February 21, 2022

expert reaction to study looking at reported vegetable consumption and risk of cardiovascular disease

A study published in Frontiers in Nutrition looks at raw and cooked vegetable consumption and risk of cardiovascular disease.

Dr Richard Francis, Head of Research, Stroke Association, said:

“This new research offers limited insights on how the amount of vegetables you eat may influence your risk of stroke.  The findings suggest that eating a diet high in raw vegetables may lower your risk of dying or being hospitalised due to a stroke (over the next 12 years or so).  However, when lifestyle and socioeconomic factors were accounted for, the links between reduced stroke risk and raw vegetable consumption were much weaker.  The study also showed that there was no link between eating more cooked vegetables and lower risk of dying or being hospitalised due to a stroke.

“This study has a number of limitations. Participants were all living in the UK, mostly white and middle-aged. None of them had heart disease when they joined the study. Where you live, your ethnicity, age and whether you’ve had heart disease can all affect your risk of stroke. So more research with broader populations over a longer period is needed.

“The links between eating vegetables and stroke risk are very complex — as shown by the conflicting results from this and other studies. What we do know is that high blood pressure is the biggest single risk factor for stroke. The best things you can do to keep your blood pressure healthy are to eat a balanced and healthy diet, exercise and stop smoking.  Getting your blood pressure checked regularly is also key to making sure your heart is healthy. If you are worried about your risk of stroke talk to your GP, who can give you support and information.”

Prof Tom Sanders, Professor Emeritus of Nutrition and Dietetics, King’s College London, said:

“Vegetables help reduce risk of cardiovascular disease probably by the contribution they make to potassium intake, which helps maintain a health blood pressure.  This paper reports the associations between self reported vegetable intake (raw or cooked) with cardiovascular disease (CVD) and total mortality in participants recruited into the UK Biobank Study.  The overall analysis showed total vegetable intake to be associated with a lower risk of CVD and all cause mortality when adjusted for potential confounding factors such as smoking habit, deprivation index, blood pressure and cholesterol lowering medication.  This finding is consistent with previous research.  Further analysis found that the benefit in terms of reduced CVD was confined to raw vegetable with no effect with cooked vegetables.  However, all cause mortality was reduced to a similar extent with higher intakes of cooked vegetables compared with raw vegetables.  The conclusion drawn by the author that cooked vegetables may not be effective in reducing risk of cardiovascular disease may not be justified especially as the group consuming the highest levels of vegetables were more likely to be receiving medication for high blood cholesterol and high blood pressure (i.e. this group was at higher risk of CVD) compared with those consuming the lowest intake which may have affected their responses to the dietary questionnaire: a phenomenon referred to as reverse causality by epidemiologists.  These findings should not be taken to indicate that eating more vegetables has no benefit to health, especially cardiovascular health.”

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

“Really the main thing that this new study demonstrates is that studies of this type can be horrendously difficult to interpret, and that the interpretation has many subjective aspects, despite all the detailed statistics.  This, and other similar studies, make it very clear that the numbers do not speak for themselves.  What’s more, the provocative top line on the press release doesn’t actually match what the study authors wrote in their research paper.  The press release headline does not give a measured picture of what the research report actually concludes, and doesn’t reflect the uncertainties that are inevitably present in a study like this.

“Overall I’d interpret the researchers’ conclusions as being that there might possibly be no protective effect of eating vegetables on the health of our hearts and blood vessels, though it’s also still possible that there is such an effect.  They are saying, more clearly than is sometimes said, that all studies like this can tell us about is associations and correlations, and not about cause and effect.  If you’re thinking that this does leave things pretty unclear on what we should do in terms of eating vegetables, then you’re probably right – at least, if you look at just this one study on its own, and if you think only about the health consequences that they studied here.  (Different studies have provided evidence about possible protective effects of eating vegetables against some cancers, for instance.)

“This study is observational.  That is, the researchers did not carry out some sort of experiment in which they determined what the participants would eat.  Instead they simply recorded what they ate, and then followed them up for twelve years and recorded whether and when they had a heart attack or a stroke, died of any cardiovascular disease (CVD, that is, heart attack, stroke, or something similar), or died of any cause.  They could then calculate whether the risk of one of these unfortunate events depended on how much people said they ate, in terms of raw and cooked vegetables.

