Monday, March 17, 2014

New Review Paper on Dietary Fat and Heart Disease Risk

A new review paper on dietary fatty acids and heart disease risk was just published by Dr. Rajiv Chowdhury and colleagues in the Annals of Internal Medicine-- one of the top medical journals (1).  The goal of the paper is to comprehensively review the studies evaluating the effect of dietary fatty acids on heart (coronary) disease.  The review covers observational and intervention studies pertaining to saturated, monounsaturated, trans, omega-6 polyunsaturated, and omega-3 polyunsaturated fats.  The paper is notable for its comprehensiveness (inclusion criteria were very lax).

Here is a summary of the results:

  • In observational studies that measured diet, only trans fat was related to cardiovascular risk.  Saturated, monounsaturated, and polyunsaturated fats were unrelated to risk.
  • In observational studies that measured circulating concentrations of fatty acids, long-chain polyunsaturated fatty acids (DHA, DPA, EPA, AA) were associated with lower risk.  The dairy-fat-derived margaric acid (17:0) was also associated with lower risk.  No other fatty acids were related to risk, including trans fatty acids.
  • In controlled trials, supplementation with omega-3 or omega-6 fatty acids did not alter risk.
The authors conclude:
In conclusion, the pattern of findings from this analysis did not yield clearly supportive evidence for current cardiovascular guidelines that encourage high consumption of polyunsaturated fatty acids and low consumption of saturated fats.  Nutritional guidelines on fatty acids and cardiovascular guidelines may require reappraisal to reflect the current evidence.
My view
Key limitations of this meta-analysis are that 1) it lumped together a variety of studies of different design and quality, and 2) it can only be as good as the studies it's based on.  Meta-analyses are considered the highest level of evidence, but I remain somewhat skeptical of them for those reasons.

This paper contains something to upset just about everyone.  Whether you believe saturated, monounsaturated, omega-6, omega-3, or even trans fat influences cardiovascular disease risk, this paper didn't consistently support any of it, instead suggesting that perhaps dietary fat composition isn't where we should be focusing our attention.

Does it challenge my own position?  Perhaps a little bit, but not very much.  After considering new evidence and reviewing old evidence, I've gradually drifted away from the view that omega-6 polyunsaturated fat contributes to cardiovascular disease.  I still think it's probably a bad idea to eat a lot of refined seed oils-- the lipid equivalent of white sugar-- but I don't see much of an argument for avoiding whole nuts and avocados.  Recent controlled trials and meta-analyses have also dampened my enthusiasm for the idea that omega-3 fatty acids have a major impact on cardiovascular disease risk.  Either the trials weren't long enough to see protection, or omega-3 isn't as powerful as we had hoped.

The data on saturated fat are consistent with my long-standing position.  If saturated fat consumption impacts cardiovascular disease risk, the effect must be small because it's difficult to detect.  However, that doesn't mean it's prudent to eat a bacon and butter diet.  These studies reflect typical dietary patterns and have nothing to say about extreme diets.  Personally I'm wary of diets very high in saturated fat because we don't have much good quality data on them, and contrary to what is claimed in some circles, it probably does not resemble what our ancestors ate.

One of the associations that was identified in this meta-analysis is that circulating margaric acid is associated with lower cardiovascular risk.  This is consistent with the findings of my review paper on full-fat dairy, suggesting that the consumption of full-fat dairy is associated with a lower risk of obesity and metabolic dysfunction (2).  We don't know whether those results reflect cause and effect.  However, we do know that there is little evidence to suggest that dairy fat is harmful to health.

Overall, this meta-analysis argues that dietary fat composition may not be a key factor in cardiovascular disease risk.  That leaves us with these important but uncontroversial factors: genetics, overall diet quality, body composition, physical fitness, hypertension, and cigarette smoking.


  1. Thanks Stephan, a sober take as usual. On the omgea 6 polyunsaturated fats point: Could we not see a meaningful association in these studies because they are mostly based in countries where the polyunsaturated fats are consumed so liberally that we cannot compare a large enough group with a very low intake, ie is there enough variability in the samples to detect an association?

