The Cholesterol Myths by Uffe Ravnskov, M.D., Ph.D.
1. Your cholesterol tells very little about your future health
Cholesterol is a peculiar molecule. It is often called a lipid or a fat. However, the chemical term for a molecule such as cholesterol is alcohol, although it doesn't behave like alcohol. Its numerous carbon and hydrogen atoms are put together in an intricate three dimensional network, impossible to dissolve in water. All living creatures use this indissolvability cleverly, incorporating cholesterol into their cell walls to make cells waterproof. This means that cells of living creatures can regulate their internal environment undisturbed by changes in their surroundings, a mechanism vital for proper function. The fact that cells are waterproof is especially critical for the normal functioning of nerves and nerve cells. Thus, the highest concentration of cholesterol in the body is found in the brain and other parts of the nervous system.
Because cholesterol is insoluble in water and thus also in blood, it is transported in our blood inside spheric particles composed of fats (lipids) and proteins, the so-called lipoproteins. Lipoproteins are easily dissolved in water because their outside is composed mainly of water-soluble proteins. The inside of the lipoproteins is composed of lipids, and here are room for water-insoluble molecules such as cholesterol. Like submarines, lipoproteins carry cholesterol from one place in the body to another.
The submarines, or lipoproteins, have various names according to their density. The best known are HDL (High Density Lipoprotein), and LDL (Low Density Lipoprotein). The main task of HDL is to carry cholesterol from the peripheral tissues, including the artery walls, to the liver. Here it is excreted with the bile, or used for other purposes, for instance as a starting point for the manufacture of important hormones. The LDL submarines mainly transport cholesterol in the opposite direction. They carry it from the liver, where most of our body's cholesterol is produced, to the peripheral tissues, including the vascular walls. When cells need cholesterol, they call for the LDL submarines, which then deliver cholesterol into the interior of the cells. Most of the cholesterol in the blood, between 60 and 80 per cent, is transported by LDL and is called ”bad” cholesterol, for reasons that I shall explain soon. Only 15-20 percent is transported by HDL and called ”good” cholesterol. A small part of the circulating cholesterol is transported by other lipoproteins.
You may ask why a natural substance in our blood, with important biologic functions, is called ”bad” when it is transported from the liver to the peripheral tissues by LDL, but ”good” when it is transported the other way by HDL. The reason is that a number of follow-up studies have shown that a lower-than-normal level of HDL-cholesterol and a higher than-normal level of LDL-cholesterol are associated with a greater risk of having a heart attack, and conversely, that a higher-than-normal level of HDL-cholesterol and a lower-than normal LDL-cholesterol are associated with a smaller risk. Or, said in another way, a low HDL/LDL ratio is a risk factor for coronary heart disease.
However, a risk factor is not necessarily the same as the cause. Something may provoke a heart attack and at the same time lower the HDL/LDL ratio. Many factors are known to influence this ratio.
What is good and what is bad?
People who reduce their body weight also reduce their cholesterol. In a review of 70 studies Dr. Anne Dattilo and Dr. P.M. Kris-Etherton concluded that, on average, weight reduction lowers cholesterol by about 10 per cent, depending on the degree of the reduction. Interestingly, it is only cholesterol transported by LDL that goes down; the small part transported by HDL goes up. In other words, weight reduction increases the ratio between HDL- and LDL-cholesterol (1). An increase of the HDL/LDL ratio is called ”favorable” by the diet-heart supporters; cholesterol is changed from ”bad” to ”good”. But is it the ratio or the weight reduction that is favorable? When we become fat, other harmful things occur to us. One is that our cells become less sensitive to insulin, so that some of us develop diabetes. And people with diabetes are much more likely to have a heart attack than people without diabetes, because atherosclerosis and other vascular damage occur very early in diabetics, even in those without lipid abnormalities. In other words, overweight may increase the risk of a heart attack by mechanisms other than an unfavorable lipid pattern, while at the same time overweight lowers the HDL/LDL ratio.
