Butter or margarine?

Another paper has been produced by the Harvard team, which likes to analyse data for associations between food and health. This one is entitled “Saturated Fats Compared With Unsaturated Fats and Sources of Carbohydrates in Relation to Risk of Coronary Heart Disease: A Prospective Cohort Study”.

The researchers have looked at data already collected for two large studies: the Nurses’ Health Study (84,628 women) and the Health Professionals Follow-up study (42,908 men). The nurses were followed from 1980 to 2010 and the men from 1986 to 2010. All were free from diabetes, cardiovascular disease and cancer at baseline. Diet was assessed by a food frequency questionnaire every four years i.e. not very accurately.

The study results were that there were 7,667 incidents of coronary heart disease (CHD) in 24 – 30 years of follow-up. The results reported that:

Higher intakes of polyunsaturated fat (PUFAs) were significantly associated with a lower risk of CHD. Comparing the highest with the lowest quintile (fifth) gave a hazard ratio (HR) of 0.80 (0.73 – 0.88).

Higher intakes of carbohydrates from whole grains were significantly associated with a lower risk of CHD. Comparing the highest with the lowest quintile gave a hazard ratio (HR) of 0.90 (0.83 – 0.98).

Higher intakes of carbohydrates from refined starches/sugars were associated with a higher risk of CHD. HR was 1.10 (1.00 – 1.21).

The team then went on to claim that replacing 5% of energy intake from saturated fats with equivalent energy intake from PUFAs, monounsaturated fats or carbohydrates from whole grains was associated with a 25%, 15% and 9% lower risk of CHD respectively.

The overall conclusion of the study was “Our findings indicate that unsaturated fats, especially PUFAs, and/or high-quality carbohydrates can be used to replace saturated fats to reduce CHD risk.”

Issue 1

The first point to always remember is that this is association, not causation. Could it be the case that people who eat more fish (PUFAs) and/or eat more cous cous (whole grains) are generally healthier than those who eat fries and burger buns? This would make PUFA and whole grain consumption a MARKER of health, not a MAKER of health.

Issue 2

The numbers are always presented as relative risk, which is misleading and should not be done. Responsible researchers present absolute risk, full stop.

There were a total of 127,536 people studied. There were 7,667 incidents of CHD over a 24-30 year period (use 27 as an average). The chance of any person having an incident in any one year was 0.22% (about 1 in 450 people). Given that the nurses were 34-59 in 1980 (and therefore 64-89 in 2010) and the men were 40-74 in 1986 (and therefore 64 – 98 in 2010), I’m surprised that the incident rate wasn’t way higher than this. And I bet it was massively correlated with age, as heart disease always is.

The data

Table 1, baseline information, was quite interesting. It presented basic data about the participants in the five quintiles of saturated fat (SFA) intake, from lowest to highest. This informed us that those in the highest SFA group consumed almost double the total fat of the lowest SFA group and almost double the monounsaturated fat (MUFA) of the lowest SFA group. The highest SFA group also consumed almost double the trans fat intake of the lowest SFA group. I expect this adversely affected the health of the highest SFA intake group. This also tells me that the highest intake SFA group were eating processed food – another confounding issue.

Table 1 also told us that total carbohydrate intake ranged from 34.2% of energy intake (for the highest fat intake women) to 55.1% of total energy intake (for the lowest fat intake men). The carb intake from whole grains was as low as 0.72% – 1.2% of daily calories for the women and slightly higher at 2-5.2% for men. The whole grain results have been calculated on a tiny base, therefore.

The three key results presented above (on PUFAs, whole grains and refined carbs) came from Tables 2 and 3 in the paper. Table 2 also found the following, but this wasn’t included in the findings:

Higher intakes of total fat were significantly associated with a lower risk of CHD. Comparing the highest quintile (fifth) with the lowest quintile gave a hazard ratio (HR) of 0.88 (0.80 – 0.96).

Higher intakes of trans fat were significantly associated with a higher risk of CHD. Comparing the highest quintile (fifth) with the lowest quintile gave a hazard ratio (HR) of 1.20 (1.09 – 1.32). This was the single most significant result of the paper, but not mentioned in the press release or the abstract.

Higher intakes of saturated fat were (not statistically significantly) associated with a lower risk of CHD. Comparing the highest quintile (fifth) with the lowest quintile gave a hazard ratio (HR) of 0.93 (0.82 – 1.05).

