Key Question: How does our diet influence risk of cardiovascular disease events?
Introduction & Context
This Sigma Statement draws together the evidence from the previous two statements in the Diet and Cardiovascular Disease Risk Series (see part one and part two). It is not intended to revisit the salient elements of those respective Statements in detail, so for the purposes of this statement we will be proceeding with a number of premises from the conclusions of the two preceding parts of the series, namely:
- A number of lipoproteins play a causative role in atherosclerosis development, and thus of cardiovascular disease (CVD) and coronary heart disease (CHD). These atherogenic lipoproteins are largely accounted for by low-density lipoproteins (LDL), but also include very-low-density lipoproteins (VLDL), intermediate-density lipoproteins (IDL), lipoprotein(a) (Lp(a)), and remnant lipoproteins. Given each lipoprotein carries one molecule of Apolipoprotein-B (ApoB), directly measuring ApoB provides a more complete assessment of all atherogenic lipoproteins in circulation.
- Lowering atherogenic lipoproteins reduces risk for CVD/CHD. The magnitude of reduction in risk correlates directly to the level of achieved lowering of atherogenic lipoproteins. For LDL-C, the optimal threshold currently appears to be around 70mg/dL (1.8mmol/L).
- Tightly controlled metabolic ward feeding studies over a 70-year period have demonstrated a hierarchy to the effects of different dietary constituents on blood lipids, as follows:
- Saturated fats (SFA) raise LDL-C and overall blood lipids to the greatest degree.
- Replacing SFA with polyunsaturated fats (PUFA) will lower LDL-C and overall blood lipids to the greatest degree.
- Monounsaturated fats (MUFA) may increase high-density lipoprotein (HDL-C) and have a slight LDL-C lowering effect; the net effect on total cholesterol may thus appear neutral.
- Dietary cholesterol independently has a minimal impact on increasing blood lipids. In the context of high dietary saturated fat intake, there may be an additive effect on blood lipids of high dietary cholesterol intake from foods high in saturated fat.
- High fibre, complex carbohydrates reduce LDL-C, however, the magnitude of effect is less than unsaturated fats, in particular PUFA.
- Refined carbohydrate adversely increase VLDL, and may influence remodelling of LDL-C into small, dense particles.
- Nutrients do not exist in a vacuum, and the most pronounced influence of diet on blood lipids is observed in the context of the relationship between different nutrients. In this regard, the Keys and Hegsted equations indicate that the most significant impact of diet on blood lipids, and CVD/CHD risk, is the relationship of SFA to PUFA, known as the ‘P:S ratio’. So for example, moving from a P:S ratio of 0.5 to a ratio of 2.0 would impart a strong lipid lowering effect.
The role of diet, therefore, in influencing risk of CVD/CHD will be discussed in the context of these conclusions, and in particular in the context of replacement nutrients. First, however, let’s deal with an important qualifier about the nature of diet-disease relationships and causation.
Indirect Effects, Over-adjustment, and Causation
This is a key point that is often missed in discussions about causation with regard to diet and disease outcomes: the effect of diet is always indirect, mediated by the impact of diet on different physiological parameters.
So, it is indeed correct to state that: dietary fat does not "clog the arteries". But atherogenic lipoproteins do! And diet influences the levels and characteristics of atherogenic lipoproteins.
It is incorrectly assumed that when looking at the relationship between diet and CVD/CHD risk, that the effect is a direct, straight-line relationship. But this isn’t the case. For example, the effect of saturated fat and polyunsaturated fat is mediated by their impact on, amongst other variables, LDL-C. The impact of sugar is mediated, amongst other variables, by the influence on de novo lipogenesis (the synthesis of fat from non-fat sources, for example, carbohydrate) and insulin resistance in the liver.
These factors are themselves influenced by mediating or supporting factors, such as what nutrient replaces another. For example, in reducing SFA whether it is replaced by PUFA or carbohydrate will differentially influence the reduction in risk associated with that change. This can be visualised using the graphic below (taken from Schisterman, Cole, & Platt, 2009), where in this context we assume that:
- ‘E’ = diet
- ‘D’ = CHD
- ‘M’ = blood lipids.
