Br. J. clin. Pharmac. (1992), 33, 125-128
Hazards, risks and reality R. E. FERNER West Midlands Regional Poisons Unit, Dudley Road Hospital, Birmingham B18 7QH
non-smokers (Doll & Peto, 1976). The risk of developing a serious reaction to a drug is also unevenly distributed, for example between slow and fast metabolizers of hydralazine. The risk of relatively common, well-defined events can be counted, but often risks have to be estimated. For example, what are the risks of a very rare but serious event such as the meltdown of a nuclear reactor? How many cancers will be caused by the red colouring in marachino cherries? How common are fatal adverse effects of a new non-steroidal anti-inflammatory drug? The techniques of risk estimation are well-established in engineering (Green, 1982). Events which might initiate an accident are identified, and then the sequence leading from the event to the accident is specified. For example, if the electricity supply to a nuclear reactor failed, it should trip the emergency generator to cool the reactor core and even if the core overheated, the radioactivity should be contained. Failures in this sequence lead to catastrophe, and each can be assigned a probability. This approach could be applied to the sequence of prescribing, from ditignosis to administration, to identify those steps where the risks are greatest. Theoretical calculations are not always soundly based, or reliable. Consider the case of the Challenger shuttle disaster, in which the 25th space shuttle exploded and its crew were killed, although the National Aeronautics and Space Administration (NASA) quoted the risk of a catastrophic failure as 1 in 100,000: once in 300 years if the shuttle flew every day (Feynman, 1988). Unmanned rockets had failed in 5 of 127 test firings, and this in retrospect pointed to the fatuousness of NASA's
Writing an editorial is a hazardous business. Hazard is, after all, 'the situation that in particular circumstances could lead to harm' (Royal Society Study Group, 1983). There is, for example, the danger of being accused of plagiarism, a risk enhanced by the number of recent publications on risk (Anonymous, 1990; Black, 1991; Bottiger, 1990; British Medical Association, 1987; Cooper, 1985; Lave, 1987; Mann, 1989). Risk here means 'the probability that a particular adverse outcome occurs during a given quantum of exposure to hazard', for example during a given time interval, per lifetime, or per editorial written (Figure 1). Risk is important to doctors, who spend their days coping with adverse events and trying hard not to replace one adverse event with another. Death, an adverse event of particular interest, comes to each of us, and so has a probability of one. The risk of death in 1989 was 1 in 88 for every living person in England and Wales (Office of Population Censuses and Surveys, 1989). Table 1 shows how this risk was partitioned among different causes. The data suggest that risks from therapy are small, but this is not the whole story. The statistics are based on death certificates, so they are uncertain. It is clear that risk is not equally distributed across the whole population, either. For example, accidents involving hang gliders are much more likely to kill their pilots than unsuspecting members of the public. Indeed, the risk for pilots may be around 1 in 600 per year of hang gliding (Royal Society Study Group, 1983). Current smokers are 10 times more likely to die from lung cancer and 17 times more likely to die from chronic bronchitis than Time = 0
TiMne i=;
P'eeie ri 0
le
0
0 n
i
Figure 1 Hazard and risk. The tiger behind bars is the hazard, since it could lead to harm. The risk is the probability that an adverse outcome will occur in unit time, or for some other specified denominator such as 'per caged tiger'. The perceived risk is the man's intuitive estimate of the risk. He may express it ('more dangerous than crossing the road') or reveal it, by avoiding the tiger's cage, even if he risks falling into the penguin pool.
125
126
R. E. Ferner Table 1 Risk of dying in 1989 in England and Wales by causea
Cause
Risk of dying from a given cause in the year 1989
Anycause Diseases of the circulatory system Neoplasm Accident and violence Motor vehicle traffic accidents Poisoning by drugs Toxic effect of carbon monoxide Fire and flames Poisoning by antidepressants Homicide Toxic effect of ethanol Railway accidents Poisoning by salicylates Assault by poison Antibiotic in therapeutic use
1 in 88 1 in 190 1 in 350 1 in 3,000 1 in 10,000 1 in 30,000 1 in 40,000 1 in 90,000 1 in 160,000 1 in 180,000 1 in 420,000 1 in 700,000 1 in 800,000 1 in 4,200,000 1 in 17,000,000
Risk of death being due to a given cause 1 in 1 1 in 2.2 1 in 4 1 in 33 1 in 130 1 in 330 1 in 450 1 in 1,000 1 in 1,800 1 in 2,000 1 in 4,800 1 in 8,000 1 in 9,500 1 in 48,000 1 in 200,000
aBased on 1989 Mortality Statistics for England and Wales. DH2 No 16, Office of Population Censuses and Surveys, Her Majesty's Stationery Office, London.
