Response to Denise Minger,
Part 3: Cherry Picking
Denise Minger, in her professed effort to tell the truth about Ancel Keys and lay out the facts as objectively as possible, repeatedly accused him of cherry-picking.
Therefore she is basically accusing him of dishonesty. He selected examples that fit his hypothesis and rejected ones that did not. His ethics as a researcher took a back seat to his agenda.
In Minger’s comments addressing my videos about him, she says I tried to selectively drop countries from the Yerushalmy and Hilleboe graph to make the diet-heart disease association look stronger. Now she is accusing me of cherry-picking, too.
Here's another problem with using that data for 22 countries to argue against the Seven Countries Study, if that’s really what the confusionists insist on doing. As I said, that data came from a statistical compilation by the FAO. That was not data used in the Seven Countries Study. The Seven Countries Study was a prospective cohort study. Researchers were dispatched within the seven countries to collect their own data using uniform standards. Individuals were studied prospectively, or over time, so they were considered cohorts. This study was not created by merely crunching someone else’s data, which is what Yerushalmy and Hilleboe did. Keys was working with far better and more useful data than those two. I'll give you an example of a problem with their data.
- Plant Positive
If you listen to and try to understand what I actually said in the video, my point was that the data from all the different countries were not necessarily of equal value, and they were not as good as the data Keys used in his Seven Countries Study. Minger missed my point, apparently. Moreover, I didn’t drop these countries, Ancel Keys did. I simply tried to show why an intelligent person like Keys might have chosen not to use some of these data. Minger’s nitpicking actually isn’t helpful to her cause. Let me show you why.
Here on the left is the graph representing the countries Keys wanted to talk about, and on the right is the Yerushalmy and Hilleboe graph of the 22 countries.
In red, I’ve circled the countries on the 22-country graph that are represented by the Keys graph on the left. These are the supposedly cherry-picked countries, the ones he used to fabricate a pattern. Just for a moment, let’s pretend Minger and the other Keys-bashers are right that these data are not representative of reality somehow. Let’s pretend they aren’t there for a moment.
Here I’ve circled in blue the countries which I addressed in my video. Mexico’s mortality data were as poor as it gets and the others were subjected to the privations of World War II. Those countries’ histories actually argue in favor of the connection between animal foods and heart disease, as I’ve said, but let’s pretend they aren’t here, either. We’re agreeing now that all these countries are unrepresentative of the diet-heart relationship, so if we ignore all the circled countries we should have a better graph.
Then Minger says I didn’t explain why Finland and Austria should stay on the graph, which is basically putting words in my mouth. Again, my purpose was not to try to make much out of all this bad data. She is pretending I set out to comment on every country on the graph for some reason. Watch my video and it should be obvious that I didn’t even try to do that. She also seems to think Japan, France and Italy should be removed because of their poor record keeping. It’s not so smart for her to say this but let’s just exclude them anyway.
All the countries she just named are in yellow. It looks like by the time you remove the countries I said were problematic, along with the ones Minger says she wants removed, to go with the original six that were supposedly cherry-picked, we’ve removed a lot of countries, haven’t we? It seems we are coming back around to my original point that this was not a very good collection of data, so it gets harder to see why Minger thinks this is such awesome material to blog about. It seems to me that this exercise creates a problem for her, not me. But the poor quality of the data here isn’t her biggest problem.
Now I’ve added green circles to the countries Keys himself left out that showed strong associations between fat and heart disease, including Finland, with its much-studied remarkably high heart disease burden during that time. Ms Minger, here’s an important question. If Keys were biased, why did he leave out countries that would have made his case stronger? Why didn’t he use Finland? Wouldn’t that have helped him a lot if he had an agenda? Isn’t this proof that your allegations of cherry-picking are entirely without foundation? Moreover, look at the countries not mentioned yet by any of us, which remain without colored circles. Even they show an increase in heart disease with increased fat!
Here they are on an uncluttered graph plus Finland, which Keys left out despite how much it would have helped his case. Anyone looking at this can see that there is an apparent relationship between fat consumption and heart disease in this data. Even if we agree that Keys cherry picked, it still looks like his point was correct. So I ask, how does this discussion help the cholesterol deniers?
It seems any way you look at it, this is a reasonably strong association, just as Yerushalmy and Hilleboe said, although it wasn’t as strong as the association for animal protein. Their observation, which Ms Minger thinks is so revelatory, is that increased animal food consumption is reflective of national wealth. Since the countries eating more animal foods would have had better health care by this logic, it would seem their higher rates of heart disease would be doubly damning for animal foods
Jeremiah Stamler was making this very point about wealth and animal food consumption back in 1958. The only counterargument Minger offers to this problem for animal food proponents like herself is that these countries would have had more accurate data on mortality, making comparisons between them and poorer countries unfair to the wealthier countries. Wealthy countries’ mortality statistics would have been better, so they would seem to have a bigger problem with heart disease on paper.
So let’s look at mortality data to include all the other likely classifications that might have absorbed mischaracterized heart disease deaths. This graph eliminates the issue of misclassification. Look closely. This also shows us more countries Keys left out that could have helped his argument, like Guatemala. There you see Guatemala all the way on the bottom, with the least problem with cardiovascular disease.
