Consultation on the health implications of alternatives to trans fatty acids
Summary of Responses from Experts
#1 While in North America LDL-cholesterol has been identified as the major atherogenic lipoprotein and therefore, as the primary target for cholesterol-lowering therapy, the impact of dietary fat and fatty acids on CHD risk has been assessed in relation to different bio-markers of atherosclerosis. How should the Task Force consider the relative importance of these bio-markers?
The question of which bio-marker is the best for assessing coronary heart disease (CHD) risk is important and is a question for which we do not have a definitive answer. High LDL cholesterol is the risk factor with the most extensive supporting evidence. The fact that the Adult Treatment Panel (ATP) III (NCEP, Circulation 2002; 106:3124-3424) designated LDL as the primary target for cholesterol-lowering therapy reinforced the dominance of LDL cholesterol level as the primary risk factor in clinical practice. However, there also is strong evidence for a causal relationship between HDL cholesterol and atherosclerosis. Any foodstuff that results in a lower HDL cholesterol level will probably increase the risk of heart disease, stroke and peripheral vascular disease. However, the latter is not 100 percent certain. It is perhaps noteworthy that ATP III (NCEP, Circulation 2002; 106:3124-3424) identified low HDL cholesterol (< 0.9 mmol/L) as an independent risk factor for CHD.
In populations with high LDL cholesterol levels, total cholesterol/HDL (or LDL/HDL) ratio seems to have the highest correlation with risk for atherosclerosis. This position is supported by Shai et al (2004), who reported that HDL cholesterol related ratios, such as total cholesterol/HDL cholesterol (TC/HDL-C), are a "powerful predictive tool independent of other known CHD risk factors". The importance of TC/HDL-C as a predictor of CHD risk also was supported by Blake and Ridker (J of Intern Med 2002;252:283-294).
While the TC/HDL-C ratio is one of the most commonly accepted risk factors by researchers, it does not predict all of the risk associated with CHD. Hence, other risk factors such as total cholesterol, LDL cholesterol, HDL cholesterol, triacylglycerides and other lipid markers, as well as non-lipid risk factors such as markers of inflammation (e.g. C Reactive Protein or CRP), lipid oxidation, endothelial dysfunction, and platelet function/clotting factors, are also important but the appropriate application of these biomarkers is uncertain. It is perhaps worth noting that an assessment of systemic inflammation markers (Blade and Ridker, J of Intern Med 2002; 252:283-294.) found CRP to be the strongest univariate predictor of the risk of CHD events, among the 12 markers measured. By contrast, based on their own results and a recent meta-analysis, Danesk et al (NEJM 2004; 350:1387-97) concluded that CRP was a relatively moderate predictor of risk of CHD and that the recommended use of CRP as a risk factor be reviewed. Nonetheless, Blake and Ridker further found that after adjusting for traditional risk factors, CRP and TC/HDL-C remained the only significant predictors of future cardiovascular events. Shai et al (Circulation 2004; 110:2824-2830) also concluded that the TC/HDL-C ratio appears to be the primary lipid predictor among postmenopausal women and that the TC/HDL C ratio as a single parameter is a powerful tool for clinical practice. On the other hand, high triacylglyceride levels together with low HDL cholesterol levels are part of what's known as the "metabolic syndrome" which is associated with an increased risk of CHD.
In fact, due to the complexity of risk factors, sometimes epidemiological studies are also considered to provide evidence of CHD risk. A recent epidemiological study (Oh et al, Am J Epidem 2005; 161:672-679) showed that dietary trans fatty acids were associated with an increase and polyunsaturated fatty acids with a decrease in mortality from CHD. Studies such as this one, which measure death from CHD as an endpoint, presumably include all risk factors. However, it is important to note that although epidemiological research provides valuable information about disease distribution and determinants of disease, it does not establish cause and effect.
In summary, all the invited experts agreed that there is enough evidence to consider TC/HDL-C as the preferred bio-marker for CHD. CRP might be a stronger biomarker, however, at present there is lack of data on the effects of dietary fats on plasma levels of CRP.
