This essay was written by Lewis Wolpert and was first published in the 1996 Mill Hill Essays.
Life is full of risks. Getting out of the bath and opening champagne bottles should be done with care. For those with bad backs a sneeze or a forward lean can be dangerous. Medical treatments can have awful side effects. Our preoccupation with risk is something of a paradox as we now live healthier and longer lives than ever before. This apparent paradox might merely reflect that our anxieties about risk have remained constant, but we now have the leisure and good health to worry about risks other than such fundamental dangers as infantile mortality, infectious diseases and starvation. There is no evidence that previous generations were less risk-conscious – their anxieties were most likely focussed on different risks. If this is true it shows at once how personal and variable perceptions and worries about risks can be, and how unreliable.
I, for example, am a cyclist and a motorist. I fasten my seatbelt when I drive and wear a helmet on my bike to reduce the risk of injury. I am convinced that these are prudent safety measures. I have persuaded many friends to wear helmets on the grounds that transplant surgeons call those without helmets, “donors on wheels”. But a book on ‘Risk’ by my colleague John Adams has made me re-examine my convictions.
Adams has completely undermined my confidence in these apparently sensible precautions. What he has persuasively argued, particularly in relation to seat belts, is that the evidence that they do what they are supposed to do is very suspect. This is in spite of numerous claims that seat belts save many thousands of lives every year. Between 1970 and 1978 countries in which the wearing of seat belts is compulsory had on average about five percent more road accident deaths than before the introduction of the law. In the United Kingdom road deaths decreased steadily from about seven thousand a year in 1972 to just over four thosand in 1989. There is no evidence in the trend for any effect of the seat belt law that was introduced in 1983; there’s actually evidence that the number of cyclists and pedestrians killed increased by about ten percent. That twice as many children were killed in road accidents in 1922 as now must not be taken as evidence that there is less risk when children play in the street today. It almost certainly reflects the care taken by parents in keeping children off the streets.
How are these figures, which are both puzzling and shocking to be explained? The answer seems to lie in our perception of risk and how we modify our behaviour. An important concept that has been developed to account for peoples’ handling of risk is the “Thermostat Model”. An individual’s propensity to take risks is influenced by their own experience and that of others and this model assumes that the degree to which we take risks varies from one individual to another. The key feature in risk taking is the balancing of perceptions of the risk and the possible rewards, and this balance may be a reflection of an individual’s particular type of personality. In general the more risks an individual takes the greater will be both the positive and negative rewards.
Of particular importance in the model is the level at which the thermostat is set. There are those who are prepared to take risks and so their thermostat setting is high while for others, who are more cautious, the setting is much lower. So, for example, a driver going round a bend in a road will be influenced by rewards and risks. These could be getting to an appointment on time, impressing his companion, his concern for his own safety and that of a child in the rear seat, the cost of damaging the car and of losing his licence, and so on. He will also have taken into account the condition of the road and the amount of traffic as well as the kind of car he’s driving.
A very important feature of this model is risk compensation: people modify their behaviour in response to what they see as changes in risks to themselves. Thus, when we “belt-up”, we drive just that much more dangerously, so that the risk of an accident increases. And while we may survive better, pity the cyclists, pedestrians and back-seat passengers. As Adams wickedly suggests, one way to prevent this is to fit all cars with a spike on the steering wheel which is directed at the driver’s heart.
Risk compensation also probably works with cyclists, but information on this is not easy to interpret. In some countries, for example, laws requiring cyclists to wear helmets seem to have made some people too afraid to continue cycling. It is claimed that twice as many lives are lost by head injuries to elderly pedestrians than to all cyclists. Should they wear helmets?
There are at least three different kinds of risk. The first is easy to experience and recognise and can be called direct; an example is the danger presented by traffic or fears of being attacked. The second contains risks that have been identified with the aid of science, and diseases like smallpox and cholera fall into this category. The third class contains what are known as virtual risks, those about which science is unsure and so includes BSE and global warming.
Most of the studies that have been carried out on risk deal with risks of the first two categories. In both of these it is possible to calculate the risk associated with a particular activity like cycling in London or the effects of smoking. At least in these cases there is some objective information on which one can base one’s behaviour.
