CANNABIS AND ROAD SAFETY

CANNABIS AND ROAD SAFETY: AN OUTLINE OF THE RESEARCH STUDIES TO EXAMINE THE EFFECTS OF CANNABIS ON DRIVING SKILLS AND ON ACTUAL DRIVING PERFORMANCE

Dr G.B. Chesher

Department of Pharmacology University of Sydney and National Drug and Alcohol Research Centre University of New South Wales.

Dr Chesher provides an extensive coverage of the latest Australian and overseas research on the impairing effects on driving of cannabis, particularly relative to those of alcohol.

Contents:

1. Executive summary

2. Introduction

2.1 Pharmacology and pharmacokinetics
2.2 Behavioural pharmacology and psychology

3. Studies using the techniques of epidemiology

3.1 Pharmacology and pharmacokinetics
3.2 Pharmacokinetics

4. A comparison of the effects of alcohol and cannabis on skill performance and driving skills

4.1 Laboratory tests
4.2 Duration of cannabis-induced impairment in laboratory tests
4.3 The effect on laboratory tasks of alcohol and cannabis in combination
4.4 Driving simulators
4.5 On-road driving

5. Epidemiology

5.1 Questionnaire based surveys
5.2 Incidence of drug detection in crash involved drivers
5.3 Attempts to assess whether or not the driver who has detectable drugs in the blood stream was culpable in the accident

6. The use of 'Responsibility Analysis' or estimation of 'culpability' to determine the role of drugs in crashes

7. Alcohol and cannabis in epidemiological studies

8. Summary (of the evidence presented above)

1. EXECUTIVE SUMMARY

There is no doubt that cannabis, smoked or taken by mouth produces a dose-related deficit in tests of performance skills as conducted in a laboratory.

Using driving simulators and on-road real vehicles, cannabis has been shown to affect driving performance. However, the effects are less severe than would be anticipated from the evidence obtained from the laboratory studies of individual tests of skills performance.

A description is given of epidemiological studies to determine the role of cannabis in road crashes. The pharmacological problems associated with these studies are described. The results of studies within the last 10 years have failed to present clear evidence for a role of cannabis in road crashes. The role of alcohol in all studies has proved to be dominant.

The evidence indicates that there is a clear difference in the mode of action of cannabis and alcohol, both pharmacological and behavioural and this is presented and the implications described.

The most recent of studies of cannabis and driving (Robbe & O'Hanlon, 1993), which was sponsored by the U.S. National Highway Safety Traffic Administration included a review of the literature. The authors' comments in summary of their literature review and of their own results include the following:

Of the many psychotropic drugs, licit and illicit, that are available and used by people who subsequently drive, marijuana may well be among the least harmful. Campaigns to discourage the use of marijuana by drivers are certainly warranted. But concentrating a campaign on marijuana alone may not be in proportion to the safety problem it causes.

2. INTRODUCTION

In this paper I will examine briefly the studies which have sought an understanding of the effect of cannabis and of alcohol on driving skills and their role in road crashes. This information has been based upon scientific data which have been collected from several scientific disciplines. I have outlined these in earlier papers and will only mention them briefly here.

The major purpose of this paper is to compare the two drugs, alcohol and cannabis and the status of the evidence as to their role in road crashes.

The determination of the legal limit for alcohol has been achieved in a scientific manner. There are pharmacological reasons why it has not been possible to follow these same techniques with drugs other than alcohol, including cannabis. This paper will draw attention to these problems.

First, we might briefly outline the nature of the evidence which has been generated to examine the effects of cannabis on driving skills and as a causative factor in road crashes. This information has been derived from the employment of three scientific disciplines:

2.1 Pharmacology and pharmacokinetics

Pharmacology is the study of the way a drug exerts its action in the body. This involves an understanding of the sites and the body systems where the drug acts and the consequences of this drug-system interaction. Information obtained from these studies can help to formulate an hypothesis as to how the drug may influence driving behaviour.

The pharmacological discipline known as pharmacokinetics studies the fate of the drug after it has been taken. It provides information as to the rate of absorption from the site of administration; the manner of its distribution in the body up to the delivery to its site of action (eg. the brain). Pharmacokinetics also studies the way the body eliminates the drug from the body and includes the understanding of the metabolism and excretion of the drug.

2..2 Behavioural pharmacology and psychology

These involve studies of the effects of the drug on human behaviour. The behaviour of relevance to this discussion concerns those skills which are (or are related to) those necessary for the safe control of a motor vehicle or other items of machinery. Psychological studies also involve the effects of the drug on mood and cognition.

The three classifications of these studies are:

(i) Those performed on specific tests of behaviour or psychological functioning (for example, tests of reaction times of various degrees of complexity; tracking; divided attention or vigilance);

(ii) Those performed in a driving simulator; and

(iii) Those performed in a real car, either in a closed course or in real traffic.

3. STUDIES USING THE TECHNIQUES OF EPIDEMIOLOGY

These studies aim to determine whether or not a causal relationship between drug use and a motor vehicle crash exists.

I shall look at each of the above factors and will compare the two drugs alcohol and cannabis in the light of current evidence. In interests of time and space I have in this summary referred to reviews of the literature and have made only a brief description of the studies themselves. A fuller description of these can of course be sourced from the original literature of the cited reviews.

3.1 Pharmacology

First, the drugs themselves. With the increase in pharmacological knowledge it is known that most drugs act upon specific receptors. A receptor is a specific site in tissues, frequently on the cell membrane, which has a specific structural affinity (shape) for a naturally occurring molecule. The interaction between receptor and the endogenous molecule is part of the body's normal, physiological functioning. Most drugs exert their activity by acting upon these receptors. Examples of such drug-receptor interactions are the opioids (morphine etc) and the opioid receptors; the antihistamines and the histamine receptors and the benzodiazepines which act on the benzodiazepine receptors. The endogenous substances that physiologically act on these receptors are, respectively, the endorphins and enkephalins on the opioid receptors; histamine on the histamine receptors; however the identification of the physiological substance for the benzodiazepine receptor has yet to be identified.