#2 Would the replacement of partially hydrogenated oils by oils rich in monounsaturated fatty acids have positive effects on serum cholesterol and lipoprotein levels and CHD risk?
There is a general consensus that the replacement of partially hydrogenated oils (including both trans and saturated fatty acids) with cis-MUFA would have positive effects on lipoproteins and CHD risk. The risk reduction depends on baseline saturated and trans fatty acid intakes.
Polyunsaturated fatty acids (PUFA) are also important components of a cholesterol-lowering and a healthy diet. Hu et al (NEJM 1997; 337:1491-1499) concluded that replacement of the SFA and TFA by MUFA and PUFA is more effective in preventing CHD than reducing overall fat intake.
Therefore, food chosen to replace trans fatty acids (TFA) and saturated fatty acids (SFA) in the diet should include a balance of unsaturated fatty acids by including both MUFA and an optimal amount of PUFA .
#3 Although the bulk of the scientific literature relating dietary fat to CHD has dealt with the risk factors surrounding atherosclerosis (viz., blood lipid levels and patterns), there is evidence implicating it in other aspects of the disease (e.g., thrombus/clot formation; cardiac arrhythmia; and in vivo oxidative stress). In addition, dietary fat has been implicated in other chronic diseases, such as cancer, diabetes and hypertension. How and to what extent should these relationships be taken into consideration in recommendations to eliminate or reduce trans fats?
Some of the most convincing evidence of a relationship between dietary fat and non-lipid CHD risk factors has been found for biomarkers of systemic inflammation and endothelial function. This evidence has come from both epidemiological and experimental studies. A limited number of studies suggest that TFA might exert adverse effect on markers of inflammation and endothelial function.
In a study by Baer et al (2004) stearic acid has also been found to result in higher fibrinogen levels whereas trans fat, oleate, and a mixture of lauric, myristic and palmitic acids had no effect on fibrinogen levels. However, this was observed at very high intakes of stearic acid (10.9% of energy), which represents a level that is unlikely to be reached by use of fats and oils high in stearic acid. In contrast, a study by Kelly et al. (Euro J Clin Nutr 2001;55, 88-96) showed that diets enriched in stearic acid did not contribute to an increase in classical risk factors for CHD and those related to thrombosis. Their results indicated that stearic acid, even at an exceptionally high intake of 19 g/day, compared to palmitic acid (22.5 g/day) had beneficial effects on thrombogenic and atherogenic risk factors. In a subsequent study, Kelly et al (Euro J Clin Nutr 2002;56:490-99) found, with the exception of a significant decrease (P<0.05) in LDL and ADP-induced platelet aggregation, there were no significant differences between high stearic acid and high palmitic acid diets on outcomes measured.
Although there is increasing interest in the relationship between dietary fat and non-lipid CHD factors, because lipid and lipoprotein risk factors do not account for the total prevalence of CHD, there are insufficient data to establish a meaningful relationship between type and amount of dietary fat and non-lipid CHD risk factors. The one exception may be the adverse effect of trans fatty acids on surrogate measures of systemic inflammation and endothelial function.
There is also growing evidence of an adverse effect of TFA intake on the risk of diabetes. The relative risk of type 2 diabetes is increased in the highest quintiles for TFA intake (Hu et al, 2001). A high intake of trans MUFA (20% of energy intake) adversely affected fasting concentrations of glucose and insulin in obese patients with type 2 diabetes (Christiansen et al. 1997).
Evidence for the possible relationship of TFA intake with cancer is however inconsistent.
#4 In Canada, it is estimated that, while meeting the essential fatty acids requirement, the ratio of linoleic acid to alpha-linolenic acid is relatively high? Many of the proposed alternatives to partially hydrogenated oils have low alpha-linolenic content. How should the Task Force consider a further increase of this ratio?
