the data (physical activity and sleep). This framing risks further amplifying an already contentious area of research whose scientific value has yet to be clearly established.
“In my view, the researchers have not given careful enough consideration of what exactly it is they are attempting to measure. The category of UPF is highly diverse and it is most unlikely that some of the foods that fall within this category mechanistically contribute to the development of type 2 diabetes. The majority of good quality studies on UPF have analysed sub-groups of UPF and have found that it is certain types of foods such as sugar-sweetened beverages, processed meats and snack foods that account for most of the negative effects on health that are observed. This effect is also observed in the current study.
“Using the NOVA category in nutrition studies is unhelpful in trying to understand the health effects of foods. Better methods exist, and could be further developed, to disentangle the effects of energy density, food texture, nutrient composition and overall diet quality on risk of developing diabetes and other noncommunicable diseases.”
Toby Lasserson, Deputy Editor in Chief of the Cochrane Library, said:
“This study looks at the relationship between consumption of ultraprocessed foods (UPFs) and risk of type 2 diabetes risk, and how things like physical activity, sleep and genetic susceptibility modify that relationship. The data come from the UK Biobank which contains health data from 500,000 people in the UK. The analyses shows that higher UPF intake was associated with an increase in diabetes risk, while higher levels of activity and adequate sleep were associated with lower risk.
“The main reason we should be cautious about these findings is that as an observational study this won’t prove causation. There are several factors that could explain the results other than a central relationship between UPF consumption and diabetes. Differences between the groups of people who consumed lower and higher amounts of UPFs from the start of the study include things gender balance, level of education, smoking status, how much they drank and their diet. Confounding is still likely even after the researchers’ attempts to adjust for these things in their analysis.
“The study is limited to people of European descent so we don’t know whether these findings might apply to other populations. As the authors themselves note in the paper, self-reported dietary recall is subject to recall bias. There is no analysis of nutritional content, so the study can tell us nothing about whether processing itself is a factor. It may be that high-UPF diets happen to be higher in sugar, for example, but this was not examined in the study.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, Open University, said:
“There are important limitations on what this study can show, and some of the wording in the press release is, in my view, very misleading. The release implies that changing UPF intake or physical activity levels can reduce the risk of type 2 diabetes. But in fact that study didn’t look at changes in UPF intake or physical activity levels, so it simply can’t tell us about the effects of such changes. The study itself did not draw any conclusions about changes in UPF intake of physical activity, because such changes weren’t measured.
“I can see how the estimates of changes in the press release were produced, and I don’t think they are reasonable measures of how much change there might be if people actually change their diet or activity levels. But the study just can’t tell us how big those changes might be.
“Also, this is an observational study, and as always there are some doubts about what causes what. Indeed the researchers rightly say in their research paper, “As with all observational studies, our findings cannot establish causality.” And, even if differences in diabetes risk are really caused by differences in UPF consumption, the differences in risk are really pretty small in my view, on the basis of the data in this study.
“The study uses data from about 57,000 participants in the UK Biobank cohort. Data from this cohort have been used quite extensively in studies of associations between UPF consumption, physical activity and the risk of various diseases. But the study is observational, and in this kind of study there are always doubts about exactly what causes what.
“The study found, for instance, that the risk of a new diagnosis of type 2 diabetes, during the follow-up period (of about 8 years on average) was higher in people who were recorded as having consumed greater quantities of UPFs, or having lower levels of physical activity.
“That could be because consuming more UPFs, or having lower physical activity, actually causes higher diabetes risk. But what was observed is an association, not cause and effect, and it could be the case that some other factors are independently associated with UPF consumption and diabetes risk, and that it is these other factors that are causing the increased diabetes risk, and not UPF consumption.
“The researchers, rightly, did take into account several other factors that might be coming into the picture, and the associations between UPF consumption, activity levels and type 2 diabetes risk were still there, albeit not so strongly. But this doesn’t get us off the hook of not being sure about cause and effect. The researchers may simply not have been able to take account of important factors, because they had no data on them.