Also smoking increases cholesterol a little. Again, it is LDL-cholesterol that increases, while HDL-cholesterol goes down, resulting in an ”unfavorable” HDL/LDL ratio (2). What is certainly unfavorable is the chronic exposure to the fumes from burning paper and tobacco leaves. Instead of considering the low HDL/LDL ratio as bad it could simply be smoking itself that is bad. Smoking may provoke a heart attack and, at the same time, lower the HDL/LDL ratio.
Exercise decreases the bad LDL-cholesterol and increases the ”good” HDL-cholesterol (3). In well-trained individuals the ”good” HDL is increased considerably. In a comparison between distance runners and sedentary individuals, Dr. Paul D. Thompson and his colleagues found that the athletes on average had a 41 per cent higher HDL-cholesterol level (4). Most population studies have shown that physical exercise is associated with a lower risk of coronary heart disease, and a sedentary life with a higher risk. It also seems plausible that a well-trained heart is better guarded against obstruction of the coronary vessels than a heart always working at low speed. A sedentary life may predispose people to a heart attack and, at the same time, lower the HDL/LDL ratio.
A low ratio is also associated with high blood pressure (5). Most probably, the hypertensive effect is created by the sympathetic nerve system, which is often overstimulated in hypertensive patients. Hypertension (or too much adrenalin) may provoke a heart attack, for instance by inducing spasm of the coronary arteries or by stimulating the arterial muscle cells to proliferate, and, at the same time, lower the HDL/LDL ratio.
Univariate and multivariate
As you see, it is not easy to know what is bad. Is it bad to be fat, to smoke, to be inactive, to have high blood pressure, or to be stressed? Or is it bad to have a lot of bad cholesterol? Or both? Is it good to be slim, to stop smoking, to exercise, to have normal blood pressure, to be emotionally calm? Or is it good to have much ”good” cholesterol? Or both? Thus, the risk of having a heart attack is greater than normal for people with high LDL-cholesterol, but so is the risk for fat, sedentary, smoking, hypertensive and mentally stressed individuals. And since such individuals usually have elevated levels of LDL-cholesterol, it is, of course impossible to know whether the increased risk is due to the previously mentioned risk factors (or to risk factors we do not yet know) or to the high LDL-cholesterol. A calculation of the risk of high LDL-cholesterol that ignores other risk factors is called a univariate analysis and is, of course, meaningless.
To prove that high LDL-cholesterol is an independent risk factor, we should ask if fat, sedentary, smoking, hypertensive and mentally stressed individuals with a high LDL-cholesterol level are at greater risk for coronary disease than fat, sedentary, smoking, hypertensive and mentally stressed individuals with low or normal LDL cholesterol.
Using complicated statistical formulas, it is possible to do such comparisons in a population of individuals with varying degrees of the risk factors and varying levels of LDL-cholesterol, a so-called multivariate analysis. If a multivariate analysis of the prognostic value of LDL cholesterol also takes body weight into consideration, it is said to be ”adjusted for body weight”. A major problem with such calculations is that we know a great number of risk factors because the more risk factors that are adjusted for, the less reliable the result will be. Another problem is that the data generated by these and other complicated statistical methods are almost impossible for most readers, including most physicians, to comprehend. For many years researchers in this area have not presented primary data, simple means, or simple correlations. Instead, their papers have been salted with meaningless ratios, relative risks, p-values, not to mention obscure concepts such as the standardized logistic regression coefficient, or the pooled hazard rate ratio. Instead of being an aid to science, statistics are used to impress the reader and cover the fact that the scientific findings are trivial and without practical importance. Nevertheless, let us have a look at some of the studies.
The ”good” one
Publications almost beyond counting have studied the prognostic value of the ”good” HDL-cholesterol. The reason is, of course, that it is hard to find any prognostic value. If HDL-cholesterol had a heart-protecting effect of real importance, it would not be necessary to use the tax payers' money to demonstrate the effect again and again in expensive studies. To be brief I shall tell you only about a few of the largest studies.
In 1986 the medical statistician, Dr. Stuart Pocock and his coworkers published a report concerning more than 7000 middle-aged men in 24 British towns (6). The men had been followed for about four years after a detailed analysis of their blood lipids. During this period 193 of the men had had a heart attack. As in most previous studies, these men had on average a lower HDL-cholesterol at the beginning than the men who did not have a heart attack. The mean difference between the cases and the other men was 2.7 mg/dl, or about 6 per cent. This difference was small of course, but thanks to the large number of individuals studied it was statistically significant.
