硬脂酸的攝取量與其他飽和脂肪酸的攝取量高度相關，硬脂酸透過促血栓形成機制會增加心血管風險
減少飽脂肪酸（SFA）攝取量時，以n-6多元不飽和脂肪酸取代SFA可能比以碳水化合物取代SFA更能降低低密度脂蛋白膽固醇（LDL-C）水平，因為前者本身就具有降低LDL-C的潛力，而碳水化合物取代SFA則不具備這種潛力。
名詞與縮寫
saturated fatty acids (SFA) 飽和脂肪酸
coronary heart disease (CHD) 冠狀動脈心臟病
LDL cholesterol 低密度脂蛋白膽固醇
Introduction
The association between total cholesterol and coronary heart disease was
firmly established in the 1950’s and the potential of diet to influence cholesterol levels
quantified in the 1960’s. The landmark Seven Countries Study provided strongly
suggestive evidence of a causal link between nature of dietary fat and coronary heart
disease (CHD) which was mediated, at least in part, via cholesterol. In the 1970’s and
1980’s causality of the association was questioned by those with vested interests, but
also by some distinguished researchers. However by the 1990’s the role of diet, and of
saturated fatty acids in particular, in the aetiology of coronary heart disease was
universally acknowledged and LDL cholesterol recognised as the principal
atherogenic component of LDL. The 2000 Canadian review updated a 1993 US FDA
review justifying a health claim relating to dietary fatty acids and cholesterol. The
specific issues relating to saturated and trans unsaturated fatty acids and LDL
cholesterol and CHD in the Canadian review are considered in more detail in part 1 of
this report; some comments in relation to the more general aspects of the 2000 review
are briefly considered here.
The process followed in the preparation of the Canadian report appears to be
entirely appropriate and most of the material relevant to the issues under consideration
has been included and correctly interpreted. However some of the related aspects are
less appropriately presented. For example a rather outdated approach is taken to the
description of cardiovascular risk factors. There is no acknowledgement that they
should be considered in aggregate in an attempt to determine absolute risk. The
document still considers them in isolation with emphasis, for example, on cut-offs for
lipid levels (table 1). The description of lipid and lipoprotein disorders was unclear,
with some of the suggested cut-offs inappropriate. However of greater importance and
relevance to the issues under consideration is the fact that the review does not clearly
distinguish the effects of different types of dietary carbohydrate when recommending
a replacement energy for dietary fat. There are repeated statements that replacing
carbohydrate for fat will result in an increase in triglyceride and reduction in the
protective high density lipoprotein (HDL). This in indeed the case when substantial
quantities of fat are replaced by monosaccharides, disaccharides, and digestible starch.
However when fat is replaced by vegetables, fruit and wholegrain cereals, rich in nonstarch polysaccharides, this deleterious effect on triglyceride and HDL appears to be
negligible. This issue may not have direct bearing on health claims suggesting the
cardiovascular benefits of reducing saturated and trans fatty acids, but is of
considerable relevance when considering the implementation of such advice.
Part 1: Critical appraisal of the Canadian Review
Part 1a): Selection and assessment of evidence
(i) Saturated fatty acids and LDL cholesterol
• The Canadian review has considered all the important studies relevant to this
question. There are a few studies relevant to the effects of the individual
saturated fatty acids which have not been included (see bullet point 6), but
these would not have influenced the overall conclusions.
• The original US review has been correctly interpreted.
• The association between saturated fatty acids and LDL cholesterol is one of
the most clearly established relationships between a nutrient intake and a
universally accepted disease indicator.
• The evidence for a causal relationship is ‘convincing’, with several metaanalyses of RCT’s carried out under metabolic and free-living conditions
yielding broadly comparable results. Regression equations suggest an
approximately 0.05 mmol/l increase in serum cholesterol for each 1% increase
in saturated fatty acids. Animal experiments confirm experiments in humans
and provide clear evidence of a mechanism. The effect of saturated fatty acids
(SFA) on total cholesterol is almost entirely explained by changes in LDL
cholesterol.
• The interpretation of the evidence cited was generally appropriate.
月桂酸比肉荳蔻酸和棕櫚酸更能顯著提高高密度脂蛋白膽固醇，可改善低密度脂蛋白/高密度脂蛋白比值。
• Three major limitations of the available evidence were in our opinion not
adequately discussed. The first relates to the effect of the individual saturated
fatty acids. The review acknowledges the different effects of the major dietary
SFA’s, but does not adequately address the differential effect on LDL
cholesterol of lauric, myristic and palmitic acids, nor the fact that lauric acid
has a more marked effect on raising HDL cholesterol than myristic and
palmitic acids, and is thus associated with a more favourable ratio of
LDL/HDL. The negligible effect of stearic acid on LDL cholesterol is
considered. This does not effect the overall conclusion since stearic acid intake
is, at least in most dietary patterns, highly correlated with intake of other
SFA’s, and in cohort studies has been shown to increase cardiovascular risk,
possibly via a thrombogenic mechanism. Second, while arguably not
important with regard to health claims aimed at improving population health,
it would nevertheless have been relevant to draw attention to the wide
individual variation which occurs in response to change in nature of dietary
fat, despite the consistent and predictable changes when considering groups of
individuals. There have been many attempts made to identify genetic
polymorphisms which might explain this heterogeneity in response, but none
appear to be important determinants, probably because polygenic factors are
responsible, and the studies have been inadequately powered to have
simultaneously examined the effect of multiple genes. Finally, and perhaps of
greatest relevance is the absence of a detailed discussion regarding the effects
of nutrients with which SFA’s are replaced. Reduction of SFA’s is likely to
have a greater LDL-lowering effect when replaced with n-6 polyunsaturated
fats than with carbohydrate because the former have, in their own right, the
potential to lower LDL, whereas the reduction seen with carbohyd replacement may simply be a consequence of a reduction in the SFA’s. An
exception to this might be if the carbohydrate is rich in soluble forms of nonstarch polysaccharide which also has inherent LDL-lowering potential.
However none of these limitations appreciably dilutes the claim that reduction
in saturated fatty acids will reduce total and LDL cholesterol in the population
at large.
• As indicated above regression equations can calculate the expected benefit in
terms of total and LDL cholesterol reduction which might be expected to be
achieved for each 1% reduction in SFA’s. This may in turn be translated into
expected clinical benefit given that each 1% reduction in cholesterol is likely
to translate into at least a 2% reduction in cardiovascular events.
(ii) Saturated fatty acids and coronary heart disease
• The first papers describing the major study relating SFA’s to CHD (the Seven
Countries study) antedated the Canadian review. However it is surprising that
the 25 year follow-up of this study (Kromhout et al, 1995) was not included
amongst the cohort studies listed on pages 31-32. This study along with the
study by Hu et al. (1999) provide the best prospective evidence of a direct link
between SFA’s and CHD. There are clinical trials which show the potential of
reducing CHD risk by reducing saturated fatty acids. Admittedly several of
these have considered dietary change other than a reduction in SFA’s, but the
Oslo Diet Heart Study (Leren, 1989) involved principally a reduction in total
and saturated fatty acids. Given that there is little evidence to suggest benefit
of reducing total fat it seems likely that the observed benefit derived primarily
from a reduction in SFA’s. The conclusions would not have been altered by
the addition of these studies.
• The US review has been appropriately interpreted.
• While the evidence linking SFA’s with CHD morbidity and mortality is not as
consistent and impressive as the link between SFA’s and LDL cholesterol, it
was considered by the Canadians to be ‘convincing’. The association is
evident in cohort studies, and RCT’s have shown clinical benefit as a result of
reducing SFA intake. Not all cohort studies have confirmed the association but
plausible explanations (especially with difficulties of dietary assessment) have
been offered for the inconsistency. The association is biologically plausible
with a clear dose-response between degree of cholesterol lowering (itself
related to extent of dietary change) and clinical benefit. The issue surrounding
whether this is indeed a convincingly causal association as distinct from a
‘probable’ one hinges around the extent to which the heterogeneity may be
regarded as sufficiently explained. We were seeking confirmation of this in the
new data.
• The interpretation of the cited evidence was generally appropriate. However
the findings from the Oxford Vegetarian Study were not quite accurately
interpreted. The appreciable reduction in CHD risk amongst the vegetarians
was attributable to the lower intake of SFA rather than abstinence of meat
(Mann et al., 1997). The overall conclusions are not influenced.
• The limitations of the available evidence are adequately considered.
• The review relies principally on two studies to quantify the benefit likely to
accrue from reduction in SFA’s.
如果用碳水化合物取代 5% 飽和脂肪酸攝取量, 冠心病風險降低 17%；如果用不飽和脂肪酸替代 5% 飽和脂肪酸，則可降低 42%。
Hu et al. (1997) estimate that ~ 5% reductionin SFA’s would lead to a 17% reduction in CHD risk if SFA’s were replaced
by carbohydrate and 42% if replaced by unsaturated oils. Kris-Etherton and
Yu (1997) estimate that an AHA Step 2 diet (saturated fat replaced by
carbohydrate) would lower CHD risk by 12% whereas isoenergetic
replacement of saturated fat with monounsaturated fatty acids would reduce
risk by 42%. Unfortunately these estimates do not take into account nature of
carbohydrate but this does not detract from the benefits which might be
expected from a reduction in intake of SFA’s.
(iii) Trans fatty acids and LDL cholesterol
• An appreciable number of studies demonstrating the effect of trans fatty acids
on LDL cholesterol had been published at the time of the 2000 Canadian
review. All major studies appear to have been included.
• Most of the dietary intervention studies examining the effects of trans fatty
acids on LDL were of high quality and published in major journals by leading
research groups. Total and LDL cholesterol appear to show a linear increase
with increasing intakes of trans fatty acids. Lipoprotein (a) is also higher on
diets high in trans fatty acids compared with other sources of fat. The
relationship between trans fatty acids and LDL cholesterol can unquestionably
be described as ‘convincing’.
• All the evidence appears to be cited appropriately.
• There are three limitations in the available evidence.
First is the limitation which applies to all dietary intervention studies, namely
that it is impossible to guarantee 100% dietary compliance in studies including
free living individuals. However the remarkable consistency of the results of
the various studies and the clear dose response shown in several suggest that
this cannot explain the association between total trans fatty acids and LDL
cholesterol.
Second, some of the studies do not fully explain how trans fatty acids were
measured, nor do all distinguish clearly between vegetable and animal sources
of trans fatty acids. This is a limitation which was not adequately discussed in
the Canadian review. However, once again, consistency of the findings
suggests that the association is a genuine one.
Third, it is not possible to assess whether the association occurs over the
whole range of intakes of trans fatty acids.
• Data from the experimental studies have not been extrapolated to estimate
potential influence on cardiovascular risk reduction by reducing intake of
trans fatty acids. However such estimates have been made on the basis of
epidemiological data. See below. (iv) Trans fatty acids and coronary heart disease
• The major epidemiological studies which have examined the association
between trans fatty acids and coronary heart disease were considered in the
Canadian review.
• When examining the various epidemiological studies, there appears to be a
degree of heterogeneity with regard to the relationship between trans fatty
acids and coronary heart disease. Since the first publication by Willett et al.
(1993) based on the Nurses’ Health Study which showed a clear association,
there have been several case control studies, one major cross-sectional study, a
Finnish cohort study (Pietinen et al., 1997) and a further longer term report
based on the Nurses’ Health Study from the U.S. (Hu et al., 1997). The crosssectional study based on the Seven Countries Study described earlier and the
two reports from the Finnish and US cohorts both confirmed the initial
observation. Two of the four case control studies did not confirm the
association. However, the Scottish study (Bolton-Smith et al., 1996) included
individuals who had had cardiovascular disease for various intervals before
being investigated and diet may well have changed after an acute event.
Indeed a study of this nature should not be regarded as admissible evidence.
The multinational EURAMIC study also found no differences between cases
and controls (Aro et al., 1995). There is no clear explanation, but it is
generally agreed that even ‘good’ case control studies may be inconclusive
when examining diet-disease relationships. Of the two case control studies
showing a positive association, one should probably also be discounted since it
was underpowered and there was no attempt to examine for confounding
variables (Siguel and Lerman, 1993). Given the impressive association
observed in cross-sectional and cohort studies which show a dose-response
relationship which is biologically plausible, it seems reasonable to support the
Canadian conclusion that there is convincing evidence that the association
between trans fatty acids and coronary heart disease is causal. The
heterogeneity is present only amongst the case-control studies which are often
discounted in the hierarchy of evidence examining diet-disease associations,
and in only a single case control study (the EURAMIC study) is the lack of an
association not readily explicable. However this is an association we were also
anxious to pursue in the more recent data.
• The cited evidence is appropriately interpreted.
• The review does indeed consider the limitations in the available evidence. The
increased risk of trans fatty acids as demonstrated in both cross-sectional and cohort studies stems principally from TFA derived from vegetable fat (trans 18:1, elaidic acid).
據估計，將所需能量的 2% 以順式不飽和脂肪酸取代反式脂肪酸，可使冠心病風險降低 50% 。此估算主要基於護理師健康研究的數據。
• It has been estimated that replacement of 2% of energy from trans fatty acids with evidence from cis unsaturated fatty acids would reduce coronary heart disease risk by over 50%. This estimate is based principally upon the data
from the Nurses’ Health Study
Part 1b): Reanalysis of pivotal studies
Hu et al., 1997:
Hu FB, Stampfer MJ, Manson JE, Rimm E, Colditz GA, Rosner BA, Hennekens CH,
Willett WC. Dietary fat intake and the risk of coronary heart disease in women. N
Engl J Med, 1997, 337(21):1491-9.
Reason for selection
This study was selected for review since it is arguably the most definitive study
demonstrating the effects of dietary fat on risk of coronary heart disease.
Summary
The Nurses’ Health Study involved over 80,000 women aged 34 to 59 years of age
who in 1980 had no known CHD, stroke, cancer, hyperlipidaemia or diabetes.
Particularly noteworthy was the fact that dietary data (obtained by means of food
frequency questionnaire) was collected not only at baseline, but was also updated
during 14 years of follow up during which time 939 cases of nonfatal or fatal
myocardial infarction were documented. Multivariate analysis included age, smoking
status, total energy intake, dietary cholesterol, percentages of energy from protein and
types of dietary fat, and other risk factors. Total fat intake was not related to risk of
coronary heart disease. Each increase of 5% of energy intake from saturated fat, as
compared with equivalent energy intake from carbohydrates, was associated with a
17% increase in the risk of CHD (relative risk 1.17). As compared with equivalent
energy from carbohydrates, the relative risk for a 2% increase in energy intake from
trans unsaturated fat was 1.93, that for a 5% increase in energy from polyunsaturated
fat was 0.62. From the data it is possible to estimate that replacement of 5% of energy
from saturated fat with energy from unsaturated fats would reduce risk of CHD by
42% and that the replacement of 2% of energy from trans fat with energy from
unhydrogenated unsaturated fats would reduce risk by 53%.
This study does indeed have some limitations. The findings are limited to
women. Nutrient intakes were based on food frequency questionnaires and inevitably
the trans fatty acid data in the food composition tables must have been incomplete.
Nevertheless the repetition of the food frequency questionnaire on 4 occasions
provided for superior data compared with most other prospective studies in which diet
is typically assessed on only a single occasion. Sample size is impressive and over
90% were followed to the conclusion of the study.
The associations between dietary intakes of fatty acids and clinical events are
compatible with the effects on lipoproteins. However the somewhat greater effects
than might have been predicted on the basis of lipoprotein mediation suggest that
enhanced atherogenic risk might also be mediated via other mechanisms, e.g.
influencing platelet aggregability, insulin sensitivity and cardiac conductivity. The
findings are also important because they place in perspective the lesser importance of
dietary cholesterol and, at least in terms of CHD, the relative unimportance of total fat
intake other than as a possible indicator of intake of saturated fatty acids.
Clarke et al., 1997:
Clarke R, Frost C, Collins R, Appleby P, Peto R. Dietary lipids and blood cholesterol:
quantitative meta-analysis of metabolic ward studies. BMJ, 1997, 314(7074):112-7.
Reason for selection
This paper was chosen since it was the first meta-analysis based on a large number of
dietary experiments using appropriate statistical methods and carried out by some of
the world’s leading epidemiologists and statisticians with appropriate nutritional
input.
Summary
The analyses were based on 395 dietary experiments carried out under strictly
controlled conditions generally in metabolic wards in order to ensure compliance.