The swapping sat fat for something else bit

The researchers estimated that the effect of replacing 5% of energy intake from saturated fats with equivalent energy intake from PUFAs, monounsaturated fats or carbohydrates from whole grains was associated with a 25%, 15% and 9% lower risk of CHD respectively.

First of all – that relative risk applies again. Remember that the incident rate for any person in any one year was 0.22%. Even if we take these relative risk estimates at face value, the incident rate would be 0.17%; 0.19% or 0.20% vs. 0.22%. Not exactly hold the front page now eh?

Second – replacing 5% of energy intake from saturated fats may not sound like much, but it’s huge (it’s not 5% of 9.6% – it’s taking 5% away from 9.6%). The intake of saturated fat (Table 2) ranges from 9.6% to 16.9% of total energy intake. The lowest quintile would need to more than halve saturated fat intake (from 9.6 to 4.6%) and the highest quintile would need to reduce it by approximately one third (to 11.9%).

Third – what would people actually be cutting back on if they cut saturated fat significantly? Figure 3.4, page 26, of the 2010 US dietary guidelines shows that the main sources of saturated fat are processed food: pizza; grain-based desserts; dairy desserts (ice cream); KFC; hot dogs; burgers; tortillas; candy; potato chips etc. Butter accounts for 2.9% of saturated fat sources and milk 7.3%.

Do you think health would improve if people halved their intake of saturated fat by halving their intake of this kind of junk? Of course it would. Would health improve by removing eggs, meat and dairy products from grass living animals from the diet? Quite the opposite.

The press release “Butter is not back” is therefore an inaccurate, misleading, cheap shot from a university that should know better. Doing this gave the ‘Harvard kings of association nonsense’ the headline they were looking for “Butter isn’t better than marg”.

Remember the most significant, not-reported, result in the whole paper? The one about transfats? Check out how liquid vegetable oils are processed into an unnaturally solidified butter-like substance and you’ll realise how dangerous this press release was.

What do we take away from this?

Sensible dietary advice remains unchanged – eat real food; don’t eat processed food. Butter, churned from natural food sources, is far better for you than hydrogenated (or some other solidifying process), bleached, deodorised, emulsified, coloured, gunge (a.k.a. margarine). Additionally: association studies, which selectively report relative risk, especially those that emanate from the Harvard ‘got it in for butter’ team are potentially bad for your health.

25 thoughts on “Butter or margarine?

    • Hi Linda
      Many thanks for that – very helpful :-)
      Best wishes – Zoe

  • I didn’t bother to look at the whole paper after reading this:


    We followed 84,628 women (Nurses’ Health Study, 1980 to 2010), and 42,908 men (Health Professionals Follow-up Study, 1986 to 2010) who were free of diabetes, cardiovascular disease, and cancer at baseline. Diet was assessed by a semiquantitative food frequency questionnaire every 4 years.”

    Seriously? Their conclusions are based on semiquantitative food questionaires every 4 years?????

  • I have written to both Flora and Benecol. While the discussion was about lowering cholesterol, references were cited. Once asked about the clinical benefits (such as reduced heart attacks etc)the discussion promptly ended.

    The manufacturers know that their products are a waste of time but rely on the “cholesterol myth” created by the medical establishment to flog drugs – a real snake-oil operation reminiscent of the 17th-19th centuries.

    A recent You-Tube presentation by Dr David Diamond likened “cholesterol lowering” to the 19th century practice of blood letting.
    Also his paper Expert Rev. Clin. Pharmacol. Early online, 1–10 (2015) is worth reading.

    The paper that is the subject of this blog
    Has so many flaws that it really is a farce. Furthermore it is NOT OPEN ACCESS. To me this suggests that the authors and the Journal (JACC) are scared stiff of letting independents accessing the contents reminiscent of Study 329. I have come to the point of not believing anything published by the “medical establishment” unless all the data has been subjected to review by independents as occurred with GSK’s Study 329.

  • another thing I don’t get – the RRs and trends for SFA and MUFA are exactly the same, so how do they get to predict any benefit from replacing SFA with MUFA?
    The RRs for SFA and refined CHO are different, so how do they get to predict no benefit from replacing refined CHO with SFA?
    The rationale and calculations for this don’t seem to be supplied in the paper.

