A study published in JAMA Pediatrics looks at childhood consumption of ultra-processed foods (UPFs) and body fat trajectories.
Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:
“The results of this study are not surprising: children who consume a lot of “ultra-processed” foods are most likely to be less healthy and more obese than their peers with lower intake. The interpretation of these results are however much more difficult. Apart from the limitations of the definition of “ultra-processed foods”, the outcome of the study is heavily confounded by socio-economic factors: children living in more deprived areas and from families with lower educational attainment and lower socio-economic status had the highest intake of ultra-processed foods. Unfortunately, these children are also at highest risk of obesity and poor health, as there are still considerable health-inequalities in the UK and socio-economic status is an important determinant of health.
“The link between ultra-processed food consumption and child-health is therefore likely more complex and needs a more holistic approach: neither a reformulation nor a ban of ultra-processed foods will address health-inequalities. This can only be achieved by changing food systems and addressing the underlying causes of health inequalities.”
Dr Duane Mellor, Registered dietitian and Lead for nutrition and evidence based medicine, Aston Medical School, Aston University, said:
“This is a large study looking at how over 9000 children in the Bristol area grew up from the age of 7 until their mid-twenties. The particular aim of this analysis was to see if consumption of ultra-processed foods is associated with body weight and fat on average 10 years later.
“This study only used information from a single 3 day diet diary (which not all children and their parents or carers completed all 3 days) when they were 7 years old to estimate their intake of ultra-processed foods. Although the researchers said this did not vary by more than 20% in most cases at the age of 10 and 13 when the diet diary were repeated, there was not enough complete data for the researchers to look at how diet habits changed over time properly. So, it is perhaps a big assumption that diet eaten at the age of 7 years is the same eaten as a teenager and into their twenties. The bigger study from which this analysis was done started with some of the children’s parents whilst they were pregnant, so some of the children included may have been included in the study for longer than others. Data from these children suggest that the diet of children tends to become healthier after the age of 7, so it would be interesting to know how much ultra-processed foods those children eating a lot of this type of foods at 7 ate when they were toddlers. Also, it is unclear if length of time each child had been taking part in this study by the time they were 7, was accounted for in this analysis. This could be important as being in a study like this could potentially influence how someone behaves and reports their behaviour, as it is known filling in a food diary for example can change how someone eats.
“Another challenge is the classification of ultra-processed food itself. This study used the NOVA classification which classifies foods into 4 groups from: unprocessed and minimally processed foods such as fresh vegetables, fruit, dairy and meat etc.; processed culinary ingredients such as butter, oils, sugar and honey; processed foods such as cheese, canned vegetables, fresh bread and beer; and ultra-processed foods which include sugar sweetened soft drinks, chocolates, ice cream, biscuits, packaged bread, breakfast cereals and jars of pasta sauce. So, it is clear that the ultra-processed foods classification includes not just most of the foods which most dietary guidelines call discretionary and unnecessary as part of a healthy diet – the ones which are to be eaten infrequently and in small amounts along with things like a jar of pasta sauce. This is a key criticism of the NOVA classification of ultra-processed foods, that it does not just includes foods which we know are not good for our health, it can include others like the pasta sauce which may encourage a child to eat more vegetables and other healthy (less processed) foods.
“So, the question is does classifying someone’s diet provide more information than looking at a diet score assessing how healthy it is such as the healthy eating index or Mediterranean Diet score, probably not. Equally, the ultra-processed food label probably give little more information than looking at intake of discretionary/ non-recommended foods or foods with added fats, sugar and salt. Although, the researchers did try and account for total saturated fat and sugar, they did not look at how they were used in the food. So, although the evidence of an association between ultra-processed food and body weight and fat was seen in this study, it is not a causal one and it does not consider any potential effects of nutrients consumed or overall diet quality.
“Overall, this study risks suggesting that all foods which are processed are bad, whereas this is probably only really the case when they are higher in fat, salt and sugar and lower in fibre. Also it makes a lot of assumptions about the 3 days of diet diary at the age of 7 and its association with the child’s growth and weight gain for the next decade.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“I did a search just now on PubMed.gov, the website run by the US National Library of Medicine that is probably the most comprehensive database of scientific papers in medicine. It turned up almost 300 publications about ultra-processed foods, over the period since about 2009 when the concept was first introduced. They aren’t all research publications – some are defining or describing the classification of foods into categories including ultra-processed, some are editorial comment, some look more like campaigning articles than real science, and a few are critical of the whole notion of ultra-processed food. But most are about research findings, and they generally report worse health outcomes of various kinds in people who consume more ultra-processed foods, or at least worse outcomes in terms of things like body fat or weight that might lead to worse health. So why do we need another study like this one? Well, in my view, this study does add to what was previously known about associations between consumption of ultra-processed foods in children and young people and their patterns of body fat. In statistical terms, what they did is sound. That said, like all these studies, this new one needs some careful interpretation, and there are some important and inevitable limitations that arise from the nature of the study.
