Development of a rational scale to assess the harm of drugs of potential misuse

Introduction

Drug misuse is one of the major social, legal, and public-health challenges in the modern world. In the UK, the total burden of drug misuse, in terms of health, social, and crime-related costs, has been estimated to be between £10 billion and £16 billion per year,1 with the global burden being proportionately enormous.2,3

Current approaches to counter drug misuse are interdiction of supply (via policing and customs control), education, and treatment. All three demand clarity in terms of the relative risks and harms that drugs engender. At present, in the UK, attitudes to policing and the punishments for possession and supply of drugs are scaled according to their classification under the Misuse of Drugs Act of 1971,4 while education and health-care provision are nominally tailored to the known actions and harms of specific drugs. Most other countries and international agencies—eg, the UN and WHO—have drug classification systems that purport to be structured according to the relative risks and dangers of illicit drugs. However, the process by which harms are determined is often undisclosed, and when made public can be ill-defined, opaque, and seemingly arbitrary. In part, this lack of clarity is due to the great range and complexity of factors that have to be taken into account in estimation of harm and the fact that scientific evidence is not only limited in many of the relevant areas but also evolves progressively and in unpredictable ways.

These qualifications apply to the evidence base of the current UK Misuse of Drugs Act, in which drugs are segregated into three classes—A, B, and C—that are intended to indicate the dangers of each drug, class A being the most harmful and class C the least. The classification of a drug has several consequences, in particular determining the legal penalties for importation, supply, and possession, as well as the degree of police effort targeted at restricting its use. The current classification system has evolved in an unsystematic way from somewhat arbitrary foundations with seemingly little scientific basis.

Here, we suggest a new system for assessing the potential harms of individual drugs on the basis of fact and scientific knowledge. This system is able to respond to evolving evidence about the potential harm of current drugs and to rank the threat presented by any new street drug.

Categories of harm

There are three main factors that together determine the harm associated with any drug of potential abuse: the physical harm to the individual user caused by the drug; the tendency of the drug to induce dependence; and the effect of drug use on families, communities, and society.5–8

Assessment of harm

Table 1 shows the assessment matrix that we designed, which includes all nine parameters of risk, created by dividing each of the three major categories of harm into three subgroups, as described above. Participants were asked to score each substance for each of these nine parameters, using a four-point scale, with 0 being no risk, 1 some, 2 moderate, and 3 extreme risk. For some analyses, the scores for the three parameters for each category were averaged to give a mean score for that category. For the sake of discussion, an overall harm rating was obtained by taking the mean of all nine scores.

Table 1. Assessment parameters

The scoring procedure was piloted by members of the panel of the Independent Inquiry into the Misuse of Drugs Act.13 Once refined through this piloting, an assessment questionnaire based on table 1, with additional guidance notes, was used. Two independent groups of experts were asked to do the ratings. The first was the national group of consultant psychiatrists who were on the Royal College of Psychiatrists' register as specialists in addiction. Replies were received and analysed from 29 of the 77 registered doctors who were asked to assess 14 compounds—heroin, cocaine, alcohol, barbiturates, amphetamine, methadone, benzodiazepines, solvents, buprenorphine, tobacco, ecstasy, cannabis, LSD, and steroids. Tobacco and alcohol were included because their extensive use has provided reliable data on their risks and harms, providing familiar benchmarks against which the absolute harms of other drugs can be judged. However, direct comparison of the scores for tobacco and alcohol with those of the other drugs is not possible since the fact that they are legal could affect their harms in various ways, especially through easier availability.

Having established that this nine-parameter matrix worked well, we convened meetings of a second group of experts with a wider spread of expertise. These experts had experience in one of the many areas of addiction, ranging from chemistry, pharmacology, and forensic science, through psychiatry and other medical specialties, including epidemiology, as well as the legal and police services. The second set of assessments was done in a series of meetings run along delphic principles, a new approach that is being used widely to optimise knowledge in areas where issues and effects are very broad and not amenable to precise measurements or experimental testing,14 and which is becoming the standard method by which to develop consensus in medical matters. Since delphic analysis incorporates the best knowledge of experts in diverse disciplines, it is ideally applicable to a complex variable such as drug misuse and addiction. Initial scoring was done independently by each participant, and the scores for each individual parameter were then presented to the whole group for discussion, with a particular emphasis on elucidating the reasoning behind outlier scores. Individuals were then invited to revise their scores, if they wished, on any of the parameters, in the light of this discussion, after which a final mean score was calculated. The complexity of the process means that only a few drugs can be assessed in a single meeting, and four meetings were needed to complete the process. The number of members taking part in the scoring varied from eight to 16. However, the full range of expertise was maintained in each assessment.

