Reader Alexander Thorn in the UK put together some impressive work using data from the British Heart Foundation and emailed it to me the other day. Let’s take a look. You can click on all the graphs for the much larger "lightbox" image, and/or click on the links below each graph to download the files.
You can download the linked file here. Explains Alex:
Taking the same European countries and plotting the incidence of the various diseases against each country ranked in ascending order of the amount of saturated fat consumed as a percentage of total calories — from 3.9% for Bosnia to 15.5% for France — I got some very definite trend lines — all sloping downward toward the countries with the greatest saturated fat intakes (France being the highest, of course).
Unlike Keys, I used all the countries with the relevant data available (no cheating!). While there is the odd outlier on both sides of the divide — some countries with high disease rates and high sat fat and some with low disease rates and low sat fat — the vast majority, I think, echo the overall trend of more disease for less sat fat that is eaten as a percentage of total caloric intake.
But Alex wasn’t satisfied there. He noted that in using DALY figures, population size of the countries would make a difference. He explains.
Just thought I’d better clarify some points re the graph. The figures for CHD, stroke and other CVD are actually DALY figures – that is Disability Adjusted Lost Years – and it is per 100,000 of the countries’ populations (I hadn’t registered this distinction at first).
The DALY figures were already ‘age-adjusted’ and include not only the years lost due to mortality from the particular disease but also disability. It is calculated – as follows – from relatively accessible incidence data (ICD codes). Years of Lost LIfe (YLLs) is based on the inputs life expectancy at age of death and age at death, while Years Lost to Disability (YLDs) is based on the inputs duration of disease/injury, disability weight of disease/injury and percentage of long-term cases.
The only possible criticism I can foresee is that the population size of each country is obviously going to have a bearing on the DALY statistics. So I have just re-entered all of the data into another spreadsheet and made adjustments for the population size of each country. This has obviously made the graph less ‘exciting’ to look at but the trend lines for all diseases still slope downward toward the countries with the highest saturated fat intake.
And, it should also be noted: they don’t slope upward, so, where in that does one imagine a testable hypothesis where increased saturated fat intake causes anything bad?
Then he had the idea of plotting on a logarithmic scale. The simplest way to explain that is that it scales to percentage change rather than absolute values. So, for example, in a graph with a vertical axis from 1-100, 10 would be halfway, since the change from 1-10 in terms of percentage is the same as the change in 10-100.
But he still wasn’t finished.
Taking the same DALYs statistics from the last graphs (but this time combining the figures for CHD, stroke and other CVD) I have plotted the countries in ascending order of the total DALYs per 100,000 head of population (left to right along the x-axis) against the macro-nutrient ratio (C:P:F) for those same countries (along the y-axis) and plotted the linear regression (or trend) lines for each macro-nutrient. The relationships are pretty clear in the bars themselves but the trend lines confirm it: The less (total) fat you eat the greater the DALYs score (line slopes down toward the higher scores at the right-hand end of the x-axis) but even more telling, the more total carbohydrate you consume the greater the DALYs score (the line slopes upward toward the right-hand end of the x-axis). Protein appears to be neutral, with a perfectly level trend line.
So, more saturated fat, better, more carbohydrate, worse, and protein is not associated.
Download the file here.
Now let’s take a look at something related: total cholesterol. It so happened that Ricardo Carvalho rang in around this same time, calling attention to some data crunching of his own.
As I’ve speculated before, a total cholesterol of 200-220 seems to be the sweet spot for total C. Ricardo is calling it 200-240 and that looks like a good call, to me. Bear in mind, this does not equal causality, only association. Any individual can die of anything with any cholesterol number, or live to 101.
What this does accomplish, however, is tantamount to utter falsification of the hypothesis that either "high" saturated fat intake or "high" total cholesterol is independently causal for heart disease, stroke, or indeed, death from any cause.
Terrific post, Richard. And big thanks to those who helped you, too.
Indeed thanks to the others. In fact, one of my larger problems these days is getting too much stuff from my readers and I have to be selective cause I can't blog everything.
Nice “problem” to have, though.
Wonder how these could find their way into mainstream Nutrition textbooks? lol
A little history lesson in the philosophy of science via Karl Popper would serve as a nice introduction to that portion of the textbook. One word: falsify. One sentence: this introduction falsifies almost all the content in nutrition books you've previously read.
Like Mark Sisson said, “Forget everything you thought you knew about diet, exercise, and health! It's time to go back to the beginning …”
Now that's some serious number-crunching. I love it. At total cholesterol of 203, I've apparently hit the sweet spot — something I can't do with a golf club.
