The health and psychological consequences of cannabis use chapter 3
3. Evidential principles 3.1 Issues in appraising health hazards The evaluation of the health hazards of any drug is difficult for a variety of scientific and sociopolitical reasons. First, causal inferences about the effects of drugs on human health are not easy to make (ARF/WHO, 1981). Even inferences about the relatively direct and transient effects of acute drug use may be complicated by individual variability in response to a standard dose of a drug, and by the fact that serious adverse effects are relatively rare. Inference becomes more difficult the longer the interval between use and alleged ill effects: it takes time for such effects to develop, and it may take considerably longer for the research technology to be developed that enables these effects to be identified and confidently attributed to the drug use rather than some other factor (Institute of Medicine, 1982). In the case of chronic tobacco use, for example, it has taken over three hundred years to discover that it increases premature mortality from cancer, and heart disease. Moreover, new health hazards of tobacco use, such as passive smoking, continue to be discovered. Second, in making causal inferences about drug use and its consequences there is a tension between the rigour and relevance of the evidence. The most rigorous evidence is provided by laboratory investigations using experimental animals, or in vitro preparations of animal cells and micro-organisms in which well controlled drug doses are related to precisely measured biological outcomes. The relevance of such research to human disease, however, is often problematic. A great many inferences have to be made in linking the occurrence of specific biological effects in laboratory animals or cell cultures to the likely effects of the drug under existing patterns of human use. Epidemiological studies of relationships between drug use and human disease have manifestly greater relevance to the appraisal of the health risks of human drug use, but this is purchased at the price of reduced rigour. Doses of drugs over periods of years are difficult to quantify in the best of circumstances. The vagaries of human memory which make quantification of consumption difficult in the case of tobacco and alcohol are magnified in the case of illicit drugs by the non-standard doses and contaminants in blackmarket drugs, and the reluctance of users to report illicit drug use. The fact that different patterns of drug use and other life-style factors are often correlated (e.g. alcohol and tobacco), makes attribution of ill-effects to particular drugs even more difficult (Task Force on Health Risk Assessment, 1986). Third, appraisals of the hazards of recreational drug use are unavoidably affected by the societal approval or disapproval of the drug in question. As Room (1984) has observed, when evaluating the impact of alcohol on non-industrialised societies, anthropologists have often engaged in problem deflation in response to the problem inflation of missionaries and colonial authorities. In our own culture, the economic interests of tobacco and alcohol industries provide a potent reason for problem deflation with these drugs. Such problem deflationists often discount the adverse effects of alcohol use, either by contesting the evidence for adverse effects, or by denying that there is a causal connection between alcohol use and particular adverse health effects. Similar processes have been at work in the appraisal of the health effects of recreational cannabis use. The countercultural symbolism of cannabis use in the late 1960s has introduced a strong sociopolitical dimension to the debate about the adverse health effects of cannabis. Politically conservative opponents of cannabis use, for example, justify its continued prohibition by citing evidence of the personal and social harms of its use. When the evidence is uncertain, as it is with many of the alleged effects of chronic use, they resolve the uncertainty by assuming that the cannabis is unsafe until proven safe. Complementary behaviour is exhibited by some proponents of decriminalisation. Evidence of harm is discounted or discredited, and uncertainties about the ill-effects of chronic cannabis use are resolved by demanding more and better evidence, arguing that until this uncertainty is resolved individuals should be allowed to exercise their free choice about whether or not they use the drug. Such approaches to the appraisal of evidence have not always been consistently applied. Both sides of the debate would reject the application of their own approaches to the appraisal of cannabis to the appraisal of the health hazards of alcohol, pesticides, herbicides, or chemical residues in food. While we do not claim to be unaffected by these processes, we will be as explicit as possible about the evidential standards that we have used, and as even-handed as we can in their application. 