CHAPTER 3 CANNABIS AND DRIVING

THC's effects on the ability of drivers to operate safely in traffic situations have traditionally been determined in two ways: from epidemiological surveys of users' involvement in traffic accidents and from empirical studies to measure the drug's influence on skills related to driving. The results obtained from both approaches are briefly reviewed for the purpose of providing a justification for our own research described in the following chapters.

3.1 Epidemiological Studies

The purpose of epidemiological studies on cannabis and traffic safety is to determine if driving under the influence of cannabis is detrimental to the safe operation of motor vehicles. Essentially, they determine if cannabis use is overrepresented in drivers who were involved in traffic accidents.

Simpson (1986) has reviewed recent epidemiological evidence regarding cannabis' role in traffic accident causality. His first concern was to determine the frequency of driving shortly after cannabis use to identify the proportion of the total driving population who may be considered 'at risk' of causing an accident for that reason. His information was derived from two sources: questionnaire surveys of adolescents (16-19 years), who were licensed to drive, and roadside surveys of recent usage among passing motorists.

He mentioned the reasonably consistent results of three questionnaire surveys conducted in the United States or Canada between 1979 and 1982. About one in six teenage drivers admitted driving while smoking or shortly afterward, and about 10% said they had done so between one and five times during the preceding month. Similar results were reported by Burns (1981): 25% of 16-21 years old American marijuana users admitted having driven after drug use. But in Valladolid, Spain, Alvarez et al. (1991)found that only 3.4% of drivers between 18 and 70 years of age admitted having driven a vehicle at least once after taking an illegal drug. Cannabis alone or in combination with another drug was used in 61.5% of these cases. The percentage of young (18-24 years) Spanish who admitted driving after illegal drug consumption was closer to those in North America. Overall this rate was 4.5% and it increased with the young Spanish drivers' educational and, presumably, socioeconomic status. The separate incidence for cannabis was not reported. These studies show that at least some, primarily young, users do drive during or shortly after cannabis smoking. Unfortunately it is not possible to generalize the results of the former studies to older drivers in North America, nor those of the latter study to other countries in Europe.

Only three surveys of recent cannabis use among drivers stopped at roadside check points have apparently been reported. According to Simpson, one yielded uninterpretable results due to technical and methodological problems. The others were widely separated in place and time. The first was completed in Canada in 1974 (Smith et al., 1975), the second in Italy in 1982 (Ferrara and Rozza, 1985). Moreover, the former relied upon the drivers' admission of use and the latter upon detection of cannabinoids in urine samples. Nonetheless, the indications of recent cannabis use given by the two sets of results were not grossly different; 4% by the first and 1.2% by the second.

Terhune (1982) tested 497 injured drivers for the presence of a wide range of drugs during treatment at the Rochester General Hospital in New York. THC in blood was detected in 9.5% of the drivers, but more than half of them also tested positively for alcohol. Chesher and Starmer (1983) found THC in 6.7% of 104 injured drivers in New South Wales, Australia, but again about half of them showed alcohol as well. Daldrup et al. (1987) examined 597 blood samples from injured drivers in the region around the German city Dusseldorf for the presence of alcohol. Blood samples having alcohol concentrations (BAG) below 0.13 g% were additionally analyzed for the presence of cannabinoids; twenty-five of the 220 blood samples (11.4%) were positive. Finally, McLean etal. (1987) found T1 IC in four out of 37 (10.8%) drivers who had survived a road accident in Tasmania. Three of them also had a BAG of .05 g%, or more. Although the latter sample was very small, the percentage of injured drivers with THC in their blood corresponds remarkably well with the other studies.

Together they agreed that THC is present in the blood of about 6-12% of drivers injured in road accidents.

These data contrast substantially with those provided by Soderstrom et al. (1988) from 1023 patients admitted as the result of vehicular and nonvehicular accidents at the Shock Trauma Center in Baltimore, Maryland. THC was found in blood in 34.7% of the patients, alcohol in 33.5%. Among automobile drivers, the numbers were 31.7% and 34.6%, respectively. Both drugs were present in about 50% of the injured drivers. It is not clear why these results differ from those of previous studies. The most plausible explanation is that residents of the Baltimore area tend, in general, to use THC more often than those in the other regions surveyed. Investigators at two other trauma centers (Philadelphia and Chicago) found similar incidence rates. However, they measured all cannabinoids instead of only THC in either blood (Lindenbaum et al., 1989) or urine (Sloan et al., 1989). The incidence of THC in trauma patients at these locations must have been much lower than in Baltimore.

