May 29, 2019

expert reaction to two studies looking at highly processed foods and cardiovascular disease

Two studies published by The BMJ demonstrate positive associations between consumption of highly processed foods and risk of cardiovascular disease and death.

Victoria Taylor, senior dietitian at the British Heart Foundation, said:

“It’s important to remember that observational studies like these can only show an association.  They cannot tell us what is behind this.  The classification of ultra-processed foods used by the researchers is very broad and so there could be a number of reasons why these foods are being linked to increased risk to our health, for example nutritional content, additives in food or other factors in a person’s life.  Before we consider making any changes to advice or policy it is important to understand this thoroughly.

“We already recommend people adopt a Mediterranean style diet which also happens to include plenty of minimally or unprocessed foods such as fruit, vegetables, fish, nuts and seeds, beans, lentils and wholegrains.  This along with exercising regularly and not smoking has been shown to be beneficial for lowering risk of heart and circulatory disease.”

Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“This is interesting new research, and the statistical analyses of the data appear comprehensive and appropriate.  But there are still some important problems in deciding what to do about the findings.  There are issues about what causes what.  There are issues about whether the findings apply to other populations.  There are issues about exactly what advice is implied by the findings.  Finally, I believe the research reports raise issues about the boundaries between scientific research and campaigning.

“Both research reports (and the press release) state clearly that they do not establish a causal link between the consumption of ultra-processed and health risks.  That’s because there are many other differences between the people who ate relatively small and relatively large quantities of ultra-processed foods, apart from their consumption of those foods.  For instance, they smoked more on average.  These other differences might be the reason for differences in risks of bad health or early death between the groups of people.  Statistical adjustments can be made to allow for some of these other differences between groups, and both sets of researchers made several such adjustments.  But one can’t be sure that everything relevant was allowed for, and adjustments simply can’t be made for things on which the researchers have no data.

“In general, no single observational study can establish what causes what.  Confidence that the ultra-processed food consumption might really be a cause of ill health should increase if more studies of more different kinds make similar findings, and more studies do keep appearing.  However, these studies, and previous ones of the same general kind, throw effectively no light on how ultra-processed foods might cause ill health.  Both of these research reports list several suggestions about how that might work, but they are no more than suggestions and, in some cases, speculations and do not arise directly from the data in these studies.  Also, the majority of the available studies that follow up groups of people, as these ones do, are actually based on the same two groups of people as these studies – the NutriNet-Santé cohort in France and the SUN cohort in Spain.  Thus they do not really provide independent confirmation that the effects apply generally.  What’s more, both these groups are far from typical even of the populations even of the countries they come from.  They are better educated than average (e.g. the Spanish cohort are all graduates), generally better off, and women predominate in both groups.  The French cohort was recruited specifically to investigate relationships between nutrition and health.  People who volunteer for such a study are unlikely to be at all typical of the whole population in terms of their concerns and food and health, and that may make their behaviour, and their health, untypical in several ways.  As one of the reports emphasises, the fact that a cohort is not typical of a population may not matter if we can be sure that the biological processes that change health risks in the group are the same as those in the general population.  But we can’t be sure of that, because we aren’t sure what the processes are.  That said, a similar piece of research was published earlier this year¹, which looked at associations between ultra-processed food intake and death rates in a US cohort of people which is much more closely representative of the US population, and it found broadly similar results to the new Spanish study and to an earlier study of mortality based on the NutriNet-Santé cohort².  Together, these studies do increase my confidence that there’s something real behind these associations, but I’m still far from sure.

“Both these new studies use data from populations that are relatively young, or middle-aged, on average, and are based on relatively short periods of follow-up.  Perhaps any patterns of risk would be different in older people or in those followed up for a longer time.  Because of the age and short follow-up, most of the participants didn’t suffer the events being studied.  In the French study of CVD risk, only 1.4% of the group had a cardio-vascular disease during follow-up.  In the Spanish study of death risk, just 1.7% died during the study.  This means that the numerical estimates of the increased risk in people who eat larger amounts of ultra-processed foods are subject to a lot of uncertainty.  The Spanish study reports an 18% increase in all-cause death risk associated with eating each additional serving of ultra-processed food.  This is already rather hard to interpret because the serving size is different for different foods and in fact the research report does not say what the sizes were.  But the data are consistent with an increase of anywhere between 5% and 33%.  If, for the sake of argument, I assume that increase really is caused by the ultra-processed food consumption, and that the increase in risk is the same, for one serving of any ultra-processed food, throughout adult life – and those are probably unjustified assumptions – that would mean that eating an additional serving of ultra-processed food would reduce life expectation by somewhere between 6 months and 3 years.  It’s not really possible to do a similar calculation for the CVD risk figures from the French survey, but again there is considerable uncertainty.  They report the increase in CVD risk associated with increasing the ultra-processed food consumption by 10 percentage points.  That would actually be a very substantial increase in consumption for the people in this study – their actual consumption of ultra-processed foods was about 17% of their total food consumption, so a 10 percentage point increase would put this right up to 27%.  The increase in CVD risk associated with this change is reported as 12%, but the data are consistent with an increase anywhere between 5% and 20%.