The ordering is not all that different, is it? Notice Finland, the heart disease leader, is still up on top with other high saturated fat-consuming nations of the day like Australia, the US, and Canada.
In this comparison of Americans and Guatemalans, the diet-heart idea is very strongly reinforced. Rural Guatemalans ate much less fat, had much lower blood cholesterol, and experienced vastly better heart health than typical Americans of the time. Yet Keys did not include them in his paper. Once again, her cherry-picking claim crumbles and lipid-hypothesis is reaffirmed.
Here’s something else that’s being missed by those who whine about cherry picking. I won’t linger on this, but the point of a cross-cultural comparison is that you are using contrasting cultures to reveal real differences. Many of the countries Keys left out were not markedly different in their dietary practices from one another. I don’t think he should be faulted for selecting appropriately contrasting cultures.
Let’s go back to Minger’s point that France, Italy, and Japan should be ignored because of their record keeping. I am more than happy to eliminate France since that country is the famous exception to the diet-heart pattern. Has Minger not heard of the French Paradox? I’m glad she is not compelling me to make this detour now. But what about Italy and Japan?
I doubt Italy could be the subject of any argument about food waste, as they were the original source for the idea of the Mediterranean diet.
A more inclusive accounting of their mortality would have put them somewhere between the US and Japan, so that would leave the diet-heart idea intact. What about Japan? Could they have actually been suffering far more heart disease than the data suggested back then?
Well, as I pointed out in The Primitive Nutrition Series, the diet of the Japanese and its relationship to their incidence of heart disease was very well-studied. Ms Minger, you don’t want to go there. There is a lot of material like this to be found. Yes, the Japanese really had much lower rates of heart disease and ate much less saturated fat.
Others beside Ancel Keys studied this. One of the difficulties I have with my chosen style of presentation is that I present a lot of information fast, whereas the confusionists like Minger will create a long blog post based on only one unimpressive paper. Why don’t they devote as much space to comparisons of genetically identical populations living in different environments that were carefully studied like the Japanese, rather than clinging to exceptional and isolated cases like the Masai?
Ancel Keys did not trick the world into believing the Japanese had lower rates of heart disease while eating their traditional diets. Even this doctor, who maintained a very cautious attitude toward epidemiology, did not doubt for a moment the accuracy of the low reported rates of heart disease in Japan.
Here, the Japanese were shown to have around one seventh the rate of fatal heart attacks as Americans. It would be difficult for a confusionist like Minger to dispute this. It’s too big a difference.
This argument she has chosen to make about cherry picking can be seen as weak without any further investigation. Here’s Minger’s logic in action. In one sentence, she says Keys never explained why he picked the countries he did. In the next sentence, she says this alleged cherry picking was shameful, terrible, and unscientific. Ms Minger, how do you know they were cherry-picked if you don’t know why he picked them? How can you say he was being unscientific even as you are speculating about his motives? There are excellent reasons to believe he did not cherry pick them. First, as I said, he chose appropriately contrasting countries. Next, he chose the countries that were likely to have the most accurate data and left out all the rest. How can anyone say he cherry picked given these facts alone?
Then, if you look at his writings, it is clear he was aware of the problem of the reliability of data and uniformity of standards in cross-cultural comparisons. Ms Minger, he said in 1953 that, “Broadly speaking, death rates ascribed to specific causes are not very reliable under the best of circumstances.” He was completely aware of these issues and factored them into his observations, noting that differences in heart disease were so great between some countries that any inconsistencies in their methods of death certification would not have altered their relationships between each other. He said American men died of heart disease at around 10 times the rate of Japanese men at the time.
He also compared the frequency of the recording of other causes of death to identify countries with similar practices to the United States. For example, Italy’s death rates were similar for other diseases like cancer and cirrhosis of the liver. Only heart disease killed at dramatically different rates in the two countries. Moreover, he was clearly aware of the effects of food supply disruptions during the war in some countries, as you can see to the right, so when Ms Minger says I tried to drop off some countries to make the associations look stronger, I was actually just looking at this data in the same rational way that Keys did. For contrast, you might ask what an irrational analysis of this data might look like.
Donald Miller gives us various examples of real cherry picking. Here he thinks he is somehow making a point by choosing different countries to show different relationships between fat and heart disease, but actually he is showing us that he has done zero research into the quality of the data, has no historical frame of reference, and has not tested his own beliefs for weaknesses. He just sees a bunch dots he can arrange on a graph. A child could do this, too, but this guy is an MD. This is the sort of intellectual firepower brought by the confusionists against Ancel Keys. It’s pretty embarrassing.
Remember, Donald Miller is using the same data that Yerushalmy and Hilleboe did, and they, the same guys Minger thinks were so awesome, showed that while there was a correlation with dietary fat, animal protein correlated even more strongly. So not even these heroes to low carbers would lend the least validity to Miller’s ridiculous graphs. This demonstrates a truly amazing fact.
Back in those days Ancel Keys did not understand the unique problems with saturated fats, but when we look at his work, we can see he mostly got the relationships right.