Research within the last five years has revealed that the cannabinoids, such as delta-9-tetrahydrocannabinol (THC) from the cannabis plant exert their effects on specific receptors known as the cannabinoid receptors. To date two cannabinoid receptors have been described and an endogenous (physiological) substance has been identified. This has been given the name 'anandamide'. It is very likely that in the near future more cannabinoid receptors will be described and more endogenous substances that act on these receptors will be identified. An historical overview of these findings has recently been published.

In contrast, the evidence strongly indicates that the drug alcohol does not act on a specific receptor, but acts more widely in a non-specific manner on the cell membranes themselves. This understanding is supported by the evidence that alcohol exerts effects on most of the tissues of the body and in excess is toxic to most tissues. The reader is referred to a recent review on this subject by Dufor and Caces.

Drugs which act upon a specific receptor produce their effects in doses measured usually as nanograms or micrograms per kilogram of body weight. Alcohol doses are measured in grams per kilogram - many hundreds of thousands times greater than those of most other drugs. Alcohol is a very non-specific drug.

Another important factor is that receptor-specific drugs exert their activity only on those cells which bear the specific receptor. In the case of the cannabinoids these receptors are found only in the brain in the basal ganglia, the cerebellum, the brain stem, thalamic nuclei, hypothalamus and corpus callosum. On the other hand alcohol affects all nerve cells to which it is delivered by the circulating blood.

Consequently it is not surprising that differences in the action of alcohol and the cannabinoids have been described in their effects on mood and behaviour. These will be discussed below.

3.2 Pharmacokinetics

The pharmacokinetics of alcohol and the cannabinoids could hardly be more different.

The apparent volume of distribution of alcohol (the volume of fluid in which the drug seems to be dissolved throughout the body) is quite low, consisting of the 41 litres of body water, providing a value of about 0.59 litres/kg. Cannabinoids, on the other hand, are very fat soluble and have a high volume of distribution which has been estimated to be about 10 litres/kg.

The meaning of these values is that the concentration of alcohol in the blood provides a reliable estimate of the concentration of the drug in the brain. This in turn provides a reliable estimate of the degree of impairment of the drinker. In addition to this, alcohol is excreted via the lungs to the breath and the blood : breath ratio is such that the determination of the alcohol in breath provides a reliable estimate of the blood alcohol concentration. It is because of these pharmacokinetic properties of alcohol that it has been possible to accumulate the epidemiological data upon which our drink-driving laws have been based.

Cannabinoids, on the other hand are lipophilic (fat loving) and are distributed in the fatty tissues of the body. When smoked, which is the most common route of administration, the cannabinoids are rapidly absorbed from the lungs into the bloodstream. Being so fat soluble the cannabinoids readily cross membranes, leave the circulation and are rapidly 'dumped' into various tissues of the body, including the brain. In this way the concentration of cannabinoid in the blood declines very rapidly as indicated in Fig 1.

As indicated in the Figure, we can describe the concentration of cannabinoid across time in the blood in the three phases: absorption, re-distribution and elimination. The steep upward curve of THC represents the inhaled THC being absorbed into the blood through the lungs; the equally sudden drop in the concentration of THC represents the drug being 'dumped' from the bloodstream into fatty tissues. This redistribution phase 'flattens' out as the 'dumped' THC re-enters the blood and is then metabolised in the liver-the elimination phase. It is important to note that the sudden decline in the concentration of THC (the psychologically active cannabinoid) in the blood does not represent drug metabolism but rather the rapid re-distribution of the drug from the blood into other tissues. The metabolism of the cannabinoids takes place when these 'dumped' cannabinoids are released back into the bloodstream whence they pass through the liver and are very rapidly metabolised and subsequently excreted.

Figure 1. The blood concentration of THC (squares) and its inactive metabolite, carboxy THC (THC Acid; diamonds) after the smoking of a marijuana cigarette. Each point is the mean of results from six volunteers, all of whom were free from cannabinoids before smoking the drug. [The 925±7mg refers to the average weight of the cigarettes and the 1.32% refers to the dry weight concentration of THC]

Figure 1 also shows the blood picture of the inactive metabolite, carboxy THC (or THC acid). It is important to note several points about the pharmacokinetics of this substance. First, in the study indicated here (Fig 1) all of the volunteers had no cannabinoids in their blood before they began smoking. Second, the THC acid is formed in the liver from the metabolism of THC, therefore its appearance in blood follows that of the parent, THC. Third, the THC acid concentration then increases and surpasses that of the parent molecule in the blood. At a time when the parent THC is in the blood at only a very low concentration, that of the metabolite is higher and exists in the blood for a longer time. Therefore, should the smoker smoke again before the parent molecule and its metabolite have been eliminated, the ratio of the concentrations of THC and of the THC acid will be different from that shown in Figure 1. This is because there will exist a higher concentration of the metabolite than of the THC in blood at the time when the next dose of cannabis is smoked.

For this reason, analytical data that provides a value only for the metabolite can only be validly interpreted as indicating recent consumption of cannabis; however the time of this consumption could be a matter of hours or days. For this reason the quantitative determination of only the metabolite is of no value to determine possible impairment.

To assess possible impairment the analyst must provide data for the active molecule, THC. And when this occurs, the only interpretation possible on present knowledge is to infer the recent consumption of the drug by smoking. To date no meaningful correlation between blood concentration of THC and impairment in laboratory tasks has been established. This point will be clarified when the results of the recent epidemiological studies are discussed below.