Details are provided regarding the search strategies as well as inclusion and exclusion
criteria. Isocaloric replacement of saturated fats by complex carbohydrates for 10% of
dietary calories resulted in a reduction in total and LDL cholesterol by 0.52 and 0.36
mmol/l respectively. Isocaloric replacement of complex carbohydrates by
polyunsaturated fats for 5% of dietary calories produces further decreases of 0.13 and
0.11 mmol/l. Replacement of carbohydrate by monounsaturated fats produced no
significant effect on total and LDL cholesterol. Avoiding 200mg/day dietary
cholesterol has the potential to further decrease total and LDL cholesterol by 0.13 and
0.10 mmol/l. The figure below (Table 2 from the original paper) shows the effect of
realistic dietary change in terms of modifying dietary fatty acids and cholesterol
without altering total energy in a fairly typical western diet. Much of the 0.76 mmol/l
reduction in total cholesterol is achieved by a reduction in LDL cholesterol (-0.62
mmol/l, i.e. a reduction of 10-15%). The effects of trans fatty acids could only be
examined in 40 of the experiments, but in this meta-analysis the effects of trans
monounsaturated fats (mainly trans C18:1, elaidic acid) on total and LDL cholesterol
did not differ appreciably from that of saturated fatty acids. As in the earlier studies
stearic acid appeared not to influence total and LDL cholesterol.
There are few limitations of this study. It could not distinguish between natural
trans unsaturates and trans unsaturated fatty acids from hydrogenated vegetable oils,
but then relatively few studies have attempted to disentangle separate effects. The
authors chose not to examine the effects of these dietary manipulations on lipoprotein
fractions other than LDL cholesterol. However this is probably the most accurate
attempt to quantify the effects of dietary modification on total and LDL cholesterol. It
is important to emphasise that the meta-analysis relates to isocaloric exchange. Since
weight loss can reduce total and LDL cholesterol appreciably greater reductions
would be apparent if the changes described here were made in conjunction with
energy restriction.
Tang et al., 1998:
Tang JL, Armitage JM, Lancaster T, Silagy CA, Fowler GH, Neil HA. Systematic
review of dietary intervention trials to lower blood total cholesterol in free-living
subjects.
BMJ, 1998, 316(7139):1213-20.
Reason for selection
This was the first systematic review of dietary intervention studies which aimed to
lower cholesterol in free living subjects. Although the review considered blood total
cholesterol, there is strong evidence to suggest that much of the changes in total
cholesterol derive from change in LDL cholesterol.
Summary
19 randomised controlled trials which lasted for four weeks or more were included.
The percentage reduction in blood total cholesterol attributable to various dietary
manipulations is shown in the figure below for those studies which continued for at
least 6 months, with an average reduction of 5.3%. When considering only the first 3
months, reduction was appreciably higher at 8.3%, but the reduction at 12 months
(5.5%) was remarkably similar to that achieved after 6 months. On the basis of
repeated food intake assessment, the targets for dietary change were rarely achieved,
but reported changes produced cholesterol changes that were consistent with those
predicted from the Keys formula.
Thus on average free-living individuals achieve roughly half the reduction in
cholesterol which might be achieved by comparable dietary manipulation in a
metabolic ward. The study provides evidence to suggest that the difference is
principally a consequence of reduced compliance. However it is important to
appreciate that even given a somewhat disappointing level of compliance reductions
in total cholesterol, which may be extrapolated to reductions in LDL cholesterol, are
such that they would be expected to produce considerable clinical benefit if the rule of
“1% reduction in cholesterol produces a 2-3% reduction in incidence of CHD” is
applied. It is noteworthy that we have been able to show changes in total and LDL
cholesterol amongst free-living individuals which are more akin to that achieved in
metabolic wards. However this required considerable ongoing advice and support for
participants in the studies (Aitken, W, in press).
Lichtenstein et al., 1999:
Lichtenstein AH, Ausman LM, Jalbert SM, Schaefer EJ. Effects of different forms of
dietary hydrogenated fats on serum lipoprotein cholesterol levels. N Engl J Med,
1999, 340(25):1933-40.
Erratum in: N Engl J Med, 1999, 341(11):856.
Reason for selection
This paper was selected since it is one of the clearest demonstrations of the effect of
trans fatty acids on lipoprotein levels, but at the same time raises the critical issue
regarding amount of TFA required to exert adverse effects on lipoproteins.
Summary
This remarkable study involved 18 women and 18 men who consumed each of 6
experimental diets in random order for 35 day periods. The foods were identical in
each diet. 30% of calories was derived from fat, two thirds of total fat provided by the
test fats: soybean oil (<0.5g TFA/100g fat), semi-liquid margarine (0.6g /100g), soft
margarine (7.4g/100g), shortening (13.6g/100g) or stick margarine (20.1g/100g). The
effects were compared with those of a diet enriched with butter with a high intake of
saturated fat (61.7g saturated fat/100g, 2.6g TFA/100g). Saturated fatty acid intake of
the experimental diets ranged from 13g/100g in soybean oil to 16-17g/100g for the
other experimental diets. Soybean oil and semi-liquid margarine resulted in 12% and
11% reductions in LDL cholesterol compared with that seen in the high butter diet.
Indeed compared with butter all experimental diets were associated with lower LDL
cholesterol. Of particular interest are levels on the remaining three experimental diets
with comparable amounts of saturated fatty acids and increasing quantities of TFA’s.
Compared with soybean oil LDL cholesterol was 3% higher on the soft margarine,
5% higher on shortening and 7% higher on the stick margarine. Ratios of total and
LDL cholesterol to HDL cholesterol were lowest in the soybean oil and semi-liquid
margarine and highest in the butter and stick margarine.
There are two major limitations to this study. First, relatively high levels of a
single fat were used particularly in the case of a stick margarine, achieving a total
TFA intake which is probably greater than that typically consumed in Australia and
NZ. Nevertheless the clear dose response indicating an increase in LDL cholesterol
with increasing intakes of TFA provides convincing evidence of the causal
association. However the clinical relevance of the findings in the context of relatively
low consumption is an issue which remains to be resolved.
Second, each experimental dietary fat included a mixture of fatty acid isomers
containing at least one trans double bond. While this does not permit evaluation of the
effects of individual isomers, it does allow the assessment of commercially available
hydrogenated vegetable fats. Hydrogenated fats derived from vegetable oil other than
soybean oil will have different fatty acid profiles. However other data suggest that
hydrogenated fats from a wider range of oils have similar effects.
These results are also important since they indicate that while shortening and
margarines, which are relatively high in trans fatty acids but low in saturated fatty acids, are preferable to butter in terms of LDL cholesterol, butter itself is associated
with relatively favourable levels of Lp(a). The most favourable lipoprotein profiles
overall are associated with diets low in saturated and trans unsaturated fatty acids and
with relatively high proportions of cis-unsaturated fatty acids
Part 1c): Consideration of the validity of the review’s conclusions
•
In our opinion the Canadian review provides a comprehensive and reasonably
accurate summary of the evidence available at the time it was conducted. The
limitations described above do not appreciably influence the overall
conclusions. However we considered that it was necessary to review in the
more recent data the issue as to whether the associations between saturated
fatty acids and trans fatty acids and coronary heart disease were appropriately
described as “convincing” rather than “probable”.
• The review does describe the circumstances under which the studies were
carried out.
• There is no doubt that the influence of total fat and cholesterol intakes on
serum LDL-cholesterol and coronary heart disease are less important than the
effects of saturated and trans fatty acids. In terms of both LDL cholesterol and
coronary heart disease, there is no evidence that reducing total fat intake
confers benefit beyond that resulting from the reduction in saturated fatty
acids. That dietary cholesterol has a less marked effect than saturated fatty
acids or total (and LDL) cholesterol has been repeatedly demonstrated since
the 1960’s. More recent studies have shown that even with intakes of dietary
cholesterol which are appreciably greater than those typically consumed, the
effect on LDL cholesterol is relatively small when intakes of saturated fatty
acids is low. There is some evidence to suggest that the effect of dietary
cholesterol is greater in individuals with apoE4E4 or apoE3E4 genotypes than
those with other apoE genotypes and amongst those with relatively high levels
of total and LDL cholesterol. However these observations do not influence the
overall conclusion. There is little evidence in epidemiological studies of a
meaningful association between intake of dietary cholesterol within the usual
range of intakes and clinical coronary heart disease.
• The report does not consider potential undesirable effects associated with the
expected dietary change. However there are no specific nutritional
requirements for vegetable sources of trans unsaturated fat principally
produced in the manufacturing process and only benefits, not undesirable
effects, would be expected to accrue from their elimination from the diet were
this feasible. Many saturated fats on the other hand contribute to total energy
intake and are also sources of other important nutrients, notably fat soluble
vitamins. However other than in the first few years of life, when substantial fat
restriction may lead to inadequate energy intakes for growth and development,
even substantial reductions of saturated fatty acids will not lead to energy
deficits (if required cis unsaturated fatty acids may compensate) or
deficiencies in fat soluble vitamins.
• We believe that there is convincing evidence for a causal link between
saturated and trans unsaturated fatty acids and LDL cholesterol. Benefit in
terms of reduced risk of coronary heart disease is likely to accrue from
reduced intakes. However it is still somewhat open to question as to whether
the evidence directly linking saturated fatty acid intake with coronary heart
disease can be described as “convincing” using current criteria or whether it
should more appropriately be described as “probable”.
• The Canadian review is comprehensive and builds appropriately on earlier
literature. Thus it is a reasonable starting point for the substantiation of a
relationship between saturated and trans unsaturated fatty acid intake and LDL cholesterol and coronary heart disease. Although the effects of the different
saturated fatty acids are not equal we acknowledge that in terms of
recommendations or claims it is inappropriate and indeed probably impossible
to distinguish between the different saturated fatty acids.
• The information that will be especially sought from the more recent data relate
to trans fatty acids and whether the adverse effects are indeed restricted to
vegetable sources and also whether the associations with LDL cholesterol and
CHD occur across the full range of intakes.
• Additionally the evidence directly linking saturated fatty acids and trans fatty
acids and coronary heart disease will be evaluated in an attempt to establish
whether this should be described as convincing or probable.
Part 2: Review of evidence released since the publication of the Health Canada
report
i. Literature search for current evidence
Search strategy:
A literature search was carried out on the OVID Medline 1999 to June Week 4 2005
database. (OVID search history included as Appendix 2.1).
Keyword and subject heading searches were made for the terms: trans fatty acid/trans
fatty acids; dietary cholesterol; fatty acids; dietary fats; saturated fat; hydrogenated
fat; meat; meat products; eggs; animal fat. Any articles identified dealing with poultry
or seafood were later ignored. Results of these searches were pooled to give 27,327
articles. This pooling grouping was then crossed with a pooled search from keyword
and subject heading searches for the terms: cardiovascular disease; coronary heart
disease; CVD; CHD; LDL; LDL cholesterol lipoproteins; LDL lipoproteins; low
density lipoprotein.
Results were limited to articles in the English language, giving a total of 1,937
articles. For ease of analysis, articles were grouped according to the categories they
were indexed under on Medline, such that this set contained 13 articles indexed on
Medline as Cochrane reviews; 304 indexed as review articles; 6 meta-analyses; 39
case-control studies; 65 cohort studies; 30 randomised controlled trials, and 1,480 not
indexed in any of these categories.
The authors individually reviewed the titles and abstracts of all Cochrane reviews,
meta-analyses, case-control, cohort and randomised controlled trial papers, and the
titles of review articles individually, selecting papers deemed of interest. Lists of
papers selected by each author were compared, and any discrepancies or differences
in paper selection resolved after consideration of the abstract. The titles of the
remainder of the 1,480 identified articles were scanned by C.B. and full text articles
subsequently obtained when deemed relevant or of importance. Any additional
articles not identified in the literature search, but referenced in papers or review
articles which were deemed of sufficient interest or impact, were also considered in
the writing of the review.
A hand search of a selection of relevant journals (see below) over the time period Jan
2000 – June 2005 was also conducted by C.B. for additional articles:
American Journal of Epidemiology; Annual Review of Nutrition; International
Journal of Epidemiology; Public Health Nutrition; European Journal of Clinical
Nutrition; American Journal of Clinical Nutrition.
ii. Analysis of current evidence, 1999 – 2005:
Saturated fatty acids (SFA’s):
Summary:
The results of studies published since publication of the Health Canada review are
generally supportive of the conclusion arrived at in that report – that increasing intake
of saturated fat is convincingly associated with raised LDL cholesterol. The main
development since the Health Canada report appears to be an increasing sophistication
of analysis based on larger pools of data with additional examination of the effects of
individual saturated fatty acids.
In addition to the further dietary intervention trials in adults, several recent
randomised controlled trials have examined saturated fat intake in childhood
following randomisation to low-fat or control diets. These studies suggest that
reduction of saturated fat intake even very early in childhood has the potential to
reduce LDL cholesterol levels and improve cardiovascular risk factors, with none
reporting adverse effects on child growth and development as a result of fat
restriction. Long term compliance may be difficult to sustain. It appears that saturated
fat reduction in young children may only produce these beneficial effects on lipid
levels in boys and not girls, with no clear explanation regarding gender differences.
Two important meta-analyses have aggregated data for the dietary intervention trials
(Mensink et al., 2003; Müller et al., 2001). When considering the effect on LDL or the
ratio of LDL:HDL there is no doubt that the most marked adverse effects are
associated with myristic and palmitic acids. While lauric acid is certainly associated
with an elevation of LDL cholesterol there is less consistency with regard to the
relative effects of this fatty acid compared with myristic and palmitic acids. However
while all three of these fatty acids are associated with increase in HDL cholesterol
when compared with carbohydrate, the effect of lauric acid is consistently the
greatest. This may to some extent mitigate the effects of LDL elevation. Although
stearic acid has no effect on LDL cholesterol, there is no doubt that in aggregate
saturated fatty acids are “convincingly” associated with LDL elevation. In terms of
substantiating a health claim, the relative effects of the individual fatty acids are not of
public health importance.
No new prospective studies of saturated fat intake and health outcomes were
identified. However further analyses of the Nurses’ Health Study, Seven Countries
Study, the British Health and Lifestyle Survey, and a cohort of Danish men and
women have been undertaken. They are generally confirmatory of an association
between saturated fat intake and coronary heart disease. However discrepancies do
exist and these are partially explicable. An analysis of the Nurses’ Health Study
cohort from 1980 to 1998 showed no significant association between saturated fat
intake and coronary heart disease. The authors ascribe this weakening of the
previously observed associations in this cohort to reductions in intakes of saturated fat
and incidence of coronary heart disease which occurred during the 18 years
subsequent to the establishment of the cohort and recording of baseline data. Data
from the British Health and Lifestyle Survey showed a significant association between
intake of saturated fat and coronary heart disease amongst women, with a stronger association observed amongst older women. A 16-year follow up of Danish men and
women reported an association of borderline significance between saturated fat intake
and coronary heart disease in women, which was observed to be stronger amongst
younger women. Both the British Health and Lifestyle Survey and the Danish cohort
study showed no association of saturated fat intake and coronary heart disease in men.
The prolonged period between the baseline dietary data and morbidity and mortality
from coronary heart disease to which intake of saturated fatty acid was related is a
likely explanation for the failure to confirm the associations convincingly
demonstrated during the earlier follow up period. However inconsistencies with
regard to gender and age are not readily explained.
Thus while there is a considerable body of evidence to suggest that saturated fatty
acids are indeed associated with coronary heart disease and there is indeed a
biologically plausible explanation, the heterogeneity observed in the recent analyses
which cannot be fully explained lead to the conclusion that the association is
“probably” causal rather than “convincingly” causal.
Saturated fatty acids and LDL-cholesterol:
Although the purpose of this report is to examine the link between dietary saturated
and trans fatty acids and LDL cholesterol, the effect on HDL cholesterol has been
commented on where available since this enables an examination of the effect of
dietary manipulation on the overall lipoprotein pattern.
Cross sectional epidemiological studies:
Very few cross-sectional studies of saturated fatty acid intake and LDL cholesterol
appear to have been undertaken in the period under review. A comparison of
omnivores and vegetarians was undertaken in Hong Kong by Lee et al. (2000).
Dietary assessment was by food frequency questionnaire recorded and analysed
without knowledge of LDL levels. Not surprisingly, total fat intake was significantly
higher in omnivores, and this was reflected in significantly higher SFA, PUFA and
MUFA intakes, as well as dietary cholesterol. Other differences were a higher protein
and lower carbohydrate intake, and lower P/S ratio, in omnivores than vegetarians.
Total serum cholesterol and LDL cholesterol were significantly higher in omnivores,
though so too was HDL cholesterol. VLDL and LDL:HDL ratio were similar in the
two groups. This study was of reasonable size and appropriate design, taking into
account the limitations of case control studies. Thus differences in lipids and
lipoproteins between omnivores and vegetarians are confirmed but no definitive
conclusions can be drawn regarding the relationship between saturated fatty acids and
LDL given the many dietary differences between the two groups.
