select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
A randomised crossover trial published in Nature Medicine looks at ultra- vs minimally-processed foods and weight loss.
Dr Dimitrios Koutoukidis, Associate Professor in Diet, Obesity, and Behavioural Sciences, and dietitian, University of Oxford, said:
“In this study, participants followed a diet high in ultra-processed or minimally-processed foods. Both groups lost a significant amount of weight, but the researchers were originally expecting the UPF group to gain weight. The researchers do not speculate on why people on a UPF diet lost weight, but it might be that having all foods free, ready-prepared, and provided, as in this study, makes people more conscious about the food they eat and reduces the number of decisions they have to make. This can allow them to follow a diet aligned with national UK guidelines.
“There was no difference between groups on most markers of heart health. Triglycerides were reduced more in the minimally-processed diet but, surprisingly, LDL cholesterol was reduced more on the ultra-processed diet. This might imply that processing is not as important for heart health if the foods already meet the standard UK healthy eating guidance. Further research is needed to better understand this.
“We know that receiving support to adhere to the diet is one of the strongest predictors of diet adherence and weight loss. In this study, participants had weekly phone calls with the researchers, but the researchers were not blinded. This might have inadvertently introduced some bias, whereby those on the minimally processed diet got more help and support than when on the ultra-processed diet.
“One of the main assumptions of the cross-over trial design employed here is that the effect of diet of the first period should not influence the outcome of the diet of the second period. The researchers found that this assumption was not met, which makes the estimated weight change biased. They have done additional analyses that suggest this concern is addressed. In this scenario, it would have been good practice to report only the weight loss of the diet of the first period. Future studies should carefully consider their design to minimise this risk.”
Prof Keith Frayn, Emeritus Professor of Human Metabolism, University of Oxford, said:
“It comes as no surprise that minimally processed foods help weight loss compared with ultra-processed foods, but it is good to have rigorous experimental evidence for this. The explanation given for the advantage of the minimally processed foods is that they provide lower energy density (fewer calories per 100 grams), which accords with what is known from other studies. The supplementary data provided with the paper show how this comes about, with the ultra-processed foods providing considerably more free sugar, although not more fat as we might have expected. The ultra-processed foods also provided less protein and more salt, making them less healthy, and perhaps helping to explain why participants ate more, since protein helps with satiety (fullness). One surprising finding was that fibre intake was somewhat greater on the ultra-processed foods than with the minimally-processed, which goes against the prevailing view, but then this study used ultra-processed foods that were designed to adhere to government nutritional advice, and were not necessarily typical of what consumers would usually purchase. Overall, the results of this study confirm that ultra-processed foods are not helpful components of a healthy diet, but they do not provide support for the idea that the processing itself is doing anything other than making them more palatable and a richer, and less filling, source of energy.”
Dr Nerys Astbury, Nutrition Scientist and Associate Professor in the Health Behaviours team, University of Oxford, said:
“Many studies have examined the association between ultra-processed food consumption and numerous health outcomes, but trials assessing consuming diets high in UPF are limited.
“This new study by Dicken et al aims to compare the effects of eating diets that adhered to the Eat-well guide recommendations made-up of minimally processed foods with comparable diets made up of mainly ultra-processed foods. The findings should be interpreted with some caution.
“Unfortunately, the trial has a critical methodological flaw, as it uses a crossover-design, which is widely acknowledged to limit the validity of dietary trials. Crossover-trials use individual participants as their own controls – thus minimising between-participant differences e.g. gender, age, ethnicity, body composition etc. which could impact the overall findings. Because of this they typically require fewer participants to be able to detect an effect than parallel group clinical trial designs. However, because the human body takes time to adapt to changes in diet, exercise sleep etc., and this sort of adaptation takes time, usually several weeks or longer, short-term diet trials with short intervention and wash-out periods, like in this study by Dicken et al can produce invalid findings when they directly compare (within-participant) the intervention/experimental and comparator periods. That’s because the order that participants receive the different diets can affect the overall results.