This illustration helps us conceptualise a major issue in analysis of prospective cohort studies examining the relationship between diet and CVD/CHD: over-adjustment. Over-adjustment occurs when the variable ‘M’ is taken out of the statistical equation. For example, this has commonly occurred in multiple recent studies purporting to find ‘no association’ between saturated fat and CVD/CHD, which equate to ‘E’ & ‘D’ respectively above. However, the studies often adjust for blood lipids (‘M’). It results in a weakened detectable relationship between ‘E’ and ‘D’, because the critical mediating factor has been taken out of the equation. The net effect is that the results are biased towards finding ‘no association’ between the exposure (‘E’) and the outcome (‘D’). This biases the results toward ‘no association’, when the relationship would be evident if examined through the lens of ‘M’, i.e., through the lens of the impact of diet on blood lipids.
Given the graphic above, it should now make sense (if it hasn't already) why we have tackled the diet-CVD topic in three separate statements, looking at three relationships (as seen in the above graphic):
- Relationship between lipids and CHD/CVD (M & D, respectively)
- Relationship between diet and blood lipids (E & M)
- Relationship between diet and CHD/CVD (E & D)
So this statement tackles that third relationship, as well as tying in the conclusions of the previous two statments, in order to answer the diet-CVD question comprehensively.
High Dietary Saturated Fat and CVD/CHD
First, as referring to ‘high’ or ‘low’ in relation to any dietary constituent can be arbitrary, it is important to contextualise what is meant by ‘high’. With regard to saturated fat, the historic context of ‘high’ is related to a saturated fat intake accounting for 16-24% of total daily calories.
The seminal Framingham study in the 1950’s established an exponential increase in risk from elevated total cholesterol (TC) levels. The CHD occurrence rate increased exponentially from 45/1,000 for those with a TC of 5.8 - 6.7 mmol/L (225-259mg/dL), to 122/1,000 in subjects with TC over 6.7mmol/L (260mg/dL). Whilst the Framingham paper did not assess dietary intake, it served, however, to highlight a fact that was established in the early epidemiology, across numerous cohorts in different countries: that where populations exhibited high blood lipid levels, they experienced high incidence of CHD. The reverse was true for populations with lower average blood lipid levels (defined at the time as <5.8mmol/L, or <220mg).
As discussed in our previous Sigma Statement, throughout the 1950’s it was consistently demonstrated in metabolic ward feeding studies that the most adverse effect on raising blood lipids occurred with increased levels of dietary SFA, and reversed when PUFA replaced SFA. This relationship between diet, blood lipids, and CHD was then examined in several cohort studies. The seminal study in this respect, the Seven Countries Study, established a relationship between saturated fat expressed as a percentage of energy, but also quantified absolute intake. So this indicated that the absolute intake of saturated fats (in grams) and the percentage of total calories made up by saturated fat were both high in populations with elevated blood lipids and high rates of CHD mortality. This relationship between SFA was strongly mediated by blood lipids, and persisted in both the 15-year and 25-year follow-ups.
Migrant studies also lent support to the relationship. Migrant studies can be highly informative, as they provide a way to examine the effects of changes in diet over time, when a population with the same genetic background moves from its habitual diet and adopts the dietary pattern of another region. In the Seven Countries Study, the Japanese cohorts were noted to consume an average of 2.9% SFA. Examining trends in dietary change in Japanese migrants was conducted in the Ni-Hon San Study. Comparing men in Japan with Japanese migrants in Hawaii and in California demonstrated an elevation of CHD risk in Hawaii and California relative to Japan that correlated with blood lipids, although the significantly elevated risk in the California group indicated that factors other than blood lipids, hypertension, and adiposity also influence risk. Interestingly, the observations were found despite the fact that a greater number of men in Japan were current smokers.