Another larger cross-sectional study was that of the EURODIAB IDDM
Complications Study (Toeller et al., 1999), principally designed with diabetes
complications in mind, but which also collected dietary information from diet history
records and blood lipid measurements in some individuals. Regression equations for
total cholesterol were based on 2762 individuals, and for LDL cholesterol on 1816
individuals, all of whom had type 1 diabetes. Energy-adjusted intake of total fat,
saturated fat, and dietary cholesterol were significantly positively associated with total
and LDL cholesterol. However, in a model fully adjusted for other dietary and clinical
measures the significance of these associations was lost. The authors ascribed this to
the effect of dietary fibre intake, which tended to decrease as saturated fat intake
increased. Thus in the final regression analysis intakes of saturated fat or dietary
cholesterol were not significantly associated with serum total or LDL cholesterol
levels. However, it was noted that for those individuals who consumed less than 8%
of energy as saturated fat the mean LDL cholesterol level was appreciably lower than
individuals with higher intakes. Additionally, intakes of dietary cholesterol in excess
of 500mg/day were associated with increased LDL cholesterol levels.
Childhood studies:
A number of randomised controlled trials and prospective studies examining the
association of total and saturated fat intake with cholesterol levels in childhood have
emerged in the last decade. However not all of these specifically measured fat, and
more importantly saturated fat, intakes and hence only two randomised controlled
trials are considered below: the STRIP and DISC studies, and the prospective
ALSPAC study. No randomised controlled trials or prospective studies specifically
examining the effect of reduction in saturated fat intake on LDL cholesterol in adults
were found in the time period under review.
Childhood studies - randomised controlled trials:
The Special Turku coronary Risk factor Intervention Project (STRIP) which began
1990 has given rise to a number of publications (Niinikoski et al., 1996; Rask-Nissila
et al., 2000; Viikari et al., 2004; Salo et al., 2000; Rasanen et al., 2004; Kaitosaari et
al., 2003; Rask-Nissila et al., 2002; Lapinleimu et al., 1995). Families of 7-month old
infants were randomised to either a control group (n = 522) or intervention (n = 540).
Intervention consisted of advice aimed at reducing the child’s exposure to
atherosclerosis or coronary heart disease risk factors, predominantly focusing on
dietary advice. The ideal diet at this age for a child was described to the parents as one
comprising 30-35% fat, with equal proportions of saturated, monounsaturated and
polyunsaturated acids making up this amount. No differences were observed between
control and intervention groups at baseline in terms of total cholesterol, HDL
cholesterol, or non-HDL cholesterol. At age 13 months a comparison of dietary
intakes was made, although given the obvious difficulties involved with children
being breast-fed this was restricted to children who were formula fed. Formula fed
intervention children (n=60 per group) were receiving significantly less dietary
saturated fat than control children (n=60). Intervention children (n=450) and control
children (n=436) at 13 months had significantly different changes from baseline in
total serum cholesterol, HDL cholesterol, and non-HDL cholesterol, with intervention
children showing lower levels for each of these measures. These comparisons were
based on non-fasting samples.
Follow-up results were published for assessments at 3 years of age (Niinikoski et al.,
1996), 5 years of age (Rask-Nissila et al., 2000) and 7 years of age (Kaitosaari et al.,
2003). Non-fasting blood samples were drawn yearly up to the age of 5 yrs, at which
age samples were drawn after an overnight fast for the first time, allowing calculation
of LDL cholesterol. The results at 5 and 7 years of age are considered below.
At 5 years of age, analysis of dietary intakes at 13 months, 24 months, 36 months, 48
months and 60 months revealed that intervention boys were consistently consuming
less total fat, saturated fat and cholesterol than control boys. Polyunsaturated fat
intakes were consistently significantly higher in intervention boys, with
monounsaturated fat and total energy both significantly lower at one of the above time
points each. Results were similar for girls, with total fat, saturated fat and cholesterol
intakes significantly lower for intervention children than controls and polyunsaturated
fat consistently higher. Monounsaturated fat intake was significantly lower amongst
intervention girls on two occasions, and total energy intakes were significantly lower
in intervention girls vs. control girls at 13 months, 48 months and 60 months.
Blood test results for boys showed that across the trial period, serum cholesterol
values of intervention boys were 6% - 10% lower than control boys, and similarly
non-HDL cholesterol 6% - 11% lower. HDL:total cholesterol ratios were similar
between groups at all time points. Analysis of LDL cholesterol at 5 years of age
showed mean levels in intervention boys were 9% lower than control boys (2.62 ±
0.55 mmol/l vs. 2.87 ± 0.61 mmol/l, p = 0.0002).
For the girls, mean serum cholesterol levels across the trial period were slightly lower
in intervention girls and bordered on statistical significance (p = 0.052). Serum nonHDL cholesterol did not differ between groups such that HDL:total cholesterol levels
were similar across groups as well. LDL cholesterol at 5 years showed no difference
between intervention and control girls (2.94 ± 0.63 mmol/l vs. 2.99 ± 0.66 mmol/l
respectively, p = 0.53).
The dietary and blood test results reported above were not based on the entire cohort;
dietary analysis from 13 months of age was based on continually smaller numbers for
each subsequent follow-up, and at 60 months was based on roughly 135 participants
per gender per treatment arm (range 126 for intervention girls – 141 for control boys).
Numbers for longitudinal blood test results ranged from 158 for lipid measures in
intervention girls to 180/181 for lipid measures in intervention boys.
Extending these findings to 7 years of follow-up (Kaitosaari et al., 2003) total fat
intake was the same amongst intervention vs. control boys, although saturated fat
intake was lower in intervention boys (11.5% total energy vs. 13.6%, p < 0.001). Both
total fat and saturated fat intakes were lower in intervention girls compared to control
girls. Serum lipid measures revealed no significant differences between control and
intervention girls, and in fact lipid measures did not even tend towards any
differences, with the lowest p-value being 0.48 for comparison of triglycerides. On the
other hand, intervention boys at 7 years had significantly lower serum cholesterol and
LDL cholesterol, and a significantly higher HDL:total cholesterol ratio than control
boys.
It appears from the results of the STRIP project that reduction of saturated fat intake
across a number of years from early infancy through childhood has the ability to
consistently reduce non-HDL and LDL cholesterol in boys, but not girls. No adverse
effects on behavioural or cognitive development of children was observed (RaskNissila et al., 2002).
Considering older age groups, the Dietary Intervention Study in Children (DISC)
(Obarzanek et al., 2001a) enrolled 663 children aged 8-10 yrs in 1987 to randomised
dietary intervention or control groups. The study was conducted in the U.S. with six
participating centres. Children were followed either until age 18 yrs (15.3%) or until a
screening visit prior to age 18 (84.7%). Mean length of follow-up was 7.4 years.
Dietary aims in the intervention group were for 28% of total energy from fat, less than
8% from saturated fat, and less than 75mg of cholesterol per 1000 kcal energy, thus
similar to the National Cholesterol Education Programme Step 2 diet. Children were
required to be mildly hypercholesterolaemic, with the average LDL after two
screening visits above or equal to the 80th percentile but below the 98th percentile for
age and sex. The intervention programme was delivered by nutritionists or behaviourists in either group or individual settings, and additionally by phone contact
at varying intervals over the study period –initially more intense with follow-up
tapering off as the study progressed. Lipid measurements were made at baseline, 1
year, 3 years, 5 years and final visit, with two different fasting samples collected 1
month apart at baseline and 3 years to reduce variance.
Mean age (SD) at baseline was 9.5 (0.72) yrs and 17.0 (0.91) yrs at final visit, and did
not differ between groups. Intakes of total and saturated fat were significantly lower at
years 1, 3, 5 and final visit in the intervention group, with dietary cholesterol also
significantly lower except at the visit.
An interesting finding in this trial was that serum LDL cholesterol fell significantly in
the usual care as well as intervention group, at least up to 5 years of follow-up, and
then climbed in both at final visit. LDL cholesterol levels were however significantly
lower in the intervention group at 1 and 3 years, but not 5 years and at final visit. In a
sub-analysis of those who attended all scheduled clinic visits (n=461) these
differences at 1 and 3 years were larger, and LDL cholesterol levels were significantly
different at 5 years but remained similar at final visit.
Regarding concerns of fat restriction influencing growth and development, again other
measures of height, weight, sexual maturation by Tanner score, age of menarche in
girls, serum ferritin, and triglycerides were not different at any points between groups.
The lack of statistical significance at the later time points in this study, which
occurred in parallel with decreasing interventional contact, are highly suggestive that
a lack of dietary adherence and compliance to dietary aims may have contributed to
the lack of significant difference between groups at these points.
These results suggest that some dietary changes occurred in the usual care group, no
doubt as a result of the knowledge that the children had marginally raised cholesterol
levels, and would have received a degree of intervention to combat this in their usual
clinical care.
Childhood studies - prospective studies:
The Avon Longitudinal Study of Parents and Children (ALSPAC) (Cowin et al.,
2001; Rogers et al., 2001) enrolled approximately 15,000 mothers; a proportion of
these enrolled in the last six months of study initiation were included in a sub-study
called Children In Focus, which forms the basis of the dietary data reported here.
Amongst these, diet records were assessed at 18, 31 and 43 months. Non-fasting
venous blood samples were obtained at 31 and 43 months and analysed for total and
HDL cholesterol. At the 18-month dietary assessment, complete records were
obtained for 517 boys and 434 girls, and at 43 months 488 boys and 375 girls.
No association was found between dietary fat intake at 43 months and blood lipid
measures at the same time. However a relationship was found with dietary intake
reported at 18 months and blood lipid measures made at 31 months, with total
cholesterol levels at 31 months significantly associated with quartile of fat intake at 18
months. There was an interaction of gender of marginal significance (p=0.052) where the effect of dietary fat on total cholesterol seemed to be greater for boys. A trend
towards a positive association between intake of dietary fat at 18 months and nonHDL cholesterol at 31 months was also observed, of borderline significance at
p=0.054 (Rogers et al., 2001). Taking dietary intakes as continuous variables rather
than quartiles, Pearson correlation coefficients were calculated for dietary intake
variables and lipid measures (Cowin et al., 2001). Saturated fat intake at 18 months
was significantly correlated with total cholesterol level at 31 months in boys (Pearson
correlation coefficient (p-value) = 0.211 (0.002)) but not girls (coefficient (p-value) =
0.015 (0.840)). The same relationship was found for calculated LDL cholesterol and
saturated fat intake in boys (correlation coefficient (p-value) = 0.174 (0.042)) but
again not in girls (correlation coefficient (p-value) = -0.009 (0.927)). These gender
differences are in agreement with those found in the STRIP study detailed above.
As dietary fats play an important role in early development there were concerns that
fat restriction could have an adverse effect on growth. However in the ALSPAC
study, no association was found between dietary fat intake and the growth measures
of height and BMI.
These three studies offer interesting insights into the effect of reduction of saturated
fat intake in children. Of particular interest is the striking sexual divergence of the
effect of dietary modification in the STRIP and ALSPAC projects. Despite significant
reductions in saturated fat intake across all time points measured in the STRIP study,
girls in the intervention group showed no reduction in LDL cholesterol measured at 5
and 7 years, or total or non-HDL cholesterol measured prior to 5 years of age. Boys
exposed to the dietary intervention on the other hand showed significant and
persistent reductions in non-HDL cholesterol as measured before 5 years of age, and
LDL cholesterol at ages 5 and 7 yrs. The reasons for this difference between boys and
girls are not known, although the authors speculate on possible hormonal influences
or the effect of differences in physical activity levels reported between boys and girls.
There are a number of potential confounding factors in the link between reduction in
dietary saturated fat and blood lipids in the studies mentioned above. As shown in the
DISC study, significant changes can occur in the population without intervention, as
occurred in the control group in this study. One possible reason for this is that subjects
were mildly hypercholesterolaemic and thus were likely to have received intervention
to reduce cholesterol levels in the course of their usual care. Dietary saturated fat
intake did fall in the usual care group, but not nearly as much as the intervention
group. A comparison of means and 95% confidence intervals of lipid measures
suggests that dietary saturated fat intakes at 5-year follow-up and the final visit, while
being significantly different, were not sufficiently different to result in statistically
significant differences in LDL between groups, as it can clearly be seen that mean
LDL cholesterol levels tended to be higher in the usual care group than intervention at
5-years and final follow-up. This absence of statistical significance could be due to a
lack of intensity in the intervention being offered during the later years of the trial,
especially given the trial was terminated prematurely due to lack of funding. The
results pertaining to 1-year and 3-year follow-up points in this study certainly suggest
that a reduction in saturated fat intake in children from ages 8 - 10 yrs can produce a
reduction in LDL cholesterol, however the 5-year and final visit results, and
significant drop in LDL cholesterol in the usual care group mean that these
differences in LDL cholesterol cannot be easily attributed to changes in saturated fat intake alone. Unfortunately, results for this study are also presented as pooled results
for girls and boys with statistical adjustment for sex, and thus cannot be compared
with the results of the STRIP and ALSPAC projects with regard to gender differences.
Thus these three intervention trials in children confirm earlier observational studies
and show the effect of reducing saturated fatty acids on total and LDL cholesterol.
However they do show the difficulty in achieving long term compliance to the type of
dietary advice necessary to bring about the required changes. The reason for the
different responses between boys and girls remains elusive.
Metabolic/Intervention studies:
One of the more important studies published in the 1999-2005 time period was the
KANWU study (Rivellese et al., 2003; Vessby et al., 2001). The study took its name
from the five centres involved – Kuopio, Aarhus, Naples, Wollongong, Uppsala – and
thus was a multinational multicentre study consisting of controlled dietary
intervention for 90 days in healthy men and women. One hundred and sixty-two
subjects were included provided they were otherwise healthy with the exception of
some being overweight. The study design consisted of randomisation to either a
saturated fat or monounsaturated fat diet (plus a further division within each of these
into placebo or n-3 fatty acid supplementation) which was followed for 3 months. The
diets were well controlled, and analysis of nutrient intakes at the end of the dietary
period showed that total energy, total fat, polyunsaturated fat, protein and
carbohydrate intake did not differ across groups. Thus the only difference in source of
dietary energy arose from substitution of saturated or monounsaturated fat. Saturated
fat intake was 17.6 ± 2.5 % total energy and 9.6 ± 1.8 % total energy in the saturated
and monounsaturated diets respectively, with monounsaturated fat intakes being 13.1
± 2.5 % total energy and 21.2 ± 4.2 % total energy respectively. Given the difference
in saturated fat intakes, dietary cholesterol also differed at 322 ± 91 mg/day vs. 254 ±
80 mg/day in saturated fat vs. monounsaturated fat diets respectively. Thus, the diets
represented an approximately 8% variation in total energy from either saturated or
monounsaturated fat.
Mean difference (95% CI, p-value) in total serum cholesterol was 0.28 (0.12 – 0.45,
p=0.0007) mmol/l and for LDL cholesterol was 0.34 (0.19 – 0.49, p=0.0001) with the
saturated fat diet being the higher for both of these measures. No effect was seen on
triglyceride or HDL levels, with respective differences (95% CI, p-value) for saturated
fat diet minus monounsaturated fat diet values of 0.01 (-0.15 to 0.17, p=0.4879)
mmol/l and -0.01 (-0.08 to 0.07, p=0.6708) mmol/l.
The Dietary Approaches to Stop Hypertension (DASH) trial had as one of its
secondary outcome measures the effects of the DASH diet on lipid levels (Obarzanek
et al., 2001b). The study was a randomised intervention conducted in four sites,
comparing a control diet with the DASH diet and a diet high in fruit and vegetables.
The DASH diet comprised a higher proportion of total energy from protein than
control, 10% less energy from total fat, and more energy from carbohydrate. Saturated
fat comprised 14% of total energy in the control diet vs. 7% in the DASH diet. Of
particular interest is that the diet high in fruit and vegetables had an almost identical
macronutrient composition and contribution of energy from MUFA, PUFA and
saturated fat as the control diet.
A strength of the study is the large sample size of 459 participants. Dietary
intervention lasted for 8 weeks and consisted of participants consuming one prepared
meal onsite per day, at which time they were provided with additional meals and
snacks to cover them until their next visit. Energy intakes were adjusted as needed so
that body weight was kept constant, and after randomisation baseline characteristics
and lipid profiles of dietary groups were not significantly different.
The DASH diet resulted in significantly lower total cholesterol relative to the control
diet, with a mean difference (95% CI) of 0.35 (0.22, 0.49) mmol/l. Likewise LDL cholesterol was 0.28 (0.16, 0.40) mmol/l lower in the DASH diet group. HDL was
also significantly lower at 0.09 (0.06, 0.13) mmol/l mean difference. These equate to
reductions of 7.3% in total cholesterol, 9.0% in LDL cholesterol and 7.5% in HDL
cholesterol. Triglyceride, total:HDL cholesterol ratio and LDL:HDL ratio showed no
significant differences between DASH and control diet groups. Separate analyses by
race, baseline cholesterol concentration and sex revealed the same trends, with the
interaction of greater reductions in total and LDL cholesterol seen in men than
women. No significant differential reduction was seen based on differing baseline
levels of total cholesterol, LDL cholesterol or triglyceride. However, greater reduction
in HDL was observed for those with higher baseline HDL levels, a finding of merit
given that for those with already low HDL levels at baseline – placing them at a
higher cardiovascular risk – any further reduction would only amplify the effect of
their already low HDL levels on cardiovascular risk.