Yet another problem arises in the interpretation of blood concentrations of cannabinoids. The pharmacokinetics of the cannabinoids are quite different when the drug is taken by mouth. Space in this discussion precludes further discussion of the pharmacokinetics after oral administration, but suffice to say the absorption of cannabinoids taken orally is slow and erratic. The absorbed THC passes through the liver and is rapidly metabolised. This results in a different proportion of THC to the metabolite, THC acid than encountered after smoking. There is a greater amount of entero-hepatic 'recycling' as some of the cannabinoids are stored in the bile in the gall bladder. These cannabinoids can later be 'recycled' and reabsorbed into the bloodstream when the gall bladder empties. In this country, most who use cannabis, smoke it.

It is also important to note that the detection of cannabinoids in a urine sample provide evidence only that the donor of that urine has been exposed to cannabis at some time in the past. It gives no indication at all of impairment or of intoxication. A frequent, heavy cannabis user may be excreting cannabinoids in urine for some weeks or in some cases, for more than a month. Those who take the drug by mouth also will be excreting the drug for a longer period.

4. A COMPARISON OF THE EFFECTS OF ALCOHOL AND CANNABIS ON SKILLS PERFORMANCE AND DRIVING SKILLS

4.1 Laboratory tests

Laboratory tests isolate specific psychological functions and determine the skill of the test subject on that function. Most studies test each volunteer on each test before and after taking the drug. For testing alcohol and cannabis, the choice of these tests rests upon an assessment of their relationship to the task of driving a motor vehicle. However, the fact is that no battery of separate tests comprehensively defines the actual task of driving. In fact, Joscelyn and others (Joscelyn et al., 1980) examined the plethora of methods employed in these studies and commented:

... many tests routinely employed have limited validity or no demonstrable relation to real-world driving. Measuring the 'same' behaviors often differ, raising questions about the comparability of experimental findings.

Laboratory tests, nevertheless do provide a 'screening' of the potential for drugs to impair specific behaviours. However, results from such laboratory testing should not form the sole basis for any judgement of the potential of a drug to impair actual driving skills or to increase the probability of an accident. For this reason, evidence for the traffic hazard associated with any drug should be confirmed by studies of actual driving (either using driving simulators or a real car) and by studies using epidemiological methods.

The data from laboratory testing of alcohol has been reviewed by Moskowitz and Austin and of the effects of cannabis by Klonoff, Moskowitz, and by Chesher. It is clear that both alcohol and cannabis cause dose-dependent deficits in the performance of specific laboratory tasks.

It is to be noted that the doses of cannabinoids in these tests are lower than those in use by many smokers of cannabis today. However, they may have been appropriate to the cannabis experience of the volunteers when these studies were conducted. In many of these studies, the volunteers were asked to rate the effect of the dose given with that of their general experience with the drug. In many (but not all) cases the doses given produced subjective effects which were as great as those generally experienced by the volunteers in their social use of the drug.

Looking at the Australian studies across time, from the 1970s to the 1990s these observations are in accord with the results expressed in a recent publication concerning the patterns of cannabis use in Australia. The earlier studies produced deficits in testing which were greater than those in the later studies. The data presented by Donnelly and Hall (1994) indicate that:

The prevalence of cannabis use seems to have been very low by contemporary standards in the early 1970s. It increased substantially throughout the 1970s and 1980s, levelled off in the late 1980s, and has probably shown a small increase in the early 1990s.

The phenomenon of tolerance to cannabis is well established and this in turn is a serious confounding variable in the studies with this drug. Tolerance develops with the regular and frequent use. This in turn depends upon the pattern of use of those in the study sample. The correlation of performance : dose : and tolerance requires further study. There is very little information available as to the change in doses used across the years since the 1970s as most data refer only to frequency of use. Studies involving high doses of cannabis should be undertaken, but with due consideration given to the degree of tolerance of the volunteers to be studied.

The Australian data presented by Donnelly and Hall indicate that:

Most cannabis use is infrequent and intermittent, with about three-quarters of adult women and two-thirds of adult men having discontinued their use, or continued to use less often than weekly. The proportion of users who are weekly users is highest in the younger age groups. Rates of weekly and lifetime use are highest among those aged 20 to 24 years, and decline markedly with increasing age.

4.2 Duration of cannabis-induced impairment in laboratory tests.

Most studies have reported a duration of cannabis-induced impairment of the order of 4 hours. On the other hand there have been three studies which have reported a longer duration of cannabis effects of between 10 to 24 hours. However, these reports have been questioned for methodological or reasons of interpretation. That of Yesavage et al. did not include a control group. Subsequently the study was repeated by Leirer et al. in an attempt to replicate this effect using a control group but was only able to show an effect up to four hours after smoking (ie. that described in the many other studies of this effect). A third study, also with a control group, did demonstrate an effect at 24 hours after smoking. The statistical significance of the effect required a statistical procedure (one tail 't' test) which is of questionable validity when there was no previous statistical proof that the effect was expected. This means that the effect was at best, only marginally significant. The study by Moskowitz et al, as described in Moskowitz's 1985 review (Moskowitz, 1985) was of a:

.... compensatory tracking task performed while simultaneously executing a visual search task as well as a critical tracking task. Performance was significantly impaired on the compensatory tracking task for more than 2 hours and upon the critical tracking task for up to 10 hours, albeit, intermittently during the period from 4 hours on. [emphasis added]

At present I think it is fair to conclude that the evidence for the long duration of cannabis induced impairment requires more study to confirm its validity. Furthermore, both tasks in which it was described are very difficult tasks. It has been argued that the use of cannabis by pilots in the 24 hours preceding flying may be more an indicator of poor judgement rather than a cause for concern about the residual psychomotor effects of cannabis.

