The major focus of this chapter is drug effects on overall crash prevalence and risk as determined from epidemiologic studies. Major categories of research of interest here are:

These categories are addressed in four sections of this chapter, as follows:

Within each section, there are two sub-sections, the first dealing with research conducted in North America (the United States and Canada), and the second with research conducted elsewhere (foreign studies).


North American Studies.

Several studies have examined the presence of drugs in crash-involved drivers in the United States. The largest of those concerned with fatally-injured drivers was the NHTSA-sponsored study of 1,882 operators of passenger cars, motorcycles, and trucks who were involved in fatal crashes in three states and selected counties in four other states during 14 months in 1990-1991 (Terhune, Ippolito, Hendricks et al., 1992). Researchers in that study collected blood specimens from a sample of all drivers who had died within four hours after the crash. Samples were analyzed qualitatively for alcohol and 43 other drugs, many which were no longer in use at the time of the study. Alcohol was the most prevalent drug, and was found in 51.5% of the crashes; other drugs were found in 17.8% of the crashes. The study found cannabinoids present in 6.7% of the drivers, cocaine in 5.3%, benzodiazepine in 2.9% and amphetamines in 1.9%. All other drugs combined were found in less than 5.0% of the drivers. Drugs were found in conjunction with alcohol in 11.4% of the cases.

Studies of fatally-injured drivers in more limited geographical areas have reported similar findings. For example, Williams, Peat, Couch, et al. (1985) found one or more drugs in 81% of 440 male drivers, aged 15-34, killed in motor vehicle crashes in California, and two or more drugs were detected in 43%. Marijuana was detected in 37% of the drivers, and cocaine in 11%. Each of 24 other drugs was detected in fewer than 5%. Except for alcohol, drugs were infrequently found alone; typically, they were found in combination with high blood alcohol concentrations.

In another California study, Budd, Muto, and Wong (1989) conducted a study in which the blood and urine of 594 fatally injured drivers from Los Angeles County were tested for drug and alcohol levels over two different time periods from 1985 to 1988. The first part of this study (November 1985 - April 1986) found that 12% of 102 drivers tested were positive for "drugs of abuse." In a follow-up study covering a longer time period (May 1987 - May 1988), 15% of 492 drivers were positive for such drugs, 8% for marijuana and 7% for cocaine. Root (1989) found "drugs of abuse" (unspecified) in 13% of 796 fatally injured California drivers tested during the 1985-1987 period. Caplan, Levine, and Goldberger (1989) reported marijuana present in 7% of the 269 fatally injured drivers tested in Maryland over an unspecified 10-month period, and cocaine in 8%. Caplan and associates also found none of the drivers positive for amphetamines, and 3% positive for phencyclidine (PCP).

Rivara, Mueller, Fligner, et al. (1989) found that about 10% of 160 fatally injured occupants in King County, Washington (Seattle) in 1986 were positive for THC, and that 2% were positive for cocaine. The paper did not provide any breakdowns by type of occupant (driver or passenger). In a study in nearby British Columbia, Jeffrey, Leslie, and Mercer (1995) reported the results of chemical analyses of the blood of 222 fatally injured drivers in British Columbia, Canada in 1991. This paper also reported the level of the drug for each drug-positive subject. Percentages of cases and mean level for each drug were: THC / THCCOOH (13%, 14.7 ng/ml); cocaine (4%, 131 ng/ml for six cases positive for cocaine + metabolite); and benzodiazepines (5%, 405 ng/ml for 10 subjects with diazepine, 1,480 NG/ml for one subject with nordiazepine). Six cases (3%) contained other drugs.

Fortenberry, Brown, and Shevlin (1986) reported the results of another study of drug presence in fatally injured occupants (this time in Alabama), indicating that almost 17% tested positively for marijuana. Sixty-four percent of the marijuana-positive victims were drivers. More than 5% of fatalities had some level of drugs (either illicit or prescription) in their blood stream. The drugs that were found and their percent of all drivers were diazepam (2.1%), methaqualone (2.2%), phenobarbital (1.0%), and propoxyphene (0.3%). (Methaqualone is a depressant no longer legal in the U. S., and propoxyphene is a narcotic analgesic.) A statistically significant association (p=0.05) was found between diazepam and low BAC.

Marzuk, Tardiff, Leon et al. (1990) found cocaine in about 14.1% of the 643 motor vehicle occupants killed in traffic crashes in New York City during the period 1984-1987. Cocaine was the only drug screened for, and its presence was determined primarily from chemical tests for the cocaine metabolite benzoylecgonine. There was no statistical difference in this percentage with respect to ridership status (driver or passenger).

The results of tests of blood and/or urine from 347 fatally injured drivers in Washington State were reported by Logan and Schwilke (1996). Drugs most commonly encountered were marijuana (11%), cocaine (3%), and amphetamines (2%), together with a variety of depressant prescription medications including narcotics, benzodiazepines, barbiturates, and anti-depressants. Trends noted included an association of depressant use with higher BACs, while marijuana use was associated with lower BACs. Marijuana use was noted to be most prominent in the 15-30 year age group, stimulant use in the 21-40 year old group, and prescription depressant use was more prevalent in the 45+ age group.

The National Transportation Safety Board and the National Institute on Drug Abuse investigated 168 fatal-to-the-driver trucking accidents in eight states over a one-year period in 1987-1988 (Crouch, Birky, Gust et al., 1993; National Transportation Safety Board, 1990a; National Transportation Safety Board, 1990b), and found a higher percentage positive for marijuana (13%) than did the above studies, and about the same percentage positive for cocaine (8%). They also found amphetamines in 7% of the drivers.

Several Canadian studies investigated drug use among fatally injured drivers. Cimbura, Lucas, Bennett, and Simpson (1982) presented the results of a comprehensive drug study carried out on specimens from drivers and pedestrians fatally injured in Ontario. Toxicological analyses were regularly performed on blood and urine and occasionally on vitreous humor, stomach contents, and liver. The analytical procedures could detect and quantitate a wide variety of drugs including such illicit drugs as cannabis. With respect to drivers, alcohol was found in 57% of the study sample and drugs other than alcohol (including salicylates), in 26%. However, in only 9.5% of the drivers were psychoactive drugs (other than alcohol) detected in the blood in concentrations that may adversely affect driving skills. Cannabinoids / THC and benzodiazepines accounted for a majority of the findings in this category, 15.7% for cannabinoids / THC and 5.7% for benzodiazepines. Narcotic analgesics were found in about 2% of the fatally injured drivers.

In a later study in Ontario, Cimbura, Lucas, Bennett, and Donelson (1990) examined the incidence and toxicological aspects of cannabis and ethanol in 1,394 fatally injured drivers and pedestrians in Ontario. The study subjects were 1,169 drivers and 225 pedestrians. THC was detected in the blood of 127 driver victims (10.9%) in concentrations ranging from 0.2 to 37 ng/ml, with a mean of 3.1 ± 5.0 ng/ml. For pedestrians, the incidence of THC in the blood was 7.6%. The incidence of THC in the driver victims in this study constituted an approximately threefold increase over the results of the earlier Ontario study. The authors attributed at least a part of the increase to inter-study differences in analytical methods for cannabinoids. In a more recent Canadian study, Mercer and Jeffery (1995) found that 13% of 227 tested fatally injured drivers in Vancouver, British Columbia were positive for marijuana and metabolites. Also, 4% were positive for cocaine, and 8% were positive for a CNS depressant (including 5% for diazepam). (These results were also reported in another paper noted elsewhere in this report.)

