Risk, science and society
One of the UK's top scientists explains why the precautionary principle - the substitution of prejudice for data - leads to irrational convictions.
- Public discussions of risk consistently overestimate dangers and undervalue the benefits we obtain by living in a complex society.
- It is difficult to modify this mindset, and there is little point in re-exploring the arguments about what is ‘acceptable’ as a risk, since the ground shifts constantly in terms of our perception. For example, food fads may be dangerous or beneficial at the same time, depending on the viewpoint of the protagonist.
- That benefits are taken for granted is probably a sign of a healthy mindset. But this complacency should have limits. Those who do not find anything worthy of note in the achievements of medicine should think, for example, about the disappearance of poliomyelitis from the UK.
- Safety is not a concept that can be extended to the whole of society, as what makes something safer for one group will often make it more dangerous for another. How do we generate appropriate attitudes? Can science help?
- The objectives of safe management of our environment depend on caution – the taking of heed, precaution – and the exhibition of prudent foresight. But they cannot include an indemnity – an assurance that no-one will come to harm from any action.
- Many non-scientific commentators adopt a position and accept data that support their argument uncritically – they do not examine it, nor do they look at unpalatable data. Uncritical belief is one definition of faith – and faith is not affected by argument. This is what we must attempt to change.
- Information is probably the answer, but it must be provided in a way that allows the requirements of society to be reflected in the content. We must seek agreement about the desirable outcome we wish to achieve – which environments we protect, what birds must be favoured, how long we should live. In all of these instances, some will hold opposing views.
- Acute episodes require prompt and effective responses, not commissions of enquiry.
In the context of the safety of our day-to-day environment, we have become highly risk-averse. Our obsession with very small risks has reached a stage that results in damage to society.
Further, the debate we have about these problems has a comparable value and intellectual content to the often-cited discussions in Byzantium about how many angels could dance on the head of a pin. It often consists of a reiteration of the prepared data of pressure groups, which often ignore unfavourable data where they are non-supportive; and a series of exhortations about what should be done to make things safe – or a statement (from the opposing viewpoint) that they already are.
It is not possible to demonstrate that something is safe. As Godel’s theorem makes clear, all that can be demonstrated is that some thing or activity has not yet shown evidence of the production of harm. Therefore, it might be proper to regard that thing or activity with caution, despite the evidence in its favour so far (1).
Further, ‘safety’ is a difficult concept. Actions that make something safer for one group will almost invariably make it more dangerous for another. In Siddal’s excellent definition – ‘Safety is the degree to which temporary ill health or injury, or chronic or permanent ill health or injury, or death, are controlled, avoided, prevented or made less frequent or less probable in a group of people’ (2) (the emphasis is my own) – this factor is made clear. As an example, airbags in cars have undoubtedly saved lives, but they have killed people too.
Anxiety about an inevitable lack of certainty about the outcomes of exposure to particular events, infections, toxicants or environmental changes leads to irrational responses, which themselves produce further problems.
If we consider the data on the effects of immunisations against the acute specific fevers of infancy and childhood, it is clear that these immunisations have been a major influence in declining mortality in the very young and have thus had a profound effect on life expectancy. More difficult to quantify, but also important, are the significant effects on morbidity in adult life.
A currently relevant example is the controversy about the MMR (measles, mumps and rubella) vaccine. Measles is highly contagious and occurs in outbreaks in communities with immunisation rates much below 75 percent. The illness will be accompanied by ear infection in 1 in 20 cases; by pneumonia or bronchitis in 1 in 25 cases (with some permanent sequelae in terms of lung disease); by convulsions in 1 in 200 cases; meningitis or encephalitis in 1 in 1000 cases; death in 1 in 2500-4000 cases; and the terrible problem of sub-acute sclerosing panencephalitis in 1 in 8000 children (3).
Anxiety about a poorly supported hypothesis regarding the association of the MMR vaccine with autism has led to a reduction in immunisation rates – such that an epidemic of the disease may be expected in the UK (outbreaks have already occurred in Sweden, the Netherlands and Ireland, with deaths recorded). This is a truly damaging response to problematic data.
Sadly, it is a repeat of an old story. Mistaken concerns about pertussis vaccine and encephalitis in the 1970s caused a drop in immunisation rates (from 79 percent to 31 percent), resulting in an epidemic with 28 deaths and a great deal of misery (see (4) for further data).