estimate. The less knowledge there is, the more uncertain the risk assessment becomes. Carcinogenesis at low doses is usually estimated from animal data at high doses. It is very difficult to know whether the extrapolation is valid. Work on vinyl chloride monomer implies that risk estimates for carcinogenesis may be too sanguine by an order of magnitude (Roberts, 1989). In the field of drugs, adverse reactions are particularly hard to predict (Hoppe & Oehme, 1989) and detect (East, 1989). The uncertainties in extrapolation from high dose to low dose persist. There are the problems of extrapolation from one species to another. Sometimes toxicity occurs in animals but not in man. For example, hydrocortisone is teratogenic in some strains of mice, in whose offspring it induces cleft palate, though there is no evidence that this is true in man. By contrast, animals can be refractory to toxic effects. Thalidomide is not teratogenic in rats; and there is no animal model for the oculo-mucocutaneous syndrome due to practolol. The rarity of serious adverse reactions to drugs poses another problem. To be reasonably certain of detecting an adverse event which occurs in 1 treated patient in 1000, and does not occur spontaneously, some 3600 have to be treated. To detect a doubling of the risk of an event which already occurs with a frequency of 1 in 1000, some 20,000 patients would need to be treated (East, 1989). This explains the delay of many years between the marketing of a drug and the recognition of adverse reactions to it (Venning, 1983). Risk of the order of 1 in 1000 went undetected for 3-7 years for pulmonary embolism due to oral contraceptives, halothane jaundice and lincomycin-induced colitis. Clinically important risks may be much smaller. The evidence is that chloramphenicol and phenylbutazone both cause aplastic anaemia in 1 in 50,000 prescriptions, whilst the nonsteroidal anti-inflammatory drug zomiperac caused one death due to anaphylaxis per 2,000,000 prescriptions (Inman, 1989). The absolute risks seem small, but the risks of using alternative drugs are smaller still. Why do people expose themselves to unnecessary
risks? Perceived risk, the intuitive feeling that an individual has about the probability of an adverse event and the harm it will cause him, is known to differ very greatly from true risk. Intuition is a bad guide. After all, we feel intuitively that Bristol is west of Edinburgh, though it is not. Even when risks are appreciated on one level, psychological defences can prevent that appreciation being turned into action. The psychological components have been extensively studied (Fischhoff et al., 1981; Slovic, 1987). There is a greater willingness to accept a risk of death if it is deferred rather than immediate (cigarette smoking rather than lightning), the result of choice rather than unavoidable (mountaineering rather than landslides), by slow attrition rather than by catastrophe (road accidents rather than Jumbo jet crashes), and familiar rather than esoteric, technological or novel (coal mining, not nuclear power stations). We are also protected in deciding to take risk by a feeling of immunity-we are all better-than-average drivers. Misperceptions can extend to the analysis of risk. The doctor may prescribe the wrong drug or dosage, the pharmacist may dispense incorrectly, the nurse may administer the drug wrongly (Ernst et al., 1991) and the patient may be particularly susceptible to some adverse effect. Only the last event is at all unlikely, but it remains the focus of attention. There are substantial difficulties inherent in making rational decisions about risk when risk perception is 'perverse', and the evidence is that rational decisions are not made. An enormous investment in fire escapes, alarms and brigades protects people against the risk of dying by fire or flames, but 454 people die of circulatory disorders for every 1 who dies by fire. Formal enquiries are held into railway accidents, whilst death in road traffic accidents is 60 times more common. When adequate data exist, risk analysis allows decisions to be improved. At its simplest a certain gain of £10 is preferable to a 0.9 chance of nothing and a 0.1 chance of £99. Similar arguments have been used to decide whether to amputate a gangrenous leg at once, or to wait (O'Brien, 1986). They have also been used in