“Dicken at al did examine the possible order-effects, and indeed clearly show that there are significant order-effects in this study. This means the order in which participants received the MPF and UPF diets did affect the findings. In the presence of these order effects, any effect in of the diet received by the participants second is likely to be invalid.
“There are many secondary/exploratory outcomes reported, and the authors have not made a correction for multiple comparisons, which is crucial to control for false positive results (statistically significant results where no real effect exists), the risk of which increases when multiple statistical tests are performed simultaneously. Many of the findings are reported within diet groups – as opposed to comparing the difference between the diet groups.
“Aside from these methodological limitations, the overall findings are thought-provoking. The most notable is the effect on weight loss. The authors emphasise that there is more weight loss in the MPF diet – but participants actually lost weight in both diet conditions. This provides evidence to refute the argument that diets high in UPF are hyper-palatable which causes over consumption and weight gain, driving the obesity epidemic.
“Whilst the authors argue that the 8-week duration is a key strength of this study, this is a relatively short period of time. To properly examine possible health impact of consuming diets containing UPF, much longer methodologically rigorous trials in free-living conditions are needed – where the effects of consuming diets containing variable amounts of UPF can be compared in real world settings, outside the controlled environment of a lab or clinic.”
Prof Alex Johnstone, Theme Lead for Nutrition, Obesity and Disease, the Rowett Institute, University of Aberdeen, said:
“This study is a robust analysis and welcomed as providing novel data from a controlled feeding study, within the UK, with two similar published RCT conducted in USA and Japan. This study was powered to examine percentage body mass (body weight) changes to compare two ad libitum weight loss diets, as UPF and MPF diets, each for 8 weeks, conducted as a randomised cross-over. Diets were provided to participants to support compliance. The study supports the current thinking that there was a significantly greater weight loss on the minimally processed diet. This was reflected with a greater reduction in fat mass, less food cravings and a lowered triglycerides measured in blood. The authors suggest nutritional differences and behavioural differences for greater weight loss on the minimally processed diet regime. Interestingly, there was a significant weight loss with the UPF group, which also met current UK Healthy Eating guidelines, suggesting that there is benefit of reducing UPF during dieting. A couple of limitations is the use of bio-impedance for the body composition assessment, which is not gold-standard approach (in contrast to dual energy x-ray absorptiometry/DXA and 4-compartment model approaches); also that energy balance was not directly measured beyond body weight and changes in body composition, assessment with energy expenditure by application of doubly-labelled water/DLW is welcomed in future mechanistic studies. The paper is good quality research and has useful discussion about the real-world significance.”
Tracy Parker, Nutrition Lead, British Heart Foundation (BHF), said:
“These findings support what we have long suspected – that the way food is made might affect our health, not just the nutrients it contains.
“The way this study was designed means it is more reflective of real-world conditions than previous research on ultra-processed foods. Unlike earlier observational studies, this was a randomized controlled trial where participants were provided with all their meals, and the diets were carefully matched to meet the Eatwell Guide – the UK’s national dietary guidelines. This allowed researchers to isolate the effect of food processing itself, making it more likely that the differences seen after eight weeks were due to how the food in their diets was processed, not just what was in it.
“The small size of the study is a limitation, and the fact that most participants were women limits how much we can generalise the findings to the general population. We also can’t be certain how closely the diets were followed, as participants self-reported what they ate during the study.
“Larger, longer-term studies will be needed to see if the greater weight loss on the minimally processed diets seen here translates into bigger improvements in risk factors, including blood pressure and cholesterol and blood sugar levels, and a reduced risk of developing heart and circulatory diseases.