3.2 Evidential desiderata The following issues must be addressed in specifying what we have taken to be the evidential desiderata in our appraisal of the health risk of cannabis use: the burden of proof; standard of proof; criteria for causal inference; preference for relevance or rigour; approaches to estimating the magnitude of risk; and the desirability of a comparative appraisal of the risks. The burden of proof concerns who bears the responsibility for making a case; those who make a claim of adverse health effects, or those who doubt it (see Rescher, 1977, chapter XII). Who bears the burden of proof determines the way in which an issue is decided in the face of uncertainty: if the burden falls on those who claim that the drug is safe, uncertainty will be resolved by assuming that it is unsafe until proved otherwise; conversely, if the burden falls on those who claim that the drug is unsafe, then it will be assumed to be safe until proven otherwise. It is by no means agreed who bears the burden of proof in the debate about the health effects of cannabis use. Proponents of continued prohibition appeal to established practice (Whately, 1846), arguing that since the drug is illegal, the burden of proof falls upon those who want to legalise it. Some proponents of its legalisation counter that this begs the question, since there was no evidence, they argue, that cannabis was harmful when its use was first made a criminal offence. Others argue that the burden of proof falls upon those who wish to use the criminal law to prevent adults from freely choosing to use a drug (e.g. Husak, 1992). We will vary the burden of proof on the basis of the state of the evidence and argument. Once a prima facie case of harm has been made, positive evidence of safety will be required rather than the simple absence of any evidence of ill effect. We will assume that a prima facie case has been made either when there is direct evidence that the drug has ill effects in animals or humans (e.g. from a case-control study), or when there is some compelling argument that it could, e.g. the inference that since tobacco smoking causes lung cancer and cannabis and tobacco smoke are similar in their constituents, it is probable that heavy cannabis smoking also causes lung cancer. The standard of proof reflects the degree of confidence required in an inference that there is a causal connection between drug use and harm. In courts of law, the standard of proof demanded depends upon the seriousness of the offence at issue and the consequences of a verdict, with a higher standard of proof, "beyond reasonable doubt", being demanded in criminal cases, while the "balance of probabilities" is acceptable in civil cases. Although these legal standards are not directly translatable into scientific practice, scientists generally require something closer to the standard of "beyond reasonable doubt" than the balance of probabilities before drawing confident conclusions that a drug causes harm. If we were to demand that such a standard be met for the health effects of cannabis, this review would be exceedingly brief. Consequently, we will relax the criteria and indicate when the evidence permits a causal inference to be made on the balance of probabilities. We will take this standard to be exemplified in the consensus of informed scientific opinion that sufficient evidence has been provided to infer a probable causal connection between drug use and a harm (e.g. Fehr and Kalant, 1983; Institute of Medicine, 1982). In the trade-off between relevance and rigour, our preference will be for human evidence, both experimental and epidemiological, rather than animal and in vitro studies. In the absence of human evidence, in vitro and animal experiments will be taken as raising a suspicion that drug use has an adverse effects on human health. The degree of suspicion raised will be in proportion to the number of such animal studies, the consistency of their results across different species and experimental preparations (Task Force on Health Risk Assessment, 1986), and the degree of expert consensus that the inferences from effects in vitro and in vivo to adverse effects under existing patterns of human use are valid. The degree of consensus on the latter point will be indicated by the views expressed in authoritative reviews in peer reviewed journals or contributions to international consensus conferences (e.g. Fehr and Kalant, 1983; Institute of Medicine, 1982). The criteria for causal inference that we will use are standard ones (see Hall, 1987), namely: 1. Evidence that there is a relationship between drug use and a health outcome provided by one of the accepted types of epidemiological research design (namely, case-control, cross-sectional, cohort, or experiment). 2. Evidence (usually provided by a statistical significance test or the construction of a confidence interval) that the relationship is unlikely to be due to chance. 3. Good evidence that drug use precedes the adverse effect (e.g. a cohort study). 4. Evidence either from experiment, or statistical or other form of control, which makes it unlikely that the relationship is due to some other variable which is related to both drug use and the adverse effect. In appraising a body of literature as a whole we determine the extent to which the evidence meets the criteria outlined by Hill (1977). Ideally, once a strong case has been made for a causal connection between drug use and an adverse health effect, the magnitude of risk needs to be estimated so the seriousness of the risk can be quantified. For example, the consumption of large amounts of water over a short period of time can kill human beings, but this is not a good reason for counselling people against drinking water. The quantities required to produce intoxication and death are so large (e.g. 30 or more litres) that only diseased or psychotic individuals consume them. The standard epidemiological measures of risk magnitude are relative