Diet and risk of coronary heart disease

Background

There are numerous well-documented risk factors for coronary heart disease (CHD) (1). Dietary factors have been suggested to be involved mainly through the influence of saturated fat intake on the risk of CHD, mediated by an increase in LDL- cholesterol (2). Other dietary factors believed to be involved in the chain of events leading to CHD are risk decreases through moderate alcohol intake, high intakes of fruits and vegetables, oily fish, fiber and nuts, and the so-called Mediterranean diet and risk increases through high intakes of trans-fatty acids and foods of high glycemic index (3).

There are a few central items to remember when discussing the health benefits of food:

1. The recommended diet must be preventive for most serious diseases.

   We can only eat one diet. If a dietary component is believed to decrease the risk of cardiovascular disease (CVD) but at the same time increases the risk of diabetes or a certain form of cancer, there is no way of a logical choice for you.

2. If we are to change dietary habits for a whole population or part of a population (e.g., persons at high risk of CHD), we should have very strong scientific evidence behind the recommendations.

   Food is one of man's strongest reward systems, and the dietary factors we prefer is determined already early in life and hard to change.

3. The present debate is more based on emotional arguments than scientific proofs.

   Today we have an intense debate on central dietary issues like high intake of carbohydrates versus high fat intake (4). We also have numerous TV programs on health aspects of foods— “you are what you eat”—without any scientific evidence for sweeping recommendations.

4. Assessment of diet has some inherited weaknesses, which work against the preformed hypothesis you are to answer.

   The assessment methods are imprecise partly due to recollection errors. This means that the correlations for the used dietary methods, for example, food frequency questionnaires, against a golden standard method, are in the order of 0.5–0.6 for central dietary components like intake of various fats or carbohydrates. This means in turn that the determination coefficient (r00B2), which describes the true impact of the factor, is in the order of 0.25–0.40. Another methodological problem is the high colinearity between many foods and dietary factors. If you eat a lot of one item, you usually eat less of another. Dietary fat intakes have high correlation coefficients, indicating that the food sources contain several fat types. This colinearity makes interpretation of multivariate analyses difficult. Which factor is giving the effect?

Today's recommendations

Dietary recommendations given by various national agencies (5) and scientific organizations (6), for the purpose of CHD prevention, are surprisingly similar, encouraging intake of fruit, vegetables, whole-grain cereals and bread, fiber, oily fish, lean meat, and low-fat dairy products. Furthermore, the intake of total fat and saturated fat should be reduced and replaced by monounsaturated and polyunsaturated fat. Trans-fatty acids should be abandoned. Food with low glycemic index is recommended, and the intake of refined sugar should be reduced. Low to moderate intake of alcohol is of significant benefit, although controversial to recommend to the public.

The question now is if these recommendations are supported by strong scientific evidence in the results of studies in nutritional epidemiology. The vast majority is observational cohort studies (3), and only a few are randomized intervention trials (7). Observational studies have inherent problems in finding causal risk factors for CHD. To increase the possibility that a factor is causally linked to a disease, epidemiologists use the Bradford Hill criteria for causality (8).

1. Strength of the statistical association.

2. Consistency of the statistical association in several studies

3. Temporality, i.e., that exposure must precede outcome

4. Coherence, data should not conflict with known risk factors for the disease

5. Biological gradient meaning that risk increases incrementally with dose of exposure

6. Random controlled trials (RCTs) or experiments support the association

7. Specificity, i.e., the exposure/risk association should be specific for one disease

8. An association should make biological sense

9. Analogy—arguing by analogy is the weakest form of evidence of causality

One systematic meta-analysis used four of the Bradford Hill criteria when evaluating the associations between dietary factors and risk of CHD (3). As mentioned earlier, this recent analysis showed:

1. Strong protective evidence (4 criteria) for vegetables, nuts, Mediterranean and high-quality dietary patterns, and harmful effects of trans-fatty acids and foods with high glycemic index.

2. Moderate evidence (3 criteria) for a protective effect of fish, omega-3 fatty acids, folate, whole grains, dietary vitamin E and C, beta carotene, alcohol (moderate and heavy intake), fruit, and fiber.

3. Insufficient evidence (≤2 criteria) is present for intake of saturated and polyunsaturated fatty acids, total fat, α-linolenic acid, meat, eggs, and milk.