Cimbura et al. (1980, 1982) analyzed the blood of 401 fatally injured drivers in the province of Ontario, Canada, during 1978-1979. THC was detected in 3.7% but alcohol was also found in most of the THC positive cases, i.e. 87%. In their subsequent study in the same area during 1982-1984, the investigators reported that THC was detected in blood among 10.9% of 1169 fatally injured drivers (Donelson et al., 1985; Cimbura et al., 1990). Again, more than 80% of THC positive blood samples also contained alcohol. Mason and McBay (1984) found THC in 7.8% of 600 drivers killed during single-vehicle crashes occurring in North Carolina during 1978-1981. Preliminary results from this study have been published by Owens (1981) who reported a 5.9% incidence rate in 169 driver fatalities during 1978-1979. As in Ontario, the vast majority of THC positive drivers had also used alcohol. Two studies employing very small samples have also been reported. In the first, 11 out of 69 drivers (15.9%) who were killed in traffic accidents in Bexar County, Texas, during 1985 had THC in their blood (Garriott etal., 1986). In the second, four out of 42 road-accident fatalities (9.5%) in Tasmania during 1983-1984 were positive for THC (McLean et al., 1987).

Disparate results were obtained by Williams et al. (1985): THC was found in 36.8% of 440 driving fatalities in California during 1982 and 1983 (80% in combination with alcohol). The reason for this disparity might have been due to the greater prevalence of marijuana use in California as compared to other American States. But it was more probably related to these investigators' selection criteria: their sample only included male drivers younger than 35 years of age. Simpson estimated that if female and older male fatalities had been included in this survey, the overall percentage showing THC would have been about 20%. Moreover, 38% of the THC positive drivers showed only trace amounts of the drug, i.e. less than 1.0 ng/ml. If these cases are viewed as false positives, according to the conventional epidemiological criterion, the number of THC positive drivers would be 22.8% in the original sample, and only 12.4% after Simpson's correction. More recent Californian data were provided by Budd et al. (1989). They assayed various drugs in the blood of drivers killed in Los Angeles County, both in a preliminary study during 1985-1986 and in a follow-up study over the next two years. Both times they found cannabinoids (Thic and/or THCCOOH) in about 19% of dead drivers (total N=313). One can only guess what the incidence rate for THC alone might have been on the basis of cases showing both the parent compound and its metabolite. Probably it was somewhere between 10 and 13%, which corresponds well with the previous studies.

The most recent data regarding the incidence of drugs in fatally injured drivers in the United States are available from a nationwide study conducted in 1990 and 1991. Blood specimens were collected from 1,882 dead drivers from 13 sampling sites, encompassing three entire states (Massachusetts, North Carolina, and Wisconsin) and selected counties in California, Nevada, Texas, and Virginia (Terhune et al., 1992). THC was found in only 4.2% of the drivers, and about 70% of the time in combination with alcohol. THC alone was found in only 19 fatalities or 1% of the total sample. These relatively low percentages may indicate a declining trend in the incidence rate of THC in fatally injured drivers in the United States, explainable by the declining prevalence of marijuana use throughout the 1980s.

With some exceptions, epidemiological studies indicate the presence of THC in roughly 4-12% of drivers injured or killed in traffic accidents. If one accepts that the population at risk is less than 4%, the injury/fatality rate must be taken to indicate that the drug's users are overrepresented among accident victims. However, there are obvious reasons to doubt whether valid estimates of the population at risk in urban North America can be derived from data that are more than 14 years old or were obtained at two locations in Northern Italy. Even if the population at risk is as small as estimated, the THC incidence among injured or dead drivers is not conclusive evidence for establishing its role as a causal factor. Alcohol was also present in the majority of survey victims showing any plasma concentrations of TI-IC. It is highly likely that these drugs in combination possess a greater risk potential than either alone. The independent contribution of THC to traffic accident causality, particularly in concentrations which are likely to be found in most users, is still dubious.

One major problem of these surveys is the common lack of sound control groups as have been used in studies of alcohol involvement in accidents (e.g. Borkenstein et al., 1974). BACs measured in fatal accident victims were compared to breathalyzer BAG estimates from randomly selected motorists passing the same accident sites at the same times and days of the week. That comparison provided a ratio of fatally injured drivers to normal controls showing a given BAG and how it varies across the range found in all drivers. This approach was feasible because nearly all (>97%) passing motorists were willing to yield breath samples; and valid, because alcohol concentration in the blood is about the same as in the brain. Many fewer motorists are prepared to yield blood samples for a variety of reasons, including that of revealing illicit drug use. Moreover, TUG concentrations in blood and brain are almost always different. As Moskowitz (1985) observed, these fundamental feasibility and validity premises which were met in alcohol surveys can not be in surveys for measuring the relationship between THC and accidents.