risk and population attributable risk. The relative risk is the increase in the odds of experiencing an adverse health outcome among those who use the drug compared to those who do not (that is, the number of times greater the risk of experiencing an effect is among those who use the drug compared with those who do not). The population attributable risk represents that proportion of cases with an adverse outcome which is attributable to drug use. The two measures of risk magnitude have different uses and implications. Relative risk is of greatest relevance to individuals attempting to estimate the increase in their risk of experiencing an adverse outcome if they use a drug. Attributable risk is of most relevance to a societal appraisal of the harms of drug use. The importance of the two measures of risk magnitude depends upon the prevalence of drug use and the base rate of the adverse outcome. An exposure with a low relative risk may have a large public health impact if a large proportion of the population is exposed (e.g. cigarette smoking and heart disease). Conversely, an exposure with a high relative risk may have little public health importance because very few people are exposed to it. Accordingly, an appraisal of the public health importance of illicit drug use must take some account not only of the relative risk of harm, but also the prevalence of use and the base rate of the adverse effect. As will become apparent in the course of this review, it is very difficult to estimate either relative or attributable risk of any probable adverse health effects of cannabis use because few epidemiological studies have been conducted. A different way of assessing the health risk posed by cannabis use has had to be used: a comparative qualitative appraisal of its risks with those of other widely used recreational drugs such as alcohol and tobacco (ARF/WHO, 1981). The motive for such comparisons is that they reduce the operation of double-standards in the health appraisal of drug use by reminding us that the drugs we currently tolerate pose major health risks. They also help to put the risk of newer drugs into perspective, so that we can use a common standard when making societal decisions about whether or not to tolerate such drug use. The task of comparison, however, is more difficult than it seems at first. First, we know much more about the quantitative risks of acute and chronic tobacco and alcohol use than we know about the health risks of currently illicit drugs. This is largely because the legal drugs have been consumed by substantial proportions of the population, over centuries in the case of tobacco, and millennia in the case of alcohol, and there have been more than 40 years of scientific studies of the health consequences of their use. The contemporary illicit drugs, by contrast, have been much less widely used in Western society, and for a shorter period, primarily by healthy young adults; there have also been few studies of their adverse health effects, and there have been even fewer attempts to quantify the risks of their use. Second, the prevalence of use of currently legal and illegal drugs is so different that any comparison based upon existing patterns of use will disadvantage the legal drugs (Peterson, 1980). In principle, this problem could be addressed by estimating what the health risks of cannabis use might be if its prevalence was to approach that of alcohol and tobacco. This approach has not been adopted here because in the absence of good data on the quantitative risks of cannabis use, a large number of contestable assumptions would have to be made to permit such estimates to be made. These obstacles provide strong reasons for cautiously interpreting comparisons of the health hazards of cannabis with those of alcohol and tobacco. They do not, however, provide insurmountable objections to such comparisons. We will accordingly make some qualitative comparisons with the health risks of alcohol and tobacco after we have considered the evidence on the adverse health effects of cannabis. References Addiction Research Foundation/World Health Organization (1981) Report of an ARF/WHO Scientific Meeting on the Adverse Health and Behavioral Consequences of Cannabis Use. Toronto: Addiction Research Foundation, . Fehr, K.O. and Kalant, H. (1983) (eds) Cannabis and Health Hazards. Toronto: Addiction Research Foundation. Hall, W. (1987) A simplified logic of causal inference. Australian and New Zealand Journal of Psychiatry, 1987, 21, 507-513. Hill, A.B. (1977). A Short Textbook of Statistics. London: Hodder and Stoughton. Husak, D.N. (1992) Drugs and Rights. Cambridge: Cambridge University Press. Institute of Medicine. (1982) Marijuana and Health. Washington DC: National Academy Press. Peterson, R (1980) (ed) Marijuana Research Findings: 1980 National Institute on Drug Abuse Research Monograph Number 31. Rockville, MD: U.S. Department of Health and Human Services. Rescher, N. (1977) Methodological Pragmatism. Oxford, Blackwell. Room, R. (1984) Alcohol and ethnography: A case of problem deflation? Current Anthropology, 25, 169- 191. Task Force on Health Risk Assessment, United States Department of Health and Human Services (1986) Determining Risks to Health: Federal Policy and Practice. Dover, MA: Auburn House Publishing Company. Whately, R. (1846) Elements of Rhetoric. Originally published 1846. (ed) D. Ehninger. Carnondale, Illinois: Illinois University Press, 1963.
Cliff Schaffer's Home Page