According to the invited experts who addressed this question, it is important to ensure that any changes to the diet to include less trans fats would not lead to decreases in the n-3 polyunsaturated fatty acid intake. Adequate intakes of n-3 polyunsaturated fatty acids, including EPA and DHA should be maintained. At present, regular canola oil is the primary source of n-3 polyunsaturated fatty acid in the Canadian diet and therefore, it is important not to reduce the consumption of this oil. The Expert Committee on Fats and Oils (ECFOL) beleives that while changes to the diet to reduce trans fats will most likely not alter current intakes of n-3 polyunsaturated fatty acid, they may increase n-6 polyunsaturated fatty acids and , the linoleic acid to alpha-linolenic acid ratio. High dietary content of n-6 polyunsaturated fatty acids is not desirable because n-6 fatty acids (i.e., linoleic acid) can interefere the metabolic conversion of alpha-linolenic acid to EPA and DHA and thereby decrease the tissue amount of these two long-chain omega-3 fatty acids. High tissue content of EPA and DHA exerts cardioprotective effects in patients with preexisting coronary heart disease and in healthy individuals. Thus, the use of oils high in cis-monounsaturated fatty acids rather than n-6 polyunsaturated fatty acids should be considered.
#5 There is a general consensus that trans fat has a more deleterious effect on risk factors for CHD than saturated fat. However, intakes of saturates are marginally high in Canada (about 11 % of total energy according to the Federal-Provincial surveys on nutrition collected between 1990 and 1999) and there is some evidence (Lichtenstein et al, 1999) that the adverse effect of trans fatty acids relative to saturated fatty acids may occur primarily at high dietary intakes of trans fat. Are there any conditions under which the replacement of trans fatty acids with saturated fatty acids could be considered? If yes, is there a threshold below which this replacement would not provide obvious benefits?
All of the invited experts agreed that the Task Force should promote all actions that would lower trans fat intake. Favourable health effects are clearly achieved when trans fat is replaced with cis-monounsaturated and cis-polyunsaturated fats. Some food manufacturers have already succeeded in replacing most trans fat with cis-monounsaturated fat in certain food categories. The switch from partially hydrogenated frying oils to frying oils high in cis-monounsaturated fat and low in saturated and trans fat in Europe shows that replacing trans fat in fast food, spreads and cooking fat is not problematic.
The primary product category that may require hard fat is baked goods (but not in all baked goods). In this category, the only viable alternative appears to be fat and oil containing a significant proportion of saturated fatty acids. However, the replacement should not lead to high intakes of saturated fat, because there is evidence from both clinical epidemiological studies that saturated fat (at least from dairy and meat) increase the risk of heart disease1.
The question of whether there is a threshold below which the replacement of trans fatty acids with saturated fatty acids would not actually be beneficial does not appear to have been systematically investigated. However, prospective cohort studies and metabolic studies may provide some relevant insight.
In reviewing the information from such studies, it must be noted that prospective cohort studies and metabolic studies tend to have complementary strengths and weaknesses. Generally speaking, prospective cohort studies have higher levels of external validity or generalizability, while metabolic studies have higher levels of internal validity. Specific strengths of prospective cohort studies when used for examining diet-disease relationships include their large sample sizes, with subjects followed over time as they self-select food intake and CHD events occur. Balancing this strength is the fact that because nutrient intakes are estimated from self-report questionnaires and nutrient databases with varying amounts of missing data, there can be considerable measurement error associated with estimates of actual levels of nutrient intake. In contrast, metabolic studies in which all food consumed is provided to study subjects (sometimes with its nutrient content determined by analysis of aliquot portions) allow for estimation of actual nutrient intake with a high level of certainty. Weaknesses of metabolic studies include small samples sizes and short periods of follow up. Study subjects have no opportunity for self-selection of their food intake and study end points are usually biomarkers of risk for CHD development (e.g. LDL/HDL cholesterol ratio), not CHD itself.
