This is the second Sigma Statement in our 'Diet and Cardiovascular Disease' series. If you have not read the previous statement, it is recommended that you do so, at it sets the stage for why blood lipids are discussed in this statement. You can read that previous statement here:
Cholesterol, Lipoproteins & Lipids: Understanding CVD Risk
This Sigma Statement will focus on the influence of diet on blood lipids. The evidence for causality between blood lipids per se and cardiovascular disease (CVD), will be discussed in the next Sigma Statement. You can join the email list to be notified when it is released.
Key Question: How does diet influence blood lipids?
Introduction & Context
In examining the effects of diet on health outcomes, it is important to understand the substitution effects of different dietary constituents. In nutrition, substitution or replacement is a vital concept because the question of whether a nutrient has a positive or harmful influence on health is invariably faced with the response: "compared to what?"
This comparison could take the form of comparing a high level to a low level of a nutrient’s intake (as a percentage of total energy); and in each case, comparing the effects of what it has replaced in the diet. For example, if a diet has 40% of its total calories coming from dietary fat, rather than 30% for example, then the inclusion of that extra dietary fat (and hence the foods that contain it) come at the expense of other nutrients/foods which could have made up that energy.
So when considering the impact of a certain nutrient, we must ask: "What foods and nutrients may be absent the diet as a result?"
These questions will be examined in more detail when we consider the evidence for the impact of diet on cardiovascular disease (CVD) and coronary heart disease (CHD), but for now it is an important factor to bear in mind as we begin to explore the influence of diet on blood lipids.
However, tightly controlled metabolic ward studies have provided insight into the effect of specific dietary constituents without any substitution (i.e. univariate effects) on blood lipids. So in this statement, we will explore the impact specifically of:
- dietary cholesterol
- saturated fat
- monounsaturated fat
- polyunsaturated fat
As a reminder, in our previous statement we explained how an overwhelming body of multiple, converging lines of evidence has established a causal role for LDL in atherosclerosis and CHD/CVD progression as a fact, beyond a hypothesis. Additionally, we talked about how the role of LDL-C in driving atherosclerosis is a cumulative, integrated exposure over the course of the lifespan. And whilst it is the number of atherogenic (ApoB-containing) lipoproteins that is the causal factor, both LDL-C and Non-HDL-C are strong risk factors and accurately predict risk in the majority of people.
So with all that, let's explore how the various nutrients impact these markers.
For many people, if we are looking at what aspects of diet may influence blood lipids including blood cholesterol levels, it may seem logical to look at the cholesterol content of foods. And certainly in the past there was an emphasis placed on avoiding foods high in cholesterol. However, you have likely heard a counter to this idea in more recent times, something to the effect of "dietary cholesterol has no effect on blood cholesterol levels". And whilst (as we'll discuss below) dietary cholesterol is not necessarily a target to worry about, to say that is has absolutely no effect on blood cholesterol is technically not entirely correct. Dietary cholesterol can influence blood cholesterol levels. However, we need to consider:
- the magnitude of effect
- interrelationships with other blood lipid-modulating dietary constituents
Meta-analyses of metabolic ward studies modelling the effects of dietary changes, indicated only a modest impact of dietary cholesterol on blood cholesterol levels. And this impact was lower than the effects of saturated fat (SFA), polyunsaturated fat (PUFA), and monounsaturated fat (MUFA). In tightly controlled feeding studies, a reduction in dietary cholesterol of 200mg (340mg/d to 140mg/d) correlates to an average reduction in blood low-density lipoprotein cholesterol (LDL-C) of a mere 0.11mmol/L (4.24mg/dL). For context, the current European Atherosclerosis Society 2019 guidelines for treating dyslipidaemia recommend achieving an LDL-C level of <2.6mmol/L (100mg/dL) , which may require a reduction of over 1.8mmol/L (70mg/dL) in some individuals. The net reduction in blood cholesterol from reducing dietary cholesterol also does not account for the magnitude of change from other dietary modifications, which have significantly greater impact. For example, foods high in saturated fat are also high in dietary cholesterol, and therefore the effect of replacing 5% of calories from SFA with PUFA would also yield a net reduction in dietary cholesterol. Consequently, replacing saturated fat with unsaturated fat, or complex carbohydrate, has over twice the effect on lowering blood cholesterol levels than reductions in dietary cholesterol.
