An article entitled “Meat consumption and mortality – results from the European Prospective Investigation into Cancer [EPIC] and Nutrition” was published on 7 March 2013 in the BMC Medicine Journal. The full PDF is available here.
The article led to the usual sensational headlines across the world – the BBC; CBC News Canada; The Herald Sun in Australia and so on. The article itself claimed “We estimated that 3.3% of deaths could be prevented if all participants had a processed meat consumption of less than 20 g/day. Significant associations with processed meat intake were observed for cardiovascular diseases, cancer and ‘other causes of death’. The consumption of poultry was not related to all-cause mortality.”
The study details
The European Prospective Investigation into Cancer (EPIC) study is being conducted in 23 centres in 10 European countries (Denmark, France, Germany, Greece, Italy, Norway, Spain, Sweden, The Netherlands and The UK). Participants were recruited between 1992 and 2000. At recruitment men were 40-70 and women were 35-70 years old. The number of participants at baseline was 448,568.
Median follow-up time for this study was 12.7 years. During the follow-up period 26,344 participants (11,563 men and 14,871 women) died. Of these 5,556 deaths were from cardiovascular disease, 9,861 cancer, 1,068 respiratory disease, 715 digestive tract disease and 9,144 other causes.
The dietary questionnaire grouped the foods relevant to this study as follows:
Red meat (beef, pork, mutton/lamb, horse, goat);
Processed meat (“all meat products, including ham, bacon, sausages, small part of minced meat that has been bought as a ready-to-eat product”); and
White meat (poultry, including chicken, hen, turkey, duck, goose, unclassified poultry and rabbit).
Information was also gathered from participants on education, medical history (including history of stroke, myocardial infarction and cancer), alcohol consumption, physical activity, smoking status, height and weight.
Here are 10 things that you may find interesting about this study:
1) I’m a member!
At least I assume I’m a member. I was approached in the 1990s when I was a member of the Vegetarian Society. The Society had been approached to get many vegetarians on board. That would make the UK arm non-reflective of the general population, but it would give a more significant vegetarian population against which to compare. The reason I express any doubt that I’m part of the survey (because I still get follow-up questionnaires) is that I am female but I was not aged 35 or over at any point in the 90’s!
Here’s an example of the questionnaire that I filled in back in the 90’s and I’m pretty sure I’ve had two since – one in the last couple of years. Notice the meat entries on page 3. We were simply asked how often we had each meat (medium serving – whatever that is): never/less than once a month; 1-3 times a month; up to 6+ times a day. You will notice from the journal article that the meat consumption is listed extremely specifically – 0-9.9 g/day; 10-19.9 g/day and so on. How does this general frequency of consumption become such specific intake?
I was a vegetarian at the baseline point of the study (which is the reference point for this article). I have not been a vegetarian for three years. So, if I die now of a heart attack, I am classified as a vegetarian death. Had I been the biggest processed meat consumer in the world in the 1990s and then saw the error of my ways the day after filling in the survey, my heart attack would be attributed to my meat consumption. That’s how these long term epidemiological studies work.
2) Deaths cannot be prevented – we are all going to die
Ha ha. This one always makes me laugh.
Every study that tries to claim this nonsense should try to estimate by how long they think life might possibly be extended if you do something – in this case, eat less than 160 g/day of processed meat. Dr Malcolm Kendrick did the maths the right way round in the brilliant The Great Cholesterol Con and reduced the headline of “50,000 lives will be saved” to – if 10 million people, at very high risk of heart disease, took a statin for thirty years, aside from any dreadful effects they may suffer along the way, they might extend life expectancy by a couple of months.”
3) When the researcher said that the meat/processed meat eaters were less healthy overall, they weren’t kidding!
Table 1 in the paper gives the baseline information for people in the lowest and highest categories of meat/processed meat/poultry consumption. Page 4 of the study says: “Men and women in the top categories of red or processed meat intake in general consumed fewer fruits and vegetables than those with low intake. They were more likely to be current smokers and less likely to have a university degree”. (We know that there is a link between education level and health and longevity). They went on to say: “Men with high red meat consumption consumed more alcohol than men with a low consumption.”
That’s one way of putting it – in fact, for men, alcohol intake wasn’t just higher – it was up to three times higher in the highest meat intake group than the lowest intake. For men, again, smoking wasn’t just higher – it was up to three times higher in the highest meat intake group than the lowest intake. Table 1 also shows that:
– Men in the highest intake group of red meat were an average 6.4 years older than those in the lowest intake group (age causes death).
– Men in the highest intake group of red meat/processed meat and poultry consumption had higher BMI’s than the lowest intake group.
– Women were also heavier in every highest intake group vs lowest. Almost twice as many women, who were in the highest processed meat intake group, were current smokers.
– The calorie intake differences were striking. Men and women in the highest category of meat and processed meat consumption consumed one and a half times the calories of their counterparts in the lowest categories of meat and processed meat consumption.
So there were a number of big differences between the groups at baseline.
The researchers should have separated Paleo dudes from Pie monsters. Those who eat red meat, vegetables, fish, nuts and seeds are a world apart from those who eat pies, pasties and Peperami pizza!