Hazards, risks and reality an approach to implicit risk analysis known as 'the standard gamble'. This asks 'what is the gamble you are prepared to take to change from the current (certain) state to a future one which you hope will be better, but which may be worse?' This technique has yielded important results. Patients with severe disability from rheumatoid arthritis would accept a high risk of death-over 80% -if there were a chance of complete cure (Rawlins, 1989). This may be compared with the approximate risks of blood dyscrasia from pheylbutazone (1 in 50,000), penicillamine (1 in 14,000) and gold (1 in 5600) (derived from Inman, 1989 and CMS Update, 1985). Thus, patients seem to be prepared to run much higher risks than we are prepared to take on their behalf. By contrast, journalists and members of the Food and Drug Administration might argue that it is better not to license new drugs at all than to run the risk of another thalidomide disaster. Alternatively, it might be argued that every possible precaution should be taken, including tests on thousands of animals and dozens of species. These approaches fail to recognise the risks inherent in not taking risks. For example, we may fail to cure diseases which could be cured or we may increase competing risks directly (zoonoses in animal house technicians, attacks by the Animal Liberation Front) or indirectly (less money to spend on more cost-effective measures). The licensing authorities are implicitly responsible for deciding what level of risk is tolerable. This may be an impossible task, because of the very wide range of individual perception . . .' a rational central decision, even if further ratified by some allowance for differences of perception, is bound to leave some people considering that they are at undue risk' (Royal Society Study Group, 1983). The risks of drugs are widely perceived as greater than they truly are. Reducing the perceived risks of drugs is going to be hard in a world where the intuitive lack of awareness of risk makes fortunes for football pools promoters. So other strategies are needed if the licensing authorities are to take reasonable risks. One strategy for reducing risk is to reduce exposure to hazards, especially in subgroups who are at high
127
risk. The licensing authorities have recognised this by requiring that data sheets include advice on prescribing in pregnancy and old age. An interesting approach to controlling risk is to emphasize the hazard. Agranulocytosis occurs in approximately 1% of schizophrenic patients treated with clozaril. The marketing company runs a service to screen full blood counts, making it hard for doctors to ignore the risk, and at the same time reducing it by detecting premonitory neutropenia (Committee on Safety of Medicines, 1990). An unsatisfactory aspect of the current system is the lack of a plan to cope with adverse reactions if they do occur. It would be much more satisfactory if there were an insurance against the possibility of an adverse outcome, and the most obvious mechanism would be nofault compensation for the rare victims of serious or fatal drug reactions, especially reactions which were previously unknown. It can be argued that natural illness is as deserving of compensation as man-made therapeutic misfortune. It is also true that patients who take drugs do so in the expectation of benefit. There are, however, benefits to others, since knowledge of the incidence of adverse effects increases with each patient taking a drug, and since the pharmaceutical companies gain, too. It seems logical that the government and the pharmaceutical industry should contribute to compensation, and that the compensation be set at a level which reflects the patient's own part in the process. If the alternative strategy for dealing with risk is to make licensing more difficult, then society is denied possible benefits. In summary, hazards and risks and their perception, assessment and management are important to the clinical pharmacologist. The licensing authorities will be in a better position to take poorly quantified risks on our behalf if there is compensation for patients who suffer adverse effects. And we should be aware that patients are sometimes prepared to take much greater risks themselves than we are on their behalves. I am grateful to Dr L. K. Harding for helpful discussion.
References Anonymous (1990). Ever so risky. Lancet, 336, 216-217. Black, D. (1991). Black looks.. at risks. J. Roy. Coll. Phys., Lond., 25, 5-6. Bottiger, L. E. (1990). Medicine and Risk. J. intern. Med., 227, 1-3. British Medical Association (1987). The B.M.A. guide to living with risk. London: Penguin Books. Cooper, M. G. (1985). Man-made hazards to man. Risk. Oxford: Clarendon Press. Committee on Safety of Medicines (1985). Blood dyscrasias. Br. med. J., 291, 1269. Committee on Safety of Medicines (1990). Clozaril induced neutropenia and clozaril patient monitoring service. Current Problems; No 30. Doll, R. & Peto, R. (1976). Mortality in relation to smoking: 20 years' observations on male British doctors. Br. med. J., 2, 1525-1536.