As a result, the effects of dietary cholesterol on blood cholesterol levels are dependent on the levels of dietary cholesterol in relation to levels of saturated fat in the diet. This fact, evident from controlled feeding studies, is also consistent with predictive equations modelling the impact of diet on blood lipids. Two controlled studies illustrate this point.
The first, a metabolic ward study, examined the effects of four diets:
- diet low in dietary cholesterol and high in SFA
- diet low in dietary cholesterol and high in PUFA
- diet high in dietary cholesterol and high in SFA
- diet high in dietary cholesterol and high in PUFA
The diets thus varied in the ratio of polyunsaturated to saturated fats (P:S), and the basal diets contained 300mg dietary cholesterol. The high cholesterol diets had 750mg or 1500mg dietary cholesterol added per day, and found that in the the low PUFA, high SFA diets (i.e., with the lowest P:S ratio), both levels of added dietary cholesterol significantly increased blood low-density lipoprotein cholesterol (LDL-C). Conversely, in the high PUFA/low SFA groups, the addition of up to 1500mg dietary cholesterol had no significant impact on blood lipids.
The second study, a controlled intervention, examined whether there was an additive effect of dietary cholesterol on blood lipids in the context of a high SFA (60g/d) or PUFA diet. Added cholesterol in the high cholesterol diets was 630mg/d, and the high-cholesterol/high-SFA diet significantly increased total cholesterol from 164mg/dL (4.2mmol/L) to 193mg/dL (5.0mmol/L) over six weeks, and increased LDL-C by 25.4mg/dL (0.65mmol/L), again highlighting the relationship between dietary cholesterol and saturated fats.
It is important to bear in mind that these effects on blood cholesterol ultimately must be assessed in conjunction with the impact of dietary saturated fats on blood lipids, because the dietary intake of both is highly correlated. Indeed, the true effect of dietary saturated fat on blood lipids in predictive equations from metabolic ward studies cannot be ascertained if dietary cholesterol is not included. While dietary cholesterol is not benign in biological effect, in the context of a low saturated fat intake (and high P:S ratio), the effect of dietary cholesterol on blood lipids is incredibly small, as is the magnitude of reduction in blood lipids achieved through reduced dietary cholesterol alone.
However, there is a correlation between dietary intake of cholesterol and saturated fat, and tightly controlled feeding trials demonstrate an additive effect on blood lipids of high dietary cholesterol when consumed concomitant with a high saturated fat intake. This is consistent with the predictive equations modelled from hundreds of metabolic ward studies conducted over the past 70-years.
As alluded to above, the most well-established adverse effect on blood lipids occurs from saturated animal fat intake. Univariate analysis of SFA in metabolic ward studies (i.e., where the impact of SFA on blood lipids is assessed in isolation), demonstrates that SFA have the most significant blood lipid-raising effect in metabolic ward studies. However, substitution analyses are more relevant for diet, and the isocaloric substitution of 5% energy from SFA for PUFA produces the greatest reduction in blood lipids, with meta-analysis of metabolic ward studies demonstrating a 14.9mg/dL (0.38mmol/L) reduction in total cholesterol.
Image 2: Clarke et al., BMJ, 1997 Jan 11; 314(7074): 112–117
That the degree of saturation of dietary fatty acids was an important determinant of blood lipid levels was evident from early metabolic ward studies comparing animal to vegetable fat. The first robust metabolic ward feeding study comparing the effects of vegetable fat and animal fat on blood lipids found that, while vegetable fat resulted in significant reductions in blood cholesterol levels, substituting the vegetable for animal fat in the patients resulted in increased blood lipids. Reverting the patients to the high vegetable fat diet lowered the concentration of blood cholesterol again. A further four month metabolic ward study demonstrated a 20% reduction in blood lipids when plant fats were substituted for animal fats.
