Wednesday, March 26, 2014

Corrections to the New Review Paper on Dietary Fat and Cardiovascular Risk

The meta-analysis by Chowdhury et al. raised quite a furor from certain segments of researchers and the popular media.  I find this reaction interesting.  I usually write about obesity, which is a topic of great interest to people, but my post about the review paper received more than twice my usual traffic.  People whose findings or opinions are questioned by the paper are aggressively denouncing it in the media, even calling for retraction (1).  This resembles what happens every time a high-profile review paper is published that doesn't support the conventional stance on fatty acids and health (e.g., Siri-Tarino et al. [2], which despite much gnashing of teeth is still standing*).  I'm not sure why this issue in particular arouses such excitement, but I find it amusing and disturbing at the same time.  This kind of reaction would be totally out of place in most other fields of science, where aggressive public media outbursts by researchers are usually frowned upon.

As it turns out, the critics have a point this time.  Significant errors were uncovered in the original version of the meta-analysis, which have been corrected in the current version (3).  These include the following two errors, one of which alters the conclusion somewhat:
  • The outcome of one observational study on omega-3 fatty acids was reported as slightly negative, when it was actually strongly positive.  This changes the conclusion of the meta-analysis, making it somewhat more favorable to omega-3 consumption for cardiovascular protection.
  • The authors left out two studies on omega-6 fatty acids.  These didn't change the overall conclusions on omega-6.

Overall, the errors had relatively little impact on the analysis, and the only meaningful effect of correcting them was a more favorable result for omega-3 fatty acids.  Some have suggested that the presence of errors in the analysis means that the whole thing is unreliable-- who knows what other errors may be present, and the paper should be retracted with much fanfare (4).  I agree that the errors suggest a less-than-rigorous approach by the authors.  However, the authors corrected the mistakes that were pointed out, and they had a relatively small impact on the overall findings.  Unless more substantial errors are uncovered, I don't see the point of complaining about the possibility that additional errors may exist.

Another limitation of the paper was brought up in the comments on my last post about it.  The paper only considered randomized trials in which people supplemented with omega-3 or omega-6 fatty acids, not those in which one type of fat is replaced by another, which are called "dietary fat modification" trials.  This is a significant limitation and I'm not sure why they decided to omit those studies.

Dietary fat modification usually means reducing the intake of saturated and trans fats, while increasing the intake of polyunsaturated omega-6 and/or omega-3 fats**.  One of the most comprehensive meta-analyses on dietary fat modification was conducted by the Cochrane Collaboration (5).  Cochrane has a well-earned reputation for rigorous meta-analyses.  Its latest meta-analysis concluded the following:
  • Dietary fat modification significantly reduced the risk of cardiovascular events by 14 percent.
  • Dietary fat modification had no significant impact on cardiovascular mortality or total mortality.
  • Low-fat diets had no impact on any cardiovascular outcome.
According to this meta-analysis, the dietary fat modification trials have identified a small but statistically significant effect on cardiovascular risk.  It's important to note the limitations of the Cochrane analysis:
  • It included some trials, such as Oslo Diet-Heart, that modified a number of dietary variables besides fat.  This means that the results actually pertain to the modification of dietary fat and other foods.
  • In many of these trials, participants decreased intake of saturated and trans fats simultaneously, and increased intake of polyunsaturated omega-6 and/or omega-3 fats (6).  Because there were so many variables changing, it's impossible to attribute effects to any individual type of fat.  
  • It did not distinguish between interventions that replaced saturated/trans fats with only omega-6 fats, and those that increased omega-3 in addition.  Two interesting meta-analyses by Dr. Chris Ramsden and colleagues have suggested that this is an important distinction: trials that replaced saturated/trans fat with omega-6 alone didn't reduce risk (and may even have increased risk), while those that replaced it with oils containing omega-3 did reduce risk (6, 7).
  • The trials this meta-analysis is based on lasted years, but they don't necessarily represent what would happen if you followed the same dietary pattern for a lifetime, beginning with people who are young and healthy at baseline.  It's likely that any effects would be exaggerated over a longer period of time.
In the end, I think all of these meta-analyses are problematic.  Meta-analyses are supposed to be the highest form of evidence; the final word.  But it turns out that the devil's in the details.  Depending on what you include or exclude, how you group the studies, and how you interpret the results, they can lead to different conclusions.  Furthermore, since they're narrowly focused on specific outcomes, they don't necessarily tell us what we want to know about overall health and well-being.  For example, a food that causes malabsorption might lead to a positive outcome in a weight loss trial, but that doesn't make it healthy in the broader sense.

