select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to study looking at plant-based, fish and other diets and COVID-19 severity

An observational study published in BMJ Nutrition Prevention & Health looks at plant-based and pescatarian diets and severity of COVID-19 infection among healthcare professionals.


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

“Since the beginning of the pandemic, there has been a lot of speculation about the impact of diet on disease risk.  This study attempts to answer this question, but there are a number of limitations that need to be considered: The study relied entirely on self-reporting, and a lot data have shown that self-reported dietary intake is unreliable.  In this study, participants were asked about their diet after they were diagnosed with COVID-19, and this might lead to further misreporting, especially among participants who are interested in a potential link between diet and disease.  Finally, the study has been conducted in different countries with widely different diets – a plant based diet in Spain or Italy is likely to be different from a mainly plant-based diet in Germany or the UK.

“The findings of the study are not surprising: people who follow a mainly plant-based diet or eat fish are often healthier when compared to a control group with a “normal” diet.  However, these results are often misleading: a plant-based or pescatarian diet is often followed by people who have an overall “healthier” lifestyle, are more physically active, better educated and belong to higher socio-economic groups – factors that are associated with better health and also a reduced risk for COVID-19 (in the UK, deprivation and social class were strongly associated with COVID-19 risk).  It is therefore impossible to infer from the data in this study whether plant based or pescatarian diets offer any protection, especially as the data presented show a lot of uncertainty.

“An interesting – and for some surprising – finding is the higher risk found in those following a low-carbohydrate diet.  The same limitations as above apply of course – and there are widely differing interpretations of “low carb”, but the data suggest that those following such a diet have a higher risk of COVID-19.  “Low carb” diets have become very popular recently, in particular to treat diabetes and other metabolic diseases.  It is therefore likely that those following a “low carb” diet are more likely to have an already higher risk for COVID-19.

“In summary, this is an interesting study, but does not provide much new information.  The current dietary recommendation emphasise the importance of plant based foods, and most data suggest that such a diet has an overall benefit on health.  There are many better and more reliable ways to reduce the risk for COVID-19.”


Dr Carmen Piernas, University Research Lecturer and nutrition scientist, Nuffield Department of Primary Care Health Sciences, University of Oxford, said:

“The study by Kim et al is an observational case control study investigating the associations between plant-based and pescatarian diets with COVID-19 severity among healthcare professionals across six high income countries.

“The research question is interesting, but the conclusions needs to be drawn cautiously in the context of several major methodological issues:

“1. The conclusions are based on a very small sample of cases, with only four cases of moderate/severe COVID-19 disease among people following a plant-based diet, and six among those following plant-based or pescatarian out of a study population of 2884 people (itself rather small to test diet-disease associations).

“2. The study population only included healthcare professionals with high exposure to COVID-19 and is not representative of the range of dietary behaviours in the wider population.  There could also be a higher chance of reporting healthier behaviours (socially desirability bias) in this group.

“3. Ascertainment of cases and controls was done through the self-reported questionnaire, which was based on experience of symptoms and/or test positivity (and only 298 cases out of 568 had a confirmed positive test).  A potential source of bias could have been introduced by considering asymptomatic COVID-19 cases as controls.

“4. Dietary exposures were broadly captured through a questionnaire which asked the type of diet followed over the previous year where participants could choose between 11 possible dietary patterns.  The questionnaire included other dietary patterns considered to be healthy that could also be plant-based e.g. Mediterranean diet, so the exposure variable is poorly defined.  There was no a priori hypothesis and no pre -registered protocol; the authors chose to highlight whole food plant-based, vegetarian and pescatarian diets on the basis of having a sufficient number of people following these diets, which limits the scope of this study to help understand dietary patterns in association with COVID-19 severity.

“5. Both dietary measures and body mass index were self-reported which can potentially introduce systematic measurement error in the estimations.  To date, body mass index has been the strongest modifiable risk factor (with a potentially causal link from Mendelian randomisation studies), and although the authors adjusted for BMI  in the final models, some residual confounding may still remain.

“Overall, this is methodologically weak and while it raises an interesting possibility that the severity of COVID-19 may be related to diet quality, a larger-scale population based study is needed to support these preliminary findings since the population studied here is unlikely to represent behaviours of the general population.”


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

“The findings of this study are not easy to interpret.  One possible explanation for them is the main one that the researchers give, that eating certain diets might possibly protect against severe Covid-19.  The trouble is that there are several other possible explanations that are also compatible with the findings, as I’ll explain.  So what might be done in the light of this new research?  All that the authors suggest, as action, is to carry out more detailed studies on possible associations between what people eat and Covid-19 severity.  That might indeed be worthwhile – though I do ask myself whether this new study was really needed to show that the more detailed studies could be worth doing.  Because of these issues, it’s not at all clear from this study that anyone should contemplate changing their diet just to give protection from serious Covid-19 – maybe it would help, maybe it wouldn’t, the study just can’t tell us.  (Of course there might well be other health reasons why a change of diet would be a good idea anyway.)