“Completely cutting UPFs out of our diets isn’t realistic for most of us. But including more minimally processed foods – like fresh or home cooked meals – alongside a balanced diet could offer added benefits too. Mediterranean-style diets, which include plenty of minimally or unprocessed foods such as fruit, vegetables, fish, nuts and seeds, beans, lentils and wholegrains, have consistently been shown to reduce our risk of heart attacks and strokes.”
Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:
“This is a very well conducted and detailed study, which contributes to the actual body of evidence we have regarding ultra-processed foods. The results confirm results from previous studies that show that diets rich in ultra-processed foods can lead to a higher energy intake when compared to minimally processed foods.
“The methodology of the study is sound, although it suffers from problems inherent to most nutritional studies. Unless study participants are housed within a research centre, researchers have to rely on participants to follow interventions and provide reliable data. This is already difficult in drug trials – where only about 80% of participants adhere to intervention and even more difficult in nutrition studies where interventions are often more complex. The authors acknowledge and address this, but it is possible that they under- or overestimate the effect size.
“A limitation of this study in the context of the discussion about ultra-processed foods is that there is no direct comparison between ultra-processed and processed food; this is quite relevant as many discussions around UPF focus on this difference and not the difference between ultra-processed and minimally processed foods.
“The most interesting result of the study is that participants on both arms lost weight – which contradicts claims that ultra-processed foods result in weight gain. Indeed, the study suggest that a diet meeting current dietary recommendations is not detrimental to weight maintenance, whether it is ultra-processed or not. This is in particular relevant as the study compared ultra-processed with minimally processed foods, where the effect size can be expected to be larger than in a comparison between processed and ultra-processed food. It is difficult to reconcile these findings with the claims in the UCL press release that ultra-processed foods are detrimental to health.
“While the study was not designed to detect changes in other disease markers, the lack of obvious differences suggests that a diet that follows current recommendations is not detrimental to health – whether ultra-processed or not.
“In summary, the study confirms a large consensus among nutrition and food scientists that ultra-processed foods are not inherently unhealthy and can be an important part of a healthy and balanced diet.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“I’m going to be critical about some of the conclusions from this research later in this comment, but I want to start by pointing out an important strength. The strength is that it’s a randomised trial. The vast majority of research on possible health effects of consuming ultra-processed foods (UPF) has been carried out by observational studies of various kinds. One can’t come to conclusions about cause and effect from single observational studies of that kind. It might eventually be possible to come to such a conclusion from a rather large number of observational studies pointing in the same direction, in detail, together with research of other kinds (in cell preparations, possibly in animal studies, and so on), but I don’t think we’re at that stage yet with UPFs. Randomised studies are generally much stronger than observational studies in dealing with cause and effect.
“This isn’t the first randomised study in humans comparing health effects of consumption of UPFs compared to other types of food. But I know of only two others, by Hall et al. and by Hamano et al. (references 20 and 21 in this research paper). Both were carried out under tightly supervised conditions, with the participants being kept on metabolic wards during the studies, so it’s a little difficult to say exactly how they relate to what people eat under more normal circumstances. Both involved small numbers of participants (20 for the Hall study, just 9 for the Hamano study). Neither really could make any attempt to be pragmatic, in the sense of having direct and immediate applicability to everyday conditions. (I’m not saying they were bad studies, just that they had some important limitations.)
“The new trial had more participants than either of the earlier trials (55 to begin with, though a few dropped out, as usually and unavoidably happens in such trials). And because the participants were living in their usual communities, and because the different diets they ate (while not their usual fare) were based on existing UK public health advice on what to eat, the conditions were much more like what might happen in the real world. That’s all good.
“I should say that I’m confident that the trial did show a real difference between diets in terms of weight loss, over the first 8 weeks when the participants were on one of the experimental diets. That’s a new finding. But, as I’ll explain, we can’t be sure the difference is due to the fact that one diet contained quite a lot of UPFs and the other contained, instead, minimally processed foods (MPF). That might be the explanation, but it might not. Also it becomes less clear what’s going on after that first eight-week period.