In my view, the most important question is "what foods should I eat"?  The answer depends on the individual, but here are my general opinions.  Eating synthetic trans fat is probably a bad idea, and it's generally found in low-quality processed food anyway.  Refined seed oils like corn and soybean oil are the nutritional equivalent of white sugar, oxidize during high-heat cooking, and also tend to be found in low-quality processed food, making them a poor choice.  I haven't seen much evidence that makes me concerned about eating the fat naturally contained in meat and dairy, so I view those as acceptable, in moderation.  Nuts and wild seafoods seem to be healthy, whether or not that relates to their fatty acid profile, so it makes sense to eat them regularly.  Extra-virgin olive oil appears to be one of the healthiest fats, so it's logical to gravitate toward it as the default added fat.

I question how much these analyses and re-analyses of dietary fat modification trials have contributed to our understanding of what we should eat.  The bulk of the evidence, including from diet modification trials, suggests that we get the best results by focusing on overall diet quality rather than specific fatty acids.

* The primary critiques were 1) that the data were over-adjusted because some of the data had been pre-adjusted for circulating cholesterol levels, and 2) Dr. Ronald Krauss has received funding from the dairy industry.  Further analysis by Dr. Krauss's team confirmed that the result was the same whether one considered studies that had been adjusted for cholesterol, or those that had not.  Dr. Krauss was actually not receiving funding from the dairy industry at the time he wrote and published the paper, but in any case this argument is weak unless specific instances of bias can be identified in the paper.

** The role of trans fats tends to be overlooked in these meta-analyses.  Typically, results are attributed to replacing saturated fat with polyunsaturated fat, when in fact trans fat was also replaced.  If we believe trans fats are harmful to cardiovascular health, then this could plausibly account for the observed effects.  Dr. Ramsden has a great discussion of this issue in one of his papers (6).


  1. "I'm not sure why this issue in particular arouses such excitement, but I find it amusing and disturbing at the same time."

    Via Dr. Noakes:

    "At any given moment there is an orthodoxy, a body of ideas which it is assumed that all right-thinking people will accept without question. It is not exactly forbidden to say this, that or the other, but it is 'not done' to say it, just as in mid-Victorian times it was 'not done' to mention trousers in the presence of a lady. Anyone who challenges the prevailing orthodoxy finds himself silenced with surprising effectiveness. A genuinely unfashionable opinion is almost never given a fair hearing, either in the popular press or in the highbrow periodicals." -- George Orwell

  2. It seems to me indefensible to recommend fat ratios based on their effects on serum cholesterol, without taking into account the effects they are having in tissues, i.e. inside cells. Hepatocytes, macrophages, and endothelial cells have LDL receptors. The latter 2 are implicated in CHD.
    Linoleic acid is required for cardiolipin in mitochondria, but linileic acid at intakes that alter LDL will increase hepatic cholesterol synthesis and LDL receptor uptake of cholesterol into tissues.
    Increasing both the tissue cholesterol pool, and whole-body cholesterol (despite the decrease in LDL).
    Excess tissue cholesterol, if unesterified, will oxidise cardiolipin. This is a determinant of mitochondrial health.
    Palmitic acid does not increase cholesterol synthesis or tissue cholesterol (despite elevating or maintaining LDL).
    Trans fats increase both cholesterol synthesis and LDL cholesterol.
    There is an optimal intake of LA that is probably not associated with significant reductions in LDL.