“What are the alternative explanations?  The first one arises because the study was observational.  That is, people weren’t asked to change their behaviour by the researchers – they just ate what they were going to eat anyway, and the researchers asked them about their diets and a whole lot of other things, and also recorded whether or not they had had Covid-19, and if so, how severe it was.  The problem is that there are other differences between people who said they ate different diets, apart from their diets.  These other differences could have been, in part or in whole, the cause of any differences in Covid-19 severity, and not the differences in diet at all.  It’s possible to make adjustments in the statistical analysis to try to take into account other factors or differences like this, and the researchers did make those adjustments for several factors including things like age, gender, smoking, physical activity and body mass index.  But the adjustments can only be done for factors on which the researchers had data, and despite using a questionnaire with around 100 items, they may not have had data on some important factors.  Indeed it’s possible that they did not adjust for some important factors on which they did have data.  For instance, they report that, in the ‘cases’ who did get Covid-19, those who said they followed a plant-based diet drank alcohol, on average, 40% fewer times a week than those who didn’t follow a plant-based diet (and that difference was too big to be explained just by statistical variation).  Some previous research has suggested that drinking alcohol may be associated with differences in Covid-19 severity.  But these researchers didn’t adjust for alcohol consumption, and they also didn’t adjust for the consumption of sugar-sweetened drinks, which was greater (on average) in people who weren’t on plant-based diets.  It’s because of issues like this that the press release, correctly, says that the study can’t establish cause, only correlations.

“It’s also important to understand exactly what correlations were found.  The correlations are not between Covid-19 severity and some direct detailed measures of what people ate, or what they said they ate (in detail), despite the fact that the questionnaire included detailed questions on how often they ate 47 food items.  Instead the correlations are between Covid-19 severity and what people said about the type of diet they had followed over the whole past year before the Covid-19 pandemic.  Respondents chose from 9 specific diets, including keto diet, Mediterranean diet, Palaeolithic diet, low carb diet, and more; or they could say Other diet, or that they followed none of the diets.  It’s not clear (to me anyway) whether respondents could choose more than one diet from the list, for instance if they had changed their diet during the past year.  The researchers then grouped what people said their diets were into three groups for analysis (and those groups didn’t include most of the specific diets on their list of 9).  People who said they had eaten either “whole foods, plant-based diet” or “vegetarian diet” were grouped under “plant-based diets”.  People who said they had eaten either one of those two diets or “pescatarian diet” were groups under “plant-based diets or pescatarian diets”.  Those who said they had eaten either “low carbohydrate” or “high protein” diet were grouped under “low carbohydrate, high protein diets”.  Together, 777 of the 2,884 people studied fell into one of those three grouped sets of diets, so obviously over 2,000 of them did not.  But those diet categories don’t seem to be really closely linked to the details of what the respondents said they ate, in the food frequency questions.  The researchers report some data on the details of what they ate, for the ‘cases’ (people who got Covid-19, whatever the severity).  People who said they followed a plant-based diet did on average, eat vegetables and nuts rather more often than did the people who didn’t say they followed a plant-based diet, and they also ate poultry and other meat less often.  But they still, on average, ate poultry and other meat 2.5 times a week, compared to about 6 times a week for those not following a plant-based diet.  Sounds like quite a bit of meat consumption for people who said they were on a plant-based diet.

“The thought in my mind about this is that the associations that were found were to do with how people named their diet, and not what they actually ate.  There’s likely to be some correlation between how they named their diet and what they ate, of course.  But someone who, say, eats plenty of fresh vegetables and whole foods, but lapses into having a pepperoni pizza once a week, might say they are on a whole-food, plant based diet, or they might feel that their pizza consumption goes against that and put themselves down as having none of the diets.  Depending on which choice they make, they will end up on opposite sides of the main comparison that the researchers made.  I do therefore wonder why the researchers chose to analyse in that way.  My concern is that people who say they are on some of the named diets, for instance a whole food plant based diet, may be thinking more about healthy behaviours than people who don’t label themselves that way, and people who are thinking more about healthy behaviours may have acted in other healthier ways because of that mindset.  Those other differences in behaviour might be what explains their lower chance of moderate or severe Covid-19.  An association between the names people gave to their diet and Covid-19 severity, rather than looking at what they actually ate, does seem on the face of it to be coming at it from the wrong end.