To put the levels of intake in both studies into more visual terms. The levels of intake in cohorts with high CHD mortality in the Seven Countries Study corresponded to:
In the Ni-Hon San Study, although the difference in CHD risk in the California cohort was not entirely explained by the traditional risk factors examined, consider the difference in diets; 6% of calories from SFA in the domestic Japanese cohort vs 26% in the Californian Japanese cohort.
In fact, the strongest weight of evidence from long-term cohort studies for the relationship between SFA, blood lipids, and CHD mortality, is observed from the Finnish population. Throughout this period of the 1950’s and 1960’s, the Finnish population suffered the highest global rates of CHD mortality, which was correlated also to the highest population-wide blood lipid levels, and a dietary SFA intake of 19-24% energy. In 1972 a concerted public health intervention targeted reducing four major risk factors across the population:
- blood lipids
- smoking rates
To achieve reductions in blood lipids, the campaign targeted SFA intake specifically through reductions in butter, which constituted the primary food source of SFA in the population. Over a 35-year period, CHD mortality decreased by 80%, of which reductions in blood lipids accounted for 67% of the mortality decrease. The population-wide reductions in blood lipids were primarily attributable to achieving a decrease in SFA intake from the average of 23% to 13%. What lends more weight to the Finnish example is that the changes in SFA intake, blood lipids, and CHD risk, occurred in the context of relatively no change in smoking rates and an increase in BMI across the population.
A number of important observations should be taken from this literature:
- The association with SFA and CHD is in the context of ‘high’ dietary intake, defined as a range of >16% energy, but most pronounced at intakes of >18-20% energy.
- The relationship between blood lipids and CHD is clear, and correlates to dietary intake, in particular a low P:S ratio (and consequently high SFA level).
- Reductions in blood lipids in populations reduces CHD mortality, and one way this may be achieved is through population-wide reductions in SFA from higher thresholds of intake.
Thus, if we take elevations in blood lipids as a primary endpoint, high SFA intake are strongly implicated in CHD risk. Further, the association may be stronger for ‘hard’ endpoints (i.e., mortality), than ’soft’ endpoints (i.e., incidence of a CVD event), with the evidence suggesting a significantly stronger associated between SFA intake weighted by persons-years of exposure (i.e., the longer the diet is high in SFA during a study period in a cohort) and CVD fatalities. In this regard, and in relation to the widespread public health recommendation of 10% threshold of energy, Hooper et al. modelled thresholds of intake relative to CVD outcomes, finding that the 10% threshold was where the greatest reduction in CVD events (note, not mortality; all events) was evident.
A final point bears stating, which is that wide variance in both dietary intake and blood lipid levels between people, and importantly within the same person, may result in difficulties in detecting associations in certain populations. Day-to-day dietary intake in individuals varies, and the blood lipid response to a meal, or day of meals, will therefore also vary. For example, someone who might habitually eat fish and intermittently eats fatty red meat would show different blood lipid responses following those meal types. As metabolic ward studies allowed for precise predictive equations for the effects of diet on blood lipids to be developed, it was also possible to develop predictions about detecting associations between diet and blood lipids at a population level. Factoring in errors in measuring dietary intake, it was possible to predict that the greater the level of difference in dietary intake, and the variance in blood lipids, the more it would be difficult to detect a true relationship between SFA and CVD/CHD. Adjusting for blood lipids only further attenuates the association.
Modelling the Substitution Effects of Different Nutrients
Given the long-established importance of the P:S ratio (ratio of polyunsaturated fat to saturated fat in the diet), the logical point of departure for this analysis is to examine the effects of substituting SFA for PUFA. At the level of epidemiology, consistent associations are noted between increasing PUFA intake and reductions in CVD/CHD risk. Meta-analysis of cohort studies indicated that each 1g/d increase in omega-3 alpha-linolenic acid (ALA) intake corresponded to a 10% reduction in CVD mortality.
There have been claims of a pro-inflammatory effect of omega-6 linoleic acid (LA), however these have never been verified experimentally (click button below for a detailed side bar on the PUFA-inflammation claims). Indeed a pooled analysis of 11 cohort studies found a 13% reduction in CHD mortality associated with substituting 5% energy from SFA with LA. Substituting 5% energy from SFA with PUFA generally was associated with a 25% reduction in CHD mortality in combined analysis of the Nurses’ Health Study (NHS) and the Health Professionals Follow-Up Study (HPFS).