Prospective cohort studies
Several large prospective cohort studies have studied relationships between dietary intakes of saturated and trans fat and coronary heart disease (CHD)2. One such study, the Nurses' Health Study, examined these relationships at 8, 14 and 20 years of follow-up, in 85,095, 80,082 and 78,778 women, respectively. Results described below are from multivariate analyses at 14 and 20 year follow-ups.
At 14 years of follow-up, Hu et al (NEJM 337;1997:1491-1499) reported that participants in the highest quintile for trans fat intake were at 27% (95% CI: 3% - 56%) increased risk for CHD (defined as CHD-related morbidity or mortality) when compared with individuals in the lowest intake quintile. In contrast, participants in the highest quintile for saturated fat intake were not at a significantly increased risk for CHD when compared with individuals in the lowest intake quintile (RR=1.16; 95% CI: 0.93-1.44). At 20 years of follow up, Oh et al (AMJ 161;2005:672-679) reported that women in the highest quintile for trans fat intake were at 33% (95% CI: 7% - 66%) increased risk for CHD when compared with individuals in the lowest quintile. Participants in the highest quintile for saturated fat intake were not at a significantly higher risk for CHD than individuals in the lowest quintile (RR=0.97; 95% CI: 0.73-1.27). The findings from these two follow up periods are similar and suggest that a high intake of trans fat is associated with a higher risk of CHD morbidity and mortality than a high intake of saturated fat.
At the 14 year follow-up Hu et al also estimated that replacement of 2% of energy from trans fat with 2% energy from unhydrogenated unsaturated fat would result in a 53% (95% CI: 34% - 67%) reduction in risk for CHD; while replacement of 5% of energy from saturated fat with 5% energy from unhydrogenated unsaturated fat would result in a 42% reduction in CHD risk (95% CI: 23% - 56%). There may be a number of reasons for the lack of a statistically significant difference between these two risk estimates. One explanation might be that nutrient intakes were estimated using self-reported questionnaire data; the sources of measurement error associated with this methodology have been well-documented.
In 1999, Ascherio et al (NEJM 1999: 340:1994) reviewed six metabolic studies3 that examined the impact of substitutions of both trans fat and saturated fat for a "control" fat4 on cholesterol biomarkers5 of risk for CHD development. Across these six studies, the percent energy from saturated fat was increased by between 4.5% and 9.9% (from a baseline of 9-11%) when saturated fat was exchanged with control fat. The percentage energy from trans fat was increased by between 3.1% and 11% (from a baseline of 0-1.4%) when trans fat was exchanged with control fat. For six of the seven comparisons made6, the effect of trans fat was significantly more hypercholesterolemic7 than the effect of saturated fat. The exception was the lowest trans fat substitution (3.1%), where the percent energy from trans fat was increased from 0.7% to 3.8%.
These results suggest that at relatively high intakes of trans fat (5.7-11.0% energy) and saturated fat (14-20.1% energy), trans fat has a more deleterious effect on cholesterol biomarkers of risk for CHD than does saturated fat. While the relative health effects of trans fat and saturated fat at lower levels of trans fat intake are also of interest, extrapolation from trans fat levels of 5.7-11.0% energy to levels of 1-3% energy is not seen as appropriate for two reasons: (1) the large variability in LDL/HDL cholesterol ratio response to equivalent substitutions of trans or saturated fat for control fat; and (2) the lack of a consistent trend in LDL/HDL cholesterol ratio response across the different dose levels of saturated fat studied.
Findings from the metabolic studies reviewed by Ascherio et al (NEJM 1999: 340:1994 are consistent with the findings of the prospective cohort study with 14 and 20 years of follow-up (Hu et al [NEJM, 337;1997: 1491-1499] and Oh et al [AMJ 161;2005:672-679]). Research using both these study designs found that at high levels of trans fat (5.7-11.0% energy) and saturated fat intake, the impact of trans fat is more deleterious than that of saturated fat (14-20% energy), whether the outcome studied is CHD events themselves, or a cholesterol-related biomarker of such risk. Conversely, no research has been done to determine whether trans fat are more deleterious than saturated fat at low levels of trans fat intake (1-3% energy).