Comparing the results from this recent, well- designed meta-analysis with the recommendations from national agencies (5) and scientific organizations (6), show that some recommendations—vegetables, fruits, whole grain cereals and bread, trans-fatty acid intake, and fish have a solid scientific base. Others like reducing total and saturated fat has no scientific evidence behind the recommendations (3,9) and are based on old observational studies with severe methodological problems.

The two major studies that have formed the basis for the recommendations to reduce total and saturated fat intake are the Seven Countries Study (10) and the Finnish Mental Hospital Study (11). More recently performed observational (12–15) and interventional studies (7) have not been able to confirm the results of these older studies. Despite this, all recommendations still contain advice to reduce total and saturated fat intake.

The Seven Countries Study

The 15-year follow-up of this observational study was published in 1986 (10). Thus, it started in the late 1960s. Fifteen centers in seven countries—mostly European—participated. Each center included 500–700 men, aged 40–59. The participants were not random samples and not representative for the background populations. Dietary intake was assessed only in 30–50 participants in each center. The dietary assessment method varied between centers. In most centers, a double-plate method was used. However, in Italy, a 7-day recall plus food diary method was used, and in the United States, a combination of a food frequency questionnaire plus one 24-hour recall was used. No analysis of variation in dietary intakes due to different methods being used is presented in the paper. As only 30–50 subjects in each center were assessed for dietary intake, the authors choose to present only the center mean intakes of, for example, energy, various fats, protein, carbohydrates, and alcohol, instead of each individual's intake values. The reason to use the group mean was the very large individual variation in dietary intake in the one and only assessment increasing the variation (SD) within each center data set. Group mean decreases the SD and increases the probability of a statistically significant association to risk of CHD.

The Seven Countries Study gave some very interesting results that shaped for decades the way epidemiologists looked at the relationship between fat intake and risk of CHD. The results were discussed in the US Congress and became the foundation for the American recommendations to decrease fat intake to prevent CHD (6). The decreased fat intake was counteracted by an increase in carbohydrates to keep energy intake constant. We will come back to the health effects of that change later. The results showed that 44% of the variation in death rate from CHD between the centers was explained by the ratio of monounsaturated fat to saturated fat intake.

Already here, epidemiologists should have reacted to the very high degree of explanation from one factor in such a complicated disease endpoint as death from CHD, where more than 200 risk factors now have been identified. Furthermore, almost all variation (96%) in death rate from CHD between the centers was explained if age, body mass index, systolic blood pressure, serum cholesterol, and number of cigarettes smoked daily was added to the above ratio. CHD death rates were lower in cohorts with olive oil as the main fat, but today we know that CHD death rates are not lower in countries with olive oil as the main fat (18); a fact that was available already in 1986 if the epidemiologists had looked for it.

Today we also know that there are numerous other risk factors that are involved in the development of CHD. Consequently, common sense would have told the scientific community that the findings above were not correct.

The study had also several weaknesses that ran undiscovered at the time of publication. It was an ecological study using population means instead of individual intakes, resulting in dramatically reduced standard deviations and hence easily achieved statistical significances in calculating risk estimates. Diet was measured in very few men in each center, making the dietary means unrepresentative for the background population. Dietary assessment methods varied between centers, making direct comparisons problematic. In spite of all these problems, the results of this study have been used for today's dietary recommendations.

The Finnish Mental Hospital Study

The authors of this study published 1979 (11) claim that it is a “controlled intervention study comparing two diets with different fat quality,” with the aim to test the hypothesis that the incidence of CHD could be decreased by the use of serum-cholesterol- lowering (SCL) diet. The diet chosen for this purpose was to replace ordinary dairy saturated fats with soybean milk and butter with margarines, both replacements high in polyunsaturated fats. However, there are also several flaws in the design of this study, and it could better be described as an observational study including patients at two mental hospitals, where the fat quality of the diet was changed halfway during a 12-year period. For example, patients (34–64 years of age) were not randomized to the two hospitals, new patients were admitted and included, while other patients were discharged and hence excluded from the study during all 12 years. Therefore, it is impossible to follow which patient got which diet during which time period, and consequently, one cannot calculate how many months each individual was “at risk.” Hence, there is no stable denominator with which to divide the number of endpoints to achieve true incidence figures. Furthermore, the diagnosis of CHD was based on death from CHD (clinical or autopsy) or appearance of certain specific ECG changes (Q-waves or ST-T-changes), which are known to be inaccurate as the criteria for a new CHD event. All statistical tests were one-sided, that is, the authors hypothesize that the effect of the intervention could only be in one direction, which is statistically very debatable and would be unacceptable in a publication of today.

