This week’s study is fabulous. It didn’t make the headlines – I chanced upon it because I get email alerts when one of my papers is cited and this one cited the 2015 article in BMJ Open Heart showing that the dietary fat guidelines were issued without RCT evidence base at the time. The title of this paper caught my eye immediately: “Global correlates of cardiovascular risk: A comparison of 158 Countries.”
The (Czech) researchers had reviewed intake of 60 foods/food groups across 158 countries and looked for relationships between those foods/food groups and cardiovascular disease deaths, raised blood pressure and raised blood glucose. Some of the foods were individual items, like oranges/mandarins or sunflower oil. Others were food groups, like meat or cereals. The researchers also created some groups to combine, for example, carbs from cereals and starchy roots or animal fat/protein. These proved very interesting.
The full review can be seen below.
The paper used a technique called “correlation”, which establishes if there is a relationship between two things and whether or not that relationship is positive or negative (i.e. both go up together or one goes up as the other goes down). It also gives us a number that tells us how strong the relationship is. The most interesting relationships found by the researchers were:
1) Raised cholesterol (defined as ≥ 5.0 mmol/L) was inversely associated with cardiovascular disease deaths in most countries. i.e. higher cholesterol levels were associated with lower deaths and lower cholesterol levels were associated with higher deaths. I found the same using World Health Organisation data for all 192 countries of the world in 2010.
2) The relationship between raised blood pressure and deaths from cardiovascular disease (CVD) was stronger in women than men. We assume that blood pressure is an important measure in all patients. It may not be equally informative.
3) The 158 countries were split into regions. “America” included 29 countries in that region. The food group “oilcrops” was strongly associated with deaths from CVD. The strongest negative relationship in this region was with eggs. This suggests that vegetable oils may be harmful and eggs may be helpful.
4) While “oilcrops” generally were associated with CVD deaths, sunflower oil specifically was associated with a number of markers of poor health: CVD deaths in European men and women and raised blood glucose in men and women in the 116 non-European countries.
5) Carbohydrates from cereals and starchy roots were particularly strongly associated with CVD deaths in Europe – for men and women.
6) There were several foods/food groups that had strong negative relationships with CVD deaths in European men and women i.e. they might be helpful to consume. All of the following were associated with lower CVD deaths: Total fat; Animal fat; Total protein; Animal protein; Total meat; Meat protein; Fruits; Coffee; Oranges and Cheese.
7) Smoking is a strong factor with CVD deaths in men, but not women. The researchers had discovered and explained this in a previous paper of theirs. Apparently women smoke more in wealthy countries where their health is generally better.
8) Glucose from cereals and starches may impact blood glucose levels more than refined sugar. The associations between cereal intake and raised blood glucose were stronger and more consistent than with refined sugar! Dr David Unwin has produced impactful info-graphics showing that cereals and bread are essentially glucose – he needs to produce more of these!
The obvious argument is that rich people eat animals and poor people eat cereals and therefore this study just tells us that rich people are healthier. The researchers addressed this by adjusting for health expenditure in each country and cereals still turned out to be associated with CVD deaths and animal foods still turned out to be negatively associated with CVD deaths. And we’re told to base our meals on starchy foods and to avoid animal foods…
The Full Review:
The study this week didn’t make the headlines. I only came across it because I get email alerts when any of my papers are cited and this paper cited the 2015 article in BMJ Open Heart showing that the dietary fat guidelines were issued without RCT evidence base at the time (Ref 1).
People can cite your papers to support a point that they are trying to make, or to critique your work, but it’s a handy alert to set up because you receive papers written by people working in a similar field. When I saw the title of this paper “Global correlates of cardiovascular risk: A comparison of 158 Countries”, I was intrigued and my intrigue was wonderfully rewarded.
This is a fabulous paper. It’s on open view (Ref 2). It’s 39 pages long and it was written by a Czechoslovakian team, which always blows me away. The work itself is impressive – to publish it in a second language is awe inspiring.