“The key problem with an observational study like this is that people who eat different amounts of vegetables will differ, on average, in terms of many other factors as well.  If it turns out that people who eat more vegetables are less likely to have a heart attack or stroke, in a given period of time, that could be because the vegetables had a protective effect on their heart and blood vessels.  Or it could be, entirely in part because of one or more of the other factors on which they differ.  For instance, it could happen that people who eat a lot of vegetables tend to be health conscious types who also take more exercise than those who aren’t so keen on vegetables, and perhaps it’s the exercise that helps their heart health and not the vegetables at all.  The other factors, that are associated with people’s diet and might also be associated with their risk of CVD, are called potential confounders, in the jargon, essentially because they can seriously muddle the picture of what’s going on.

“It’s possible to make statistical adjustments to allow, to a certain extent, for potential confounders.  Making such adjustments is routine in this kind of research.  The researchers here allowed for the participants’ age, sex and ethnicity, and also made adjustments for a lot of potential confounders, including their educational level, the level of deprivation of the area where they lived, whether they had high blood pressure or diabetes, their physical activity level, BMI, smoking and alcohol consumption, whether they took various food supplements, took certain prescribed drugs, how much they ate of several other types of food, and more.  The snag is that one can never be sure that all the potential confounders have been adjusted for, and anyway you can’t make adjustments for factors that you have no data on.  So you just can’t ever be sure in this type of research that you have dealt with all the confounding.  In this study, despite the top line on the press release, the chance of dying or having a heart attack or stroke during the follow-up period was generally found to be lower in people who ate a lot of vegetables, particularly raw vegetables, than in people who ate few vegetables or none at all, even after all the adjustments.  That could be because eating raw vegetables really does cause a reduction in CVD risk.  Or it could be because all the careful adjustments did not deal with all the confounding, so that there is some so-called ‘residual confounding’.  You just can’t tell from this single study which of these is true.  That’s why it’s usual to say that a single observational study cannot, on its own, establish cause and effect.  It can find evidence of correlation or association, but not causation.

“However, in this case the researchers go rather further and say that “Residual confounding is likely to account for much, if not all, of the observed associations.”  They do still state that “one cannot rule out the possibility of a true causal protective effect”, which is why the top line of the press release is very misleading.  The researchers are actually concluding that their study just can’t really tell us whether or not eating raw vegetables protect people’s heart and circulation.  But what’s their basis for pointing out so explicitly that there might be nothing going on other than residual confounding?

“It’s that they present the results both with and without most of the adjustments that they made, and they find a marked difference between the adjusted and what they call the ‘basic’ results.  (The basic results are adjusted for the participants’ age, sex, ethnicity, and the region where they live, which are very likely to affect CVD risk, but for nothing else.)  For instance, for the outcome of a new occurrence of CVD, the basic analysis found that the risk of CVD in a given period in those who ate the largest total amount of vegetables was a quarter less than in those who ate the smallest amount.  But after the full adjustment, the reduction in risk was much less, only a tenth.  The researchers looked at a particular statistical measure of the change produced by the adjustments, concluded that the change was very marked, and from the decided that it was quite likely that the change would have been even larger if they had been able to adjust for everything that might have been a confounder, so that the effect of vegetable eating on CVD risk could entirely disappear.

“That’s certainly a possibility, but really it’s only giving us more evidence that you can’t draw cause-and-effect conclusions from an observational study.  There are other possibilities than everything being residual confounding.  The researchers don’t mention the possibility that confounding can actually mask a real effect, as well as making a non-effect look real.  That may not be very likely, but it can’t be ruled out.  If there’s some unadjusted factor that actually makes CVD more likely, and it happens also to be higher in those who eat fewer vegetables, then that other factor would tend to work in the opposite direction to any true protective effect of eating more vegetables and might mask it.  Also, there might be a consequence of the fact that the researchers adjusted for people’s consumption of other foods, such as fruit, fish and meat.  The effect of this adjustment is that the adjusted comparison is really being made on the basis that the same amount of fruit, fish and meat is eaten, which in statistical terms makes sense.  But it doesn’t allow for the fact that, if people changed their diet to eat more vegetables, they might well end up eating less of some of the other foods.  And there could be measurement issues, as the researchers point out.  The intakes of vegetables and other foods were recorded when the participants joined the study, so they can’t take into account any dietary changes after that during the twelve-year follow up.  I’m not saying that any of these alternatives actually happened – I’m simply using them to point out that one really can’t tell from this study whether or not eating vegetables actually has a cause-and-effect association with cardiovascular health.

