Dear JULES,
Meat & disease – again!
Executive Summary
* Yet another epidemiological paper has been published reporting associations between meat intake and different conditions.
* This note goes through the study in a specific and general way. This note tries to arm you with rebuttals for when people try to use nutritional epidemiology to attack your real food, nutrient dense diet.
* The study used a large differential to compare low and high meat intake. Associations between meat intake and higher incidence of heart disease, pneumonia, diverticulitis, colon polyps and diabetes were claimed. There were many more non-findings than findings. Attention was not drawn to this fact. There were also three inverse associations, which were not given the prominence of the positive associations. Higher meat intake was associated with lower incidence of atrial fibrillation, lower incidence of cataracts and lower incidence of iron deficiency anaemia.
* The usual three flaws of epidemiology applied: association not causation; relative not absolute risk; and the healthy person confounder. This note goes through them for this study and as general principles to use for rebuttals.
* As an additional flaw, this study failed to adjust for energy intake despite reporting that BMI was a significant confounder (when BMI was adjusted for, associations with meat and health risks were substantially reduced).
* This note closes with five options for rebutting the next nutritional epidemiology paper that is waved at you – if you want to do this.
Background
Five people have now contacted me about this week’s note. I didn’t address it immediately as I thought that people would be fed up with yet another epidemiological paper attacking meat or eggs or animal food generally. However, clearly there is an appetite (lolz) for continued dissection of such papers. A couple of the emails explained why – every time a paper like this makes headlines, those of us who eat nutrient dense foods (meat, fish, eggs, and dairy) get it waved in our faces along with vitriolic “there – I told you so” kind of comments. People want a retort to attacks such as this. This week’s note will cover the paper that readers wanted dissected, but I’ll cover it in a different way. I’ll try to give you the ammunition to have a rapid response to this, and similar, studies when they are next waved at you.
Introduction
I presented at a conference for the Institute for Optimum Nutrition (ION) in July 2013. It was one of the hottest days on record and the conference was held in an eco-building that didn’t believe in air con. The room was in the mid-30s (95f) by the time we arrived at 8am and it steadily rose throughout the day. I was thankfully on first, but my slot was 90 minutes. By the end of speaking to an informed and engaging audience for an hour and a half I was parched and dying to jump in a cold bath. The ION are an enlightened bunch and my messages about nutritional myths were well received, but there are always exceptions. A conventional dietician leapt to her feet to ask the first question. It was something about the diet heart hypothesis – fat will kill us and saturated fat even quicker – that kind of thing. I was wanting a drink rather than having to answer a question and then the wonderful Dr John Briffa jumped in. He said “you look like you need a drink Zoë, do you mind if I take this question?” I said, “please do” and stepped aside to get some tepid eco water.
John then taught me the best lesson I have learned at any conference. He didn’t answer the question – he asked questions back. The dietician had made sweeping statements about fat, cholesterol, and heart disease as part of her question (taking the opportunity to educate the audience lest they go away thinking that fat is not trying to kill us). John took the first part of her statement and asked, “what’s the evidence for that?” He took the second part and did the same and so on. She was then on the defence, not the attack. Every time she tried to answer him, she made another statement which he then challenged. He was completely charming throughout and he had her wrapped up in knots in barely a couple of minutes. My huge learning was – don’t defend, attack! The dietary guidelines are theirs to defend and ours to challenge.
We’ll go through the study before we come to the rebuttals, but we’ll try to ensure that the rebuttals don’t rely on you having to have a dissection before you rebut.
The study
Attacks on animal foods are so regular that I wonder if there’s collusion between US and UK researchers. US researchers (especially Harvard) constantly interrogate the Nurses’ Health Study and the Health Professionals Follow-up Study to come up with yet more tiny associations between intake of animal foods and different conditions. UK researchers are more likely to use the European Prospective Investigation into Cancer (EPIC) study or the UK Biobank study.
This latest paper used the UK Biobank study (Ref 1). It used data from 474,985 middle-aged adults who were followed up for an average of 8 years. The study noted up front that participants who consumed meat regularly (defined as three or more times a week) were less healthy people overall (they had “more adverse health behaviours and characteristics than participants who consumed meat less regularly”). The study also noted up front that “the positive associations observed for meat consumption and health risks were substantially attenuated after adjustment for body mass index (BMI).” i.e. after we adjusted for differences in BMI, associations between eating meat and poor health largely fell away.