What the researchers did
How many times have we looked at the attacks on real food (red meat, eggs, dairy products etc) emanating from the Harvard School of public health and lamented the fact that they never seem to investigate processed food or even all food? This Czech study has done what we have been longing for a research team to do for some time.
The Czech researchers took average food intake of 60 food items between 1993 and 2011 for 158 countries and matched this to indicators of Cardiovascular Disease (CVD). The countries comprised: 42 from Europe; 47 from North Africa, Asia & Oceania; 29 from America and 40 from Sub-Saharan Africa. The researchers also assessed obesity rates, health expenditure and life expectancy to get as full a picture as possible of the health of each nation, alongside the food that is consumed. The CVD indicators prioritised were not the usual cholesterol levels (although these turned out to be fun), but raised blood pressure (defined as systolic ≥ 140 or diastolic ≥ 90 mmHg), CVD mortality and raised blood glucose (defined as ≥126 mg/dL (7.0 mmol/L) history of diabetes, or on medication).
The primary outcome grabbed my interest too. The researchers noted that many factors varied between the very different countries assessed – from short life expectancy to religious customs. However, they found that “regardless of the statistical method used, the results always show very similar trends and identify high carbohydrate consumption (mainly in the form of cereals and wheat, in particular) as the dietary factor most consistently associated with the risk of CVDs.”
The food categories analysed
The paper noted that this research had been made possible by an online version of the Food and Agriculture Organization Corporate Statistical Database (FAOSTAT) (Ref 3). The researchers noted the limitations of such data. Food availability does not equal food consumption. We can never know what is thrown away, but there is no reason to think that this would disproportionately affect one food over another. There would also be differences in data accuracy across 158 countries, but, again, no reason to think that this would be unique to a food or a country (Note 4). The minimum average consumption of any food was set at 5 grams a day. If a food were consumed less often than this, it wasn’t included in the analysis. If items were missing from too many countries, these were also excluded, as the objective was to compare country health markers with country food intake. Examples of foods where data were missing from many countries were bananas, onions, palm oil and soybean oil. The latter two were a particular shame as another oil – sunflower oil – did throw up some interesting findings.
The study was able to review 60 food items. Fourteen of them were basic indicators of fat and protein intake, or their combinations (animal fat, animal protein, animal fat and animal protein, total energy, etc.). Another six basic indicators were calculated by the researchers as a % of total calories: i) energy from carbohydrates in cereals (CC energy); ii) energy from carbohydrates in starchy roots (SRC energy); iii) the combination of CC energy and SRC energy; iv) energy from carbohydrates and alcohol; v) energy from alcohol; and vi) energy from all plant food (excluding alcohol). The addition of energy from alcohol and carbohydrates was rationalised by the fact that distilled alcohol is made from starchy carbohydrates. I wasn’t convinced that this was a useful category and so I excluded it from findings that I’ve quoted in this post.
This left 40 items – 16 described major food groups e.g. alcohol, cereals, fruits etc – the remaining 40 were individual food for which data were available for the countries. It was noted that the most consumed individual foods were wheat (198 g/day), potatoes (100 g/day), rice and beer (both 79 g/day). That was the first staggering finding – across 158 countries in the world, the average consumption of wheat totalled almost 800 calories per day (198 grams at approximately 4 calories per gram).
All health marker information came from the World Health Organisation (Ref 5). Life expectancy data came from the World Bank (Ref 6).
The most interesting relationships/associations found
The primary statistical technique used in the study and the one mentioned in the title was “correlation” (The Pearson linear correlation was used). Correlation analysis produces a number between -1 and +1. Minus one is a perfect negative relationship; plus one is a perfect positive relationship. A negative relationship would exist between, for example, the number of layers worn and temperature: the higher the temperature, the fewer layers worn. A positive relationship would exist between ice cream sold and temperature: the higher the temperature, the more ice cream sold.
The paper was absolutely full of correlation numbers – some were more surprising and thus more interesting than others. (I’ll use the words relationship, or association, instead of correlation, from now on, as they are words with which we’re more familiar):
1) Raised cholesterol (defined as ≥ 5.0 mmol/L) was inversely associated with CVD deaths in most countries.