    I'm reminded of the classic study design where you compare people who smoke 10 cigarettes a day with those who smoke 9, you are then very unlikely to see an association with lung disease and smoking in that cohort.

    Just a thought.

  2. "I still think it's probably a bad idea to eat a lot of refined seed oils-- the lipid equivalent of white sugar-- but I don't see much of an argument for avoiding whole nuts and avocados." I find it interesting, though, that the results didn't reveal any negative effects of PUFAs on cardiovascular health. Considering how much refined seed oils people tend to consume these days, one would expect to see some pattern in these results. I'm really confused, to be honest.

  3. The advice to eat from a wide variety of foods and not too much of any holds true. As in financial investing, spreading the eggs into many baskets is a sound choice.

  4. Insightful. Thanks for the information. It's good to see people changing their mind with evidence.

  5. Nothing there (at least as you report it) to offend me.
    The data for n=6 from oils and spreads, and n=6 from nuts, seem to point in opposite directions.
    Which shouldn't really surprise anyone, there is more to both these foods than n=6.
    Based on Boyd Eaton's Paleo game paper, and the recent mammoth fat analysis, it looks as if Paleo animal fat was around 1/3 PUFA, 1/3 MUFA, and 1/3 SFA.
    The PUFA was relatively high in DHA, EPA, DPA, AA etc - i.e very diverse and balanced very long-chain fatty acids, that you only find in animals.
    In that regard, the paper seems to back up the paleo premise somewhat. Diets that have less SFA, more PUFA today tend to do so by replacing animal fats with vege oils. So they will be low in these very-long-chain PUFAs.
    But at the end of the day, weak evidence of any sort proves nothing, except the truth of what Dr Aseem Malhotra said: "saturated fat is not the major issue".

  6. I've been predicting for almost two years now that full fat dairy will go the way of the egg!

    Thanks for the review Stephan, I value your analysis, and I don't know about you but I stopped suggesting fish oil daily supplementation sometime ago. If a person wants to do so I see no problem but I'm not completely convinced of the benefit. Could be wrong but I guess one day we may know.

  7. The problem with such broad meta analysis is that you can glop away the important bits.

    Total omega 6 in diet correlates in part with the omega 6 in what we feed to animals. The book Inflammation Nation suggests that this is the more important source of inflammatory omega 6 oils since humans don't build the longer chains efficiently. (Note that Mr. Fish Oil, Barry Sears, actually advocates a bit of GLA to get over that first bottleneck.)

    Another problem is what you do with those omega 6 heavy oils. Do you eat them as nuts, mayo, salad dressing or use them as a frying oil. And we can divide frying between deep frying and a bit of oil squirted on a restaurant's frying sheet. The latter gives me the same greasy mouth feels as partially hydrogenated oils. I wonder if the interior of my body has as much difficulty cleaning out this varnish as my mouth does.

    My little rant is not proof of either hypothesis. I'm just pointing out that some uses of omega 6 oils could be quite harmful without moving the aggregate numbers enough to be conclusive. Someone doing a meta analysis has to look for evidence for/against these hypotheses to come to any useful conclusion.

    Another quick bit: the Eskimos and Japanese eat a lot of their protein raw. I wonder if this pattern of under cooking applies more generally to eaters of omega 3 oils (or cold weather peoples for that matter). Raw fish or gently cooked meat are both rather different chemically from the charred formally-beef patties served at burger joints. But they all can fit under the heading "animal protein" if a scientist so chooses.

  8. Stephan- thank you as always for the informative writing.

    question- is "circulating concentration" different from "tissue concentration" ? trying to reconcile circulating concentrations of linoleic acid not being a factor against other things i've read, like the chart showing the correlation between tissue % HUFA and CHD incidence.

    it does seem like full-fat dairy is being vindicated this past year. is interesting to see the science evolve.

  9. I've never understood why the Paleo movement keeps demonizing n-6 fats. I have yet to see really good evidence to claim they are that dangerous. Most studies on insulin resistance, that I've seen, show that replacing SFAs with PUFAs reduces insulin resistance. I have also yet to see good (human) data that the fragile PUFAs would indeed be oxidized in the body and contribute to inflammation. Some time back I went over several studies at PubMed on this and couldn't find any good data to support that point.

