From Ancel Keys & the

Diet-Heart Hypothesis to LCHF,

May Not Be a Huge Leap

By Prof. Grant Schofield and George Henderson

Ancel Keys has become a kind of cartoon villain for dietary reformers for various reasons – allowing ecological epidemiological comparisons to dominate his thinking, attacking John Yudkin’s sugar hypothesis with made-up ecological claims, and basically (with a lot of help) bullying his rivals out of their labs and grants so that their science couldn’t undermine his and that of his mates.

Is there redemption? Did we get something wrong?  Is he a good guy after all?

Ancel Keys had a first-class mind, and this was the main reason he won his battles. His work should not be ignored, any more than we should ignore the work of George Bernard Shaw or H.G. Wells because their ideas of social improvement led them eventually to support eugenics and dictators like Stalin. In this paper on the Roseto effect, Keys performs a valuable function – he debunks an early example of what would today be called ‘psychobollocks’.[1] keys-roseto-2

The Italian immigrants of Roseto ate a diet high in red meat and animal fat (lard), and also high in starchy carbohydrate, the source of which no-one seems to have recorded. They avoided olive oil and had a seemingly low rate of heart attacks, supposedly due to relatively stress-free lives (the evidence for which was questionable). Keys shows that most of this supposed reduction in heart attacks is due to variation in coding the causes of death. And what is left over, he attributes to the Italians’ diets before immigration, and the fact that Italian immigrant diets had more monounsaturated fat and less saturated fat than the usual American diets. We also note (which Keys didn’t) that the inhabitants of Roseto drank wine instead of the soft drinks drunk by other Americans.

In this paper, Keys lays out the diet-heart hypothesis as it existed in 1965 –

But it is desirable to clarify the dietary problem.

STOUT et al. [2] discredit the hypothesis that dietary fat is important in atherogenesis
and its clinical complication in the form of coronary heart disease. Actually, the hypothesis that most proponents offer is not as simple as they indicate. One statement of the hypothesis is: “Atherogenesis in the coronary arteries is promoted by increasing concentration of cholesterol in the ß-lipoprotein in the blood plasma and this cholesterol concentration is raised by increasing the proportion of dietary calories supplied by saturated fatty acid, the poly-unsaturated fatty acids having a weaker opposing influence”.

A more modern and updated version of this, based on evidence from 60 controlled diet trials: [2] Just remember that Apo B is in LDL and similar particles, and tends to correlate with the small dense, and potentially harmful LDL sub-fraction. Keys knew about these, and they are resurfacing again today as an important marker of arterial health. Apo A1 is in the HDL sub-fraction which is associated with benefit.

Replacement of carbohydrates with SFAs did not change apo B concentrations. The cis unsaturated fatty acids, however, decreased apo B, and this effect was slightly stronger for PUFAs. SFAs and MUFAs increased apo A-I concentrations relative to carbohydrates. PUFAs did not significantly change apo A-I concentrations.

Apo B correlates to non-HDL cholesterol in people eating the normal diet, and less precisely to LDL cholesterol, whereas Apo A1 correlates with HDL cholesterol.[3]

How did we get from this to a low fat diet?

Below we show you how. The low fat diet has potential to provide some benefit in improving particles. But the low carb diet is much better.

If we compare the 10% SFA, 30% fat dietary guidelines diet with a typical LCHF diet with unrestricted saturated fat, and fat from a variety of sources, we can see that the LCHF diet actually has the potential to lower Apo B much more than the low fat diet, despite having twice as much saturated fat. It also has more potential to increase Apo A1. The absolute levels of Apo B and Apo A1, and the ratios between them, are actually better predictors of heart disease risk than are LDL, HDL and their ratios.[4]

The dietary guidelines diet:

10% SFA
20% UFA (5% PUFA)


Ignoring protein, only 20 % of this diet will lower Apo B, only 25% will increase Apo A1

LCHF diet

20% SFA
50% UFA (5% PUFA)

50% of this diet will lower Apo B, 65% will raise Apo A1.

For some reason, when Ancel Keys came to test his hypothesis in 1968 in Minnesota, people were so opposed to increasing fat in the diet that he could only try to lower Apo B by replacing about 9% of the SFA in the diet with an equal amount of linoleic acid from corn oil (an intervention which wouldn’t have increased MUFA).[5] This gave him a very small leverage in terms of Apo B compared to our example, offset by a negative effect on Apo A1.