Table 2 claims that adjustments have been made to Table 1’s data for age, gender, education, weight, height, energy intake, alcohol, physical activity, smoking – everything. I have to assume that this has been done although some of the differentials (three times) will take some ‘ironing out’.
4) The death rates are nothing to lose sleep over
Assuming that Table 2 is accurately reflecting meat consumption alone – there is one interesting observation to be made straight away. In every case (men and women, red meat, processed meat and/or poultry), it is better to consume some meat than none. The lowest intake group is not the best for mortality in any circumstance.
Table 2 has the figure that made the headlines – the “44% greater risk”. Page 4 notes: “Participants with an intake of 160+ g red meat/day had a hazard ratio (HR) of 1.37 compared with individuals with an intake of 10-19.9 g/day.” When the other factors were taken into account, the HR became 1.14. “The association for processed meat was stronger than for red meat. In the multivariate model [i.e. after allowing for other factors], the HR for high (160 g/day) versus low intake was 1.44.”
Table 2 also shows that (as the abstract confirmed) there is no association between poultry consumption and all-cause mortality in either direction. No positive or inverse association – there’s just no relationship. I will ignore poultry from now on, therefore (with just one wee mention to follow).
Let’s have a look at the death rates in my own little table to summarise numbers all over the paper:
|Death rate p.a. (*)||0.72%||0.36%||0.46%||0.10%||0.17%||0.19%||0.46%|
|1.44 ‘risk factor’||0.67%||0.14%||0.25%||0.28%||0.67%|
|1.38 ‘risk factor’||0.53%||0.11%||0.20%||0.22%||0.53%|
(* The death rate per annum takes the average – median – follow-up of 12.7 years to calculate the death rate in any one year during the study). The relevance of the 1.38 ‘risk factor’ will become clear with the next point.
The first thing to note is that men have twice the death rate of women.
The second thing to note is that the overall death rate is 0.46% in any one year. That’s 4.6 people in 1,000. And those people were aged 40-70 at baseline (men) and 35-70 at baseline (women) making them between 47-48 and 82-83 at median follow-up. If fewer than 5 in 1,000 of that age group were dying each year, I’m not going to worry.
The deaths for different causes were not broken down by gender. Taking them at the top level, the cardiovascular disease (CVD) death rate was 0.1% i.e. 1 in 1,000 each year and the cancer death rate (that’s the health topic that EPIC is all about) was 0.17% i.e. 1.7 in 1,000 each year (17 in 10,000). I’m really not losing sleep over those numbers.
All this nonsense that we hear “One in three people die of heart disease!” “One in three people die of cancer!” Of the people who died in this 12-13 year study, 5,556 out of 26,344 deaths were from CVD – that’s 21% – one in five. 9,861 out of 26,344 were from cancer – that’s 37% – that’s more like one in three. But it’s one in three deaths reported. Not one in three people. If you have two friends, it’s not the case that one of the three of you is going to die over the next year. You need 1,000 friends (aged between 35 and 83) to lose 1-2 of them over the next year.
Those “one in three” headlines are to make sure that you give money to heart and cancer charities. They are not fair and honest numbers.
5) The researchers cooked the books – part 1
Let’s look in more detail at Table 2 – it is fascinating…
The following extract looks at red meat and processed meat only (poultry shows nothing other than you should eat some rather than none). It takes the column claiming that all other factors have been accounted for (called HR (c) – hazard ratio (c)). It shows us where the 1.44 number comes from:
|Col 1||Col 2||Col 3||Col 4||Col 5|
|Red meat||N cases||HR (c)||Reset!||%||Cum %|
|0 to 9.9||3,175||1.07||1.00||12.05%||12.05%|
|10 to 19.9||2,774||1.00||0.93||10.53%||22.58%|
|20 to 39.9||6,459||1.01||0.94||24.52%||47.10%|
|40 to 79.9||8,935||0.99||0.93||33.92%||81.02%|
|80 to 159.9||4,639||1.03||0.96||17.61%||98.63%|
|0 to 9.9||6,236||1.04||1.00||23.67%||23.67%|
|10 to 19.9||4,683||1.00||0.96||17.78%||41.45%|
|20 to 39.9||7,301||1.03||0.99||27.71%||69.16%|
|40 to 79.9||5,997||1.09||1.05||22.76%||91.93%|
|80 to 159.9||1,904||1.21||1.16||7.23%||99.15%|
The number of cases (col 1) is in Table 2 and the Hazard ratios (col 2) are in Table 2. So columns 1 and 2 in the table above come directly from Table 2 in the paper. I have added in columns 3, 4 and 5.
Watch what they’ve done with the baseline (1.0). This has been taken as the second lowest meat intake group in all cases. Hence hazard ratios have not been calculated with reference to up to 9 g/day. They have been calculated relative to 10-19.9 g/day in the case of red and processed meat (and 5-9.9 g/day in the case of poultry).
My column 3 gives the hazard ratios had the lowest intake of meat been taken as the baseline (1.0) and not the second lowest – as suited the ‘researchers’.