Real (if limited) problems with vaccines exist. The disappearance of poliomyelitis from the UK, following the introduction of the killed Salk vaccine and the eradication of the disease by the use of subsequent attenuated live vaccine, has conferred inestimable benefits on us all (Figure 1).
However, the dangers of live vaccines are demonstrated by a recent outbreak of paralytic polio in the Dominican Republic and Haiti, resulting from use of the oral polio vaccine (5). A live vaccine has the capacity to alter in a way that allows the possibility of infection in particular individuals. Nevertheless, the risk/benefit analysis produces a clear result for us all, as this event appears to occur at a rate of less than 1 in 2,000,000 doses. This is a clear demonstration of the inability to intervene in any situation in a way that produces benefits for all, and penalties for none.
Non-rational responses may determine that predictable and adverse outcomes follow well-intentioned but uncritical decisions: as demonstrated by the response to recent rail accidents and their related deaths. As Figure 2 from The Economist indicates (6), if the ban on high-speed travel resulted in a significant increase in road travel (as it appeared to), then it is likely that death rates from travel will have increased, so great is the difference in road and rail safety in terms of distance covered.
Part of our irrationality is a disregard of benefit – but this has a number of components. Some part is straightforward – we consider the probability of a child dying of measles to be so remote that we can ignore it. Other components are more complex, and can be regarded as societal benefits of which we are aware, but to which we do not relate directly.
Figure 3, from data used by Stephen Jay Gould to make a different point (7), illustrates the increase in weight of batters and pitchers in the American baseball leagues over a period of around 100 years. In this time, the rules of the game have not changed in a way that would affect the optimal physique of successful players (in the way that recent changes in rugby football have done), and the apparatus used in the game has not altered radically (as in golf). What the graph illustrates is the effect of better nutrition in a population: the triumph of US agriculture and the effect of effective food distribution.
Similar changes occur in the population as a whole, and are reflected in the increase in height in the UK, a process which started later here than in the USA, and that is only now becoming asymptotic, in terms of the curve of rate of gain in height per decade. The change is more evident in males who (in biological terms) gain greater reproductive advantage from increase in size; in females, an earlier onset of menses is the best-defined comparable effect. Increased foot size may reflect greater skeletal size, which confers benefits in the ease and success in delivery of a viable fetus.
It is not surprising that we do not note these benefits, as they are not measurable for the individual. It is incontestable that the two factors of reduction in neonatal and infant mortality and better nutrition underlie the biologically recent and dramatic increase in life expectancy – from 44.1 years in males and 47.6 years in females in 1900, to 75.1 and 80 years respectively in 1999. But individual appreciation of this benefit will be limited.
Nature seldom confers an unmitigated benefit. Risks are attached to better growth – it has been suggested that the earlier onset of periods seen in adolescence indicates developments that may underlie the increase in carcinoma of the breast that is seen in some countries. However, this effect is not marked when compared with the real risk of good quality food – overnutrition.
Apart from the problems of obesity, Lutz (8) and Hart (9) have argued convincingly that neither natural nor additive carcinogens explain food-related carcinogenic risks – overnutrition per se is the problem. A great deal of animal data supports this conclusion (see also Shimokawa (10)); it is well established that animals with a restricted diet develop fewer ‘spontaneous’ neoplasms than those fed ad libitum.
Benefits as ephemera
It is a normal and healthy thing to take most of our advantages for granted. It is important to remember, however, that these advantages may disappear very rapidly if we cease to apply the techniques that have proved to be of value. For example, our food supply is not secured without care; crop losses from infestations or locusts continue to occur in other countries, and recent problems with Colorado beetle have had dramatic effects in the former Soviet Union.
Suggestions that organic production of vegetables and fruit may confer a net benefit on consumers, by avoiding the use of herbicides and pesticides, are difficult to sustain. Apart from the well-documented dangers to farmers of a consequent increase in the use of machinery (11), there are specific difficulties for populations in this approach.
Organic crops present problems with quality (of varying importance – scab, sooty blotch, size variation in apples), and in the production of a consistent and predictable supply to consumers. Other problems exist for the farmer, since organic production is demanding and leads to specialisation in the production of a particular crop. This produces a subsequent vulnerability to epidemics, or to singular events in the microclimate.
There is another problem with organic crops. Most accept the greater wastage implicit in the use of this methodology, but few realise that it should be greater still. This is illustrated in Figure 4, derived from data reported by Mongeau et al (12) on the production of carcinogens (furocoumarins) in infected parsnips.