The actual results of the study are even more problematic. SCL diet definitely lowered the serum cholesterol. However, the hospital (N) with the SCL diet during the first 6 years had already at the start of the “controlled” study significantly lower serum cholesterol. The authors gave no explanation for this surprising fact. The main results concerning CHD events, are given in Table I, showing the data for men. The numbers of CHD events were very small even when cases developing ST-T changes were added. Only the difference between the two groups in coronary deaths plus Q-waves in Hospital N was significant in two-sided tests. There were no significant differences for women in incidence of CHD (data not shown). The study, thus, showed that serum cholesterol could be lowered by use of a very extreme diet replacing saturated fats with polyunsaturated fats, but this change in diet did not seem to have a significant effect on the incidence of CHD events, neither in men nor in women.

Table I. Incidence of manifestations of coronary heart disease.

Criterion	D		MD		IMD
Diet period	SCL	Norm	SCL	Norm	SCL	Norm
Hospital N
Number	4	5	4	16	11	23
Incidence	3.7	4.3	3.7	13.9	10.3	20.3
/1000 man-years
p-value	0.42		0.005		0.03
Hospital K
Number	2	7	4	10	14	24
Incidence	2.3	8.0	4.7	11.5	16.8	28.3
/1000 man-years
p-value	0.06		0.06		0.06

D = coronary death, MD = major ECG change (Q-wave) or coronary death, IMD = intermediate or major ECG change and coronary deaths.

Diet period: SCL = soya bean oil diet Norm = Normal diet.

Data for men. No statistical significant differences in women.

It is surprising that so many national agencies and scientific organization have totally overlooked these serious design flaws of the Finnish Mental Hospital Study and still use the results of this study to support their views on the risks of saturated fat intake for the development of CHD.

Later observational cohort studies

The Health Professionals Follow-up Study

This was a large (34,463) cohort study (12) of male adult (40–70 years of age) health professionals that was followed for more than 10 years, and hence had many CHD endpoints (1702). Diet was assessed by Food Frequency Questionnaires (FFQ) repeated several times during follow-up. The authors used multivariate logistic analyses with adequate adjustments, in comparing CHD risk between the 5th and the 1st quintiles of various fat intakes. The results showed no significant association between either total fat intake, or intake of saturated fat and risk of CHD. The intake of trans-fatty acids significantly increased, and the intake of monounsaturated and/or polyunsaturated fat significantly decreased the risk of CHD.

The Nurses’ Health Study

This large (78,778) observational cohort study of 30–55-years-old women was published in 2005 (13) by the same group as the previously described the Health Professionals Follow-up Study (HPFS). Diet was assessed using the same FFQ repeated several times. A similar multivariate logistic analysis, as in the HPFS with adequate adjustments for confounders including fat subclasses, was performed to compare CHD risk in 5th and 1st quintiles of the various fat intake categories. The main results were very similar to those found in the HPFS.

The Malmö Diet and Cancer Study

This study started in 1991 and the results on nutritional intake and risk of CVD were published between 2004 and 2007 (14–16). It was a large (28,098 men and women) prospective cohort study, using an elaborate diet assessment method combining a food diary and a large FFQ (168 items). The cohort was followed for 9 years yielding 1556 CVD endpoints. A multivariate logistic analysis with adequate adjustment including fat subclasses was used to compare the risk of CHD in 4th and 1st quartiles of various fat intakes. The results showed that there was no significant relationship between the intake of total fat, the intakes of saturated, monounsaturated, or polyunsaturated fat, and the risk of CHD. Trans-fatty acids were not possible to separate in this study.

It can be concluded that the three studies mentioned above showed essentially the same results—no association between total or saturated fat intake and risk of CHD/CVD. Two recent meta-analyses of Mente et al. (3) and Siri-Tarino et al. (9) support these findings.

It can be concluded that in spite of these unequivocal results from these large, well-designed studies, the recommendation to the population is still to decrease total and saturated fat intake, and there are no data from observational studies to support these recommendations.

