Academic ResearchDiabetesRed Meat

Red meat & diabetes?

There is an article widely reported in the media today (11 August 2011). The original research was published in the American Journal of Clinical Nutrition. You can see the abstract for free and the article then costs $12. I bought the article, so that I can comment on the full picture and not the abstract and certainly not on the basis of the usual dreadful reporting that goes on in the UK media – if not elsewhere.

The study

The team looked at three studies for which there was food questionnaire information available:

– 37,083 men in the Health Professionals Follow Up Study (1986-2006);

– 79,570 women in the Nurses Health Study I (1980-2008);

– 87,504 women in the Nurses Health Study II (1991-2005).

In total 13,759 incidents of type 2 diabetes were recorded from 4,033,322 person-years of follow-up. That’s a 0.34% incident rate to start with. Hardly justifying the headline “Diabetes threat from two slices of bacon a day.”

The conclusion

You can see the conclusion in the abstract: “The pooled HRs (95% CIs) for a one serving/d increase of unprocessed, processed, and total red meat consumption were 1.12 (1.08, 1.16), 1.32 (1.25, 1.40), and 1.14 (1.10, 1.18), respectively. The results were confirmed by a meta-analysis (442,101 participants and 28,228 diabetes cases): the RRs (95% CIs) were 1.19 (1.04, 1.37) and 1.51 (1.25, 1.83) for 100 g of unprocessed red meat and for 50 g of unprocessed red meat, respectively. We estimated that substitutions of one serving of nuts, low-fat dairy, and whole grains per day for one serving of red meat per day were associated with a 16–35% lower risk of T2D. ”

There must be an error with the two words that I have highlighted in red.  Is the article really saying that eating 100g of unprocessed red meat has 1.19 risk (presumably relative to not eating any unprocessed red meat) but that eating 50g of unprocessed red meat has a 1.51 risk? i.e. claiming that people who eat 100g of unprocessed red meat have ‘a 20% greater risk of diabetes’ but people who eat half this amount have ‘a 50% greater risk’?! Do they mean unprocessed at the first mention and processed at the second? (I’ve emailed Frank Hu – watch this space. Update – email back by return, fair play! Confirmation that this IS an error and AJCN will be asked to correct).

The overall conclusion is: “Our results suggest that red meat consumption, particularly processed red meat, is associated with an increased risk of T2D.”


1) On P2 of the full study we have the significant error at the outset. Under the heading “Assessment of meat consumption”, we have this telling passage: “Questionnaire items in unprocessed red meat consumption included ‘beef or lamb as main dish’, ‘pork as main dish’, ‘hamburger’ and ‘beef, pork or lamb as a sandwich or mixed dish’, and items on processed red meat included ‘bacon’, ‘hot dogs’, and ‘sausage, salami, bologna, and other processed red meats.'”

We reach the fundamental issue, which renders the study futile, in this one passage. Real food supporters define unprocessed meat as that which has been naturally reared and processed meat as any and every other meat. Take the Weston Price Foundation definition of real meat for any study. Real unprocessed meat comes from animals that have been living their entire life freely outdoors grazing on (ideally fast growing) grass in rain and sunshine. These animals must have been eating grass, which they are designed to eat and not grain which they cannot digest. Unless they have been chewing the cud, which, as ruminants they are designed to do and pre-digesting vitamin D blessed grass for those who cannot digest cellulose – humans – there is no point in consuming them.

Hamburgers are not real meat. Presumably a lamb curry takeaway qualifies as “beef or lamb as main dish” – this is not real meat, as real food supporters would define it. The fundamental point of the study is about red meat – processed and unprocessed. To make any relevant claims, the study should have looked at those who eat no meat (every single other factor unchanged), those who eat real meat (every single other factor unchanged) and those who eat processed meat (every single other factor unchanged). The fact that other factors cannot be held constant is one of the major reasons why the UK Food Standards Agency had to admit (in the context of fat and heart disease studies):

“However, the ideal controlled dietary trial for prevention of heart disease (a long-term intervention trial with differing levels of saturated fatty acids and measuring coronary disease endpoints) has not yet been done and it is unlikely ever to be done”.