I found the same using World Health Organisation data for all 192 countries of the world in 2010 (Ref 7). For men and women in Europe, North Africa, Asia & Oceania and America, raised cholesterol was associated with lower CVD deaths. The inverse relationship was particularly strong in Europe where it was an identical -0.79 for men and women. That’s strong.
2) The relationship between raised blood pressure and deaths from CVD was stronger in women than men.
The relationship between raised blood pressure in women and deaths from cardiovascular disease (CVD) was 0.69, which is fairly strong. The relationship between raised blood pressure in men and deaths from CVD was 0.42. That’s not so strong. This was explained by life expectancy. Men in a number of the 158 countries may not live long enough to need to worry about blood pressure or dying from CVD!
3) There were far fewer clear and strong relationships between foods and raised blood pressure, CVD mortality and raised blood glucose in men and women from the 29 American countries than in the men and women from the 42 European countries (See Note 8 for explanation).
The only strong and significant positive relationship between American CVD mortality and any of the 60 foods/food groups was with “Oilcrops”. The relationship was 0.65 in men and 0.6 in women. The only negative relationship, stronger than 0.5, for American CVD mortality was with eggs. This was the case with both men (-0.51) and women (-0.58) i.e. higher egg consumption was associated with lower CVD deaths.
4) Sunflower oil manufacturers would not like this paper.
While Oilcrops generally were associated with CVD deaths, sunflower oil specifically was associated with a number of markers of poor health: CVD deaths in European men and women and raised blood glucose markers in men and women in the 116 non-European countries. Even after adjustment for smoking and health expenditure in each country, sunflower oil intake was positively associated with higher blood pressure and higher CVD deaths in all men and to an extent in women.
5) Carbohydrates from cereals and starchy roots were particularly strongly associated with CVD deaths in Europe.
The strongest relationship for European men between any food/food group and CVD deaths was with energy from carbohydrates in cereals and starchy roots. This relationship was 0.77 (Note 9).
The strongest relationship for European women between any food/food group and CVD deaths was with energy from carbohydrates in cereals and starchy roots. This relationship was 0.82.
6) There were several foods/food groups that had strong inverse relationships with CVD deaths in European men and women.
I noted every inverse relationship stronger than (-)0.7 (which is high). The following were found to be strongly associated with lower CVD deaths: Total fat; Animal fat; Total protein; Animal protein; Total meat; Meat protein; Fruits; Coffee; Oranges and Cheese.
Some of the relationships were almost as high as will ever be seen in statistical analysis. In European women, the relationship between intake of total fat and animal protein and low CVD deaths was 0.87. It was 0.86 for total fat and total protein and it was 0.85 for total fat.
7) Smoking is a strong factor with CVD deaths in men, but not women.
In a smaller sample of 115 countries, smoking was added to the analysis to see how the food ‘risk factors’ of cereals and wheat ranked against this known unhealthy behaviour. For men, the relationship (correlation) between smoking of any tobacco product and CVD mortality was 0.53. The relationship with cereals was 0.42 and for wheat alone it was 0.33. Not so far behind smoking.
In contrast with men, women’s smoking prevalence was negatively related to CVD deaths (-0.4). The researchers shared that this had been explained in a previous paper of theirs (Ref 10) – the incidence of smoking among women generally is much lower and women smoke more in wealthy countries where their health is generally better.
8) Glucose from cereals and starches may impact blood glucose more than refined sugar.
Refined sugar was associated with raised blood glucose in men and women in all countries outside Europe combined, but not in Europe. In European men and women, cereals and starches, rather than refined sugar, were associated with raised blood glucose. Doctors like David Unwin have produced info-graphics showing how much sugar there is in our daily bread, or cereal. This study supports the need to continue to educate starch pushers about this fact.
Are cereals a maker or a marker of bad health?