The lack of separate control groups can be circumvented by the use of a culpability index. The index is the ratio of the percentage of drivers with detectable drug levels and deemed culpable to that of drug-free drivers from the same sample who were likewise responsible for their accident. Warren et al. (1981) reanalyzed the data from Cimbura et aL (1980) showing that 52% of the drug-free fatally injured drivers were deemed culpable compared to 90% of those with evidence of recent cannabis use. This yielded a culpability index of 1.7, the same as for alcohol. Results of two other studies (Terhune, 1982; Donelson et al., 1985) were consistent with these findings. On the other hand, Williams et al. (1985) and Terhune et al. (1992) failed to find a significantly elevated culpability index for THC users. In the latter study 58% of the drivers in whom only THC was detected were deemed culpable, whereas 68% of the drug-free drivers were. The difference in favor of THC users was not significant. In contrast, dead drivers showing only alcohol were culpable in 92% of all cases. Those showing both drugs were slightly more often responsible for causing the accident (95%). It should be noted, however, that the usually low frequencies of culpable drivers showing THC alone prohibit any definite conclusion.
In summary, epidemiological research has shown that some people do drive after cannabis use and that drivers involved in accidents often show the drug's presence. However, alcohol has been a severe confounding factor in all surveys of injured or killed drivers. For this and other reasons given above, the independent contribution of THC to their accidents remains exceedingly obscure.

3.2 Laboratory Studies of Skills Related to Driving

Much research has gone on to determine the drug's effects on isolated psychological functions and skills related to driving since cannabis' revival in the 1960s. Owing largely to the fact that its methodology already existed and could be economically applied, this research preceded epidemiological surveys and actual driving studies. Yet one of the major drawbacks of laboratory research is that no performance model comprehensively defines the actual driving task. As a consequence, different researchers have employed different part-task models to design a plethora of tests. This was clearly illustrated by Joscelyn et al. (1980) who inventoried the methods applied in the drugs and driving research and came up with long lists of tests, tasks, and response variables. They noted: "many tests routinely employed have limited validity or no demonstrable relation to real-world driving. Methods measuring the 'same' behaviors often differ, raising questions about the comparability of experimental findings (p. 34)."

This does not mean that laboratory studies are useless for assessing the drugs and driving problem. On the contrary, they provide the earliest evidence concerning a particular drug's hazard potential for driving. In the context of well designed experiments, drugs that produce large performance impairments in many different tests can be considered potentially hazardous whereas drugs that fail to produce any impairment can be considered safe. Yet results obtained in the laboratory should not form the sole basis for the final judgment of a drug's potential to impair driving performance and jeopardize traffic safety. They should be confirmed, if possible, by epidemiological surveys and actual driving studies.

Two recent reviews of the literature pertaining to the effects of marijuana on skills related to driving contain most of the information pertinent to this dissertation (Moskowitz, 1985; Chesher, 1986). No attempt will be made to replicate their comprehensive efforts. Instead their major conclusions will be mentioned along with certain contradictions and omissions in the available data base. Where appropriate, the results of more recent studies are also mentioned.

Visual functions

Visual acquisition of information is the first step in the process of controlling a motor vehicle. For that reason it seems appropriate to first answer the question of whether either visual functions or oculomotor coordination is impaired by THC. In general, studies reviewed by Moskowitz have shown little or no effect for 0-6 hours of ingested THC doses up to about 300 Ag/kg, or inhaled doses up to about 15 mg (210 Ag/kg in a 70 kg person) on static or dynamic visual acuity, binocular fusion, lateral phoria, glare recovery, color vision or saccadic and ballistic eye movements involved in visual tracking. Stapleton et al. (1986) reviewed the effects of psychotropic drugs on eye movements and came to a similar conclusion: unlike alcohol, marijuana has relatively little effect on oculomotor functions.

A curious finding was reported by Schwin et al. (1974) who measured critical flicker fusion frequency (cFF) in subjects who inhaled 15 mg THC. This treatment produced a rise in CFF, the opposite of what one would expect for any drug with depressant CNS properties. Yet Block et al. (1992) found a CH decrease after subjects smoked a marijuana cigarette containing 19 mg TI-IC.

Fine and Gross Motor Control

Several studies have shown that even very low inhaled THC doses (e.g. 10 jag/kg) increase hand or whole body instability in tests such as maintaining the position of a stylus in a hole without touching the sides, or balancing on a platform supported by a central fulcrum. These impairments are dose-related so it seems clear that one of the drug's effects is to impede, or otherwise interfere with the neuromuscular-proprioceptive servocontrol loop. Unless the marijuana user is also a watchmaker or tightrope walker, it seems unlikely that he would suffer greatly from this effect in any practical task, including driving. However, the sensitivity of motor coordination measures and their systematic change with THC dose, led Moskowitz to suggest (p. 328) that one or more might prove useful for identifying states of marijuana intoxication that are incompatible with safe driving.