Plus – the second critical point related to the so-called unprocessed meat – what is “beef or lamb as main dish” eaten with? rice? potatoes? carbs? What are hamburgers eaten with? burger buns? chips? ketchup? carbs? What are “beef, pork or lamb as a sandwich” eaten with – bit of a clue there – bread, likely hydrogenated fat margarine, emulsified mayonnaise, and, no doubt, more ingredients in the bread alone than in the varieties of real meat available to humans.

It doesn’t matter what the survey did or concluded next – they did not measure real meat vs processed meat or isolate this consumption from any other macro nutrient.

2) Diabetes is a condition related to blood glucose levels and insulin – either the body’s failure to release insulin to respond to a rise in blood glucose levels (type 1) or because cells are unresponsive (resistant) to insulin released (type 2). The macro nutrient most relevant to diabetes therefore is carbohydrate. Fat has no relevance and the relevance of protein is debatable, but negligible compared to that of carbohydrate. So, what is most likely to have any impact on diabetes – the processed (don’t call it unprocessed) hamburger, or the bun, fries and ketchup? To claim an association between one part of food intake and not the whole is meaningless.

The report even notes that they conducted a sensitivity analysis  with “adjustment for other major dietary variables (whole grain, fish, nuts, sugar-sweetened beverages, coffee, egg, potatoes, fruit and vegetables, all in quintiles).” Why not take the dietary questionnaires (however unreliable these notoriously are) and run an association with all dietary carbohydrate and incidence of diabetes over time. Biscuits, cakes, confectionery, bread, sugary cereals, pizza – the 400 calories of sugar and 700+ calories of flour that the average American eats daily. Are those eating more than their share getting more than their ‘share’ of diabetes?

This study was called “Red meat consumption and risk of type 2 diabetes…” Notwithstanding that it is about processed meat and even more processed meat and association with type 2 diabetes, when will we see the study “Carbohydrate consumption and risk of type 2 diabetes…”?

3) To be fair – the study did not claim causation. Research papers rarely do. They propose association and let the media run the “just two rashers of bacon can increase the risk…” headlines. Association does not mean causation – either way round. We should not be able to jump from association to ‘meat consumption causes diabetes’ any more than we should be able to jump from association to ‘diabetes causes meat consumption’. Not only is causation usually assumed, a direction of causation is assumed. The people who developed diabetes may have all worn blue socks – does that mean that wearing blue socks increases the risk of anything?

4) Table 2 of the report negates the idea that there is any trend. Data is presented for the three studies, for five different levels of meat intake. The serving sizes are determined relatively, by quintile, so, as an example, for the Health Professionals Follow-up Study for what the study assumes to be unprocessed red meat, the five different serving categories are 0.17 servings per day, 0.43, 0.65, 0.94 and 1.44 servings per day). Finally, three models are presented for each of these different meat intake levels, by study:

– an age adjusted model;

– a model adjusted for age, alcohol consumption, physical activity, smoking, ethnicity, menopausal stage for women, family history of diabetes/hypertension/hypercholesterolemia, quintiles of total calories and a dietary score for diabetes that the team made up (more on that below);

– a model adjusted for everything above and BMI.

All of this is then done for their opinion of unprocessed red meat, processed red meat and total red meat (that should be total processed meat). You can see that this is indeed a multi variate model!

There is not a steady trend between meat intake and the incidence of diabetes in every model variant of every study. For the Health Professionals Follow-up Study (HPFS), increased (what they call) unprocessed red meat intake from 0.65 servings per day to 0.94 servings per day was shown to have a reduced incidence of diabetes in all three models.

It is not clear how the dietary score for diabetes impacted the assumptions and therefore results, but the team “created a low diabetes risk score as a diet low in trans fat and glycaemic load and high in cereal fiber and the ratio of polyunsaturated to saturated fat.” What the ratio of two fats, which nature can put naturally in foods in different proportions and man can put very unnaturally in foods in different proportions has to do with the risk of diabetes I do not know.