The paper claimed that higher consumption of unprocessed red and processed meat combined was associated with higher risk of many different conditions. Notice how unprocessed and processed meat typically get lumped together to damn one in the name of the other. Watch out for definitions too. Unprocessed red meat in US studies typically includes burgers (probably the main source of red meat in the US diet). This doesn’t tend to happen in the UK – burgers are processed meat in UK studies. The supplemental file in this Biobank study reported that processed meats were bacon, ham, sausages, meat pies, kebabs, burgers, and chicken nuggets. However, this leaves meat dishes (curries, takeaways, ready meals etc) in the unprocessed red meat category. Grass fed beef and lamb is never isolated in these papers. What we eat has not been studied.
The specific claims were that – per 70 g/day higher intake of unprocessed red and processed meat – there was a 15% higher risk of ischaemic heart disease, 31% higher risk of pneumonia, 19% higher risk of diverticular disease, 10% higher risk of colon polyps and a 30% higher risk of diabetes (Ref 2). The intake differential used was large – 70 g/day extra was more than the difference between the lowest intake group (meat once a week or fewer times) and the highest intake group (meat five times a week or more). The supplemental file informed us that the lowest intake group averaged 10 g/day (try to even weigh that out) and the highest intake group averaged 79 g/day. The researchers picked a big difference to try to achieve results.
The three flaws in epidemiology
1) This is association and not causation.
The highest association that they came up with was a 31% higher association for meat intake and pneumonia. I looked for a plausible mechanism and the paper suggested that it might be possible that higher availability of iron in red meat might be associated with higher risk of infection. Clutching at straws came to mind. Far more plausible was the next suggestion: “It is also possible that hospital admission for pneumonia is a marker for co-morbidity and overall frailty.”
Ditto with the claimed 30% association between meat intake and diabetes. It makes no sense that a food with zero carbohydrate content will raise the risk of diabetes (Ref 3). The foods that are eaten with burgers, maybe (fries, bun, milkshake, fizzy drink) – but not the meat itself.
2) This is relative risk.
The absolute risk is always tiny compared to the relative risk. The relative risk makes the headlines. In this paper, there were 6,350 diagnoses of pneumonia among 459,309 people over 8 years. That’s 1 for every 579 person years. A 31% higher incidence is 1.31 diagnoses for every 579 person years.
3) The healthy person confounder.
Table 1 in the main paper told us that the typical person who consumed meat 0-1 times a week was female, university educated, in paid employment, with a lower BMI, less than half as likely to be smoking 15 or more cigarettes a day, more likely to be active, more likely to be a non-drinker, more than twice as likely to eat oily fish etc. As we always say, these factors are adjusted for, but you can’t adjust for an entire person/lifestyle.
Extra issues with this study
In my experience of analysing epidemiological papers, the three flaws of epidemiology always apply and there is usually at least one other flaw beyond these.
The main additional issue with this study is that it does not appear to have adjusted for energy intake. Energy intake is mentioned only once in the supplemental file and this was to report the extreme ranges of energy intakes used to exclude participants. If a woman reported an energy intake of below 500 or above 3,500 calories a day, she was excluded from the study. If a man reported an energy intake of below 800 or above 4,000 calories a day, he was excluded from the study.
The main paper reported energy intake three times – in the context of noting it as an issue. The paper reported that “confounding for energy balance could not directly be accounted for.” The paper’s best effort was to adjust for BMI and to admit that there may still be some confounding with energy intake. This is significant. We have reviewed so many epidemiological studies where the lowest intake group had an absurdly low energy intake, and the highest intake group had a much more plausible (or high) intake. This then makes a mockery of any comparisons between the lowest and highest intake groups because it is as if the lowest intake group has forgotten to declare most of their intake in the food frequency questionnaire (Ref 4).
The energy intake is especially important in this paper because the abstract (summary) of the paper notes that the results were much reduced when BMI was adjusted for. Maybe if energy intake had been adjusted for (as it should have been) the results would have been reduced even further (or become non-significant).
Before we leave this paper specifically, it’s important to note that many non-associations were found. Non-findings are also important, but they are rarely highlighted. The figures in the main paper confirmed that there was no association between an additional 70 g/day of unprocessed red and processed meat intake and all the following: hemorrhagic stroke; venous thromboembolism; varicose veins; hemorrhoids; GERD (gastroesophageal reflux disease); gastritis and duodenitis; hernia; enteritis and colitis; gallbladder disease; osteoarthritis; kidney stones; urinary tract infections; enlarged prostate; female genital prolapse; uterine fibroids; cellulitis; and carpal tunnel syndrome. (I have no idea why those conditions were chosen for review).