Drug presence in non-fatally injured drivers has been reported for patients admitted to trauma units in several locations. Dischinger and Birschbach (1990) examined the incidence of drugs (other than marijuana) seen in approximately 220 motor vehicle operators admitted to the Maryland Institute for Emergency Medical Services Systems over the period January-June 1988. They obtained positive test results for drugs other than alcohol as follows: amphetamines (0%); barbiturates (2.2%); cocaine (4.4%); methadone (0.4%); opiates (2.6%); and PCP (1.8%). Soderstrom, Triffilis, Shankar, et al. (1988) tested 393 car drivers only for alcohol and marijuana admitted to the same trauma unit in 1985-1986, finding 16% positive for marijuana alone, and 17% for marijuana plus alcohol.

Kirby, Maull, and Fain (1992) performed drug screens on 164 injured drivers admitted to a trauma center in Knoxville, Tennessee during early 1988. Drugs found were: marijuana (32%); benzodiazapine (12%); cocaine (5%); opiates (5%); amphetamines (2%); and barbiturates (1%). The King County, Washington study by Rivara and associates cited above also studied drug presence in 452 non-fatally injured motor vehicle occupants brought to an emergency room and then admitted to a hospital, finding that 25% of these subjects were positive for THC, and 8% were positive for cocaine. Again, there was no breakdown by type of occupant (driver or passenger).

Waller, Blow, Maio, et al. (1995) studied 717 drivers who came to two emergency rooms in southeastern Michigan for treatment following a motor vehicle crash. The study found that marijuana, cocaine, and/or opiates were present in 14% of the drivers. This study was unusual in that it is the only North American study of drugs in non-fatally injured drivers involving drivers who had presented at an emergency room but were not necessarily admitted to a hospital following initial examination and treatment.

In the most recent U.S. study of drugs among emergency room patients, Lillis, Good, Kwong, et al. (1999) found that 6.4% of 888 patients were positive for THC, 6.2% for cocaine or cocaine metabolite, and 4.1% for benzodiazepine. Few details about the study (including which emergency departments participated) are provided in the paper.

In a Canadian study, Stoduto, Vingilis, Kapur, et al. (1993) examined the incidence of alcohol and drugs in a sample of 339 drivers admitted to the Regional Trauma Unit at the Sunnybrook Health Science Center in Toronto, Ontario over a 37 month period. Drugs alone found were: marijuana (14%); benzodiazapines (12%); cocaine (5%); morphine (5%); codeine (4%); barbiturates (3%); pethidine (3%); diphenhydramine (2%); and pheniramine (2%). Drugs and alcohol combined were found in 17% of the drivers.

Foreign Studies

Pertinent more recent studies were conducted in Australia, Great Britain, Scandinavia, Belgium, France, The Netherlands, and Switzerland. (The reader is referred to a paper by Tornros (1990) for a review of earlier studies.) Three studies were concerned with fatally injured drivers. Two of these (Everest and Tunbridge, 1990; Everest, Tunbridge, and Widdop, 1988) were conducted in England and Wales. The studies found a wide variety of drugs, but only a small percentage of the drivers had used the drugs. In the first-cited study of 520 car drivers, such drugs were found in just 6.7% of the drivers: marijuana (2.3%); benzodiazepines (2.1%); anticonvulsants (0.4%); antihistamines (1.4%); and phenothiazines, antidepressants, and hypnotics (0.2% each). In the second-cited study, drugs of all types affecting the central nervous system were found in only 7.3% of the 330 drivers. Marijuana was present in approximately 1% of the drivers, and diazepam and/or its metabolite nordiazepam in less than 1% of the drivers. Other drugs were present in even smaller percentages of drivers.

A third study of fatally injured drivers (Gjerde, Beylich, and Morland, 1993) was conducted in Norway in 1989 and 1990 and examined the incidence of alcohol and other psychoactive drugs in 159 car drivers. The drugs found were: amphetamines (0.6%); antiepileptics (0.6%); barbiturates (0.0%); benzodiazepines (13.8%); marijuana (5.0%); cocaine (0.0%); muscle relaxants (1.3%); opiates (3.1%); and other drugs (0.6%). Of a total of 79 drivers fatally injured in single-vehicle accidents, 21.5% were positive for drugs other than alcohol.

In a study of the relative risk of driving while impaired by cannabis, Swann (2000) reported the results of tests of the presence of the active ingredient of marijuana, Delta-9-THC, in fatally injured drivers in New South Wales, Australia. He found that, in 4.3% of the 544 fatalities, cannabis was the only drug present, the driver was fully responsible for or contributed to his or her own death, and the levels of Delta-9-THC were sufficiently high to indicate that the driver was impaired.

Some other Australian studies have been concerned with drug presence in drivers in non-fatal crashes. A paper by Perl, Hodder, Havi, et al. (1990) reported drug test results of 612 drivers who had been hospitalized as a result of a traffic crash. They found that 37% of the drivers were positive for one or more of 56 different drugs found. Of interest here were: amphetamine / methamphetamine (1%); tranquilizers (2%); morphine (<1%); and marijuana (3%).

Two papers by Longo, Hunter, Lokan, et al. (2000a; 2000) documenting the results of a later Australian study used data on the tests of blood samples collected from 2,500 non-fatally injured drivers involved in road crashes. The blood samples were analyzed for the presence of alcohol, cannabinoids, benzodiazepines and stimulants. A total of 22.6% of drivers tested positive for at least one drug including alcohol. Either alone or in combination with other drugs, cannabinoids were found in 10.8% of the samples, benzodiazepines in 2.7% and stimulants in 1.3%. The authors reported that "a small number" of cannabinoid-positive drivers tested positive for tetrahydrocannabinol (THC, the main active ingredient in marijuana), but "most drivers" tested positive for the inactive metabolite.

Sjoegren, Bjoernstig, Eriksson, et al. (1997) reported the results of drug tests of 130 injured motor vehicle drivers who were hospitalized and 247 fatally injured drivers who were autopsied from May 1991 through December 1993. The tests were performed in Umea, Sweden and Gothenburg, Sweden. Benzodiazepines (3.5%), opiates (3.7%), and tetrahydrocannabinol (1.9%) were the most commonly drugs detected overall. Amphetamines, barbiturates, and other drugs were detected in smaller percentages of drivers.

A Belgian study (Schepens, Pauwels, Van Damme et al., 1998) sought to determine levels of alcohol and drugs of abuse in weekend drivers injured in car crashes. The study tested blood and urine samples of 211 drivers injured in weekend car crashes and involved five collaborating hospitals in Flanders. All injured weekend drivers admitted to the emergency units from July 1, 1994, to June 30, 1995, were included in the study sample. Twelve percent of the sampled drivers were positive for drugs, either in combination with alcohol or alone. Amphetamines were the most common drug (3.3%), followed by benzodiazepines (2.9%), cannabis (2.4%), and opiates (1.9%). Barbiturates and cocaine appeared less often, about 1% for each drug.