Healthy roots produce little of these toxins, but when infected they are found in greatly increased amounts throughout the parsnip, not simply in the visually altered areas (fungicide-treated roots do not develop significant levels). Cutting out the damaged bits of spoiled vegetables is not sensible, any more than cutting the eyes out of sprouting potatoes would be if the concern were the ingestion of potentially dangerous toxins (in sprouting potatoes, glycoalkaloids are the problem).
We are very much given to ‘solutions’. Eating the right diet and a proper exercise programme will apparently allow us to live forever. These convictions provide a background against which ill-founded claims attain the status of decrees that all will be well, if only a particular course of action is followed.
Consider the example of soy proteins. Newspaper articles repeatedly emphasise the estrogenic effects of soy, which are presumed to be beneficial. The nonsense that Japanese women do not get hot flushes is found in many articles. The firm conviction that soy proteins are good for you leads to a disregard for the studies of Setchell et al (13). These studies show levels of estrogens in male infants that are fed soy-based infant formulae, which are 6 to 11 times the levels that have hormonal effects in adults. Calculations show that, in early life, levels 13,000 to 22,000 times normal plasma estradiol concentrations may be reached.
Further, genistin produces structural chromosomal alteration in human lymphocytes (14). An epidemiological study by North et al (15) has shown that hypospadias (a defect of formation of the male urethra) is increased in frequency in the male offspring of women who eat soy proteins in pregnancy.
We have here all of the ingredients for a food scare (hormonal effects, what is usually interpreted by the press as evidence of carcinogenicity, and data suggesting teratogencity in humans). Yet soy is still a Good Thing in 1066 and All That terms. Whether it is or is not a Good Thing is far from clear. But the uncritical recommendation of soy as a dietary supplement is foolish.
In the same way, we have accepted that the data that indicate that anti-oxidants are useful in preventing cell damage, or inhibiting various pathogenetic processes in experimental circumstances, can be translated into dietary recommendations.
Trials of anti-oxidants have been unrewarding. In one study of vitamin E supplementation in approximately 7000 men and 2500 women at high risk of heart disease, no evidence of an effect could be found, with no significant differences in the numbers of deaths from cardiovascular causes or stroke (or death from any cause, (16)). Most studies have been ineffectual, and in some there has apparently been a demonstration of adverse effects in the test group (17, 18) for good scientific reasons – see Paoloini et al (19) and (20).
The general problem is that the data that justify interventions in defined experiments are inadequate justifications for dietary interventions that take the form of advice to populations. Two points should be made here.
First, the adverse events that we seek to avoid are often hazards and are thus unquantified. They are ‘exposures that can, in particular circumstances, lead to harm’.
In most instances, the concerns relate to chronic diseases in which the causal mechanisms are unclear (arthritis), or, if well understood, where the nature of the environmental exposure-causing concern is poorly defined (carcinoma of the colon). Links between dietary factors and cancer are often sought: but where, as in carcinoma of the colon, the event targeted is a low-frequency occurrence (in population per year terms), it is not surprising that interventions based on encouraging people to eat fibre, in order to prevent carcinoma of the colon, have proved ineffectual.
The ‘particular circumstances’ referred to in the definition of hazard include genetic background (often identified from family history), and singular exposures, such as an unusual environment or diet.
Second, risk – if properly defined as ‘the probability that a particular adverse event occurs during a stated period of time, or results from a particular challenge’ – is very difficult to measure in low-frequency events in populations. The standard technique of looking for changes in rates of those events that have been documented following an intervention is, therefore, rarely successful.
This general conceptual problem about risks has been further confounded by the proposed use of prejudice as a substitute for data, in the form of the precautionary principle (PP). This is damaging since it eschews the scientific method in the way that it deals with data.
Holm and Harris (21) have defined the PP in this way: ‘When an activity raises threats of serious or irreversible harm to human health or the environment, precautionary measures that prevent the possibility of harm shall be taken even if the causal link between the activity and the possible harm has not been proven or the causal link is weak and the harm is unlikely to occur.’
This principle should never be used in evaluating data relating to populations. It arbitrarily changes the weight that is given to evidence from different investigations on an uncertain basis, and represents the antithesis of science. If you decide that some data are more important than others, you are deciding that you know what is best and, in the framework we are discussing, that the outcome of particular interventions will be beneficial.
Apart from soy and anti-oxidants, what about changes in abattoir practice and foot-and-mouth disease, or the decision to alter the rendering practice (temperature and solvent use) that may have contributed to the BSE problem?
What are we trying to do?