“However, there are some important limitations, and in my view they go beyond the limitations mentioned by the researchers in their paper. The researchers found that, on average, participants lost more weight in the periods when they were on a diet based on minimally processed foods, than when they were on a diet based on UPFs. While it’s encouraging that, on average, the participants did lose some weight on both of the diets, it’s difficult to see how this really means that in general a diet based on less food processing is better for weight loss.
“That’s because only two different diets were used. As this is a randomised trial, there’s good evidence that the average weight loss was larger on this particular MPF diet than on this particular UPF diet (on the first 8-week period at any rate – more on that below), and that this difference was caused by the difference in diets. But because there was only one diet of each kind, it gets difficult to generalise this conclusion to all possible MPF and UPF diets that fit in with the national dietary advice. There might be something specific about the diets that were used in this study, that isn’t necessarily connected with the degree of processing of the foods and drinks in them. The results might have been different with other diets, that still differed in how much processing their ingredients had had. Or maybe the results would have been similar with all reasonable diets with these different levels of processing. The point is, that because the researchers didn’t examine findings on other diets, we just can’t tell anything about this. More studies, and a lot of them (costing a lot of time and money) would be needed to learn enough about this aspect.
“It’s worth mentioning, I think, that there were differences between the two diets that went beyond whether or not they contained UPFs. For example, they differed in the way they were prepared.
“The items in the UPF diet were ‘sourced from leading UK supermarkets’ and ‘were not culinary preparations’. That would indeed help to ensure that the UPF diet in this study was ‘representative of UPF available in the UK’, as the researchers state.
“However, the MPF diets were ‘culinary preparations of individual ingredients’, as the research paper puts it. That is, they were somehow prepared in a kitchen, under instructions from the researchers, that wasn’t the kind of place where UPF food is produced. The researchers say they did this, rather than using shop-bought items, to be sure that the classification as being minimally processed was correct. But people in real life (not in a study) who are trying to use minimally processed foods outside a study like this would probably buy lots of the minimally processed ingredients themselves, and cook them at home, rather than having them all delivered from another kitchen as happened in this study. Or they might also buy some prepared foods from supermarkets, being careful to ensure as far as possible that they are actually minimally processed. So it doesn’t sound to me as if the MPF diet in this study was necessarily very representative of what people, avoiding everything but MPF, would eat in the UK.
“That feeds through to an aspect the researchers comment on, that in this study the UPFs were provided in their normal branded packaging, which could have affected the portion sizes that people chose, or indeed their feelings towards the UPFs. This, and aspects of the fact that the participants didn’t obtain raw materials for the MPF diet, or cook the food, themselves, might somehow have been related to the finding that participants gave lower flavour and taste ratings to the MPF diet than the UF diet.
“So is this study a fair test of UPF diets and MPF diets in general, as they would be eaten in the UK by people sticking to the government Eatwell diet recommendations? I don’t think so, and that means we have to be careful in deciding what the results mean, and the findings shouldn’t be generalised too far. In this respect, it’s also important to note that this trial can’t tell us much about which aspects of UPFs might have led to the smaller average weight losses. Is it something about all, or most, UPFs, or only about a few of them? Again we can’t say. This trial is a start in finding out, but only a start.
“I’ll now turn to another, rather more statistically technical, issue that concerns me about this study. Why I think this is important is that it casts huge doubt on the estimates of weight loss over the course of a year, that are given in the paper and the UCL press release. To be fair, those annual weight loss estimates aren’t a principal outcome of the study – they are referred to as ‘Exploratory outcomes’. But they are given some prominence in the UCL press release. To explain fully what’s going on needs a complicated explanation – I’ll spare you most of the details (though I’ve put some of them in a Technical Statistical Note below, in case you’re interested).
