* A study has just been published with findings from a crossover trial where 36 participants consumed plant-based ‘meat’ vs animal meat 2-3 times a day.
* The study was funded by the plant-based ‘meat’ company Beyond Meat.
* The primary outcome of interest was TMAO (trimethylamine-N-oxide). TMAO is known to increase following consumption of meat, eggs and especially fish. Choosing TMAO as the primary outcome of interest was highly likely to find in favour of plant-based ‘meat’.
* TMAO did increase, but interestingly only in the 18 people who did the animal phase first, followed by the plant phase. Those who did the plant phase first and the animal phase second, had no significant increase in TMAO.
* The paper implied that TMAO is harmful but did not provide evidence for this. There is limited evidence (from population studies) of an association between TMAO and cardiovascular disease (CVD), but causation cannot be claimed.
* A number of other researchers, who have investigated TMAO and disease, have concluded that it might be a marker for something else but “TMAO is not the bad guy of heart attacks and stroke that we’d previously thought.”
* The most significant ‘black swan’ in the TMAO/disease debate is that fish is the richest source of (preformed) TMAO and fish is not claimed to be a risk factor for death or avoidable disease – usually on the contrary.
* This was a well-executed study, with multiple measurements, but it found nothing of real significance. The funder will likely be content, however.
I spotted this week’s story on twitter in Dr Christopher Gardner’s timeline. I had some communications with Dr Gardner in February 2018 about his low carb vs low fat study (Ref 1). Gardner himself is a “longtime vegetarian” (Ref 2). Gardner is the senior author on a paper that has just been published in the American Journal of Clinical Nutrition (Ref 3). He tweeted a summary of the findings as follows:
“Better with plant: TMAO, LDL-C, weight
“Better with animal: nada
”No diff: IGF-1, HDLC, TG, gluc, ins” (Ref 4).
The conflicts are important to note up front. In what follows, CDG is Chris D. Gardner: “CDG received funding for the study from Beyond Meat in the form of an unrestricted research gift made to Stanford University.”
Gardner also asked in a tweet “can we get some love for the study name/acronym? Study With Appetizing Plantfood – Meat Eating Alternative Trial SWAP-MEAT” (Ref 5). So, a study funded by meat alternatives was entitled SWAP-MEAT (for alternatives). I detect confirmation bias.
The aim of the study was to compare the effect of consuming plant-based ‘meat’, as opposed to animal meat. The study involved 36 participants, of whom 67% were female and 69% Caucasian. The average age was 50 and the average BMI was 28. The participants were randomised to either the plant or animal intervention for the first phase. They were instructed to consume ≥2 servings per day of plant-based ‘meat’ or animal meat for 8 weeks and then to swap over to the other intervention. They were supposed to keep all other foods and drinks as similar as possible between the two phases. This was a crossover trial, therefore. There was no washout period – the participants started the other intervention as soon as they had finished the first (this may have been important – we’ll come back to this).
The primary outcome chosen was TMAO (we’ll explain this in a minute). The secondary outcomes chosen were fasting insulin-like growth factor 1, lipids, glucose, insulin, blood pressure, and weight.
Gardner’s tweet was a good summary. There were no differences in fasting insulin-like growth factor 1, HDL-Cholesterol, Triglycerides, glucose, insulin, or blood pressure, and so we can ignore these from now on.
Let’s quickly dismiss the other two findings (cholesterol and weight were statistically significantly different), so that we can concentrate on TMAO.
Cholesterol: Average LDL-Cholesterol was lower at the end of both diets than at baseline (this was not highlighted) (Ref 6). It had fallen by more at the end of the plant phase of the diet than at the end of the animal phase of the diet. Table 2 in the Gardner paper listed the ingredients in the plant products and the animal products. The animal products contained beef, pork, and chicken with some salt, herbs, and spices and not much else. The plant products contained pea protein isolate, soy protein isolate, canola oil, sunflower oil, coconut oil, rice flour, rice protein, bean protein, sugar, maltodextrin, potato starch, potassium chloride and all sorts. Some of the plant products contained plant sterols; the animal products don’t. These will lower cholesterol, but increase the risk for heart disease (Ref 7).
