What, exactly, is the Dietary Guidelines for Americans Committee's case against saturated fat?

[Edit 7/04/2015]

This analysis of the observational evidence cited in support of the US Dietary Guidelines recommendation to limit saturated fat to 10% or less replaces the version I posted earlier, but I have kept that version and you can still read it lower down this post.

Dietary Guidelines for Americans Committee Report 2015

Critique of the evidence for restricting saturated fat, with emphasis on the evidence from meta-analysis (including prospective cohort studies and RCTs) of substitution of saturated fat with other sources of energy.

1) The claim that the evidence for substituting PUFA for SFA is “strong and consistent”, which refers to the first two Bradford Hill criteria, is incorrect.

"Strong and consistent evidence from RCTs and statistical modeling in prospective cohort studies shows that replacing SFA with PUFA reduces the risk of CVD events and coronary mortality.
For every 1 percent of energy intake from SFA replaced with PUFA, incidence of CHD is reduced by 2 to 3 percent."
(Part D. Chapter 6: Cross-Cutting Topics of Public Health Importance. Lines 603-606)

Bradford Hill defined a strong association as having a factor of 2 or more; the association between CHD mortality and PUFA for SFA substitution in meta-analysis is, at best, approximately 0.80. The claim of consistency is also not accurate; meta-analysis of this question pools studies which have shown both positive and negative associations between PUFA and CHD mortality and/or events. The existence of more than one study, within each meta-analysis, in which increasing PUFA and decreasing SFA was associated with increased CHD, refutes the claim of consistency.

2) The method of sub-group meta-analysis which was given particular emphasis by the Committee is less than 6 years old and its interpretation is still an open question.

“Regarding saturated fat, Question 5 was answered using the NHLBI systematic review and related AHA/ACC Guideline on Lifestyle Management to Reduce Cardiovascular Risk, which focused on randomized controlled trials (RCTs), as well as existing SRs (systematic reviews) and MA (meta-analysis) addressing this question published in peer-reviewed literature between January 2009 and August 2014.
Particular emphasis was placed on reviews that examined the macronutrient replacement for saturated fat.”
(Part D. Chapter 6: Cross-Cutting Topics of Public Health Importance. Lines 84-89)

As noted by the Committee, meta-analysis of prospective cohort studies and RCTs shows no independent association between SFA and CHD mortality; the method of sub-group meta-analysis which instead compares stepwise substitutions of PUFA for other nutrients dates from the 2009 study by Jakobsen et al.[1] The post-script of the Skeaff and Miller meta-analysis, 2009, testifies to the novelty of the Jakobsen et al. methodology, and how it was welcomed by two experienced epidemiologists who could not within their own analysis find evidence to support their anti-SFA position.[2] PUFAs are essential nutrients with countless bioactive metabolites, and are not just energy sources, and those energy sources that lack essentiality – SFA, MUFA, and CHO – seem to all stand in much the same relation to PUFA in these meta-analyses. The results of these meta-analyses may reflect both the essentiality and functionality of PUFA, and the effect of a reduction in energy from other sources, rather than the harmfulness of these energy sources per se.

3) Notwithstanding the 2 previous points, the evidence from meta-analysis of energy substitution, taken at face value, does not support a limit on SFA.

“Farvid et al. found dietary LA intake is inversely associated with CHD risk in a dose-response manner: when comparing the highest to the lowest category of intake, LA was associated with a 15 percent lower risk of CHD events (pooled RR = 0.85; 95% CI = 0.78 to 0.92; I²=35.5%) and a 21% lower risk of CHD deaths (pooled RR = 0.79; 95% CI = 0.71 to 0.89; I²=0.0%). A 5 percent of energy increment in LA intake replacing energy from SFA intake was associated with a 9 percent lower risk of CHD events (RR = 0.91; 95% CI = 0.86 to 0.96) and a 13 percent lower risk of CHD deaths (RR = 0.87; 95% CI = 0.82 to 0.94).”
(Part D. Chapter 6: Cross-Cutting Topics of Public Health Importance. Lines 577-589)

