Monday, February 20, 2012

The “pork paradox”? National pork consumption and obesity

In my previous post () I discussed some country data linking pork consumption and health, analyzed with WarpPLS (). One of the datasets used, the most complete, contained data from () for the following countries: Australia, Brazil, Canada, China, Denmark, France, Germany, Hong Kong, Hungary, Japan, Mexico, Poland, Russia, Singapore, Spain, Sweden, United Kingdom, and United States. That previous post also addressed a study by Bridges (), based on country-level data, suggesting that pork consumption may cause liver disease.

In this post we continue that analysis, but with a much more complex model containing the following country variables: wealth (PPP-adjusted GNP/person), pork consumption (lbs/person/year), alcohol consumption (liters/person/year), obesity (% of population), and life expectancy (years). The model and results, generated by WarpPLS, are shown on the figure below. (See notes at the end of this post.) These results are only for direct effects.

WarpPLS also calculates total effects, which are the effects of each variable on any other variable to which it is linked directly and/or indirectly. Two variables may be linked indirectly, through various paths, even if they are not linked directly (i.e., have an arrow directly connecting them). Another set of outputs generated by the software are effect sizes, which are calculated as Cohen’s f-squared coefficients. The figure below shows the total effects table. The values underlined in red are for total effects that are both statistically significant and also above the effect size threshold recommended by Cohen to be considered relevant (f-squared > 0.02).

As I predicted in my previous post, wealth is positively associated with pork consumption. So is alcohol consumption, and more strongly than wealth; which is consistent with a study by Jeanneret and colleagues showing a strong association between alcohol consumption and protein rich diets (). The inclusion of wealth in the model, compared with the model without wealth in the previous post, renders the direct and total effects of alcohol and pork consumption on life expectancy statistically indistinguishable from zero. (This often happens when a confounder is added to a model.)

Pork consumption is negatively associated with obesity, which is interesting. So is alcohol consumption, but much less strongly than pork consumption. This does not mean that if you eat 20 doughnuts every day, together with 1 lb of pork, you are not going to become obese. What this does suggest is that maybe countries where pork is consumed more heavily are somewhat more resistant to obesity. Here it should be noted that pork is very popular in Asian countries, which are becoming increasingly wealthy, but without the widespread obesity that we see in the USA.

But it is not the inclusion of Asian countries in the dataset that paints such a positive picture for pork consumption vis-à-vis obesity, and even weakens the association between wealth and obesity so much as to make it statistically non-significant. Denmark is a wealthy country that has very low levels of obesity. And it happens to have the highest level of pork consumption in the whole dataset: 142.6 lbs/person/year. So we are not talking about an “Asian paradox” here.

More like a “pork paradox”.

Finally, as far as life expectancy is concerned, the key factors seem to be wealth and obesity. Wealth has a major positive effect on life expectancy, while obesity has a much weaker negative effect. Well, access to sanitation, medical services, and other amenities of civilization, still trumps obesity in terms of prolonging life; however miserable life may turn out to be. The competing effects of these two variables (i.e., wealth and obesity) were taken into consideration, or controlled for, in the calculation of total effects and effect sizes.

The fact that pork consumption is negatively associated with obesity goes somewhat against the idea that pork is inherently unhealthy; even though pork certainly can cause disease if not properly prepared and/or cooked, which is true for many other plant and animal foods. The possible connection with liver problems, alluded to in the previous post, is particularly suspicious in light of these results. Liver diseases often impair that organ’s ability to make glycogen based on carbohydrates and protein; that is, liver diseases frequently lead to liver insulin resistance. And obesity frequently follows from liver insulin resistance.

Given that pork consumption appears to be negatively associated with obesity, it would be surprising if it was causing widespread liver disease, unless its relationship with liver disease was found to be nonlinear. (Alcohol consumption seems to be nonlinearly associated with liver disease.) Still, most studies that suggest the existence of a causal link between pork consumption and liver disease, like Bridges’s (), hint at a linear and dose-dependent relationship.


- Country-level data is inherently problematic, particularly when simple models are used (e.g., a model with only two variables). There are just too many possible confounders that may lead to the appearance of causal associations.

- More complex models ameliorate the above situation somewhat, but bump into another problem associated with country-level data – small sample sizes. We used data from 18 countries in this analysis, which is more than in the Bridges study. Still, the effective sample size here (N=18) is awfully small.

- There were some missing values in this dataset, which were handled by WarpPLS employing the most widely used approach in these cases – i.e., by replacing the missing values with the mean of each column. The percentages of missing values per variable (i.e., column) were: alcohol consumption: 27.78%; life expectancy: 5.56%; and obesity: 33.33%.


R. K. said...

Another oddity. FAOSTAT shows a lower per capita consumption of pigmeat in Denmark when I look at the 2007 data. (49.7 kg or 109.6 lbs). The NationMaster data appears to be older and the comments about 'pork consumption' indicate some skepticism about the data (click on 'Definition'.)

Ned Kock said...