Of interest are the findings relating to the fruit and vegetable diet, in which no
significant differences were seen in any lipid measures compared with the control
diet. Given their identical compositions in terms of macronutrients and types of
dietary fat this is perhaps not surprising, and contributes to the evidence which
suggests that the higher fibre and lower dietary cholesterol intake do not have an
appreciable effect on lipid measures.
With regard to smaller, shorter studies published between 2000 and 2005, Judd et al.,
(2002) conducted a randomised dietary crossover trial in 50 men – arguably still a
reasonable sample size for a study of this design. The results of the study are more
pertinent to the role of dietary trans fatty acids, hence the study is described in detail
below under ‘Trans fatty acids and LDL cholesterol: Metabolic/Intervention studies’.
One of the six diets under investigation provided 18% of total energy from lauric,
myristic and palmitic acids combined, with another 2.7% from stearic acid. This diet
produced one of the highest total serum cholesterol levels of the diets under
investigation, significantly more than diets with high amounts of carbohydrate, oleic
acid, or stearic acid. Similarly, LDL cholesterol was significantly higher on this diet
than seen on diets high in carbohydrate or oleic acid.
Poppitt et al. (2002) conducted a smaller trial in 20 healthy young men. The main aim
of the study was to compare the effects of two diets differing only in the type of
butter, one a standard butter, the other produced by modifying bovine diets to produce
a butter lower in saturated fat and higher in poly- and monounsaturated fats. The two
diets did not differ in macronutrient composition, however the modified butter diet
had significantly lower saturated fat (15% total energy compared to 20% total
energy). The modified butter fat diet had lower amounts of C10:0, C14:0 and C16:0
than the control but C12:0 and C18:0 content was greater. The modified butter diet
also provided significantly less dietary cholesterol. Given the same composition of the
diets in terms of macronutrients, the modified butter diet thus represented a reduction
in saturated fat with increases in PUFA and MUFA. The study was a randomised,
cross-over trial in which not only was all food and drink provided, but subjects also
lived in the metabolic unit of the study site for the diet periods hence guaranteeing a
very high level of dietary compliance. Dietary intervention periods were 3 weeks,
separated by a minimum washout period of 4 weeks. At randomisation there were no significant differences in baseline variables for either
of the dietary groups, and importantly there was no change in body weight in either
group across dietary periods which would otherwise influence lipid profiles.
In terms of lipid profiles, the only significant treatment effects between normal and
modified butter diets were that the modified butter diet resulted in lower total
cholesterol and LDL cholesterol levels than the standard butter diet. Blood samples
were collected throughout the treatment period as well as at baseline and the end of
dietary periods. Significant differences in total and LDL cholesterol between diets
achieved statistical significance only at the end of the 3 week dietary intervention
period.
The amounts of saturated fat used in this study (15% and 20% total energy) are
relatively high compared to current usual recommendations. However the replacement
of 5% of total energy from saturated fat for polyunsaturated and monounsaturated fat
represents a realistic change for free-living individuals, especially in the case of
having a particularly high intake of saturated fat.
In a number of studies the effect of replacing dietary saturated fat has not surprisingly
varied depending on which dietary components replace energy lost by reduction in
saturated fat. In two trials, Hodson et al. (2001) examined the effect of replacing
dietary saturated fat with either n-6 polyunsaturated fatty acids or monounsaturated
fat. The two separate trials were both very similar in design: a randomised cross-over
dietary intervention, with subjects switching between two diets consumed for 2 and a
half weeks each without an intervening washout period.
In trial I (n = 29) a diet high in saturated fat was compared with an n-6
polyunsaturated fatty acid diet. Trial II (n = 42) compared the same saturated fat diet
with a diet high in monounsaturated fat. All diets had total fat intakes of
approximately 30% of total energy. Subjects were Human Nutrition students, and
were asked to self-select a background diet which would be modified to achieve the
dietary goals for each phase of the trials. Subjects were provided with a recipe book
for ideas and provided with specific spreads and oils depending on diet allocation. For
the high saturated fat diet, subjects were asked to use butter as a spread and in baking
and cooking and to consume high fat dairy foods. In the n-6 polyunsaturated fat diet
participants were provided with safflower and sunflower oils to use as a spread and in
baking and cooking, and similarly subjects on the monounsaturated fat diet were
given canola oil and spreads. Dietary intake was assessed by 3 day weighed diet
record.
Participants in trial I (saturated vs. n-6 polyunsaturated fats) were not significantly
different in age, weight and BMI from those in trial II (saturated vs. monounsaturated
fats). Mean saturated fat intake (SD) assessed by diet records in saturated fat diets
were 17.5 (4.2) % in trial I and 17.7 (4.2) % in trial II, compared with 8.5 (2.8) % on
the polyunsaturated fat diet in trial I and 8.4 (2.8) % on the monounsaturated fat diet
in trial II. Proportion of energy from total fat was comparable between the diets,
although total energy intake was reduced on both the polyunsaturated and
monounsaturated diets relative to the saturated fat diet, with 1.6MJ (358 kcal) less on
the polyunsaturated and 1.7MJ (406 kcal) less on the monounsaturated fat diets,
apparently largely due to changes in alcohol intake. Mean body weight however did
not change on the mono- or polyunsaturated diets relative to the saturated fat diets.
Total cholesterol levels were significantly reduced on the poly- and monounsaturated
fat diets compared with the saturated fat diet, by 19% and 12% respectively (p<0.001
for both comparisons). LDL cholesterol was also significantly lower, with a reduction
of 22% (p<0.001) on the polyunsaturated fat diet, and 15% (p<0.001) on the
monounsaturated fat diet.
Metabolic/Intervention studies – individual saturated fatty acids
A number of studies conducted prior to 2000 suggested that individual fatty acids
exert differing effects on lipid profiles. In the time period under review in this report,
only a handful of studies which specifically examined individual fatty acids were
identified. Unfortunately the majority of these were underpowered, or there were
methodological issues such as weight loss, and thus it was not possible to draw
definitive conclusions from their results.
In a study conducted in Colombia, Bautista et al. (2001) examined the effects of palm
oil, high in palmitic acid, in the context of high and moderate cholesterol diets.
Twenty-eight healthy male university students were randomised to a 4 x 4 doubleblind dietary cross-over trial. The four diets were: high palm oil-high cholesterol; high
palm oil-low cholesterol; moderate palm oil-moderate cholesterol; and low palm oilmoderate cholesterol. All diets contained 57% of energy from carbohydrates, 12%
from protein and 31% from fats. Being university students, diets were prepared in the
cafeteria of a university residence and subjects were served breakfast, lunch and
dinner on-site, with staff keeping records of dietary compliance. In all other respects
subjects were free-living individuals and were also allowed to consume food outside
of the study, intake of which was recorded and nutrient intake estimated from
Colombian food comparison tables. Diets were consumed for 4 week periods. All
subjects completed the study, and 90% of all diets provided were consumed by
participants; compliance did not differ between the four diets or with diet order. Foods
consumed outside of the study were mostly bottled drinks, sweets and bread. Across
the study period an average but non-significant reduction in body weight of 1.1 kg
was observed (p=0.09).
Considering the high vs. low dietary cholesterol comparison, high cholesterol diets led
to a significant increase in total and LDL cholesterol, but no change in HDL
cholesterol. Similarly high palm oil vs. low palm oil produced an increase in total and
LDL cholesterol but no effect on HDL cholesterol. The effects and level of
significance of dietary palm oil was greater than dietary cholesterol, with the high vs.
low cholesterol diets producing a 0.21 mmol/l increase in total cholesterol (p=0.009)
and 0.16 mmol/l increase in LDL cholesterol (p=0.047), whereas the high vs. low
palm oil diets produced an increase in total cholesterol of 0.39 mmol/l (p<0.001) and
LDL cholesterol of 0.38 mmol/l (p<0.001).
One aspect considered in depth by the authors was the variability in responses
between individuals. Large variation in individual response to a high cholesterol diet
has been reported in a number of other studies, and was also seen here. Lipid response to the high palm oil diet varied amongst participants but was less striking than the
variability in responses to dietary cholesterol. Furthermore, responses to dietary
cholesterol correlated with responses to dietary palm oil, with Pearson’s correlation
coefficients of 0.46 (p=0.02) for total cholesterol and 0.41 (p=0.03) for LDL
cholesterol. Extent of total cholesterol response to dietary palm oil was also correlated
with baseline total cholesterol, a relationship not observed when considering total
cholesterol response to dietary cholesterol.
Other data published in recent years regarding the influence of individual saturated
fatty acids include those of Tholstrup et al. (2003), who observed an increase in HDL
cholesterol within 24 hours of a myristic acid enriched diet, but not after a stearic acid
enriched diet. Kelly et al. (2002) observed a significant decrease in LDL cholesterol
after a diet enriched with stearic acid, but not a diet enriched with palmitic acid. These
studies were both cross-over design in healthy young males, although were also both
small with 9 and 10 subjects.
Predictive modelling equations:
Mensink et al. (2003) conducted a meta-analysis of 60 controlled trials. Articles
published between January 1970 and December 1998 were included, and had to meet
the following criteria: food intake clearly described with dietary fatty acids as the sole
variable; dietary cholesterol constant between comparison diets; parallel, cross-over
or Latin-square design; diet adherence for at least 13 days to allow time for changes in
lipid measures to take place; adults with no disturbances of lipid metabolism or
diabetes included. A total of 60 studies were selected for the meta-analysis.
The estimated regression coefficient (95% CI) for the effect of replacement of 1% of
total energy from carbohydrates with 1% of energy from SFA’s on LDL cholesterol
was 0.032 (0.025, 0.039) mmol/l. Corresponding regression coefficients for total
cholesterol and total:HDL cholesterol ratio were 0.036 (0.029, 0.043) mmol/l, and
0.003 (-0.008, 0.013) respectively. Thus, replacement of carbohydrates with SFA’s
was predicted to significantly increase LDL cholesterol and total cholesterol, although
produce no significant change in total:HDL cholesterol ratio.
Intakes of individual saturated fatty acids were reported for 35 studies, with mean
intakes being: 1.1% of total energy for lauric acid, 1.3% for myristic, 6.2% for
palmitic, and 3.0% for stearic.
Lauric acid appeared to be the most potent individual saturated fatty acid when
considering the increase in LDL and total cholesterol, although this was also true for
HDL cholesterol, such that replacement of carbohydrates with lauric acid actually
decreased total:HDL cholesterol ratio. No effect of myristic, palmitic or stearic acids
was predicted on total:HDL cholesterol ratio when these fatty acids replaced
carbohydrate constituting 1% of total energy.
Regarding the effect of substitution of 1% of total energy from carbohydrates with
individual saturated fatty acids on total cholesterol and LDL cholesterol, there was a
trend toward greater increases with decreasing chain length. Regression coefficients
are shown in the table below.
It should be noted that stearic acid was not associated with a change in total or LDL
cholesterol when replacing carbohydrate. Also, as mentioned above, while lauric acid
resulted in the greatest increase in LDL cholesterol, it had a greater relative effect on
HDL cholesterol so that it was predicted to decrease total:HDL cholesterol ratio
relative to carbohydrates. None of the other individual saturated fatty acids influenced
total:HDL cholesterol ratio.
Thus in summary the results of this meta-analysis suggest that myristic acid has the
greatest detrimental effect on the overall lipid profile, followed by palmitic acid.
Stearic acid was not predicted to significantly alter lipid levels, and while lauric acid
showed the greatest ability to raise LDL cholesterol, total:HDL cholesterol ratio is
lower.
Müller et al. (2001) also published predictive equations relating to the effect of
individual dietary saturated fatty acids on LDL cholesterol. The key focus of the paper
was on trans fatty acids, but analyses relating to the effects of lauric, myristic and
palmitic acids were included.
Predictive models were derived from the results of four studies carried out from 1993
– 1998 by the authors at the University College of Akershus in Norway. These studies
involved Latin-square or crossover designs with participants receiving the
experimental different diets in random order. Blood samples were taken after 3 weeks
of dietary intervention in three studies and after 2 weeks in the fourth. One study
involved only male participants and the other three only female participants, with the
number of participants ranging from 16 to 31, mean ages 22 to 31 years, mean BMI
23 to 26 kg/m2
, and mean baseline serum cholesterol 4.44 to 5.35 mmol/l. Thus the
studies were relatively small and conducted in healthy young normal weight
participants. In the final regression analysis, the results for four overweight
participants in one of the trials were excluded due to significant weight loss of 1.7 – 7
kg during the study.
Predictive equations were generated which included lauric (C12:0), myristic (C14:0)
and palmitic (C16:0) acids, trans fatty acids, and also included terms for oleic,
linoleic and alpha-linolenic polyunsaturated fatty acids (using coefficients derived by
Yu et al. (1995)). Stearic acid (C18:0) was not included in the analysis as the authors
considered that there was sufficient previously published evidence that stearic acid
had no effect on total or LDL cholesterol levels.
Regression coefficients (SD) for the effect of each 1% change of total energy from
individual saturated fatty acids on LDL cholesterol were: 0.01 (0.17) for lauric acid;
0.071 (0.011) for myristic acid; and 0.047 (0.032) for palmitic acid. Similar results
were obtained for total cholesterol.
The developed models were tested against other datasets by applying them to the
results of seven dietary studies which included the fatty acid composition of the
experimental diets (including trans fatty acid intakes) and were deemed to be welldesigned. The Pearson correlation coefficient between predicted and observed total
cholesterol in these studies was r = 0.981. A correlation coefficient specifically for
LDL cholesterol in these studies is not given.
It should be noted that there was a large uncertainty for the coefficient for lauric
(C12:0) acid in this study, and the results for this fatty acid should thus be interpreted
with caution. The models developed seem to behave as robust predictors of change in
lipid profile when compared with results from seven other dietary interventions.
The authors concluded that C12:0 did not raise serum total cholesterol as much as
myristic (C14:0) or palmitic (C16:0) saturated fatty acids. When considering isoenergetic comparison of the individual saturated fatty acids, myristic acid was
predicted to cause the greatest increase in LDL cholesterol followed by palmitic acid.
These results are generally consistent with those of Mensink et al. (2003) in as much
as myristic and palmitic are clearly the most important fatty acids in determining
elevation of total and LDL cholesterol. The intake of lauric acid in most populations is
lower and the effect on LDL appears to be less consistent. However it is of interest
that this fatty acid increases HDL to a greater extent than the other saturated fatty
acids and is this associated with a relatively favourable ratio of total:HDL cholesterol
SFA’s and coronary heart disease:
Cross sectional epidemiological studies:
The EURODIAB-IDDM study (Toeller et al., 1999), outlined above under ‘Saturated
fatty acids and LDL cholesterol: cross sectional epidemiological studies’, also
gathered data on diagnosed cardiovascular disease. Out of a sample of 2868
individuals with type 1 diabetes, 286 cases of CVD were identified. As with the link
between dietary saturated fat and LDL cholesterol, odds ratios for cardiovascular
disease adjusted for total energy intake were significantly higher with increasing
intakes of total fat, saturated fat and cholesterol. However in multivariate analysis
taking into account other dietary and clinical variables these associations were no
longer statistically significant.
A cross-sectional study of saturated fat intake and risk of nonfatal MI was conducted
in Costa Rica by Kabagambe et al. (2003). Four hundred and eighty-five male and
female survivors of first MI were enlisted as cases and 508 controls matched by age,
sex and area of residence were identified. Dietary information was collected with the
use of a previously developed and validated food frequency questionnaire,
administered during home visits by trained personnel.
Cases reported significantly higher intakes of total fat, animal fat, saturated fat,
individual saturated fatty acids, cholesterol and energy than controls. Mean saturated
fat intake (SD) as percentage of total energy was 11.7 (2.8) % for controls and 12.4
(2.8) % for cases. Multivariate models for risk of MI were generated using the highest
quintile of intake vs. the lowest quintile with adjustment for established CVD risk
factors and dietary confounders. Attempts were also made to identify the effects of
individual saturated fatty acids, pooling lauric and myristic acid due to their low
intake levels.
Relative risk (95% CI) of MI for the highest quintile of total saturated fat intake vs.
lowest was 3.00 (1.54 – 5.84). Odds ratios (95% CI) for MI for a 1% change in total
energy from lauric plus myristic acid were 1.08 (0.65 – 1.81), 1.16 (1.01 – 1.32) for
palmitic acid, and 1.98 (1.08 – 3.65) for stearic acid. The corresponding OR for a 1%
increase in energy from total saturated fat (reflecting mainly palmitic, this being the
most abundant individual fatty acid) was 1.12 (1.03 – 1.21).
