Are vegetarians & vegans less likely to get Covid-19? – Zoë Harcombe

Executive summary 

* Population studies typically involve tens or hundreds of thousands of people. This population study involved 702 people in Brazil – 80% of them women.

* It claimed that vegetarian and plant-based diets were associated with a lower incidence of Covid-19 than omnivore diets. There were no claims that diet made any difference to the severity or duration of Covid-19.

* There were many issues with the paper.

* The diet was self-reported. The incidence of Covid-19 was self-reported. We don’t know if the self-reported Covid-19 was based on self-diagnosed symptoms or a(n unreliable) test.

* All groups in this study consumed animal foods. The term ‘plant-based’ diet implies vegan but there weren’t enough vegans to study. Participants were divided into omnivores, flexitarians (who consumed meat up to 3 times a week and could consume unlimited fish, eggs and dairy – what I would call omnivores) and vegetarians/vegans combined (and thus a group consuming eggs and dairy).

* The usual healthy person confounder was present and very relevant to Covid-19. Those in the so-called ‘plant-based’ group were far less likely to be black, far less likely to be overweight or obese and far less likely to have comorbidities – all known risk factors for Covid-19. These factors were adjusted for, but we can’t adjust for a completely different person.

* The claimed plausible mechanism for why those eating more plants (because eating only plants wasn’t studied) might self-report lower Covid-19 incidence did not withstand scrutiny.

* Those more concerned about Covid-19 are more likely to test for it (more tests = more incidence) and more likely to think that they have it (symptom worry). Those with risk factors (ethnicity, BMI, comorbidities etc) are more likely to be concerned about Covid-19 and thus more likely to self-report that they have it. Those just happened to be the people over-represented in the omnivore group. Notwithstanding that the omnivore group just happened to be more omnivorous than the other groups.

Introduction

This time last year we reviewed a paper called “COVID Vaccine Hesitancy and Risk of a Traffic Crash” (Ref 1). This had plenty of flaws, but it was a contender for one of the strangest titles for a Monday note yet. One year later I woke up to the headline “Vegetarians and vegans ‘less likely to get Covid’” (Ref 2). My first thought was – how would that work?

My friend in Australia woke up to a similar headline “People who eat plant-based diets could be at lower risk of getting COVID-19: study.” Well done Australia for including that very important word “could” (Ref 3). However, far more caveats than this were needed.

The paper behind the headlines was published in BMJ Nutrition, Prevention & Health (Ref 4). The lead author was Acosta-Navarro. The study was conducted in Brazil and by a team from Brazil. The introduction to the paper narrated “The initial epicentre of the COVID-19 pandemic was the Huanan seafood wholesale market in Wuhan, China where man was contaminated by zoonotic transmission.” The paper did not start factually, therefore, as there are other hypotheses about the origin of SARS-CoV-2.

There were a number of issues with this study. We’re not going to reach the 14 that were easy to find with last October’s red meat and diabetes study, but I’ll approach this note in the same way – reporting issues as we go along (Ref 5).

The study

The study was a population study but small relative to typical population studies. It included 702 people whereas many of the population studies we review include tens or hundreds of thousands of people. People were recruited for this study through social media networks and the internet during the period 18 March to 22 July 2022. Approximately 80% of people in Brazil are estimated to use social media, so one fifth of the population was excluded from this study (Ref 6). It was not clear from the paper if the recruitment period was also the time period for self-reported Covid-19, or if people were reporting if they had ever had Covid-19. The discussion section reported that the study was carried out in Brazil’s third wave of Covid-19 (March to July 2022) but gave no further clarity on time scales.

Participants received an online questionnaire about sociodemographics, lifestyle, past medical history, and eating patterns. They were divided into two groups – omnivorous and plant-based, according to their self-reported dietary pattern. The paper admitted “a basic food frequency questionnaire served as a tool for validation of the main self-reported dietetic pattern.”

Issue 1 – dietary questionnaires are flaky at the best of times. A basic one is even more so.

The researchers defined omnivorous participants as “those who consumed any food of animal origin.” That was reasonable until you saw the definition of the so-called plant-based group. This included “flexitarian” and “vegetarian” subgroups and was divided into these two subgroups for the key results. The flexitarian group was defined as “individuals who consumed meat at a frequency ≤3 times a week.” Flexitarians might also have been consuming fish, eggs and dairy daily. That’s also an omnivore, therefore. The vegetarian group included people who meet the definition of vegetarian (people who consume eggs and dairy, but not meat or fish) and it included vegans (people who consume no animal foods). The paper admitted “We combined the vegan and lacto-ovo-vegetarian groups because of the relatively lower sample size of the vegan group.” i.e. there weren’t enough purely plant-based (vegan) people to study.

