This muddle is just the latest to beset epidemiology, the science of "dying by numbers". Epidemiology looks for links between a disease and its victims' lifestyle and environment. It has a noble lineage. Dr John Snow's study of cholera outbreaks in 19th-century Soho showed that people died after drinking from particular wells. He realised that cholera was transmitted by infected water - an insight that revolutionised medicine. In recent times, epidemiology has focused on non-infectious disease. It has unmasked the dangers of smoking, asbestos, ultraviolet light and many other risks. In other words, epidemiology is more than just "lies, damn lies and statistics".
Now it seems that most of the high-risk hazards have been discovered. A recent report in the American journal Science entitled "Epidemiology faces its limits" points out that no more than a dozen or so agents have been repeatedly linked to cancer: cigarette smoke, alcohol, radiation, a few drugs, a handful of occupational hazards, such as asbestos, and three viruses. The search to uncover the smaller risk factors continues, as the huge number of epidemiology studies published every year shows. However, recent studies have often been contradictory and have linked cancer with anything from yoghurt to stress. Epidemiology is grinding to a halt and the public is losing confidence. Is there a way forward?
To carry out an ideal study you would first select a large group of people from a wide range of backgrounds. Expose half of them to the potential risk you are interested in, and then wait and see how many more people in the risk group die than those in the control group. However, the ethical problems of putting healthy people in possible danger in this way are obvious. So, instead, epidemiologists compare the lifestyles and environments of people suffering from a disease with those of unaffected people.
The problem is that many diseases, such as heart disease, are linked to low income, since low income is linked to many other factors. Thus, poor housing, smoking, poor diet and alcohol consumption may all appear to "cause" a whole host of diseases, although only one, or even none, of these factors may be the true cause.
But what can be done instead? The challenge for the future is to uncover the smaller risks that remain significant because they are widespread and affect many people. Several researchers have suggested that for a study to be taken seriously, it should demonstrate that a factor increases the risk of disease at least three or four fold. But smaller risks are still important - if they are real. The size and especially the design of the study should also be considered critically, but, ideally, an effect should be consistently demonstrated in a number of studies using a variety of research methods. Most importantly, a plausible biological mechanism for the effect must exist if the link is real, and proof for this mechanism requires an experimental approach in the laboratory. Indeed, the recent fear that two boys who have developed brain-wasting Creutzfeld-Jakob disease may have caught it from beef infected with BSE points out the limits of epidemiology. Two cases are not enough for epidemiologists to start looking at the cause of their illness. The biological evidence, however, is much more worrying. BSE can be transmitted from sheep to cows and on to cats and zoo animals. Evolutionarily speaking, are we distant enough to escape infection?
The real strength of epidemiology is, paradoxically, its biggest weakness. It functions in the real world, dealing, as Sir Richard Body, an Oxford University epidemiologist puts it, "with things that acutely affect people. The controversy it may cause is nothing to be frightened of." Unfortunately, something that starts as a robust debate in the scientific world may cause a public panic. The Science report speaks about how scientists and the media may dangerously amplify a study's importance. Researchers feel they have to show that a risk exists to get their work published and justify their funding. (After all, who would remember the scientist who discovered that vegetables did not cause cancer? And who remembers the study that showed that saccharine does not cause bladder cancer?) The scientific journals then play up the results to the newspapers to get their names in lights, and many of the newspapers will, in turn, as last week's events showed, cause as big a sensation as they can. The end result is that the public is not well served, and the scientific community starts to lose its authority.
As far as the Pill and thrombosis scare is concerned, what lessons can we apply? Three studies have shown that women taking third-generation oral contraceptives are more likely to have a thrombosis. The findings therefore have consistency. However, the numbers studied in two of the studies were small. One compared 4,000 women with cardiovascular disease with 12,000 healthy women. Another compared 1,000 women with thrombosis with 3,000 controls. It is difficult to see how such studies could detect that certain brands double the risk of thrombosis from 1:7,000 to 1:3,500. The third study apparently looked at "hundreds of thousands of women taking the Pill" and is therefore more convincing. Unfortunately, like the other two studies, it has not been published, so it is difficult to assess its validity.
Should the Government have publicised the findings of these reports? Professor Walter Spitzer and Dr Hershel Jack, the scientists in charge of two of them, don't think so. As the British Medical Journal pointed out last week, even if the third-generation Pills do increase the risk of thrombosis, this danger may be more than outweighed by the lessened risk of heart attacks and strokes, which are far more likely to be fatal.
This case illustrates the conundrum facing epidemiologists who risk being accused of either sensationalising their work or covering it up. Whatever happens, we can be sure that epidemiology will continue to be controversial because, in the words of Sir Richard Body, "it deals with real people, real worries and real diseases".Reuse content