Even where risk can be calculated objectively for any particular event, different individual’s perceptions of that risk can be, and often are, quite different. For example, with respect to transport much greater attention is given to accidents that occur in planes, ships or trains, compared to those on the road, in motor cars, on bicycles or walking. There are two related reasons. In being transported by train or plane you are putting your trust in the organization or people who run them. One does expect them to have taken all the necessary precautions. But when you drive a car, for example, which is on all grounds much more dangerous, you have the conviction that you are in control. Thus the responsibility now rests on you or the driver who is usually yourself or someone you trust. Yet the actual number of deaths for one thousand million kilometers travelled is less than one for airlines and trains, around five for car drivers and passengers, one hundred for motor cyclists, fifty for cyclists and seventy for pedestrians. Even these figures disguise that we walk enormously fewer kilometers than we fly, so that what might be perceived as a greater danger in walking is in fact misleading.
There are thousands of potentially dangerous chemicals in our environment. Some natural, many manufactured. Testing how toxic they are for humans is fraught with difficulty. For example, tests are done on animals to find out if a substance will cause cancer and these can take up to four years and the costs begin to approach a million pounds. Even then there is no absolute guarantee that the results of the tests on animals will apply exactly to humans.
There are deaths associated with medical treatments. Anaesthesia deaths with surgery for example run at about five per million. By contrast some now abandoned medical treatments like flecamide for heart patients ran into the frightening thousands per million treated. It is worth remembering that all drugs have side effects. A tragic example is the drug thalidomide which was taken by pregnant mothers to help them sleep and which was tested without ill effects on rodents, but caused limb deformities in human babies at very low doses.
Some chemicals in the environment have known effects on health and the risk associated with this can be assessed. The most reliable way to find out what, if any, are the effects of an environmental agent if they are not already obvious is to record effects on representative sections of the population. Nuclear radiation is known to cause cancer, particularly leukemia. It is possible to calculate the effect of this radiation to which we are all naturally exposed in the United Kingdom, and it turns out to be responsible for about four percent of all deaths due to cancer. Sunlight’s effect is more striking and is responsible for the majority of skin cancers including melanoma. Smoking is about five hundred times more dangerous than pregnancy complications but the average loss of life expectancy is fifty years for women who die in childbirth but only fifteen years for men who die from smoking-related diseases. For men involved in traffic accidents the risk is twenty times less than smoking but the loss of life expectancy is thirty years.
As far as sports are concerned rock climbing is over one hundred times more dangerous than skiing. Tennis I’m pleased to say has so few fatalities that there are no statistics. There is a one hundred times greater chance of being killed by an accident in one’s home than by a terrorist bomb and even less by a fire in a public building. But just look at the extensive fire regulations which must be followed in hotels and department stores even though compared to car travel the risk of death from such fires is four thousand times less.
Our perception of risk is complex and may bear little relation to the objective probabilities. For example surveys have shown that there is a perception among the general population that childhood vaccination is ten times more dangerous than it actually is and the possibility of death from stroke is underestimated by a factor of ten. Also our common sense approach to risk is very unreliable. The risk of losing a few pounds on the lottery is very high and most people know that; but it is hard to understand why they think that if the pot is greater it is worth risking more. And who would not think it wise to bet on red in roulette if black had just come up ten times in a row. Some perceptions are quite irrational. For example I know mothers who when taking on a new au pair are very frightened that they will run off with the child if there has been even one such case reported. If I hear that a cyclist I know has been injured I am much more cautious than normal although my risk remains the same. Statistics have little impact on such risk assessments.
A major error in personal risk assessment probably relates to our inability to make sound judgments when the amount of information is limited. For example, a little while ago there was a suggestion in the United States that a game called ‘Dungeons and Dragons’ was risky since it might lead to teenage suicide. The evidence in support of this claim was that twenty-eight teenagers who regularly played the game had committed suicide. But the average suicide rate for teenagers nationwide in the United States is about one in ten thousand. Since some three million teenagers played the game, the number of suicides that might be expected among the players was three hundred so there was no evidence of a significant link between playing the game and suicide.