“The research paper makes it clear what data was collected from the participants. Data on UPF consumption were collected early in the study, in some cases going back to when the participants were recruited in 2006-2010. That’s some time before the current UPF definition was put forward, and there are in any case issues about how accurate the measures of UPF consumption actually are. But a key point is that only one measure of UPF consumption was used for each participant, even if it was an average of more than one dietary assessment made for UK Biobank.
“Likewise, the researchers collected a number of measures of physical activity and of sleep characteristics on each participant. These measures, made between 2013 and 2015, were based on readings from a body-worn accelerometer (basically a more accurate version of what many people collect from things like Fitbits). But again, a key point is that the research used just one data point on each measure of activity and sleep, for each participant.
“(The conference abstract and poster do not report the results on sleep – not really an issue as one never has time for everything in a conference presentation.)
“The research also used a measure of the genetic susceptibility to type 2 diabetes. Problems are sometimes reported with such so-called polygenic risk scores, and rather few details are given in the paper of how the score was constructed in this case. But it’s certainly potentially useful to see whether the associations between UPF consumption, physical activity, and so on, on the one hand, and type 2 diabetes risk on the other hand are different in people with different genetic susceptibility levels. The researchers found that the general pattern of higher diabetes risk in people who consumed more UPFs, or who had less physical activity, was there at all levels of the genetic score.
“But, if I haven’t confused you too much already, you’ll have noticed that each participant provided only one UPF consumption measure, one data point for each of the physical activity and sleep measures, and obviously one genetic risk score. So the study simply couldn’t look at how changes in UPF consumption or the activity measures might affect the risk of a new type 2 diabetes diagnosis during follow-up.
“The top line of the press release, however, says “Reducing ultra-processed food (UPF) intake and being physically active can cut risk of type 2 diabetes, regardless of genetic make-up.” It’s talking about what would happen if people changed their UPF intake and physical activity levels. But the study couldn’t consider such changes because it didn’t measure them.
“What is actually being done to produce these measures of change is something different. The study measured, for instance, the risk of type 2 diabetes during follow-up in people consuming different amounts of UPFs. And then the press release is making the assumption that, if someone changes their UPF consumption to a given lower level, their type 2 diabetes risk will fall to the level that people who had that lower level of UPF consumption at the start of the study would have.
“The trouble is that it seems to me very unlikely that changes in potential risk factors like UPF consumption would work like that, even if the differences in diabetes risk are actually caused by the UPF consumption (and as I’ve explained, we can’t be confident that they are caused in that way).
“Surely, for instance, someone who consumed high levels of UPFs when they entered the study, kept on consuming at that level till close to the end of follow-up, but then suddenly reduced their UPF intake, would not have much change in their diabetes risk during follow-up. Things are more complicated – just as if someone gives up smoking, their lung cancer risk will reduce, but not instantly to the level of someone who never smoked, or someone who gave up decades ago.
“In my view some of the wording in the rest of the release is too causal. It does acknowledge that “the findings are observational and so cannot prove cause and effect”, but that means that language like “can cut the risk” and “seems to reduce” is too definite, and doesn’t fully take into account the inevitable limitations of the study that was actually done.
“How big is the change in risk anyway? Well, not very big. In this study, out of a group of 1,000 participants, about 16 had a new diagnosis of type 2 diabetes during the average 8-year follow-up. Now consider a group of 1,000 people, similar to first group in every respect except that each of them had a diet containing a level of UPF consumption that’s lower by 10% as a proportion of their total diet. (I mean, for example, that someone whose diet consisted of 40% of UPFs in the first group of 1,000 is replaced by someone whose diet had 30% of UPFs.) Then the number who would have a new type 2 diabetes diagnosis during follow-up would be about 15 – and there’s some statistical uncertainty about that figure. Not a big change at all.”
Prof Alberto Fiore, Professor in Food Chemistry and Technology, Abertay University, said:
“This study did not test whether reducing ultra-processed foods prevents diabetes. It averaged several self-reported 24-hour dietary records, placed many nutritionally different foods into a single NOVA category, and related that average to later diabetes diagnoses. This approach mixes sugary drinks and nutritionally poor snacks with fortified cereals, wholegrain breads, plant-based dairy products and other processed foods that can supply fibre, protein and micronutrients. The overall average therefore tells us very little about whether processing itself matters.