Also in Belgium, another paper documenting the BTTS study (Belgian Toxicology and Trauma Study) was performed between January 15th 1995 and June 15th 1996 (Charlier and Plomteux, 1998). The purpose was to investigate how frequently the drivers involved in road accidents were driving under the influence of psychotropic drugs. Two thousand fifty-three blood tests were performed of which 207 were in the Liege region. The national level data indicated presence of the following drugs: amphetamines (3.0%), barbiturates (1.3%), benzodiazepines (8.5%), cannabis (6.0%), cocaine (0.7%), opiates (1.1%), and propoxaphene (0.5%).

A French case-control study examined the prevalence of opiates, cocaine metabolites, cannabinoids and amphetamines in the urine of drivers injured in traffic crashes (Marquet, Delpla, Kerguelen et al., 1998). Subjects were from emergency departments of five hospitals and consisted of 296 "drivers" aged 18 to 35 and 278 non-crashed "patients" in the same age range. Screening for drugs in urine was performed by fluorescence polarization immunoassays in each hospital. Each positive result was verified using gas chromatography-mass spectrometry (GC-MS), in a single laboratory. The tests showed cannabinoids (active ingredient) in 16% of the drivers, opiates in 10%, and cocaine and amphetamines in less than 1% each. By contrast, an earlier study of "road users" (car / truck drivers, motorcyclists, bicyclists, moped riders and pedestrians) in the Netherlands (Vis, 1988) found only 5% of the 282 subjects tested positive for "cannabinoids or opiates." (3.5% were positive for medicinal drugs taken from a list of the 50 most frequently prescribed drugs, mostly benzodiazepines or their metabolites.)

Ward and Dye (1999) summarized the results of 20 epidemiologic studies of cannabis involvement world-wide. Their results indicate that cannabis was detected in as few as 2.5% to as many as 38% of the crash-involved drivers. These wide differences are due to cultural differences in different countries (U.S., Jamaica, and Australia), time differences (from 1982 to 1998), the threshold criterion for drug presence (THC or metabolite and plasma concentration level used for indication of drug presence), and the definition of the subject sample (all fatalities vs. injured drivers only vs. fatal young males).

Finally, a study in New Zealand examined the long-term traffic-crash effects of cannabis use (Fergusson and Horwood, 2001). The study sought to determine the association between cannabis use and traffic accident risk among those who reported driving a motor vehicle between the ages of 18 and 21 years. The subjects were a birth cohort of 907 New Zealanders, and data were collected on annual frequency of cannabis use over the period from 18 to 21 years; annual rates of traffic crashes during the period 18-21 years; and measures of driver behaviors and characteristics. The researchers found statistically significant relationships between reported annual cannabis use and annual rates of crashes in which driver behaviors contributed to the crash, indicating that those using cannabis more than 50 times per year had estimated rates of crashes that were 1.6 times higher than the rate for non-users. However, when the crash rates were adjusted, driver behaviors and characteristics related to cannabis use (drink driving behavior, risky or illegal driving behaviors, driver attitudes, and driver sex) the association disappeared. The authors concluded that "although cannabis use was associated with increased risks of traffic accidents among members of this birth cohort, these increased risks appear to reflect the characteristics of the young people who used cannabis rather than the effects of cannabis use on driver performance."

The most striking feature of the foreign studies is the wide variance among the countries studied with respect to the percentages of drivers with given drugs. The two British studies and the Australian study found much smaller percentages of drivers positive for drugs of abuse than did the Norwegian study, which had percentages more in the range of those found in the North American studies.


North American Studies

Studies of this type fell into two categories, (1) those that performed chemical tests of drivers stopped at some roadside location, and (2) those that surveyed drivers and asked them about their use of drugs prior to driving.

Only one U.S. study of the first category was found in our literature search, a study of 317 tractor-trailer truck drivers stopped at a truck weighing station in Brownsville, Tennessee during one week in December 1986 (Lund, Preusser, Blomberg et al., 1988). The presence of a wide variety of drugs was studied. Drugs and drug types found included: marijuana (15%); cocaine (2%); prescription stimulants (5%); and non-prescription stimulants (12%).

Several surveys have been conducted to obtain self-reports of driving after using drugs. In an earlier study, Hingson, Heeren, Mangione et al. (1982) performed anonymous random digit dialing telephone surveys of nearly 6,000 16-19 year old respondents in Massachusetts and upstate New York in 1979-1981. These surveys explored frequency of, among other things, driving after using marijuana. They found that 17% of the respondents reported driving at least once in the previous month after using marijuana, and 4% had done so after using "psychoactive drugs." In a 1982 survey of 623 seventh and tenth graders of age 16 or greater in the Boston, Massachusetts area, Wechsler, Rohman, Kotch, et al. (1984) found that 17% had driven after smoking marijuana during the school year just ending.

The National Household Survey on Drug Abuse (NHSDA) conducted by the Federal Government is another source of self-reported information on drinking, drug use, and driving in the United States. This survey has been conducted periodically since 1971, and the 1996 survey contained a special Driving Behaviors Module funded by NHTSA. A summary of the design and findings of the 1996 survey drawn from this module is contained in a government report (Townsend, Lane, Dewa et al., 1998).

The Driving Behaviors Module involved 11,847 personal interviews in a nationally representative sample of households. The respondents were individuals age 16 and older reporting that they had driven a motor vehicle in past 12 months, and whether they had driven within two hours after drug or alcohol use. The report by Townsend and associates indicated that 3.7% of the respondents had driven within the past 12 months after using marijuana, 1% had driven after using cocaine, and 1% had driven after using tranquilizers. Less than 1% was reported for stimulants and sedatives. Among those who had driven within two hours after using marijuana, about one in four had driven six or more times during the past month.

A more recent Canadian study reported the results of a roadside survey conducted from August 9 to August 29, 1999 in order to determine drug use among Quebec drivers (Dussault, Lemire, Bouchard et al., 2000). The survey used a two-stage stratified sampling procedure, and included 147 sites representative of the Quebec driving population. During both daytime and nighttime, a total of 5,507 drivers participated in the survey, among which 95.9% provided a breath sample and 41.4% a urine sample. Among those who refused to provide a urine sample, 70.1% agreed to provide a saliva sample. Altogether, 82.5% of the drivers provided either a urine or a saliva sample. Regardless of time of day, a BAC above the illegal limit (.08) was found in 0.8% of the breath samples. During time period surveyed, drugs were found in the following proportions: cannabis (5.22%), benzodiazepines (3.66%), cocaine (1.09%), opiates (1.08%), barbiturates (0.35%), amphetamines (0.07%), PCP (0.03%).