The Committee’s report quotes Farvid et al. selectively (above). The Farvid et al. meta-analysis found that 5 percent energy intake from LA replacing the same amount of energy from carbohydrates was associated with a 13% reduction in CHD mortality.[3] This is exactly the same as the reduction in CHD mortality associated with 5% energy intake from LA replacing the same amount of energy from SFA. A similar conclusion can be drawn from Jakobsen et al. and Mozaffarian et al. with regard to total PUFA (in fact this exact point – that PUFA can be substituted for either SFA or carbohydrate - is made by Dariush Mozaffarian in presentations).[4]
There are two additional notes relating to the substitution meta-analyses so far; the substitution of PUFA for carbohydrate is for all carbohydrate, a mixture of refined and unrefined, and the substitution of PUFA for CHO is slightly superior to substitution of PUFA for SFA with regard to some endpoints.
The Farvid et al. analysis is not cited as evidence in the section relating to carbohydrate. If energy substitution sub-group meta-analysis is considered meaningful evidence in favour of SFA restriction, why is it not discussed in the context of carbohydrate?

4) There is no discussion of an upper limit to benefit from PUFA.

The current average intake of LA by Americans is over 7% of energy, and this would likely be higher were it not for the restrictions on total fat recommended by previous DGA Committees. The average intake of SFA is, at 11%, close to the recommendation of 10% or less. A 5 percent substitution of energy from LA for energy for SFA would result in an LA intake of over 12% and a SFA intake of 6%.
There are countries in the world that have long had similar fat intakes, and these are the countries of the former Soviet Union, where sunflower oil has been the main cooking fat since Tsarist times.
These countries have some of the highest incidences of CHD mortality in the world.[5] In Poland, a former satellite of the Soviet Union, the replacement of sunflower oil with rapeseed oil was followed by a sharp reduction in coronary mortality. Although the abstract of the epidemiological study that reports this change suggests this effect has been due to an increased intake of ALA, the change also saw a significant reduction (estimated reduced by one half to two thirds) of the total PUFA in cooking oil and of its LA content (reduced by two thirds or more). This study was co-authored by Walter Willet of Harvard School of Public Health, who also co-authored the Farvid et al. and Jakobsen et al. meta-analyses.[6]
The suggestion is that there may be evidence relating to the upper limit of LA safety available, and that this is a subject for discussion, not only with regard to CHD mortality but also with regard to non-CHD and all-cause mortality.

5) There is no evidence of a benefit from fat substitution on all-cause mortality.

If substitution of PUFA for saturated fat reduces CHD without adverse effects on other outcomes, we would expect overall mortality to be reduced. Death is measured with less error than any other disease-specific outcomes. Focus on overall mortality avoids the risk of concluding that an intervention improves one endpoint, but, in reality, is offset by harm to another. For example, a treatment may reduce CHD but increase cancer incidence, so that the effect on overall mortality is neutral. This is possible in those meta-analyses of energy substitution where only CHD endpoints are reported.[7] The 2012 Cochrane review by Hooper et al., of randomised studies designed to test the hypothesis that saturated fat influences CVD, showed no association between treatment arm and overall mortality (pooled relative risk 0.98, 95%CI: 0.93–1.04, 71,790 participants, 4292 deaths). The analysis by Mozaffarian et al. found discordance between CHD mortality (RR 0.80) and total mortality (RR 0.98, non-significant). The question of whether the reduction in CHD mortality associated with substitution of energy from PUFA for an equivalent amount of energy from SFA is also associated with an increase in non-CHD mortality from all causes should be resolved.
The Government may not have a mandate for decreasing the rate of morbidity and mortality from one disease by a recommendation that increases morbidity and mortality from other causes; the evidence for, and the legal and ethical implications of this question should be part of the discussion.

6) The explanation given for the lack of benefit from substituting MUFA for SFA in meta-analysis is not supported by evidence and amounts to special pleading.

“Evidence is limited regarding whether replacing SFA with MUFA confers overall CVD (or CVD endpoint) benefits. One reason is that the main sources of MUFA in a typical American diet are animal fat, and because of the co-occurrence of SFA and MUFA in foods makes it difficult to tease out the independent association of MUFA with CVD.
However, evidence from RCTs and prospective studies has demonstrated benefits of plant sources of monounsaturated fats, such as olive oil and nuts on CVD risk.”
(Part D. Chapter 6: Cross-Cutting Topics of Public Health Importance. Lines 617-621)

The use of “one reason” and “because of” incorrectly implies that the claim has been tested. These explanations are not given in the meta-analyses of Jakobsen et al. or Mozaffarian et al. which found a lack of evidence of benefit for replacing SFA with MUFA, but were first proposed as unsupported speculation in Martijn Katan’s 2009 editorial response to the Jakobsen meta-analysis, which is more nuanced than the Committee’s statement above, including a discussion of the confounding factors associated with animal fat consumption.[9] That there are evident benefits from both olive oil and nuts, traditional foods which supply bioactive components other than fats, should not be interpreted as evidence that the results of meta-analysis of the cohort studies available are more unreliable with regard to MUFA than with regard to SFA, PUFA, or other macronutrients.