A further controlled feeding study comparing subjects habitually consuming a low-fat, low-cholesterol diet, compared to two men with coronary heart disease (CHD), found that feeding a background diet of 50g animal fat from butter, tallow, and beef, in controlled amounts led to significant increases in blood lipids, effects which were not observed when olive oil was provided up to 60% of total calories. Adding 100g extra fat to the background diet from vegetable oil sources of unsaturated fat significantly reduced blood cholesterol levels, which was reversed on removal of the additional unsaturated fat. Beveridge et al., in a 1955 metabolic ward study comparing vegetable to animal fat, and with varying levels of added cholesterol, found that a diet containing 58.5% calories from animal fat (butter), significantly increased blood lipids, while feeding vegetable fat led to a pronounced reduction. Importantly for much of the criticism of the lipid-hypothesis (the scientific theory of a causal relationship between high blood cholesterol levels, atherosclerosis, and heart disease), these early metabolic ward studies had already established two facts:
- That the lipid-raising effects of different fats was largely independent of dietary cholesterol.
- That the P:S ratio was a fundamental determinant of the impact of fat composition on blood lipids.
However, early studies indicated that not all fat sources had the same impact on blood lipids, even within the same class of fat subtype, and different high-saturated fat containing foods, like coconut and cows milk, were also shown to increase blood lipids to differing extents. In fact, certain researchers questioning the lipid hypothesis did so on the basis of experiments using whole-milk. While not aware of this fact at the time, current literature indicates that the impact of dairy fat from whole-milk sources on blood lipids is not equivocal; several studies have found that, at the same level of saturated fat, butter will have a more significant lipid-raising effect than cheese. This may relate to their degree their odd-chain length* (C15:0 and C17:0 in particular), and the presence of the ‘milk-fat globule membrane’, which may exert protective effects through downregulating cholesterol synthesis in the liver.
*Side Bar: Fatty Acid Types - Click button below for detailed explanation
Indeed, meta-analyses of controlled feeding studies indicate that the greatest lipid-raising effects are observed for the even chain C12:0, C14:0, and C16:0 saturated fatty acids, with C18:0 having a lesser impact. However, scrutiny of this issue through the lens of individual fatty acids is overly reductionist, and fails to reflect the fact that food is the fundamental unit in nutrition. SFA is the most significant contributor to the predictive equations of dietary impact on blood lipids, and the strongest predictor of increased blood lipid levels. The extent of this effect, positively or negatively, will be mediated by the substitution effects of other dietary constituents, in particular PUFA.
The effects of monounsaturated fatty acids (MUFA) have historically been more difficult to fully elucidate, due to many studies using broad term comparatives, i.e., ‘animal’ vs. ‘vegetable’ fat, when in fact MUFA exist in both sources. This relationship was evident in the early metabolic ward feeding studies examining the impact of dietary fatty acids on blood lipids, where MUFA were shown to have little independent effect on blood lipids; however, the variations in MUFA levels were correlated with variations in SFA. Thus, increasing energy from one source may also concomitantly increase energy derived from the other. However, a meta-analysis of 248 metabolic ward studies could not find a correlation between MUFA and either SFA or PUFA, and found no independent effect of MUFA on blood lipids.
The lack of an independent effect was confirmed in the Clark et al. meta-analysis of 395 metabolic ward studies in univariate analysis (image 2 above). However, multivariate analysis modelling the effects of isocaloric replacement of 5% energy from SFA indicated an increase in high-density lipoprotein cholesterol (HDL-C), and a small reduction in total cholesterol (TC).
The difficulty teasing out the effects of MUFA is again a matter of relativity; What is the impact of the relationship between MUFA intake compared to other dietary constituents? What is the effect of isocaloric replacement?
In their meta-analysis of 60 controlled intervention feeding studies, Mensink et al. demonstrated a modest LDL lowering effect when MUFA are compared to carbohydrate. And the Clark meta-analysis demonstrated a modest reduction in TC when MUFA replace SFA. An issue which may mask the effects of MUFA per se may be use of the TC:HDL ratio as a surrogate outcome. MUFA raises HDL-C to a greater degree than PUFA, but PUFA lowers LDL-C to a greater degree than MUFA. Thus the effect on the TC:HDL ratio when either MUFA or PUFA replace saturated fat is comparable. However, while the Mensink et al. meta-analysis based their analysis around the TC:HDL ratio as a more sensitive predictor of risk, more recent analysis, in particular from the Emerging Risk Factors Collaboration have found that the TC:HDL ratio does not offer any superior marker of CVD/CHD risk prediction.