Foreign Studies

Two studies were found, one in Germany and the other in Italy. The German study was documented in an article by Krueger, Schulz, and Magerl (1995), and the Italian study in an article by Zancaner and associates (1995). The data in the study by Krueger and associates were taken from a roadside survey in Germany in 1992. Chemical analyses were performed on 2,234 saliva samples, with the adjusted results indicating the following percentages of drivers nationwide testing positive for given drugs: benzodiazepines (2.7%); opiates (0.7%); marijuana (0.6%); barbiturates (0.6%); amphetamines (0.08%); and cocaine (0.01%).

The study by Zancaner and associates also involved a roadside survey. The survey was conducted in 1994-1995 to determine drug usage of drivers in the Veneto region of northeast Italy. According to the abstract in the paper, the study involved 1,237 drivers, including 265 who were suspected of driving under the influence of drugs. Both clinical and toxicological assessments were made, the latter involving chemical tests of both urine and blood. The study concluded that 10.6% of the subjects were "under the influence of drugs of abuse or psychoactive drugs," with the most frequently abused substances being (in order of frequency) cannabinoids, cocaine, amphetamines, opiates, benzodiazepines, and barbiturates. Unfortunately, the paper did not describe sample selection techniques, so we do not know what driver group is represented by the participating drivers.

The recent review by Ward and Dye (1999) of epidemiologic literature worldwide pertaining to cannabis and highway safety found that the few roadside surveys of THC in drivers have yielded fairly low rates: 4% in Canada in 1974, 1.2% in Italy in 1982, and only 0.6% (based on the less sensitive saliva test) in Germany in 1992-1994.

De Gier (2000) summarized the literature on the prevalence of illicit drugs in non-crashed drivers in different European countries. Although a total of 23 studies published in the time period 1990-1998 were collected, only four large-scale studies had large enough samples to be included in de Gier's summary. The results indicate that cannabis and opiates, the most frequently detected drugs, were detected in less than 1% in the general driver population. Benzodiazepines amounted to 3.6%.

Self-reported data are available from surveys in several countries. For example, Del Rio and Alvarez (1994) surveyed 1,500 Spanish drivers about their drug use and driving. Drugs for which responses were solicited were cannabis, amphetamines, tranquilizers, opiates, cocaine, psychedelic drugs, and inhalants. They found that, of these, cannabis was the drug used most frequently before driving in the past year (1.7% of the respondents), and that cocaine was next highest at 1%. Percentages for each of several other drugs were less than 0.5%. Lesch, Lentner, Mader, et al. (1989) surveyed 1,234 Austrian drivers about their use of several classes of drugs. Their data indicate that 2.6% were "possibly" using the following drugs while driving: CNS depressants, 2.4%; stimulants, 9.6%; analgesics or spasmolytics; and antidepressants or neuroleptics, less than 1%.


North American Studies

Two U.S. studies were found that conducted a formal assessment of drug-crash risk. The first study (Terhune, Ippolito, Hendricks et al., 1992) used the responsibility-analysis approach (4) and found that no increased crash risk was associated with marijuana or cocaine alone, but that multiple drug use may be associated with increased responsibility.

In the second U.S. study, Leveille, Buchner, Koepsell, et al. (1994) used a matched case control study design to compare drug use habits of 234 crash-involved older drivers (65+ years old) in the Seattle, Washington area, relative to 447 control subjects matched on age, sex and county of residence who were not involved in a crash during that period. They failed to find an over-involvement in crashes for people taking benzodiazepines or sedating antihistamines, relative to people not taking this drug. Although they were aware that other studies found an association with crash rates, they were not able to provide a good explanation for their failure to find one, except to note that there is a large variance in the prolonged effects of different benzodiazepines, and that their sample may have been too small to yield the expected effect. In this study there was an elevated crash risk for opioid users (RR=1.8). In both the crash-involved group and the control group, codeine was the most commonly used analgesic (61% and 51% of all analgesics, respectively). Note, however, that the time course and effects of orally ingested codeine may be very different from those of intravenously taken heroin.

Another U.S. study was performed by Ray et al. (1992) on a cohort of 16,262 drivers of age 65-84 registered in the Tennessee Medicaid program. They found that those taking benzodiazepines had a relative risk of an injury crash of 1.5, and those taking tricyclic antidepressants had a relative risk of 2.2. Persons taking high dosages of these drugs experienced higher relative risks, 2.4 for diazepam and 5.5 for amitriptyline. However, they did not find any increase in relative risk for people taking oral opioid analgesics.

Hemmelgarn, Suissa, Huang et al. (1997) conducted a cohort study of 224,734 Canadian drivers 67-84 years old. They obtained relative over-involvement rates of 1.45 (after initial use) and 1.26 (after continued use) in injury crashes for users of long-life benzodiazepines. However, they found no over-involvement for drivers using short-life benzodiazepines.

In another Canadian study, Neutel (1995) assessed the risk of hospitalization for injuries received in traffic accidents after a first prescription for benzodiazepines was filled. Saskatchewan Health supplied study populations of 78,000 adults who received benzodiazepine hypnotics, 148,000 who received benzodiazepine anxiolytics, and 98,000 control subjects. These populations were monitored for 2 months after the index prescription fill-date for hospitalizations due to traffic crashes. Analysis showed an odds ratio (OR) of 3.9 (1.9 to 8.3) for persons taking benzodiazepine hypnotics and an OR of 2.5 (1.2 to 5.2) for those taking Benzodiazepine anxiolytics, with regard to hospitalization due to traffic crashes within four weeks after the prescription was filled. Within two weeks after the prescription was filled, the OR had risen to 6.5 (1.9 to 22.4) for hypnotics and 5.6 (1.7 to 18.4) for anxiolytics. After 1 week, the ORs were even higher (9.1 and 13.5), but the confidence limits were wide. The highest risk groups were the youngest age group (20 to 39 years old) and males.

Foreign Studies

Drummer (1995) used data from some 1,000 fatal crashes Victoria, New South Wales, and Western Australia to develop fatal-crash risk factors for several drugs. Again, the responsibility analysis approach was used. Drummer computed odds ratios for drugs / no-drugs for each drug and found that only alcohol gave a statistically significant odds ratio greater than one (odds ratio=7.6, p <0.001). The odds ratio for cannabis approached significance (p=0.065) and was actually less than one (0.60), suggesting a beneficial effect of marijuana use. The odds ratios for drivers with stimulants, benzodiazepines, and opiates were 2.0 for each drug, but were not anywhere near statistical significance (p=0.217, 0.295, and 0.220, respectively).

The study by Longo, Hunter, Lokan et al. (2000) cited earlier in this report analyzed the causal role of alcohol, cannabinoids, benzodiazepines and stimulants in crashes involving 2,500 injured Australian drivers. The responsibility analysis approach also was used in the analysis. Benzodiazepine use was associated with higher culpability when those with very low concentrations were excluded (percentage ratio 3), but THC was not associated with increased culpability. Relatively few drivers tested positive for stimulants and there was no clear evidence of greater culpability.

However, Swann (2000) obtained quite different results with regard to THC in another Australian study of fatally injured drivers. In this study, the drivers were tested for the presence of Delta-9-THC, which only persists in the body for hours after use and allows the identification of drivers who were impaired by cannabis at the time of their death. By contrast, other Australian studies showing no culpability identified cannabis users by measuring Carboxy-THC, which can remain detectable in body fluids for weeks after cannabis use. Since impairment after cannabis use only persists for hours, the bulk of the cannabis users identified by Carboxy-THC would not be impaired, and responsibility studies did not show an increased risk of an accident for this group. These studies identified drivers who had consumed cannabis, not necessarily drivers who were impaired by cannabis.

