In The Great Cholesterol Con, Dr Malcolm Kendrick analysed some World Health Organisation (WHO) data. The WHO has extensive data from almost 200 countries on more health measures than you could imagine - definitely worth a look one rainy afternoon. This is where Kendrick presented the world with two different Seven Country Studies. Kendrick took the seven countries with the lowest saturated fat intake and then the seven countries with the highest saturated fat intake.
Cholesterol-lowering-enthusiasts may need to read this twice - but he found: "Every single one of the seven countries with the lowest saturated fat consumption has significantly higher rates of heart disease than every single one of the countries with the highest saturated fat consumption."
As Kendrick's two unbiased seven country studies showed - there is not even an association between saturated fat and heart disease - let alone causation. However, Keys published his Seven Country Study and the rest, as they say, is history.
The next chapter in The Great Cholesterol Con goes on to look at cholesterol and heart disease (and overall death rates) and quoted many great studies where it is shown that lower cholesterol is associated with higher mortality. However, it did leave me thinking - having run the data on saturated fat and heart disease, let's just run all the data on the cholesterol and heart disease and get to the bottom of this hypothesis from all parts of the allegations.
Cholesterol & Mortality
It actually didn't take that long - not even a couple of hours one Saturday afternoon. You go to the WHO statisitics area of their web site ( https://apps.who.int/infobase/Comparisons.aspx ) and then pick data for cholesterol from risk factors (how judgemental to start with) and then look under: Global burden of disease (mortality); All causes; Non communicable diseases and then G Cardiovascular disease (shortened to CVD). CVD deaths include ischemic heart disease and cerebrovascular disease - that means fatal heart attacks and fatal strokes to lay people. You find the most recent year where you can get both sets of data to compare like with like. This turns out to be 2002. You download their very user friendly spreadsheet data (CSV) - cut and paste it into a spreadsheet application and then try to remember how the heck to do scatter diagrams!
The WHO data is split into men and women. I first did the scatter diagrams for average (mean) cholesterol levels and CVD deaths. Then I ran the Pearson correlation coefficient on these numbers. This gives us the term called "r". "r" tells us if there is some kind of a relationship: an r score of 0 would indicate no relationship; an r score of 1 would indicate a perfect relationship. A negative r score is called an inverse relationship e.g. the price of concert tickets is likely to be inversely related to the number of concert tickets bought - fewer tickets being bought at higher prices.
The "r" score for men revealed that there was a small relationship of 0.13 - however this relationship was inverse. The diagram and correlation shows that higher cholesterol levels are associated with lower CVD deaths and lower cholesterol levels are associated with higher CVD deaths. In women, the relationship is stronger - to the point of being meaningful. The r score was 0.52 - but, again, inverse. For women, higher cholesterol levels are quite significantly associated with lower CVD deaths and lower cholesterol levels are quite significantly associated with higher CVD deaths. Please note that I have added r squared on the graphs below (the spreadsheet application can do this for us) and it can confirm that you've got your r numbers right and r squared tells us the strength of any relationship we have observed.
All you need to do is to look at the lines going down to the right and wonder how on earth we ever got away with telling people that cholesterol causes heart disease. High cholesterol is associated with lower heart disease and vice versa - for all the data available in the world. High cholesterol is not even associated with high heart disease, let alone does it cause it.
It gets worse. I then kept the cholesterol information and changed the death rates to total deaths - all deaths from any cause - cancer, heart disease, diabetes, strokes - all deaths. You can see the diagrams for men and women again below. This time there is a significant relationship for both men and women: 0.66 for men and 0.74 for women - again inverse. There is a significant association between higher cholesterol levels and lower deaths and lower cholesterol levels and higher deaths for men and an even more significant relationship for women.
I removed the outliers (obvious ‘off-the-line' data points) and reran the data and it made not one iota of difference. With 192 data points showing such a strong trend, a couple, or a handful, of countries really makes NO difference.
This is serious. I've shown it to a couple of academics (Professor sort of things) with whom I've been having great debates, as I want to see what the view is from people who wholly believe the fat/cholesterol/heart/death hypothesis. (Kendrick talks in his book about what happened when he showed an intelligent colleague his two seven countries studies and the evidence was just dismissed instantly). It is most useful to know what the resistance arguments will be before starting to invite the resistance. The two arguments I got back were:
1) "Ah yes - but this is only an association."
Ah yes - but a) we changed global dietary advice back in 1977-1983 on the back of an association in Seven (carefully hand picked) Countries that miraculously became a causation even when the association was far from established and b) it is an association that's the opposite to the one that the world currently holds true and c) that's what epidemiology is supposed to be about - establish an association and then investigate if there could be any causation or useful learnings. So, let us go out with a new paradox - that high cholesterol is associated with low deaths and then see what dietary advice emerges.
2) "But that's total cholesterol - the key thing is the ratio of good to bad cholesterol."
Oh boy! The chemical formula for cholesterol is C27H46O. There is no good version or bad version. HDL and LDL are not even cholesterol, let alone good cholesterol or bad cholesterol. They are lipoproteins and they carry cholesterol, triglyceride, phospholipids and protein. Do you think that taxis are people? Or do you think that they are carriers of people (and luggage, and pets and fresh air and other things).