“What seems to be new about this study is as follows. It followed up around 9,000 children in south-west England over an average period of 10 years, and during that time, measures were taken of how certain body measurements were changing. The study concentrated on adiposity, that is, on aspects of the fat content of their bodies. One of the measures they used was BMI (body mass index), which measures people’s total body weight in relation to their height. That can be affected by the amount of non-fat tissues such as muscle mass, because it all contributes to total weight. So the researchers also looked at other measures such as FMI, the fat mass index, which is calculated in much the same way as BMI but using a measurement of the weight of fat in the person’s body, not the total weight, in relation to their height. The researchers than looked at how these measures changed in these children over the years as they grew up into young people, and related that to the proportion of the weight of their diet that consisted of ultra-processed foods. Previous research has looked at similar things in children and young people, but involving fewer participants and generally for shorter follow-up periods, and the results were not always clear-cut.
“The new study found that BMI, and some other measures of body fat content, increased faster over the years in children whose diet contained a greater percentage of ultra-processed foods at the study baseline (which, for most of the participants, was when they were 7 years old). Many of their comparisons are between those whose proportion of ultra-processed foods at age 7 was in the lowest fifth of all the participants, and those who were in the highest fifth. There were really quite large differences in the consumption of ultra-processed foods between those two groups – for the lowest fifth, under a quarter of the weight of their diets consisted of ultra-processed foods, but for the highest fifth, it was over two-thirds. At age 7, the participants’ BMI varied very little with their consumption of ultra-processed foods. Then, across all the participants, BMI increased by, on average, 0.55 per year of age. But it increased rather faster in participants who consumed the greatest proportion of ultra-processed foods at age 7, compared to those who ate the smallest proportion. The difference isn’t huge; average BMI increased by only 0.06 per year more in the highest consumption group than the lowest. But that builds up over time, so that by age 24, those who consumed a high proportion of ultra-processed foods when they were 7 had an average BMI over one point higher than those who consumed a low proportion. There was a roughly similar pattern with the fat mass index (FMI). The picture was different with total body fat as a percentage of total weight – that was already higher at age 9 (when its measurements started) in those that ate higher proportions of ultra-processed food, and those differences persisted over the follow-up period but did not change much over time. But with the lean mass index (LMI), that relates the body weight excluding fat to the person’s height, there were no clear differences related to the consumption of ultra-processed foods at any age. This is quite complicated, but it’s indicating, roughly, that there is an association between the amount of fat in the bodies of young people and the amount of ultra-processed food they ate, but no evidence of an association with the weight of other body components than fat.
“That seems to be clear enough evidence; what are the complications? Well, one obvious one is that the researchers couldn’t directly measure associations between consumption of ultra-processed food and the participants’ health. The measures like BMI are not direct health measures, though they have been found in many other studies to be related to health outcomes. And all five of the fifths into which the participants were divided, in terms of ultra-processed food consumption at age 7, ended up with an average BMI at age 24 of over 25, and 25 is usually taken as the upper limit of ‘normal weight’ and the lower limit of ‘overweight’, though those limits have been criticised. At those levels, a higher BMI, as was found in those who ate larger amounts of ultra-processed food, may indeed be bad for health, though this study could not show that itself. Another caveat is that some children’s diets might have changed a lot over time, in terms of the amount of ultra-processed food they ate, but the researchers based their analysis on the consumption at baseline, which was age 7 for most participants. The researchers acknowledge this, and they did also measure what the children ate at age 10 and 13, but they report that there were generally only small changes in ultra-processed food consumption over those ages, so that using only the baseline level wouldn’t have affected their findings much.
“Perhaps the biggest issue, though, is that the study is observational. That is, the researchers did not decide how much ultra-processed food the participants ate. That would clearly be impossible to do over 10 or more years of follow-up. Instead, they recorded how much was eaten, and recorded the various bodily measures and how they changed over time. The problem is that children who ate different amounts of ultra-processed food at age 7, and their families, would have differed in other ways, not just in how much ultra-processed food they ate. The associations between body fat measures and ultra-processed food consumption might be caused by these other differences, and not by the ultra-processed food consumption at all. It’s possible to make statistical adjustments to allow for other differences like this, and these researchers did so, for a range of factors including the child’s sex, ethnicity, birth weight, physical activity level, and a measure of the deprivation of the area where they lived, together with several characteristics of the mother. That does tend to increase confidence that what’s causing the differences in the patterns of BMI and so on is differences in ultra-processed food consumption, to some extent anyway, but it can’t establish cause and effect for certain. There may well be some other factors, for instance some other features of the children’s or the family’s lifestyle, that are the real causes of the BMI associations, but couldn’t be adjusted for because the researcher had no data on them. The research report, and the press release, acknowledge this possibility.