Swann's study found that drivers who test positive to Delta-9-THC and have no other psychotropic drug or alcohol present have a relative risk (as shown by odds ratio) of 6. In 4.3% of the 544 fatalities, cannabis was the only drug present, the driver was fully responsible for or contributed to his or her own death, and the levels of Delta-9-THC were sufficiently high to indicate that the driver was impaired.

Barbone, McMahon, Davey, et al. (1998) analyzed the crash involvement of an extensive cohort of 40,400 UK residents who had benzodiazepine prescriptions during their three-year study period (1992-1995). In their sample, 19,386 were involved in a "first road accident" during that period, 1,731 were actual users of prescribed drugs, and 916 were users of benzodiazepines. Their principal findings and conclusions were that (1) users of anxiolytic benzodiazepines were at increased risk of experiencing crashes (average odds ratio = 1.62), and (2) crash risk was dose related. Also, the risk associated with benzodiazepine decreased with increasing driver's age (2.66 for people over 30) and increased when the driver also used alcohol (8.15 for those with positive BAC). Without alcohol, the odds ratio for a crash was 1.52 which was still significantly different from 1.0. On the other hand, no over-involvement was obtained for hypnotic benzodiazepines. This could be because (1) those are taken mostly at night before sleep; whereas anxiolytic benzodiazepines are taken during the day, often before driving, and (2) the case crossover method underestimates the effects the longer people have been taking the drug - and older people are often taking it chronically.

Marquet, Delpla, Kerguelen, Bremond, et al. (1998) performed a case-control study involving opiates, cocaine metabolites, cannabinoids, and amphetamines in France. Subjects were recruited from persons admitted for treatment in emergency departments of five hospitals nationwide and comprised 296 "cases" aged 18 to 35 and 278 non-traumatic "controls" in the same age range. Screening for drugs in urine was performed by fluorescence polarization immunoassays in each center, and each positive result was verified using gas chromatography-mass spectrometry (GC-MS), in a single laboratory. Statistical analysis comprised single-step logistic regression and simultaneously took account of confounding factors and the final differences in prevalence values between the two populations or different subgroups. The study's data indicate that only cannabinoids and amphetamines had odds ratios differing significantly from 1, 1.94 for cannabinoids and 0.57 for amphetamines (suggesting a beneficial effect).

Finally, the French Benzodiazepine/Driving Collaboration Group (1993) conducted a study that looked specifically at at-fault crash-involved drivers and pedestrians. Their results were consistent with those of the U.S. study by Leveille and associates summarized above, in the sense that they, too, failed to find an odds ratio significantly greater than 1.0 between injured drivers and pedestrians who were responsible for the crash vs. those who were not responsible for the crash, once the effects of BAC were held constant. One problem with their data is that most crashes were of young males drivers at night, while most of the benzodiazepine users were adult females driving during the day - so the two groups had different risk levels and exposure to begin with.


North American Studies

Several earlier studies dealt with drugs in drivers arrested in the United States for traffic violations. White, Clary, Graves, et al. (1981) reported the results of blood tests stemming from some 72,000 California drivers arrested for impaired driving in the 1970s. Percentages of tests indicating the presence of various classes of drugs in drivers with BACs < .10, were as follows: sedative/hypnotic, 30%-47%; phencyclidine (PCP), 79%; and morphine, 62%. Another study of drivers arrested in St. Louis, Missouri in the mid-1980s yielded the following percentages: phencyclidine, 47%; marijuana, 47%; benzodiazepines, 22%; barbiturates, 15%, opiates, 11%; and cocaine, 9% (Polkis, Maginn, and Barr, 1987). A third study in Washington, DC reported in 1992 found 39% of arrested drivers positive for cannabinoids, but only 9% positive for phencyclidine (Sutton and Paegle, 1992).

A fourth study of U.S. drivers arrested for traffic violations dealt with quite a different population (Brookoff, Cook, Williams et al., 1994). This population had the following characteristics:

The study involved 175 consecutive cases of such drivers who were stopped on 46 consecutive summer nights in Memphis, Tennessee. Urine tests for cocaine or marijuana metabolites were performed for 150 of the drivers who agreed to provide a sample. These subjects also received a standard behavioral impairment evaluation which was later compared with the results of the drug screening test. The authors reported that 59% tested positive for either marijuana (33%), cocaine (13%), or both (12%). The authors then state that those who tested positive were "intoxicated" by those drugs.

Taken at face value, these conclusions do not flow from the data presented and are misleading. First, although many of the subjects exhibited decrements on field sobriety tests, it is a leap of faith to say that they were "intoxicated" by those drugs. Second, in the text of the paper, it is reported that 30% of the marijuana positives were not confirmed by laboratory analysis, yet they appear as positives in the main table presented in the paper, the abstract, and in most of the discussion. Revising the conclusions to say that 59% (possibly adjusted down to 53% to reflect unconfirmed positives) of drivers police suspected of drug-impaired driving were positive for the marijuana or cocaine metabolites would make them more nearly in the same "ball park" as those flowing from the DEC evaluations of Preusser and associates and of Adler and Burns, summarized later below.

The most recent U.S. paper in our review (Walsh, Cangianelli, Buchan et al., 2000) reports the results of a study in which DUI Officers (Tampa Police Department, Tampa, Florida) were trained to use two rapid immunoassay devices to test driving-under-the-influence (DUI) suspects for recent drug use. In addition to routine determination of BAC, urine specimens were collected by police officers from persons placed under arrest for suspicion of DUI. Two hundred twenty-seven urine specimens were collected and analyzed by one of two "on-site" immunoassay kits being evaluated. The arresting officer conducted all analyses, but re-analyses were performed on all specimens testing positive on-site, and on 10% of on-site negatives. The results indicated that 30% also tested positive for one or more illegal drugs, and 55% of those who passed the Breathalyzer test with legal levels of alcohol (i.e., BAC<.08) were positive for one or more illegal drugs. Marijuana and cocaine were the primary drugs detected (19% and 16%, respectively). Narcotics and amphetamines were found in less than 1% of the specimens.

Another source of information about drugs in drivers arrested in the United States for traffic violations is NHTSA's Drug Evaluation and Classification (DEC) Program (See see reference). That program uses trained Drug Recognition Experts (DREs) to determine drug usage by looking for such signs and symptoms as dilated / constricted pupils, horizontal gaze nystagmus, and time estimates, among others. Four studies were found that documented evaluations of DEC programs. The most comprehensive of these was the study by Preusser, Ulmer, and Preusser, (1992) which evaluated DEC programs in selected sites in the states of Arizona, California, Colorado, New York, and Texas over various periods during 1986 to 1991. A total of 1,711 cases were evaluated, with laboratory tests available for 1,469 of these. The population dealt with was a subset of drivers already suspected of DWID. The data from this study suggest that such drivers who are classified as drug-impaired by DREs comprise some 1-3% of drivers arrested for DWI. Of the 1,469 drivers tested by a laboratory, the following drugs were found: marijuana (42%); CNS stimulants (36%); CNS depressants (16%); narcotic analgesics (13%); and PCP (5%).