[1] Major types of dietary fat and risk of coronary heart disease: a pooled analysis of 11 cohort studies. Jakobsen MU1, O'Reilly EJ, Heitmann BL, et al. Am J Clin Nutr. 2009 May;89(5):1425-32. doi: 10.3945/ajcn.2008.27124.

[2] Dietary Fat and Coronary Heart Disease: Summary of Evidence from Prospective Cohort and Randomised Controlled Trials. Skeaff CM, Miller J. Ann Nutr Metab. 2009;55:173–201 DOI: 10.1159/000229002

[3] Dietary linoleic acid and risk of coronary heart disease: a systematic review and meta-analysis of prospective cohort studies. Farvid MS, Ding M, Pan A. et al. Circulation. 2014 Oct 28;130(18):1568-78. doi: 10.1161/CIRCULATIONAHA.114.010236.

[4] Effects on coronary heart disease of increasing polyunsaturated fat in place of saturated fat: a systematic review and meta-analysis of randomized controlled trials. Mozaffarian D, Micha R, Wallace S.  PLoS Med. 2010 Mar 23;7(3):e1000252. doi: 10.1371/journal.pmed.1000252.

[5] European Cardiovascular Disease Statistics: British Heart Foundation Health Promotion Research Group, 2008. Allender S, Scarborough P, Peto V, Rayner M.

[6] Rapid declines in coronary heart disease mortality in Eastern Europe are associated with increased consumption of oils rich in alpha-linolenic acid. Zatonski W, Campos H, Willett W. Eur J Epidemiol. 2008;23(1):3-10. Epub 2007 Oct 23.

[7] Chewing the saturated fat: should we or shouldn't we? Thornley S, Henderson G, Schofield G. N Z Med J. 2014 May 23;127(1394):94-6.

[8] Reduced or modified dietary fat for preventing cardiovascular disease. Hooper L, Summerbell CD, Thompson R, et al. Cochrane Database Syst Rev. 2011 Jul 6;(7):CD002137. doi: 10.1002/14651858.CD002137.pub2.

[9] Omega-6 polyunsaturated fatty acids and coronary heart disease. Katan MB. Am J Clin Nutr. May 2009 vol. 89 no. 5 1283-1284.doi: 10.3945/​ajcn.2009.27744. 

[original version]

The 2015 DGA committee has released a 571 page document which is meant to inform the next dietary guidelines.[1] Changes are that % fat vs carbohydrate is no longer prescribed and cholesterol is no longer subject to a limit.
However, the old limit of 10% energy from saturated fat remains in place. Low fat or no fat dairy is the only dairy you're allowed. Meat? However poor or aged you may be, you should eat less of it. Although consumption of added sugars and refined grains is of concern, it always takes secondary place to the established evils of saturated fat and sodium.
Most of the document is dreadfully written and repetitively displays a circular logic. The healthy diet pattern (there are three of these, but they are interchangeable) is healthy (because it outperforms, slightly, a dummy version of the SAD diet); the healthy diet pattern avoids certain foods; ergo, these foods are not part of a healthy diet (even though they weren't an important part of the dummy SAD diet either).
Thus the verdict is repeated many, many times, and the prosecution does its summing up, and only then is the evidence presented. I'm familiar with this evidence. It's presented dishonestly.

Firstly, there is a nolo contendere acceptance of the evidence that saturated fat does not independently correlate with cardiovascular disease.Regarding saturated fat, Question 5 was answered using the NHLBI systematic review and related AHA/ACC Guideline on Lifestyle Management to Reduce Cardiovascular Risk, which focused on randomized controlled trials (RCTs), as well as existing SRs (systematic reviews) and MA (meta-analysis) addressing this question published in peer-reviewed literature between January 2009 and August 2014.
 Particular emphasis was placed on reviews that examined the macronutrient replacement for saturated fat.
The analysis of these pretends to be applying a truncated version of the Bradford Hill criteria. There's a good summary of these criteria and examples of their application here. There are 9 criteria and the first two are Strength of the Association and Consistency.