This study was of sufficient size to show potential links between dietary saturated fat
and nonfatal MI, and additionally was well-conducted. Increasing saturated fat intake
was clearly shown to raise risk of MI, and this also held true for individual saturated
fatty acids. It is interesting to note, as previously mentioned in Part 1 of this report,
that although the above metabolic studies generally do not support a link between
stearic acid and LDL cholesterol, it emerged as conferring the greatest risk per
percentage of total energy of all the individual saturated fatty acids (with the effects of
lauric or myristic acid analysed as a combined group). This suggests that stearic acid
may exert effects on cardiovascular risk other than via effects on lipid profile. Prospective studies:
Oh et al. (2005) reported on the effects of dietary fat intake in the Nurses’ Health
Study. A previous paper from the same study (Hu et al., 1997), described above under
the heading ‘Part 1b: Reanalysis of pivotal studies’, reported on dietary fat intake and
risk of CHD over 14 years of follow-up. This paper extended the time period to 20
years of follow-up. Dietary intake was assessed via posted food frequency
questionnaires, the first being a 61-item questionnaire sent out in 1980, which was
extended to 116 items in 1984, followed by similar questionnaires in 1986, 1990,
1994 and 1998. The endpoint was nonfatal myocardial infarction or fatal CHD
between 1980 and June 1, 2000. After exclusion of those with incomplete FFQ’s or
previous diagnosis of CHD, cancer, diabetes or hypercholesterolaemia before 1980,
78,778 women were included in the analysis. Over the 20-year follow-up period 1,241
cases of nonfatal MI and 525 CHD deaths were recorded (total 1,766 incident cases).
From 1980 to 1998 fat intake fell from 39% of total energy to 29%. Saturated fat
intake ranged from 10.1% of total energy in the lowest quintile to 17.6% of total
energy in the highest quintile of intake.
A comprehensive multivariate relative risk calculation was made between saturated
fatty acid intake quintiles, calculated based on a cumulative average intake from all
available food frequency questionnaires, and incident CHD including numerous
possible confounding factors.
Multivariate relative risk of incident coronary heart disease was 0.97 (0.73-1.27) for
the highest vs. the lowest quintile of saturated fat intake.
This is in contrast with a previous 14-year follow of the Nurses’ Health Study (Hu et
al., 1997) which did indeed show an association. In this previous analysis each
increase of energy intake of 5% from saturated fat, replacing carbohydrate, was
associated with a relative risk (95% CI) of coronary heart disease of 1.17 (0.97 – 1.41,
p-value = 0.10). The authors suggest that the lack of association in this latest analysis
(Oh et al., 2005) could be due to declining intakes of saturated fat and incidence of
coronary heart disease weakening the associations found previously.
Boniface et al. (2002) reported on 16-year follow-up of the 1984-85 Health and
Lifestyle Survey in Great Britain. Initial response rate to a face-to-face interview,
physical measurements, and self-reported food frequency questionnaire was 73.5%.
Ninety-seven percent of these were flagged for on-going monitoring of death
certificate information. This study took into account coronary heart disease deaths
amongst those aged 40-75yrs at the initial interview up until December 2000. Cox
survival analyses for survival without CHD death relative to fat intake were made for
men and women separately as well as combined, after adjustment for confounding
factors of age, alcohol intake, smoking, body shape (BMI or waist-hip ratio), exercise,
blood pressure, social class and deprivation index.
An association was found between total dietary fat and saturated fat and CHD death
amongst women but not for men. Multivariate relative risk (95% CI) of CHD death
for an increase of 100 g/week of saturated fat was 1.40 (1.09 – 1.79) for women. The corresponding risk estimate for men was 1.00 (0.86 – 1.18). An analysis according to
age suggested that this relationship was stronger in older women.
A similar 16-year follow up study was reported by Jakobsen et al. (2004). 3,686
Danish men and women were identified through their participation in four other
cohort studies: the 1914 and 1936 cohorts, and the MONICA-I and MONICA-III
cohorts. Median length of follow up was 16 years (range 7-22 years). Participants
filled in 7 day weighed food records (or a small proportion interviewed by dietitian)
and hospitalisation and death records obtained from the National Patient Registry and
Cause of Death Registry.
Cox’s proportional hazard regression models were developed and a comprehensive
fully-adjusted model included confounding factors of fat intake, total energy intake,
protein, other types of dietary fatty acids, family history of MI, smoking, physical
activity, education, alcohol intake, dietary fibre, and dietary cholesterol, as well as
further adjustments for blood pressure and BMI.
Three hundred and twenty-six cases of fatal or nonfatal CHD were identified (98
women and 228 men). Hazard ratio for development of CHD in women with an
increase of 5% energy from saturated fat (replacing carbohydrate) was of borderline
significance at 1.36 (0.98, 1.88). A further examination of the data according to age
group revealed an appreciable and significant increase in risk in the younger women
(hazard ratio (95% CI) for a 5% increase in saturated fat intake = 2.68 (1.40, 5.12))
but not in the age group over 60 years (HR (95% CI) = 1.22 (0.86, 1.71)). As in the
Boniface et al. study, no significant association between saturated or total fat intake
and risk of coronary heart disease was seen in men.
Despite their similar designs and length of follow-up it is interesting to note that
Boniface et al. reported a stronger association in women 60 yrs of age and older,
whereas Jakobsen et al. found a stronger effect amongst women under 60 yrs, with no
significant effects seen in women over 60 yrs. There is no obvious explanation for this
discrepancy nor why the relationship should be present in women but not in men.
A follow-up of the Seven Countries study was published in 2000 which considered
the effect of dietary and lifestyle variables on all-cause mortality after 25-years of
follow-up (Kromhout et al., 2000). It is interesting to note that despite all the
limitations of this study the model that best predicted observed all cause mortality
rates was one which included saturated fat intake, vitamin C and smoking. Based on
this analysis, a 5% reduction in saturated fat intake at a population level would lead to
a 4.7% reduction in age-adjusted all cause mortality.
Thus although there is support in the more recently published data for the association
between saturated fatty acids and cardiovascular disease, there is heterogeneity in the
results of the studies, particularly with regard to the different effect in men and
women and in different age groups. There is no clear explanation for this
heterogeneity.
Trans fatty acids:
Summary:
There has been much debate regarding the link between dietary TFA and adverse
health outcomes since the time of the Health Canada report. Numerous editorials,
comments and reviews were retrieved while searching the evidence published since
1999; a selection are summarised in table 2.2 at the end of this document. In some
countries government action and legislation regarding food labelling and trans fatty
acid content of industrially produced foods has taken place on the basis of reports
from agencies such as the Danish Nutrition Council, European Food Standards
Agency, and the US Food and Drug Administration.
Numerous dietary intervention and metabolic studies published between 2000 and
2005 on the effects of dietary trans fatty acids support their role in increasing LDL
cholesterol. While there is still some uncertainty as to whether this increase is more,
less or equal to that of saturated fat, the association between intakes of trans fatty
acids and LDL cholesterol is unquestionably “convincing”. In addition, while some
studies covered in the time period under review showed no significant effect of
dietary trans fatty acids on HDL cholesterol levels, the majority document a decrease
in HDL cholesterol when trans fatty acids replace dietary carbohydrate, and metaanalyses confirm a more adverse effect on overall lipoprotein profile (LDL:HDL
ratio) than that conferred by saturated fatty acids. For the purposes of substantiation of
a health claim the much debated issue regarding the relative adverse effects of trans
and saturated fatty acids on LDL cholesterol is of little relevance.
Two cross-sectional studies examining dietary trans fatty acids and LDL cholesterol
have been published during the period 2000 – 2005; one carried out in Costa Rica, the
other a multinational European collaboration, the TRANSFAIR study. The Costa
Rican study found no evidence for an association between trans fatty acid intake and
LDL cholesterol. The TRANSFAIR study however reported an inverse relationship
between total trans fatty acid intake and LDL cholesterol levels, with differing effects
for different trans isomers: t18:1 and t18:2 were inversely associated with LDL
cholesterol levels, while t14:1 (n-9) and t22:1 showed a positive association. It is
noteworthy that the mean intakes of trans fatty acids reported in both of these studies
were very low, with trans fatty acids contributing less than 1% of total energy. It is
conceivable that there may be a threshold effect whereby low levels of trans fatty acid
intake do not effect lipid levels as do higher intakes. However regardless of the
explanation of these findings, such cross-sectional studies do not negate the findings
of the more powerful intervention studies.
Recent evidence from the major prospective cohorts the Nurses’ Health Study and
Zutphen Cohort study show a clear link between dietary trans fatty acid intake and
incidence of fatal and nonfatal cardiovascular disease. The report from the Zutphen
Cohort Study also examined the pooled effect of dietary trans fatty acids from these
two studies and the previously published Health Professional’s Study and AlphaTocopherol Beta-Carotene Cancer Prevention Study. There is also a clear positive
association between dietary trans fatty acids and cardiovascular disease across these
four large cohort studies, with evidence for a dose response relationship.
Two case control studies considered in the report from Health Canada found no
association between adipose trans fatty acids and incidence of myocardial infarction
or sudden cardiac death (Roberts et al., 1995; Aro et al., 1995). In contrast three
similar studies reported since 2000 have reported a clear association of increased risk
of nonfatal or fatal myocardial infarction with increasing dietary intakes or adipose
tissue content of trans fatty acids. No explanation for this disparity is offered by the
authors of three recent studies, nor indeed could we find a definitive one. It should be
noted that in the EURAMIC study (Aro et al., 1995) a trend towards a positive
association between adipose trans fatty acids and acute myocardial infarction was
present after exclusion of two outlying centres with low adipose trans fatty acid
levels. The study by Roberts et al. is the only one considered here, and in the Health
Canada report, to examine the link between adipose tissue trans fatty acids and
sudden cardiac death. Thus samples from cases were taken during post-mortem
examinations (median time of 24 hours after death) rather than from survivors of first
acute MI. Furthermore samples were collected from the anterior abdominal wall and
not from abdomen or buttock tissue aspirates as in other similar studies. Whether
these differences in methodology could account for the differences in results is
unknown.
Despite the strength of the evidence from the cohort studies the heterogeneity in the
data suggest that at the present time it may be more appropriate to describe the
association as “probable” rather than “convincing”.
Trans fatty acids and LDL-cholesterol:
Cross sectional epidemiologic studies
Kim et al. (2003) examined trans fatty intake and LDL particle size in a crosssectional study of adults in Costa Rica. Participation rates were high at 85% for men
and 93% for women, resulting in a final sample size of 202 men and 212 women.
Dietary information was assessed using a semi-quantitative food frequency
questionnaire. Trans fatty acid intake was higher in urban areas than rural, and also in
women than men, although intakes in this population were particularly low in general,
with intakes barely reaching 1% of total energy. Although specifically examining the
effect of trans fatty acid intake on LDL particle size, the authors did not find a
significant correlation between trans fatty acid intake measured by food frequency
questionnaire and plasma LDL or HDL cholesterol levels*
. Thus, it is conceivable that
low levels of trans fatty acid intake may not appreciably influence LDL cholesterol
levels.
Another important cross-sectional epidemiologic study was the multi-national
European TRANSFAIR study which included an investigation of the relationship
between trans fatty acids and cardiovascular risk factors (van de Vijver et al., 2000).
Subjects were 327 men and 299 women aged 50 – 65 yrs in Finland, Iceland,
Netherlands, France, Sweden, Spain, Greece, and Portugal. Response rates were
variable between countries (20 - 66% for men and 25 - 72% for women). Those who
had recently altered their diet or were on certain medications were excluded. Blood
lipids were determined from a fasting venous sample. Total TFA amounted to 2.40
g/d or 0.87% total energy in men and 1.98 g/d or 0.95% total energy in women.
Lowest intake of TFA was reported in Portugal at 0.57% total energy and highest in
Iceland at 1.65% total energy.
In half of the men and women participating, adipose tissue samples were collected
from the buttock for the purposes of validating food frequency questionnaire data.
Significant correlations between dietary intake and adipose tissue measurement were
0.67 for total trans fatty acid intake, 0.36 for t16:1 intake, and 0.63 for t18:1 intake (p
< 0.01 for each).
Linear regression models were developed separately for men and women, using TFA
intake as a continuous variable and serum lipid levels as the outcome measure. Mean
total cholesterol levels were 5.7 ± 1.0 mmol/l for men and 6.2 ± 1.1 mmol/l for
women. Mean daily intake of total trans fatty acids was low in this study, with women
having a slightly higher intake than men at 0.95 ± 0.55 % total energy in women vs.
0.87 ± 0.48 % of total energy in men. An analysis of lipid levels across quintiles of
trans fatty acid intake revealed no association between total trans fatty acid intake
with LDL or HDL cholesterol, or LDL:HDL ratio. After adjustment for other fatty
acids, total cholesterol was found to have an inverse relationship with total trans fatty
acid intake.
* The authors also point out that in a previous study by Lichtenstein et al. (1999) (covered in
the Health Canada review) no association was found between trans fatty acid intake and LDL
or HDL cholesterol levels when comparing groups with a low intake of trans fatty acids (0.55% of total energy vs. 0.91% of total energy; see analysis of Lichtenstein et al., 1999
paper in Part 1 above).
Linear regression models for individual trans fatty acids were developed. Positive
associations between fatty acid intake and LDL cholesterol were found for t14:1 (n-9)
and t22:1. An inverse association was found for t18:1 and t18:2 after adjustment for
intake of other fatty acids. Regarding effects on HDL cholesterol, the only isomer
with a significant association was t16:1 (n-9) which showed an inverse relationship
with HDL cholesterol levels. The combined effects of these isomers on LDL and
HDL cholesterol levels were seen in regression models for LDL:HDL ratio, with the
three above mentioned individual trans fatty acids, t14:1 (n-9), t16:1 (n-9) and t22:1,
showing positive associations with LDL:HDL cholesterol ratio.
The TRANSFAIR study is the only published study reporting an inverse association
of either total or LDL cholesterol with trans fatty acid intake. Positive associations
with LDL cholesterol levels were reported for t14:1 (n-9) and t22:1, however these
particular trans fatty acids contribute a very small proportion of dietary trans fatty
acids. Potential limitations of the TRANSFAIR study include the relatively low
response rate in some countries, with the lowest rates for any country being 20% for
men and 25% for women, thus those surveyed may not be a representative sample of
the population, especially relevant as while a being multinational study conducted
across Europe, a sample size of 329 men and 299 women would not even be a
particularly large study in any of the individual countries alone.
While the low response rate may to some extent account for the findings, the most
plausible explanation for the absence of an association between trans fatty acids and
total and LDL cholesterol may be the very low intakes of trans fatty acids in both the
Costa Rican and the TRANSFAIR studies. There is no clear explanation as to why
t14:1 (n-9) and t22:1 differed from other trans fatty acids.
Prospective studies
A Finnish prospective study of a low-saturated fat, low-cholesterol diet in children
also provides some evidence on the effect of dietary trans fatty acids in children. Salo
et al. (2000) published a paper examining the effect of dietary trans fatty acids in the
STRIP project (described above on page 25 under ‘Saturated fat and LDL cholesterol:
Childhood randomised controlled trials’). Given that in the intervention group
margarine intake was encouraged, it was assumed that trans fatty acid intake would
higher amongst those participants than in the control group.
In the entire cohort at 3 years of age (402 intervention children, 411 control) intake of
trans fatty acids were indeed significantly different between arms, although low in
both groups, at 0.8% total energy in the intervention group vs. 0.6% total energy in
the control group (p<0.001). Although HDL levels were on average lower in
intervention children, so too were non-HDL cholesterol levels, no doubt due to the
lower saturated fat intake in this group, and HDL:total cholesterol ratios were almost
identical. Importantly, while differences were found between the two groups it was
not possible to distinguish between the effects of saturated fatty acids and trans fatty
acids. Thus, while no specific detrimental effects could be ascribed to trans fatty acids
in this study the results should be interpreted with caution. One possible positive
interpretation of these findings it that low intakes of trans fatty acids do not appear to
be detrimental to lipid profiles of young children (nor for that matter on other
cognitive or behavioural parameters measured in the study (cited in accompanying
editorial by Deckelbaum and Williams, 2000)).
Metabolic/Intervention studies
Vermunt et al. (2001) examined the effects of dietary alpha-linolenic trans fatty acid
on plasma lipids and lipoproteins in 91 healthy men, in a multi-centre dietary
intervention carried out in Scotland, France, and the Netherlands. After a 6-week runin period on a ‘trans-free diet’ (less than 1g trans fatty acids per 100g total fatty
acids), subjects were randomized to either a low trans or high trans fatty acid diet for
the following 6 weeks. Subjects were advised to avoid other trans fatty acid
containing foods during the study period and dietary intake was assessed by 4 day diet
records at the end of both run-in and experimental periods (weeks 5 and 11). Subjects
on the high trans diet consumed an average of 1.41 g/d of trans alpha-linolenic acid.
