3

votes

Animal dietary fat study

Answered on August 19, 2014
Created January 17, 2011 at 5:10 AM

I don't have access to journal articles at the moment. Does anyone else have access to this who can give an opinion?

http://www.ncbi.nlm.nih.gov/pubmed/19561318

Medium avatar

(10611)

on June 29, 2013
at 04:15 PM

Add vegan to the scrap pile and I'm with you 100%. Our ancestors didn't live on fad diets.

Medium avatar

(10611)

on June 29, 2013
at 04:12 PM

I'm tired of seeing that old saw about correlation a d causation. The people that present it use it as a rhetorical device. Correlation is an empirical tool used to optimize a response. Harris Benedict's BMR equation is a good example using population: as you age and lose weight, your metabolism goes down. Linear correlation. Are there special snowflakes? Yes, but the goodness of fit is never perfect. The question of correlation vs causation can be tested very easily by losing weight. Or by getting older...

4e40d2b9e1a762949a25b958762aa10d

(762)

on January 17, 2011
at 03:30 PM

The manuscript is available for free download in PubMed Central. Manuscript format differs from the final in that the figures are at the end rather than embedded throughout the text. But, all the content is still there.

  • 034c678bff434ab3781e3f1771018af9

    asked by

    (279)
  • Views
    1.3K
  • Last Activity
    1427D AGO
Frontpage book

Get FREE instant access to our Paleo For Beginners Guide & 15 FREE Recipes!

3 Answers

best answer

5
47a42b6be94caf700fce9509e38bb6a4

(9647)

on January 17, 2011
at 08:02 AM

Hi, I just had a look at the article (but couldn't download it for some reason), and it's pretty much what you would expect it to be, and pretty much along the lines of what Drew suggested: just an epidemiological study, and subject to all the weaknesses of such studies.

But first of all, it seems like the surveys were done as responsibly as such things can be done: they had an enormous sample (500,000+) and threw away a fair number of the surveys for good reasons as well. But the problem of misreporting obviously remains. One amusing bit is that they threw out some of the extremes:

We further excluded 9778 subjects who had reported extreme values (ie, more than two interquartile ranges above the 75th percentile or below the 25th percentile on the logarithmic scale) for total energy intake (n = 4205), total fat intake (n = 741), or percent energy from total fat (n = 4832).

I'm guessing those of us paleos who like to get around 70% of our calories from fat wouldn't have made the cut.

Anyhow, back to the point. The surveys were "prospective" in that they were performed before any incidence of pancreatic cancer rather than after (and apparently there was a dearth of information on pancreatic cancer, particularly because it is hard to detect early and has a very high mortality rate). So they looked at a whole bunch of data that had been gathered for other purposes (the "NIH-AARP Diet and Health Study"), and then cross-checked all the participants with other databases that have information on cancer. Better than what had come before, but still just an observational thing.

So they found positive associations between dietary fat and pancreatic cancer: higher rates for higher-fat eaters somewhere on the order of 1.5 to 1 from the highest quintile of fat consumption to the lowest quintile of fat consumption, and they found similar results for saturated fat consumption and animal fat consumption (as opposed to vegetable). This of course doesn't tell us anything. At this point we could say that this is just the "healthy people effect." The people who listen to public health advice (eat less fat) are also the people who are healthy: they eat better otherwise, they exercise more, they eat more fruits and veggies (I hear that this is something non-paleos do to make up for lack of meat consumption, ha ha), and they might be happier and healthier in any number of physical and psychological ways.

But the experimenters controlled for a few other things and still found the positive associations. As it says in the abstract, they controlled for total energy intake, smoking history, body mass index, and diabetes. So if your increased fat intake meant that you were eating more calories then they compared people of equicaloric diets and re-examined the data. (I know it's much more complicated than that, but I'm just learning about the statistics stuff.) They did the same with BMI, diabetes, and smoking. Here's the jargon:

We did not find evidence of effect modification by BMI, self-reported diabetes, or smoking history ( P interaction > .5 in most instances; data not shown).