  10. Hi Stephan,

    How do you define "overall diet quality"?

    I second George Henderson's comment.

  11. I notice that they did not include any dietary intervention with modified fat intakes (replacing SFA with PUFA or MUFA). A recent Cochrane review (Hooper et al. 2012) of interventions concluded that modification of fat intake
    results in a significant 14% reduction in CHD incidence. The only RCT's in Chowdhury et al. are trials with supplements.

    Which reminds me about a nice quote from Skeaff & Miller's 2009 metaanalysis: "... the evidence from cohort studies of dietary intake of fats and CHD is mostly unreliable (with a few exceptions) because most studies have ignored the effects of measurement error and regression dilution bias."

  12. It being St Pat's Day +1,

    when I saw a report of this on my Irish Times' website, I saw this future marketing slogan:



  13. The key to it all is something called exercise. Doesn't matter what we eat as long as we burn as many or more calories than we consume.

  14. Stephan,
    Per your review with Kratz and Baars:
    "...these findings may provide a rationale for future research into the bioactive properties of dairy fat and the impact of bovine feeding practices on the health effects of dairy fat."

    You know any published studies since, based on such research?

    I recall your analysis of the Masai, of their high dairy saturated fat intake, without apparent ill effect.

    It matters to me and mine because our land (S. Appalachians) has pockets of good soil, but oft between rock on hilly ground. Grows great grass but tough to till. So traditionally, dairy cows/goats, and a hog or two for whey/skim milk, turns this grass into high quality human foods.

    But many here have become scared of foods they grew up on, in part due to the modern, 'learned' mantra against sat fat. Even scared to drink Jersey milk in any quantity.

    So am wondering: With how much confidence may I share that dairy fat (esp. grass-based?) appears to protect against CV risk? Any studies underway to examine possible cause and effect? Also matters to me because I consume about a gallon day, in various forms, and it keeps my kids running!

  15. Hi Sam,

    It's possible.

    Hi Carl,

    I didn't mention it in the post for brevity, but I think oxidation is another reason to avoid polyunsaturated oils, at least for high-heat applications. High-oleic sunflower and safflower oils have been developed to improve heat stability and they're commonly available today. They have a fatty acid composition similar to olive oil.

    Hi Right Angle,

    Circulating concentration is not the same as tissue concentration, but the two should be correlated.

    Hi Seppo,

    The RCT evidence implicating saturated fat in insulin resistance in humans is weak. Most of the longer-term RCTs showed no effect of fat modification on insulin sensitivity. This is a topic I've written about a few times. Here are two examples:

    Hi raphi,

    Depends on who you ask. I would define it as an omnivorous whole food dietary pattern including plenty of unrefined plant foods. Low in soda, junk foods like pizza, chips, cake, cookies, white flour, sugar, etc.

    Hi Erik,

    Good point. It's been a while since I read the Cochrane review but didn't they include the Finnish mental hospital trial? That was a really poor study that yielded a huge effect, and it seems to have a large influence on the outcome of meta-analyses that include it.

    Hi pawpaw,

    I wish I could give you a definitive answer, but unfortunately nearly all of the data that pertain to this question are observational, and they must be viewed with caution. I'm not aware of any ongoing RCTs on full-fat vs. nonfat or vs. no dairy that are collecting outcomes like heart attack, diabetes, and obesity risk.

    The data in our paper, overall, didn't support the idea that full-fat dairy protects against heart attacks-- the associations were neutral overall. However, the studies that used fatty acid biomarkers tended to be more favorable toward dairy fat.

    It is a lot of work for no definitive answers, unfortunately. I still feel that our review paper provides a rationale for RCTs, but until they're performed we can't say anything definitive.

  16. I generally agree with the spirit of your "overall diet quality": omnivory & the absence of junk food. I doubt many 'bien pensant' would disagree...however, the definition you gave is more about what it's not than about what it is. It seems to me that the bar has been set a little low.