Adler and Burns (1994) provided additional details on a subset of this population of drivers in their evaluation of the Arizona DEC program. In this study, 484 drivers examined by certified DREs from the Phoenix Police Department during January 1989 through May 1993 were laboratory-tested. Drugs found in the laboratory tests were: marijuana (34%); opiates (28%); cocaine (24%); benzodiazepines (22%); amphetamines (18%); barbiturates (7%); PCP (5%); and others drugs (30%).

The third DEC-documented program providing data on drug presence in drivers suspected of driving under the influence of drugs (DUID) was conducted in Virginia during the period 1988- 1990 (Jernigan, 1992). A total of 1,199 blood samples were submitted for analysis. The laboratory analysis found the following drugs: marijuana alone (17%); PCP alone (14%); other drugs alone (9%); and multiple drugs (18%). Finally, Tomaszewski, Kirk, Bingham, et al. (1996) evaluated the results of DRE assessments in Denver, Colorado, where urine screens found cannabis (67%), narcotics (5%), benzodiazepines / barbiturates (10%), stimulants (33%, including cocaine metabolites), and a few other drugs (<2%).

It should be kept in mind that the DEC subjects for whom the toxicological analyses were performed were drivers who (1) had been arrested for DUI, (2) were suspected of having been impaired by drugs other than alcohol, and (3) had been evaluated by the DREs. Thus, the above findings apply only to this restricted group of DUID suspects rather than to all drivers arrested for DUI.

Finally, we cite an interesting study by Marowitz (1994) which deals with a much different group of arrested drivers, but which, nevertheless, sheds some light on the overall drug-crash problem. This study compares the driving records of 106,214 persons arrested for drug offenses in California in 1989 with the records of 41,493 persons from the general driving population. Extensive statistical analyses were performed on the collected data, with careful attention to accounting for differences between the drug group and the non-drug group. No information was available on actual drug presence in the subjects.

The author concluded that drug arrestees committed significantly more (two to three times as many) traffic violations and significantly more (1.34 to 1.66 times as many) traffic crashes during the study period as did the general driving population. With respect to traffic crashes, the author concluded that drug arrestees had more traffic crashes for the year prior to the arrest and the year following the index arrest.

Other analyses were also conducted, including some aimed at estimating the effect of drugs on crash risk. Of these, the analysis of single-vehicle crashes (which are more likely to be caused by driver error) was especially interesting, indicating the drug-arrestees had a significantly higher percentage of such crashes than did the control group. The weighted mean number of single-vehicle crashes for all drug arrestees was 2.47 times that of the control group. The author also found that the drug arrestee group had significantly more injury/fatality crashes. Nevertheless, the author was cautious about attributing increased traffic-crash risk to drug usage.

The extensive statistical analyses of the extremely large amount of data collected support the author's conclusions about increased traffic violations and crashes among drug arrestees. However, there is insufficient evidence in this study to conclude that the higher number of traffic arrests and traffic crashes among drug arrestees is due to driving while impaired by drugs. The author's findings on single-vehicle crashes and injury crashes not withstanding, support for the hypothesis that drugs used by drivers increases traffic-crash risk requires an unwarranted leap to the conclusion that the drug arrestees also drive more while impaired by drugs and that such drug-impaired driving is the cause of their increased number of traffic violations and crashes.

In sum, the studies discussed in this section do not provide much information about drug use among drivers in general who are stopped or arrested for traffic violations, although "ballpark" estimates for drivers arrested for DWI appear to be in the 1-10% range. However, these studies do indicate strongly that relatively high percentages of such drivers who are also suspected of "drugged" driving by the police and are evaluated by drug recognition experts are positive for a number of drugs that could impair driving performance. One small study in Tennessee also suggests that a large percentage of drivers interdicted by the police for other driving behaviors associated with drug impairment may also be positive for marijuana or cocaine. Finally, a very large study of driver records in California suggests that drivers arrested for a number of non-traffic drug-related offenses have increased numbers of traffic violations and traffic crashes.

Foreign Studies

The largest number of foreign studies in this category have been conducted in Norway. This review covers nine Norwegian studies based on data dating back to 1978. Typically, these studies are based on analyses of blood samples from drivers suspected of driving while under the influence of alcohol or drugs. The samples were screened by the Norwegian National Institute of Forensic Toxicology. Unfortunately, the articles do not always describe the sampling protocols.

Bjorneboe, Bjorneboe, Bugge, et al. (1988) reported data from the years 1978, 1983 and 1986, indicating that there was a 19% increase in the number of samples submitted from 1983 to 1986. Of the 14,350 samples collected in 1986, 789 were tested for drugs. The authors state that 82% of these samples were found to be drug-positive compared to 54% and 88% of the 426 and 445 samples tested in 1978 and 1983, respectively. There was an increase in amphetamine use from 1978 to 1986. Benzodiazepines and marijuana levels were found in high levels during all the time periods. In 1987, the drugs found were: amphetamines (23%); diazepam (31%); flunitrazepam (25%); opiates (8%); marijuana (42%); and others (15%).

Christophersen, Bjorneboe, and Gjerde (1990) present data for the period November 1986 to February 1988. They indicate that the 270 samples analyzed for drugs were randomly selected from the drivers suspected of driving while under the influence of alcohol (DWI) or drugs (DWID). Drugs found were: amphetamines (4%); cocaine (0%); benzodiazepam (23%); and marijuana (47%). The percentages were about the same for drivers arrested for DWI and DWIDs.

Data from Gjerde, Christophersen, and Morland (1992) document some of the Norwegian findings for the 1989-1990 period which generated 380 blood analyses. Their report focuses on amphetamines which were found in 17% of the sample. The authors also examined impairment in 284 cases. Forty-nine of 81 amphetamine-positive drivers with no other drug were evaluated clinically for impairment, and 78% were found to be impaired. Data on the presence of other drugs are limited, with breakdowns being given only for 25 cases of arrests resulting from traffic crashes.

Later Norwegian data on drug presence among drivers suspected of drugged driving were collected in 1993 and are reported in two studies (Christophersen, Beylich, Bjorneboe et al., 1995; Morland, Beylich, Bjorneboe et al., 1995). In the first-cited study, the authors found drugs (primarily benzodiazepines, marijuana, amphetamines, and opiates) in 30% of the 1,197 suspected drivers tested. Of the 362 cases with drugs, 206 (57%) had benzodiazepines, 147 (41%) had THC, 81 (22%) had amphetamines, and 58 (16%) had opiates. As a percentage of total cases tested, these numbers are 17%, 12%, 7%, and 5%, respectively. The second-cited study was based on drug analyses of blood samples from 394 drivers suspected of drugged driving and involved in non-fatal crashes in Norway in 1993. The most prevalent drugs were benzodiazepines, cannabis, opiates, and amphetamine. Percentage of cases with positive results for such drugs were 13.7%, 7.6%, 4.1%, and 4.3%, respectively. The authors report evaluating all cases for possible impairment "based on drug concentrations," but do not describe their criteria for impairment. They state that more than 75% of the drug-positive drivers were "impaired" or "likely impaired."