Reaction Time

Simple reaction time is only slightly affected by even relatively high THC doses. In various complex reaction time test paradigms, where the association between the given stimulus and correct response varies between trials and requires an intervening decision on the subject's part, THC's effects have been sometimes large, sometimes small but significant, and occasionally absent. Generally however, the variability of complex reaction times increased under the influence of inhaled THC doses in the range, 100-250 rig/kg.
Moskowitz was somewhat at a loss to explain the contradictory reports concerning THC's effects on complex reaction time. These seemed to depend upon the particular test paradigms employed by various groups of investigators. Their diversity without any clear indication of what information processing stages each one measures, leaves one in doubt concerning the meaning of these results. Moskowitz could conclude that moderate THC doses impair neither the initial perception of an unambiguous and anticipated signal, nor the final motor responding speed. However, the elevated variability in complex reaction times suggested to him impairment in the subjects' mechanism of attention to the task.

Some idea of the complexity of measuring THC's influence on complex reaction time, as well as the suggestion of a fundamental drug effect, can be gathered from a study of Chesher et al. (1986). These investigators employed a mental rotation task, i.e. the 'little men' test. A figure of a man holding an object in one extended hand appeared on a computer terminal. On successive trials the figure could be upright or inverted, facing forward or backward and its hand holding the object could be the right or the left. The subject's task upon seeing each new figure was to make one of two button pressing responses to indicate which hand held the object. The frequencies of correct and incorrect responses were measured along with the associated reaction times. There were 16 experimental conditions defined by separate combinations of ingested alcohol (0.00, 0.25, 0.50 and 0.75 g/kg) and inhaled THc (0.00, 2.5, 5.0 and 10.0 mg). A different group of 20 subjects, who had previously used both drugs, participated in each condition. Chesher's (1986, p. 114) description of the difference between the drug's effects is illuminating.

"The results for the mental rotation test indicated an interesting qualitative difference between the two drugs. The nature of the alcohol or marijuana effect differed according to whether the subject's response to the mental rotation task was correct or incorrect. The drug effect for those responses which were correct was similar to those for the other reaction time measures - an increase in reaction time for alcohol but not for marijuana. Indeed, the lowest dose of marijuana indicated a. trend towards antagonism of the effect of alcohol. However, the reaction times, and their standard deviations, for the items for which the subjects' responses were incorrect, indicated a qualitatively different drug effect. The mean reaction time for these responses showed a highly significant linear increase across the marijuana doses but not across the alcohol doses. Associated with this, alcohol produced a significant, linear effect on increasing the number of errors; marijuana did not. Furthermore, in this measure also, there was a trend for an antagonism of the alcohol effect by the lowest dose of marijuana. One possible interpretation of these different drug effects is that a 'speed-accuracy trade-off' occurred in the alcohol condition and a more cautious approach was applied in the marijuana condition. One might assume that the linear increase in reaction time for these responses under the influence of marijuana could be due to a greater time in pondering the problem. Possibly when faced with a similar problem under the influence of alcohol, the subject is more likely to take a risk and make a hasty guess at the answer. This behaviour resulted in a significant linear increase in the error rate."

This demonstrates what may be a fundamental and specific THC effect: increased caution that results in slower responses to an ambiguous situation.

Tracking

Moskowitz's review leaves no doubt that THC impairs every type of laboratory tracking performance. This was shown, without exception, in eight separate studies. Degraded tracking performance occurred shortly after inhaled TI-IC doses as low as 3 mg (43 Ag/kg, for 70 kg persons). After a 200 jig/kg dose, tracking performance impairment persisted continuously for four hours and intermittently for the next four hours. Barnett etal. (1985) found marijuana-induced impairment in a critical instability tracking test for seven hours after smoking, but neither the magnitude nor duration of impairment was related to the administered THC dose (i.e. 100, 200, or 250 Ag/kg).

Surprisingly, in view of the strength and consistency of the effects of TI-IC on laboratory tracking performance, the same effects have been difficult to replicate in tracking tests more resembling car driving (below).