5) The final point to make, as is the case with all presentation of numbers from studies to achieve maximum impact (and likelihood of media reporting) is that there are lies, damned lies and statistics. If you buy two lottery tickets each week, I can halve your chance of winning by allowing you only to buy one. Your odds of winning are now (for example) one in fourteen million instead of two in fourteen million – still absolutely naff all. However I have halved your chance of winning. Imagine I halved your ‘chance’ of developing diabetes in a similar, playing with numbers, kind of way…

Let’s take some actual numbers from Table 2 as an example:

Processed red meat, age adjusted model, from the HPFS…

– 0.02 servings of processed red meat per day (one serving is 28g so that’s half a gram of processed meat!) is associated with 340 incidences of diabetes in 138,550 person years. That’s an incidence rate of 0.25% or 1 in c. 400 people.

– 0.12 servings of processed red meat per day (that’s 3 grams – can any of you actually measure your intake to that level?) is associated with 409 incidences of diabetes in 121,238 person years. That’s an incidence rate of 0.34% or 1 in c 300 people.

(This is raw data – adjusted for age only – not adjusted for smoking, exercise, calorie intake, family history of diabetes, weight etc i.e. impossible to isolate two different minute meat intakes and assume that this is the only difference.) This aside, this is presented as – if at 0.02 servings per day you have a 1.00 ‘risk’ of developing diabetes, at 0.12 servings per day you have a score of 1.38 i.e. a 38% higher risk. There’s the headline “3 grams of bacon a day and you have a 40% greater risk of diabetes.” There’s how numbers are played with to frighten the life out of you and to make sure that you have sugary cereal for breakfast instead of eggs from grass living chickens.

The bottom line

This is a study about processed meat and even more processed meat and observed associations between intake of each and incidence of diabetes. No biochemical pathway is proposed for how fat/protein is supposed to impact a condition of glucose/carbohydrate handling deficiency. The obvious connection between the buns, chips and ketchup being consumed with the hamburger has not been made.

To let the media have the last word – the Mail article tells us how it is: “There is now widespread evidence that red meat drastically (my emphasis – couldn’t resist it) increases the likelihood of major health problems including heart disease, strokes, and some types of cancer.”

Accuse modern, processed meat of all this and more – all modern food in fact. Hang processed ‘food’ generally for crimes against human health – heart disease, strokes, cancer, diabetes, Alzheimers – all modern illness. Throw the book at the man-made horrors. But, if we really think that nature put all the essential fats, essential amino acids, full range of B vitamins, fat soluble vitamins, iron, calcium, magnesium and zinc in red meat and was trying to kill you at the same time, we wouldn’t be here today!

9 thoughts on “Red meat & diabetes?

  • What is your rationale for defining hamburger & lamb curry as not “real meat”? The other ingredients mixed into the patty or in the curry, should not disqualify the meat component, in my opinion.

    • Hi Lindy
      I just think if you have a category for real meat and then one for processed, the real meat should be just that. Who knows what is in the burger or curry? I also think that there’s a reason these have been put in ‘real meat’ and that’s because that category wouldn’t have shown any results had they not been in there. Sadly I’ve been doing this too long to trust any of them!
      Best wishes – Zoe

  • Pingback: The Usual Bad Science

  • Bit late reading your article and all the comments. My husband’s type 2 diabetes is fully controlled by diet. A lot of red meat and ‘bad’ processed red meat, ie ham, bacon, sausage and so on. Plenty of green vegetables and also some cheese and full fat milk. Limited carbs. No more than 50 or 60 gramms a day.

    The optician finds no signs of diabetes in his eyes, at all and NO signs of the alleged retinal bleed about nine months ago. Was this a hoax by the NHS retinal screening unit. He has highish cholesterol as statins made him very ill indeed. May have contributed to the onset of T2D. However his bloods and liver tests, all of it, are spot on. He is 71 years of age.