East, M. 0. (1989). Prediction of post-marketing safety from clinical studies. In Risk and consent to risk in medicine, ed. Mann, R. D. Carnforth: The Parthenon Publishing Group. Ernst, M. A., Buchanan, A. & Cox, C. (1989). Drug errors: A judgement of errors. Nursing Times, 87, 26-30. Feynman, R. P. (1988). What do you care what other people think? Further adventures of a curious character. London: Unwin. Fischhoff, B., Lichtenstein, S., Slovic, P., Derby, S. L. & Keeney, R. L. (1981). Acceptable risk. Cambridge: Cambridge University Press. Green, A. E. (1982). High risk safety technology. Chichester: John Wiley & Sons. Hoppe, L. D. & Oehme, F. W. (1989). Chemical risk assessment: A review. Vet. Hum. Tox., 31, 543-554. Inman, W. H. W. (1989). PMS and 'Brass Tacks'. In Risk and consent to risk in medicine, ed. Mann, R. D. Carnforth:
128
R. E. Ferner
The Parthenon Publishing Group. Lave, L. B. (1987). Health and safety risk analyses: Information for better decisions. Science, 236, 291-295. O'Brien, B. (1986). What are my chances doctor?-A review of clinical risks. London: Office of Health Economics. Rawlins, M. D. (1989). Trading risk for benefit. In Risk and consent to risk in medicine, ed. Mann, R. D. Carnforth: The Parthenon Publishing Group. Roberts, L. (1989). Is risk assessment conservative? Science, 238, 1553.
Slovic, P. (1987). Perception of risk. Science, 236, 280-285. Royal Society Study Group (1983). Risk assessment. London: The Royal Society. Venning, G. R. (1983). Identification of adverse reactions to new drugs. II (continued): How were 18 important adverse reactions discovered and with what delays? Br. med. J., 286, 365-368.
(Received 2 October 1991, accepted 14 October 1991)
Hazards, risks and reality R. E. FERNER West Midlands Regional Poisons Unit, Dudley Road Hospital, Birmingham B18 7QH
non-smokers (Doll & Peto, 1976). The risk of developing a serious reaction to a drug is also unevenly distributed, for example between slow and fast metabolizers of hydralazine. The risk of relatively common, well-defined events can be counted, but often risks have to be estimated. For example, what are the risks of a very rare but serious event such as the meltdown of a nuclear reactor? How many cancers will be caused by the red colouring in marachino cherries? How common are fatal adverse effects of a new non-steroidal anti-inflammatory drug? The techniques of risk estimation are well-established in engineering (Green, 1982). Events which might initiate an accident are identified, and then the sequence leading from the event to the accident is specified. For example, if the electricity supply to a nuclear reactor failed, it should trip the emergency generator to cool the reactor core and even if the core overheated, the radioactivity should be contained. Failures in this sequence lead to catastrophe, and each can be assigned a probability. This approach could be applied to the sequence of prescribing, from ditignosis to administration, to identify those steps where the risks are greatest. Theoretical calculations are not always soundly based, or reliable. Consider the case of the Challenger shuttle disaster, in which the 25th space shuttle exploded and its crew were killed, although the National Aeronautics and Space Administration (NASA) quoted the risk of a catastrophic failure as 1 in 100,000: once in 300 years if the shuttle flew every day (Feynman, 1988). Unmanned rockets had failed in 5 of 127 test firings, and this in retrospect pointed to the fatuousness of NASA's
Writing an editorial is a hazardous business. Hazard is, after all, 'the situation that in particular circumstances could lead to harm' (Royal Society Study Group, 1983). There is, for example, the danger of being accused of plagiarism, a risk enhanced by the number of recent publications on risk (Anonymous, 1990; Black, 1991; Bottiger, 1990; British Medical Association, 1987; Cooper, 1985; Lave, 1987; Mann, 1989). Risk here means 'the probability that a particular adverse outcome occurs during a given quantum of exposure to hazard', for example during a given time interval, per lifetime, or per editorial written (Figure 1). Risk is important to doctors, who spend their days coping with adverse events and trying hard not to replace one adverse event with another. Death, an adverse event of particular interest, comes to each of us, and so has a probability of one. The risk of death in 1989 was 1 in 88 for every living person in