– For red meat this means that every level of intake is better than the lowest level other than the highest level. i.e. any red meat consumption between 10 and 160 g/day is better than less than 10 g/day and >160 g/day. The single ‘best’ level of red meat consumption (the lowest HR) is 40-79.9 g/day.
– For processed meat, the 1.44 HR headline number would have been 1.38 if the true lowest intake were used as the baseline. The two categories covered by 10-39.9 g/day are better than the lowest intake. Only the 160+ g/day stands out as different. Which brings us to…
6) The researchers cooked the books – part 2
Check out columns 4 and 5 in my table above. Normally, studies like this put consumption into quartiles or quintiles. As an example, where a study divides the groups into quintiles, the researchers take the lowest fifth consumption of red meat and then the next lowest and then the middle of the five groups then the second highest and then the highest. This study has not done this. They have grouped the intakes unevenly with the following outcomes:
– For red meat, 98.6% of deaths are covered by the five lowest consumption groups. Only 1.37% of deaths occur in the 160 g/day group.
– For processed meat, 99.15% of deaths are covered by the five lowest consumption groups. Only 0.85% of deaths occur in the 160 g/day group.
This tells us a number of things:
i) The study groupings have been manipulated to present a result (I don’t see how else I can phrase this).
ii) The headline could (should?) have been “Not even 1% of deaths, over 12.7 years, occur among those who eat over 160 g processed meat per day”!
iii) 70% of deaths are occurring in the bottom three groups out of six of processed meat intake. 30% of the deaths are occurring in the top three intake groups. (i.e. you’re ‘better off’ in the upper half of consumption than lower half).
iv) Only the top (engineered group) intake of processed meat consumption shows an association worth looking at further and this accounts for fewer than 1% of the total deaths in 12.7 years.
This is like doing a survey about alcoholism and mortality and making the top group so small that it includes Billie Holiday and George Best and making headlines on this basis.
It’s bad science!
7) Other headlines could/should have been
– “Men are twice as likely to die as women!”
– “EPIC study shows you’ve got a one in a thousand chance of dying from heart disease.”
– “Whatever you do, eat meat. Between 10-160 g/day of red meat or your death rate goes up!”
– “70% of deaths in a 12-13 year study were in the lowest half of intake groups of processed meat consumers.”
– “Lies, damned lies and statistics: Researchers cherry pick a baseline to secure global headlines.”
– “Researchers can’t show a relationship between processed meat and death rates in 99.15% of their data, but they still base the headlines on 0.85%”
– “Bloggers are fed up having to waste time analysing nonsense manipulated by people with an end in mind.”
8) It always has to be said – association is not causation
Don’t forget, every study like this can only ever make allegations of association, not causation. There is no association between poultry and death rates or red meat and death rates. Credit where it’s due – at least this was possibly the first study to distinguish between red meat and processed meat – even though the media slipped into interchangeable language almost immediately.
When the correct baseline is taken into account only the top two processed meat intake groups need looking at further. If the consumption had been fairly allocated to quintiles then we would probably have nothing to look at (they didn’t isolate 0.85% of the data for nothing.)
IF any association is left at the end of all the manipulation, it’s still only association. No causation has been proven and there are far more obvious associations for me than the one being implied (processed meat and deaths).
i) People who eat processed meat eat processed food generally, so they are unhealthier all round (as all the cofactors confirmed); and ii) Processed meat comes in pies, pastries, bacon sarnies, covered with brown sauce/tomato sauce etc. The obvious relationship is that the pastry/flour/sugar that accompanies processed and not real meat is the intake causing harm.
The EPIC study has data for all food intake – why do they keep picking on meat? Why don’t they review the association between pies, pasties, bread, pastries etc and deaths or sugar and deaths or processed food and deaths? They are obsessed with demonizing meat to the point of missing the search for stronger and more important health connections.
9) It needs to be reiterated – absolute risk, not relative risk, is what matters
These studies always present relative risk and not absolute risk. The report claimed that those eating more than 160g of processed meat a day (don’t forget the bread and sauce) – were 44% (try 38%) more likely to die over a typical follow-up time of 12.7 years than those eating about 20g (note their careful use of language to exclude the up to 20 g/day group).
As I’ve shown in this post, the overall chance of dying in any one year was 0.46%. This is 4.6 people in 1,000. Even if any meat consumption could be proven as a risk factor (i.e. causation, not association and causation due to processed meat and not the pies and bread that processed meat comes in), the people in the highest intake group (accounting for not even 1% of the deaths) would have a death rate of 6.3-6.7 (38 vs 44%) people in 1,000. Hardly hold the front page now!
10) And finally…
Lest we forget that this EPIC study is all about cancer and that it has been trying to prove the importance of fruit and veg consumption since its inception, I have to share the comment on page 3 of the article: “Additionally adjusting for fruit and vegetable consumption did not appreciably change the observed associations and was not included in the main models.” i.e. fruit and veg consumption made no difference to the numbers.
The EPIC study discovered this in 2010 – fruit and veg intake makes no difference to mortality – as I have blogged on here. So much for 5-a-day.