    Until there is a working definition of "diet quality" (even if imperfect) a lot of nutrition/health research will be talking past each other [simply because there'll be multiple definitions muddling the waters; sort of an "to each his own" paradigm at present.

    Do you agree with my sentiment?
    Where would you start when trying to define "diet quality"?
    Do you think Matt Lalondes attempt (via nutrient density & other variables) is a good one?

    I would define it as an omnivorous whole food dietary pattern including plenty of unrefined plant foods. Low in soda, junk foods like pizza, chips, cake, cookies, white flour, sugar, etc

  17. Stephan: The Cochrane analysis excluded the Finnish Mental Hospital Study and any cluster randomised trials. But it was included in Skeaff & Miller (2009) and Mozaffarian et al. (2010).

  18. The Cochrane meta-analysis included Oslo, even though they didn't considered it to be "free of systemic difference in care" and "free of dietary differences other than fat".

    Generally speaking, the fat modification (SFA >> PUFA) trials also decreased trans fats and/or junk food in the experimental group

  19. Many years ago I went to a talk by Richard Doll on cancer. In the question time he was asked about saturated fat, and said he never ate butter because it causes heart disease. His colleague Richard Peto had some data suggesting this. As it happened, a study had just been published in Science showing that dairy products protect against heart disease. I said so and he was not pleased. I talked to him afterwards, trying to mollify him, but it was no good. I was an enemy.

    A few years ago I tried communicating with Richard Peto about heart disease, telling him about the association with copper deficiency. Somehow this association has escaped notice here in Oxford. I got nowhere, and concluded that Peto was just as frightened of Richard Doll as I was.

  20. And, what about cancer?

    The effect of fat, generally, and omega 6 polyunsaturated fat, in particular, on cancer has not been studies as extensively as its effects on CVD--but what evidence there is suggests cancer might be a real concern. Look at the Lyon Diet Heart study or the Veterans Trial. Look at the numerous animal studies.

    I acknowledge that in recent years, new data cast doubt on the previously assumed mechanism for omega 6's potential adverse effects (i.e., it now appears very likely that the mechanism is not a result of too much AA leading to inflammation). That said, there are other plausible mechanisms, like epigenetic effects or oxidation/depletion of endogenous antioxidants.

    Depending on the mechanism, you might be right that nuts or avocados are fine... On the other hand, it also seems there is little risk in keeping omega 6 consumption reasonably low, so why not limit nuts and avocados out of caution?

  21. The two humans trials I mentioned:



  22. Steven Hamley: Yes, the Oslo study (the first one, by Paul Leren) was meant as a "cholesterol-lowering" diet, not particularly a high PUFA-diet. The participants were encouraged to eat more fish etc., but in the end about 70 % of their fat intake came from soybean oil.

    I'm wondering about the usefulness of measuring biomarkers of MUFA acid as a measure for intake. We convert oleic acid in vivo from desaturation of stearic acid, and it has been suggested that endogenously synthethized MUFAs have different effects from MUFAs from the diet.

    Perhaps that is why the correlation coefficient between circulating and reported MUFA was only 0,05?

  23. Finnish Mental Hospital Study is included in this new meta-analysis also (see supplements). An interesting nuance is that the analysis of prospective cohorts showed null results for saturated fat in both cholesterol (lipid) adjusted and non-adjusted models. Siri-Tarino's meta-analysis was criticized because it adjusted for cholesterol. Now we know that non-adjusted analysis give the same results.

    It's a pity that the authors did not run the substitution analyses for cohorts (SFA->PUFA) like in the pooled analysis by Jakobsen et al. 2009. I think it matters a lot what you have in place of sat fat.

  24. I look to new metrics of LDLp, CAC & CIMT to be used in diet trials. I have posted my results on LCHF diet without confounding co-variable of weight change. As a Lipidologist I question value of Good cholesterol vs. Bad cholesterol and large vs small particles.