The objectives of safe management of our environment depend on caution – the taking of heed, precaution – and the exhibition of prudent foresight. But they cannot include an indemnity – an assurance that nobody will come to harm from any action.
Reasons for this have been touched upon, but a further and increasingly important factor that begins to influence our decision-making is the greater understanding of our own genetic variability. Many parts of the population are at different risks from different exposures because of their genetic make-up.
In the eastern Mediterranean there is a high incidence of glucose-6-phosphate dehydrogenase deficiency, and individuals with this deficiency will develop an haemolytic anaemia when given some drugs. This is not a reason to ban sulphonamides (a trigger for the haemolytic process) – but it is a reason for caution in their use.
This idea of an individual vulnerability makes regulation for populations difficult, although it may well account for the failure of some mass dietary interventions. It is a further example of issues relating to population benefit and individual harm.
In a major area of health concern – degenerative vascular disease – individual variability in susceptibility is increasingly recognised, and the role of nutrition in early life is known to be critical. We first became interested in the effects of early life on vascular development, and thus the long-term risks of degenerative vascular disease in individuals, in the 1970s, when I was able to study a number of cases of the single umbilical artery syndrome (22).
Although not certain of the arrangement of the vessels at that time (the cases we studied were living and had been identified by sectioning their umbilical cord), it was evident to us that the plan of the circulation was altered. All blood to the placenta must flow through one common and internal iliac artery to the placenta, rather than the normal two.
In children known to have had one umbilical artery, we measured vessel performance in the segments of the arteries of the trunk and leg on each side during life, and were able to show a profound difference in the two legs. One side had a performance typical of an elastic artery, the other of a muscular vessel (in the segment considered both should be elastic – like the aorta). We did not know which side had failed to develop an umbilical artery, but assumed that the high-flow side would be elastic in type, and the low-flow side muscular, to ‘match’ their functional status.
Subsequent studies of cases of the syndrome, where death had occurred before or soon after birth, confirmed that the altered flow during development had altered the structure of the vessels in a fundamental way, with a typical elastic structure being replaced by a muscular vessel on the low-flow side.
This is a gross example where the alteration of flow is marked. There is good evidence from a number of studies from our own and other groups, eg, Martyn (23), which have shown that there is a close link between developmental status and vessel performance, in population terms.
The important epidemiological studies of Barker and his colleagues (24), (25), (26) first documented the non-homogeneous distribution of ischaemic heart disease in the UK, and related this distribution to birth weight and weight at one year. Records were used that often covered a 40- to 50-year period (see Figure 5), and a powerful association between Standardised Mortality Rates for ischaemic heart disease and these weights was found. These effects have been found in other countries and cultures.
Better nutrition, via its effects on maternal nutrition, helps to explain the fall in premature ischaemic heart disease mortality seen in the USA and UK – a change that antedated emphasis on lipid intake.
Thus we grow vessels that match our blood pressure and which will be variably susceptible to disease. Measuring these changes in populations will better enable us to advise a properly defined risk population, and this ‘targeting’ may improve the efficiency of this kind of advice. It is clear that we could better direct our advice to those who are at risk of cardiovascular disease.
Ebrahim and Smith (27) studied the effectiveness of multiple risk-factor intervention in reducing cardiovascular risk factors, total mortality, and mortality from coronary heart disease among adults affected by the disease. They made a systematic review and meta-analysis of randomised controlled trials (on stopping smoking, taking exercise, dietary advice, weight control, use of anti-hypertensive and cholesterol-lowering drugs), and were able to pool more than 900,000 man-years of observation.
The pooled effects of multiple risk-factor intervention on mortality were insignificant, although a small benefit of treatment may have been missed by their methodology. The evidence suggests that such interventions implemented through standard health education methods have limited use in the general population. We need to give more support to better defined risk groups.
I am not sanguine about the effects of this type of persuasion, for in general we are bad at long-term adherence to (often good) advice. This is true even with life-threatening disease. In coronial practise, up to 40 percent of suddenly dead epileptics are found not to have taken their medication on toxicological study – a startling finding when one considers the social difficulties and humiliations that may accompany fits.
Changes in society
Personal attitudes vary, and are clearly changing in an important way. Barsky and Borus (28) have noted that somatisation – the reporting of somatic symptoms that have no pathological explanation – has increased rapidly in complex societies.