4.3 The effect on laboratory tasks of alcohol and cannabis in combination

The effect of this drug combination has been reviewed and only an outline will be given here.

There is very clear evidence from numerous studies of the effect of alcohol and of cannabis on the performance of specific tasks in the laboratory. Both drugs produce a dose related impairment on these tasks and the effect of the drugs when given in combination is essentially additive. Although of more academic than practical interest is the evidence as to the nature of this additivity. Several studies have observed a trend that the effect of cannabis plus alcohol is less than additive, meaning that 1 + 1 is less than 2. In the most recent study, Dauncey et al. reported this effect, found to be statistically significant, and termed it to be a 'de-intensification'. In the light of the present knowledge of the quite different mode of action of cannabis and alcohol such an interaction is not necessarily surprising.

What is quite surprising and important however, is the result of a study by Perez-Reyes. For pharmacological reasons the researchers studying the alcohol-cannabis interaction administered the drugs such that the peak of blood concentration of both drugs occurred as near as possible at the same time. Such is the thinking of the pharmacologist! Indeed Perez-Reyes and his colleagues had reported such a study showing an additive decremental effect of the drug combination. Interestingly, in their later study they had the volunteers smoke marijuana (placebo; 1.7% and 3.58% THC) before they commenced drinking alcohol (0.85g/kg) over a period of 30 mins. This would have produced a BAC of the order of 0.1g%. Their results showed a dose-dependent effect for cannabis and the characteristic effects expected for the one dose of alcohol. However, no significant interaction between the two drugs was recorded. The authors concluded:

The lack of interactive effects, particularly on psychomotor performance, highlights the influence that the order of administration of the companion drug has on its interaction with the reference drug.

4.4 Driving simulators

A driving simulator is also a laboratory based apparatus. It is important to realise that it is only a simulation of real life driving and driving simulators vary greatly in the degree to which they can simulate the real event. It is fair to say that all but the most sophisticated and extremely expensive simulators are to the test subject, still a laboratory piece of equipment. They lack realism both in the dynamics of car driving and in the visual presentation of the road and other traffic. Nevertheless they are able to present simulated dangerous presentations to which the driver must respond. The effects of cannabis on performance in a driving simulator have been reviewed and a summary only is given here.

The early driving simulator studies, for the driver, were not interactive with the 'driving scenery' which was generally a film of the road to be covered and the driver had little or no control over the presented imagery.ÊÊ These studies showed no significant effects of marijuana on car control. However marijuana did produce the following effects, namely:

(a) An increase in decision latency before starting, stopping or overtaking;

(b) Impaired monitoring of a speedometer; and

(c) Reduced risk-taking behaviour in tasks requiring a decision to overtake a vehicle in the presence of an oncoming car.

Later simulator studies with apparatus with a more realistic driving dynamics and an interaction between 'scenery' and the driving manoeuvres did show marijuana effects on car control. The study by Smiley et al. found that cannabis increased lateral position variability, headway variability, and caused the 'driver' to miss more signs that indicated the need to follow another route. On the other hand, cannabis caused the subjects to drive in a more conservative manner inasmuch as they maintained a longer headway when car following, refused more opportunities to overtake a vehicle in front and when they accepted this opportunity, they began to do so at a greater distance from the approaching vehicle. The effects of alcohol (at about 0.08g% BAC) in this study were surprisingly small.

Another and very similar study by Stein et al. showed alcohol effects were as one would expect and significantly affected practically every performance parameter. Alcohol (at about 0.1g% BAC) was associated with significantly increased 'accidents' (hitting obstacles or exceeding road edges by a full car width) and 'traffic tickets' (exceeding speed limit by 32 'radar checks'). Alcohol was also associated with increased lane deviations, speed variability, response times to signs, and errors in sign recognition. In contrast, cannabis was associated with few changes. The mean speed travelled was lower and two measures of steering control changed significantly. Alcohol and cannabis in combination were associated with more adverse reactions than alcohol alone. Alcohol was consumed first and the performance testing was begun 15 minutes after the end of cannabis smoking.

4.5 On-road driving

Driving studies with a real car, conducted in an open field, of course present a more realistic experience of a motor vehicle than do simulators. However they usually require the driver to undertake manoeuvres that are not necessarily part of normal driving - such as weaving between cones. Those studies undertaken in on-road traffic naturally require great care on the part of the experimenter to avoid dangerous driving. Therefore these studies are restricted in the measures that can be realistically taken. They are somewhat akin, for both the experimenter and the test driver, to a driver undertaking a test for a driving licence. Indeed, experimental studies of the effects of drugs using in-car performance have been described by Smiley as being really a simulation of real driving.

On-road driving studies vary considerably in their experimental design and in the tests of driving employed. In this paper, only the broadest outline of the results is given in the interests of brevity. Reviews of these studies have been presented and published. The reader is referred to the original studies or to the cited reviews for more information.

There have been to date, seven on-road studies to examine the effects of cannabis on driving performance. Each of these is outlined below:

1. Klonoff studied volunteers in a closed course as well as in-traffic on a city road. The closed course study comprised eight tests and the response scores rested essentially on the number of cones struck. Testing was conducted in 4 blocks, each of 5 trials. The first three were taken as practice and the fourth, after drug treatment, were the test trials. The anticipated scores in the fourth block were determined by regression analysis on the assumption that the rate of learning or performance would continue at the same rate. Using this technique the author concluded that there was an impairment under cannabis. While the mean of the impairment was not large, the trend was clear.