The Bridges study, cited in the post and published in 2003, lists pork consumption in Denmark at around 65 kg/p/y. That is about 143 lbs/p/y.

In the Bridges study sample Denmark is not the biggest consumer of pork. That honor goes to Austria, which also has low obesity levels and high life expectancy.

Pål Jåbekk said...

Very interesting Ned. I would have liked to see wealth added to many other epidemiological nutrition data to see how it confounds. Also, I think the Danes are getting quite fat. Over 40% of males there are overweight.

js290 said...


As an aside, Asians suffer from poor glucose control and insulin resistance that may not necessarily manifest itself as obesity.

R. K. said...

I see my confusion. They used FAOSTAT data from 1996, whereas I was looking at 2007 data.

But then, that raises another question. Was pigmeat consumption consistent in Denmark over the years?

Looking at years 1961 to 2007, in kg, the mean was 50.5, the max was 70.6, and the min was 29.65.

Ned Kock said...

Hi Pal. The overweight percentages tend to be quite a lot higher than the obese percentages. I think that 40% is a good guess for the “adult” male population that is “overweight” in Denmark. The percentage of obese is quite a bit lower. This article has a graph that illustrates the differences:

john said...

This is cool to see in combination with Jaminet's posts. Someone referenced the Weston A Price article about pork preparation, which is interesting; Paul mentioned bacon's apparent protectiveness from his data; I linked in the comments a study about bacon protecting rats against colon carcinogenesis (and keeping them lean)--just some more information to consider. I'm sure things like cut and pig's diet matter a lot as well. "Pork" doesn't tell us too much.

I have heard stories of people experiencing inflammation reactions (swollen gums, joint pain) with pork, but it hasn't happened to me.

Ned Kock said...

Hi John. The WPF article you referred to may be this one:

I am wary of studies that show “negative results” soon after something is eaten. If you look at various health markers immediately after resistance exercise, you may be able to make a “convincing” case for it being unhealthy.

Ned Kock said...
This comment has been removed by the author.
Ned Kock said...

I often run into people who say that they’ll have a headache if they don’t eat anything for 4 hours or more. And guess what? They do have headaches, and feel physically ill, from fasting for 4 hours or more.

That happens until they realize that there is no reason for them to have the headaches.

They stop having them, even when they fast for 20 hours or more.

If people believe strongly that a food will cause them harm, frequently they will display physical manifestations of health problems after eating that food ...

john said...

Those are good points. You are implying that the found effect was influenced due to Weston A Price "believers"?

Though that was a post-meal observation, there does seem a be a positive effect of bacon, which makes it seem like the omega-6 content is not what causes problems, if there are problems at all.

Jaminet has posted his conclusion now, so there's something else to perhaps consider.

Ned Kock said...

Hi John. The psychological effect comment is unrelated to the WPF study. I’ll check Paul’s post. Thanks.

Ned Kock said...

My understanding of Paul’s pork trilogy is that the most likely culprits are pathogens that are inherent to pork, with pig organs and blood being particularly problematic in that respect.

Well, I am not convinced, and it is not only because of the population data analyses in my last two posts. We simply cannot ignore the traditional Okinawans, the longest-living traditional population in the world, who eat pigs from head to tail as a dietary staple:

As for the country level data analyzed in this and the previous post, it is quite possible that pork consumption, at the country level, is a marker for choline consumption from pig products. If that is true, then avoiding products containing pig organ meat would not be a very good idea.

Ned Kock said...

One more comment related to confounders for hepatitis E. Some studies, like the one below, may give the impression that pork consumption is a major cause of HE:

From the article linked above: “Antibodies against hepatitis E virus (anti-HEV) were found in 248 Swedish and Danish patients between 1993 and 2007 … Patients with hepatitis of unknown etiology should therefore be investigated for anti-HEV even if they have not been outside Europe, since infections acquired from pigs or other animals should be taken into consideration.”

Still, the main cause of HE is known to be poor sanitation. Moreover, the percentage of incidence in the population suggested by the article linked above would be inconsistent with claims based on studies like Bridges’s, cited in my two last posts.

Finally, when you read the article linked above more carefully, you see things like this: “Among patients with known country of infection, most were infected in Asia, mainly on the Indian subcontinent.”

Well, pork consumption in India is fairly low compared with Denmark …

Conway Calin said...

Nice Post,
kamagra online
kamagra oral jelly
buy kamagra jelly

online bookie said...

very interesting thanks for writing

Imran Hosen said...

Too Stressed about your FAT???
This is an Amazing Weight Loss & Health Tips E-Book
Your life can 'Recover'!! I have got the result from this E-Book. I reduced my fat from 220 pound to 180 pound within 3 weeks that is why I'm sharing with You.
Get this 100% free Weight Loss & Health Tips E-Book, Which will help you to drop the extra fat within 3 weeks. So Download the E-Book......Thanks
Weight Loss
Lose Weight Fast
How To Lose Weight Fast
Weight Loss Tips