  25. A recent Zoe Harcombe blogpost -

    has the following:

    "The Department of Health Dietary Reference Values remains the basis for UK dietary advice.[vii] It has not been updated in the past 22 years, despite epidemics of obesity and type 2 diabetes having developed in this time. This document recommends that “the population’s average intake of non-milk extrinsic sugars [free sugars] should not exceed about 60 g/day or 10 per cent of total dietary energy.” "

    This is pretty close to the Kitavans' daily intake of CHO from fruit, allowing for differing average body mass.

    (Thank you that series, way back when. I enjoyed it.)


  26. Are you aware of Heisenbug who argues that underlying the cigarettes and fiber is a change to colonic bacterial ratios?

  27. A practical question: I have been buying olive oil for most cooking, and even making mayo. It is more expensive that the other cooking oils, this study seems to indicate that I am wasting money??

  28. Thanks Stephen for this information. The nutritional guidelines and cardiovascular guidelines help me a lot, as always keep the healthy living. It is true that the key to it all is exercise even you eat a lot as long as you burn more calories!

    Deborah Koval

  29. Hi Stephan, can you please give your take on this very different opinion on the Chowdhury work?

    Apparently there was a lot of adjusting and fudging of numbers in order to get the results Dr. Chowdhury got.

  30. RLL, really the study is saying that decisions should be based on more than just sweeping guidelines unleashed in the media. Olive oil has benefits when olive oil itself is compared to other oils, which wasn't the case in these studies.

    Seppo, studies that purport to show that replacing SFA with PUFA reduces CDH risk are actually comparing people with low PUFA intakes (some of whom are probably EFA deficient) with people with high intakes. Essential fatty acids are essential, but adding amounts of linoleic acid high enough to lower serum cholesterol might be counter-productive as this increases cholesterol synthesis as well as uptake.

    "although cholesterol synthesis increases with high PUFA intake, the lowering of plasma LDL-C observed with PUFAs is likely due to other mechanisms, including redistribution of cholesterol between plasma and tissue pools and upregulation of the LDL receptor. In contrast, the observed increases in plasma cholesterol concentrations due to SFAs do not appear to be related to a rise in cholesterol synthesis. However, when intake of TFAs [trans-fatty acids] was compared to palmitic acid, an increase in both cholesterol synthesis and plasma LDL-C concentrations was observed, suggesting that cholesterol synthesis contributed to the higher concentration of circulating LDL observed with TFA intake"

  31. Stephan,

    My friend Karl Kaiyala mentioned you today at lunch. I am an epidemiologist and oral medicine specialist at UW. I found out in June of last year that I was APOE4 carrier. It helped explained that even though I had a healthy lifestyle as a competitive triathlete and relatively low sat fat diet, my LDL was always elevated. I went on a low sat fat diet and my LDL became normal.

    You may have seen the recent paper of a small RCT at the Seattle VA published in JAMA Neurology in August 2013 showing that people with mild cognitive impairment and healthy controls showed improvement in Alzheimers biomarkers like amyloid when they were fed a strict low sat fat, low glycemic index diet for 30 days. The effect in APOE4 carriers or homozygotes showed an even larger effect, which are people like me.

    What this tells me is that the days of saying "I like this diet" or " That diet is good" are probably over. It depends upon your underlying genotype, and other factors, which diet may be optimal to prevent disease and maintain wellness for any particular person. This is just the start of this movement, in my opinion.

    The group of APOE4 + is relatively small, about 13% of the US population. It is an ancient polymorphism that does not allow efficient clearing of saturated fat, compared to the 2 and 3 types, which are most people. Compared to other SNP's, it is a powerful gene with high heritability. I am new to your interesting and informative blog and I hope I am not repeating what someone else has said.

    Mark Drangsholt

  32. Hi Mark,

    Thanks for your comment. I agree with your point. I think that's a profound limitation of most biomedical research: significance is based on sample averages and individual variability is often ignored.

    I find it really interesting to look at the RCTs of this or that diet, and compare them to the anecdotes people report online. A number of people claim that a low-carb, higher-saturated fat diet generally improved their lipid profile, including reducing LDL. Others find that their LDL goes through the roof. If you look at the RCTs, LDL doesn't usually change much with that type of diet on average.