Yet Donald Miller, fifty years later and with all the research that has happened since then, is getting it wrong. Decades later, with a huge body of scientific research just sitting there to educate him if he would only look at it, Miller still has not advanced beyond a simple game of connect the dots.
This same data Keys used was reexamined by Jolliffe and Archer in 1959, who found the most important factor in accounting for different rates of ischemic heart disease between countries was saturated fat. The authors of this paper, who were more restrictive in the data they used, also found a significant negative correlation for cereals with ischemic heart disease. Are you seeing that, Ms Minger?
Minger had a few other comments in this paragraph. She says my rationale for removing Denmark from the graph is bunk. She quotes a slide of mine as saying, “Denmark didn’t have any reduction in heart disease mortality during the war.” Minger is right about this – I am about to correct myself - but she didn’t help her case by bringing this up.
I had two slides on Denmark. Here is one. This makes no reference to death rates so this can’t be the one she is talking about. It seems to imply lower consumption of a number of animal products during the war there.
Here’s the other slide that I used. This must be the one to which she is referring. It says, “In Denmark, where the total fat consumption certainly declined but where the butter and egg consumption increased no fall was noticed in the death rate from arteriosclerosis during the war, but there was some increase during the two years immediately following the end of the war.” So there is an apparent disagreement between my slides regarding the consumption of animal foods during the war. This raises two points.
First, this illustrates the exceptional circumstances in Denmark, which may have explained why Keys left it out of his comparison. The other issue is that her correction of me here doesn’t help her case. She is asking us to look closely at a slide that said in Denmark, heart disease mortality remained the same during the war and increased afterwards, and it also says their consumption of butter and eggs increased. The author wrote this in 1950, before saturated fat’s dangers were understood, but he found reason to be suspicious of animal fats. Now we can read this and say, well of course heart disease mortality didn’t decline in Denmark. They just kept right on eating eggs and butter! He says at the bottom that in Denmark, the consumption of animal fats was risking public health.
After having examined this further, I see now this was indeed the case in Denmark. She has corrected me. I said their consumption of animal foods dropped during the war, but that was incorrect. I thank her for this correction because it further supports the lipid hypothesis. Denmark’s export market failed during the war, resulting in increased consumption of artery-damaging butter and eggs, so they kept right on dying from heart disease just like they had been before the war.
Here is what took place in Denmark during the war years. On the left you see deaths from atherosclerosis. On the right you see egg consumption. The trend lines track closely. The connection between fatty animal foods and heart disease is demonstrated yet again.
I misinterpreted this slide because I had not considered the fact that Denmark had been a major exporter of animal foods, so while their production may have declined, their consumption did not. I didn’t make this distinction at the time.
Here are some news accounts from the war describing their surpluses of eggs and meat. No wonder their heart disease deaths continued apace.
It’s true, then, Ms Minger. My point about Denmark was bunk. I was in error, and I appreciate this correction, and I will happily give you credit for it.
But what matters is not whether I or Colin Campbell or Ancel Keys or anyone else is right all the time. What matters is that people understand the real science and history of cholesterol, and that is what confusionists like you try every day to obscure.
In this vein, she also raises Finland and Austria. Once again, I didn’t say they belonged on the graph. Yerushalmy and Hilleboe did. Keys left them out. I don’t understand her point. Does she really want to talk about Finland?
As I said in my video number 38, Finland was experiencing extraordinary levels of heart disease, leading to a major public health campaign to improve their risk factors. Saturated fat consumption eventually went down there. Lives were saved. The campaign was successful.
What about Finland during the war? Here you see a line representing their atherosclerosis mortality rates, the second line from the bottom. Sure enough, deaths from heart disease fell during the war and increased afterward. Once again, her beliefs don’t seem to be supported by the facts.
And what about Austria? Well, Minger herself says tuberculosis infections there were the major factor causing a drop in deaths from heart disease. She is here referencing a doctor named Broda Barnes who studied autopsies in Graz, Austria during this time. Those who died from tuberculosis also had diseased arteries.
Barnes thought tuberculosis was causing hypothyroidism, resulting in heart disease.
Minger may not know that hypothyroidism causes an elevation of cholesterol, including LDL bad cholesterol. So again, I don’t see how she is arguing against the lipid hypothesis in any way here. Fine, chalk up the drop in recorded heart disease deaths to tuberculosis. But think about this for a moment.
I did not mention Austria at all, I didn’t think the 22-country graph was worth much to begin with, yet she says I didn’t explain why it should stay on the graph even though she has given reason herself why it wouldn’t belong there. Are you following this? Once again, Keys left it off and she seems to agree with his decision. Look at the top and you’ll see she also mentions Norway in this blog post. Is she implying that tuberculosis might have affected heart disease mortality in Norway, too? If that is what she wants to imply, she would be wrong.
Look at the red line here. That represents the drop in tuberculosis they experienced back then due to vaccinations. Therefore, there is nothing here in the data from Norway that a cholesterol confusionist like Minger can twist and misrepresent to obscure the diet-heart relationship.
Minger also expressed objections to my China Studies videos. I’ll get into that next.