by Steve Hickey
Epidemiology points the way, giving an indication that a relationship between smoking and cancer exists, but by itself it does not provide a scientific explanation for that relationship. Given enough factors (or perhaps enough epidemiologists), spurious relationships can always be found. For example, since 1950, the number of televisions has increased in line with the level of carbon dioxide in the atmosphere. Can we conclude, therefore, that carbon dioxide causes televisions? The answer is clearly “no”—an example of the statistical rule that “correlation does not imply causation.” In common language, just because two factors occur together does not necessarily mean that one causes the other.
Hill’s Rules
Bradford Hill and Edward Kennaway, of St. Bartholomew’s Hospital, London, investigated the epidemiology of smoking and lung cancer in 1947. Richard Doll joined the investigation slightly later. Doll later became a consultant to chemical and asbestos companies, who funded his research. His findings were highly criticized for underestimating the harm such products caused.7 Ultimately, Doll’s reputation suffered when this commercial involvement and potential bias was exposed, but Hill is regarded as having been “the world’s leading medical statistician.”8
It is perhaps sobering to remember that, at the time, smoking was common and few believed it could be associated with the disease. Reportedly, however, the Germans had already identified smoking as a cause of cancer.9 over the next forty years, it became apparent that smoking twenty-five or more cigarettes a day increased the risk of lung cancer by twenty-five times while lowering bodily reserves of vitamin C.10 We can now state clearly that smoking tobacco promotes cancer, because, in addition to the epidemiology, we have an explanation of the processes involved from the basic physics, chemistry, and physiology.
Hill was aware of the limitations of epidemiology—unless the statistics are rigorously applied and the limitations of the data openly exposed, epidemiology can mislead more often than it informs. For this reason, Hill provided a set of criteria or rules that must be met before a causal link could be inferred.
• Plausibility: The measured correlation must be biologically plausible and there must be a rational, theoretical explanation for the phenomenon. This rule means that epidemiology (indeed, all clinical sciences) should conform to the underlying physiology and biochemistry.
• Strength of association: The observed relationship or correlation must be strong. Weak relationships constitute feeble evidence. Unfortunately, many medical claims on chronic illness and nutrition are based on weak relationships, such as the link between dietary cholesterol and heart disease. This rule might be taken as a caution against relationships that are commonly expressed as “percentage risk.” It could also be another way of saying that if you need an enormous population to detect an effect, then that effect is probably too small to worry about.
• Timing: The proposed cause should precede the effect; also, the effect should be consistent with time. If the consumption of a substance varies over the years, then so should the associated disease. If the cause is removed and reintroduced, then the effect should vary.
• Dose-response relationship: The effect should increase with the intake or dose of the proposed causal factor (the more exposure to the substance, the greater should be the measured effect). We might add that the results should not be extrapolated outside the range of doses covered in the trial. This is a critical error in trials of vitamin C, where the dose employed is often less than 1 percent of the intake claimed to be effective. We will see how misrepresentation of dose levels has led to medicine’s failure to investigate the claims for vitamin C and the common cold.
• Consistency: The relationship should be consistent when trials are repeated. The idea may be considered weak or even abandoned if subsequent results refute the suggestion. This rule suggests that the finding needs to be replicated, which is the basis of the scientific method.
• Coherence: The claimed effect must be consistent with scientific knowledge and should not conflict with other theories. If someone else has an alternative idea, it may provide a better explanation.
• Analogy: A commonly accepted phenomenon in one area can sometimes be used in another. The levels of ascorbic acid in animals that synthesize the vitamin are equivalent to an intake of several grams a day for a human. By analogy, the range of vitamin C intakes studied should encompass these doses but it seldom does.
• Experimental evidence: A proposed relationship should be shown independently, by experiment. Supporting experimental evidence from biochemistry, physics, or physiology greatly increases the plausibility of the proposed link.
• Singular cause: There should be only one cause rather than a list of risk factors. This final requirement is a blow to the proponents of multiple epidemiological risk factors for disease.
According to Hill, a pioneer of both the randomized clinical trial and epidemiology in medicine, all these criteria should be met before causation is assumed. Scientists in other disciplines might suggest that these conditions would be a minimal requirement for even a tentative indication of a causal relationship. These rules are basically common sense applied to data obtained from social and statistical studies of populations.
Epidemiology can be a powerful scientific tool, but it is often poorly applied. Hill’s rules are rarely applied in modern epidemiology and the result is a continuous flood of apparently contradictory information. For example, in 1981, it was announced that coffee causes cancer of the pancreas and this might explain a large proportion of the cases in the United States.11 Later, this well-publicized study was largely refuted.12 Indeed, there are now indications that coffee might prevent other cancers.13 True science tries to find underlying mechanisms and models explaining phenomena by cause and effect. Limitations of currently fashionable methods of medical investigation can severely hamper scientific progress.