These observations have been confirmed in controlled interventions. Compared to diets with high (18-20%) SFA intake, intervention diets rich in PUFA significantly lower blood lipids, and the substitution of SFA with PUFA reduces CVD risk. Mozaffarian et al. demonstrated in a meta-analysis of controlled feeding studies that substituting 5% energy from SFA with PUFA reduced LDL-C by 10mg/dL (0.25mmol/L), and for each 5% energy from SFA replaced with PUFA, CVD risk decreased by 10%. This is consistent with the meta-analysis of 15 RCTs by Hooper et al., which found replacement of SFA with PUFA led to a 15% reduction in risk for CVD events. Further, the most pronounced reductions in CVD mortality risk were evident with an average reduction of SFA of 8% from baseline levels of >18%. Thus, converging lines of evidence provide robust evidence that the substitution of SFA with PUFA leads to the greatest reduction in blood lipids, and in CVD/CHD risk and mortality. This is consistent with the predictive equations modelled on hundreds of metabolic studies on the impact of diet on blood lipids.
The effect of monounsaturated fats has historically been more difficult to elucidate. The body of evidence from metabolic ward studies indicated an overall neutral effect, but that was largely determined by examining TC as an outcome. Mensink et al. demonstrated that MUFA lower LDL-C and triglycerides, and other interventions indicate that plant-derived MUFA may increase HDL levels. This latter point warrants exploration, as MUFA are provided in mixed diets from both animal and plant sources, and recent evidence suggests that the source may be a reason why previous literature found mixed associations between MUFA and CVD/CHD risk. A meta-analysis of 11 cohort studies by Jakobsen et al. found that substituting 5% energy from SFA with MUFA significantly increased risk for myocardial infarction (MI). However, the primary source of MUFA in those cohorts was animal fat, and further trans-fatty acids (TFA) were included in the definition, and quantification, of MUFA.
This illustrates an important point; high animal fat diets will often translate into a concomitant high level of both SFA and MUFA in analysis. More recent research has conducted analyses having regard to the difference between sources of MUFA. In their analysis of the NHS and HPFS cohorts, Li et al. demonstrated that the replacement of 5% energy from SFA with plant-derived MUFA was associated with a significant reduction in CHD risk. A more nuanced analysis of these cohorts in 2018 demonstrated that the 5% isocaloric replacement of SFA with total MUFA was associated with an 8% decrease in CHD, conversely, however, substituting SFA with animal-derived MUFA resulted in no reduction in risk. Due to the high correlation between animal-derived MUFA and SFA, analysis of substituting of 5% energy from a mix of animal-MUFA and SFA with plant-derived MUFA was associated with a 19% reduction in relative risk for CHD. This effect of plant-derived MUFA has been confirmed in food-based intervention studies, in particular the OmniHeart Trial and the Lyon Diet-Heart Study.
Overall, it is evident that the primary benefit from MUFA is derived from plant sources. While the substitution of SFA with PUFA will lead to the greatest reductions in blood lipids and CVD/CHD risk, the requirements for essential fatty acids is relatively small relative to total energy needs, and thus any additional replacement of SFA in the diet beyond PUFA would preferentially be derived from plant-derived MUFA. This encapsulates that the totality of evidence supports replacing saturated fat with unsaturated fat in the diet.