Intervention studies on fat intake and risk of CHD

Unfortunately, there exists only one well-designed intervention study exploring the effect of long-term decrease in fat intake on risk of CHD, the Women's Health Initiative (WHI) Randomized Controlled Dietary Modification Trial (7). This study of 48,835 women aged 50–79 years, assigned 19,541 women to an intervention diet designed to lower total fat intake to 20% of calories and increase intakes of vegetables/fruits and whole grains. Of total, 29,294 were assigned to receive their usual diet. The participants were followed for 8.1 years, and CHD and stroke events were recorded. Mean fat intake decreased during intervention by 8% and saturated fat intake by 3%. Intake of vegetables/fruits and grains increased.

The formal intention-to-treat analysis showed no significant effect of the intervention diet compared to the usual diet (HR 0.97, CI 0.90–1.06). In spite of this very clear negative finding, the authors state the following: “The trial is not a test of the dietary guidelines recommended for prevention of CHD (5,6).”

It is extremely surprising to find that the authors state that their findings, although not statistically significant, could be in line with earlier “intervention studies.” The first to be mentioned is the Finnish Mental Hospital Study, which we have shown to have serious flaws in design and analysis. It is also surprising to see that the authors conclude that the results suggest that a diet lower in saturated fat and trans-fat intake and higher in intakes of vegetables and fruits and polyunsaturated fat is of significant benefit in preventing CHD. Never give up your beloved hypothesis even if your own data from the best trial by far contradicts it. Collective scientific suicide is the word for it.

Contrary to this statement, we conclude that intervention studies do not support a preventive effect on CHD by lowering total and saturated fats.

Alcohol intake and risk of CHD

During the last two to three decades, numerous reports have convincingly shown that a light to moderate intake of alcohol significantly reduces the risk of developing CHD and also protects patients with CHD from deteriorating. Two recent meta-analyses (17,18) clearly show that an intake of 1–3 drinks/day (no difference between wine, beer, or spirit) reduces the risk for CHD by 20–50%. The risk of developing cardiac insufficiency is also reduced by ∼30%.

There is no difference in protective effect between men and women, but women are recommended to have a smaller intake (≤1 drink/day) than men (≤2 drinks/day). In one of the studies (18), a significant reduction of all deaths was also shown in people with a light to moderate intake of alcohol.

The results of the Mente meta-analysis, using Bradford Hill's criteria for causality, also showed that both light to moderate (1–3 drinks/day) and heavy consumption (>3 drinks/day) was associated with a 30% lower risk of CHD.

In light of these very coherent findings, it is curious that the Mente meta-analysis does not at all discuss the important finding that a moderate alcohol intake reduces the risk of CHD with ∼30%, which is in bright contrast to the comments on other findings in their study. We are aware that there are important negative effects of heavy drinking, but we think that the medical profession has a moral and ethical responsibility also to inform the public in a balanced way about the positive effects of light to moderate alcohol consumption.

Comments and conclusions

This review of the scientific basis for the current guidelines for prevention of CHD shows:

1. The recommendations to increase intake of vegetables and fruits, whole grain, oily fish, and fiber seem to be supported by acceptable scientific proof.

2. Increases in lean meat and low-fat dairy products have no solid basis.

3. The recommendation to reduce total fat intake to below 30 E% (of energy intake) and saturated fat intake to below 7–10 E%, and to replace it by monounsaturated and polyunsaturated fat, is not supported by scientific facts.

4. Decrease in trans-fatty acids is motivated by strong scientific evidence.

5. That light to moderate intake of alcohol (1–3 drinks/day) significantly reduces the risk of CHD for both men and women is strongly supported by scientific data.

6. There seem to be good scientific evidence to reduce food with high glycemic index and refined sugars.

These conclusions are valid for the prevention of CHD. However, they may not be valid for other widespread diseases such as cancer, diabetes mellitus, or rheumatic diseases. Consequently, it is of utmost importance to include the effect on such common diseases too, if a specific diet should be recommended to the public. To our knowledge, such recommendations based on scientific valid evidence do not exist today. As food intake is one of our strongest reward systems, general advice on food choices to entire populations, with the aim to reduce burden of our most serious diseases, should be based on strong scientific proof of beneficial effects on global health. Such scientific evidence is not at hand today.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.