Changes in our attitudes have reduced the tolerance of mild symptoms and benign infirmities, and the threshold for seeking medical attention has been lowered. They also point out that physiological responses to anxiety (hyperventilation, palpitations) may be interpreted as illness. There are other drivers for this, but it is clear that this group of individuals are different from the rest of us (29) and that, interestingly, anti-depressants may resolve their problems, much as they may help in chronic pain (30, 31). Their symptoms are real.
For these, and other, reasons it is probably worth avoiding the term ‘acceptable risk’. The concept is applicable (and valuable) in the workplace, where exposures are measurable and understood, but it cannot work where we are dealing with hazards.
Even when real risks are defined, the view of what any individual sees as acceptable will change in any real-life situation, where those who are injured (or believe themselves to be injured – it makes little difference) will redefine what is acceptable in terms of their own experience.
I began by talking about angels and the head of a pin. In fact, there was no such debate in Byzantium, but the story persists. In the same way, other stories have persisted.
There was no childhood cancer from Alar (a much-criticised pesticide) (32), or as a result of radioactivity from Sellafield, but the problems are continually re-presented. What there was was the uncritical acceptance of a belief or beliefs promulgated by pressure groups – in effect, a declaration of faith. This type of declaration is not affected by data, and more or better studies do not seem to change views.
The characteristics of the hazard involved in generating anxiety will affect the extent to which the idea that there is a problem gains credence, in the public view. Nuclear power and nerve gas rate high in ‘dread’, while sunbathing and bicycle riding have low dread scores and are perceived as non-risky, despite considerable public knowledge of the long-term harm of the one, and the injuries associated with the other.
Baird (33) found that knowledge tended to increase risk tolerance, but there are studies which suggest the opposite for radon (34), and there are also data showing a lack of relationship between factual knowledge and risk perception in female college students with regard to toxic shock syndrome (35).
In general, there is an acceptance of risk if there is a perceived benefit – for herbal medicines, for example (36), despite their clearly documented dangers or dangerous interactions (37).
How to deal with the problem that a scare, once started, may persist and cause damage to our community? One way is for society to agree a set of objectives about any perceived risk – what values are unacceptable in differing contexts relating to age, sex, or occupation, for example, in the way we appear to have done for polio vaccine.
The alternative is to be rational in the scientific sense, and with this in mind the Social Issues Research Centre formed a working party that considered how scientific issues should be discussed by the media (38). It touched on the quality of the work and the reputation of the workers involved in the issue, the extent of peer review, the nature and reputation of the institution involved in the work, and so on.
I am sure that this is a helpful process and may work, for example, in those health-related issues where a directly interpretable outcome can be measured, such as cure or remission rates in chronic disease, relief of symptoms or of hypertension. This science-based approach must form the basis of decisions about regulation of xenobiotics (drugs, pesticides and industrial chemicals).
However, the evidence is that this approach will not work once an environmental or long-range hazard (in temporal terms) has been identified by a pressure group. A major problem is that the media will not undertake the rational analysis of the relevance of the data for population risks, in the way suggested many years ago by Bradford Hill (39), but will latch on to a story. Those data not supportive of the effect that has attracted attention (autism and MMR vaccine, or neoplasia from exposure to depleted uranium, say), or of a supposed benefit (soy proteins), will be disregarded.
Further, we are possessed by the difficulty that, in a liberal society, we would mostly support the concept that people should be free to do things that may harm themselves – but it is clear that we do not extend this permissiveness to causing harm to others. So perhaps we should not accept the possible return of rubella or measles – we might move to the position of some states in the USA where entry to school is conditional on an appropriate immunisation history. This is not a novel concept for the UK – we insist that medical students have to be immunised against Hepatitis B before beginning the course.
Presently, our society presents a strange mixture of attitudes. As a taste for public and ostentatious displays of compassion grows, there is a comparable enthusiasm for diminishing individual liberties by being prescriptive about some aspects of our life, including dietary preferences. I have pointed out that these, in particular, are often ill-founded, and may replace perceived hazard with established risk.
This determination to do good also extends to a firm view that no activity, regime, diet or game should do harm. I believe that we will have to develop a collective tolerance for individual damage, if we are not inadvertently to produce injury to the mass of the people by overzealous concern for the individual. It is our responsibility to affect the thinking of opinion-formers and, if we are serious about attempting to influence events, to do two (uncomfortable) things.
We must be ready – if possible – to respond quickly and in a ‘non-scientific’ context to current scares. (How much uranium do you have in your body anyway? How can the increment of depleted uranium 238 be important?) Secondly, some of us must avoid generating anxiety in order to increase our chance of getting the next grant.