“I think there’s a reasonably simple explanation based on Extended Data Figure 1, which appears near the end of the copy of the paper that was circulated to journalists. Frame (a) of this Figure shows that there was a large difference between the average weight losses on the two diets in the first period (labelled ‘First diet’), but only a very small difference, well within the limits of random statistical variability, in the second period. Overall this diagram shows that, in the first period, participants lost more of their bodyweight if they happened to be on the MPF diet than if they happened to be on the UPF diet. But there’s hardly any difference between the weight losses on the two diets for people that were on them in the second period.
“The calculations of what the weight losses might be on the two diets over a whole year, given in the paper and the UCL press release (stated to be 13 percentage points for men and 9 percentage points for women on the MPF diet, compared to 4 percentage points for men and 5 percentage point for women on the UPF diet) is, I believe, based on the average losses over the two 8-week periods taken together. I don’t think that makes much sense. If people kept up these diets for a whole year, would we see the weight losses falling off considerably over time, and the difference between the losses also falling off over time, as in in Extended Data Figure 1? I think we probably would. Or would we see everything averaging out, to give the annual weight losses stated in the paper, even though it clearly doesn’t all average out nicely over just two 8-week periods? I can’t envisage that happening over an entire year.
“And would we see the difference between the average weight losses on the two diets be like the average of the two eight-week periods shown in Extended Data Figure 1, so about 1 percentage point for the whole of a year? Or would it be considerably larger than that for the first eight weeks or so, and then close to no difference, possibly for the whole rest of the year? If the former, yes, we’d maybe get annual figures like those given in the paper and the press release – but if the annual pattern is more like a continuation of Extended Data Figure 1, we wouldn’t get such big estimates of annual weight loss, and we’d get a considerably smaller difference between the annual weight losses on the two diets.
Technical statistical note
“The trial had what’s called a two-period crossover design. This jargon means that each participant actually ate both the MPF and the UPF diet for two separate periods (of 8 weeks in this study), with a four-week gap in between the two periods of 8 weeks. Ideally this four-week gap (the so-called washout period, in which people were asked to return to their usual diet) would mean that, by the start of their second 8-week period on the researcher-imposed diet, the specific effect of their diet in the first 8-week period would have worn off, more or less.
“The other two randomised studies that I mentioned earlier (Hall et al. and Hamano et al.) also used two-period crossover designs, though with no washout period at all in the Hall trial. If a crossover trial is appropriate and works well, it gets over an issue in all randomised trials. This issue is that individual participants can differ considerably in their responses to the randomised treatments, just because people aren’t all the same. In a crossover trial like this one, each participant’s weight loss in their period on the UPF diet is compared directly to their weight loss on the MPF diet, and then these comparisons for different people are put together to produce the estimate of a difference between the diets. Differences between individuals effectively cancel out. This means you get more statistical information from a given number of participants, than you’d get in a trial where one group had only the UPF diet and another group had only the MPF diet.
“If everyone gets both diets, where does the randomisation come in? Well, it can happen that everyone’s responses differ, on average, between the first diet they have and the second diet they have, regardless of whether they had the UPF or the MPF diet first. So the random part is in allocating participants to their first diet. Some had MPF first and then UPF, and some had UPF first and then MPF.
“This randomisation allows the researchers to estimate the difference between average weight loss on the two diets, regardless of which one they had first, and also the difference between the average weight loss in the two treatment periods, regardless of which diet they had in which period. And the researchers made those estimates.
“Where things get awkward is if the difference between the weight losses on the two diets is not the same in the second treatment as in the first treatment period. (In the jargon, this is called a treatment by period interaction, and they are trouble.)
“There are all kinds of reasons why this might happen. One is that the washout period is not long enough, so that there are lingering effects of the first diet into the second treatment period, and that those lingering effects are different depending on which diet they had in the first period. But that’s only one possibility – there are others, and in this kind of crossover design there’s no real way of telling what the explanation is.