The city traffic study was conducted rather in the manner of a driving test by a driving examiner. The subjects drove for about 45 minutes on a course of 16.8 miles after being given their dose of cannabis. A strong trend towards impaired performance was indicated by the lower scores given by the examiner on judgement and concentration after the higher dose of cannabis.

2. Hansteen et al. conducted a closed course study in which subjects were required to drive six times around a 1.1 mile course set out on an airfield. The course was set out with cones and poles and the number of these hit were counted. The course involved curves and straight sections and drivers were required to undertake various manoeuvres. The mean number of struck objects per lap increased from a mean of 13.2 in the placebo condition, 13.4 in the low cannabis dose, 16.8 for the high cannabis and 17.4 for the alcohol dose (BAC 0.07g%). The effects for the high cannabis dose and the alcohol dose achieved significance.

3. Casswell conducted a closed course study in which the behaviours sampled were more typical of those for real driving, than for the studies outlined above. Driving behaviours recorded included overtaking, responding to road signs, making a hairpin turn and driving through a narrow gap. A subsidiary reaction time task was also included to monitor attention. Driving behaviour under cannabis, alcohol and the combination was tested. After alcohol, and alcohol plus cannabis, the subjects showed poorer tracking performance and drove at increased speed over various segments of the course, including the hairpin bend, and the straight section. Under alcohol alone, the speed through the narrow gap was also increased.

On the other hand, marijuana alone was not accompanied by steering or tracking errors. The mean speed dropped significantly after cannabis, both on the hairpin bend and on the straight section of the course.

Casswell suggested that drivers under the influence of cannabis appeared to compensate for what they perceived as being an adverse effect on driving. Compensation was exhibited by driving more slowly. This contrasted with the effects of alcohol. The increased reaction times to the subsidiary task under cannabis suggests an effect on attention. The extent of this effect was of the same order as that measured by the author in another study after 8 hours of continuous driving.

4. Attwood conducted a study on a closed course constructed on an airfield and, like Casswell, used measures appropriate to real driving including acceleration, following a lead car which varied its speed and responding to 'traffic signals'. The drug effects (alcohol, and two doses of cannabis alone and together with alcohol) recorded were not particularly robust, even with a complicated multivariate analysis which did distinguish the treatment conditions from each other.

5. The study by Peck and colleagues (Peck et al., 1986) from the California Department of Motor Vehicles, is best summarised by the authors' own summary.

Approximately 80 volunteer male marijuana and alcohol users received one of four experimental treatments: (1) marijuana, (2) alcohol, (3) marijuana and alcohol, or (4) double placebo.

After consumption, each subject drove a vehicle over a test course which simulated a number of real-world driving conditions.

Four post-drug runs were involved, separated by one hour intervals. The subject's performance was rated by an in-car examiner, outside observers, and computerised vehicle measurements.

Blood and urine specimens were extracted after each run to establish levels of tetrahydrocannabinol (THC), serum carboxy, and alcohol. A variety of multivariate statistical techniques were applied in evaluating treatment effects.

Both marijuana and alcohol had significant effects on driving performance, and the effects were particularly detrimental under the both-drugs treatment. The effects of marijuana were more rapid than those of alcohol and somewhat less severe for most tasks.

In this study cannabis was smoked after the consumption of the alcohol dose. In discussing their results and comparing them with other studies, they had this to say:

There is a vast amount of empirical evidence documenting the effects of marijuana on a wide array of human performance measures-cognitive, psychomotor and affective. Although the literature has clearly established that marijuana affects all three domains and results in detriments in the ability to perform many psychomotor and cognitive tasks, the evidence is somewhat more equivocal on the question of actual driving skill and even more equivocal on the question of those aspects of driving skill that are related to safety and accident avoidance. [Emphasis that of Peck et al.]

6. Smiley et al. tested the effects of cannabis (placebo and two doses) and alcohol (placebo and BAC of 0.05 g%) in combination and the effect of alcohol alone (BAC 0.08g%) on driving in a closed course study using an instrumented car.

The high dose of cannabis significantly increased headway and headway variability (ie distance from a car in front). Alcohol alone at the BAC 0.05g% produced an increase in speed, both in the straight sections of the road and in curves. In her review of her own study, and those of others, Smiley (Smiley, 1986) concluded:

In conclusion, marijuana does appear to impair driving behaviour. However, this impairment is mediated in that subjects under marijuana treatment appear to perceive that they are indeed impaired. When they can compensate, they do, for example, by not overtaking, by slowing down and by focussing their attention when they know a response will be required. Unfortunately, such compensation is not possible where events are unexpected or where continuous attention is required. Effects on driving behaviour are present shortly after smoking but do not continue for extended periods. [emphasis added]

7. The most recent and most comprehensive study of the effect of cannabis on driving on city roads and a public highway is that conducted in The Netherlands and was sponsored by the U.S. National Highway Safety Traffic Administration. An intelligent departure in methodology in this study from the others reviewed here is that the dose of cannabis used was determined in a pilot study using the volunteers who were to take part in the main study. The aim was to estimate the dose these volunteers generally use on a social occasion. Accordingly socially appropriate doses (for these subjects) were chosen for the driving study. Three driving studies were then performed. The first was conducted on a closed section of a public highway with no traffic; the second on a highway with traffic and the third in city traffic. The measure they have found to be of significance is the standard deviation of lateral position on the roadway (SDLP). It is a measure of the 'automatic' function of information processing in the driving task. Cannabis, in all tests produced a dose-related increase in the SDLP. Mean speed was somewhat reduced under cannabis as was the headway distance from the lead vehicle in the test in highway traffic.