The most recent Norwegian data indicate an increase in the number of drivers suspected by the police as being under the influence of drugs. A review by Christophersen and Morland (1997) found the most commonly detected drugs among 3,329 blood samples from drivers suspected by the police as driving under the influence of drugs were benzodiazepines (37%), THC (31%), amphetamine (30%), and narcotics (12%). Multi-drug use was frequently found (>60%). The occurrence of amphetamine also increased considerably from that found in prior studies. Christophersen and Morland (1997) concluded that the frequency of drugged drivers apprehended in roadside traffic was at least 10-fold higher in Norway than in most other countries, and that this over-representation was probably due "mainly to differences between national road traffic acts and the level of attention to the problem, and not to national differences in the prevalence of drugged driving."

Another study of the 1995 Norwegian data by Skurtveit, Christophersen, and Morland (1999) found that 71% of the drivers influenced by amphetamine in 1995 were drivers who had been arrested earlier because of impaired driving. More than 60% of the drivers apprehended in 1995 for driving under the influence of amphetamine had alcohol, THC or benzodiazepines in their initial sample. Christophersen, Abotnes, and Skurveit (2000) examined the 1995 data further for the prevalence of benzodiazepines in the 3,343 drivers who were apprehended by the police for suspicion of influence by drugs, finding that benzodiazepines (which represented some of the most frequently detected drugs) were detected in approximately 30% of the cases. In 8% of the cases, one benzodiazepine only was detected, and the blood drug concentrations in most of these cases were above therapeutic levels. In the remainder of the cases, one or more benzodiazepines were combined with illegal drug(s) (73%), other prescribed drugs (10%), or alcohol (15%).

Christophersen and associates also found the frequencies of benzodiazepines detected among drivers from different Norwegian counties correlated with benzodiazepine prescriptions from the same area. Further, 62% of the drivers had been arrested in the past 11 years for the same reason and there were 5.6 cases per rearrested driver. Alcohol was most frequently detected for those arrested for the first time before 1992, while benzodiazepines or illegal drugs were most frequently found for those with the first arrest during 1992 - 1995. They concluded that "our study shows that apprehended drivers using benzodiazepines are mainly represented by drug abusers, combining prescribed and illegal drugs and/or alcohol. A treatment program or other reactions, are thus necessary in addition to fines, prison penalty and suspension of driving license."

An earlier Norwegian study of rearrest recidivism (Gjerde, Bjorneboe, Bjorneboe et al., 1988; Gjerde, Bjorneboe, Christophersen et al., 1988) examined the driving records 100 drivers first arrested in 1983 and followed through 1988. Fifty of the 100 drivers were initially arrested for drunken driving and 50 for drugged driving. After three years, 34% of the drug group had recidivated compared to 20% of the drunk group. After five years, 50% of the drug group had recidivated compared to 32% of the drunk group.

Another Scandinavian study, this time in Denmark, (Christensen, Nielsen, and Nielsen, 1990) reviewed data on 461 cases that police had suspected of driving under the influence of drugs. Police provided the Medicolegal Council with the result of a clinical examination. Based on the results of the clinical examination, an estimation was made as to whether the driver would have been influenced by the drug. The samples were also screened for 100 different legal and illegal drugs. Apparently, the study was looking for a measure of the relative risk of various drug groups. This was done, we surmise, by comparing the percentage of crash-involved drug-driving suspects in a given drug group with the percentage of crash-involved drug-driving suspects in all drug groups.

Drugs found in drivers suspected of drugged driving were: benzodiazepines (65%); opiates (38%); and antidepressants and anticonvulsants (12.6% each). From an analysis of their "relative-risk" data (relative risks not presented in the article) and a cursory review of the literature on drug effects on behavior, the authors concluded that there is ". . . a likely traffic danger by persons taking drugs, mainly barbiturates, benzodiazepines, and cyclic antidepressants," and that ". . . drivers under the influence of opiods do not contribute excessively to accidents." These conclusions should be regarded with some skepticism, not only because of a lack of information about specific drugs and chemical analysis methods, but also because of the highly selective nature of the samples. The results of the analyses of relative risk should definitely not be taken at face value, and the sweeping conclusions about "traffic danger" of various drug categories do not follow at all from the design and results of the study.

More recently, Steentoft and Worm (1996) investigated the frequency of benzodiazepines, morphine, amphetamine and cocaine in Danish traffic cases. The subjects were 294 drivers whose blood alcohol samples were negative. Drugs were detected in 27% of the cases, but analyses for drugs were only requested by the police in less than 10% of the cases. Twenty-three percent of the cases were positive for benzodiazepines with diazepam being by far the most frequently occurring drug followed by flunitrazepam. Morphine was detected in 7% of the cases, amphetamine in 5% and cocaine in only one case. Compared with a similar investigation in 1983, the frequency of benzodiazepines increased from 15% to 23%.

Holmgren, Loch, and Schubert (1985) analyzed the urine of Swedish drivers stopped by police for suspected DWI/DUID. Their analysis revealed that one third of the stopped drivers had some drugs-related substances in their urine - especially benzodiazepines and cannabis. None had any narcotics. Among those who were stopped but did not have detectable alcohol odor or detectable BAC levels, 91% had one or several "traffic hazardous drug substances in their body fluids." Of the legal drugs, the most common were benzodiazepines followed by analgesics. The most common illicit drugs were cannabis, followed by CNS stimulants and opiates. Of the drug-positive drivers, the police suspected 49% of DUID (rather than DWI), the medical doctors who interviewed them and took their urine suspected 78% of DUID, and a total of 99% of the drug-positive drivers were suspected of DUID by either or both the police and the doctors.

The prevalence of drugs in DWI/DUID suspects not only varies by country and culture, but also changes over time. Lillsunde, Korte, Michelson, et al. (1996) analyzed the blood of 298 Finnish drivers suspected of DWI/DUID in 1979 and 332 Finnish drivers suspected of DWI/DUID in 1993. They found drugs "hazardous to traffic safety" in 7.0% of the drivers in 1979 and in 26.8% of the drivers in 1993. Benzodiazepines were the most frequently detected drugs in both years: 6% of the cases in 1979, and 22.9% in 1993. Of the benzodiazepines, diazepam was the most common (75%), and oxazepam was the second (36%) (34% had multiple benzodiazepines). In half of the benzodiazepine cases, the concentration exceeded the limit that "can be evaluated as causing possible driving impairment". Interestingly, unlike other studies, they did not find an increase in benzodiazepine frequency with age. In the 1993 sample, THC was detected in only 8 drivers, constituting 2.4% of the sample. Seven of these 8 drivers were under 35 years old. No narcotics were found in any drivers in either 1979 or 1993.

