On March 4th 2014, articles started to appear on line. “Animal protein-rich diets could be as harmful to health as smoking” said the Guardian. The Daily Mail captured the age dimension more accurately with “Eating lots of meat and cheese in middle age is ‘as deadly as SMOKING‘”.
The source of the media headlines is this article in Cell Metabolism. The full article is available on free view.
The study reviewed data for 6,381 adults aged 50 and over (average age 65) using American public health data (NHANES III). The participants were followed for up to 18 years, giving 83,308 person years worth of data.
Average calorie intake was reported as 1,823 per day (which already suggests under-reporting). This was 51% carbohydrate (by calorie intake); 33% fat intake and 16% protein intake. Most of this protein intake (11 of the 16%) was reported as protein from animal sources.
Instead of dividing the people into three equal groups, the percentage of calorie intake in the form of protein was used to allocate subjects to one of three groups:
– High protein intake was categorised as more than 20% of calories from protein (1,146 people);
– Moderate protein intake was categorised as 10-19% of calories from protein (4,798 people);
– Low protein intake was categorised as fewer than 10% of calories from protein (437 people).
This is interesting in itself. Normally groups are divided so that equal numbers of people fall into each group. The most common grouping used is tertiles (3 groups), quartiles (4 groups) or quintiles (5 groups). If subjects had been divided into tertiles, 2,127 people would have been in each of the 3 groups and the percentage of protein would have been the outcome – not the input (e.g. we may have found that one third of people had protein intake lower than 14% of calories; one third had protein intake between 14-18% and the final third had protein intake higher than 18%). The method used here is not wrong, but we are not comparing equal groups and this will have a difference when relative risk comes into play – as it will.
Association between protein and mortality
This is a direct quotation from the article (my emphasis): “Using Cox Proportional Hazard models, we found that high and moderate protein consumption were positively associated with diabetes-related mortality, but not associated with all-cause, CVD [cardiovascular], or cancer mortality when subjects at all the ages above 50 were considered.”
i.e. when we looked at the 6,381 over 50 year olds there was not even an association with protein intake and all-cause mortality, or CVD mortality, or cancer mortality.
There was a relationship with diabetes mortality and protein intake, but the numbers were so tiny (one death from diabetes in one group) that this was not considered important.
And that could have been the headline – “There is no association between protein intake and mortality” – but then there would be no headline.
After finding no overall association, the researchers spotted a pattern with age and split the information into participants aged 50-65 and participants over 65. They then found (direct quotation again): “Among those ages 50–65, higher protein levels were linked to significantly increased risks of all-cause and cancer mortality. In this age range, subjects in the high protein group had a 74% increase in their relative risk of all-cause mortality (HR: 1.74; 95% CI: 1.02–2.97) and were more than four times as likely to die of cancer (HR: 4.33; 95% CI: 1.96–9.56) when compared to those in the low protein group.”
This means that there was an equal and opposite result for the over 65 group. If all the participants together showed no association and one section of the group are then separated out to show a positive association, the remaining section of the group must have a negative association. That’s the law of averages. Sure enough, the 3,342 people over the age of 65 were far less likely to die from any cause if they were in the moderate or high protein intake group. Cancer mortality for the low protein group was two and a half times the cancer mortality for the high protein group.
Cardiovascular (CVD) mortality was about the same for the high protein group and the low protein group for the 50-65 year olds. CVD mortality was then much lower for the moderate protein intake group than for the low protein intake group for 50-65 year olds. For the over 65 year olds, the highest protein intake group was the best one to be in for all-cause mortality, cancer mortality and CVD mortality.
The fact that the headlines chose to claim “protein will kill you in middle age” rather than “protein will save you in old age” just highlights the nonsense and bias.
The usual errors
There are two facts that every study fails to clarify:
1) Association does not mean causation (just because we observe singing in the bath, it does not mean that being in the bath causes singing any more than singing causes being in the bath); and
2) Relative risk is a poor measure when absolute risk can be reported instead. (You can double your chance of winning the lottery by buying 2 tickets. Your relative chance is twice as high as it was before. Your absolute chance was 1 in 14 million and is now 2 in 14 million. You’re still not going to win the lottery!)
This study has absolute risk numbers and should share them. There could be 4 deaths in 1,000 people from cancer in the high protein group and 1 death from cancer in the low protein group. This meets the headline “four times as likely to die of cancer”, but it’s hugely different to having a 1 in 1,000 chance of dying vs. a 1 in 250 chance of dying – neither of which is going to lose you any sleep at night. I’ve emailed Dr Longo to ask for the raw data on death rates to see what the absolute risk is. (And remember – this is still only in the 50-65 age group and will be the other way round in the over 65s).
Protein vs animal protein
The study claims to have adjusted for protein in general vs. animal protein to conclude that animal protein is the harmful factor and not protein per se. Call me suspicious, but I always check for conflicts of interest and the lead researcher, Dr Longo, has declared interests in (actually, he’s the founder of) L-Nutra – a company that makes ProLon™ – an entirely plant based meal replacement product.