  • I’m wondering if some of the corporate donors to Harvard were complaining about the recent studies that said that saturated fat was not bad for us and saying how happy they and then Harvard would be if a study came out that refuted those studies. It is just a thought.

      • That’s absolutely true, and it was always the case with “substitution” studies, which tell us nothing about saturated fat per se.
        But I think there’s something else going on. I think this is a rearguard action, and that the position defended in this hastily constructed paper is going to be abandoned.
        A rearguard action serves to delay one’s attackers and also, if successful, serves a moral purpose, in this case demonstrating that the authors weren’t wrong by their own lights and within their own terms of reference, even if they will need to widen their terms of reference and lift their game for their credibility to have a chance of survival in the long term.

        Let us take this paper at face value – who will gain by increasing wholegrains and PUFA by 5%E?
        We can say with most probability that this applies to the lowest quartile of intake in each category, and that there is very little basis for applying it to the upper quartile – there is no hypothetical test of such an application in the data.
        What will we gain most by replacing? First, trans fats (1.20), but there is very little of this to replace even in the upper quartile of intake. Second, refined CHO (1.10), and there is more of this available to replace than saturated fat (0.93).

        On the face of it, replacing refined CHO with anything other than trans fats will give more benefit than replacing SFA.
        This will eventually be the new position, but meanwhile they have laid down a smokescreen during their rearguard action to hide the exact location of the new position.
        After all, they’re not dug-in there yet.

        • “Let us take this paper at face value – who will gain by increasing wholegrains and PUFA by 5%E?
          We can say with most probability that this applies to the lowest quartile of intake in each category, and that there is very little basis for applying it to the upper quartile – there is no hypothetical test of such an application in the data.”

          If the data really can be extrapolated credibly for added PUFA and wholegrains only for the lowest quartile, by comparing the RRs between lowest and highest quartiles, and we have no measure at all of what increases would mean to the upper quartile, then the supposed benefit from the proposed substitution is only a quarter of that claimed by the authors. It certainly can’t have the full value.

          Unless they think the lowest quartile of intake of PUFA and wholegrains and highest intake of SFA and refined CHO represents the whole US population.

          • Harvard/Willett are notorious for blurring the differences between Omega 3 and O6 fats too, I predict major differences even from replacing some O6 with O3.

            ISTR Stephan Guyenet listing some early papers that showed negative results from increasing Omega 6 back in the day when there was an anatomically correct (non-excess) intake. Only when the majority of the population was overdosed on O6 did the studies showing “advantages” start to appear, suggesting it may adversely affect the metabolism of Real Fats. Rose Corn Oil Trial comes to mind. Of course since he became a Professional Obesity Researcher and deleted a bunch of his old blog posts this may not still be there.

  • Zoe, another worthless association study to further confuse people when reported in newspapers. I wonder what motivates the publication of such rubbish? Any publication is good?

    The Credit Suisse report refers to margarine as solvent-extracted vegetable oils. Sounds lovely. The report says that American consumption of these oils has risen by 169% since 1961 and that we now consume too much omega 6. These oils are believed to be inflammatory.

    So, butter made from milk or solvent-extracted vegetable oils? I think I’ll stick with the butter.

  • Ah, yes. Yet another “Saturated Fat is Bad For You” article coming out of the Harvard Medical School. These people are biased, in my opinion; they always seem to present “research” against fat and animal foods- (unless it’s seafood). In my opinion, the only way you can scientifically ascertain if one substance is better than another- (in this case, unsaturated fat versus saturated fat)- is if you observe a group of people all eating the SAME diet, minus the one element being tested- (types of fat, in this case). Otherwise, how can you prove anything? There are far too many variables.

    One blatant problem for me: the foods they suggested are “high in saturated fats” are actually mostly high in polyunsaturated and trans fats- (pizza, cakes, ice cream, KFC). Butter is relatively expensive these days, and to make a profit, most “junk food” is higher in soy oil and other fake fats. Sure, there may be some dairy- (like cheese in pizza)- but which part of the food is the detrimental element? Even most American ice cream is no longer just made with milk and cream; most all of them contain soy oil. (I know this because I have a soy intolerance and am highly aware of which foods I can and can’t eat). Speaking of which, it looks like the trans fat issue was quite muddled in this paper. They found trans fat to be a problem, and yet the headlines were all about saturated fat. Once again, if you’re going to vilify saturated fats, in order to “prove” anything, you have to separate out the foods that are just high in saturated fat, not trans fat as well. Has there been any correlation between extra virgin coconut oil and heart disease? I doubt it.