#6 There is growing evidence that the individual long chain saturated fatty acids do not have an equal effect on the risk factors for CHD. While it is generally recognized that myristic acid is the most hypercholesterolemic saturated fatty acid, there is still debate on the relative benefits of stearic, lauric and palmitic acids. Should recommendations regarding replacement of trans with saturates take into consideration that all saturates may not have the same effect on CHD risk?
Most of the randomized metabolic studies suggest that lauric, myristic and palmitic acids are the most LDL-cholesterol raising saturated fatty acids relative to carbohydrates, and that stearic acid is either neutral or slightly hypocholesterolemic. However, the impact of these long chain saturated fatty acids on CHD is not altogether clear. These four saturated fatty acids increase HDL cholesterol to different extents and consequently result in different levels of TC/HDL-C ratios. Lauric acid would reduce the ratio to a greater extent than stearic and myristic acids, while palmitic acid would raise the ratio. All experts referred to these conclusions primarily based on the meta-analysis performed by Mensink et al in 2003 (AJCN 2003; 77:1146-1155). On the other hand, Hu et al (NEJM, 337;1997:1491-1499) , based on the data from the prospective cohort study of 80 082 women in the Nurses Health Study estimated that an increase of 1% energy from lauric + myristic acids, palmitic and stearic acids increases the relative CHD risk by 14% (significantly), 3% (non-significantly) and 9% (significantly) respectively.
Some studies have shown that linoleic acid mitigates the hypercholesterolemic effect of palmitic acid. Unfortunately, there are no data on the effects of linoleic acid on other saturated fatty acids and therefore, it is not known whether the mitigating effect of linoleic acid pertains only to palmitic acid. At this point, the possible dampening effect of n-6 and n-3 polyunsaturated fatty acids on the cholesterol raising effects of individual saturated and trans fatty acids cannot be taken into consideration in recommendations to eliminate or reduce trans fats.
One invited expert informed the Task Force that fully hydrogenated soybean oil (i.e. primarily tristearin) interesterified with soft oil would elevate LDL/HDL-cholesterol ratio and fasting blood glucose levels relative to natural saturated fat. (Unpublished data) In contrast to this assertion, a study by Snook et al. (Euro J Clin Nutr 1999:53:597-605) demonstrated that myristic, palmitic and stearic, fed as synthetic triglycerides (i.e. trimyristin, tripalmitin and tristearin), were not particularly different from those effects of natural fats and oils on blood cholesterol levels, except myristic acid, which was not as hypercholesterolemic as expected. A study by Nestel et al. (Am J Clin Nutr 1998;68: 1196-1201) found that effect of stearic acid-rich, structured triglyceride on plasma lipid concentrations was not different from a palmitic acid rich diet. In another study, a high-stearic acid diet that provided 5% energy as stearic acid, compared to oleic acid rich diet, did not impair glucose tolerance and insulin sensitivity in healthy women (Louheranta et al. Metabolism 1998; 47: 529-534).
In the absence of more data on the effect of fully hydrogenated oil interesterified with soft oils and also on the relative effects of individual saturated fatty acids on the various risk factors for CHD, most experts consider that it is prudent to ensure that replacements of partially hydrogenated oil not lead to large increases in saturated fats whether they are derived from natural oils and fats or fully hydrogenated fats.
#7 As consumers are turning away from trans fats and products made with partially hydrogenated oils, there is a temptation in some cases (e.g., some baked goods, cookies) to turn back towards alternatives such as butter and tropical oils which are major sources of saturated fats. Based on your analysis, would there still be an overall net health benefit to Canadians if partially hydrogenated oils were effectively eliminated from our food supply but substituted, in some instances, with butter and tropical oils?