He did this despite being on record as sceptical about the value of high linoleic acid (PUFA) intakes. Why? Maybe he had painted himself into a corner with his original ecological studies correlating fat with heart disease mortality. Maybe he had to compromise with the Harvard crew, Hegsted and Stare, who had already sold out to sugar interests. Even for someone as influential as Keys, setting up a big study was a co-operative effort that required the support of most of the major players in the field. In any case, the Minnesota study didn’t show that replacing SFA with linoleic acid reduced CHD mortality.

So, are there studies that support our hypothesis, that LCHF reduces CHD risk without SFA restriction? Just as in Keys day, few people can get funding for a long-term, large study that increases fat in the diet. The PREDIMED study is probably the only example; in this case, the diet arms with an increase in MUFA from olive oil (an extra 50g per day), or an increase in total unsaturated fats from nuts and olive oil (6 servings per week and 32g per day respectively), experienced lower risk of major CVD events compared to controls.[6] (Those in the control group received small nonfood gifts, which was nice.) This is not a test of a high-SFA diet, as the subjects were complying with a supposed Mediterranean diet (i.e. not any real one but the version created recently by academics), but it does test the effect of increases in MUFA.

The median follow-up period was 4.8 years. A total of 288 primary-outcome events occurred: 96 in the group assigned to a Mediterranean diet with extra-virgin olive oil (3.8%), 83 in the group assigned to a Mediterranean diet with nuts (3.4%), and 109 in the control group (4.4%). Taking into account the small differences in the accrual of person-years among the three groups, the respective rates of the primary end point (major CVD events) were 8.1, 8.0, and 11.2 per 1000 person-years. The unadjusted hazard ratios were 0.70 (95% confidence interval [CI], 0.53 to 0.91) for a Mediterranean diet with extra-virgin olive oil and 0.70 (95% CI, 0.53 to 0.94) for a Mediterranean diet with nuts as compared with the control diet (P=0.015, by the likelihood ratio test, for the overall effect of the intervention). HRs for all but stroke were non-significant after multivariant adjustment.

There are a few recent epidemiological studies showing reduced CVD risk in higher-fat populations, with little or no adverse effect of SFA, including the high-quality Malmö Diet and Cancer Study. (n=28,098; 1250 deaths).[7]

For men, a significant trend towards lower cardiovascular mortality in upper quartiles of total fat intake was observed (P = 0.028) with the RR for men in the fourth quartile being 0.65 (CI 0.45–0.94, P = 0.023) (Fig. 1). No difference was observed between quartiles of saturated fat intake for men. Having relatively high intakes of monounsaturated or polyunsaturated fats compared with saturated fats did not show benefit for either sex.


Cardiovascular mortality men, Malmo

Although this association only exists for men in Malmö, this is important for 2 reasons – men have a much higher cardiovascular disease risk compared to women (3-4x in Sweden during these years), and cholesterol and LDL-cholesterol – the imprecise Apo B proxys – correlate with CVD risk in men in this age group, much more so than in women. Mean fat consumption by men in the upper quartile was 47.7% of energy – a respectable amount.

Similar findings came from the recent Harvard paper on fat and mortality in the combined Nurses’ Health Study and Health Professionals’ Follow-up Study (we wrote about this a while back).[8]

After adjustment for known and suspected risk factors, dietary total fat compared with total carbohydrates was inversely associated with total mortality (hazard ratio [HR] comparing extreme quintiles, 0.84; 95% CI, 0.81-0.88; P < .001 for trend).

Again, the correlation between total fat and CVD mortality for men, but not women, is statistically significant: HR 0.86 (0.79-0.93) p.=<.001.
This was from a study of US health professionals, amongst whom bias against fat and saturated fat was presumably at epidemic rates
And, just to show we’re not biased the other way, here’s a follow-up the PREDIMED study in the high risk cohort – where saturated fat was indeed associated with CVD mortality (okay, it was probably just the sign of poor compliance overall) – a high intake of total fat was still associated with reduced risk. HR 0.58 (0.39, 0.86).[9]

And, in memory of Keys, here’s a recent ecological (between-country) study: [10]

We found exceptionally strong relationships between some of the examined factors, the highest being a correlation between raised cholesterol in men and the combined consumption of animal fat and animal protein (r=0.92, p<0.001). The most significant dietary correlate of low CVD risk was high total fat and animal protein consumption. Additional statistical analyses further highlighted citrus fruits, high-fat dairy (cheese) and tree nuts. Among other non-dietary factors, health expenditure showed by far the highest correlation coefficients. The major correlate of high CVD risk was the proportion of energy from carbohydrates and alcohol, or from potato and cereal carbohydrates. Similar patterns were observed between food consumption and CVD statistics from the period 1980–2000, which shows that these relationships are stable over time. However, we found striking discrepancies in men’s CVD statistics from 1980 and 1990, which can probably explain the origin of the ‘saturated fat hypothesis’ that influenced public health policies in the following decades.