“Since this study, and almost all the other studies of associations between ultra-processed food consumption and measures related to health, are observational and so can’t establish cause and effect, can we ever know whether eating large amounts of ultra-processed foods is bad for us, and in particular whether we should therefore try to change what we eat? Or that, as the researchers on this study recommend, “More radical and effective public health actions that reduce children’s exposure and consumption of UPFs are urgently needed to address childhood obesity in England and internationally.”? How can we know that any such actions would in fact reduce childhood obesity, if we can’t determine cause and effect? It’s pretty obviously impossible to carry out large-scale randomised trials*, where people are assigned diets at random, at least not over the period right from age 7 to age 24 as in this study. So all the studies will be observational. In that situation, cause and effect is established, not by any single study, but by building up evidence of various sorts over a lot of studies, that points in roughly the same direction. That would need to include evidence of a dose-response relationship – the more ultra-processed food is eaten, the worse the health-related outcomes – and this new study did provide some evidence of that, as have others. It would also need to discuss evidence of how any causal effect of ultra-processed foods on health might actually work in the body, and that could be pretty complicated since the range of foods that count as ultra-processed is rather wide and there could be several different mechanisms involved in the body. But it’s beyond my personal expertise as a statistician to put all these diverse sources of information together – there are statistical aspects of doing that, but also non-statistical understanding of nutrition and how food is processed by the body.
“If in fact the association is one of cause and effect, though, we’d need clear advice on how people should try to change their diet, and the rather complicated definition of what is an ultra-processed food won’t make that easy. Examples can be given – the press release mentions fizzy drinks, frozen pizzas, mass-produced bread and some ready meals. But the list of ultra-processed foods is much longer than that – it also includes breakfast cereals, sugared yoghurts, and a lot more. Bread and cakes might or might not be ultra-processed, depending on who made them (what exactly counts as mass-produced?) and what the detailed ingredients are. What would need to happen to food labelling? How does any of this relate to what I might cook myself at home? In general, if there really is a need for better public health actions in response to results like these, what might those actions be? How should the cost issues* be dealt with?
* My search of PubMed did turn up two randomised trials involving ultra-processed foods. One was mostly about the effects of different labelling systems and isn’t really relevant here. But there was one randomised study from 2019 comparing two diets – see https://www.cell.com/cell-metabolism/fulltext/S1550-4131(19)30248-7. Twenty people (who weren’t ill) were admitted to a research hospital. Each of them was given access to one diet for two weeks, and then the other diet for two weeks. One diet consisted largely of ultra-processed foods, while the other contained very few. The diets were matched in terms of their calorie and nutrient content. (The random element is that half of the participants, chosen at random, got the ultra-processed diet first, and the rest got the other, to allow for the possibility that the order mattered.) Participants were allowed to eat as much as they liked, and their food consumption and various other measurements were made. Participants, on average, ate more calories, carbohydrates and fat when they were on the ultra-processed diet, and put on weight on that diet, but lost weight when on the other diet. This does tell us something, but the circumstances are so far removed from the real world that it doesn’t tell us much. Perhaps the most telling point is that the researchers calculated how much it would cost, per week, to prepare 2,000 kcal per day of ultra-processed or unprocessed meals from their diets, using prices from their local supermarket (in the US) – it was $106 for the ultra-processed meals, compared to $151 for the unprocessed meals. Another example of how cost can be a serious barrier to healthy eating. (Devotees of American eating might or might not want to look at the 44 pages of illustrations of the meals available of the two diets, in the supplemental information for this paper at https://www.cell.com/cms/10.1016/j.cmet.2019.05.008/attachment/f7d43756-3f67-4557-8322-59a9d143d63c/mmc1.pdf.)
‘Association Between Childhood Consumption of Ultraprocessed Food and Adiposity Trajectories in the Avon Longitudinal Study of Parents and Children Birth Cohort’ by Kiara Chang et al. was published in JAMA Pediatrics at 16:00 UK time Monday 14 June.
Dr Duane Mellor: “I do not have any conflicts of interest to declare.”
Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee. However, my quote above is in my capacity as an independent professional statistician.”
None others received.