“I also wonder about the researchers’ choice to concentrate on comparing people who got moderate to severe Covid-19 with people who did not get Covid-19 at all, rather than comparing people who got Covid-19 at any level of severity with people who didn’t get Covid-19.  Both comparisons would have been interesting.  However, comparing people who got more serious Covid-19 with those who didn’t get it at all does mean that the comparison takes into account both any effect of the diet on the chance of catching Covid-19, and also any effect of the diet on the severity of Covid-19 in people who did catch it.  That perhaps makes the conclusions more difficult to tease out.  The researchers do also, in passing, make a statistical comparison between people who got very mild to mild Covid-19 and people who got moderate to severe Covid-19, leaving out those who did not get Covid-19 at all, and that comparison does not directly involve the chance of catching Covid.  However, they do not discuss those differences in detail, and I think no adjustments were made for other factors that could be influencing the results.  I mention these choices of analysis really only to point out that there are always such choices, and it’s always worth thinking about how different the findings might have been if the researchers had made different choices.  It’s generally considered to be good practice to  register in advance what statistical choices will be made, before the data have been collected and analysed, to avoid any doubt that the choices were made after looking at the data, but the researchers don’t report that such pre-registration had been done in this study.

“Finally, as the researchers point out, this study was done using a set of health care workers on the Covid front line, so the findings might not apply to other groups.  Also, the participants are not typical of all health care workers with substantial exposure to Covid-19 – 95% of the participants were doctors, and most (over two-thirds) were male.”


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

“As is now well known, COVID-19 is a viral illness characterised by a variety of outcomes, ranging from symptom-free infections to severe and potentially fatal disease.  The risk of severity and death increases steadily with age, but, within age-groups, the mechanisms that determine the severity of disease amongst infected individuals are poorly understood.  It is important therefore to search for environmental and life-style factors that may influence the vulnerability of individual patients to severe disease.  In this study the authors recruited health care workers, from five European countries and the USA, who were at high risk of occupational exposure to COVID-19, and then assessed both their dietary habits and their experience of infection, duration, and severity of disease.  Dietary habits were assessed by asking the participants to state whether they followed any of a set of pre-defined dietary patterns, and by means of a short food-frequency questionnaire.  There were no effects of dietary pattern on the risk of contracting COVID-19, but amongst the participants who did become infected, those who reported following plant-based diets and plant-based diets plus fish, had 73% and 59% lower odds of moderate to severe disease.  Strikingly, participants who reported consuming low-carbohydrate, high-protein diets were nearly four times more likely to experience moderate to severe disease compared to those reporting a plant-based diet.  It should be noted however that the range of uncertainty around these estimates was high, and the statistical significance of the differences was relatively low.

“Diets based on high levels of plant foods and low levels of meat are associated with lower risks of several non-infectious conditions, including cardiovascular disease and some types of cancer, but it is unusual and intriguing to see evidence of protective effects against an acute infectious disease such as COVID-19.  However, as is always the case with observational studies of this type, it is impossible to be sure that the observed correlations reflect real causal mechanisms.  The authors speculate that their observations may be attributed to a beneficial effect on the immune system caused by higher levels of micronutrients in plant-based diets, but unfortunately their experimental design could not provide any data to support this idea.  In short, this study is interesting, but not conclusive.”


Prof Francois Balloux, Professor of Computational Systems Biology and Director, UCL Genetics Institute, UCL, said:

“The study reports that doctors eating plant-based or pescatarian diets tend to be at significantly lower risk of developing severe COVID-19 symptoms upon infection.  The sample size is decent, and the analyses look competently performed.  Though, further validation may be required to confirm a direct, causal between diet and COVID-19 illness severity.  Indeed, unaccounted lifestyle variables correlated with diet might influence general health of the subjects of the study, and hence how well they coped with COVID-19 infection.”



‘Plant-based diets, pescatarian diets and COVID-19 severity: a population-based case–control study in six countries’ by Hyunju Kim et al. was published in BMJ Nutrition Prevention & Health at 23:30 UK time on Monday 7 June 2021.

DOI: 10.1136/bmjnph-2021-000272



Declared interests

Prof Gunter Kuhnle: “No conflicts to declare here.”

Dr Carmen Piernas: “No conflicts.”

Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee.  I am also a member of the Public Data Advisory Group, which provides expert advice to the Cabinet Office on aspects of public understanding of data during the pandemic.  My quote above is in my capacity as an independent professional statistician.”

Dr Ian Johnson: “No conflicts to declare.”

None others received.

in this section

filter RoundUps by year

search by tag