England and Wales (Office of Population Censuses and Surveys, 1989). Table 1 shows how this risk was partitioned among different causes. The data suggest that risks from therapy are small, but this is not the whole story. The statistics are based on death certificates, so they are uncertain. It is clear that risk is not equally distributed across the whole population, either. For example, accidents involving hang gliders are much more likely to kill their pilots than unsuspecting members of the public. Indeed, the risk for pilots may be around 1 in 600 per year of hang gliding (Royal Society Study Group, 1983). Current smokers are 10 times more likely to die from lung cancer and 17 times more likely to die from chronic bronchitis than Time = 0
TiMne i=;
P'eeie ri 0
le
0
0 n
i
Figure 1 Hazard and risk. The tiger behind bars is the hazard, since it could lead to harm. The risk is the probability that an adverse outcome will occur in unit time, or for some other specified denominator such as 'per caged tiger'. The perceived risk is the man's intuitive estimate of the risk. He may express it ('more dangerous than crossing the road') or reveal it, by avoiding the tiger's cage, even if he risks falling into the penguin pool.
125
126
R. E. Ferner Table 1 Risk of dying in 1989 in England and Wales by causea
Cause
Risk of dying from a given cause in the year 1989
Anycause Diseases of the circulatory system Neoplasm Accident and violence Motor vehicle traffic accidents Poisoning by drugs Toxic effect of carbon monoxide Fire and flames Poisoning by antidepressants Homicide Toxic effect of ethanol Railway accidents Poisoning by salicylates Assault by poison Antibiotic in therapeutic use
1 in 88 1 in 190 1 in 350 1 in 3,000 1 in 10,000 1 in 30,000 1 in 40,000 1 in 90,000 1 in 160,000 1 in 180,000 1 in 420,000 1 in 700,000 1 in 800,000 1 in 4,200,000 1 in 17,000,000
Risk of death being due to a given cause 1 in 1 1 in 2.2 1 in 4 1 in 33 1 in 130 1 in 330 1 in 450 1 in 1,000 1 in 1,800 1 in 2,000 1 in 4,800 1 in 8,000 1 in 9,500 1 in 48,000 1 in 200,000
aBased on 1989 Mortality Statistics for England and Wales. DH2 No 16, Office of Population Censuses and Surveys, Her Majesty's Stationery Office, London.
estimate. The less knowledge there is, the more uncertain the risk assessment becomes. Carcinogenesis at low doses is usually estimated from animal data at high doses. It is very difficult to know whether the extrapolation is valid. Work on vinyl chloride monomer implies that risk estimates for carcinogenesis may be too sanguine by an order of magnitude (Roberts, 1989). In the field of drugs, adverse reactions are particularly hard to predict (Hoppe & Oehme, 1989) and detect (East, 1989). The uncertainties in extrapolation from high dose to low dose persist. There are the problems of extrapolation from one species to another. Sometimes toxicity occurs in animals but not in man. For example, hydrocortisone is teratogenic in some strains of mice, in whose offspring it induces cleft palate, though there is no evidence that this is true in man. By contrast, animals can be refractory to toxic effects. Thalidomide is not teratogenic in rats; and there is no animal model for the oculo-mucocutaneous syndrome due to practolol. The rarity of serious adverse reactions to drugs poses another problem. To be reasonably certain of detecting an adverse event which occurs in 1 treated patient in 1000, and does not occur spontaneously, some 3600 have to be treated. To detect a doubling of the risk of an event which already occurs with a frequency of 1 in 1000, some 20,000 patients would need to be treated (East, 1989). This explains the delay of many years between the marketing of a drug and the recognition of adverse reactions to it (Venning, 1983). Risk of the order of 1 in 1000 went undetected for 3-7 years for pulmonary embolism due to oral contraceptives, halothane jaundice and lincomycin-induced colitis. Clinically important risks may be much smaller. The evidence is that chloramphenicol and phenylbutazone both cause aplastic anaemia in 1 in 50,000 prescriptions, whilst the nonsteroidal anti-inflammatory drug zomiperac caused one death due to anaphylaxis per 2,000,000 prescriptions (Inman, 1989). The absolute risks seem small, but the risks of using alternative drugs are smaller still. Why do people expose themselves to unnecessary