But this was a univariate analysis and as mentioned, the difference could therefore be explained by many ways. A multivariate analysis adjusted for age, blood pressure, body weight, cigarette smoking and non-HDL-cholesterol reduced the difference to an insignificant 0.9 mg/dl, or 2 per cent. This means that those who had suffered a heart attack had a lower HDL-cholesterol mainly because they were older, fatter, had a higher blood pressure and smoked more than those who had not had a heart attack. Dr. Pocock and his colleagues concluded that a low HDL-cholesterol level is not a major risk factor for coronary heart disease.
Their results were challenged in 1989 by nine American scientists headed by Dr. David Gordon. They had analysed the predictive value of HDL-cholesterol in four large American studies, a total of more than 15,000 men and women (7). They thought that the British scientists had used an incorrect way to adjust their figures. If another formula is used, the American researchers wrote, HDL-cholesterol is a much better predictor.
But in one of the four studies, analyzed by Dr. Gordon and his colleagues, the number of fatal heart attacks was identical in the first and second HDL tertile (individuals were classified into three groups, or tertiles, according to their HDL-cholesterol). In one of the studies the number of fatal cases was identical in the second and the third tertile, and in one study more deaths were seen in the third tertile (those who had the largest amount of the ”good” cholesterol) than in the second tertile. And these figures were the unadjusted ones.
After adjustment for age, cigarette smoking, blood pressure, body weight and LDL-cholesterol the differences were even smaller. In three of the four studies, the differences lost statistical significance. And remember that the figures were not adjusted for physical activity or mental stress, not to mention the risk factors we do not know yet.
Dr. Pocock and his colleagues returned with a new analysis later the same year, now using the same way of analysing as had Dr. Gordon and his colleagues. At that time the participants in the study had been followed for 7.5 years and a total of 443 heart attacks had occurred. This is the largest single HDL study to date (8).
This time a difference was noted between the HDL cholesterol of the heart patients and the others. The difference was small but statistically significant, even after adjustment for the five risk factors mentioned. However, the largest difference was noted for total cholesterol. The authors therefore concluded that a determination of HDL-cholesterol may be of marginal additional value in screening and in intervention programs for risk of coronary heart disease. They could also have added that they did not adjust for all risk factors so that the difference could as well be due to the heart patients being, for instance, more stressed or less active physically than the others.
The ”bad” one
”LDL has the strongest and most consistent relationship to individual and population risk of CHD, and LDL-cholesterol is centrally and causally important in the pathogenetic chain leading to atherosclerosis and CHD”. These words you will find in the large review Diet and Health (9).
Reviews by distinguished scientific bodies are supposed to meet high standards. Therefore, you are probably wondering how the authors of Diet and Health, an official, most authoritative and supposedly reliable review from the National Research Council in Washington, had reached their conclusion about LDL-cholesterol. Four publications were mentioned.
In 1973 Dr. Jack Medalie and his coworkers published a five-year follow-up study of 10,000 Israeli male government and municipal employees (10). But the Israeli study did not support the words of Diet and Health, because total cholesterol, not LDL-cholesterol, had the strongest relationship to risk of coronary disease.
The second paper claimed by the Diet and Health-authors was a 1977 report from the Framingham Study by Dr. Tavia Gordon and her colleagues (11). This study concerned HDL cholesterol, however. Only logistic regression coefficients (a statistical concept unknown to most doctors) for coronary disease on LDL-cholesterol were given, and one of the conclusions of the paper was that ”LDL-cholesterol ...is a marginal risk factor for people of these age groups” (men and women above 50 years). Some of the coefficients were indeed low. For women above the age of 70 it was negative, which means that women at that age ran a greater risk of having a heart attack if their LDL-cholesterol was low than if it was high. Thus, there was no support either from Gordon's paper.
Also, the third paper (12) concerned HDL-cholesterol only. No support again.