Issue 2 – all groups in this study consumed animal foods. There was no vegan group, although that’s what “plant-based” implies. The omnivore, flexitarian and vegetarian groups all consumed animal foods.

There were 424 omnivores, 87 flexitarians and 191 vegetarian/vegans. We don’t know how many vegans there were – other than not enough.

The characteristics table

As this is a population study, we turn first to the characteristics table. Table 1 informed us that 80% of participants were female. These were evenly distributed between the omnivore and so-called plant-based groups and so 80% of both these groups were female. Females were clearly more attracted to participating in this study than males. The average age of all participants was 37 years old, and this also didn’t vary significantly between diet groups. (This study was not generalisable far beyond young Brazilian women).

Ethnicity was the first significant difference between the omnivore and ‘plant-based’ groups. There was an error in the table. The proportion of people in the ‘plant-based’ group who were white was 72%. The table reported it as 60.9%. This means that the numbers in that part of the table didn’t add up to 100%. That’s a sloppy error that should have been spotted (by the researchers or the one peer reviewer). There were more than four times as many black people in the omnivore group than the ‘plant-based’ group.

The educational level was also significantly different with far more people in the ‘plant-based’ group having postgraduate level qualifications and far fewer having stopped their education at school. The omnivore group had almost double the incidence of pre-existing conditions as the ‘plant-based’ group. The omnivore group was significantly less likely to do physical activity and significantly more likely to be overweight or obese.

Interestingly, none of the measures intended to avoid Covid-19 were different between the groups. There was no significant difference in the vaccination status of the two groups. The vaccination rate was 98% across all participants. That’s the highest I’ve seen in any paper reporting this. Table 1 also recorded whether people had been following a normal life during Covid-19 or reducing contact with other people or isolating completely at home. There were no significant differences between these measures between the two groups.

(By the way, I would love to have access to the raw data as it could be cut in a different way to answer a different research question. The vaccination rate was too high to tell us anything, but the degree of isolation and incidence of Covid-19 (albeit self-reported) could have been interesting.)

Issue 3 – all of this was self-reported. Not even BMI was measured by the researchers.

Issue 4 – we have the usual healthy person confounder, which we see in all population studies. The person in the ‘plant-based’ group was better educated, less likely to be overweight or obese, less likely to have any current health conditions and more likely to be physically active. These are adjusted for, but we can’t adjust for an entirely different person.

Incidence of Covid-19

Table 1 reported the incidence of Covid-19 and whether the illness was mild or moderate to severe. Those were the only options given to describe how bad the person perceived their illness to be. I’ve extracted the relevant numbers below. Remember the self-reported issue and note that these numbers are before any adjustments.

Despite virtually everyone being vaccinated, 47% of participants self-reported getting Covid-19 (that comes from 53% self-reported not getting Covid-19). Not getting Covid-19 was split between the groups (before any adjustment) as 48% for the omnivore group and 60% for the ‘plant-based’ group.

Figure 2 presented the results that were in the abstract of the paper. After adjusting for BMI, physical activity and pre-existing medical conditions etc, the ‘plant-based’ group had a 39% lower incidence of Covid-19 than the omnivore group. (This was presented as an odds ratio of 0.61 with a confidence interval of 0.44 to 0.85.) Figure 2 further split the ‘plant-based’ group into flexitarians and vegetarian/vegans. This time the flexitarian group (omnivores basically) showed no difference from the omnivore group and the vegetarian/vegans had a very similar result to the one above for the whole ‘plant-based’ group. The odds ratio for the vegetarian/vegans, compared to the omnivores, was 0.61 and this was a significant result (OR 0.61, 95% CI 0.42 to 0.88). This made me wonder if there would have been any results whatsoever if the flexitarians had been combined with the omnivore group, as they should have been.

There were no significant differences between the groups for severity of illness or the self-perceived length of the illness. There was no claim that diet made any difference as to how bad Covid-19 might be – just whether someone got it or not.

The claim & plausible mechanism

The only claim being made in this paper is that those less omnivorous than the omnivores self-reported getting Covid-19 less often. There are so many issues – where to start?

Issue 5 – Covid-19 incidence was subjective. The paper did not mention testing for Covid-19 (albeit Covid-19 testing is notoriously unreliable). Participants may have tested or may have self-diagnosed from symptoms – we don’t know. This was all subjective.

Issue 6 – subjective Covid-19 incidence is a function of other factors. Given that people were being encouraged to test regularly world-wide, it is reasonable to assume that the incidence of Covid-19 in this study was a function of how often each person tested for Covid-19. The risk factors for Covid-19 included ethnicity, BMI and comorbidities. It is thus reasonable to assume that black, overweight or obese, people with existing conditions would have rightly been more concerned about Covid-19 than white, slim, healthy people and thus would likely have tested more. The more one tested, the more one would test positive.