    It seemed as if they downplayed the refined carbohydrate/sugar issue as well; there was more of a “plug” for whole grains than a warning against refined grains. And I noticed they didn’t recommend we cut our sugar intake a certain percent, like they did with saturated fat.

  • Another point – these studies only used FFQs, repeated every few years.
    So first you have the inbuilt error of the FFQ, then you have the variation between them – Fat dropped from 39%E to 29%E over the course of the Nurses Health Study – so that gets averaged out.
    Surely all the RRs are within whatever the margin of error for that sort of process must be.

  • This is fantastic Zoe.
    The whole thing comes across as a lie – no-one’s eating more than a trace of butter, so why present it as evidence about butter?
    No-one’s substituting anything, so why present it as evidence about substitution?
    No-one’s eating 5%E more wholegrains than anyone else, so why predict what would happen if they did?

    One question – what was the quintile range for PUFA? If we add 5%E to the upper quintile, what crazy non-Keysian figure do we arrive at?

    • Hi George
      That’s a really good point – I was trying to work out from the paper what they had done to ‘estimate’ the swaps – I gave up!

      The PUFA quintiles are presented by nurses and HPFS:
      Nurses: 3.9% 4.9% 5.6% 6.4% 7.4%
      Men: 4.4% 5.2% 5.8% 6.4% 7.5%
      So add 5 points to each of these.

      Given how tiny the whole grain consumption was – hiking this by 5% of total energy would have multiplied the lowest intakes three fold – all fictitious as you say!

      Bye for now – Zoe

      • I say non-Keysian because in the 7 Countries Study, which is still the touchstone for these people, Keys et al say that the range of PUFA in the “natural diets” of the populations they studied was 3-7% from linoeic acid and a very little extra from other polyenes. (Nice word polyenes, saturated fats have alkane chains, unsaturated bonds are alkene).
        Being generous, we’ll say a range of 4-8%, which is exactly what we have here. There’s no authority to increase this in the 7 countries study.
        Interestingly, at least 2 of the trials in the 2015 Hooper et al meta-analysis of saturated fat reduction (WHI) involved reductions in PUFA from near 8% to near 4%. This did not increase CHD mortality or events. These were among the subgroup of larger, longer, and more recent trials, which showed a null effect.
        If there is a real effect of PUFA, this result is consistent with earlier studies correcting a relative deficiency, and modern populations having such high baseline PUFA levels that changes have no effect.

        • Keys also said – in the summary volume of the Seven Countries Study – “Serum cholesterol averages and CHD incidence rates were not found to be related to the percentage of calories provided by protein or polyunsaturated fatty acids in the diet” (I-194)

          But then he said that smoking and exercise weren’t related either, so he could have got everything wrong!

  • ‘As Nina Teicholz reports in her recent book, Willett is a big proponent of mythical olive oil drenched “Mediterranean Diet”…’

    I strongly recommend the section of her book in which Ms Teicholz describes how journalists and other opinion-formers were encouraged to praise the “Mediterranean Diet”. Luxury hotels, slap-up five-star dinners, superb open-air buffets, sun, sea and sand… and all free of charge, compliments of the food manufacturers (olive oil, etc.) and Mediterranean governments.

    And, of course, there IS no such thing as “the Mediterranean Diet”. People eat very different diets around the Mediterranean, and of course Ancel Keys’ original observations of the diet in Crete, which started the whole thing off, were inadvertently collected during the Lenten fast! Either Keys was extremely incompetent, or extremely dishonest (or both, of course).

    • Wasn’t Willett one of those who was wined and dined? (If I remember correctly?) In any case, he’s been very anti-saturated fat for a long time, so I, too, was not surprised to see his name on the paper. He also made an appearance in an English documentary about meat- (I believe it was called “Should We Eat Meat?”- something like that)- where he was one of the dissenting voices, due to “heart disease”. Funny, as most red meat is mainly high in monounsaturated fat, like olive oil. (Except lamb; which is, interestingly enough, quite popular in the: Mediterranean!) So many paradoxes, ho hum…..