Finally, atrial fibrillation and flutter, cataracts and iron deficiency anaemia were significantly lower in people who consumed 70g day more of unprocessed red and processed meat.
Rebuttals
I used to be on a couple of governing bodies (boards), and it meant I attended black tie kind of dinners quite regularly. You’d be sat on a table of strangers and the usual opening to people on either side of you was “what do you do?” or “what brings you here?” Sometimes I shared that I was an HR Director and sometimes I shared that I wrote diet books. If the latter, the most common comment back was “so you’ll be watching what I eat then.” I honed a reply “I don’t mind what you eat. I mind that you know what you should eat.”
You don’t need to rebut anyone’s view on nutrition. If someone wants to wave a “meat is bad” paper at you, you can choose to smile nicely and ignore them. They are not your problem. However, if you do want a rebuttal, hopefully one of these may work:
1) My favourite rebuttal is called the Surgeon Captain Peter Cleave (1906-1983) rebuttal. Cleave famously said: “For a modern disease to be related to an old-fashioned food is one of the most ludicrous things I have ever heard in my life.” The question to ask is “Please can you explain to me how the food that we have been eating since time began is responsible for today’s diseases?”
2) Starting off with agreement helps. “I agree with you that processed anything is not good for health – be it meat or bread – but please show me the evidence that grass fed meat is trying to harm me?”
3) Questions follow from the three flaws of epidemiology:
a) “Please can you tell me the difference between association and causation?” This is to make them realise that they cannot claim meat causes disease. (A handy example to have to hand is the sharks and ice cream one… deaths from shark attacks are associated with ice cream sales. Shark attacks do not cause people to eat ice cream any more than ice cream sales cause shark attacks. However, on a sunny day, more people go in the water – where they can get attacked by a shark – and more people eat ice cream.)
b) “Please can you tell me what the absolute risk difference was?” (Trial lawyers have a rule “never ask a question to which you don’t know the answer”, but in this case you’ll be fine because the person attacking you will not know what the absolute risk difference was.) You can always remember the number in this note to use as an example (“The last study I looked at had an absolute difference of 1 for every 579 person years. A 31% higher incidence is 1.31 diagnoses for every 579 person years- who cares?”) or use a generic example “the difference is often about 1 in 1,000 and 25% higher is 1.25 in 1,000 – are you bothered?”
c) “Please describe for me the person most likely to be consuming the most (processed) meat vs the person consuming the least.” If they can’t do this, you can do it for them. The person who consumes the least is female, younger, more educated, and more affluent. She doesn’t smoke, drinks moderately (if at all), eats oily fish, and she has a yoga teacher on the days she’s not horse riding! (wink)
Then you give the Gary Taubes close. Gary is the person who has best described what nutritionally epidemiology is all about. Nutritional Epidemiology essentially compares an overall healthy person with an overall unhealthy person. Nutritional Epidemiology then claims that – if only the unhealthy person ate like the healthy person, they would be as healthy as the healthy person! That is nutritional epidemiology in a nutshell (lolz again).
I have used the following two images in presentations to make this point. Nutritional epidemiology tries to argue that if only the second couple ate like the first family, they would be as healthy as the first family. You could always just keep these two images on your phone as the only rebuttal you may ever need for vitriolic paper waving!
Until the next time
All the best – Zoë
References
Ref 1: Papier et al. Meat consumption and risk of 25 common conditions: outcome-wide analyses in 475,000 men and women in the UK Biobank study. BMC Medicine. March 2021. https://bmcmedicine.biomedcentral.com/articles/10.1186/s12916-021-01922-9
Ref 2: Confidence intervals were reported as: “…higher consumption of unprocessed red and processed meat combined was associated with higher risks of ischaemic heart disease (hazard ratio (HRs) per 70 g/day higher intake 1.15, 95% confidence intervals (CIs) 1.07–1.23), pneumonia (1.31, 1.18–1.44), diverticular disease (1.19, 1.11–1.28), colon polyps (1.10, 1.06–1.15), and diabetes (1.30, 1.20–1.42)…”
Ref 3: https://www.zoeharcombe.com/2019/07/plant-based-diets-diabetes/
Ref 4: Remember this paper reviewed in November 2020. The average calorie intake of group 1 was 1,112. The average calorie intake of group 4 was 2,828 – 2.5 times higher. The dietary intake data were thus completely unreliable, but the researchers made claims despite this. https://www.zoeharcombe.com/2020/11/eggs-diabetes-and-chinese-adults/
