Recently, however, several papers have presented evidence suggesting that carbohydrates may be involved in the etiology of heart disease. A study of 72 premenopausal women between the ages of 25 and 44 in switching from a 0.35 fat diet to a 0.21 fat diet with 0.65 carbohydrates (half simple, half complex) and found that cholesterol levels jumped 0.15 to

0.33 depending on the LDL subclass pattern of their parents.²² Another study showed that a 0.60-carbohydrate/0.25-fat diet leads to higher plasma triacylglycerol, very-low-density-lipoprotein (VLDL) and VLDL-cholesterol than a 0.4-carbohydrate/0.45-fat diet.¹⁹ The sugar:starch ratio in each diet was

0.33:0.66. Case control study data by Knopp et al.²³ shows that lowering fat from 0.35 to 0.22-0.27 while increasing carbohydrates from 0.45-0.46 to 0.52-0.60 for those with hyper-cholesterolemia lowers LDL cholesterol but raises triglycerides after 12 months. Gaziano et al. present data on the relative risk ratios for case controlled studies of triglyceride levels, showing a superlinear relationship.²⁴ Evidence from clinical trials presented by Oliver²⁵ indicates that increasing the intake of unsaturated fats is more likely to lead to reduced ischemic heart disease than reducing saturated fat intake. Katan²⁶ also shows that low-fat, high-carbohydrate diets do not reduce ischemic heart disease, probably since triglycerides levels rise. Finally, a recent study shows that carbohydrates increase the risk of coronary heart disease in women compared with mono- and polyunsaturated fats, but not in comparison with saturated fat.²⁷ No distinction was made between simple and complex carbohydrates in these studies. Thus, there is sufficient reason to suspect that sugar may, indeed, be involved in the etiology of heart disease.

Methods

To investigate whether sweeteners can be considered a risk factor for heart disease, the multi-country statistical approach was employed. In this approach, national disease statistics are compared with components of the national dietary consumer supply values in a multi-variate model. The components with statistically-significant associations with the disease are then examined further for links to that disease to establish their role as risk factors.

While this approach has been successfully applied in the past, it has fallen out of favor in the medical research community for several reasons. One, called the ecological fallacy, is that it is hard to guarantee that some unknown confounding variable is not affecting the analysis.²⁸ Confounding between animal fat and sugar⁶ played an important role in the Keys-Yudkin debate. In addition, in studies using mortality from disease, the medical care delivery system of the country can affect the statistics. Also, the national consumer dietary supply²⁹ is not necessarily a good measure of the diet that actually pertains to those in the disease study. Finally, the approach is rather coarse in that dietary components that comprise a small fraction of larger components of the diet are generally not broken out. For example, the various fatty acids are not listed separately.

In defense of the multi-country approach, it has generally yielded results which are consistent with what is or has been accepted regarding dietary risk factors for various diseases. For example, this approach has been successfully used to link low-fiber diets to appendicitis;^30,31 high-fat, low-carbohydrate, high-lipid diets to diabetes mellitus;³² dietary sugar to heart disease;⁵ animal fat to CHD;¹ and dietary fat to colon cancer³³ and Alzheimer’s disease.³⁴ Successes with the multi-country approach are often verified by congruence between epidemiologic, clinical, and laboratory research findings.²⁸ For dietary factors, the Food Balance Sheets of the Food and Agriculture Organization (FAO)²⁹ are used. The FAO regularly publishes a detailed summary for many components of the diet for 179 countries in three-year intervals through 1994.³⁵ The values represent food available to the consumer, and do not take into account losses due to spoilage, wastage, etc. Thus, they overestimate the amount actually consumed. In the U.S.A., the reduction was recently estimated to be

0.25. However, it is assumed that similar reduction factors are found in other countries used in the study. Sweeteners in their tables include sugar, non-centrifugal; sugar (raw equivalent); sweeteners; and honey, with 0.95-1.00 of the calories coming from sugar.