Vitamin C and the Difficulties of Social Medicine
The increasing emphasis on risk factors and medicine as a social science can prevent proper investigation of the effects of vitamin C. Analysis of small, minimally effective doses has resulted in mixed findings and has spread confusion about ascorbate’s true potential. The public release of such results leads to a gradual discrediting of the scientific method (misapplied epidemiology). Almost anything one does is supposedly implicated in one illness or another, causing anxiety and confusion in the minds of the public. Most people’s vitamin C intake is low and population studies are based on these intakes. The high doses that might be effective in disease prevention are rarely investigated. Consequently, clinical trials and epidemiology are unlikely to show a clear link with vitamin C, even if shortage of it caused the disease.
Over the last half century, vitamin C studies have separated into those that investigate large doses and conventional research limited to micronutrient intakes. A factor central to the establishment’s misrepresentation of the role of vitamin C has been the rise of social medicine, the idea that disease is a product of our social activities. An epidemiologist studies the population to investigate the habits of people and to contrast the diseased with the healthy. In one study, people with heart disease may have consumed more animal fat, a second study suggests lack of vitamins C or E, while another investigation implicates sugar. Subjects like environmental biology and ecology have the reputation for being soft sciences. However, the very nature of these subjects involves complexity and it can be hard to make simplifications and obtain general laws. When Charles Darwin claimed that “I am turned into a sort of machine for observing facts and grinding out conclusions,” he was describing a particular form of genius that cannot be replaced by statistical analysis.
The design of accurate epidemiological studies presents great difficulty.14 The first problem with epidemiology is choosing what factors are to be measured. This is a fundamental problem, not appreciated by many researchers. The problem of having a large group of potential factors is given a highly descriptive name in decision scienc
e—the curse of dimensionality. Paradoxically, beyond a certain point, using more factors decreases the predictive accuracy of the statistics: the more individual measures a study includes, the less accurate will be the result.
Suppose we wish to find dietary factors, such as vitamin C, that might relate to heart disease. There are thousands of potential candidates, from apples to zinc. Measuring each of these has an associated cost, as well as presenting practical difficulties. It is hard to determine accurately how much salt 10,000 people each consume over a period of a year, for example. It is clearly impractical to sample and measure every item in an individual’s diet. Even if the population were highly constrained—for example, just soldiers consuming army food—the problems remain. Even soldiers have a choice of the food that they actually eat. Soldier A might hate broccoli, but could have a sweet tooth and love apple pie, whereas Soldier B never eats dessert but gets a food package every month from his family.
Often a researcher’s solution is to simply ask the subjects what they eat by questionnaire and then to estimate the amounts of component nutrients from standard tables. For example, the investigator might assume an apple weighs 100 grams and contains 25 mg of vitamin C. However, even in this case that is apparently easy to estimate, values are subject to large errors. Apples come in different sizes and varieties, and they are stored for varying periods and subject to dissimilar treatments and methods of transport. Some are organic while others are mass-produced, and the amount of vitamin C in fruits and vegetables varies widely. Moreover, a subject reporting one apple a day may be confused, have a poor memory, or simply be lying.
The Framingham Study is one of the biggest medical investigations of the twentieth century and was highly influential on the growth of social medicine.15 Before Framingham, clinical investigations of disease tended to be small or descriptive case reports.16 Framingham began with support from the newly created U.S. National Heart Institute and the initial report of this long-term study, published in 1961, covered the first six years of following risk factors in the development of heart disease.17 The results suggested that high blood pressure, smoking, and high cholesterol levels were in some way associated with heart disease. These risk factors emerged, but provided little insight into the disease. The follow-up has continued and over fifty years of data collected from Framingham residents has helped generate over 1,000 scientific papers.18
Framingham identified some risks associated with heart disease and stroke, but it also created a revolution in medicine by changing the way we view the origins of disease. It is claimed that the Framingham research dispelled the myth of a single cause of heart disease and started the now popular concept of risk factors. The results persuaded many researchers to concentrate on these risk factors instead of looking for a direct cause of the disease, such as chronic scurvy.19 Unfortunately, Framingham data do not provide information that can help decide whether heart disease is a result of long-term vitamin C deficiency.
Back to Basic Science
History illustrates how medicine can go wrong by concentrating on epidemiology and clinical trials at the expense of primary scientific methods. In the nineteenth century, doctors attributed tuberculosis (TB), known as consumption or the white plague, to a combination of heredity or constitutional factors together with the miasma, or smells, in the environment. These risk factors, which seemed to explain how the disease was found to run in families, were thought to be the cause of TB.
It is easy to feel smug in the hindsight we have due to the benefit of accruing scientific knowledge. We now know that tuberculosis is a result of an infection that can lie dormant for years. There is a higher risk of contagion in confined spaces. Poorly nourished people with a low vitamin C intake, living in the same house as a TB sufferer, would have a greatly increased probability of catching the disease. With tuberculosis, the risk factors were a byproduct of the infectious process, but were mistaken as causes.
Bacteria cause tuberculosis, and it was Robert Koch who described the bacillus causing TB, Mycobacterium tuberculosis, in 1882. Mycobacterium tuberculosis is a slow-growing, aerobic bacterium that can divide every 16–20 hours. Koch discovered the cause of TB in the laboratory, after developing a new technique to stain the bacteria for microscopic identification. He won the Nobel Prize in Physiology or Medicine in 1905 for this discovery. Koch’s success illustrates how disease mechanisms are discovered using basic biological science rather than social medicine. By concentrating on basic experimental science, Koch was able to demonstrate the single primary cause. Once the cause was understood, it explained the association with the risk factors.