The test under city driving conditions was conducted under one dose of cannabis and as a comparison, subjects were also tested under alcohol at a BAC of 0.04g%. Results in this test showed that this modest dose of alcohol, but not cannabis, produced a significant impairment of driving performance relative to placebo. Alcohol impaired driving performance but subjects did not perceive it. Cannabis did not impair driving performance yet the subjects thought it had. After alcohol, there was a tendency towards faster driving and after cannabis, slower.

This research group has conducted many studies with the same methodology and has accumulated much data on the effects of other drugs. They therefore were able to indicate the extent of the impairment on the measure of SDLP. The greatest effects of cannabis in this study were 3.7 and 2.9cm. In other studies drugs, for example diazepam (Valium), or lorazepam (Ativan), produced increases of 7 and 10cm respectively. The authors commented:

In so far as its effects on SDLP are concerned THC was just another moderately impairing drug.

The authors go on to say that the effects of cannabis differ qualitatively from those of other depressant drugs, especially alcohol:

Very importantly our city driving study showed that drivers who drank alcohol overestimated their performance quality whereas those who smoked marijuana underestimated it. Perhaps as a consequence, the former invested no special effort for accomplishing the task whereas the latter did, and successfully. This evidence strongly suggests that alcohol encourages risky driving whereas THC encourages greater caution, at least in experiments.

Finally, Robbe contrasted the effects of cannabis when measured with laboratory based, individual tests in the laboratory, with those conducted in an on-road vehicle:

The results of these studies corroborate those of previous driving simulator and closed-course tests by indicating that THC in single inhaled doses up to 300 µg/kg has significant, yet not dramatic, dose-related impairing effects on driving performance. They contrast with results from many laboratory tests, reviewed by Moskowitz (1985), which show that even low doses of THC impair skills deemed to be important for driving, such as perception, coordination, tracking and vigilance. The present studies also demonstrated that marijuana can have greater effects in laboratory than driving tests. The last study, for example showed a highly significant effect of THC on hand unsteadiness but not on driving in urban traffic.

5. EPIDEMIOLOGY

The studies outlined above indicate that cannabis does cause dose-dependent effects on laboratory based tests of human skills. Furthermore, studies utilising driving simulators and on-road driving also indicate a degree of cannabis induced impairment of driving skills. However in these cases the extent of the impairment indicated from laboratory studies is not replicated in the simulator or in-car studies.

The effects of alcohol on the other hand can be demonstrated both in laboratory studies and in simulated or on-road driving at very much the same dose levels. Explanations for these differences between alcohol and cannabis have been suggested and rest essentially upon the difference in the awareness by the drug taker of the presence of drug impairment. This in turn may be explained by the present understanding of the quite different ways alcohol and cannabis are known to act on the brain.

Also mentioned above and in other publications our present laws on alcohol and driving have been based upon the scientific principles outlined here and in particular on the results of epidemiological studies. It is pertinent therefore to discuss briefly the nature of the epidemiological studies undertaken to date with cannabis and road crashes.

Epidemiological studies with alcohol are greatly facilitated by the pharmacokinetics of that drug. Alcohol is excreted in the breath and the ratio of the concentration on the breath and in the blood is relatively constant. Therefore the determination of the concentration of alcohol in the breath (by a 'breathalyser') provides a reasonably and acceptably accurate indication of the blood concentration. It is unfortunate therefore that cannabinoids are not excreted on the breath and the concentration of cannabinoids that can be detected on breath represent only that contained in the 'dead-space air' in the upper respiratory tract. The cannabinoids so detected do not correlate in any way with the blood concentration. In addition to this the blood concentration of cannabinoids do not show any useful relationship to the degree of impairment or the degree of subjective effects of the drug. The blood concentration of alcohol on the other hand does exhibit a reasonable correlation with the degree of impairment.

These properties of cannabis mean that the determination of the role of cannabis in road crashes by the same techniques of the case-control study as used for alcohol, is not an easy task. The pharmacokinetics of cannabis make this an exceedingly difficult task. The difficulty is not only related to the poor correlation between blood concentration and impairment, but also because it requires the collection of a blood sample-from both the crash case and the controls. The collection of the latter sample is likely to involve a high refusal rate, and this alone would almost certainly invalidate the study. One does not know the reason for the refusal!

The studies that have been undertaken to date can be described within three groups and these are:

(i) Questionnaire based surveys;

(ii) Incidence of drug detection in accident involved drivers; and

(iii) Attempts to assess whether or not the driver who has detectable drug in bloodstream was culpable in the accident.

Studies along the lines outlined above have been reviewed by Simpson.

5.1 Questionnaire based surveys

Questionnaire based surveys by definition depend upon self report data and their reliability is questionable. Furthermore, the incidence of cannabis use and the likelihood of a driver admitting to such use is likely to change across time.

5.2 Incidence of drug detection in crash involved drivers

This technique involves the analysis of blood or urine samples taken from crash involved drivers. The detection of cannabinoids in urine provides information only that the drug has been consumed within the last day or even month. It provides no indication at all of impairment. Therefore only the analysis of a blood sample is likely to be helpful. However, the detection of cannabis in a blood sample does not itself prove impairment or crash culpability. This fact has been well expressed by Compton as follows:

Knowing only the frequency with which crash-involved drivers use drugs does not allow one to know the danger posed by the drugs. It may simply reflect the general drug usage pattern in the driving public at large. For example, finding that 30% of crash-involved drivers have nicotine in their blood does not imply that nicotine was involved in the occurrence of their crashes. It may be that 30% of the general driving population smokes cigarettes and the smoking of cigarettes is unrelated to crash occurrence. Finding that a drug was overrepresented in crash-involved drivers (as compared to non-crash involved drivers) would strongly suggest it played a role in increasing crash risk. However, this approach requires knowing the drug usage rate of the general driving public, something we do not know and can not easily determine.