"Strong and consistent evidence from RCTs and statistical modeling in prospective cohort studies shows that replacing SFA with PUFA reduces the risk of CVD events and coronary mortality.
For every 1 percent of energy intake from SFA replaced with PUFA, incidence of CHD is reduced by 2 to 3 percent."

Part D. Chapter 6. page 16. (p451 doc)

Problem #1 - The evidence is not strong; in the meta-analysis by Dariush Mozaffarian et al., which is a meta-analysis supportive of substitution with PUFA, the average reduction in coronary mortality for 5% substitution was 0.80.[2]

Nowhere does the correlation attain the strength that Bradford Hill asked for, a factor of two or greater.
When the correlation is closest to one, as here, it can only be called weak. In fact it is even weaker than that, because the effect in primary prevention is not significant.

Emphasizing the benefits of replacement of saturated with polyunsaturated fats, Mozaffarian et al., 2010 found in a MA of 8 trials (13,614 participants with 1,042 CHD events) that modifying fat reduced the risk of myocardial infarction or coronary heart disease death (combined) by 19 percent (RR = 0.81; 95% CI = 0.70 to 0.95; p = 0.008), corresponding to 10 percent reduced CHD risk (RR = 0.90; 95% CI = 0.83 to 0.97) for each 5 percent energy of increased PUFA. This magnitude of effect is similar to that observed in the Cochrane MA. In secondary analyses restricted to CHD mortality events, the pooled RR was 0.80 (95% CI = 0.65 to 0.98). In subgroup analyses, the RR was greater in magnitude in the four trials in primary prevention populations but non-significant (24 percent reduction in CHD events) compared to a significant reduction of 16 percent in the four trials of secondary prevention populations.

From Ramsden et al. BMJ 2013 

Problem #2 - The evidence is not consistent, because there is more coronary mortality in some PUFA substitution studies, less in others, and no difference in others again.
You cannot use meta-studies as evidence of consistency!

The DGA committee also draw on the Harvard et al. meta-analysis by Farvid et al.[3]|

Consistent with the benefits of replacing SFA with PUFA for prevention of CHD shown in other studies, Farvid et al., 2014 conducted an SR and MA of prospective cohort studies of dietary linoleic acid (LA), which included 13 studies with 310,602 individuals and 12,479 total CHD events (5,882 CHD deaths). Farvid et al. found dietary LA intake is inversely associated with CHD risk in a dose-response manner: when comparing the highest to the lowest category of intake, LA was associated with a 15 percent lower risk of CHD events (pooled RR = 0.85; 95% CI = 0.78 to 0.92; I²=35.5%) and a 21% lower risk of CHD deaths (pooled RR = 0.79; 95% CI = 0.71 to 0.89; I²=0.0%). A 5 percent of energy increment in LA intake replacing energy from SFA intake was associated with a 9 percent lower risk of CHD events (RR = 0.91; 95% CI = 0.86 to 0.96) and a 13 percent lower risk of CHD deaths (RR = 0.87; 95% CI = 0.82 to 0.94).
Once again, the word "consistent" is abused. Individual studies are not consistent, and this is a meta-analysis (which is supposed to include all the relevant studies) so the concept of consistency does not apply. In what sense is an average consistent?

However an even larger deception is taking place in this selective quotation from Farvid et al., because that paper also concludes that a 5 percent of energy increment in LA intake replacing energy from carbohydrate intake is associated with similar benefits as replacing SFA.
Every meta-analysis that tells you that there is no benefit from replacing SFA with CHO, but a benefit from replacing SFA with PUFA, is saying the same thing, but Farvid et al. finally spelled it out.

9 cohort studies evaluating substitution of LA for carbohydrate showed that substituting 5% energy intake from LA for carbohydrates lowered risk by about 10%. A slightly lower risk benefit was seen for substitution of LA for SFA.This systematic review and meta-analysis suggests that risk of CHD decreases with higher dietary LA intake, when replacing either carbohydrate or saturated fat.