Mean (SD) body weight increased in subjects on the high trans diet by 0.2 (1.1) kg,
and fell in those on the low trans diet by 0.6 (2.4) kg, and these changes were
significantly different between groups (p = 0.03). After correction for changes in body
weight and intakes of saturated and monounsaturated fats, the high trans group
showed a non-significant increase in plasma LDL, and a non-significant decrease in
HDL. Although these changes failed to reach significance, at p = 0.10 and p = 0.22
respectively, LDL to HDL cholesterol ratio increased by a mean change of 8.1% (1.4
– 15.3%, p = 0.02) on the high trans diet vs. low trans diet. Likewise, the treatment
effect of high vs. low trans diets on total cholesterol to HDL ratio was an increase of
5.1% (0.4 – 9.9%, p = 0.03). These differences were seen in each of the three study
centres, and were evident after three weeks consumption of experimental diets,
providing evidence that the specific trans fatty acid alpha-linolenic acid in itself has
an adverse effect on blood lipids.
Judd et al. (2002) conducted a large well-controlled study examining the effect of
trans fatty acid intake on plasma lipids and lipoproteins in men, concentrating on the
relative effects of dietary replacement of saturated fats, carbohydrate, or
polyunsaturated fatty acids with trans fatty acids. Fifty generally healthy men were
enrolled and randomly assigned to six experimental diets in a Latin-square design.
Average age was 42 yrs and mean BMI 26.2 kg/m2
. Diets were followed for 5 weeks
each. All meals were provided, with breakfast and dinner meals being consumed onsite, and carry-out lunch meals and all weekend and snack foods provided. Six diets
were used which consisted of a carbohydrate enriched diet (CHO), deriving 30.5% of
total energy from fats, and five other diets varying in fat content. The other five diets
aimed to provide 38.9% total energy from fat (actual range 38.2% - 39.9%) and thus
derived approximately an extra 8% of total energy from fat and 8% less from
carbohydrate than the CHO diet, and varied principally in the composition of dietary
fats: oleic acid diet (OL) which was enriched to provide 8% of total energy from oleic
acid; LMP diet which comprised 8% enrichment with lauric, myristic and palmitic
acids; a diet with 8% enrichment from stearic acid (STE); a diet of 8% enrichment
with trans fatty acids (TFA) of a range of t18:1 isomers; and a diet enriched with 4%
from stearic acid and 4% from trans fatty acids (TFA/STE). There was no significant
effect of diet order on lipid measures.
TFA, TFA/STE, and LMP diets were all associated with significantly higher total
cholesterol levels than the other three diets with respective means (SD): 4.991 ± 0.088
mmol/l; 4.981 ± 0.088 mmol/l; 4.957 ± 0.088 mmol/l. LDL cholesterol was again
higher on TFA and TFA/STE diets than on the other diets, with respective means
(SD) of 3.36 ± 0.08 mmol/l and 3.32 ± 0.083 mmol/l. The TFA, TFA/STE, and STE diets were all associated with low HDL levels, respectively: 1.159 ± 0.041 mmol/l;
1.174 ± 0.041 mmol/l; 1.156 ± 0.041 mmol/l. Total cholesterol to HDL ratios were
thus significantly highest on the TFA and TFA/STE diets at 4.5 ± 0.1 and 4.4 ± 0.1
respectively.
Those dietary effects which were of significance at the p = 0.01 level were
summarised as percentage changes with respect to substitution of other dietary
components in Table 6 in the original paper. Those of interest to this report are
summarised in the table below.
Of particular relevance to this discussion is the confirmation that large quantities of
trans fatty acids increase LDL to a greater extent than that of saturated fatty acids.
However once the intake of trans fatty acids is as great as 4% total energy, little
further increase in LDL cholesterol is seen when intake of trans fatty acids is greater.
Selected results from Table 6 of Judd et al., 2002.
One study which directly compared the relative effects on serum lipids of saturated
vs. trans fatty acids was de Roos et al. (2001). Twenty-nine volunteers (men and
women) were given 2 different diets – a high trans and high saturated fat diet – for
four weeks each in a randomised crossover design. The high saturated fat diet had
slightly higher total fat content (41.0% of total energy vs. 37.4%), and derived 22.9%
of total energy from saturated fat vs. 12.9% of total energy in the trans fatty acid diet.
The trans diet had a fairly substantial level of trans fatty acid intake at 9.4% of total
energy, with the saturated fat diet providing only 0.4% of total energy from trans fatty
acids. Serum LDL cholesterol and triglycerides were the same on both diets. Mean
(SD) serum HDL cholesterol was 1.87 (0.46) mmol/l on the saturated fat diet and 1.49
(0.33) mmol/l on the trans fatty acid diet, reflecting a mean decrease (95% CI) of 0.39 (0.28, 0.50) mmol/l or 21%. Total cholesterol concentrations differed solely as a result
of the difference in HDL concentrations.
Dyerberg et al. (2004) conducted an 8 week dietary intervention study in healthy
males. Study design was a randomised, double-blind parallel dietary intervention for 8
weeks with 12 weeks of follow-up. Diets under analysis were a trans fatty acid
enriched diet, an n-3 polyunsaturated fat enriched diet, and a control fat diet. Seventynine participants completed intervention with mean age across diet groups ranging
from 35.3 yrs to 39.2 yrs. Intakes at the end of intervention revealed the trans diet
group were consuming 6.8% of total energy from trans fatty acids vs. 0.9% in the
control group. Controls consumed higher amounts of saturated, monounsaturated and
polyunsaturated fatty acids, but total fat intake was comparable, at 33.9% in the trans
diet group vs. 34.8% in the control group. Experimental outcomes included platelet
membrane fatty acid composition as measured by gas chromatography, and heart rate
variability measured by 24 hour Holter monitor.
Not surprisingly, platelet membrane fatty acid composition was significantly higher in
t18:1, t18:2, and total trans fatty acid content after consumption of a diet high in trans
fatty acids, and interestingly t18:1 levels were still significantly higher than baseline
at follow-up 12 weeks after finishing the diet (mean (SD) platelet membrane t18:1 as
percentage of total fatty acid content 0.91(0.03) at baseline vs. 1.1 (0.04) at 12 weeks
follow-up).
With regard to effect on plasma lipids, the high trans diet group showed a decrease at
end of dietary intervention in HDL levels compared with baseline, and the change was
significantly different from the changes in HDL seen pre- vs. post-diet in the control
group.
One potential weakness of this study is that due to the randomised parallel design,
random allocation resulted in the trans diet and control diet groups differing in some
respects. While BMI, proportion of non-smokers, physical activity and fish intake
measures were similar, mean age in controls was slightly older (37.6 ± 10.6 yrs vs.
35.3 ± 10.3 yrs) and baseline lipid measures differed substantially, with controls
having higher total and LDL cholesterol, lower HDL cholesterol, and hence
significantly different LDL:HDL ratios.
Two small studies identified between 2000 and 2005 which made use of deuterium to
measure metabolism and catabolism of blood lipids are described below. Whilst small
studies, they are included in this report as they are of a very few studies which
examined potential biological mechanisms which could lead to differences in lipid
levels between diets, by examining HDL, LDL and cholesterol synthesis and
catabolism rates.
A small study conducted by Matthan et al. (2004) used a randomized dietary crossover design with three different diets followed for a period of 5 weeks each. Only
eight postmenopausal women were included in the study, which limits the extent to
which the results can be generalized to the greater population. The diets differed in
that two-thirds of the fat content was derived from either soybean oil (unsaturated fat
diet), stick margarine (hydrogenated fat diet) or butter (saturated fat diet). The dietary
composition of macronutrients was kept relatively, although not exactly, consistent across the three diets, as shown in the table below, with dietary fat providing
approximately 29% of total daily energy intake
Measurement of HDL and LDL catabolism was achieved by an experimental
protocol, carried out after the 5 week consumption of each diet, whereby a gradual
intravenous infusion over 15 hours of deuterated leucine (5,5,5-2
H3-L-leucine) was
combined with hourly ‘meals’ – these were identical small servings which in total
provided the daily caloric intake for each dietary phase. Blood samples were collected
via a second intravenous line for measurement of lipid and lipoprotein characteristics
of interest.
The hydrogenated fat diet produced levels of total and LDL cholesterol intermediary
between the other two diets, such that compared to the unsaturated fat diet, total
cholesterol was 7% higher on the hydrogenated fat diet, and 11% higher on the
saturated fat diet; likewise LDL cholesterol was 12% higher on the hydrogenated fat
diet, and 15% higher on the saturated fat diet. However a significant increase in HDL
was observed on the saturated fat diet relative to the unsaturated, which did not occur
with the hydrogenated fat diet, such that the hydrogenated fat diet showed the poorest
ratio of total cholesterol to HDL cholesterol.
The results of lipid production and degradation comparisons, as measured by the
infused deuterated leucine experiment at the end of each diet phase, showed that the
fractional catabolic rate of HDL was significantly higher on the hydrogenated fat diet
compared to the other two diets. HDL production rates did not differ between diets,
suggesting that the reduced HDL levels seen on the high trans fatty acid diet arose
from increased HDL catabolism rather than lower HDL production rates.
Regarding LDL measures, LDL production rates based on apoB-100 were similar on
all three diets, with both the saturated and hydrogenated fat diets showing a
significantly lower LDL fractional catabolic rate, thus giving rise to the higher LDL
levels seen on these two diets relative to the unsaturated fat diet. Importantly, the
saturated fat and hydrogenated fat diets were not significantly different in terms of
either LDL production or fractional catabolic rates. This suggests that saturated and trans fatty acids produce a similar effect of raising LDL cholesterol, however the
poorer lipid profile observed on the hydrogenated fat diet in this study appears to have
arisen via trans fatty acids increasing HDL catabolism.
Another small study making use of deuterium was conducted by French, Sundram and
Clandinin (2002) in 10 Malaysian women. A diet high in trans fatty acids gave
significantly greater rises in LDL cholesterol than a diet high in saturated fat (11.5%
increase over high-saturated fat diet, p < 0.05). In this study deuterium was used to
measure cholesterol fractional synthetic rate. A high trans diet produced a
significantly greater fractional synthetic rate of free cholesterol than a diet high in
saturated fat. The fractional synthetic rate of total cholesterol was also greater on a
high trans diet in 8 out of the 10 subjects than on a high saturated fat diet, but did not
reach statistical significance.
Idris and Sundram (2002) examined the effects of dietary trans fatty acids in
Cynomolgus monkeys, allowing tight control of dietary intake. Nine monkeys were
randomly rotated through four dietary regimes for 6 weeks each in two different
phases. In the first phase, different blends of oils were the main difference between
these four diets, with only one diet having significant amounts of trans fatty acids,
derived from partially hydrogenated soybean oil, which is frequently used for human
consumption. In the diet high in trans fatty acids, total trans fatty acids contributed
9.3% of total energy intake, with 7.4% from t18:1(n-9), 1.1% from t18:1(n-11), and
0.5% from t18:1(n-13). For the second phase, the same four dietary regimes were
used, with the addition of dietary cholesterol at 0.1% of total energy to each diet.
The high trans diet (without added cholesterol) produced significantly higher total
cholesterol, LDL cholesterol and significantly lower HDL cholesterol, resulting in the
poorest LDL:HDL ratio of all four diets. The addition of dietary cholesterol in the
second phase of the experiment resulted in the high trans diet producing significantly
higher total cholesterol and LDL cholesterol than without dietary cholesterol (also
seen in the other three diets) although no decrease in HDL cholesterol was seen, such
that the LDL:HDL ratio was no longer the poorest of the four diets. The authors give
no possible explanation as to why the addition of dietary cholesterol appears to have
ameliorated the HDL-lowering effects of dietary trans fatty acids. These results also
seem to suggest that, at least in non-human primates, dietary trans fatty acids are
much more atherogenic when consumed as part of a diet which is also high in
cholesterol.
Predictive modelling equations
An authoritative meta-analysis of the effects of dietary trans fatty acids was published
in 2003 in the American Journal of Clinical Nutrition (Mensink et al., 2003). As
described above under ‘Saturated fatty acids and LDL cholesterol: Predictive
modelling equations’, articles published between January 1970 and December 1998
were included, and had to meet the following criteria: food intake clearly described
with dietary fatty acids as the sole variable; dietary cholesterol constant between
comparison diets; parallel, cross-over or Latin-square design; diets adherence for at
least 13 days to allow time for changes in lipid measures to take place; adults with no
disturbances of lipid metabolism or diabetes included. A total of 60 studies were
selected for the meta-analysis, of which only 8 were included that specifically
measures trans fatty acids. Thus in essence the modelled effect of dietary trans fatty
acids was only based on these 8 studies, with models that estimated the influence of
trans fatty acids by predicting the changes in lipid measures that would occur with
replacement with polyunsaturated or monounsaturated fatty acids or carbohydrates.
The relative effects of each of these other replacement dietary components was based
on the modelling results of the full meta-analysis of 60 trials. In the 8 studies that
measured trans fatty acids intakes ranged from 0% to 10.9% of total energy.
Trans fatty acids emerged as the most detrimental of the fatty acids examined in the
study (which included an examination of the effects of the individual saturated fatty
acids). Regression coefficients (95% CI) for the effect of trans fatty acids vs.
carbohydrate of 1% of energy were 0.022 (0.0005, 0.038) for total:HDL cholesterol
ratio, 0.031 (0.020, 0.042) for total cholesterol, and 0.040 (0.020, 0.060) for LDL
cholesterol. No effect on HDL cholesterol ratio was predicted (regression coefficient
(95% CI): 0.000 (-0.007, 0.006).
It was predicted that thus replacement of trans fatty acids accounting for 1% of total
energy would decrease the total:HDL cholesterol ratio by 0.019 when replaced with
saturated fatty acids, 0.048 for cis monounsaturated fatty acids, and 0.054 for cis
polyunsaturated fatty acids.
Given these results, replacement of trans fatty acids constituting 1% of total energy
with carbohydrate would have the same effect on total:HDL cholesterol ratio as
replacement of 7.3% of saturated fatty acids with carbohydrate. The authors state that
given an ‘average US diet’ which derives 2.6% of total energy from trans fatty acids
and 13% from saturated fatty acids, replacement of the 2.6% of trans fatty acids with
carbohydrates would have a greater beneficial effect on total:HDL cholesterol ratio
than replacing the total amount of saturated fatty acids with carbohydrate. The authors
conclude that ‘the replacement of trans fatty acids with unsaturated fatty acids from
unhydrogenated oils is the single most effective measure for improving blood lipid
profiles.’
Another study which produced predictive equations on the effects of dietary trans
fatty acids on LDL cholesterol was that of Müller et al. (2001). Their impetus was that
most published studies of predictive equations for LDL cholesterol were based only
on the effects of either total saturated fatty acids, or individual saturated fatty acids,
ignoring the influence of trans fatty acids. The design of the study is included above
under ‘Saturated fatty acids and LDL cholesterol: predictive modelling equations’.
One weakness of the study was that the dataset from which the models were derived
consisted only of four studies which had been carried out from 1993 – 1998 by the
authors at the University College of Akershus in Norway.
Trans fatty acids were incorporated into predictive equations which included lauric
(C12:0), myristic (C14:0) and palmitic (C16:0) acids. Stearic acid (C18:0) was not
included in the analysis. Trans fatty acids from partially hydrogenated soybean oil
(labelled ‘TRANS V’) were distinguished from trans fatty acids from fish oil (labelled
‘TRANS F’), and regarded as consisting of the individual trans fatty acids: TRANS V:
t16:1, t18:1; TRANS F: t16:1, t18:1, t18:2, t20:1, t20:2, t22:2, t24:1.
Regression coefficients (SD) for the effect of each 1% change of total energy per
trans fatty acid on LDL cholesterol in mmol/l were 0.025 (0.042) for TRANS V, and
0.043 (0.028) for TRANS F. In the final predictive equation, comparing isoenergetic
amounts of dietary fatty acids, the saturated fatty acid myristic acid produced the
greatest rise in LDL cholesterol followed by palmitic acid. Trans fatty acids derived
from fish oil (TRANS F) closely followed palmitic, in turn closely followed by trans
fatty acids from vegetable oil (TRANS V). Similar results were obtained for the
prediction of total cholesterol. The final equation is given below.
Trans fatty acids and coronary heart disease:
i. Cross sectional studies of trans fatty acid intake and nonfatal MI
Three cross sectional studies examining the link between dietary and/or adipose tissue
trans fatty acids and nonfatal MI were identified since publication of the Health
Canada report. They will be described below in order of publication. Each of these
had similar experimental designs including subjects who had a first nonfatal MI and
data relating to trans fatty acid intake derived from food frequency or dietary
questionnaires or adipose tissue samples.