Attention

Marijuana smoking has two demonstrably adverse effects on attention. One is seen in dual-task situations and the other in vigilance tests. Moskowitz himself was the first to clearly demonstrate these effects. His earliest study (Moskowitz et al., 1972) was designed to measure the effects of 50, 100 and 200 fig TI-IC/kg on subjects' detection responses to peripheral signals while they were counting light flashes as these appeared on a central display. All three doses caused the subjects to detect fewer peripheral signals in a dose-related manner, and the degree of impairment increased with the cadence of counting. With various embellishments, the adverse effects of similar THC doses upon divided attention have been seen again in several studies published since 1985. Barnett et al. (1985) found dose-related impairment in both a central compensatory tracking task and a peripheral visual search-and-recognition task following THC doses of 100, 200, and 250 Ag/kg. Performance in a divided attention task involving pursuit tracking in the upper half of the video monitor, and, searching for a target number, in the lower half, was not affected by smoking a marijuana cigarette containing either 1.3 or 2.7% THC (Heishman et aL, 1989). Yet smoking 25 puffs of a marijuana cigarette containing 3.55% THC produced a significant prolongation of response latencies to target numbers in the same test (Azorlosa et al., 1992). In another study, subjects had to react to critical changes of two-digit numbers displayed in the central field of vision and one-digit numbers displayed in the periphery by pressing one of four response buttons or a foot pedal. Smoking a marijuana cigarette containing 2.4% THC reduced the subjects' overall accuracy and prolonged their reaction times (Perez-Reyes etal., 1988). Marks and MacAvoy (1989) employed a divided attention task in which subjects had to react to a break in the regularity of a flashing central light as well as to each flash of one of ten peripheral lights that were spaced at 15° intervals along a horizontal perimeter. Decrements in both central and peripheral signal detections were found after inhaling 5.2 but not 2.6 mg THC.

Sharma and Moskowitz (1973, 1974) further demonstrated the effects of THC on sustained attention as it is traditionally measured in prolonged signal detection or vigilance tasks. Their paradigm was adopted from the famous `Mackworth Clock Test'. Subjects viewed a number of lights arranged in a 30.5 cm (12 in) diameter circle. These were illuminated singly in a regular sequence, except occasionally when the progression skipped a particular light. The discontinuity in the sequence was defined as a signal to which the subject responded. Normally subjects show a 'vigilance decrement' or decline in the percentages of detections as a function of time on watch, as did subjects in both experiments after smoking a placebo cigarette. In the first experiment, the decrement accelerated after subjects had inhaled 50, 100 and 200 Ag THc/kg. Changes from placebo were significant after the two higher doses. One condition in the second experiment replicated the results for the highest dose condition in the first. Another demonstrated that the same decrement occurred when the subjects were required to respond, following every flash, to indicate their perceptions of signals and non-signals. The stated purpose of the new condition was to show that the subjects were in fact losing their ability to discriminate between stimuli and not simply ceasing to respond because of wandering attention.

Conclusion

Doses of THC that have been administered in laboratory studies are less than those commonly found in marijuana cigarettes and are apparently sought by drug users to achieve the desired 'high'. Indeed both Moskowitz and Chesher admitted that their highest laboratory doses may be considerably less than those contained in cigarettes manufactured for consumption by normal users. Thus, it should be clear from their brief review that doses normally taken by users possess the potential for impairing at least their neuromuscular coordination, tracking skills and attentional functions. Yet there is at least some suggestion from laboratory studies that the subjects attempted to compensate for these deficiencies, in so far as they could, by employing behavioral strategies that maintained the integrity of some performance parameters at a cost to others.

3.3 Driving Simulator Studies

Early studies by Crancer et al. (1969), Rafaelsen et al. (1973a), Ellingstad et al. (1973) and Moskowitz et al. (1976) utilized the filmed ride approach where subjects had little or no control over the presented imagery. Dott (1972) used a different approach for measuring subjects' decisions to pass a preceding car, or not, in the presence of an opposing vehicle portrayed as models on a continuous belt. Doses of inhaled THC varied from about 3 to 22.5 mg (43-321 p.g/kg for 70 kg, or 154 lb, persons). Smiley (1986) reviewed these early studies to conclude that THC had (1) not affected vehicle control, (2) increased decision latency before starting, stopping or overtaking, (3) reduced the willingness to accept a risk during passing maneuvers, and (4) impaired speedometer monitoring. Except in the case of one individual who, after inhaling 12 mg THC, repeatedly drove through stop lights during a filmed ride, no particular sign of dangerous driving behavior was observed.