risks? Perceived risk, the intuitive feeling that an individual has about the probability of an adverse event and the harm it will cause him, is known to differ very greatly from true risk. Intuition is a bad guide. After all, we feel intuitively that Bristol is west of Edinburgh, though it is not. Even when risks are appreciated on one level, psychological defences can prevent that appreciation being turned into action. The psychological components have been extensively studied (Fischhoff et al., 1981; Slovic, 1987). There is a greater willingness to accept a risk of death if it is deferred rather than immediate (cigarette smoking rather than lightning), the result of choice rather than unavoidable (mountaineering rather than landslides), by slow attrition rather than by catastrophe (road accidents rather than Jumbo jet crashes), and familiar rather than esoteric, technological or novel (coal mining, not nuclear power stations). We are also protected in deciding to take risk by a feeling of immunity-we are all better-than-average drivers. Misperceptions can extend to the analysis of risk. The doctor may prescribe the wrong drug or dosage, the pharmacist may dispense incorrectly, the nurse may administer the drug wrongly (Ernst et al., 1991) and the patient may be particularly susceptible to some adverse effect. Only the last event is at all unlikely, but it remains the focus of attention. There are substantial difficulties inherent in making rational decisions about risk when risk perception is 'perverse', and the evidence is that rational decisions are not made. An enormous investment in fire escapes, alarms and brigades protects people against the risk of dying by fire or flames, but 454 people die of circulatory disorders for every 1 who dies by fire. Formal enquiries are held into railway accidents, whilst death in road traffic accidents is 60 times more common. When adequate data exist, risk analysis allows decisions to be improved. At its simplest a certain gain of £10 is preferable to a 0.9 chance of nothing and a 0.1 chance of £99. Similar arguments have been used to decide whether to amputate a gangrenous leg at once, or to wait (O'Brien, 1986). They have also been used in
Hazards, risks and reality an approach to implicit risk analysis known as 'the standard gamble'. This asks 'what is the gamble you are prepared to take to change from the current (certain) state to a future one which you hope will be better, but which may be worse?' This technique has yielded important results. Patients with severe disability from rheumatoid arthritis would accept a high risk of death-over 80% -if there were a chance of complete cure (Rawlins, 1989). This may be compared with the approximate risks of blood dyscrasia from pheylbutazone (1 in 50,000), penicillamine (1 in 14,000) and gold (1 in 5600) (derived from Inman, 1989 and CMS Update, 1985). Thus, patients seem to be prepared to run much higher risks than we are prepared to take on their behalf. By contrast, journalists and members of the Food and Drug Administration might argue that it is better not to license new drugs at all than to run the risk of another thalidomide disaster. Alternatively, it might be argued that every possible precaution should be taken, including tests on thousands of animals and dozens of species. These approaches fail to recognise the risks inherent in not taking risks. For example, we may fail to cure diseases which could be cured or we may increase competing risks directly (zoonoses in animal house technicians, attacks by the Animal Liberation Front) or indirectly (less money to spend on more cost-effective measures). The licensing authorities are implicitly responsible for deciding what level of risk is tolerable. This may be an impossible task, because of the very wide range of individual perception . . .' a rational central decision, even if further ratified by some allowance for differences of perception, is bound to leave some people considering that they are at undue risk' (Royal Society Study Group, 1983). The risks of drugs are widely perceived as greater than they truly are. Reducing the perceived risks of drugs is going to be hard in a world where the intuitive lack of awareness of risk makes fortunes for football pools promoters. So other strategies are needed if the licensing authorities are to take reasonable risks. One strategy for reducing risk is to reduce exposure to hazards, especially in subgroups who are at high
127