As a third criticism, how plausible is this claim - "for every 1 percent of energy intake from SFA replaced with PUFA, incidence of CHD is reduced by 2 to 3 percent"? With no safe upper limit set or implied.

For every 1 percent? Is the reduction the same for the 1st% and the 20th%?* And what of the observation that higher PUFA % intakes (like lower SFA % intakes) tend to be reported by those under-reporting calories? Is the correlation the same for absolute intakes (grams/day)?How is the suggestion to be placed in context? The calculations begin at 1%E as LA, yet the average dietary intake of LA in the USA was over 7% in 1999.[4] Is the case against saturated fat now to be based on chasing a PUFA target that for all practical purposes has already been met?

The graphic from Farvid et al. above shows that there is less data above 6-7% LA and correlations become less reliable. As Ancel Keys would have predicted - dietary intake of LA above 7% is not a usual part of natural human diets, and the range of intakes in the 7 Countries study was 3-7%. 
We are still in the "weak" range of correlation, meaning there could always be another explanation for what we are seeing. And we do not have all the data. The countries of the former Soviet Union have very low SFA intakes (6-7%) and very high LA intakes (unknown, but sunflower oil is the main cooking fat), and these countries have some of the highest rates of CHD mortality in the world. If we had reliable cohort data from these countries, what then?
And what of the elephant in the room of PUFA celebration - the lack of any association with all-cause, age-adjusted mortality? If PUFA substitution prevents CHD deaths, and CHD deaths are a major part of all deaths, then PUFA substitution should reduce all deaths. If it doesn't, then either the reduction in CHD mortality is illusory, or PUFA (or something associated with it) is causing more death from other causes. It doesn't.[5] Well there is a small, non-significant reduction, and the theory is that if this were multiplied to infinity by more and more studies it would attain significance and be interpreted as saving thousands of lives. As long as the new studies didn't come from parts of the world like Azerbaijan and, well, most of the rest of the world. But that the idea that a tiny association magnified means anything in a world of uncertainty, unreliability, and alternative explanations (known and hidden confounders) is nothing but clutching at straws.

Has all this effort and expense and messing with peoples' lives only had the result of sweeping the problem of CHD under the carpet of death from other causes?

There is also the following curious passage on MUFA. Remember that the lipid hypothesis recommends replacing saturated with unsaturated fats.

Evidence is limited regarding whether replacing SFA with MUFA confers overall CVD (or CVD

endpoint) benefits. One reason is that the main sources of MUFA in a typical American diet are

animal fat, and because of the co-occurrence of SFA and MUFA in foods makes it difficult to

tease out the independent association of MUFA with CVD.
However, evidence from RCTs and prospective studies has demonstrated benefits of plant sources of monounsaturated fats, such as olive oil and nuts on CVD risk.

That's some special pleading. Suddenly the methods used to separate SFA and PUFA, which the argument has depended on so far, are not good enough to separate SFA and MUFA, because they do not give the desired results. (The use of the words "one reason" and "because" above implies that these explanations have been tested; they have not; they appear in the literature as speculations). Animal fats - pork, and especially chicken - are a major source of PUFA in the US diet. Canola and other high oleic oils are sources of MUFA.
Nuts and olive oil, real high-fat foods, do seem to show benefits that don't show up when MUFA alone is measured. MUFA has been a big disappointment to epidemiologists, it lowers cholesterol when substituted for SFA, but this is not associated with a reduction in coronary disease. So the bogey of animal fat is invoked, without much justification and without the whiff of a mechanism to explain why oleic acid from canola oil should differ from oleic acid from beef (after all, cholesterol has just been acquitted). 

I know anecdotal evidence  has low admissibility, but all evidence is evidence of something. All over the internet and print media people will tell you that eating a lot less carbohydrate and more fat, sometimes more saturated fat, has improved their lives and their health. Doctors are saying this about their patients too.
Where are the blogs where people rave about how replacing butter with margarine has fixed their health problems? Millions of people take statins - where are the stories from statin users about the improvements to their lives? You will find more negative stories from statin users online. You might find stories of improved cholesterol, but where is the increased vitality and reversal of obesity and type 2 diabetes? Oh, wait.
You might conclude from this that any association between improvements in cholesterol and improvements in health is not necessarily a linear or temporal one. There is perhaps stronger evidence for the idea that improvements in health are temporally associated with improvements in cholesterol.