Smiley et al. (1981) conducted the first study using an interactive simulator with accurate visual imagery, though not moving base dynamics. The simulated tasks contained in a 45—minute scenario included curve following, reacting to wind gusts, car following, route selection from signs, avoiding an obstacle which appeared in front of the simulated vehicle and passing. A visual choice reaction time was also superimposed on driving. Three groups of marijuana users smoked cigarettes containing 0, 100 and 200 Ag/kg THC on two occasions per dose, once with and once without alcohol. The quantity of alcohol consumed varied between groups to reach intended blood concentrations of 0.00, 0.05 and 0.08 g%, respectively. To ensure high motivation, good driving was rewarded and blatant errors, such as crashes, were penalized financially. The test began 15 minutes after the cessation of smoking. Both THC doses increased lateral position variability and the highest dose increased speed variability during curve following. Both increased headway variability, and the highest, lateral position variability during car following. Both doses caused the subjects to miss more signs indicating the need to follow another route. The high dose caused the subjects to hit the roadway obstacle more often than placebo, and also, to react slower to the subsidiary task. Yet both THC doses caused the subjects to drive in a more conservative manner. They maintained a longer headway while car following, refused mon, opportunities to pass, and when they did, began this maneuver at a greater distance from the approaching vehicle. Alcohol's effects in this study were generally less than THC's. Chesher (1986) was puzzled by this, calling the alcohol effect "surprisingly small" and its interactive effect with THC, "unclear". Certainly it is so that BACs of 0.08 g% and below have been enough to substantially degrade drivers' control of vehicular lateral position in real driving tests (Louwerens et al., 1987; Ramaekers et aL, 1992a).
Stein et al. (1983) conducted two studies of alcohol and marijuana effects using a driving simulator and a 15-minute test scenario that were very similar to those employed by Smiley et al. (1981). The former administered the two drugs in complete crossover designs. THC doses of 0, 50 and 100 Ag/kg THC were combined with BACs of 0.00 and 0.10 g% in the first study. The same BACs were combined with 100 and 200 tig/kg in the second. This time alcohol had the expected adverse effect on practically every performance parameter, THC had little effect in the first study and little in the second in spite of the higher dose. The latter did cause the subjects to operate at generally lower speeds, however. The combination of drugs produced widely different individual reactions. After the highest THC dose, the combination produced more adverse reactions than alcohol alone.

3.4 Actual Driving Studies

A number of studies on marijuana's effects upon actual car driving have been reported since 1974. All studies but one were carried out on courses closed to other traffic. Klonoff (1974) conducted the exceptional study wherein 64 subjects drove on a closed course and 38 also participated in a city driving test. In his first study subjects were assigned to one of three groups that were treated with (1) placebo, (2) 4.9 mg THC, and (3) 8.4 mg THC. They undertook eight tests: a slalom, two tunnel tests, a funnel test, a backing up, turning in a corner, a risk judgment test and an emergency braking test. Except for the latter two, the performance measure was number of cones hit. Subjects performed 20 trials in four blocks of five. Treatments were administered between the third and fourth block and each subject's performance was related to his/her performance predicted by means of regression analysis over the first three blocks. Performance after placebo was as predicted, but after marijuana, significantly worse, though not much. The low dose impaired performance in two tests (tunnel and corner) and the high dose in five (slalom, both tunnel tests, funnel and risk judgment).

Subjects in the city driving test were divided among four groups who were treated with placebo and marijuana, on separate occasions a week apart. The respective groups' treatments were (1) placebo followed by 4.9 mg THC, (2) the same in reverse order, (3) placebo followed by 8.4 mg THC, and (4) the same in reverse order. After smoking a placebo or marijuana cigarette, the subjects drove for 45 minutes over a 16.8 mi (27.0 km) route on city streets while aspects of their performance were rated by a professional examiner using an abbreviated version of the British Columbia Department of Motor Vehicles' standard driver's licensing test. All subjects were allowed to complete the test which indicates that their performance never became dangerously unsafe under the drug's influence. Nonetheless, the examiner rated the subjects' performance as significantly worse on scales of judgement and concentration following the highest but not the lowest dose. The majority showed some impairment, but 32% after the low dose and 16% after the high dose performed significantly better than they had following placebo suggesting qualitative differences between the drug's effects in different subjects.

Hansteen et al. (1976) tested sixteen subjects in four conditions, (1) placebo alcohol +placebo marijuana, (2) placebo alcohol + marijuana (THc dose of 21 rig/kg), (3) placebo alcohol + marijuana (THc dose of 88 fig/kg), and (4) alcohol (BAc 0.07 g%) + placebo marijuana. Subjects were instructed to drive through a 1.1 mi (0.7 km) course delineated by traffic cones as quickly as possible but without exceeding 30 mph (19 km/h). Performance was measured shortly after smoking and three hours later. Number of cones hit, 'rough handling' (superfluous and/or awkward movements as observed by an accompanying investigator), and driving time were scored. More cones were hit and more time was taken to complete each lap after consuming the higher THC dose, but no increase in rough handling was observed. Alcohol, on the other hand, adversely affected both performance measures and diminished the time taken to complete each lap. The authors concluded that the drug effects on performance were not dramatic since no major differences were found between conditions with respect to observer ratings.