If the older epidemiological studies available to the original diet-heart enthusiasts didn’t show the same protective effects of total fat intake, this is perhaps because there was more trans fat in the food supply a few decades ago, or perhaps because modern studies are better controlled for carbohydrate quality and other factors.

The higher intakes in these studies of usual diets are usually around 40-50% fat, so they’re not actually low carb diets, but they are getting close, whereas our hypothetical LCHF example was 70% fat (which still allowed a generous hypothetical 75g of carbohydrate per day).

However – if Key’s hypothesis, and the prevailing appearance in his day of fat and SFA as associated with CVD mortality in epidemiology, was able to launch dozens of long-term studies of fat and saturated fat reduction for prevention of CVD, why is it that today, with a more up-to-date version of that hypothesis and at least as much epidemiological and experimental support for it, there are no trials, and instead we’re getting reheated leftovers of the old trials and their feeble and uncertain results, in the form of endless meta-analyses of the same data sets?

If you’ve read this far, you may appreciate this French documentary about the diet-heart hypothesis, which has some cool old footage of Ancel Keys and his crew.
There’s a particularly shocking bit in the middle where Jerry Stamler goes mad and starts telling us to throw away egg yolks. This marks the exact point at which the diet-heart hypothesis became an official licence to produce thousands of profitable low fat processed foods, with no regard for their actual nutritional value

[1] Keys A. Arteriosclerotic heart disease in a favored community.  J chron Dis. 1966, Vol. 19, pp. 245-254.
[2] Mensink RP, Zock PL, Kester ADM, Katan MB. Effects of dietary fatty acids and carbohydrates on the ratio of serum total to HDL cholesterol and on serum lipids and apolipoproteins: a meta-analysis of 60 controlled trials. Am J Clin Nutr May 2003
vol. 77 no. 5 1146-1155.

[3] de Nijs T, Sniderman A, de Graaf J. ApoB versus non-HDL-cholesterol: diagnosis and cardiovascular risk management. Crit Rev Clin Lab Sci. 2013 Nov;50(6):163-71.
[4] Walldius G, Jungner I. The apo B/apo A-I ratio – a new predictor of fatal stroke, myocardial infarction and other ischaemic diseases – stronger than LDL and lipid ratios. Atherosclerosis. 2006;7(suppl):468. Abstract Th-W50.6.
[5] Ramsden CE, Zamora D, Majchrzak-Hong S et al. Re-evaluation of the traditional diet-heart hypothesis: analysis of recovered data from Minnesota Coronary Experiment (1968-73). BMJ 2016; 353
[6] Estruch R, Ros E, Salas-Salvadó J et al. Primary Prevention of Cardiovascular Disease with a Mediterranean Diet. N Engl J Med 2013; 368:1279-1290.
[7] Leosdottir M, Nilsson PM, Nilsson JA, Månsson H, Berglund G. Dietary fat intake and early mortality patterns – data from The Malmö Diet and Cancer Study. J Intern Med. 2005 Aug;258(2):153-65.
[8]  Wang DD, Li Y, Chiuve SE et al. Specific Dietary Fats in Relation to Total and Cause-Specific Mortality.JAMA Intern Med. 2016 Aug 1;176(8):1134-45.
[9] Guasch-Ferré M, Babio N, Martínez-González MA, Corella D et al. Dietary fat intake and risk of cardiovascular disease and all-cause mortality in a population at high risk of cardiovascular disease. Am J Clin Nutr. 2015 Dec;102(6):1563-73. doi: 10.3945/ajcn.115.116046.
[10] Grasgruber P, Sebera M, Hrazdira E et al. Food consumption and the actual statistics of cardiovascular diseases: an epidemiological comparison of 42 European countries. Food & Nutrition Research [S.l.], v. 60, sep. 2016.
A nice quote from reference [2]:
The efficacy of replacing SFAs with carbohydrates depends on the effects on body weight in the long term, and that effect is uncertain.

Leave a Reply