The effect of carbohydrate (CHO) may be considered somewhat similar in analysis to the effect of MUFA: i.e. it is dependent on source and type. Certain prospective cohort studies that have purported to find no effect of replacing SFA with CHO have generally failed to separately quantify the effects of complex, fibrous carbohydrate (C-CHO) with refined CHO/added sugars (R-CHO). Because they have diametrically opposed effects (substituting SFA with C-CHO will moderately reduce risk of CVD/CHD; substituting with R-CHO will result in similar risk), a failure to analyse the respective effects separately results in a cancelling out effect, and ‘null’ association. In the Li analysis of the NHS and HPFS, the authors found that the substitution of 5% energy from SFA with C-CHO was associated with an 11% reduction in risk for CHD. Interventions with wholegrain CHO sources consistently demonstrate reductions in blood lipids, reduced blood glucose levels, and increased insulin sensitivity. It is hypothesised that these effects result from the synergistic properties of the bran, endosperm, and germ remaining intact in the whole grain. Interventions with very high dietary fibre intake also confirm a direct modulating effect of fibre on blood lipids and blood glucose. While cumulatively the data supports a reduction in CVD/CHD risk from replacing SFA with C-CHO, it should be noted that the magnitude of effect is smaller in comparison to the effect of replacing SFA with unsaturated fat.
The Hierarchy of Substitution for Reducing CVD/CHD Risk
A number of conclusions can be made, bringing this series and body of evidence together. First, the relationship between SFA and CVD/CHD is most pronounced at high - 16%-24% - levels of dietary intake. Secondly, this relationship is strongly and primarily mediated by the impact of diet on blood lipids. Third, population-wide interventions that are successful in reducing SFA intake are accompanied by a decline in CVD/CHD mortality and events. Fourth, the magnitude of effect is mediated by the nutrient replacing SFA in the diet, with the following hierarchy evident:
- Polyunsaturated fats (from plant and marine sources);
- Monounsaturated fats (from plant sources);
- Unrefined, complex/wholegrain carbohydrates.
The negative impact of TFA is becoming less of a factor due to their large-scale removal from the food supply. The replacement of SFA with R-CHO does not modulate CVD/CHD risk. While this has been interpreted to mean R-CHO is more of a risk than SFA, the more appropriate conclusion is that a diet high in one, either, or both, is deleterious for cardiovascular health.
The final point which is important to state here is that the totality of evidence no longer supports an emphasis on the total fat content of the diet as a determining factor, in particular evidence from the Mediterranean diet indicates that high total fat diets may be cardio-protective, however, it is instructive to note that such diets contain an average of 6-9% SFA and are comprised predominantly of UFA.
And that concludes this statement on diet and cardiovascular disease risk. Let's re-cap on the most important points from all three statements in this series...
Key Ideas From This Series
- An overwhelming body of multiple, converging lines of evidence has established VLDL, IDL, LDL, and Lp(a), to all be pro-atherogenic lipoproteins.
- ApoB provides a measure of the actual number of particles for all such atherogenic lipoproteins listed in point 1 (as each particle contains one ApoB molecule).
- Therefore, there exists a causal role of low-density lipoprotein (LDL), and/or ApoB-containing particles, in atherosclerosis and CHD/CVD progression.
- A measure of total cholesterol does not fully explain risk as it provides a measurement of cholesterol content within particles, rather than number of actual particles. However, a high total cholesterol measure may still be useful in risk screening or at a population level.
- Similarly, LDL-C measures the cholesterol content of LDL particles, rather than number of particles. In cases of discordance, where LDL-C and LDL particle number do not correlate, LDL-C may over-/under-estimate risk. However, in the majority of cases LDL-C is able to accurately capture risk.
- While high HDL levels have been shown to be protective in epidemiological studies, evidence remains lacking that there are therapeutic benefits to direct interventions aiming to increase HDL-C.
- While high TGs have strong associations with CHD/CVD, this association is not evident once non-HDL is adjusted for, indicating that it is in fact that it is triglyceride-enriched lipoproteins that are atherogenic, rather than TGs per se.
- While several claims in both the popular and scientific literature have been made downplaying the role of LDL-C, in particular referencing the ‘low TG, high LDL-C but also high HDL’ phenotype, there is little to no evidence that such a phenotype mitigates the atherogenic potential of high ApoB levels in circulation.
- Remnant lipoproteins warrant consideration, although the precise role of remnant lipoproteins in atherosclerosis remains to be fully elucidated.
- The relationship between SFA and CVD/CHD is most pronounced at high (>16% of daily energy) levels of dietary intake. This relationship is strongly and primarily mediated by the impact of diet on blood lipids.