The DGAC are a bunch of brainy people, familiar with the evidence (some of them anyway), presenting a summary of this evidence to non-specialists - the Secretaries of the U.S. Department of Health and Human Services (HHS) and the U.S. Department of Agriculture (USDA).
How honest is their case?They present the observational evidence as being stronger than it is, and they suppress an important finding of this evidence which would contradict their saturated fat recommendation.
After all, if 7% PUFA is where the benefit lies (which is endlessly debateable and certainly not a case I'd personally want to make, especially in light of the all-cause mortality association), who eating either a standard American diet or one of the healthy "Healthy" DGA diets doesn't have a few % CHO to spare? And in that case, if you're willing to trade some sugar for some nuts, then where is the evidence against SFA? The observational evidence, weak though it was in terms of consistency and strength of association, just flew out the window.

Bye bye.


*Appendix 1

Walter Willet of Harvard, co-author of the Farvid et al. study, also put his name to this study, about a decline in CHD mortality in Eastern Europe where rapeseed oil has been substituted for sunflower oil.[6] Sunflower oil is about 44-75% PUFA, as LA, rapeseed oil supplies 15-30% PUFA, 15-20% LA.[7,8] This is evidence for the hypothesis that restricting PUFA or LA reduces CHD mortality.
Consistency.

Appendix 2

The following PUFA recommendations are listed on the website of the Linus Pauling Institute.

Upon request of the European Commission, the European Food Safety Authority (EFSA) proposed adequate intakes (AI) for the essential fatty acids LA and ALA, as well as the long-chain omega-3 fatty acids EPA and DHA (62). EFSA recommends an LA intake of 4% of total energy and an ALA intake of 0.5% of total energy; an AI of 250 mg/day is recommended for EPA plus DHA.
(note: this is about the average NZ intake)

The World Health Organization recommends an acceptable macronutrient distribution range (AMDR) for omega-6 fatty acid intake of 6-11% of energy and for omega-3 fatty acid intake of 0.5-2% of energy. Their AMDR for EPA plus DHA is 0.250-2 g/day (the upper limit applying to the secondary prevention of CHD).
(note - this requires use of seed oils and intensive fishing. The WHO are zealots for the diet-heart hypothesis).

The International Society for the Study of Fatty Acids and Lipids (ISSFAL) recommends for healthy adults an LA intake of 2% energy, ALA intake of 0.7% energy, and a minimum of 500 mg/day of EPA plus DHA for cardiovascular health.

[1] Scientific Report of the 2015 Dietary Guidelines Advisory Committee. link
[2] Effects on coronary heart disease of increasing polyunsaturated fat in place of saturated fat: a systematic review and meta-analysis of randomized controlled trials. Mozaffarian D, Micha R, Wallace S.  PLoS Med. 2010 Mar 23;7(3):e1000252. doi: 10.1371/journal.pmed.1000252.

[3] Dietary linoleic acid and risk of coronary heart disease: a systematic review and meta-analysis of prospective cohort studies. Farvid MS, Ding M, Pan A. et al. Circulation. 2014 Oct 28;130(18):1568-78. doi: 10.1161/CIRCULATIONAHA.114.010236.
[4] Changes in consumption of omega-3 and omega-6 fatty acids in the United States during the 20th century. Blasbalg TL, Hibbeln
JR, Ramsden CE, et al. Am J Clin Nutr. 2011 May;93(5):950-62. doi: 10.3945/ajcn.110.006643. 

[5] Chewing the saturated fat: should we or shouldn’t we? Thornley S, Henderson G, Schofield G. NZMJ 23 May 2014, Vol 127 No 1394; ISSN 1175 8716
[6]  Rapid declines in coronary heart disease mortality in Eastern Europe are associated with increased consumption of oils rich in alpha-linolenic acid. Zatonski W1, Campos H, Willett W. Eur J Epidemiol. 2008;23(1):3-10. Epub 2007 Oct 23.

[7] http://www.chempro.in/fattyacid.htm

[8] Chemical composition and stability of rapeseed oil produced from various cultivars grown in Lithuania. Dainora Gruzdienė, Edita Anelauskaitė.
http://www.icef11.org/content/papers/epf/EPF278.pdf