Weight: Average weight was higher at the end of both diets than at baseline (this was not highlighted). Baseline weight, on average, was 78kg. Weight was, on average, 78.7kg at the end of the plant phase and 79.6kg at the end of the animal phase. Weight at the end of the plant phase was 1kg lower than at the end of the animal phase. This was statistically significant.
Table 1 reported the serving size, calories, and macronutrients for the 7 plant options and the 6 animal options. The average calorie intake for the plant options was 164; it was 206 for the animal options. The participants consumed an average of 2.5 servings a day of plant-based ‘meats’ and 2.6 servings a day of animal meats. Participants could choose different plant and animal options, but all options were delivered to participants and if all were consumed, daily calorie intake could be 125 higher in the animal group. The supplemental file reported that the animal group consumed, on average, 130 calories more a day (from the 24-hour recall interview with the dietician.) While the 3,500 calorie theory is a myth, such differentials could make a difference over 8 weeks.
TMAO: The main result claimed was that average (mean) TMAO concentrations were significantly lower overall after the plant phase than the animal phase. However, the order of diets had a significant effect. TMAO concentrations were significantly lower for plant among the 18 people who did plant second, but not for the 18 people who did plant first (the numbers for these differences are in the table at the end of this note). We’ll come back to this too. First, we need to understand what TMAO is and what is being claimed…
What is TMAO?
Trimethylamine N-oxide (TMAO) is an organic compound in the class of compounds called amine oxides. TMAO is produced by gut microbes from dietary components such as choline, betaine, and carnitine. These nutrients are found most abundantly in animal foods, such as red meat, eggs, fish, and poultry. As bacteria in the gut feast on choline, betaine, and carnitine, they produce a substance called trimethylamine (TMA). The liver takes that TMA and converts it into TMAO.
You can immediately see that choosing TMAO as the primary outcome for a study comparing plant and meat intake ensured that a result would be found. TMAO is far more likely to be made following red meat intake than after plant-based meals. A study funded by fake ‘meat’ chose the primary outcome of the experiment to find in favour of fake ‘meat’.
Is TMAO harmful?
The paper described TMAO as an “emerging risk factor for CVD” (cardiovascular disease). It cited articles that reported that red meat intake raises TMAO concentrations, but it didn’t cite any articles that claimed that TMAO concentrations are harmful.
The press release about the article on the Stanford web site stated:
“Gardner calls TMAO ‘an emerging risk factor,’ meaning there seems to be a connection between higher levels of TMAO and an increased risk of cardiovascular disease, but the connection has yet to be definitively proved. Two precursors to TMAO, carnitine and choline, are found in red meat, so it’s possible that individuals who regularly eat beef, pork or lamb for dinner will simply have higher levels of TMAO. ‘At this point we cannot be sure that TMAO is a causal risk factor or just an association, Gardner said” (Ref 8).
We could leave it there and note that the researchers knew that red meat would raise TMAO, but they presented no argument that this was harmful. However, red meat, TMAO, mortality and CVD is an interesting and important subject, so let’s look a bit closer…
There is one meta-analysis, from 2017, which examined TMAO and CVD (Ref 9). It was a meta-analysis of population (cohort) studies and so it can only establish associations, not causation. The paper reported that higher circulating TMAO was associated with a 23% higher risk of cardiovascular events (data from 5 studies) and a 55% higher risk of all-cause mortality (data from 10 studies). Those are reasonable sized numbers but they’re still association, not causation and they’re still relative not absolute risk. There were no raw data numbers in the meta-analysis. A 55% difference could be 1.55 in 1,000 vs 1 in 1,000. Additionally, this meta-analysis was strongly determined by studies that examined TMAO intakes at the extreme and such diets typically have extreme and implausible dietary intakes, as we most recently showed in this post (Ref 10).
The claimed association between TMAO and CVD has been explored further. Dr James DiNicolantonio, with whom I have written papers, reported the observed association with the counter argument that nutritional intake of TMAO and its precursors (choline etc) don’t correlate with cardiovascular risk (Ref 11). i.e. claiming that ‘dietary components raise TMAO raises CVD’ fails if the diet items don’t raise CVD directly. (A number of us have been making the same argument about the diet-heart hypothesis for many years. The argument that ‘saturated fat raises LDL-cholesterol causes heart disease’ fails when the saturated fat/heart disease direct connection fails).