“One interesting and important aspect is that, as far as I can tell, similar interpretations could be made about previous studies of this kind, which raises questions about what any of them might really mean.  For instance, this new research report refers to two previous major observational studies, involving hundreds of thousands of participants: the EPIC study and the PURE study.  Both looked (among several other things) at possible associations between raw vegetable intake and all-cause mortality, and found an association between higher consumption of raw vegetables and lower risk of death in a given period of time.  Both also published their estimates before and after making the usual kind of statistical adjustments.  And both found differences between the  unadjusted and adjusted estimates on a very similar scale to this new study.  (I can’t compare them using exactly the same statistical measure as in the new study, because those figures are not given in the EPIC and PURE reports.)  However, though both of the older studies do mention the possibility of residual confounding, neither of them goes as far as the new one, to suggest explicitly and quite strongly that their findings might well entirely be residual confounding.

“That leaves open the question of why the new study should make its suggestion about residual confounding so prominent, while the older studies do not.  I suspect this is likely to be a change in how researchers tend to see such findings, rather than in real major differences between what was done, or found, in the different studies.  In many ways, that’s good, but it does make it awkward in trying to work out how these findings could relate to advice about what we should eat.  Vegetables might be good for us in terms of heart and blood vessel health, or they might not, according to this study, and (by implication) the same goes for previous similar studies too.  And maybe different individuals should eat different combinations of foods – a study like this one that looks at overall averages can’t tell us much about that anyway.

“The researchers do suggest a way out of this, though it will need yet more research.  They suggest using research and statistical methods that explicitly aim to find evidence about cause and effect, such as mendelian randomisation (which uses genetic information as well as recording what people eat and what happens to their health).  I think that’s a good suggestion.

“But overall, really whatever a single observational study like this one found, it should never be used in isolation to determine what’s going on or what we should do.  That needs evidence from several more studies, on what goes on inside the body when we eat vegetables, and more.

“One further piece of clarification about the press release.  It says, “The risk of dying from CVD was about 15% lower for those with the highest intake compared to the lowest vegetable intake. However, this apparent effect was substantially weakened when possible socio-economic, nutritional, and health- and medicine-related confounding factors were taken into account.”  Just to make it clear, the 15% reduction is the figure after these potential confounders have been taken into account.  The reduction in risk before they were taken into account is about a third.  But it’s still true that the authors would argue that even the 15% reduction may all be due to residual confounding, and not be because of a true protective effect at all.”

Dr Dipender Gill, NIHR Clinical Lecturer in Clinical Pharmacology and Therapeutics, St George’s, University of London, said:

“This observational study finds that after correcting for potentially confounding factors, the association between vegetable consumption and cardiovascular disease is substantially reduced.  The authors conclude that there is a need to reassess the burden of cardiovascular disease that is attributed to low vegetable intake in high-income populations.

“However, there are important study limitations that mean we can’t infer what causes what in the complex relationship involving vegetables, cardiovascular risk and lots of other things.  As the authors acknowledge, there is the possibility of residual confounding affecting the association estimates.  Furthermore, many of the considered confounders that were adjusted for may actually represent mediating mechanisms.  For example, vegetable consumption may reduce cardiovascular risk by lowering blood pressure and bodyweight, and improving glycaemic control.  By adjusting for such traits, the authors may inadvertently be negating some of the mechanisms by which vegetable consumption is exerting beneficial effects.