The fourth reference was to the National Cholesterol Education Program, which produced another large review without original data (13). One of its conclusions was that ”a large body of epidemiologic evidence supports a direct relationship between the level of serum total and LDL-cholesterol and the rate of CHD.” The large body of evidence was to be found in three references. The first one was another large review without original data, Optimal resources for primary prevention of atherosclerotic disease (14), with Dr. Kannel as the first author. I shall return to their review below.
The next reference was yet a large review (15), but nothing in that review was said about the connection between the LDL-level and the incidence of coronary heart disease.
The last reference was an analysis of various lipoproteins as risk factors in the Honolulu Heart Study (16). The conclusion of that paper was that ”both measures of LDL-cholesterol were related to CHD prevalence, but neither appeared to be superior to total cholesterol”.
Before I discuss Kannel's review I shall mention another conclusion in the National Cholesterol Education Program: ”The issue of whether lowering LDL-cholesterol levels by dietary and drug interventions can reduce the incidence of CHD has been addressed in more than a dozen randomized clinical trials”. This is a most misleading statement because at that time, in 1988, only four randomized trials including LDL-cholesterol analysis had previously been published (17), and only in one of them the number of heart attacks was lowered significantly.
Let me now return to the review by Kannel and colleagues, the one used as evidence by the authors of The Cholesterol Education Program, which in turn was used as evidence by the authors of Diet and Health. Almost nothing was written about LDL-cholesterol in Kannel's review except for the following (page 164A): ”Longitudinal studies within populations show a consistent rise in the risk of CHD in relation to serum total cholesterol and LDL-cholesterol at least until late middle-age”.
A little more cautious conclusion than in Diet and Health, it may seem, but even for this prudent statement the evidence was weak. References to six studies were given. In two of them LDL-cholesterol was not analysed or mentioned at all (18); in two reports LDL-cholesterol was only correlated to the prevalence of heart disease (19); in one report two tables was aimed at the subject (tables 8 and 9) and showed that the predictive power of LDL-cholesterol was statistically nonsignificant (20); in one study LDL-cholesterol was predictive for heart disease, but only for men between 35 and 49 and for women between 40 and 44 (21).
In conclusion, the ”large body of evidence” was cooked down to one single study, which showed a predictive value for LDL-cholesterol but for a few age groups only. LDL-cholesterol is neither centrally nor causally important, it has not the strongest and most consistent relationship to risk of CHD, it has not a direct relationship to the rate of CHD, and it has not been studied in more than a dozen randomized trials.
But how then has the idea of the bad cholesterol emerged? As mentioned in the National Cholesterol Education Program, there are two main reasons. First, there was the discovery of a defective LDL-receptor in familial hypercholesterolemia and its consequence, the extremely high level of LDL-cholesterol in the blood of individuals with this disease. The discoverers, Nobel prize winners Michael Brown and Joseph Goldstein, suggested that the high LDL-cholesterol was the direct cause of the vascular changes seen in such individuals and also suggested that a similar mechanism was operating in the rest of us (22). Second, feeding experiments in animals raised the animals' LDL-cholesterol and produced vascular changes that have been called atherosclerosis by the experimentators.
These arguments are weak, however. If LDL-cholesterol were the devil himself LDL-cholesterol would clearly be a better predictor than total cholesterol, because the latter include also the ”good” HDL-cholesterol. And experiments on animals can only be suggestive and cannot prove anything about human diseases. Besides, the vascular findings in laboratory animals do not look like human atherosclerosis at all, and it is impossible to induce a heart attack in animals by diet alone (23). And finally, findings pertaining to people with a rare genetic error in cholesterol metabolism are not necessarily valid for the rest of us (24).
Thus, the experimentors claim support from unsupportive epidemiological and clinical studies, and the epidemiologists and the clinicians claim support from inconclusive experimental evidence. The victims of this miscarriage of justice are an innocent and useful molecular construction in our blood, producers and manufacturers of animal fat all over the world, and millions of healthy people who are frightened and badgered into eating a tedious and flavorless diet that is said to lower their bad cholesterol.
Read also:
Ravnskov U. High cholesterol may protect against infections and atherosclerosis recently published in Quarterly Journal of Medicine (2003;96:927-34).
.