This second set of assessments covered the 14 substances considered by the psychiatrists plus, for completeness, six other compounds (khat, 4-methylthioamphetamine [4-MTA], gamma 4-hydroxybutyric acid [GHB], ketamine, methylphenidate, and alkyl nitrites), some of which are not illegal, but for each of which there have been reports of abuse (table 2). Participants were told in advance which drugs were being covered at each meeting to allow them to update their knowledge and consider their opinion. Recent review articles5,6,7,15–18 were provided.

Table 2. The 20 substances assessed, showing their current status under the Misuse of Drugs Act

Occasionally, individual experts were unable to give a score for a particular parameter for a particular drug and these missing values were ignored in the analysis—ie, they were neither treated as zero nor given some interpolated value. Data were analysed with the statistical functions in Microsoft Excel and S-plus.

Results

Use of this risk assessment system proved straightforward and practicable, both by questionnaire and in open delphic discussion. Figure 1 shows the overall mean scores of the independent expert group, averaged across all scorers, plotted in rank order for all 20 substances. The classification of each substance under the Misuse of Drugs Act is also shown. Although the two substances with the highest harm ratings (heroin and cocaine) are class A drugs, overall there was a surprisingly poor correlation between drugs' class according to the Misuse of Drugs Act and harm score. Of both the eight substances that scored highest and the eight that scored lowest, three were class A and two were unclassified. Alcohol, ketamine, tobacco, and solvents (all unclassified at the time of assessment) were ranked as more harmful than LSD, ecstasy, and its variant 4-MTA (all class A drugs). Indeed, the correlation between classification by the Misuse of Drugs Act and harm rating was not significant (Kendall's rank correlation −0·18; p=0·25; Spearman's rank correlation −0·26, p=0·26). Of the unclassified drugs, alcohol and ketamine were given especially high ratings. Interestingly, a very recent recommendation from the Advisory Council on the Misuse of Drugs that ketamine should be added to the Misuse of Drugs Act (as a class C drug) has just been accepted.19

Figure 1. Mean harm scores for 20 substances

Classification under the Misuse of Drugs Act, where appropriate, is shown by the colour of each bar.

We compared the overall mean scores (averaged across all nine parameters) for the psychiatrists with those of the independent group for the 14 substances that were ranked by both groups (figure 2). The figure suggests that the scores have some validity and that the process is robust, in that it generates similar results in the hands of rather different sets of experts.

Figure 2. Correlation between mean scores from the independent experts and the specialist addiction psychiatrists

1=heroin. 2=cocaine. 3=alcohol. 4=barbiturates. 5=amphetamine. 6=methadone. 7=benzodiazepines. 8=solvents. 9=buprenorphine. 10=tobacco. 11=ecstasy. 12=cannabis. 13=LSD. 14=steroids.

Table 3 lists the independent group results for each of the three subcategories of harm. The scores in each category were averaged across all scorers and the substances are listed in rank order of harm, based on their overall score. Many of the drugs were consistent in their ranking across the three categories. Heroin, cocaine, barbiturates, and street methadone were in the top five places for all categories of harm, whereas khat, alkyl nitrites, and ecstasy were in the bottom five places for all. Some drugs differed substantially in their harm ratings across the three categories. For instance, cannabis was ranked low for physical harm but somewhat higher for dependence and harm to family and community. Anabolic steroids were ranked high for physical harm but low for dependence. Tobacco was high for dependence but distinctly lower for social harms, because it scored low on intoxication. Tobacco's mean score for physical harm was also modest, since the ratings for acute harm and potential for intravenous use were low, although the value for chronic harm was, unsurprisingly, very high.

Table 3. Mean independent group scores in each of the three categories of harm, for 20 substances, ranked by their overall score, and mean scores for each of the three subscales

(click to view larger size)

Drugs that can be administered by the intravenous route were generally ranked high, not solely because they were assigned exceptionally high scores for parameter three (ie, the propensity for intravenous use) and nine (health-care costs). Even if the scores for these two parameters were excluded from the analysis, the high ranking for such drugs persisted. Thus, drugs that can be administered intravenously were also judged to be very harmful in many other respects.

Discussion

The results of this study do not provide justification for the sharp A, B, or C divisions of the current classifications in the UK Misuse of Drugs Act. Distinct categorisation is, of course, convenient for setting of priorities for policing, education, and social support, as well as to determine sentencing for possession or dealing. But neither the rank ordering of drugs nor their segregation into groups in the Misuse of Drugs Act classification is supported by the more complete assessment of harm described here. Sharply defined categories in any ranking system are essentially arbitrary unless there are obvious discontinuities in the full set of scores. Figure 1 shows only a hint of such a transition in the spectrum of harm, in the small step in the very middle of the distribution, between buprenorphine and cannabis. Interestingly, alcohol and tobacco are both in the top ten, higher-harm group. There is a rapidly accelerating harm value from alcohol upwards. So, if a three-category classification were to be retained, one possible interpretation of our findings is that drugs with harm scores equal to that of alcohol and above might be class A, cannabis and those below might be class C, and drugs in between might be class B. In that case, it is salutary to see that alcohol and tobacco—the most widely used unclassified substances—would have harm ratings comparable with class A and B illegal drugs, respectively.