In this study, regression analyses were performed for 1986 mortality rate data for acute myocardial infarction (AMI) and other ischemic heart diseases (OIHD) for four 10-year age groupings from 35 to 74 for 33 countries³⁶ (Appendix 1, p. 83). Countries were initially screened for life expectancy in 1970 greater than 69 years and populations over 1 million. Only Eastern European countries (Bulgaria, Hungary, Poland, Romania, and the USSR) were intentionally excluded because the high heart disease mortality rates in these countries indicate a systematic difference with respect to other countries. Dietary supply data for 1982 (averaged from 1979-1984) and 1973 (averaged from 1972-74) were also used in the analysis.²⁹

Results

The AMI mortality data for those aged 65-74 were first used in a multiple linear regression analysis with alcohol, animal fat, calories, cereals, cigarettes,³⁷ fat, fish, fruit, life expectancy,³⁸ starchy roots, stimulants, sweeteners, and vegetables. Only animal fat and sweeteners were found to be statistically significant (p<0.05). Even with smaller subsets of variables, these two were still the only variables which yielded significant results. However, for linear regressions, five variables were found to be statistically significant for males age 65-74: sweeteners (r² = 0.347, p <0.001); animal fat (0.587, <0.001); fat (0.509, <0.001); calories (0.169, 0.017); and cereals ((-)0.398, <0.001), where (-) denotes an inverse association. For AMI and females age 65-74, only four variables were found to be statistically significant: sweeteners (0.486, <0.001), animal fat (0.333, <0.001), cereals (0.271, 0.002), and fat (0.259, 0.003). Four of these five variables have been mentioned in the past as having some possible association with heart disease. The model examined was that animal fat and sweeteners together comprise the primary dietary risk for AMI. The p value for cereals in multiple regressions showed it to be not statistically significant.

Scatter plots of the highest linear associations for AMI are given in Table 1

(p. 98) and shown graphically in Figures 1 and 2 (p. 99). For males, dietary animal fat supply for 1973 has a significantly higher correlation for AMI mortality rates than does either animal fat for 1983 or sweeteners. Possibly the effect of animal fat builds up slowly through the years. For females, dietary sweetener supply has a much higher correlation with AMI mortality rates than does animal fat, with the results rather independent of year.

Figures 3 and 4 (p. 100) present the statistical results for AMI for the four age groups considered. For males, animal fat is the primary dietary risk factor, but sweeteners seem to help, with the F value (F = t², t = student t-test) not changing much. For females, sweeteners is the primary dietary risk factor, with animal fat contributing. The F value is nearly the same for both calculations for age 65-74, but drops to about half for the pair compared to sweeteners alone for age 35-44.

The OIHD mortality data were first tested with a number of factors with similar results as for AMI. Thus, the same model, that animal fat and sweeteners are the primary dietary factors raising the risk of OIHD was adopted. The results are presented in Table 1 and Figure 5 (p. 101). First, the statistical associations for the various age groups are similar to those for AMI. The combined animal fat/sweetener association Table 1. Statistical results (r² and p) for AMI and OIHD mortality rates (WHO)³⁶ for 1973 animal fat and 1983 sweeteners (FAO).²⁹

p values: * <0.001; ** <0.05; *** >0.05

generally decreases with age of the groups, except for females 35-44. Second, animal fat generally has a higher association than sweeteners for men. The exceptions are for males 65-74 and men 35-44 with the 1982 dietary data. For women, sweeteners are always found to be much more highly associated 98 with OIHD than animal fats, which often had inverse associations, although not statistically significant ones. Animal fats from 1973 were found to be more highly associated with OIHD in men than the 1982 values. Figure 2

(p. 79) shows linear regression data for sweeteners for 65-74-year-old females.

r² = 0.608, p<0.001

0.704 x sweeteners (cal/day).

r² = 0.486

animal fat-1973

animal fat-1983

sweeteners-1983

r² = 0.572

Data from Spain for trends of CHD mortality³⁹ can be used as a further check. Regression analyses were run for the current year dietary data as well as for two, four and six years prior to the mortality data. The results are shown in Table 2 (p. 78) for current year dietary data. The results support the findings of the geographical distributions for females. However, for males, only the current year data show a strong association with animal fat. For other years, sweeteners have a much higher r², even higher than for females for four and six years prior diets. The authors of the study did not examine sweeteners as a possible factor.