Casswell (1979) was the first who included a subsidiary task to simulate the demands for monitoring the environment. Thirteen males were tested in three treatment sessions receiving alcohol and marijuana treatments twice in each session and drove for 35 minutes after each treatment. Treatments included (la) alcohol (0.10 g% BAc) + placebo marijuana, (lb) placebo alcohol + marijuana (6.25 mg THc), (2a) double placebo, (2b) placebo alcohol + marijuana (6.25 mg THc), (3a) alcohol (0.05 g% BAc)+ marijuana (3.12 mg THc), and (3b) alcohol (0.05 g% BAc) + marijuana (3.12 mg THc). Subjects' tasks included overtaking, driving on straight sections, through a hairpin bend, and through narrow gaps, while responding to road signals, traffic signals, and auditory signals in the subsidiary task. Alcohol alone and in combination with marijuana produced more coarse steering corrections, higher speed and increased lateral position variability. Marijuana alone was associated with lower driving speed and prolonged reaction times in the subsidiary task. Reaction times were also prolonged by the combination of marijuana and alcohol. The authors said that drivers under the influence of marijuana appeared to compensate for what they felt were the adverse effects of the drug by maintaining control effort, and decreasing speed to reduce the required rate of information processing. Alcohol, in contrast, appeared to produce more risky behavior.

Attwood et al. (1981) also employed normal driving tasks on a closed course. Eight males participated in a within-subjects design, receiving (1) double placebo, (2) alcohol (0.08 g% BAc)+ placebo marijuana, (3) placebo alcohol + marijuana (150 Ag/kg THc), and (4) alcohol (0.04 g% BAc) + marijuana (75 Ag/kg THc). The driving tasks were performed on an airfield runway and included: maintenance of a constant lateral position and velocity, maintenance of a constant headway while following a lead car that varied in speed, bringing the car to a smooth stop at a traffic signal, and deciding whether or not to overtake a preceding vehicle in the presence of an approaching car. The latter maneuver was, however, not actually undertaken. Various measures, as speed, lateral position, acceleration and headway, were taken but the number of significant comparisons were no more than expected by chance. All measures were then subjected to a discriminant analysis that separated overall treatment effects. Overall driving performances after all drug treatments were significantly worse than following placebo when tested in this multivariate analysis. Smiley (1986) suspected that the lack of univariate effects was attributable to the low number of subjects and the lack of a subsidiary task.

Peck et al. (1986) assigned 84 subjects in equal proportions to four treatment conditions: (1) double placebo, (2) alcohol (0.08 g% BAc) + marijuana placebo, (3) marijuana (19 mg THc) + alcohol placebo, and (4) both drugs combined. If these subjects could have inhaled all of the drug available in the cigarette, one weighing 70 kg (154 lb; population average) would have received a dose of about 270 Ag/kg. Because of the remaining butt, the actual THC dose probably never exceeded 250 Ag/kg. The subjects were tested four times in complete replications of a driving test battery beginning shortly after drug administration and continuing at hourly intervals thereafter. Ratings of the subjects' driving proficiency were obtained from driving licence examiners who rode with the subjects or observed them from static positions at points along the course; and, by California Highway Patrol officers who followed the subjects' vehicle in a police car. A computerized system within the subjects' vehicle recorded their use of controls, speed and lateral position relative to course delineation. A risk acceptance test was included to measure the subjects' willingness and ability to drive through gaps wider and narrower than the vehicle. Other tasks involved stopping in response to signals, making a forced lane change and driving through pylons in a chicane. Finally, a standard police field sobriety examination and two standard laboratory tests (tracking and time estimation) were administered to the subjects outside of the vehicle. Several hundred measures of performance were obtained. No dramatic performance failures were reported as an effect of either drug or their combination. In general, the number of significant drug effects on particular measures were about what one might expect given the total number of statistical tests.

The investigators resorted, like Attwood et al., to multivariate statistical analysis of their data. Twelve performance measures were combined in discriminant analysis, which significantly separated the effects of each drug or their combination from placebo's. The THC effect was significant over all four replications of the tests, being greatest in the first trial. Alcohol's effect was greatest in the second trial and slightly greater than THC in every one. The combination of THC and alcohol produced significantly more impairment then did either drug alone in the first and third trials. Field sobriety checks by the police and ratings of the subjects' driving proficiency by experts failed to show any effect of THC, though these did reveal the effects of that drug in combination with alcohol. Practically the only indication of a serious effect of THC was provided by the officers following the subject's vehicle in a police car. They reported that they would have stopped the subject for suspicion of being intoxicated on 32% of all THC trials (alcohol 50%, both drugs 60%). But they also said they would have stopped 15% of the placebo treated subjects. This either indicates that the subjects were exceptionally poor drivers, or were made to appear so under conditions of the test, or that the officers were responding to cues that they ordinarily would have ignored in real driving conditions.