Data not shown! Anyhow, these first two might spell a little bit of trouble. Adjusting for BMI would seem to eliminate the exercise effect. And adjustment for diabetes might seem to eliminate the sugar/flour effect. That is, it might be difficult to be able to explain increased pancreatic cancer among those with higher fat intake by pointing out that the people with higher fat intake are the same people who exercise less. OK, well, exercise schmexercise. What about diabetes? Since there's such a strong link between sugar/flour and diabetes, this might make it difficult to explain increased pancreatic cancer among those with higher fat intake by suggesting that people with higher fat intake are the same people who eat more sugar and other refined carbohydrates. Trouble?

I would guess it's not trouble enough. I'd bet that if you went back through the data and looked for associations with sugar and refined carbohydrate that you'd find something. (Sounds like a job for Denise Minger! This statistics jargon is impenetrable to me.) There are also so many other health factors, like the aforementioned fruit and veggies. They didn't control for that, and getting all your vitamins and minerals has to be pretty big.

Anyhow, the editorial in the same issue of the journal and even the press release caution against misinterpretation of the results, echoing what the authors themselves say (so I guess we can put even more blame on the press, big surprise):

[The authors] do note, however, that there is insufficient epidemiological and laboratory evidence to confirm the importance of animal fats or even that meat is the important factor, as opposed to other dietary or lifestyle preferences associated with meat consumption.

Well, there we go.

4
Af842c68e3d07fa0e35b4274f3acaeec

on January 17, 2011
at 05:54 AM

I realize this doesn't answer the question, but I would just like to note a few things about this study that leads me to put very little confidence in studies like this.

  1. Food questionaires are notoriously unreliable and ambiguous. The questions they ask are very inaccurate predictors of people's actual intakes. This is especially true when people fill out these surveys after the fact, rather than keeping a dietary journal, which have themselves been shown to be 30% inaccurate at times. Their food categories are often misleading as well, wrapping in things like pizza as red meat.

  2. They show CORRELLATION, which must not be confused with CAUSATION. The rate of ice cream sales and crimes in an area are moderately correlated, does this mean that one causes the other? Absolutely not. A third variable, temperature is likely the cause in this. Correlations, unless they show results like lung cancer and smokers (smokers are 20X more likely to get lung cancer) should only be used to discover trends that can be tested in experimental trials, the kind that one may draw causal relationships from, sometimes.

So, to me this article is at best something to study further, and at worst, misinformation that will be used by the media to scare people about fat.

Medium avatar

(10611)

on June 29, 2013
at 04:12 PM

I'm tired of seeing that old saw about correlation a d causation. The people that present it use it as a rhetorical device. Correlation is an empirical tool used to optimize a response. Harris Benedict's BMR equation is a good example using population: as you age and lose weight, your metabolism goes down. Linear correlation. Are there special snowflakes? Yes, but the goodness of fit is never perfect. The question of correlation vs causation can be tested very easily by losing weight. Or by getting older...

0
0afe351ed011e1f914c9ad32e0a150c1

on June 29, 2013
at 03:31 PM

Statistically, you always throw out the extremes. All studies are "observational", not everything applies to everyone! There are people who are "immune" to arsenic poisoning, but very very few. If you believe that smoking causes cancer (observed at a rate of about 46%) then the proof here is positive, excessive animal fat consumption does indeed cause cancer. Want another study of a people who lived a pure palio life? Check the life expectancy of the Aleuts. Even if you toss out the infant mortality (which palio people love to do) your chances of reaching age 50 are less than 50%. It's not conventional wisdom, it is FACT. And just who is it that thinks palio man was slim and trim? or that the females didn't have more fat than the males? In the end a BALANCED diet works, not some rehashed bad diet (palio, Atkins, no-carb etc.) all have failed miserably!

Medium avatar

(10611)

on June 29, 2013
at 04:15 PM

Add vegan to the scrap pile and I'm with you 100%. Our ancestors didn't live on fad diets.

Answer Question


Get FREE instant access to our
Paleo For Beginners Guide & 15 FREE Recipes!