“While the study contributes to the overall body of evidence, it should not discourage individuals from consuming vegetables.  There is an abundance of other evidence strongly supporting the health benefits of a balanced diet that is rich in vegetables.”

Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:

“This is a large observational study with two very interesting results:

1) People who eat more vegetables are generally healthier.

2) This study didn’t suggest that it was eating vegetables that makes those people healthier in terms of cardiovascular disease.

“These results are not surprising, and it is well known that healthy people tend to follow a healthier lifestyle and it is fiendishly difficult to explore cause and effect.  In this study, people consuming high amounts of vegetables were better educated, less likely to smoke and more likely to be physically active.

“In their analyses, the authors could show that a beneficial effect of fruit and vegetable intake disappeared when adjusting for expected confounders.

“But does this mean that there is no benefit in eating vegetables?

“It doesn’t, because it is much more complicated.  There are a number of limitations when interpreting the results:

1) Measuring food intake is notoriously difficult, and the limitations of self-reporting are well known.  Study volunteers tend to over-report foods that they believe are healthy, and this will affect results.  Indeed, some studies using more objective markers showed much stronger associations between fruit and vegetable intake and health.

2) People who don’t eat vegetables need to eat something else – and when estimating the health effect of eating vegetables it is important to consider what they replace.  Replacing a sugary snack with carrot sticks is likely to improve health – and have a beneficial effect on CVD risk.  That wouldn’t be the case when replacing a whole-grain snack with carrots.

“Such substitutions are notoriously difficult to investigate in observational studies.

3) There is not one “vegetable”, but they are all very different.  An analysis based on intake of total vegetable intake is more likely to look at behavioural factors than the effect of the actual food.  Vegetables have a wide range of nutrients and bioactive that are known to be beneficial for health: fibre, vitamins, minerals and plant bioactives that many believe to be healthy.  Dietary recommendations usually stress the importance of a wide range of different fruits and vegetables: eating exclusively cucumbers or kales is unlikely to have the same beneficial effect as eating a lot of different vegetables.  Unfortunately it was not possible to investigate the variety of vegetables consumed in this study.

“In summary, this is a very interesting study – but not one that should be used as a justification to stop eating vegetables.”

Dr George Savva, Statistician, Quadram Institute Bioscience, said:

“This paper confirms that people who eat more vegetables are less likely to suffer from cardiovascular disease and death, but also highlights how difficult it is to attribute this effect to any specific cause, whether it is the vegetables themselves that are beneficial or another related dietary or socioeconomic factor.  Diet and other health related factors are closely interlinked with many knock-on causal effects, so it’s important to be extremely cautious when trying to interpret the effect of just one factor in isolation.”

Victoria Taylor, Senior Dietitian at the British Heart Foundation (BHF), said:

“Although this study found that eating more vegetables wasn’t associated with a lower risk of heart and circulatory diseases once other lifestyle and other factors were taken into account, that doesn’t mean we should stop eating vegetables.

“Fruit and vegetables remain an important source of vitamins, minerals and fibre, which make them an essential part of a healthy, balanced diet.

“There are multiple factors in our diet and lifestyle that can influence our chance of developing cardiovascular disease, and we need to consider all of these when it comes to lowering our overall risk.  Adopting a whole diet approach to the way we eat – such as the traditional Mediterranean style diet – together with addressing our lifestyle, which includes not smoking, being physically active and managing our weight, are important to reducing the risk of heart and circulatory conditions.”

Prof Janet Cade, Nutritional Epidemiology Group, School of Food Science and Nutrition, University of Leeds, said:

“This is of course interesting, coming from a very large cohort study.  However, a key consideration is the appropriateness of the measure used for vegetable intake.  The key exposure measure used here was extremely simplistic, representing just two questions asking people to report how much raw and cooked vegetables they eat on average a day.  There lies a problem.  The word ‘average’ is used in many different contexts to represent the mean, median or mode.  So what people have reported could be any one of these.  Participants would have to undertake some internal mathematical calculations to work out an average intake according to their understanding of the word average.