Furthermore, any comparisons of the incidence of cannabis detections in crash-involved drivers with those of non-crash involved drivers should be collected from a comparable population and at the same time. The patterns of cannabis use vary not only across time but also across populations.

Therefore studies reporting the incidence of drugs in the blood of crash-involved drivers is essentially meaningless without some control of the incidence of drug use in non-crash involved drivers. Nevertheless, such studies have been reported and are reviewed by Simpson who summarised that:

Marijuana users are certainly among drivers who are injured in road crashes (suggested by the presence of cannabinoids in urine);

More importantly, recent use, as indexed by the presence of THC in blood, is evident in perhaps less than 10% of injured drivers; and

When cannabis is detected, there is an 80% chance that alcohol will also be found.

5.3 Attempts to assess whether or not the driver who has detectable drugs in the bloodstream was culpable in the accident

Of the first attempts to assess culpability has been an ongoing series of data collected by McBay of fatal, single vehicle crashes. Culpability in single vehicle crashes is assumed to be that of the driver (assuming no mechanical fault can be found) and the choice of fatal crashes assumes that death occurred shortly after the accident; meaning that drug metabolism ceased at death and therefore the blood sample from the dead body will represent the blood picture at the time of the crash. Cannabis was detected in 7.8% of 600 such cases, but 88% of these also contained alcohol in concentrations which of themselves could have accounted for the crash.

6. THE USE OF 'RESPONSIBILITY ANALYSIS' OR ESTIMATION OF 'CULPABILITY' TO DETERMINE THE ROLE OF DRUGS IN) CRASHES

In the absence of a separate control group (as used in the assessment of crash probability with alcohol as described above) an alternative of a 'culpability index' is currently being employed in drug studies. The basic construct is first to formulate a means of determining the responsibility or culpability of a driver involved in a crash. There have been several means of constructing this 'culpability index' and this must be done with each of the accident cases by observers who have no information as to the drug status of each driver. The responsibility (or culpability) ratio is then determined as the proportion of drug-bearing drivers who were determined to be culpable, to the non-drug bearing drivers who were deemed to be culpable. The null hypothesis predicts a culpability ratio of 1.00 (ie, the drug has had no causal relationship with crashes).

To date there have been six studies employing this technique (two of which have involved the re-analysis of earlier generated data). These are briefly outlined below:

1. Warren and others re-analysed the data of Cimbura and found a culpability index for cannabis of 1.7, the same as that found for alcohol. However, the original data comprised a total of 484 drivers and pedestrians, 3.7% of whom were positive for cannabis. However, 88% of these people were also positive for alcohol. This left a very small number from which to assess a culpability ratio for cannabis alone.

2. Terhune also has previously collected data independently re-analysed to estimate a culpability ratio. All BACs over 0.10% were judged significantly more culpable than the drug-free group. The cannabis group also had a higher culpability ratio than the drug-free group, but this was only marginally significant (58.8% vs 34.4%). This estimation was also compromised by the small sample size for cannabis only (n=17). The cannabis plus alcohol group was analysed separately.

3. Donelson began a very ambitious project but was unfortunately thwarted by funding problems which precluded the complete analysis of the collected data. However, a random sample of 415 cases was analysed. The results cautiously suggested a finding consistent with those of Warren et al. and Terhune above.

4. Williams et al. in a study involving 440 cases, demonstrated as in the above studies that alcohol had a higher culpability ratio compared with culpable drug-free drivers (92% vs 71%). However, those drivers in whom only cannabis was detected were less likely to be responsible for the crashes (53% vs 71%).

5. Terhune et al. reported a very comprehensive study involving 1 882 cases. They found that alcohol was the dominant drug in fatal crashes, although the basic focus of their research was to describe the effect of drugs other than alcohol. They reported that fully 40% of the drivers had only alcohol in their systems and another 11% had alcohol combined with drugs. Among the drivers with BACs at or above 0.10% (n=625) their responsibility rate:

... was an extraordinary 94%, well above that found for any other single substance.

Of cannabis, the authors stated that while cannabinoids were detected in 7% of the drivers, the psychoactive agent THC was found in only 4%. Of the drivers with only one substance in their system, only 1.1% had cannabis alone, either as the THC the psychoactive compound or had the inactive metabolite carboxy THC. The presence of the inactive metabolite and the absence of detectable THC infers less recent ingestion of cannabisÑassuming an efficient analysis.

The THC only drivers had a responsibility rate below that of the drug-free driversÑie. as with the study by Williams et al. (1985) they were considered to be less likely to have been a cause of the crash than the drug-free drivers.

The report also indicated the range of THC concentrations found in the blood. There were 109 cases of THC alone; of these, 22.9% contained what the authors called a 'trace' ie. 1 to 2 nanograms THC per millilitre of blood (ng/ml); 69.7% contained 'low' concentrations between 3 to 19 ng/ml; and 7.3% contained a 'high' concentration of equal to or greater than 20 ng/ml.

6. Drummer reported a study of 1 045 fatalities in New South Wales, Victoria and Western Australia and used the technique of responsibility analysis (culpability index).

As with other studies, the dominant drug was alcohol, being found overall in 36% of all driver fatalities, 33% of which were over the legal limit of 0.05g%. Cannabis was found in 11% of cases of which 56% (n= 63) also contained alcohol (mean BAC 0.16 g% ± 0.08g%). There was no significant difference in the BAC of the alcohol only drivers and those with alcohol plus cannabis.

Assessment of the culpability ratio by Drummer provided the same result as those of Williams et al. and Terhune et al; there was a trend to a decrease in relative risk when either THC or the metabolite carboxy THC was measured in blood or urine. The relative risk was 0.6 relative to drug-free drivers, although this was not significant statistically.