How often have we heard that so and so’s father smoked thirty a day until the age of eighty without a day’s illness. As if this has any relevance to the risk of smoking damaging one’s health. Isolated cases are not relevant. Worse still, beware the doctor who says that he knows that clinical trials have shown that a particular drug treatment is ineffective or damaging but that in his experience it works well. We are all too easily persuaded by own own ability to make sound judgments. I have very few colleagues who doubt their own ability to select the best medical students. How, one should ask, would they know?
The most reliable way to assess any medical treatment is by random clinical trials. The best and most satisfactory technique for doing this is called double blind testing in which the treatment is given to some patients and not to others. The patients are assigned randomly by the controller of the trials to one or other treatment, and neither the patients nor the doctors know who has been given which treatment; both, in this sense, are blind. This anonymity is essential because if either group knows what is going on, this can affect the outcome in all sorts of subtle ways.
While trusting our doctors perhaps we should all begin to take more responsibility for what treatment they offer us. Many of the treatments provided have never been validated by clinical trials and their effectiveness and risks must remain, for the moment, in doubt. A good beginning might be for all of us to enquire how enthusiastic our doctors are about evidence-based medicine. And if you are committed to alternative medicine do not expect an informative response.
Social scientists have identified four different types of attitude that people have with respect to risk and how it should be handled. Individualists try to control their own environment but oppose controls; egalitarians are also against controls but regard nature as something to be obeyed; hierarchists believe that nature must be managed and so like controls being imposed; and fatalists, who just try to duck when necessary.
Governments and their advisors have to balance such attitudes with the costs of precautions against actual risk and this can be very difficult. In making decisions about how to handle risk, they must make an assessment of the risk, the cost, and public perceptions – the balancing of which is itself risk ridden. So engineers and inspectors must make very careful assessments of what can go wrong. Their assessment of risk in complex situations like a nuclear plant or a chemical factory may well be quite a way out, but what else can they do? There is no complex situation which is risk-free; nevertheless there is a moral obligation to make that risk as low as possible within the constraints of money and reality. Cars would be driven much more safely, but very much more slowly with that spike sticking out of the steering wheel.
BSE presents a very clear example of the problems associated with virtual risks where there are few hard objective data on which to base probabilities and so precautions. This leads people to choose their behaviour even more freely and in ways that reflect their personalities. For the view that eating beef could be dangerous might, on a common sense view, appear to be based on flimsy evidence. The key finding was the sudden appearance of a quite new form of Creutzfeldt-Jakob disease. But there are thus far less than twenty cases. The science of neither the cause of the disease nor the way the infection shows itself over time, fit with common sense. It is generally accepted that BSE came from cattle eating infected sheep. The proposed infectious agent – and there is as yet no incontrovertible evidence – is thought to be a protein molecule called a prion which is quite unlike a virus or bacterium. But prions are known to have much difficulty crossing from one species to another, so what is the basis of the suggested danger that it could cause a new Creutzfeldt-Jakob disease in humans? Simply that at present there is no other explanation than it is due to infection with a BSE agent.
Less than twenty cases seems to be a very small number, so why all the anxiety? The anxiety comes from an understanding of the pattern of development of such diseases which have so called long incubation times. Mathematical predictions of how many cases will appear over the next few years are complex but could strike fear even into the heart of a sceptic. It may be that there will be just a few more cases but it is just as possible that there will be hundreds of thousands. No one knows.
BSE thus raises very difficult problems as to how such issues should be handled. Because of the extreme dangers one has to balance the risks against all social and financial costs of taking precautions. Government is about such difficult choices. Fire regulations require enormously expensive constructions, though fires are rare. Building the Thames Barrier is a precaution for a very rare event. The social responsibility of scientists is to make public information that can affect our lives, not to make ethical or political decisions as to what to do. It is not for the scientists to decide, and it is precisely such difficult decisions that politicians must take. It is also their moral and legal duty to take precautions.
Stopping at traffic lights the other day I mentioned to a fellow cyclist, a stranger, how dangerous cycling was. I had just narrowly escaped being run over. “Yes”, he replied, “but I bet you don’t eat beef. Yet the risk of being damaged by cycling is so much greater”. “Of course” I said, “but so are the benefits. I would be really depressed if I were to give up cycling, but giving up beef for a year or so until the evidence is clearer is no trouble at all. It’s a matter of risk benefit”. Happily he agreed, and we pedalled our separate ways, my helmet firmly on my head. Life is a risky business.