Pedersen et al. (2000) conducted a case control study of men and postmenopausal
women admitted for a first acute MI in Oslo, Norway in 1996. Patients with previous
MI or serious disease such as diabetes, which may alter their dietary patterns, or
significant recent weight loss, were excluded. Participation rates were high with only
5% of those asked to participate refusing. Controls were matched for age, sex and
geographical location. Adipose tissue samples were collected from the buttock, and
validated food frequency questionnaires completed during interviews with
participants. Power calculations were based on the EURAMIC study (Aro et al., 1995;
reviewed in the Health Canada report) and 112 cases and 107 controls were recruited.
Two patients had previously diagnosed coronary heart disease but were included as
they reported not changing their diet. Odds ratios were calculated for the highest
versus the lowest quintiles of dietary intake, with consideration of possible
confounders including age, sex, waist-to-hip ratio, current smoking, and family
history of CHD.
The correlation between adipose tissue trans fatty acid and dietary trans fatty acid
intake was found to be 0.295. Adjusted odds ratio (95% CI) for a first MI for the
highest vs. the lowest adipose tissue trans fatty acid levels were 2.25 (0.78 – 6.48),
with a significant trend across groups at p=0.03.
Levels of trans fatty acids in this study were highly correlated with levels of linoleic
and linolenic fatty acids, probably due to a common dietary source such as margarine.
Adjustment of the above odds ratios for content of either linoleic or linolenic acids in
the adipose tissue abolished the significance of the increased risk associated with
higher trans fatty acids.
A 2003 paper in the Journal of Nutrition (Baylin et al., 2003) examined the
association between trans fatty acid intake and non-fatal myocardial infarction in a
Costa Rican population. Subjects included men and women who attended hospitals in
the metropolitan area of San Jose, Costa Rica, and were diagnosed as having had an
acute MI by two independent cardiologists. Cases were excluded if they had had any
previous admission related to cardiovascular disease, or if they died during the
hospitalization arising from the MI. Five hundred and thirty cases were matched by
age, sex, and area of residence to 531 controls. Cases and controls were both visited in
their homes for collection of dietary and health information, clinical measures, and
adipose tissue biopsy. Participation rates were exceptionally high in both groups with
97% of cases and 90% of controls agreeing to participate. The association of trans
fatty acids with nonfatal MI was principally based on adipose tissue trans fatty acid,
with a biopsy from the upper buttock. Food frequency questionnaires were administered to examine possible confounding of results by other dietary factors for
which good biomarkers of intake were not available. The major source of 18:1 and
18:2 trans fatty acids in the Costa Rican diet is considered by the authors to be
derived from partially hydrogenated soybean oil, used by over 40% of subjects.
Multivariate odds ratios for nonfatal MI were determined using conditional logistic
regression including the possible confounding factors: age, sex, residence, income,
history of diabetes, history of hypertension, physical activity, smoking, years living in
the same house, adipose alpha-linolenic acid, intake of alcohol, vitamin E, saturated
fat and total energy. Further adjustments for BMI, waist-to-hip ratio, multivitamin
use, folate of fibre intake were also examined and found not to influence OR’s of any
of the associations with trans fatty acids reported.
Total adipose tissue trans fatty acid was associated with an increased risk of MI, with
the highest quintile of adipose tissue TFA exhibiting a multivariate OR (95% CI) for
MI of 2.94 (1.36 – 6.37) when compared to the lowest quintile of adipose tissue TFA.
The trend across quintiles was highly significant at p = 0.004.
When considering individual trans fatty acids the most striking effects were seen for
t18:2. An increased risk associated with t16:1 was also apparent in the highest quartile
of intake. No effect was seen with t18:1.
The same trend of increased risk of nonfatal MI with adipose tissue TFA has been
reported by Clifton et al. (2004), also published in the Journal of Nutrition. This study
had a very similar design, but with the added interesting feature of coinciding with the
withdrawal of trans fatty acids from commercially available margarine.
Cases were men and women admitted for heart disease between 1995 and 1997 at any
of the four major hospitals in Adelaide. Subjects with a previous diagnosis of heart
disease, angina or diabetes were excluded. All blood samples were taken within 4hr of
development of chest pain. Adipose tissue samples were taken from abdominal
subcutaneous fat at the umbilicus level within 2 weeks of discharge. Food frequency
and other lifestyle questionnaires were administered by a dietitian before discharge.
Statistical analysis was performed using logistic regression, including the potential
confounders: total energy intake, % energy as fat, protein, carbohydrate, saturated fat,
polyunsaturated fat, BMI, age, sex, blood pressure, lipids (however HDL was not
measured), smoking, job classification, and all other measured fatty acids.
209 cases and 174 controls were recruited. Total adipose TFA was higher in cases
than controls, a similar effect was seen for each of the trans fatty acids t18:1 (n-9),
t18:1 (n-8) and t18:1 (n-7). In logistic regression adipose tissue t18:1 (n-7) was an
independent predictor of first heart attack.*
* This interpretation and discussion of the results of Clifton et al. (2004) are based on the published
erratum in J Nutr 134:1848 (2004) which states: ‘throughout the article “trans 18:1(n-11)” should read
“trans 18:1(n-7)” and “trans 18:1(n-10)” should read “trans 18:1(n-8)”’.
Using dietary data from food frequency questionnaires, the highest quintile of TFA
intake had a risk of MI twice that of the lowest quintile, however this was abolished
after correction for saturated fat intake. Thus, adipose TFA appeared to be a more
powerful predictor than TFA from dietary intake In June 1996, a year after commencement of recruiting, food manufacturers in
Australia withdrew TFA from the production of popular margarine brands. The
authors state that prior to this, margarine provided approximately 50% of the dietary
intake of TFA. Thus an analysis pre- and post-withdrawal has interesting bearing on
the influence of TFA in the food supply.
Dietary margarine intake assessed by food frequency questionnaire did not differ
before or after this withdrawal in either cases or controls. When an analysis was done
of biopsies taken before June 1996, cases had a higher level of trans 18:1 (n-9) and
18:1 (n-7) than controls. For those samples taken after June 1996, cases and controls
did not differ in any of the adipose tissue TFA’s.
Of interest is the observation that: t16:1 fatty acid levels were found to be higher in
cases than controls, indicating a possible role of animal derived TFA in the increased
risk of MI. However, although the highest quintile of TFA intake from dietary data
had a risk of MI twice that of the lowest quintile, this group also had a higher
saturated fat intake, and after correction for saturated fat, this association disappeared.
The studies described above examining the link between adipose tissue trans fatty
acids and myocardial infarction all suggest a powerful connection between the two. It
is noteworthy that the European EURAMIC study (covered in the Health Canada
report) found no association between adipose tissue t18:1 fatty acids and MI in cases
and controls across Europe. However, two Spanish centres displayed adipose fatty
acid profiles quite different from other centres with low mean adipose t18:1 levels and
high mean adipose oleic acid levels. When these two centres were excluded, a positive
association was observed suggesting an increased risk of MI with higher adipose t18:1
levels. The authors concluded that they ‘cannot exclude the possibility that the
contribution of trans fatty acids to risk of MI is significant in countries with high
intakes of trans fatty acids’. Thus the only truly negative case control study is that by
Roberts et al. considered in the Health Canada report.
Lemaitre et al. (2002) reported a study examining the association of trans fatty acids
with myocardial infarction, though unlike the above studies, using red blood cell
membrane fatty acid content as a marker of dietary trans fatty acid intake rather than
adipose tissue content. The study was conducted in King County, Washington, US,
with a specific interest in whether trans 18:1 isomers or trans 18:2 isomers were more
strongly associated with MI. Cases were identified between October 1998 and June
1999. Response rate amongst cases was 77% and 59% for controls, giving final
samples sizes of 179 cases and 285 controls. Mean age (SD) of cases was 59.5 (10.3)
yrs and controls 57.8 (10.3) yrs, with over 80% of both groups being male.
Red blood cell (RBC) membrane fatty acid content was measured by gas
chromatography, which detected nine different fatty acids in samples: the t18:1
isomers (n-9), (n-10), (n-11), (n-12) and a mix of (n-6) to (n-8); the t18:2 isomers (cis
9, trans 12) and (trans 9, cis 12); and the t16:1 isomers (n-7) and (n-9). Of note is that
no conjugated linoleic acid (CLA) isomers were detected in membrane samples.
Dietary intakes were not measured amongst all cases or controls, but instead in a subsample a 17-item food frequency questionnaire was given to spouses of cases and
controls. The red blood cell membrane total trans fatty acid level was related to intake assessed by questionnaire amongst 111 controls with r = 0.5, p < 0.001. The different
trans 18:1 and 18:2 isomers were additionally correlated to intake of different foods.
An analysis of 4-day food records completed by 55 controls showed no correlation of
RBC membrane trans fatty acids and total energy intake or percentage of energy from
saturated fat.
Odds ratios were calculated based on increases in RBC membrane trans fatty acid
content equal to the interquartile range of controls. After adjustment for traditional
risk factors, an increase of one interquartile range of membrane t18:1 was
significantly associated with increased risk of MI, although membrane trans fatty acid
content was found to be inversely related to membrane content of the long-chain n-3
polyunsaturated fatty acids docosahexaenoic acid (DHA) and eicosapentaenoic acid
(EPA), which was also different between cases and controls. After further adjustment
for membrane DHA and EPA levels, the association of membrane t18:1 with MI was
no longer significant. However the corresponding OR (95% CI) after the same
adjustments for an interquartile range increase in membrane t18:2 was highly
significant at 2.66 (1.58 – 4.46). In a model assessing risk of MI which included both
membrane t18:1 and t18:2, only t18:2 independently emerged as significantly
associated with MI.
Analysis across quartiles of membrane t18:2 was also suggestive of a linear doseresponse relationship, with OR (95% CI) of MI for increasing quartiles compared with
the lower quartile of 1.41 (0.66 – 3.00), 2.39 (1.07 – 5.35) and 4.22 (1.65 – 10.8).
i.b. Methodological considerations:
Adipose tissue TFA and dietary intake:
In order for a health claim regarding dietary trans fatty acids and CHD to be made on
the basis of the studies reported above, which have derived associations using adipose
tissue trans fatty acids, it is important to demonstrate that adipose tissue trans fatty
acids are a reliable marker of dietary trans fatty acids, and that dietary trans fatty
acids are incorporated into adipose tissue.
A pioneering study by Schrock and Connor in 1975 demonstrated a high level of
incorporation of dietary 18:1 trans fatty acids into adipose tissue in New Zealand
white rabbits, albeit at a particularly high trans fatty acid intake (Schrock and Connor,
1975).
A similar study into the effect of controlled dietary fatty acid variation and adipose
tissue composition in humans wasn’t conducted until 2002, when Baylin et al.
provided a comprehensive validation of adipose tissue fatty acids as biomarkers of
dietary trans fatty acid intake (Baylin et al., 2002). These researchers analyzed the
respective adipose and dietary correlations of 35 detectable fatty acids using adipose
samples from the upper buttock. Trans fatty acids in general were the most highly
correlated fatty acid group. Aside from t14:1(n-5) and t16:1(n-7) which had relatively
low correlations of 0.09 and 0.12 respectively, t18:1 isomers had correlations between
0.30 and 0.43, with total t18:1 having a pooled correlation coefficient of 0.43. Trans
18:2 isomers had higher correlations between 0.53 and 0.61. The only other fatty
acids with comparable correlation coefficients were linoleic (18:2(n-6)) at 0.58 and
alpha-linolenic (18:3(n-3)) at 0.34.
In this study sample, total trans fatty acids accounted for 3.1% of total adipose fatty
acids, and trans fatty acid contributed 4.8% of total dietary fat and 4.0 g/d of energyadjusted dietary intake; these levels are comparable with dietary intake reported by a
number of studies worldwide, suggesting these findings could be widely generalized
to other populations.
In epidemiological studies such as those above which have measured both dietary
intake and adipose tissue trans fatty acid levels, correlations between the two have
varied across studies. In the Nurses’ Health Survey, two such comparisons between
adipose trans fatty acid content and trans fatty acid intake estimated from a FFQ were
undertaken, producing correlations of 0.50 and 0.41 (Garland et al., 1998). Cantwell
(2000) provides a review of studies which have correlated adipose tissue trans fatty
acids with dietary assessment, with correlation coefficients ranging from 0.29 (Hunter
et al., 1992) to 0.51 (London et al., 1991). Lemaitre et al. (1998) examined
correlations between adipose and dietary trans fatty acids, finding that after
adjustment for energy intake, age and BMI, correlation coefficients for trans fatty
acid isomers were generally higher amongst men. For total trans fatty acids, the
correlation coefficient for women (n=27) was 0.52, and for men (n=24) 0.76. Trans
16:1 fatty acids gave the lowest correlation of the trans fatty acids for both sexes at
0.30 for women and 0.42 for men.
Site of adipose tissue collection: abdominal vs. buttock adipose tissue In the three studies summarised above linking adipose tissue trans fatty acids and
nonfatal MI published since 1999, Clifton et al. (2004) used abdominal adipose
samples whilst both Pedersen et al. (2000) and Baylin et al. (2003) used buttock
adipose samples. As shown by Baylin et al. (2002), buttock adipose tissue trans fatty
acids are highly correlated with dietary intake. While one can assume abdominal
adipose composition would not differ appreciably, it would useful be to have data to
quantify this. The only study identified by the authors of this report to compare
abdominal vs. buttock adipose tissue content is that of Mamalakis et al. (2002).
Unfortunately this study did not include dietary intake data to examine whether
dietary trans fatty acids are preferentially deposited into abdominal or buttock adipose
tissue. Moreover, this study was conducted in school aged children (mean age 13 yrs)
and thus the findings are of limited relevance to the link between adipose tissue trans
fatty acids and MI in adulthood, as the effect of changes in circulating hormones postpuberty on adipose tissue composition is unknown. Briefly, the study found that
abdominal trans fatty acids were significantly higher than buttock trans fatty acids in
girls, but not boys, and also in the pooled gender analysis. Although girls and boys
had the same abdominal trans fatty acid content, girls had significantly lower buttock
trans fatty acid content than boys.
Red blood cell (erythrocyte) membrane fatty acid composition as a biomarker of TFA
intake
In the study by Lemaitre et al. (2002) the intake of dietary trans fatty acids measured
by FFQ in 111 controls gave a good correlation with red blood cell membrane trans
fatty acid content (r = 0.5, p< 0.001), suggesting red blood cell membrane content is a
valid biomarker of trans fatty acid intake. The results of other studies are supportive
of this. Barnard et al. (1990) showed dietary trans fatty acids were incorporated into
red blood cell membranes in monkeys. Brosche et al. (1986) showed changes in red
blood cell membrane trans fatty acid content mirroring changes in dietary intake in
elderly subjects. Emken et al. (1979) reported uptake of over 24hrs of deuterium
labelled trans fatty acids into red blood cell membranes in young men, and
additionally noted that uptake into red blood cell membranes was less, and slower,
than uptake into plasma phospholipids.
ii. Prospective studies of trans fatty acid intake and coronary heart disease
An important prospective study released since the Health Canada report is that of the
Zutphen Elderly Study (Oomen et al., Lancet 2001). In this Dutch study, a significant
source of dietary trans fatty acid was partly hydrogenated fish oil, in contrast to the
partially hydrogenated vegetable oils more frequently used in the US (Hulshof et al.,
1999).
The Zutphen cohort was established in 1960 with 878 men born between 1900 and
1919 and formed the Dutch component of the Seven Countries Study. Dietary surveys
were conducted in 1985, 1990 and 1995. Clearly an appreciable number of the
original cohort had died by this time, and further exclusion of those with diagnosed
MI or angina resulted in dietary information being available for 667 men in 1985, 435
in 1990, and 225 in 1995. Trans fatty acid intake was determined with the use of timespecific tables of Dutch foods. National data on composition of edible fats was
available for 1985 and 1990 from the Wageningen University in the Netherlands, and
for 1995 from the large European TRANSFAIR study. Vital status was determined
from municipal registries and cause of death from either Statistics Netherlands,
hospital discharge data, or general practitioners. During the 10 year follow up period,
98 cases of coronary heart disease were identified, with 49 of these being cardiac
deaths (46 primary cause of death; 3 secondary cause of death).
Trans fatty acid intake in this cohort was higher than reported for other studies
conducted internationally, with total trans fatty acid intake in 1985 accounting for
4.3% of total energy, and falling to 1.9% by 1995.
Cox’s proportional hazard analysis was used to calculate relative risks, with a fullyadjusted analysis taking into account age, energy intake, BMI, smoking, alcohol
intake as a categorical variable, vitamin supplements, intake of saturated fatty acids,
monounsaturated fatty acids, polyunsaturated fatty acids, cholesterol and fibre.