    Same with body weight. Most overweight people lose body fat on a low-carb diet, while a few actually gain fat. On average, they lose fat, so that's what we hear about.

    A lot of the variability must be genetic. APOE4 is one of the few alleles we've identified that seems to have a major effect on lipids. We still have a lot to learn. The big question is how do we predict who does best on which diet. Unfortunately the GWAS studies have mostly been disappointing in this department.

    PS- I did see the study you alluded to. The limitation in my view is that the experimental diet differed in many ways from the control diet, so it's hard to know what the relevant factor was. It does suggest that Alzheimer's progression may be sensitive to diet though, which is pretty cool.

  33. Hi again. As predicted by the Plant Positive blog post I sent you, there has now been published this article on the errors in Chowdhury's work:

    I'd be interested in your take on all this.

  34. A few years ago I added a bunch of butter to my diet and it took my total cholesterol from 181 to 393. I then cut out butter altogether and it went right back down to where it was after ~6-8 months. That told me everything I needed to know about the subject. I suppose I may be an APOE4 carrier as well.

    In any event, the odds of me simulating familial hypercholesterolemia via butter consumption and somehow sidestepping the health consequences of such a large LDL pool seemed highly unlikely. Keeping 200mg/dL from oxidizing seems doable, but 400?

    Relying too much on the results of studies can be very dangerous. Although, when I looked into it afterward, I did find a study that found the fatty acid composition of butter (high myristate) to be a potent cholesterol-booster, with coconut oil being next in line.

  35. Travis, 23andme confirmed me as APO E4/E4, and over the past 20 years, my total cholesterol has crept up from the 190s to the low 230s. My diet used to be heavily based on whole grains and beans, per new-age hippie vegetarian dogmas, and now it's moderate carb, highish fat with very little starch. My take on APOE4 is that it doesn't pigeonhole all carriers into a rigidly defined set of physiological patterns. For all the talk of APOE4 and lipids, the only thing that screws me over is excessive amounts of carbohydrate.

  36. Given the poor quality and dubious intent of this meta-analysis I would hesitate to draw any definitive conclusion from it.

    There have been significant errors pointed out (the use of _completely wrong_RR_, omission of major studies), leading to hasty corrections, but given the magnitude of the errors and the general lack of accuracy they suggest the calls for retraction are getting louder each day:

    IMO, this paper is just another unfortunate example of the many shabby meta-analysis apparently designed to make big media headlines and grab the maximum amount of attention instead of advancing our understanding of a given subject (in fact, such studies often effectively spread confusion and distract the public from the real state of knowledge, which drives researchers like Willet, having done years of tedious work to expand that knowledge, understandably mad). Such papers should ring an alarm bell regarding today's standards of scientific publishing.

    Mind you: Willet is the last one, you would suspect of being biased against dietary fat. In fact he always opposed the indiscriminate demonization of dietary fats even at a time when the overwhelming scientific consensus was against him. He was among the first to acknowledge the data rehabilitating full-fat dairy as heart-healthy and anti-obesogenic.

  37. If the danger with pufas do not show up until after several years would not a number of studies have to be excluded. In my mind most of the natural fats we eat are well supplied with antioxidants, also added to many omega-6 oils. After 5 years consumption Chris MasterJohn pointed out that cancer started to increase in the n-6 group, in the "veterans study". Not before. If each gram n-6 require x mg antioxidants per day to not spoil in the body, 6 years high consumption of n-6 requires a daily intake of antiox twice that of 3 years to preserve existing n-6 in use and store in the body. Is this similar to the "empire effect" that made the fall of the Roman empire inevitable?
    Is the effect of large pufa body stores similar in cancer and CVD?

  38. "We DO know that there is little evidence to suggest that dairy fat is harmful to health."

    To cardiovascular health, perhaps.

    But I thought I should flag up the growing evidence of possible associations between even relatively moderate dairy fat intake and prostate cancer:

    "Conclusions: These results support the hypothesis that dairy products and calcium are associated with a greater risk of prostate cancer."

    About a one-third increased risk for 2.5 or more servings a day vs 0.5.

    Wishing you lifelong health

    Ivor Goodbody


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.