Cereals, wheat and energy intake from cereals and starchy carbohydrates are positively associated with deaths from cardiovascular disease. Total fat and total protein, especially from animals, are inversely associated with deaths from CVD. There is clearly an argument to be made that more affluent people can afford steak, rather than bread, and more affluent people tend to be healthier. This argument would have cereal intake as a marker, not maker, of bad health.
The researchers answered this argument by analysing the 60 foods/food groups and health markers for all 158 countries together adjusted for health expenditure per capita. (They also adjusted for smoking). This is an ideal way to adjust for any ‘affluence = health’ confounder. The researchers found that the relationships between cereals, wheat, and energy from cereals and starchy roots combined, maintained a positive relationship with CVD deaths in men and women even after the adjustment for health expenditure.
This is why the researchers concluded: “regardless of the statistical method used, the results always show very similar trends and identify high carbohydrate consumption (mainly in the form of cereals and wheat, in particular) as the dietary factor most consistently associated with the risk of CVDs.”
That’s the bottom line. Cereals, wheat and carbohydrate intake from starchy foods remain associated with deaths from CVD even when health expenditure adjustments are made. This finding is even more troubling alongside the fact that we average a staggering 800 wheat calories a day. Total fat, animal fat and animal protein remain inversely associated with CVD deaths, even after health expenditure adjustments.
Historical literature is full of references to the gentry eating meat and peasants eating bread. There is still a sense today that steak is for the rich and pasta is for the poor. This study suggests that for the poor to eat pasta and the rich to eat steak would be a gross health inequity. It also suggests that the dietary guidelines of so-called developed nations – telling us, as they do, to base meals on starches and to avoid animal foods – are a gross health injustice. Cereals, wheat and starches are not associated with good health. We need to help poorer nations to have access to better food and we need to stop advising affluent nations to eat like impoverished ones!
For those of you who like tables
CVD mortality in men
|Strongest positive relationship(s)
Energy from carbs in cereals & starchy roots (0.77)
The only significant finding was:
|Strongest negative relationships(*)
Total fat & animal protein (-0.81)
(*) All above 0.7 listed for Europe
CVD mortality in women
|Strongest positive relationship(s)
Energy from carbs in cereals & starchy roots (0.82)
The only significant finding was:
|Strongest negative relationships(*)
Total fat & animal protein (-0.87)
Ref 1: Harcombe Z, Baker JS, Cooper SM, et al. Evidence from randomised controlled trials did not support the introduction of dietary fat guidelines in 1977 and 1983: a systematic review and meta-analysis. Open Heart 2015.
Ref 2: Grasgruber P, Cacek J, Hrazdíra E, Hřebíčková S, Sebera M. Global Correlates of Cardiovascular Risk: A Comparison of 158 Countries. Nutrients 2018. (https://www.ncbi.nlm.nih.gov/pubmed/29587470) or (http://www.mdpi.com/2072-6643/10/4/411)
Ref 3: Food Balance Sheets. Available online: .
Note 4: Where there were obvious differences, these often had a religious explanation e.g. a country where religion dictates that pork and/or alcohol, for example, is not consumed.
Ref 5: Noncommunicable Diseases. Available online: .
Ref 6: TheWorld Bank. Data Catalog. Available online: .
Ref 7: http://www.zoeharcombe.com/2010/11/cholesterol-heart-disease-there-is-a-relationship-but-its-not-what-you-think/
Note 8: I corresponded with the authors to understand why this was the case. The explanation was that there will be “some problems with the accuracy of data and even with the fact that men from many American countries simply do not reach the age, when raised blood pressure starts to manifest itself. As a result, when you compare countries with so different population characteristics and different quality of healthcare, the correlation coefficients are inevitably lower than in Europe, or even insignificant.”
Note 9: The carbohydrates and alcohol combined group was fractionally higher, but, as mentioned above, I didn’ consider this a useful group to include.
Ref 10: Grasgruber, P.; Sebera, M.; Hrazdira, E.; Hrebickova, S.; Cacek, J. Food consumption and the actual statistics of cardiovascular diseases: An epidemiological comparison of 42 European countries. Food Nutr. Res. 2016.