The favourable effect of MUFA specifically on blood lipids may be a question of the magnitude of increase over SFA. A meta-analysis of controlled feeding studies by Cao et al. compared moderate vs. lower fat diets where the difference in total fat was derived from increasing MUFA (23.6% vs. 11.4%, respectively), while SFA and PUFA were relatively constant. Compared to the lower fat (i.e., lower MUFA) diets, the higher MUFA intake resulted in a mean increase of HDL of 2.28mg/dL (0.05mmol/L). Triglycerides were also significantly reduced.
Cumulatively, the data from controlled feeding studies indicates that the primary effect of MUFA is increasing or preserving HDL-C levels, with a modest LDL-lowering effect. The magnitude of increasing HDL-C and improved overall lipid profiles appears to be dependent on the extent of MUFA increasing at the expense of SFA, and the food sources providing the diet's MUFA.
Due to the pronounced relationship between PUFA, SFA, MUFA, and blood lipids, much of the impacts of PUFA have been highlighted to a degree already. A cumulative body of evidence (including isolated effects of PUFA on blood lipids and isocaloric replacement of energy from SFA with PUFA) demonstrates unequivocally that PUFA have the greatest impact on reducing LDL-cholesterol, triglycerides, and overall blood lipid profiles.
The magnitude of effect of PUFA on blood lipids has long been recognised to correlate to SFA. The ratio of polyunsaturated to saturated fat (‘P:S ratio’) was established from metabolic ward studies in the 1950’s, reflecting the greatest magnitude of lowering blood lipids occurs with a reduction in SFA and concomitant increase in PUFA. Metabolic ward studies by Kinsell et al. and Ahrens et al. demonstrated that alternating periods of substituting animal fat for vegetable fat led to significant increases followed by decreases, respectively, in blood lipids. The controlled feeding studies by Keys et al. manipulating dietary fat composition within diets ranging from 9-44% dietary fat established the ‘Keys Equation’, based on the findings that SFA increased blood lipids by twice the amount that PUFA lower them. The equation, which remains valid, supports the P:S ratio as the primary determinant of blood lipid levels in humans.
it was established through the 1950’s that the process of industrial hydrogenation altered the chemical composition of PUFA from a cis to a trans configuration, forming trans fatty acids (TFA). It was also found that TFA increase serum lipids to the same extent as SFA. In this regard, notable studies that subsequently questioned the beneficial effect of PUFA over SFA, in particular the Sydney Diet-Heart Study, were confounded by the fact that the nominally PUFA-based interventions in fact contained hydrogenated, trans-PUFA, thus resulting in adverse effects on blood lipids and CHD risk.
The body of evidence from the early metabolic ward studies indicating the primary importance of PUFA, and the P:S ratio, in determining blood lipid responses in humans, has been confirmed by a voluminous body of 248 and 395 metabolic ward studies incorporated into respective meta-analyses. This is supplemented with converging evidence from randomized controlled trials which supports lower blood cholesterol levels from PUFA-rich diets compared to diets high in SFA (18-20%). Mozaffarian et al. demonstrated in a meta-analysis of controlled feeding studies that the isocaloric replacement of 5% energy from SFA with 5% PUFA reduced LDL-cholesterol by 10mg/dL (0.25mmol/L) and TC by a mean of 29mg/dL (0.75mmol/L), corresponding to a 12% reduction in heart disease risk for each 18mg/dL (0.46mmol/L) reduction in TC.
Similar to dietary fat, the effect of carbohydrate on blood lipids is dependent on type and food source. Broadly speaking, carbohydrates have a neutral influence on blood lipids, when compared to the benefits derived from substituting saturated for unsaturated fats. Dietary fibre has the most well-established blood lipid-modulating effect of carbohydrate. Soluble fibre and beta-glucans [found commonly in oats, barley, and legumes] exhibit the greatest lipid-lowering effect, with an average 3-5% LDL-C reduction, and range of 4-17% reduction from 5-17.5g fibre a day. It should be noted that the magnitude of effect differs relative to fibre type and dose, however, overall a linear relationship exists between levels of intake and magnitude of LDL-C reduction.