“Crucially, when the authors separated the UPF category into different types of food, ultra-processed drinks were the only subgroup that remained significantly associated with diabetes after full adjustment. We have long known that frequent consumption of sugary drinks is a marker and contributor to poor-quality diets associated with diabetes and other non-communicable diseases. Re-labelling these drinks as “UPFs” does not demonstrate that industrial processing is the cause, nor is it a major new discovery.
“The relevant public-health message is about diet quality, not avoiding everything classified as ultra-processed. Some NOVA category 4 foods – including fortified wholegrain cereals, wholegrain breads, plant-based milk and yoghurt – can contribute fibre and essential nutrients and may help people follow healthier diets. Separate research involving more than 200,000 US adults found that overall diet quality was associated with the risk of developing type 2 diabetes or cardiovascular disease and with the risk of death, while these associations did not substantially differ according to food-processing level.
“While physical activity was measured more objectively (recorded objectively with a wrist accelerometer for seven days) and had a stronger association with diabetes risk, it remains observational and cannot prove causation.
“The paper is a large, peer-reviewed prospective cohort study with careful statistical adjustment and useful sensitivity analyses. However, it remains observational, does not measure dietary change and cannot establish that reducing UPFs would reduce diabetes risk. Its data more convincingly reinforce the familiar association between poor dietary quality – particularly drinks with unfavourable nutritional profiles – and diabetes than they establish NOVA processing level as an independent cause. The headline stating that reducing UPF intake can cut diabetes risk is misleadingly causal.”
Seamus Higgins, Associate Professor in Food Process Engineering at the University of Nottingham, said:
“This is a substantial prospective study of nearly 57,000 UK Biobank participants followed for almost eight years, strengthened by repeated dietary records, objectively measured physical activity and sleep, genetic data, and extensive adjustment for potential confounders. The findings add to growing evidence that higher consumption of ultra-processed foods (UPF), classified using NOVA, is associated with an increased risk of developing type 2 diabetes, with each 10% higher UPF intake associated with a 7% higher risk.
“Importantly, the findings remained consistent after adjustment for overall diet quality and across several sensitivity analyses. However, NOVA covers a diverse range of foods and, when individual UPF categories were examined, ultra-processed drinks were the only subgroup that remained significantly associated with diabetes after full adjustment.
“Overall, this is a well-conducted and useful addition to the evidence linking higher UPF consumption with type 2 diabetes, but the appropriate language remains “associated with” rather than claiming that reducing UPF has been proven to directly reduce diabetes risk.”
Paper: ‘Associations of Ultraprocessed Food Intake, Wearable-Measured Lifestyle Factors, and Genetic Susceptibility With Incident Type 2 Diabetes’ by Wenxin Xu was published in Diabetes. DOI: 10.2337/db26-0395
Abstract: ‘Joint associations of ultra-processed food, wearable-measured physical activity and genetic susceptibility with incident type 2 diabetes’, to be presented at the European Association for the Study of Diabetes (EASD) in Milan..
Declared interests
Dr Marie Spreckley: “Dr Marie Spreckley is a postdoctoral researcher at IMS Epidemiology, University of Cambridge, where her research focuses on nutrition and weight management. She has no financial relationships with food or beverage companies, pharmaceutical companies, or other commercial organisations, and has received no industry funding for her research.”
Dr Rosiered Brownson-Smith: “None.”
Dr Beverley O’Hara: “I have no conflicts of interests to declare.”
Toby Lasserson: “No COIs.”
Prof Kevin McConway: “I have no conflicts of interest to declare.”
Prof Alberto Fiore: “I do not have any conflict of interest and I do not get any payment from any source except the one from my university.”
Seamus Higgins: “Previously held technical and leadership roles within the food industry, including with Nestlé and Premier Foods. Since joining the University of Nottingham approximately ten years ago, he has had no employment, consultancy, advisory, financial, or other commercial relationships with these companies and reports no competing interests in relation to this article.”