Fish is the richest dietary source of preformed TMAO and yet associations between fish and CVD/death are not claimed. A meta-analysis of population studies concluded that dietary choline has no significant impact on risk for CVD or CV mortality and yet choline is a precursor to TMAO (Ref 12). Likewise, a 2016 meta-analysis failed to associate consumption of eggs – a rich source of choline – with increased CVD risk (Ref 13).
James’ conclusion from this (and the rest of the research in this very interesting paper), was that elevated TMAO “must be viewed as a marker for other factors that both raise TMAO and confer increased risk for vascular disease and diabetes.” James and his co-researchers hypothesised that hepatic (liver) insulin resistance was the issue. Their recommendation was that management of metabolic syndrome and reversal of obesity would alleviate the risk associated with elevated TMAO.
A much cited 2016 article by Velasquez et al called “Trimethylamine N-Oxide: The Good, the Bad and the Unknown” noted that TMAO levels are determined not just by consumption of dietary substances (choline, betaine, and carnitine) but by gut flora, liver enzymes and kidney function (Ref 14). Velasquez et al also noted that the ARIC (Atherosclerosis Risk in Communities) and EPIC (European Investigation into Cancer) studies did not report increased CVD risk with increasing dietary intake of choline. They also noted the fish anomaly – that the food richest in TMAO is not claimed to be a CVD risk. Indeed, the He et al meta-analysis concluded that fish consumption is inversely associated with fatal coronary heart disease (Ref 15). Velasquez et al also questioned if TMAO was a marker or a bystander rather than an issue.
Landfald et al had a paper published in 2017 called “Microbial trimethylamine-N-oxide as a disease marker: something fishy?” (Ref 16). They also challenged the fish paradox – that TMAO has gained attention as a possible biomarker for CVD and yet fish and seafood contain considerable amounts of TMAO and are generally seen as cardioprotective. Their overall conclusion was: “We suspect that the TMAO story may be a red herring.”
The most recent paper on this topic, from Australian researchers, published in May 2020, concluded that “Essentially, we found no direct link between TMAO and the extent of atherosclerosis either in humans or in the laboratory” and “This showed TMAO is not the bad guy of heart attacks and stroke that we’d previously thought” (Ref 17). This team reported that levels of TMAO could be increased with so-called ‘healthy’ (high fibre) and so-called ‘unhealthy’ (high choline) diets and that TMAO levels correlated with plaque instability rather than the extent of atherosclerosis (Ref 18).
The twist in the finding
Back to the SWAP-MEAT paper. Having explained the cholesterol and weight findings, the main result claimed was the TMAO increase in the animal phase. We now know that this was no surprise and, indeed, the research team knew that this would happen. This passage was in the introduction to the paper: “Recent trials have reported that red meat intake raises TMAO blood concentrations.”
But remember, the issue over the order of interventions? TMAO levels were (statistically) significantly lower for the plant phase among the 18 people who did plant second. They were not significantly lower for the 18 people who did plant first. Table 3 in the paper told us the baseline TMAO levels for participants; Table 4 told us the results after 8 weeks of each phase; Figure 3 gave further detail on the results at the end of each phase. The TMAO levels were as follows (Note 19):
TMAO went down and then up, but stayed lower than baseline when plant was done first. These results were not statistically significant and could have happened by chance. TMAO went up and then down, ending up lower than baseline when animal was done first. These results were statistically significant. The paper debated whether there should have been a washout period, but suggested that the absence of a washout period may have revealed this ‘order effect.’ The researchers hypothesised that the 8-week vegetarian diet could have altered microbiomes “in such a way as to prevent the production of TMAO when the animal phase came second.” However, the analysis of the microbiomes in the study found no such changes.
An alternative hypothesis is that the randomisation just happened to get three times as many men in the animal first group. The average weight of those men just happened to be approximately 13kg greater than the men in the plant first group, giving an average BMI of 29.7 vs 26.2 (from Table 3). We may have observed a metabolic/obesity response, with TMAO as the marker – as hypothesised by James and colleagues.