There is unanimous agreement among the invited experts that butter is not a good replacement for partially hydrogenated oils. Butter, compared to all the other solid dietary fats including palm oil, palm kernel oil, and coconut oil as well as compared to margarines and shortenings with low to moderate levels of trans fatty acids, has been shown to have adverse effect on the TC/HDL-C ratio (Mensink et al, AJCN 2003; 77:1146-1155 and Lichtenstein et al., NEJM 1999; 340:1933-40). Oils such as palm kernel oil, coconut oil, and palm oil might be better substitutes than butter and other animal fats.
As a general conclusion, the Task Force promotes all actions that lower TFA intake. Results of metabolic and epidemiological studies consistently show that TFA are more harmful than any other type of fat.
Favorable health effects are achieved even if TFA are replaced by saturated fat and even more so if replaced by cis-monounsaturated and cis-polyunsaturated fatty acids. It is important to maintain adequate intakes of polyunsaturated fatty acids to get the benefits of n-6 and n-3 fatty acids, and to limit the cholesterol-raising saturated fatty acids.
In this respect, high MUFA oils are considered as the first choice for an alternative to partially hydrogenated oils and could be used for frying. Coconut oil, palm kernel oil, and palm oil, and fully hydrogenated/interesterified oils can be considered as replacement but not as primary replacements. Butter is not seen as a good replacement because of its greater tendency to increase the ratio of TC/HDL-C compared to palm, palm kernel and coconut oils. (Mensink et al, AJCN 2003; 77:1146-1155).
1 Lichtenstein et al (N Engl J Med. 1999.340(25):1933-40) / Wood et al (J Lipid Res. 1993. 34(1):1-11) / Kris-Etherton et al (Metabolism. 1993. 42(1):121-9) / Hu et al (Am J Clin Nutr. 1999. 70(6):1001-8).
2 (a) Oh et al (AMJ 161;2005:672-679) / Hu et al (NEJM, 337;1997:1491-1499) / Willet et al (Lancet 1993;341:581-585); (b) Ascherio et al (BMJ 1996;313:83-90); (c) Pietinen et al (AJE 1997;145:876-887); (d) Bolton-Smith et al (EHJ 1996;17 :837-845); and (e) Oomen et al (Lancet 2001;357:746-751).
3 (Zock and Katan [JLR 1992;22:399-410]; Judd et al [Lipids 2002;37:123-131]; Judd et al [Am J Clin Nutr 1994;59:861-868]; Sundram et al [J Nutr 1997;127:514S-520S]; Mensink and Katan [NEJM 1990; 323:439-445]; and Nestel et al [J Lipid Res 1992;33:1029-1036])
4 Four studies used oleic fat as the control fat (Judd et al [Am J Clin Nutr 1994;59:861-868]; Judd et al [Lipids 2002;37:123-131]; Nestel et al [J Lipid Res 1992;33:1029-1036]; Mensink and Katan [NEJM 1990; 323:439-445]), one study used cis 18: blend as the control fat (Sundram et al [J Nutr 1997;127:514S-520S]) and one study used linoleate fat as the control fat (Zock and Katan [JLR 1992;22:399-410]).
5 In terms of the cholesterol ratios, one study failed to calculate a cholesterol ratio (Nestel et al [J Lipid Res 1992;33:1029-1036]); in this study, both LDL and HDL levels were significantly worse for the trans versus the saturated fat diet. Of the five studies that calculated cholesterol ratios, LDL:HDL was calculated in two (Mensink and Katan [NEJM 1990; 323:439-445] and Sundram et al [J Nutr 1997;127:514S-520S]), TC:HDL was calculated in two (Judd et al [Lipids 2002;37:123-131]; Judd et al [Am J Clin Nutr 1994;59:861-868]), and HDL:LDL was calculated in one (Zock and Katan [JLR 1992;22:399-410).
6 While six studies were reviewed, Judd et al [Am J Clin Nutr 1994;59:861-868] administered a moderate trans fat diet as well as a high trans fat diet. Thus, the total number of trans fat-saturated fat comparisons was seven.
7 The term "hypercholesterolemic" is used to refer to a significant unfavourable change in any cholesterol parameter (LDL, HDL, LDL:HDL, TC:HDL, HDL:LDL).