Smiley et al. (1987) tested the effects of marijuana (0, 100 and 200 tig/kg Tfic) in combination with alcohol (0.00 and 0.05 g% BAG) and alcohol alone (0.08 g% BAG) on driving in a closed-course study. Treatments were administered to groups of nine males over a three hour period in a party-like atmosphere in the evening. Subjects drove shortly after smoking as well as on the following morning. Driving tasks included maintenance of a constant lateral position at 80 km/h (50 mph), curve following, car following, route navigation, obstacle avoidance, and emergency decision making. Additionally, subjects had to perform a subsidiary task requiring visual monitoring. The high THC dose resulted in increased headway and headway variability. Alcohol alone at the 0.05 g% BAC level produced increased speed. Number of subsidiary task detections decreased at 0.05 g% BAC but increased at 0.08 g% BAG. Smiley's (1986, p. 133) conclusion from her own and previous studies was as follows:

" 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. Where they can compensate, they do, for example, by not overtaking, by slowing down and by focusing 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."

3.5 General Conclusion

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 iig/kg has relatively minor effects on driving performance, certainly less than BAGS in the range 0.08-0.10 g%. In contrast to this, laboratory studies have repeatedly shown performance impairment occurring after inhaled doses as low as about 40 Ag/kg. These became large and persistent after 100-200 tig/kg doses. Tracking, divided attention and vigilance test performance were particularly vulnerable to THC's effects.

It is exceedingly difficult to explain the disparity in results obtained by laboratory tests and in driving simulations. Rather than try, it seems better for the moment to assume that both sets of results are valid for the particular circumstances under which they were obtained (this issue is discussed in the General Discussion, Section 9.4.4). It demonstrates, however, that performance decrements obtained under the artificial and non-life threatening conditions in the laboratory do not automatically predict similar decrements in driving simulations that are closer to real-world driving.

This all leaves the effects of THC on actual driving performance an open question. Authors' conclusions have gone both ways. In 1977, for example, one author concluded from the literature that "cannabis is, as should have been anticipated, a hazardous drug for the road user" (Milner, 1977, p. 2), whereas another concluded that "there is no evidence that marijuana is a significant public safety problem" (McBay, 1977, p. 97). The situation was unchanged in 1985 as illustrated by the following quotations:

"It should be clear from the above review that there is more than sufficient experimental evidence to conclude that marihuana seriously impairs psychomotor performance required for driving . . . Any situation in which safety both for self and others depends upon alertness and capability of control of man-machine interaction precludes the use of marihuana." (Moskowitz, 1985, p. 342)

"Several investigators have reported that marijuana reduces risk-taking propensity and driving speed. Because of these compensating tendencies, it is presently not possible to assess the net impact of marijuana as a causal agent in traffic accidents. Although some increased risk appears likely, the magnitude of the risk remains obscure . . . Many of the laboratory marijuana studies which have shown the greatest psychomotor impairment have utilized tasks that are only abstractly related to driving . . . it does not necessarily follow that performance on a highly novel and complex laboratory task designed to magnify performance decrements is correlated with actual 'real world' performance in a vehicle. The fact that attempts to measure response to simulated accident situations have not consistently detected a marijuana-induced decrement, even at high dose levels, underscores the need for more research." (Peck et al., 1986, pp. 152-153)

Because epidemiological research seems too handicapped by practical constraints to demonstrate the relationship between THC and traffic accidents, and since previously applied experimental approaches provide different indications concerning the strength of that relationship, it seemed timely to investigate cannabis' effects on driving performance in the real traffic environment.

As mentioned above, only one study has been conducted in actual traffic before this program started (Klonoff, 1974) . But Moskowitz (1985) and Smiley (1986) rightly criticized the method used by Klonoff for measuring driving performance on the grounds that the examiners' reliability was never determined and that the scoring instrument had never been shown to provide measures related to driving safety. Smiley questioned, for example, whether ratings of posture and irritability are relevant for good driving performance. Finally, Klonoff administered relatively low THC doses to his subjects. The effects of high doses of THC on driving in real traffic still needs to be determined.

The studies reported in this dissertation were conducted to escape these limitations. First, the effects of low, moderate and high THC doses on highway driving were determined, both in the absence and presence of other traffic. Second, Klonoff's city driving study was replicated, with some modifications with regards to the employed procedures and with the addition of another group of subjects who undertook the same driving test but then under the influence of a low dose of alcohol. The next chapter summarizes the designs and general methods for the studies that were so conducted. Thereafter, each study is reported separately in chapters that follow.