The broad category of ‘complex carbohydrates’ was examined in the comprehensive metabolic ward study meta-analysis by Clark et al. The reduction in blood lipids from isocaloric substitution of 10% energy from SFA with complex carbohydrate was found to be 20.07mg/dL (0.52mmol/L). However, the magnitude of lipid-lowering from the same isocaloric replacement of SFA with unsaturated fat (MUFA + PUFA) is three times that of substitution with complex carbohydrates.
A common framing of the effects of higher carbohydrate diets is the increase in fasting triglycerides (TGs) relative to other dietary fats. This is often misconstrued as an adverse effect. However, in the context of higher-fibre carbohydrate diets that reduce LDL-C concentrations, this is a misnomer. High TGs (176 to 880 mg/dL) are an issue in the context of concomitant high LDL-C, as the atherogenic load consists of the cholesterol in LDL, very low-density lipoproteins (VLDL), intermediate-density lipoproteins (IDL), and chylomicron (CM) remnants. However, evidence from familial chylomicronaemia indicates that isolated high TGs in the context of low LDL-C is not atherogenic per se, as cholesterol is primarily transported in CMs and large VLDL in this context, with these lipoproteins being too large to penetrate the layer of cells lining the inner arterial wall. Conversely, in the case of low TGs (<176mg/dL) in the context of elevated LDL-C, the primary atherogenic lipoprotein is LDL-C. Thus, the suggestion that phenotype of ‘high LDL-C but also low TGs’ as non-atherogenic is misleading.
Intake of added/free sugars may have deleterious impacts on blood lipids, driving what is termed the “atherogenic lipoprotein phenotype”: high blood LDL-C, low HDL-C, high circulating triglycerides (TGs), and a remodelling of LDL into small, dense lipoprotein subparticles. This may occur through an elevation in hepatic TG synthesis from de novo lipogenesis, precipitating an overproduction of VLDL and hepatic insulin resistance, and driving a cycle of elevated circulating lipids and impaired clearance. Controlled feeding studies have demonstrated significantly adverse effects on blood lipids from free sugars. While many of these studies are mechanistic and use doses not reflective of habitual intake, Livesey and Taylor demonstrated that 50g monosaccharide fructose may have adverse effects on blood lipids in the post-prandial state. However, at this juncture the majority of evidence suggests an indirect effect of added sugars on blood lipids, mediated by energy excess driving increased visceral adiposity and liver fat accumulation. Future research may be more informative by focusing on the post-prandial period for quantifying the effects of diet on blood lipids.
The totality of evidence supports replacing SFA with unsaturated fats for improving blood lipid profiles, based on:
- MUFA increasing or preserving HDL levels
- PUFA decreasing LDL
- Both MUFA and PUFA decrease TG’s (with the magnitude of effect being greater for the latter than the former)
Robust evidence exists for a beneficial effect of high dietary fibre intake from complex carbohydrate, while simple, added sugars have deleterious impacts of blood lipids, particularly underpinned by increased hepatic fat.
Therefore, to achieve a blood lipid profile conducive to cardiovascular and metabolic health, the cumulative weight of evidence supports a combination of:
- a diet low in SFA
- dietary fat intake coming predominantly from unsaturated fat
- a high dietary fibre level derived from complex carbohydrate sources
- a low free sugar intake
Thank you for putting out material on these topics at such a high level of detail. Extremely informative.
I am curious, however, why your work makes no reference to the many recent studies indicating no association between dietary saturated fat and CHD.
Are you able to comment at all?
Apologies in advance if I have not read/interpreted your work correctly.
I won’t speak on Alan’s behalf, so I’ll let him answer you directly himself. However, are there particular studies you find compelling?
It seems that most of the “no association” that gets mentioned in relation between saturated fat and CHD, relates back to a few meta-analyses (particularly the Siri-Tarino and Chowdhury ones). The problem with these papers specifically, and the “no association” rhetoric based on meta-analyses, are something that I know Alan has addressed. If I remember correctly it may have been mentioned in our Heart Disease podcast episode as well, and the issue of finding no association in bad meta-analyses was discussed on the recent Causality episode.