Fully-adjusted relative risk (95% CI) for coronary heart disease amongst the highest
tertile of trans fatty acid intake was 2.00 (2.07 – 3.75) when compared to the lowest
tertile, with a significant trend across tertiles (p=0.03).
An analysis was also conducted examining the effect of a difference at baseline of 2%
of energy from total trans fatty acids. For a diagnosis of coronary heart disease a fully
adjusted RR (95% CI) of 1.28 (1.01 – 1.61) was found, and for fatal coronary heart
disease 1.33 (0.96 – 1.86).
When considering trans fatty acids from different sources, the relative risks (95% CI)
of coronary heart disease for a difference at baseline of 0.5% of total energy from
either ruminant trans fatty acids, manufactured t18:1 acids, and other manufactured
trans fatty acids were similar at: 1.17 (0.69 – 1.98), 1.05 (0.94 – 1.17), and 1.07 (0.99
– 1.15) respectively.
The authors conducted a pooled analysis of their results combined with three previous
prospective studies – the Nurses’ Health Study 14 year follow-up (Hu et al., 1997),
the Health Professionals follow-up study (Ascherio et al., 1996), and the AlphaTocopherol Beta-Carotene Cancer Prevention Study (Pietinen et al., 1997) all of which measured dietary trans fatty acid intake and incidence of CHD, and which are
each covered in the Health Canada report. The pooled fully adjusted relative risk
(95% CI) of coronary heart disease based on an increase of 2% of energy in trans
fatty acids at baseline was 1.25 (1.11-1.40).
Oh et al. (2005) reported on the effects of dietary fat intake after 20 years of follow-up
in the Nurses’ Health Study (methods detailed above in ‘Saturated fatty acids and
coronary heart disease: prospective studies’). Trans fatty acid intake decreased
slightly over the follow-up period from 2.2% of total energy in 1980 to 1.6% in 1998.
A comprehensive multivariate relative risk calculation (also detailed above) was made
between trans fatty acid intake quintiles and incident CHD. Relative risk (95% CI) of
incident CHD for the highest quintile of trans fatty acid intake (median 2.8% of total
energy) vs. the lowest quintile (median 1.3% of total energy) was 1.33 (1.07 – 1.66).
The fourth highest quintile (2.2% of total energy) was of borderline significance at a
RR (95% CI) of 1.19 (0.99 – 1.44), and a highly significant trend across quintiles
existed at p=0.01. Risk of CHD differed by age and BMI group, with women younger
than 65 yrs and women of BMI < 25 kg/m2
at greater risk of CHD, although in both
cases the interaction of CHD risk and age/BMI was not significant.
Part 3. Relevance to Australia and New Zealand
Information on population intakes of saturated fat, and food sources of saturated fat
such as meat and dairy products, is readily available in both countries from National
Nutrition Surveys conducted at fairly regular intervals. Information on trans fatty acid
intakes is sparse. Whilst the respective National Nutrition Surveys of Australia and
New Zealand include data on the intakes of industrially produced foodstuffs, such as
margarines, and meat from ruminant animals, which are common sources of dietary
trans fatty acids, there have been few attempts to derive estimated intakes of trans
fatty acids from these. Given this paucity of population-level data, other published
measures of trans fatty acid intake in the Australasian population measured in specific
groups as part of scientific studies are also considered below.
Data on saturated fat intake from the most recent Australian Nutrition Survey
(McLennan and Podger, 1995a + b), which is now a decade old, indicated a mean
saturated fat intake for males aged 19 yrs and over of 39.0 g/day or 12.7 % total
energy, and 26.7 g/day or 12.7% total energy for females aged 19 yrs and over. Total
fat intake contributed approximately a third of total energy across all ages. Regarding
food sources of dietary saturated fat, milk and milk products were the largest
contributors to dietary saturated fat intake accounting for 26.7% and 27.2% of
saturated fat intake for males and females respectively. Meat (including poultry) was
the second highest contributor at 22.5% and 18.7% of saturated fat intake for males
and females respectively. Fats and oils, potentially also a large source of dietary trans
fatty acids, accounted for 8.9% of saturated fat intake for both genders. The values
quoted here are for persons aged 19 yrs and over.
Data from the latest New Zealand Nutrition Survey regarding saturated fat intake are
also somewhat out of date, the survey having been conducted in 1996/97 (Russell et
al., 1999). Mean intake of saturated fat was 15% of total energy for both males and
females and similar across ages. Absolute intake of saturated fat was higher for men
given their higher total energy intake. Proportion of energy from total fat intake was
higher for Maori females (36% total energy) compared with females of NZ European
and other ethnicities (34% total energy). Mean saturated fat intake was also higher for
Maori women (39 g/day, 16.0% total energy) than NZ European women and women
from other ethnicities (31 g/day, 14.7% total energy). Likewise Maori men had a
higher proportion of total energy derived from total fat (37% total energy) than men
of NZ European and other ethnicities (35% total energy), however proportion from
saturated fat was comparable across ethnicities, at 52 g/day or 15.7% total energy for
Maori men compared with 49 g/day or 15.2% total energy for men of NZ European or
other ethnicities. Median intake of total fat or saturated fat did not appear to be related
to NZ deprivation index score. Saturated fat intake data for Maori should be
interpreted with caution due to the smaller number of Maori studied. With regard to
comparisons over time, total fat intake had decreased from the previous Life In New
Zealand survey (Russell and Wilson, 1991) conducted in 1989, from 37.5% to 34.9%
of total energy. However comparison between the surveys should be interpreted with
caution because of slight differences in methodologies.
Thus it appears that amongst the populations of Australia and New Zealand there
exists a fairly substantial potential for reduction of saturated fat intake.
The study by Clifton et al. (2004) described in Part 2 under the heading ‘Trans fatty
acids and coronary heart disease: cross-sectional studies’ was conducted in Adelaide
between 1995 and 1997. The authors report that the trans fatty acid content of many
commercially produced margarines dropped in June 1996. Given the importance of
margarines as a source of trans fatty acids, intakes are likely to be lower after this
time than before. However estimated dietary intakes of TFA for participants enlisted
pre- and post-withdrawal are not presented separately but pooled. It appears that twothirds of participants were enlisted after June 1996. Total mean (SD) TFA intake is
given as 3.52 (1.82) g/d or 1.2 % of total energy for cases and 3.01 (1.29) g/d or 1.1 %
total energy for controls. Approximately 10% of trans fatty acids were derived from
animal sources for both cases and controls, approximately 23% from dairy sources,
and the remaining majority from margarines and mixed dishes. Given that the data
from some of these subjects was collected prior to June 1996, it is reasonable to
believe that actual intakes post-June 1996 were slightly lower than the total intakes
reported here. Mansour et al. (2001) also conducted a very small study in 10
volunteers from Geelong, Australia, in which usual mean (SD) intake of trans fatty
acids was 4.4 (0.9) g/day or 1.9 % total energy. Of course, the data from both these
studies are now outdated.
The paucity of reliable estimates of population intakes of trans fatty acids in
Australasia has been raised before by Food Standards Australia New Zealand
(Landells, 2005). Apart from the estimates of Clifton et al. and Mansour et al. the
most recent estimates on intakes in Australia were noted to have been published in
1994 from simulated Australian diets or based on food frequency questionnaire data
collected in 1987, and for New Zealand from the 1989/1990 Life In New Zealand
(LINZ) survey. Mean trans fatty acid intakes from simulated Australian diets were
calculated to range from 6.4 g/day or 2.5 % total energy for males and 4.4 g/day or 2.1
% total energy for females, to 13.6 g/day or 5.3 % total energy for males and 10.5
g/day or 5.1 % total energy for females, based on different assumptions about the
trans fatty acid content of available foods. The authors conclude that intakes ‘are
likely to be less than 2 – 2.5 % energy’ (Noakes and Nestel, 1994). Data from the
LINZ survey suggest intakes in New Zealand to be slightly less at 5.4 g/day for males
and 3.4 g/day for females. Based on the mean total energy intakes published from the
LINZ survey (Wilson et al., 1990) of 10,400 kJ for men and 6,800 kJ for women,
these figures equate to 1.9 % total energy for both sexes. Thus although current
intakes may not be excessive it is necessary to ensure that levels do not increase and
indeed benefit may accrue from reduction to even lower levels.
Another important factor to consider in the case of Australia and New Zealand is the
multi-cultural nature of their populations. A great degree of dietary variation between
ethnic groups has been documented in both countries (Ferguson, 2002; Harris et al.,
Where intakes of saturated or trans unsaturated fatty acids as percentages of total energy
intake have not been given in the original papers or reports described above, they have been
converted from intakes in grams per day using the conversion factors of 1g trans fatty acids =
9 kcal or 17 kJ, based on the conversion factors for fat reported in McCance and
Widdowson’s Composition of Foods (Food Standards Agency, 2002).
2004; Metcalfe et al., 1998; Russell et al., 1999; Swinburn et al., 1998; Thomson and
Shaw, 2002). Given the high coronary heart disease rates amongst the indigenous
populations and some more recent migrant groups, higher intakes of saturated fatty acids and possibly trans fatty acids may represent an even greater cause for concern
than is the case for European populations.
Part 4. Other relevant biomarkers of disease outcome
Nutritional factors including saturated fatty acids and trans unsaturated fatty acids
have been related to several biomarkers for coronary heart disease other than
lipoproteins.
Perhaps one of the most striking associations is the “convincing” relationship between
saturated fatty acids and insulin resistance. Insulin resistance is a key abnormality
underlying the metabolic syndrome and type 2 diabetes, critically important
determinants of cardiovascular disease, and has been shown in prospective studies to
predict coronary heart disease (Zethelius et al., 2005). Carefully controlled nutritional
intervention studies have clearly shown the ability of saturated fatty acids (when
compared with the effect of monounsaturated fatty acids) to increase insulin
resistance, measured by the intravenous glucose tolerance test (KANWU study,
Vessby et al., 2001). There is no doubt about the relationships between markers of
inflammation (e.g. CRP, TNF-α, interleukin 6) and coronary heart disease (Wilson et
al., 2004; Pai et al., 2004). There is a reasonable body of evidence derived from crosssectional studies which links dietary saturated fat and CRP (King et al., 2003), though
the findings are not entirely consistent (Fredrikson et al., 2004). One intervention trial
carried out in Canada has shown the potential of dietary modification to reduce CRP
levels (Jenkins et al., 2003). However in addition to substantial reduction of dietary
saturated fatty acids, the trial involved substantial increases in dietary fibre and
changes in other fatty acids. Thus evidence for the association between saturated fatty
acids and inflammatory markers, while of considerable interest, is for the present most
appropriately regarded as “possible”.
Trans unsaturated fatty acids have also been linked with markers of inflammation in
cross-sectional analyses in participants in the Nurses’ Health Study (Mozaffarian et
al., 2004). Intakes of trans 18:1 and trans 18:2 (but not trans 16:1) were positively
correlated with TNF-α. The associations of TFA consumption with coronary artery
disease risk are greater than would be predicted by effects of serum lipoproteins
alone. In this cross-sectional study relations between TFA intake and TNF receptor
concentrations were partly attenuated by adjustment for serum lipid concentration
suggesting a further independent effect mediated by promotion of the inflammatory
response.
In a further analysis based on the same study, Lopez-Garcia et al. (2005) additionally
found associations between TFA intakes and CRP as well as E-selectin and soluble
cell adhesion molecules (sICAM and sVCAM). Thus while there appears to be
evidence for a link between trans fatty acids and markers of endothelial function it is
insufficient to form conclusions.
Obesity and increased tendency to thrombosis are unquestionably associated with
increased risk of cardiovascular disease. High intakes of saturated fatty acids have
been linked with both of these states, but it has not been possible to disentangle the
effects of saturated fat from total fat, once again leaving the evidence classified as
insufficient.
Part 5. Overall conclusions
a) Public summary
Coronary heart disease (CHD) mortality rates have fallen during the last several
decades, but it remains a major cause of serious illness and death in adults in Australia
and New Zealand. The underlying pathology in most cases is atherosclerosis which
involves an accumulation of lipoproteins, platelets, monocytes, endothelial cells, and
smooth muscle cells in the walls of arteries, following damage to the layer of cells
lining the artery. Atherosclerosis results in narrowing of the arteries and consequently
reduction in the blood supply to heart muscle. A clot or thrombus may be
superimposed on the atherosclerotic lesions, leading to a total obstruction to the blood
supply and consequently death of the section of heart muscle supplied by the artery.
This process leads to coronary thrombosis or myocardial infarction, whereas reduction
of the blood supply leads to angina. The pathology is believed to result from an
interaction between genetic and environmental factors. However it is noteworthy that
cholesterol derived principally from low density lipoproteins is an important
constituent of the atherosclerotic plaque, and total and LDL cholesterol are the most
clearly established of the many potentially modifiable risk factors for CHD. LDL
cholesterol is the major contributor to total blood cholesterol, hence total cholesterol
is often used as a surrogate for LDL cholesterol since it is more easily measured. A
clear dose response effect is apparent for total cholesterol in prospective
epidemiological studies examining determinants of CHD, and CHD is extremely
uncommon in populations with low mean cholesterol levels. Randomised controlled
clinical trials have shown a benefit in terms of CHD risk reduction that is proportional
to the reduction in cholesterol levels, regardless of whether this is achieved by drug
therapy or dietary modification. Thus, LDL cholesterol is the most convincing of the
biomarkers for CHD and it is generally accepted that measures able to reduce LDL
cholesterol will reduce CHD risk.
The starting point for the present review was a similar process undertaken in Canada
in 2000 by Ratnayake and McDonald. The report updated a 1993 United States report
and reaffirmed the observation, first made in the 1950’s, that saturated fatty acids
were important determinants of total and LDL cholesterol. The association can
unquestionably be described as ‘convincing’, with much evidence derived from
randomised controlled trials and a clear dose response effect apparent with increasing
amounts of SFA. The report indicated the differential effect of the different dietary
saturated fatty acids on lipoproteins, with stearic acid having a negligible effect on
LDL compared with the cholesterol raising effect of lauric, myristic and palmitic
acids. However given the fact that stearic acid often co-exists with the other saturated
fatty acids, and that stearic acid may promote thrombogenesis and so enhance
atherosclerotic risks, this does not detract from the overall association and certainly
not from the benefits in terms of LDL cholesterol of lowering total intake of saturated
fatty acids. The Canadian review did not consider in detail the appreciable variation in
individual response to a reduction in saturated fatty acids, though the population risk
reduction which would be expected to accrue from reduction in saturated fatty acid
intake should not be underestimated. Finally, it should be noted that the extent of LDL
cholesterol reduction achieved by lowering intake of saturated fatty acids is dependent
upon the source of replacement energy. Replacing saturated fatty acids with polyunsaturated fatty acids would result in appreciably greater reductions in LDL
cholesterol than replacement with either carbohydrate or monounsaturated fatty acids.
Not replacing a reduction in saturated fatty acids, partially or totally, and resultant
weight loss would also result in additional reduction of LDL cholesterol.
There is rather less direct evidence for the association between saturated fatty acids
and CHD. While the studies generally suggest a relationship between saturated fatty
acids and coronary heart disease and while there are certainly several plausible
hypotheses, there are some inconsistencies in the data which cannot all be easily
explained. Thus while we believe a reduction in saturated fatty acids is highly likely
to reduce not only LDL cholesterol and other coronary heart disease risk factors but
also cardiovascular disease, the current evidence for the direct association between
saturated fatty acids and coronary heart disease is arguably more appropriately
described as “probable” rather than “convincing”.
Far fewer data exist for trans fatty acids. However a series of well conducted studies
including randomised controlled trials show an association between trans fatty acids
and LDL cholesterol. There are however two major limitations regarding the studies
covered by the Canadian review and more recent data. Many of the studies do not
distinguish between animal (largely occurring naturally) sources of trans fatty acids
and vegetable sources, largely produced by the hydrogenation of vegetable derived
oils. Furthermore it is not clear whether the effect of trans fatty acids on LDL
cholesterol is biologically meaningful at low levels of intakes, such as that likely to be
found in Australia and New Zealand. These limitations do not preclude the conclusion
that the association between trans fatty acids and LDL cholesterol is a “convincing”
one.
Fewer data exist relating trans fatty acids directly to coronary heart disease than is the
case for saturated fatty acids. However, there are also fewer inconsistencies than is the
case with studies linking saturated fatty acids and coronary heart disease. While one
case control study found a lack of association between trans fatty acids and sudden
cardiac death, three recent case control studies are confirmatory of an association, and
another suggestive of a trend towards a positive association. Despite the apparent
strength of evidence some inconsistencies remain, and we therefore believe it may be
more appropriate to describe the association between trans fatty acids and coronary
heart disease as “probable” rather than “convincing”.