“In this trial, whatever caused it, there was indeed a treatment by period interaction. That’s easiest to see in Extended Data Figure 1. As I’ve explained, overall this diagram shows that, in the first period, participants lost more of their bodyweight if they were on the MPF diet than if they were on the UPF diet. But there’s very little difference between the weight losses on the two diets for people that were on them in the second period. (The diagram also shows that, whatever diet people were on, the average weight loss in the second period was less than in the first period. But that’s not what complicates the interpretation – it’s that the difference between the losses on the two diets is very different in the second period to what it was in the first period.)
“Because the difference between the weight losses on the two diets is so different between the two periods, I’d argue strongly that it makes little sense to average out those two differences to give an overall average across the two periods. It would be better to consider the two periods separately. Indeed the researchers do that, though I’d say that they don’t report all the results clearly in the main research paper.
“They report (in the Primary Outcome part of the Results section) that, when they just used the data from the first period, the difference between the percentage weight loss on the MPF diet and the UPF diet was 1.86 percentage points, with a larger weight loss on the MPF diet on average, and that this was a bigger difference than what you’d expect to see simply through random variability.
“What they don’t explicitly report on the main paper, and I can’t even find it in the Supplementary Tables (apologies if I missed it somewhere), but which can be seen in Extended Data Figure 1, is that the difference between the weight losses on the two diets based only on data from the second period is very much smaller, probably only somewhere around a tenth of a percentage point. The difference between the weight losses on the two diets, using only data from the second period, is almost certainly too small to rule out the possibility that it’s just due to random statistical variability. You can tell that from the so-called ‘error bars’ in the diagram.
“Averaging out the first-period difference of 1.86 percentage points greater weight loss on the MPF diet, and the second-period difference of almost zero between the average weight losses on the two diets, to give an overall difference in average weight losses of about 1 percentage point, makes little sense in my view. But those predictions of annual weight loss over a whole year appear to depend exactly on this sort of averaging.”
‘Ultraprocessed or minimally processed diets following healthy dietary guidelines on weight and cardiometabolic health: a randomized, crossover trial’ by Samuel J. Dicken et al. will be published in Nature Medicine at 16:00 UK time on Monday 4 August 2025, which is when the embargo will lift.
DOI: 10.1038/s41591-025-03842-0
Declared interests
Dr Dimitrios Koutoukidis: “I have led publicly-funded investigator-led clinical trials where the weight loss intervention has been provided free of charge to participants by Oviva, Nestle Health Science, and Second Nature. I have not received any personal benefits from these associations.”
Prof Keith Frayn: “I have no commercial interests related to this topic.
I am the author of books on metabolism and energy balance including ‘A Calorie is a Calorie’, Piatkus, Jan 2025.”
Dr Nerys Astbury: “No conflicts.”
Prof Alex Johnstone: “Voluntary committee member for the Nutrition Society, The Nutrition Society and British Nutrition Foundation.
Current consultancy with Horizon European Committee as Committee Chair in Agri-Food Systems review panel; Scottish catering company for nutrition analysis.”
Tracy Parker: “One of the co-authors (Mark Hamer) has funding from BHF for a separate study (https://www.propassconsortium.org/), which is mentioned in the acknowledgements.”
Prof Gunter Kuhnle:
“Employment:
– University of Reading, Professor of Nutrition and Food Science,
– University of Reading, Director of the Chemical Analysis Facility (includes a range of industry customers).
Member of the Advisory Committee on Novel Foods & Processes (ACNFP), Scientific Advisory Committee of the BNF,
– former member of the Committee on Toxicity (COT), EFSA Panel on Additives and Nutrient Sources.
Grant income:
– Current:
– UKRI BBSRC Transforming UK Food System (TUKFS) “FoodSequal”.
– Previous (before 2020):
– Mars, Inc. – research on the health effect of flavanols.
Other:
– Family owned vineyard.”
Prof Kevin McConway: “I have no conflicts of interest.”
This Roundup was accompanied by an SMC Briefing.