Discussion

Supporting data for the role of sweeteners in the etiology of heart disease also comes from the island of St. Helena (15.57°S, 5.42° W). As reported Shine et al.⁴⁰ and discussed by Hoffer,⁴¹ people there exercise a lot, don’t smoke much, and have low fat intake, but still have a very high incidence of heart disease. Their sugar intake had increased to 125 pounds per person per year.

Thus, these analyses suggest that epidemiologic data support the finding of the 1960s that dietary sucrose is a risk factor for heart disease. Returning to the studies of Keys^1,15 and Yudkin,^5-7 we see that they did not use sophisticated statistical analyses techniques, such as multivariate analyses, so were unable to examine carefully the separate roles of animal fats and sweeteners. In addition, they did not study the mortality rates of females, where the sugar effect is larger.

While associations are useful in trying to make the link between diet and disease, the case is strengthened considerably when mechanisms can be found, such as a step-by-step linkage between dietary components and disease. The proposed mechanism linking sugar to heart disease is the synthesis of serum triacylglycerols (triglycerides or TG) by the liver, followed by increased production of very low density lipoprotein (VLDL).

Sucrose is a disaccharide which splits into glucose and fructose in the first metabolic step.⁴² Fructose proceeds through several steps in the liver before generally being synthesized into phosphoglycerides and TG. Triglycerides, in turn, are incorporated into plasma lipoprotein (colloquially, cholesterol): VLDL is 50-60% TG, 20% cholesterol and cholesterol esters (CCE), 15-20% phospholipids (PPL), and 5-10% protein. LDL, derived from VLDL, is 8% TG, 22-40% CCE, 22-30% PPL, and 20-22% protein; and high density lipoprotein (HDL), formed in both the liver and intestines, is 5-8% TG, 16-25% CCE, 24-30% PPL, and 40-52% pro-tein.^42,43 Thus, the more fructose in the diet, the more production of TG, leading to more VLDL and LDL. Both TG and VLDL are considered to be high risk factors for IHD.²⁴

Syndrome X was defined in 1988⁴⁴ as a cluster of abnormalities occurring in nondiabetic persons that increase the risk of coronary heart disease.⁴⁵ The components of syndrome X were originally identified as including some degree of glucose intolerance, a high TG and low HDL concentration and an increase in blood pressure. Recently, atherogenesis has been added to the syndrome.⁴⁵

The results of this paper identify simple sugars as the primary cause of serum TG and VLDL. This conclusion is strongly supported by the results of Ref. 10, which show that for 16 Asian and South American countries, primarily from data on the armed forces, the cholesterol levels are directly proportional to the fraction of calories from simple sugars (mono and disaccharides) (r=0.81) and inversely related to the fraction of complex sugars (polysaccharides) (r=-0.72), with a very weak dependence on the fraction of calories from fat (r=0.25). Hudgins et al.⁴⁶ provide recent support for this conclusion through clinical studies. Barker et al.⁴⁷ also point out that LDL-cholesterol levels are not necessarily affected by dietary fat.

Based on the principles of syndrome X, the national dietary data were used to see whether carbohydrates other than the simple sugars play a role in the etiology of AMI. For this analysis, the carbohydrate portion of cereals and starchy roots were included in a multiple linear regression with animal fat and sugars for those aged 65-74. The p values for the carbohydrates from cereals and roots were found to be