“What, if anything, should be done about these results?  Both sets of researchers do point out that more research is needed to find out more clearly what is going on.  Given that the category of ultra-processed foods is rather wide and diverse, and contains some quite subtle distinctions (for example, bread might not be ultra-processed, but it counts as ultra-processed if it is mass-produced and packaged, and/or contains emulsifiers), it will be difficult to be clear on what, if anything, might be causing the increased risks.  It will probably be different with different foods.  If advice is to reduce consumption of ultra-processed foods, then any health benefit might depend on what people eat instead.  They might well not eat the same patterns of non-ultraprocessed foods as did the people in these studies who had chosen less processed diets.  They might alter their other behaviours in ways that are good, or bad, for their health.  We just can’t tell yet.

“Finally, it’s clear to me, from the way that these research reports are written, that the researchers themselves do generally believe that we should eat fewer ultra-processed foods.  That’s fine, and they know much more about the subject than I do – but in the Abstract of the French research paper, in the Conclusions section, the authors say that “public health authorities in several countries have recently started to promote unprocessed or minimally processed foods and to recommend limiting the consumption of ultraprocessed foods.”  This is true, in relation to France and Brazil at least, but that is not a conclusion from this piece of research, and in my view, putting such a statement in the Conclusions section of the abstract of a research paper is overstepping the boundary between scientific reporting and campaigning.”

1 Kim H, Hu EA, Rebholz CM. Ultra-processed food intake and mortality in the USA: results from the Third National Health and Nutrition Examination Survey (NHANES III, 1988-1994). Public Health Nutr 2019:1-9. 

2 Schnabel L, Kesse-Guyot E, Allès B, et al. Association Between Ultraprocessed Food Consumption and Risk of Mortality Among Middle-aged Adults in France. JAMA Intern Med 2019. doi:10.1001/jamainternmed.2018.7289

Dr Gunter Kuhnle, Associate Professor in Nutrition and Health, University of Reading, said:

“These are two very well and thoroughly conducted studies which show that very high intake of ‘ultra-processed’ foods are associated with higher disease risk.  Increases in cardio-vascular disease risk were rather modest and only found in people whose diet included more than 30% ultra-processed foods (less than a quarter of the study population) – however no increases in death from heart disease or cancer were found.  Indeed, increased mortality was only observed in those consuming four or more servings of ‘ultra-processed’ food per day.

“These studies are important, as they show that there is an association between the consumption of ‘ultra-processed’ foods and health which warrants further investigation.  A study published previously suggests that these foods encourage overeating, and ‘hyperpalatabilty’ of many – but not all – foods in the ‘ultra-processed’ category could explain to some extent the observed associations.

“The one major limitation of these studies is the focus on ‘ultra-processed’ foods.  While the term is convenient to create the image of ‘unhealthy food’, it is neither specific nor useful to inform public health or give dietary advice.  While ‘ultra-processed’ food is commonly assumed to be food that is extensively processed and where “numerically the majority of ingredients are preservatives; stabilizers, emulsifiers, solvents, binders, bulkers; sweeteners, sensory enhancers, colors and flavors; processing aids and other additives” (from the NOVA classification), it actually contains foods which undergo few processing steps, such as hamburgers, crisps or chips, or contain preservatives that have been used for centuries such as preserves.  It is also not obvious why salami is considered to be ultra-processed, yet cheese, which often requires considerably more processing steps and additives, is not.  The classification combines a wide range of foods with very different potential impacts on health, which limits its usefulness as a basis for recommendations.

“It is likely that the consumption of ultra-processed foods is a marker for an overall ‘unhealthy’ lifestyle.  Indeed, it is noticeable that the population with the highest intake of ‘ultra-processed’ food had a generally rather unhealthy lifestyle and was more likely to smoke and less likely to be physically active.  Dealing with these underlying causes is more important than simply addressing the symptoms.”

Paper 1: ‘Ultra-processed food intake and risk of cardiovascular disease: prospective cohort study (NutriNet-Santé)’ by Bernard Srour et al. was published in the BMJ at 23:30 UK time on Wednesday 29 May 2019.

Paper 2: ‘Association between consumption of ultra-processed foods and all cause mortality: SUN prospective cohort study’ by Anaïs Rico-Campà et al. was published in the BMJ at 23:30 UK time on Wednesday 29 May 2019.

Declared interests

Victoria Taylor: “None.”

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 Gunter Kuhnle: “I have no conflict of interest here.”