  • “Diabetes is a condition related to blood glucose levels and insulin – either the body’s failure to release insulin to respond to a rise in blood glucose levels (type 1) or because cells are unresponsive (resistant) to insulin released (type 2). The macro nutrient most relevant to diabetes therefore is carbohydrate. Fat has no relevance and the relevance of protein is debatable, but negligible compared to that of carbohydrate.”

    This is almost exactly parallel to saying that cholesterol is a kind of fat, and that therefore, consumption of carbohydrates is irrelevant. Type 2 Diabetes comes from insulin resistance. It’s not clear what causes insulin resistance. But there is at least some evidence that elevated free fatty acids has an impact on it, and also on the functioning of the pancreas. And even that says nothing about diet, because there is not any necessary relationship between free fatty acids and the consumption of fat.

    People have been misinformed for years about the relationship between dietary fat and cholesterol. Assuming that there is no relationship between diabetes and the consumption of fat, simply because fat doesn’t trigger the production of insulin, may be exactly the same sort of mistake. Body chemistry may be more complicated and surprising than you think.

  • There is an excellent article by Dr Jack Kruse (a neurosurgeon) linking magnesium deficiency to obesity and diabetes:

    “Low intracellular Magnesium (Mg) levels that causes the genesis of insulin resistance peripherally. And we have known it for a long time but have done little in clinical medicine to treat it. This is why so few people know about it. Peripheral leptin and insulin resistance (at muscles and fat cells) occurs first for this to happen but the depletion of Mg always predates insulin resistance. So when blood insulin rises, you lose intracellular Mg and this feedback loop makes the peripheral cells even more insulin resistant because we can’t make insulin or let it act properly on target cells. Simultaneously, arteries constrict decreasing blood flow and glucose and insulin to the target tissues. This worsens the Mg deficits going forward. This feedback loop continues daily in all diabetics until the brain becomes leptin resistant (LR) at the hypocretin neurons. Once this occurs, epigenetic switches are thrown in the hypocretin neurons for leptin and insulin signaling at the brain level…….again this occurs because of the poor Mg concentration at these cells and our DNA/RNA then become “hard wired” for a diabetic metabolism at the brain level. This allows for the human brain to become exquisitely sensitive to the dopamine reward of foods. The reward of foods are important when the brain is already LR because this is an outflow only tract of the hypocretin neurons in the hypothalamus. … The real issue is what is happening to the main nucleus in the brain that controls all energy balance as insulin wrecks the periphery and the liver. That nucleus is the hypocretin neurons. … Obesity is a brain disorder. It also fully explains why no macronutrient is the major cause of obesity. And it ironically, shows why carbs appear to be the major macronutient that causes obesity. Its because of insulin’s affect on Mg directly. I fully understand why we have the Kitavins and the Taubian view of obesity. Its called an incomplete out of context story that can be fully understood by biochemistry.

    There are other ways to obesity and LR too……see the omega six pathway with a heavy preponderance of trans- omega six fats.”

  • There are too many confounding factors in this observational study for it to of any use.

    BTW 204,157 people studied, of which 13,759 developed T2D is 6.74%. The person-years of follow-up is not relevant. The population is the people not the person years.

  • Another self reporting study that shows maybe a change, yet on self reporting engineering studys we know that they are out by 40 to 50% when scaled to a measurable parameter, ie total products sold, so the any self reporting study is somewhere between useless and dangerous. Traffic studies are real fun to analysis, and then project into the future and spend millions on roads based on those projections. Oh well, that is the real world.

    I often see the 30% under reporting of consumption around in food studies, but I think that is lite. We cannot measure the calories in actual food to 20%, so now we guess at weight of portions, especially at home.

    I used a scales for a while, as an ex-obese person, and I suggest the self reporting error is plus 100% some of the time. Atwater said approximate, and we are now better today, in terms of meaningful measurement. The computer generates two decimals points in tests that are plus/minus 20%. OK. And then compare 1.19 and 1.51 % as different risks. Caca el toro.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.