So the study would have us believe that animal protein increases cancer mortality in people between the ages of 50 and 65, but then magically reverses this ‘causation’ at 65 such that you’d better be in the high protein group or you’ll be dropping dead like flies. This just doesn’t make sense.
I could go into a discussion of quality animal protein (meat, eggs and dairy from pasture living animals) vs. processed animal protein (fast food burgers with white buns and ketchup, mass produced chickens, low-fat sugared yoghurts) and so on, but this cannot explain why animal protein across all the people surveyed would be allegedly harmful before the age of 65 and protective thereafter. By the way – do eat quality animal produce and don’t eat processed anything (meat or otherwise), but that’s just a general health golden rule. It can’t explain this study.
Of mice, not men
The researchers turn away from John and Jane Doe to Mickey and Minnie Mouse to try to explain the results. Dr Longo is well known for his mice experiments (he was one of the chaps whom Michael Mosley interviewed when he did his Horizon programme on Intermittent Fasting).
The researchers thus did some experiments on mice. They gave some 18 week old male mice a diet with either low (4-7%) or high (18%) protein intakes. (We don’t know if either fat or carbohydrate made up the difference for the different protein intakes). They implanted melanoma cells in Mickey Mice (gave them cancer in effect) and then looked to see how the cancer progressed over the next 39 days, while the mice were fed either high or low protein intakes. Tumour incidence was reported as 100% for the high protein group and 90% for the low protein group after 25 days. The discussion that followed centered around a term you may recall from the Horizon programme – IGF-1 – Insulin-like Growth Factor.
This could have led to the headline – “Male mice given cancer cells get cancer”, but it is being used as an explanation for the observations in the human study. The hypothesis being put forward is that protein intake increases IGF-1 and that IGF-1 helps our bodies grow and it may therefore help cancer to grow. We have not even proven if protein intake determines IGF-1 in humans and therefore the hypothesis falls over at the first hurdle. The theory then suggests that IGF-1 falls with age. So are they then saying that protein over the age of 65 doesn’t impact IGF-1 and doesn’t therefore impact growth or cancer? It still just doesn’t make sense.
The final twist was that the researchers found no significant difference when they gave animal vs. plant protein to mice. So this cannot justify the headlines condemning meat and cheese. Additionally – bang goes the ProLon™ PR!
What should we take from this?
* Humans don’t need that much protein. As a rule of thumb we need approximately 1 gram of protein per 1 kilogram of body weight. Body builders and pregnant women may benefit from more, but we don’t need that much.
Having said this, protein has a substantial metabolic advantage over carbohydrate and fat and can help with weight loss as a result[i]. Hence more than 1 gram per kilogram of body weight is not a problem – so long as the intake comes from real food and not from fake shakes.
* Protein is in virtually every food provided by nature (oils and sucrose being the only two exceptions and they’re not really food). Nature provides fat/protein combinations – meat, fish, eggs, dairy products – and carbohydrate/protein combinations – grains, pulses, fruits, vegetables. Rarely do foods have fat/protein and carb in good measure (nuts and seeds being the exceptions). Why would nature put protein in everything if it were out to get us?
* This study has made an interesting observation and that’s it. It has not provided a plausible explanation. As for the smoking comparison – this is a stunt to grab headlines – not appropriate for researchers who want to be taken seriously. Smoking presents an absolute risk – provide the same numbers for my grass-grazing roast dinner if that’s what you’re claiming.
* Should you ditch meat and dairy as a result of this? Not unless you want to deprive yourself of essential fats, complete protein and invaluable quantities of vitamins and minerals. The golden rule of diet remains unchanged and that is – eat real food! This means meat, eggs and dairy from pasture living animals; fish; nuts and seeds; vegetables and fruits in season. Enjoy whole grains and starchy veg only if you are normal weight – limit these fattening foods if not. Red wine and dark chocolate and what more could a man want? Or mouse!
p.s. just had a thought a couple of hours after this was posted. Where are the vast majority of deaths going to be among the 6,381 people who were over 50 when the study started? In the 50-65 year old group or in the 65+ year old group? The latter of course. So, notwithstanding that we have no plausible mechanism, the ‘advantage’ of animal protein is thankfully in the group that will benefit most! Looking forward to the raw data…
[i] Eric Jequier, Institute of Physiology, University of Lausanne, Switzerland, found that the thermic effect of nutrients (thermogenesis) is approximately 6-8% for carbohydrate, 2-3% for fat and 25-30% for protein. I.e. approximately 6-8% of the calories consumed in the form of carbohydrate are used up in digesting the carbohydrate and turning it into fuel available to be used by the body. In contrast, 25-30% of the calories consumed in the form of protein are used up in digesting the protein and turning it into fuel available to be used by the body. (Eric Jequier, “Pathways to Obesity”, International Journal of Obesity, (2002).)