Many thanks for you comment!
First, this particular Statement was focused on the impact of diet on blood lipids, as the primary risk factor for CVD/CHD. We cover the associations between diet and actual CVD/CHD separately, in the final Statement: https://sigmanutrition.com/diet-cvd/
In particular, we cover the issue of overadjustment for blood lipids obscuring true associations between saturated fat and CHD in studies finding ‘no association’.
A second issue we detail in that Statement is the lack of definition for “high”, which also may obscure associations if a cohort has low absolute intake, or low % intake. In studies where is evident as >15-18%, the relationship with CHD is more clear.
A final point to bear is that many of these studies are meta-analysis, which is a recipe for general disaster because the included studies often vary in the comparison of ‘high vs low’ to such a degree that the ‘no association’ finding is practically meaningless.
Let me know if you’ve further questions, but hopefully the final Statement answers your questions!
All the best,
Hey Alan and danny. Can i get all the references that used in this statement. Specially added sugar and effect on lipid. Thanks
You should find all of the relevant references hyperlinked to the primary paper from the body of the Statement 🙂
I wasn’t able to find the references can you simplfy it. Sorry for the extra work
In the Statement, you’ll see parts of the text are highlighted in red. They are hyperlinked. If you click on those bits of text, it will open a new tab with the relevant study.
Hi – really enjoy your work and the podcasts – just halfway through the latest Prof Chris Packard one. I am not sure if I am misinterpreting it, but my reading is that the reduction in TC (0.76 mmol/l) from PUFA replacement for SFA as cited by Mozaffarian et al. (2010) meta-analysis was for the overall intervention (ie ~10% difference in [PUFA]), rather than per 5% substitution? I’m also interested in your interpretation of the Mozaffarian et al. (2010) study in terms of changes in LDL-C from PUFA. There is no LDL-C data presented in their results, but it is stated in the discussion ‘In short-term feeding trials, each 5%E of PUFA replacing SFA lowers low-density lipoprotein cholesterol (LDL-C) by 10 mg/dl, without an appreciable reduction in HDL-C’ – do you think they are referring to their own findings in that meta-analysis, or previous research?
Good to hear from you, and dlad you’re enjoying the work, thanks for the kind words. Having looked back over that particular section of the Statement, I think I’ll reword it to be more clear: the risk reduction was per the substitution analysis, but the effect of TC/LDL was derived from the weighted inverse variance of the trials (although strangely, they did not display in a figure…headscratch). Thanks for bringing this up.
In terms of the effects of substitution on LDL-C/TC, however, that is a very well-established body of literature going back to the 1950’s metabolic ward studies. They do reference the Mensink et al. 2003 analysis of 60 controlled trials in support of this (linked below), but there are two other prior papers which synthesised the metabolic ward research that are also worth reading:
Clarke et al. (1997) meta-analysis of metabolic ward studies.
Hegsted et al. 1993 also synthesis of the metabolic ward research.
Mensink et al. analysis of 60 controlled trials.
thank you very much for all your great work and your “translation” of science to “us normal people”.
Hoping my question is not to stupid:
It seems to me that the measure used is ration (two or more types of lipids) or percent (of total lipids)? What about absolute numbers – contextualised with persons’ weight, total calory intake etc.? I’m asking because it’s easy for me to have an say 50:50 ratio or greater of SFA:PUFA / MOFA as I’m (small) on low calories and eating mostly little fat. So one greek joghurt or peace of cheese and the ratio shoots up negatively.
Hope my question makes sense?
I understand, this question in fact comes up regularly as it relates to individual energy intake and body size.
I would think less about ratios, and think more about food selections. If greek yogurt and cheese are your main sources, this is not necessarily something to worry about based on our current understanding of dairy fats. If the remainder of your fat intake is from foods like olive oil, oily fish, nuts, etc., then you’re nailing the concept of fat balance we were discussing without even thinking of ratios 🙂
Hope that helps?
thanks very much for your answer
I feel confirmed in my understanding and choices 😉
all the best to you