Participants were asked to assess the harm of drugs administered in the form that they are normally used. In a few cases, the harms caused by a particular drug could not be completely isolated from interfering factors associated with the particular style of use. For example, cannabis is commonly smoked as a mixture with tobacco, which might have raised its scores for physical harm and dependence, among other factors. There is a further degree of uncertainty resulting from polydrug use, especially in the so-called recreational group of drugs that includes GHB, ketamine, ecstasy, and alcohol, for which adverse effects could be attributed mainly to one of the components of commonly used mixtures. Crack cocaine is generally deemed to be more dangerous than powdered cocaine, but they were not considered separately in this study. Similarly, the scores for the benzodiazepines might have been biased in the direction of the most abused drugs, especially temazepam. Individual scoring for particular benzodiazepines and for the various forms in which other drugs are used would be more appropriate should this or any other system of harm classification be used in a formal setting.

In view of the small numbers of independent scores, we did not think that estimation of correlations between the nine parameters was legitimate. There is quite likely to be some redundancy—ie, the nine parameters might not represent nine independent measures of risk. In much the same way, the principal components of the parameters were not extracted, partly because we thought that there were insufficient data and partly because reduction of the number of parameters to a core group might not be appropriate, at least until further assessment panels have independently validated the entire system.

Our analysis gave equal weight to each parameter of harm, and individual scores have simply been averaged. Such a procedure would not give a valid indication of harm for a drug that has extreme acute toxicity, such as the designer drug contaminant MPTP (1-methyl 4-phenyl 1,2,3,6-tetrahydropyridine), a single dose of which can damage the substantia nigra of the basal ganglia so severely that it induces an extreme form of Parkinson's disease. Indeed, this simple method of integrating scores might not deal adequately with any substance that is extremely harmful in only one respect. Take tobacco, for instance. Smoking tobacco beyond the age of 30 years reduces life expectancy by an average of up to 10 years,9 and it is the commonest cause of drug-related deaths, placing a huge burden on health services. However, tobacco's short-term consequences and social effects are unexceptional. Of course, the weighting of individual parameters could be varied to emphasise one facet of risk or another, depending on the importance attached to each. Other procedural mechanisms, such as those of multi-criteria decision analysis,20 could be used to take account of variation of ranking across different parameters of harm. Despite these reservations about the interpretation of integrated scores and the need for further consideration of the weighting of parameters of harm, we were greatly encouraged by the general consistency of scores across scorers and across parameters of harm for most drugs.

Our findings raise questions about the validity of the current Misuse of Drugs Act classification, despite the fact that it is nominally based on an assessment of risk to users and society. The discrepancies between our findings and current classifications are especially striking in relation to psychedelic-type drugs. Our results also emphasise that the exclusion of alcohol and tobacco from the Misuse of Drugs Act is, from a scientific perspective, arbitrary. We saw no clear distinction between socially acceptable and illicit substances. The fact that the two most widely used legal drugs lie in the upper half of the ranking of harm is surely important information that should be taken into account in public debate on illegal drug use. Discussions based on a formal assessment of harm rather than on prejudice and assumptions might help society to engage in a more rational debate about the relative risks and harms of drugs.

We believe that a system of classification like ours, based on the scoring of harms by experts, on the basis of scientific evidence, has much to commend it. Our approach provides a comprehensive and transparent process for assessment of the danger of drugs, and builds on the approach to this issue developed in earlier publications5–8,11,12,21,22 but covers more parameters of harm and more drugs, as well as using the delphic approach, with a range of experts. The system is rigorous and transparent, and involves a formal, quantitative assessment of several aspects of harm. It can easily be reapplied as knowledge advances. We note that a numerical system has also been described by MacDonald and colleagues23 to assess the population harm of drug use, an approach that is complementary to the scheme described here, but as yet has not been applied to specific drugs. Other organisations (eg, the European Monitoring Centre for Drugs and Drug Addiction24 and the CAM committee of the Dutch government25) are currently exploring other risk assessment systems, some of which are also numerically based. Other systems use delphic methodology, although none uses such a comprehensive set of risk parameters and no other has reported on such a wide range of drugs as our method. We believe that our system could be developed to aid in decision-making by regulatory bodies—eg, the UK's Advisory Council on the Misuse of Drugs and the European Medicines Evaluation Agency—to provide an evidence-based approach to drug classification.