“Secondly, the list of confounding factors included in the fully adjusted models was extensive, with the potential for overadjustment to occur.  For example, it was unclear why the authors chose to adjust for fresh fruit intake.  In any of these associations adjusting for highly correlated variables, such as fruit and vegetables could lead to adjusting out the effect of vegetables.

“Whilst residual confounding is a possibility, more likely, in my view, is that the measure of exposure was too rough, with considerable measurement error included.  Measurement error can lead to bias, either diluting or exaggerating the apparent effect size.  This is a particular problem in the area of nutrition epidemiology where measuring long-term dietary intake is prone to error.

“The conclusion for me should be that this study highlights the need for rigorous assessment for exposure measures of diet to be able to study the effect of this on disease risk.  In future, more reliable techniques for measuring actual intake, such as online tools including myfood24 would help us advance in this field.  Providing the opportunity to measure food and nutrient intakes in large population studies in detail.  These new technologies were not available at the start of the Biobank project and unfortunately doubt has to be cast on the reliability of findings from the use of simple questions expecting users to express an average intake value.”

Dr Ian Johnson, Nutrition researcher and Emeritus Fellow, Quadram Institute Bioscience, said:

“One major limitation of almost all research on food and human health is the near impossibility of undertaking double-blind placebo-controlled intervention trials, which are the gold standard for other aspects of medical research.  Researchers are therefore obliged to undertake observational studies on the general population.

“This paper describes a very strong prospective investigation of cardiovascular disease in a large population of UK citizens whose diet and lifestyle were recorded at the beginning of the investigation before any disease emerged.  The authors observed statistically significant protective associations between raw vegetable consumption and both the incidence of, and death from, cardiovascular disease, which were consistent with many previous studies on this topic.

“However, the authors have also shown that these effects tended to disappear when a range of other aspects of lifestyle and diet were taken into account, including exercise, meat intake, smoking and excessive alcohol intake, all of which have previously been shown to be associated with the risk of heart disease.

“The authors themselves acknowledge that they cannot rule out the possibility of a true protective effect of raw vegetable consumption, but their results do seem to show that other aspects of a healthy diet and lifestyle may be more important.  The authors also acknowledge that there is evidence from other research for protective effects of higher vegetable intake against other common diseases.

“This study raises interesting questions about the relative importance of the many different aspects of healthy diet and lifestyle, but it should not discourage consumers from following public health recommendations to consume diets high in vegetables of all types.”

Prof Naveed Sattar, Professor of Metabolic Medicine, University of Glasgow, said:

“There is good trial evidence that eating foods rich in fibre such as vegetables, can help lower weight, and improve levels of risk factors known to cause heart disease.  The present observational study cannot overcome such evidence and its conclusions can be debated since the authors may have over adjusted for factors that account for lower intake of vegetables.  The authors correctly admit this possibility.

“In short this paper should in no way change advice to eat at least 5 portions of fruit and vegetables a day.  Many living in the UK fall well short of this, sadly, and more needs to be done to encourage better intake of vegetables.  In fact, I suspect we may have underestimated the importance of a healthy diet on health and disease in general.  We are good at treating with preventative drugs, but the UK needs to up its game in helping people improve lifestyles, including diet and activity – that is the big goal going forwards post pandemic.  How we best do this, though, requires more thought from Government and health agencies.”

‘Raw and Cooked Vegetable Consumption and Risk of Cardiovascular Disease: a Study of 400,000 Adults in UK Biobank’ by Qi Feng et al. was published in Frontiers in Nutrition at 05:00 UK time on Monday 21 February 2022.

DOI: 10.3389/fnut.2022.831470/full

Declared interests

Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee.  My quote above is in my capacity as an independent professional statistician.”

Dr Dipender Gill: “I am an NIHR Clinical Lecturer in Clinical Pharmacology at St George’s, University of London.

I am employed part-time by Novo Nordisk, unrelated to the subject of the current study.”

Prof Gunter Kuhnle: “No COI to declare.”

Dr George Savva: “No conflicts of interest.”

Prof Janet Cade: “I am Director of Dietary Assessment Ltd which supports myfood24.”

Dr Ian Johnson: “No conflicts.”

Prof Naveed Sattar: “No COI.”

None others received.