  3. I liked the initial report. Imperfect, yes, and perhaps incomplete, but the point was that the demonization of "saturated" was unwarranted, and wrong. At the same time, the study reinforced the idea that "trans-fat," an "unnatural" man-made fat, was just plain bad.

    So what's left here: fat from meat and dairy is ok, as is olive oil, a remnant from the old list, and a "natural" fat. Also apparent is the potential of "grading" oils based on how they are made. No one "makes" the oils in meat or dairy.

    Now, if you want to go further, then let's have a fight over which type of fat you should eat. That's not how I understood the point of the study, which seemed to me to say two things: it said that saturated fat was not bad, and that fat of any sort was preferable in terms of calories/nutrition to carbs/sugars. It seems implicit that if those saturated fat calories are not harmful, then those calories will displace some carb/sugar calories.

    The idea that any mistake in the study (or dispute regarding the data) renders the entire conclusion void and meaningless is amusing. If consistently applied that should render useless about 90% (or more) of human activities here on Planet Earth.

  4. It doesn't bother me at all when errors are found provided the paper is edited to account for the new information. This is what peer review is all about.

  5. Yes, it's both amusing and disturbing to read expert comments in media. For example, tone of professor Mozaffarian has changed since early last week. Last week his opinions were in sharp contrast to what Willet has said this week. Kris-Etherton pointed out at Foodnavigator's site " ... lack of adverse consequences of trans fatty acids in the Annals paper is very surprising". What? Half of the media coverage still seem to contain those errors present in the first release of the paper. What a mess.

    The fat replacement trials from 1960s were done with extremely (unrealistically) high intakes of omega-6 fatty acids (typically 10-16 E%). Currently intake of omega-6 FAs globally is around 6-7 E%. We know from other fields of nutrition, that more is not always better when pursuing optimal health. Perhaps the optimal intake of omega-6 FAs is around 5-8% of energy? It's hard to get there if you do not use seed oils, margarine or nuts.

    As a side note, corn and soybean oils are not nutritionally equivalent to sugar . Both contain vitamin E and essential fatty acid, linoleic acid. Soy oil is also a source of ALA, another (semi)essential fatty acid. Sugar does not contain any essential micronutrients in relevant amounts. So at least technically speaking there is difference between sugar and seed oils.

  6. Hi Stephan, don't you think we should acknowledge that Patty Siri-Tarino gets her research funding from the National Dairy Council? Her work should therefore be treated with a fair amount of skepticism.

    Here are two videos that helped me track down some of the research papers and check them for myself, instead of relying on what the Dairy Council thinks is OK for me to eat:

  7. Hi Reijo,

    I don't really know what the optimal n-6 intake is, but I wouldn't feel comfortable using highly refined foods to try to reach a value that may or may not be optimal. I take your point that refined seed oils do have some nutritional value, but it's almost impossible to become essential fatty acid deficient and there are better ways to get vitamin E. If you eat unrefined food instead of refined seed oils, you'll end up with plenty of EFAs and vitamin E.

    Refined seed oils are much less nutritious than any unrefined food, and less nutritious even than partially refined foods like extra-virgin olive oil. The non-essential, non-triglyceride components of olive oil (e.g., polyphenols) are thought to be important for its health benefits, and these are mostly removed by refining.

    Hi Gunther,

    Funding source alone is not a rational basis for discarding scientific data.

    Many of the primary proponents of the diet-heart hypothesis are directly or indirectly funded by industry, including seed oil and statin manufacturers. I don't ignore their work on that basis. Do you?

  8. "Funding source alone is not a rational basis for discarding scientific data."

    In fact Chowdhury was funded, in part, by the Heart Foundation UK, who one might have thought would have preferred a different outcome.
    Often, only the people who make a product are interested in testing it. If you ignore that data, you're not left with much.

  9. hi stephen,

    I see a link to an article you wrote abot vitmin k a lot, but the post is no logner available...

    Is is possible you re post your thoughts on teh matter or, of they have changed, post a new article?