7. ALCOHOL AND CANNABIS IN EPIDEMIOLOGICAL STUDIES

The relative risk for drivers with alcohol plus cannabis was also greater than that for the control group, but this culpability ratio was no different from the alcohol only group. Also in this study (as indicated above), there was no significant difference in the BAC of the alcohol-only drivers and those with alcohol plus cannabis.

The same finding was reported by Terhune who also suggested that the high levels of alcohol are primarily responsible for the increased crash risk.

Therefore the effects of alcohol in road crashes are really profound. The studies reviewed here using the method of 'responsibility analysis' have confirmed the information already established by the case-control methods-that alcohol is the dominant drug associated with risky and dangerous driving and road crashes.

There have been suggestions throughout the studies reviewed here that the crash responsibility rates associated with the low BAC plus other drug, might be higher than in the low alcohol-only groups. The interaction of other drugs and alcohol (including cannabis) require further study using epidemiological techniques. One must remember the description by Perez-Reyes of the effect of the order of administration of alcohol and cannabis in these interaction studies.

8. SUMMARY (OF THE EVIDENCE PRESENTED ABOVE)

The most recent of the reports of studies of the effects of cannabis on actual driving performance included a summary of the published literature on marijuana and driving. They concluded this review with the following paragraph:

The foremost impression one gains from reviewing the literature is that no clear relationship has ever been demonstrated between marijuana smoking and either seriously impaired driving performance or the risk of accident involvement. The epidemiological evidence, as limited as it is, shows that the combination of THC and alcohol is over-represented in injured and dead drivers and more so in those who actually caused the accidents to occur. Yet there is little if any evidence to indicate that drivers who have used marijuana alone are any more likely to cause serious accidents than drug free drivers. To a large extent, the results from driving simulator and closed-course tests corroborate the epidemiological findings by indicating that THC in single inhaled doses up to 250 µg/kg has relatively minor effects on driving performance, certainly less than BACs in the range of 0.08Ê-Ê0.10g%.

Apart from the above, a very important finding in the reviewed studies is the difference in the drug users' awareness of the effect of the drugs alcohol and cannabis. Alcohol use is accompanied by increased confidence, an impairment of judgement to the extent that driving behaviour becomes more risky, with faster speeds and a greater willingness to take risks. Cannabis use on the other hand, is accompanied by compensatory driving behaviour, including a reduced willingness to take risks and slower driving speeds. Indeed the compensation was described by Robbe and O'Hanlon in the following manner:

The task of driving has been described as a 'self-paced' task. That is, drivers choose their own levels of task difficulty. There is a difference therefore between a driver's skills performance, as measured in individual laboratory tasks and driver behaviour. Driver performance, or skills performance is what a driver can do. Driver behaviour is what a driver actually does. Driving skills (or driver skills performance) differ very widely within a community. Some of us may be extremely cautious and others much less so. The correlation between driver skills and crash probability is not as great as many may imagine. For example, it is held by many that superior driver skills lead to reduced crashes and this led to the concept of 'advanced driver training'. Indeed, an editor of a road magazine claimed: 'I have for many years claimed that the licensed racer is far safer than ordinary chaps, on the grounds of practised skills, mental ability, cognisance of hazards in driving, keen interest in driving as well, and so on.'

In order to examine the possibility that unusually skilled drivers really did have different on-the-road driving records from the average driver, a comparison was made of the on-the-road driving records of a group of licensed racing drivers with those of other drivers matched for such characteristics as sex and age, etc. What they found was that in all measures of traffic violations including crashes, speeding violations, other moving violations as well as non-moving violations, the rates for the racing drivers exceed those of the comparison drivers, in most cases by a considerable margin.

In the light of the above, Terhune et al. asked the following questions:

A nagging question which qualifies conclusions from epidemiological studies of drugs in crashes is: If certain drugs are linked to elevated crash risks, how much of the elevation is due to characteristics of the people who use these drugs?

For example, Terhune in a literature review remarked that research revealed a striking similarity between the personal correlates of marijuana use and the correlates of crash involvement. Rebellious, deviant, youthful males were prominent among marijuana users and among those in crashes. Jessor et al. also addresses these issues.

A general conclusion made by Robbe and O'Hanlon when discussing the results of their study and of their review of the literature is worth citing here as a general conclusion to this review:

In summary, this program of research has shown that marijuana, when taken alone, produces a moderate degree of driving impairment which is related to the consumed THC dose.

The impairment manifests itself mainly in the ability to maintain a steady lateral position on the road, but its magnitude is not exceptional in comparison with changes produced by many medicinal drugs and alcohol.

Drivers under the influence of marijuana retain insight in their performance and will compensate where they can, for example, by slowing down or increasing effort. As a consequence THC's adverse effects on driving performance appear relatively small. Still we can easily imagine situations where the influence of marijuana smoking might have an exceedingly dangerous effect ie, emergency situations which put high demands on the driverÕs information processing capacity, prolonged monotonous driving, and after THC has been taken with other drugs especially alcohol.

We therefore agree with Moskowitz's conclusion that 'any situation in which safety both for self and others depends on alertness and capability of control of man-machine interaction precludes the use of marijuana'.

However, the magnitude of marijuana's relative to many other drugs' effects also justify Geringer's (1988) conclusion that 'marijuana impairment presents a real, but secondary, safety risk; and that alcohol is the leading drug-related risk factor'. Of the many psychotropic drugs, licit and illicit, that are available and used by people who subsequently drive, marijuana may well be among the least harmful.

Campaigns to discourage the use of marijuana by drivers are certainly warranted. But concentrating a campaign on marijuana alone may not be in proportion to the safety problem it causes.