This paper’s conclusion could have been those more worried about Covid-19 were more likely to get Covid-19 – with the mechanism being that they tested more often.

The researchers for this study needed to offer a plausible mechanism for why people who were ‘plant-based’ self-reported Covid-19 less often than those who were more omnivorous. The explanation offered in the discussion section of the paper was related to diet and immunity. The researchers hypothesised that due to “the high intake of some key nutrients and phytochemicals in groups following plant-based diets, it is plausible that a difference in immune status might be observed between plant-based and omnivorous dietary patterns.” This doesn’t make sense given the fact that the essential nutrients that the body needs are found more densely and in the right form in animal foods, not plant foods (Ref 7).

This hypothesis also failed when the data in Table 3 were taken into account. While Table 2 showed the ‘plant-based’ group consuming more fruit, vegetables, cereals, beans, nuts and seeds etc than omnivores, Table 3 was a comparison within the same diet group. Table 3 reported the food consumption of people in each group according to whether or not they self-reported getting Covid-19. If the ‘fruits and vegetables aid the immune system’ argument held, the omnivores eating more fruits and vegetables would have less incidence of Covid-19 than the omnivores eating less plant foods. This was not the case. There were no significant differences between those who self-reported Covid-19 in the omnivore group and those who didn’t in terms of any food intake – not cereals, beans, fruits, vegetables, nuts and seeds, vegetable oils, dairy foods, eggs or meat. The same applied to the vegetarian/vegan group and the same applied to the flexitarian group. There was no relationship between higher plant consumption (or lower animal food consumption) and incidence of covid-19 within groups.

This tells us that plants were not a plausible mechanism. This also tells us that we have the classic issue that we so often see in population studies:

Issue 7 – plant intake is describing a healthy person; the plant intake did not make the person healthy. The healthier person has less reason to suspect that they might have Covid-19 and thus less reason to test for it or to self-report it.

Issue 8 – there was no plausible mechanism for the claimed result.

In Table 3, there was one food group that did make a difference: sweets and desserts. In the omnivore group, those who consumed sweets and desserts less than once a week were significantly less likely to self-report having had Covid-19. There’s that healthy person again.

The limitations section clarified that “as an observational study, we are unable to confirm a direct causal association between diet and COVID-risk or infer specific mechanisms.” This did not stop the conclusion asserting “Our results suggest that a plant-based diet and mainly vegetarian diet may be considered for protection against infection with COVID-19.” This also did not stop the ‘box out’ on the front page claiming “In light of these findings and findings of other studies and because of the importance of identifying factors that can influence the incidence of COVID-19, we recommend the practice of following plant-based diets or vegetarian dietary patterns.”

Suggest, consider and recommend all you like; this study has not found what it claimed. A more robust hypothesis would be that self-reported Covid-19 is associated with concern about Covid-19 (people concerned about Covid-19 test for it and/or worry that they have it). Those with more reason to be concerned about Covid-19 (based on ethnicity, BMI, comorbidities etc) would be more likely to self-report Covid-19, therefore. These same people were more likely to be self-reported omnivores. This study supports the ‘Covid-19 concern’ hypothesis. It hasn’t even tested its own plant hypothesis, as there weren’t enough vegans in Brazil to do this.

References

Ref 1: https://www.zoeharcombe.com/2023/01/does-not-having-a-covid-jab-increase-your-risk-of-a-traffic-crash/
Ref 2: https://www.thetimes.co.uk/article/32c99624-2d2b-41fa-9dcd-cbb781f77738?shareToken=5bed16ac3c93325cdee97b4c8bd7c086
Ref 3: https://www.9news.com.au/world/vegans-and-vegetarians-less-likely-to-catch-covid-virus/91b94212-5579-4cc7-ad11-ffb5df5fbd87
Ref 4: Acosta-Navarro et al. “Vegetarian and plant-based diets associated with lower incidence of COVID-19.” BMJ Nutrition, Prevention & Health. 2024.
https://nutrition.bmj.com/content/bmjnph/early/2024/01/02/bmjnph-2023-000629.full.pdf
Ref 5: https://www.zoeharcombe.com/2023/10/red-meat-type-2-diabetes/
Ref 6: https://oosga.com/social-media/bra/
Ref 7: https://www.zoeharcombe.com/2023/07/the-basics-of-nutrition-macronutrients/
https://www.zoeharcombe.com/2023/07/the-basics-of-nutrition-vitamins/
https://www.zoeharcombe.com/2023/07/the-basics-of-nutrition-minerals/