risk. The licensing authorities have recognised this by requiring that data sheets include advice on prescribing in pregnancy and old age. An interesting approach to controlling risk is to emphasize the hazard. Agranulocytosis occurs in approximately 1% of schizophrenic patients treated with clozaril. The marketing company runs a service to screen full blood counts, making it hard for doctors to ignore the risk, and at the same time reducing it by detecting premonitory neutropenia (Committee on Safety of Medicines, 1990). An unsatisfactory aspect of the current system is the lack of a plan to cope with adverse reactions if they do occur. It would be much more satisfactory if there were an insurance against the possibility of an adverse outcome, and the most obvious mechanism would be nofault compensation for the rare victims of serious or fatal drug reactions, especially reactions which were previously unknown. It can be argued that natural illness is as deserving of compensation as man-made therapeutic misfortune. It is also true that patients who take drugs do so in the expectation of benefit. There are, however, benefits to others, since knowledge of the incidence of adverse effects increases with each patient taking a drug, and since the pharmaceutical companies gain, too. It seems logical that the government and the pharmaceutical industry should contribute to compensation, and that the compensation be set at a level which reflects the patient's own part in the process. If the alternative strategy for dealing with risk is to make licensing more difficult, then society is denied possible benefits. In summary, hazards and risks and their perception, assessment and management are important to the clinical pharmacologist. The licensing authorities will be in a better position to take poorly quantified risks on our behalf if there is compensation for patients who suffer adverse effects. And we should be aware that patients are sometimes prepared to take much greater risks themselves than we are on their behalves. I am grateful to Dr L. K. Harding for helpful discussion.
References Anonymous (1990). Ever so risky. Lancet, 336, 216-217. Black, D. (1991). Black looks.. at risks. J. Roy. Coll. Phys., Lond., 25, 5-6. Bottiger, L. E. (1990). Medicine and Risk. J. intern. Med., 227, 1-3. British Medical Association (1987). The B.M.A. guide to living with risk. London: Penguin Books. Cooper, M. G. (1985). Man-made hazards to man. Risk. Oxford: Clarendon Press. Committee on Safety of Medicines (1985). Blood dyscrasias. Br. med. J., 291, 1269. Committee on Safety of Medicines (1990). Clozaril induced neutropenia and clozaril patient monitoring service. Current Problems; No 30. Doll, R. & Peto, R. (1976). Mortality in relation to smoking: 20 years' observations on male British doctors. Br. med. J., 2, 1525-1536.
East, M. 0. (1989). Prediction of post-marketing safety from clinical studies. In Risk and consent to risk in medicine, ed. Mann, R. D. Carnforth: The Parthenon Publishing Group. Ernst, M. A., Buchanan, A. & Cox, C. (1989). Drug errors: A judgement of errors. Nursing Times, 87, 26-30. Feynman, R. P. (1988). What do you care what other people think? Further adventures of a curious character. London: Unwin. Fischhoff, B., Lichtenstein, S., Slovic, P., Derby, S. L. & Keeney, R. L. (1981). Acceptable risk. Cambridge: Cambridge University Press. Green, A. E. (1982). High risk safety technology. Chichester: John Wiley & Sons. Hoppe, L. D. & Oehme, F. W. (1989). Chemical risk assessment: A review. Vet. Hum. Tox., 31, 543-554. Inman, W. H. W. (1989). PMS and 'Brass Tacks'. In Risk and consent to risk in medicine, ed. Mann, R. D. Carnforth:
128
R. E. Ferner
The Parthenon Publishing Group. Lave, L. B. (1987). Health and safety risk analyses: Information for better decisions. Science, 236, 291-295. O'Brien, B. (1986). What are my chances doctor?-A review of clinical risks. London: Office of Health Economics. Rawlins, M. D. (1989). Trading risk for benefit. In Risk and consent to risk in medicine, ed. Mann, R. D. Carnforth: The Parthenon Publishing Group. Roberts, L. (1989). Is risk assessment conservative? Science, 238, 1553.
Slovic, P. (1987). Perception of risk. Science, 236, 280-285. Royal Society Study Group (1983). Risk assessment. London: The Royal Society. Venning, G. R. (1983). Identification of adverse reactions to new drugs. II (continued): How were 18 important adverse reactions discovered and with what delays? Br. med. J., 286, 365-368.
(Received 2 October 1991, accepted 14 October 1991)