    Kind regards,


  10. Hi Stephan, I appreciate your reply.

    I never said to ignore the work based on the funding, but to regard it with skepticism and a much higher level of scrutiny than independent work.

    But since you mentioned diet-heart, I do think a much larger body of independent research exists proving diet-heart than there exists independent research proving otherwise. Doesn't it seem weird to you that almost ALL the research "disproving" diet-heart are Big Dairy, Cattle or Atkins funded? When you dig deeper, you find Lustig is Atkins-funded, Ronald Krauss receives his funding from the National Dairy Council, the National Cattlemen's Beef Association and the Atkins Foundation.

    What do you think of tobacco research? Do you use the same rule there? Any doctor in the 80s who read industry-funded research without skepticism and hence didn't advise their patients against smoking probably feels like an idiot today, if not an accomplice to murder.

    BTW there is also plenty of archeological and antropological evidence proving that high-saturated-fat eating Inuit and nomadic Russian and Central Asian peoples were not healthy and indeed suffered from metabolic diseases, kidney stones and CVD at early ages.

  11. Hi Michael,

    I took certain articles down that I feel are below my current quality standards. I have no immediate plans to re-write that one.

    Hi Gunther,

    Wasn't industry research on tobacco conducted in large part by the tobacco companies themselves, rather than by independent researchers with tobacco company funding? Things have changed a lot in the research world since then anyway, due to concerns about funding source bias. Generally, for a researcher at a respected academic institution to accept such funding, the funding organization cannot be involved in the conduct or analysis of the experiments.

    It's not easy to get grant funding in science, and if your findings undermine the prevailing hypotheses of the field, it's even harder. You may not recall this but Krauss was an esteemed lipid researcher for a long time prior to his foray into the CVD effects of saturated fat. He sought industry funding after his findings began to challenge diet-heart, likely because his colleagues (who determine NIH funding) wouldn't fund the work. This situation is common.

    It's the same as the pro-diet-heart researchers who are funded by Unilever or Pfizer. Industry funds people who are doing work that's relevant to their interests. It doesn't mean the researchers are doing the bidding of industry, just that their interests are aligned.

    You said that "almost ALL the research "disproving" diet-heart are Big Dairy, Cattle or Atkins funded". I'm not sure where you got that idea but I'm pretty confident it's not correct. As George mentioned above, Chowdhury was partially funded by the Heart Foundation UK. And the term "disprove" is a strong one-- I don't think any single study has claimed to disprove the diet-heart hypothesis. Specific studies do strengthen or weaken it.

  12. Hi Stephan, I just looked over full text of paper. I think you've done a good job summarizing and discussing. One thing that hasn't gotten much attention is that one of the most robust results was an inverse association between circulating margaric acid (some kind of dairy fat marker) and CVD. Circulating long-chain omega-3s and 6's also appear protective, although apparently adding them specifically as supplements isn't. I would summarize by saying that the observational evidence suggests that fish and dairy consumption are associated with decreased cardiovascular disease.

    Overall, the effect sizes are not all that impressive, with a couple of exceptions the confidence intervals overlap zero. Given this, it would be nice to know the details of their "random effects" estimate to understand the extent of pooling in their summary relative risk estimates. A hierarchical Bayes' approach would almost certainly do a better job accounting for uncertainty than their vaguely specified "random effects" model...

  13. My biggest issue was the wording of the conclusion: "Current evidence does not clearly support cardiovascular guidelines that encourage high consumption of polyunsaturated fatty acids and low consumption of total saturated fats"

    What does high consumption of PUFAs mean, and who said that? The IoM set the PUFAs DRI at 5-10% of calories, and listed potential damage from lipid peroxidation as a reason for that upper limit.

    I agree that how heated this debate gets is ridiculous. It seems to be an issue of 'can i eat my bacon', which has always been answered with 'yes', in moderation and while consuming a nutritionally adequate diet outside of that.


Comment moderation is in effect. The core criteria for publication are that a comment must add value, and must be respectful. For more information, see this post.