Intake of meat, eggs and especially fish should raise TMAO levels. This was known to the researchers. TMAO was well chosen as the primary outcome in a study funded by fake ‘meat’. The paper implied that higher TMAO was a bad thing, but didn’t offer evidence for this. There is not much evidence elsewhere for this either. Any claim that TMAO is harmful is seriously undermined by the fish paradox. The fish issue aside, TMAO is at best a marker of a bad diet and/or a marker of poor health. This study was a significant undertaking (testing everything from fasting insulin to self-reported flatus), but one is left thinking that it’s of value to fake ‘meat’ and no one else.
Ref 1: https://www.zoeharcombe.com/2018/02/low-carb-vs-low-fat/
Ref 2: https://med.stanford.edu/news/all-news/2020/08/plant-based-meat-versus-animal-meat.html
Ref 3: Crimarco et al. A randomized crossover trial on the effect of plant-based compared with animal-based meat on trimethylamine-N-oxide and cardiovascular disease risk factors in generally healthy adults: Study With Appetizing Plantfood—Meat Eating Alternative Trial (SWAP-MEAT). AJCN. August 2020. https://academic.oup.com/ajcn/advance-article-abstract/doi/10.1093/ajcn/nqaa203/5890315?redirectedFrom=fulltext
Ref 4: https://twitter.com/GardnerPhD/status/1293240268792016897
Ref 5: https://twitter.com/GardnerPhD/status/1293243915244576771
Ref 6: Baseline LDL-C on average was 122 mg/dL. LDL-C was, on average, 109.9 mg/dL at the end of the plant phase and 120.7 mg/dL at the end of the animal phase.
Ref 7: Harcombe & Baker. Plant sterols lower cholesterol but increase risk for coronary heart disease. OnLine Journal of Biological Sciences. 2014. https://thescipub.com/abstract/ojbsci.2014.167.169
Ref 8: https://med.stanford.edu/news/all-news/2020/08/plant-based-meat-versus-animal-meat.html
Ref 9: Qi et al. Circulating trimethylamine N-oxide and the risk of cardiovascular diseases: a systematic review and meta-analysis of 11 prospective cohort studies. J Cell Mol Med. 2018. https://pubmed.ncbi.nlm.nih.gov/28782886/
Ref 10: https://www.zoeharcombe.com/2020/05/dietary-fat-mortality/
Ref 11: DiNicolantonio et al. Association of moderately elevated trimethylamine N-oxide with cardiovascular risk: is TMAO serving as a marker for hepatic insulin resistance. BMJ Open Heart. 2019. https://openheart.bmj.com/content/6/1/e000890
Ref 12: Meyer & Shea. Dietary choline and betaine and risk of CVD: a systematic review and meta-analysis of prospective studies. Nutrients. 2017. https://www.mdpi.com/2072-6643/9/7/711
Ref 13: Alexander et al. Meta-analysis of egg consumption and risk of coronary heart disease and stroke. J Am Coll Nutr. 2016. https://www.tandfonline.com/doi/full/10.1080/07315724.2016.1152928
Ref 14: Velasquez et al. Trimethylamine N-Oxide: The Good, the Bad and the Unknown. Toxins (Basel). 2016. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5127123/
Ref 15: He et al. Accumulated evidence on fish consumption and coronary heart disease mortality: A meta-analysis of cohort studies. Circulation. 2004. https://pubmed.ncbi.nlm.nih.gov/15184295/
Ref 16: Landfald et al. Microbial trimethylamine-N-oxide as a disease marker: something fishy? Microb Ecol Health Dis. 2017. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5444358/
Ref 17: https://www.hri.org.au/news/heart-study-debunks-meat-metabolite-myth
Ref 18: Yen Chin Koay et al. Plasma levels of trimethylamine-N-oxide can be increased with ‘healthy’ and ‘unhealthy’ diets and do not correlate with the extent of atherosclerosis but with plaque instability. Cardiovascular Research. May 2020. https://academic.oup.com/cardiovascres/advance-article-abstract/doi/10.1093/cvr/cvaa094/5817823?redirectedFrom=fulltext
Note 19: Notes: P values for significance were
– p=0.23 (not significant) for plant first (2.5 ± 0.4) compared with animal second (3.0 ± 0.6)
– p=0.007 (significant) for animal first (6.4 ± 1.5) compared with plant second (2.9 ± 0.4)
– p=0.012 (significant) for the aggregation of the two phases – plant (2.7 ± 0.3) vs animal (4.7 ± 0.9).