  • Watching TV recently, I have been astounded to notice that Flora is still continuing with its ridiculous (if it were not so harmful) campaign to get people to eat margarine “to lower their cholesterol”. Of course, it’s a classic “proxy goal” – everyone has known for at least 20-25 years that the amount of cholesterol you eat doesn’t affect the amount in your blood, AND the amount in your blood doesn’t affect your health (unless it’s too low, which is definitely harmful).

    So Flora deceitfully (and, I’m sure, knowingly) advertise the proxy goal of lowering cholesterol – because many people, after decades of systematic brainwashing, “know” that cholesterol is bad for you.

    Someone should be put in prison for a very, VERY long time.

    • Tom, I couldn’t agree more.

      It’s actually been known since the 1960s that dietary cholesterol has no influence on blood cholesterol. I think even Ancel Keys admitted this very early on.

      • Keys did indeed. Here are a couple of paras from a literature review I have on all this:

        Keys undertook several trials to investigate if dietary cholesterol raised blood cholesterol levels and concluded that it did not. “It is concluded that in adult men the serum cholesterol level is essentially independent of the cholesterol intake over the whole range of natural human diets. It is probable that infants, children and women are similar” (p.54) (Keys et al., 1956).
        The most definitive statement made by Keys on this subject was recorded in a 1954 Symposium on Atherosclerosis: “The evidence – both from experiments and from field surveys – indicates that cholesterol content, per se, of all natural diets has no significant effect on either the cholesterol level or the development of atherosclerosis in man” (p.182 – original emphasis) (Keys and Anderson, 1954).

        The interesting thing about the second statement is that Keys admitted that the cholesterol content of all natural diets had no effect on atherosclerosis (not just cholesterol levels). To test this he needed to feed humans dietary cholesterol (he did a large number of such experiments). Dietary cholesterol is only found in animal foods (meat, fish, eggs, dairy). De facto, feeding animal foods had no effect on atherosclerosis or cholesterol levels. Not many people have spotted that…

        Keys, A., and Anderson, J. T. (1954) The relationship of the diet to the development of atherosclerosis in man. In: D. o. M. S. National Research Council, (ed.) Symposium on atherosclerosis. Washington, pp. 181-196.
        Keys, A., Anderson, J. T., Mickelsen, O., Adelson, S. F., and Fidanza, F. (1956) Diet and Serum Cholesterol in Man: Lack of Effect of Dietary Cholesterol. The Journal of Nutrition. Vol.59(1), pp.39-56.

        • When talking about serum cholesterol, most people fail to mention the genetic association between high-LDL alleles and CVD risk, which accounts for 50% of the correlation between LDL and CVD and is diet-independent.
          Of the remaining 50%, the effect of carbohydrate, and the symptomatic effect that other diseases can have on LDL, probably account for most of it.
          Most people who have heart attacks don’t have high cholesterol.
          Almost everyone who has a heart attack has a glycemic disorder according to at least one of the criteria for diagnosing diabetes.
          So where does the nutritional establishment put its effort?

    • Hear hear, Tom! That Flora Pro-Activ ad really irritates me, with the sixtysomething couple who find that she is “as bad” as he is in respect of cholesterol. I always shout at the TV “go and read the ingredients!” It’s very expensive to buy, and, of course, totally unnecessary, and I’m astounded that Flora are allowed to continue to show it, given the recent publicity surrounding blood cholesterol and dietary cholesterol. Are Flora still allied with the British Heart Foundation?

  • Thanks for the autopsy Zoe. The paper now requires a decent burial…

    The zombie lie of “saturated fat causes heart disease” refuses to die amongst the academics at Haavaaad and they continue to try and eat our brains.

    As usual, the paper is behind a pay-wall so mere mortals cannot even access the “hidden” information you have reported.

    This shameful paper resorts to all of the standard tricks of poor base data, multiple confounders, reporting of RR instead of AR, and failing to report data in the abstract which does not support their hypothesis.

    As Nina Teicholz reports in her recent book, Willett is a big proponent of mythical olive oil drenched “Mediterranean Diet” so I am not surprised to see him as one of the authors. It would appear that the zombies ate his brain years ago…

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.