It is a collective responsibility. Overemphasis on benefits or risks directed towards a sector of the economy or group within the population may result in harm to the many. Too many examples of this phenomenon already exist and we should avoid adding to them.
Professor Sir Colin Berry is professor of morbid anatomy and histopathology at Queen Mary, University of London. He is speaking on the subject of ‘The future of risk‘ at the spiked-conference Panic attack: Interrogating our obsession with risk, on Friday 9 May 2003, at the Royal Institution in London.
In praise of bad habits, by Dr Peter Marsh
Food science, by Tom Sanders
(1) M Guillen, Bridges to Infinity, Rider, London, Melbourne, Sydney, Auckland, Johannesburg, 1983
(2) E Sidall Risk, Fear and Public Safety. Atomic Energy of Canada Ltd, 1980
(3) Health Education Authority. MMR Immunisation Fact sheet, February 1997, p1
(4) H. Bedford, and D. Elliman, British Medical Journal, 2000, 320, 240-243
(5) Epidemiol Bull, 2000, 21(4), 13
(6) Anonymous, The Economist, 25 November 2000, p35-36
(7) SJ Gould, Life’s Grandeur. 1st ed Johanthan Cape, London, 1996, p244
(8) WK Lutz, and J Schlatter, Carcinogenesis, 1992, 13(12), 2211-6
(9) DA Hart, DA Newumann, and RT Robertson, Dietary restriction: Implications for the design and interpretation of toxicity and carcinogenicity studies, ILSI Press, Washington DC, 1995, 396
(10) I Shimokawa, et al. Ageing, 1996, 8, 254-262
(11) CL Berry, Proc. Roy. Inst., 1996, Oxford Science Publications, Oxford. p. 125-143
(12) R Mongeau, et al, Food and Chemical Toxicology, 1994,32, 265-271
(13) KDR Setchell, et al., Lancet, 1997, 350, 23-27
(14) SE Kulling, et al., Archives of Toxicology, 1999, 73, 50-54
(15) K North, J Golding, and the Alspac Study Team, BJU Intern, 2000, 85, 107-113
(16) Heart Outcomes Prevention Study Investigators, New Eng. J. Med., 2000, 342, 154-160
(17) GS Omenn, et al., J. Natl. Cancer Inst., 1996, 88(21), 1550-9
(18) GS Omenn, et al., IARC Sci. Publ., 1996, 136, 67-85
(19) G. Paolini, et al., Nature, 1999, 398, 760-761
(20) D-W Ziang, et al., J. Natl. Cancer Inst., 1999, 91, 60-66
(21) S Holm, and J Harris, Nature, 1999, 400, 398
(22) CL Berry, Brit. Heart J., 1978, 40(7), 709-17
(23) CN Martyn, et al., Brit. Heart J., 1995, 73(2), 116-21
(24) DJP Barker, Fetal and infant origins of adult disease, BMJ Publications, London, 1992
(25) D Ley, H Stale, and K Marsal, Acta Paediatrica, 1997, 86(3), 299-305
(26) H Martin, et al., Circulation, 2000, 102(22), 2739-44
(27) S Ebrahim, and GD Smith, British Medical Journal, 1997, 314(7095), 1666-74
(28) AJ Barsky, and JF Borus, J.A.M.A., 1995, 274(24), 1931-34
(29) GH Hall, WT Hamilton, and A. Round, J. Roy. Coll. Phys. Lond., 1998, 32, 44-45
(30) PG O’Malley, et al., J. Family Pract., 1999, 48(12), 980-90
(31) P Onghena, and B Van Houdenhove, Pain, 1992, 49(2), 205-19
(32) CL Berry, J. Roy. Coll. Phys. Lond., 2000, 34, 486-491
(33) BRN Baird, Risk Analysis, 1986, 6, 425-436
(34) CJ Kennedy, C.K. Probart, and S.M. Dorman, Health Educ. Quart., 1991, 18, 319-329
(35) SL Colbry, and FR Boutsen, Health Values, 1993, 17, 26-30
(36) E Ernst, Am. J. Med., 1998, 104, 170-8
(37) SC Piscitelli, et al., Lancet, 2000, 355, 547-548
(38) Social Issues Research Center, Code of Practise/Guidelines on Science and Health Communication, 2000
(39) AB Hill, Proc. R. Soc. Med., 1965, 58, 295-300.
This is an edited version of a speech given at the Royal Institution, London.
To enquire about republishing spiked’s content, a right to reply or to request a correction, please contact the managing editor, Viv Regan.