So
I guess when I look at the data in the first set of graphs I can't help but notice that there's definitely a trend as far as economic status of the country that goes along with the trend of increased heart disease. France, netherlands, etc on one end and on the other is bosnia, georgia etc. This in my mind is the greatest criticism of the analysis of the statistics.
Not to say I don't agree with the hypothesis I just am taking a critical eye when I look at the data.
Yes, however, all of the primitive people's studied over the last couple of hundred years were in excellent health (Weston Price, et al).
It's only when the primitive people's are introduced to modern foods rather than hunting, gathering, and or being pastoral themselves does health degrade. So, they go from sourcing high quality nutrition on their own to being too poor to afford modern high quality food, so they eat the flour, sugar, frankenoil based stuff cause they can afford it, and health deteriorates.
That is a good point. I have spilt the data up – since most of the ‘poorer’ countries are between the far left and just over the midway point on the x-axis and most of the ‘wealthier’ countries are from just over the midway point and on to the far right of the x-axis – and replotted the graphs and the trends for the DALY scores still slope downwards the more saturated fat is consumed for both sub-sets. However, I will concede the degree of the slope is not as steep for the wealthier countries.
“Which Cholesterol Level Is Related to the Lowest Mortality in Population with Low Mean Cholesterol Level: A 6.4-Year Follow-up Study of 482,472 Korean Men ” – http://aje.oxfordjournals.org/content/151/8/739.full.pdf
The graphs from Alexander Thorn are really interesting but then comes this: He “made some adjustments for the population size of the country.” It’s bad enough that he doesn’t say what these adjustments (more than one?) were; what’s even worse is that he has apparently multiplied the rate/person by the total number of persons. That makes no sense at all.
Seth, glad to see you here, fellow n=1 experimenter.
Let me dig this up, and I’ll email Alexander to make sure he knows of your complaint. He did send me a barrage of graphs in quick succession and did himself point out problems and/or inconsistencies.
I’ll follow up tomorrow.
Seth:
I emailed Alex and here was his reply:
“I see what you both mean. I was a little too ‘slap-happy’ with playing about with the figures at one point. To answer the question about multiplying cases per 100,000 by population size – that is not exactly what I did. DALYs are disability adjusted lost years per 100,000 of population due to the listed diseases rather than the absolute incidence of those diseases per 100,000. So I could foresee the counter-argument that small countries with small populations may skew the trends. So I found some census figures for the overall populations for each country and divided those by 100,000 and multiplied the lost years by the resulting figure to get an absolute amount of years lost for each country. Hope that clarifies what I did and why.”
Thanks, Richard. Let me repeat that this is really interesting data. Yeah, it’s a perfectly reasonable concern that the small countries are different from the large countries. Changing the data so that you plot “absolute amount of years lost for each country” doesn’t solve the problem, however, because that is a measure where a large country (with, let’s say, 10,000,000 people) will obviously have a larger number than a small country (with say 1,000,000 people). It makes the problem worse, not better.
Here’s a better way to look at these data: make a scatterplot with X = percent saturated fat and Y = disability adjusted lost years per 100,000 population. To see if large and small countries are different, divide the countries into two groups splitting at the median of their population sizes. Then plot large countries with one color of dot (say, red) and the small countries with a different color of dot (say, blue). That will show you if the large and small countries show the same trend. If you want to see even more clearly, fit separate lines to the two sets of points and plot those lines in different colors.
I have done this now – thanks for the suggestion – and sent the results to Richard. The trends are still evident – the more saturated fat in the diet the fewer years lost to disease.
Alex (& Seth):
Got the graphs and will probably do a new entry, but it may take a couple of days due to being particularly busy on a project (and cooking spaghetti sauce — to go on spaghetti squash, of course).
I’ll drop a not here when it’s up.
Seth:
Sorry it took so long, but the new graphs are now up.
https://freetheanimal.com/2009/11/drilling-down-saturated-fat-epidemiology.html
Thanks for letting me know.
I am surprised at the protein in the diet.
http://www.impactaging.com/papers/v1/n10/full/100098.html
Dietary restriction (DR) without malnutrition is widely regarded to be a universal mechanism for prolonging lifespan. It is generally believed that the benefits of DR arise from eating fewer calories (termed caloric restriction, CR). Here we argue that, rather than calories, the key determinant of the relationship between diet and longevity is the balance of protein to non-protein energy ingested. This ratio affects not only lifespan, but also total energy intake, metabolism, immunity and the likelihood of developing obesity and associated metabolic disorders. Among various possible mechanisms linking macronutrient balance to lifespan, the nexus between the TOR and AMPK signaling pathways is emerging as a central coordinator.
I would think if you took protein as a percentage of calories you might get another answer. Can you seperate animal from vegetable like grain protein? Lower protein and high fat might be the better option. Eric