John Snow, a London doctor, is sometimes described as the father of epidemiology. However, Snow’s work on the London cholera epidemic in the 1850s was based on a new, theoretical understanding of the cause of disease. In the mid-nineteenth century, the primary risk factor for infection was thought to be bad smells. The name malaria, literally meaning “bad air,” comes from this idea and remains as the modern name for the mosquito-borne disease. It is easy to see how infection was associated with bad smells—drinking contaminated malodorous water might cause disease and infected wounds would often release a putrid smell. While doctors assumed that bad smells caused infectious diseases like cholera, Snow’s approach was an early form of germ theory.20
This led him to stop an outbreak of cholera in London.21 By tracing the number of local cholera victims on a map, Snow noticed an association with a local well in Broad Street, Soho. Other people also produced maps of infection, using the data to support the miasma theory of disease.22 Patients with the disease were marked with spots on the maps. More than fifty years before Snow, Valentine Seaman had used so-called spot maps to report deaths from yellow fever in New York.23 Both the doctors who believed in contagion and their opponents used these maps to advance their respective causes.24
Snow’s achievement arose because of his theoretical understanding of infection. His use of the map and “epidemiology” was to provide data to support his idea that special animal poisons spread infections. We now call these poisons “germs.” The water in the Soho well was favored for its clarity and taste. For a doctor following the miasma theory of disease, it would be an unlikely place to look for the cause, because clear fresh water is odorless and thus inconsistent with the miasma theory of disease. Snow’s adoption of the germ theory led him to make his discovery. He removed the pump from the well and the epidemic subsided.
Close to John Snow lived another pioneer in health, the engineer Joseph Bazalgette (1819–1891). Like Snow, Bazalgette was to lead the way in disease prevention and, arguably, he made the larger practical contribution. He designed and built the London sewers to rid the city of the intolerable smell and miasma of the river Thames, and thus prevent disease.25 Sewers brought sanitation to London and, eventually, to the world. Serendipitously, Bazalgette’s response to the miasma theory prevented the deaths of millions. Working from the unsound association of disease with bad smells, engineering may have saved more lives than any physician. Bazalgette’s ultimate achievement came about through serendipity, a correct result based on faulty reasoning.
Today, we expect a response to illness based on solid scientific evidence and, particularly, the biological mechanisms underlying the disease. It is clearly insufficient to have a limited practical understanding based on risk factors and hope we end up, like Bazalgette, preventing disease by chance. Snow showed the way, not by mapping the disease, which was an established approach to following epidemics, but by utilizing his knowledge of disease mechanisms.
The mechanisms for the antiviral and anticancer actions of massive doses of vitamin C are already established.26 When combined with clinical observations from physicians reporting remarkable effects, there is sufficient data to demand clinical trials. Snow’s work was based on a new understanding of disease. But even now, people promoting the risk factor approach appropriate his success, describing him as the founding father of modern epidemiology.
In Snow’s day, his germ-based ideas had few supporters compared with the miasma theory. Similarly, clinical trials of vitamin C in large doses do not fall within the mindset of current medical opinion, which is dominated by ideas of science as sociology.
Whenever medical problems and disease are described in terms of risk factors, the biological understanding is deficient. Finding new scientific explanations is difficult. It often relies on the ability of an individual scientist to see through the mess of risk factors in order to provide a simple model or theory. Most medical problems have a simple explanation. Shortage of vitamin C may be the underlying cause of heart disease, arthritis, and many other chronic conditions facing modern humans, but medicine’s risk factor approach to studying disease today is unlikely to be able to confirm or refute this idea in the foreseeable future.
Limiting True Progress
Social science and genetics now have increasing preeminence in medicine to the detriment of true progress. This emphasis on social medicine is a relatively new phenomenon, popularized since the 1950s. Unfortunately, the further medicine gets from a physiological or biochemical understanding of disease, the less scientific progress is made. Since this new approach was adopted, medical progress on the leading causes of disease has slowed.
The idea that disease is a product of our social activities dominates current medicine. Social medicine is a pragmatic form of clinical science that downgrades the importance of theory, biochemistry, and physiology. Large-scale studies have become a preferred form of research—identifying the relative importance of minor risk factors. While such detailed scientific information gathering from populations is useful, it can also constrain the progress of medical science. Scientific advancement typically depends on developing new theories based on interpretation of the physical evidence. A major advance in molecular biology in the latter half of the twentieth century was the double helix model for the structure of DNA, which resulted from the work of Rosalind Franklin, Frances Crick, and James D. Watson in the early 1950s. This simple theory enabled rapid advances to be made in genetics and cell biology. Currently, however, medicine seems to abhor biological theory and understanding, preferring to study statistical relationships between social influences, risk factors, and disease. Such science fails to describe the mechanisms involved. Consequently, the understanding needed to solve medical problems becomes increasingly difficult to achieve.