- Population-wide interventions that are successful in reducing SFA intake are accompanied by a decline in CVD/CHD mortality and events.
- The magnitude of effect is mediated by the nutrient replacing SFA in the diet, with the following hierarchy illustrating greatest effect:
- Polyunsaturated fats (from plant and marine sources);
- Monounsaturated fats (from plant sources);
- Unrefined, complex/wholegrain carbohydrates.
- Trans fats have negative impacts on health. However, this is becoming less of a factor due to their large-scale removal from the food supply.
- The replacement of SFA with refined carbohydrate does not alter CVD/CHD risk. Therefore a diet that is high in either or both nutrient(s), is deleterious for cardiovascular health.
- The total dietary fat content of the diet is not a determining factor in atherosclerosis progression. Diets high in total fat intake may still be neutral or cardio-protective, depending on the sources and composition of that dietary fat. For example, high-fat Mediterranean diets, which can be concluded to be “heart healthy”, contain an average of 6-9% SFA and are comprised predominantly of unsaturated fat sources.
- To achieve a ‘healthy’ blood lipid profile, from a cardiovascular health perspective, the cumulative weight of evidence supports a diet that:
- Is low in saturated fat (< 10% of calories)
- Has it’s dietary fat coming predominantly from unsaturated fat
- Has a high dietary fibre level
- Is low in free sugars/refined carbohydrate
We recorded a review/recap podcast episode in order to address some of the questions, counterpoints and clarifications related to this topic. You can listen here or find the podcast on your podcast app of choice here.
If you haven't read our previous statements in this series, you can find them here:
To get notified of the release of future Sigma Statements, make sure you are on the Sigma email list, which you can join here.
I have listened to your content as well as Alan as a guest on other podcasts and appreciate your focus on science and not mincing words. I heard Peter Attia state on one of his recent podcasts that, in reviewing all of the studies he could get his hands on, he found no strong evidence for a relationship between SFA in the diet and ASCVD. After hearing that, I made my way back to your website to read your research. I believe your interpretation and conclusions around the science. I am diligently following the recommendations and will continue to test lipids to see if I can bring down relevant markers through diet and exercise. By the way, I feel great having cut out all but a few grams of SFA per day. I’d love to listen to Alan and Peter discuss this topic!
Thanks for the kind words. I haven’t heard the specific episode you’re referring to, so can’t speak directly to that. However, any time I’ve heard Attia discuss lipids and CVD, I have found him to be of the same position as Alan and I (as well as any respectable lipid expert!), specifically that LDL/ApoB is causal in atherosclerosis.
However, if he has gone on to say there is no good evidence for a relationship between high SFA and CVD, I’d find that odd. As our statements outline, the increase is risk is due to an increase in LDL-C (or ApoB). This happens to varying degree in individuals, but overall it’s pretty clear that high intakes of SFA increase LDL-C, compared to lower intakes. Any evidence we typically see cited to support the claim that saturated fat has no association tends to be studies with significant problems (e.g. Siri-Tarino 2010, Chowdhury 2014, etc.).
Hopefully this clarifies our position, that is to say: high intakes of SFA will carry greater risk than lower intakes, due to their ability to increase LDL-C and ApoB count.
Thanks for this great series of articles. My doctor recommends a medication called Ezetimibe which is meant to lower the absorption of saturated fat from your diet. It doesn’t change the amount of cholesterol your liver produces but for most of us its our lifestyle overtime that contributes most to cvd
what are your thoughts on this?
also you have said “Population-wide interventions that are successful in reducing SFA intake are accompanied by a decline in CVD/CHD mortality and events. ”
Why does Ezetimibe not lower CVD/CHD mortality and events? are the studies not long enough?
Yes, ezetimibe is a commonly prescribed drug with consistent evidence for safety and efficacy. Often it is used as an adjuvant therapy with statins, or can be used alone.
Which studies are you basing the conclusion that it doesn’t impact events on? There are conflicting results, but likely can be explained by methodological issues or which outcomes we look at.