Perl, Hodder, Havi, et al. (1990) examined drug presence in 233 drivers arrested in 1987-1989 for drunk driving and suspected of drugged driving in New South Wales, Australia. The study found that 70% of the drivers were drug-positive with the following frequencies of occurrence for the various drugs: narcotics (43%); benzodiazepines (35%); marijuana (30%); and stimulants (10.5% ). Additional data from 1989 through 1993 (Perl, Mascord, Moynham et al., 1995) on a wide variety of drugs found that about 80% of the drivers were drug-positive. A breakdown of 200 of the drug-positive cases by drug type indicated marijuana was the most common drug with 159 occurrences (79.5%). Opiates and other narcotic analgesics were found in 41% of the drivers, amphetamines in 17%, and minor tranquilizers in 26%. Some of the characteristics of the study group are also presented.

In an interesting Swiss study, Augsberger and Rivier (1997) examined the epidemiologic and analytical laboratory records of 661 living drivers suspected of driving under the influence of a drug (DUID) during a 13-year period from 1982 to 1994. A traffic crash had occurred in 254 (40%) of the records, 273 (43%) drivers were suspected of DUID during police controls, and 95 (15%) drivers were suspected of DUID because of their erratic driving. The authors found one or more psychoactive drugs (including alcohol) in 92.8% of the samples, including: cannabinoids (57%), opiates (36%), ethanol in (36%), benzodiazepines in (15%), cocaine (11%), methadone (10%), and amphetamines (4%). The majority (58%) of cases had two or more drugs in biological samples.

The last of the foreign studies in this category was conducted in Slovenia (Zorec-Karlovsek and Lokar, 1988). The authors report the use of alcohol, "trigonics" (not a term in use in this country), and benzodiazepines in Slovenian drivers arrested for suspected alcohol intoxication. Three hundred urine samples from the 36,613 arrestees were analyzed for the presence of psychotropic substances by chromatographic methods, enzyme-immunological and color test methods. Benzodiazepine use was confirmed in 1%, opiates in 0.33%, and marijuana in 0.33% of the urine samples. Benzodiazepine use was claimed in 0.9% of the 3,105 individuals making statements about drug use but, interestingly, was confirmed in only about half of the lab-tested drivers claiming to have used benzodiazepines in their statements, with diazepam being the most frequently abused agent.

As with the U.S. studies, foreign studies of drugs in drivers stopped or arrested for traffic violations have been concerned primarily with such drivers who were already suspected of drug impairment prior to lab testing for drug presence. The lone exception seems to the Slovenian study which tested a sample of all drivers arrested for DWI and found that some 2% were positive for the psychotropic drugs screened for. The studies of drivers suspected of drugged driving are difficult to compare because of inconsistencies in defining drug classes but do indicate high percentages of such drivers were positive for a number of drugs that could impair driving performance.

In all of the above studies, drug presence was measured given suspected impairment. However, none of these studies established a direct relationship between drug presence and specific impairment. To overcome this limitation, a study of the relationship between drug presence and specific behavioral impairments in drivers stopped for suspected traffic violations, is needed. Two studies that have done this are discussed below.

In the first study, Neuteboom and Zweipfenning (1984) tested the mental and physical performance of a sample of 906 Dutch drivers who were arrested for DWI/DUID and who admitted to medical officers that they took benzodiazepines. These drivers were most of the 3.2% of 38,203 drivers who were stopped for suspected impaired driving during 1981-1982. The authors compared these drivers' "performance" to that of drivers with identical BACs but without any drugs. The measures of performance were holistic measures of physical condition and mental condition as perceived by medical officers. Physical condition was noted in terms of gait ("steady" or "unsteady") and "mental condition" was noted in terms of behavior ("restrained" or "uncontrolled"). As expected, they found that the frequency of both physical and mental impairments increased with increasing BACs. More important for this report, however, they also found that among those drivers with BACs below .20, those who also took benzodiazepines were more likely to be impaired than those with alcohol only, but for drivers with BACs above .20, benzodiazepine did not seem to increase impairment beyond that already explained by alcohol alone. The authors also found that 9.7% of the 38,203 drivers had used drugs before driving.

Of these, 8.2% used "non-medical drugs," and of these, 40.2% used heroin (17.1%), methadone (20.1%), or both (3.0%); and 6.2% used cocaine (3.6%), and other stimulants (2.6%). Thus, of the total sample of suspected impaired drivers, only one-twentieth of one percent (0.005%) used a stimulant. This percent is probably an underestimate since all the drugs listed were based on admission or medical officers' opinions without the benefit of a chemical test. And, of the total sample of suspected impaired drivers, less than one-third of one percent (0.32%) used a narcotic.

In the second study, Kuitunen, Meririnne, and Seppala (1994) found that of 387,770 stopped for DWI/DUID in Finland over a period of 6 years (1987-1992), a total of 130 drivers tested positive for diazepam only (i.e., only one third of one percent of all stopped drivers!). These were further divided into two groups: chronic vs. acute drug users (based on the blood diazepam / nordiazepam ratio). All drivers were given the Finnish Clinical Test for Drunkenness (CTD), which is similar to the U.S. Standardized Field Sobriety Test (SFST). The CTD consists of a set of motor coordination tests (walking with open and closed eyes, observing the gait while turning, touching finger-to-finger, and collecting small objects), balance test (Romberg test with open and closed eyes) ocular saccadic movement (nystagmus), mental ability (backward counting by fixed subtraction, and time orientation), and behavioral tests (speech, pulling oneself up, and overall behavior). Performance on each test was scored on a scale of 1-4, where 1 is normal performance and 4 is highly unstable or incorrect. The results showed that - except for slurred speech - chronic use of diazepam did not cause any significant impairments. On the other hand, acute use was significantly associated with impairment in most of the tests, including walking with open and with closed eyes, the pooled motor subtests, and the pooled behavioral subtests.


The epidemiologic literature on drugs and driving has continued to grow since the 1980 review of drugs and highway safety. However, most of the new studies are in two areas: drug presence in drivers involved in traffic crashes and drug presence in drivers suspected of drugged driving violations. Only one U.S. study was found that examined drug presence in on-the-road drivers not involved in crashes, and that study was concerned only with drivers of large trucks. No study assessed drug-crash risk by comparing the drug use of drivers who were just involved in crashes with that of a similar group of drivers who were not just involved in crashes. However, two U.S. studies and one Canadian study used data from driver records and surveys to determine risk factors associated with the use of selected drugs.

Figure 5-1: Percent of Fatally Injured Drivers Tested Positive for Various Drugs -
Reviewed North American Studies

image - click "d" for longdescd

The literature indicates that chemical tests of drivers in crashes were performed most often for narcotics, benzodiazepines, barbiturates, cocaine, amphetamines, and cannabis. The range and means of the percentage of fatally-injured drivers who were positive for these drugs in the North American studies we reviewed are shown above in Figure 5-1. Cannabis had the highest percentages, ranging from 7% to 37% with a mean of 14%. The mean percentages of each of the other five drugs amounted to about 5% or less.

Figure 5-2 depicts how the mean percentages of drug-positive fatally injured drivers in North American studies compares with the mean percentages in foreign studies. Except for narcotics and barbiturates, the foreign studies show lower percentages than North American, considerably lower in the case of cannabis (14% for North America compared to only 2% for foreign). It should be kept in mind, however, that there were only four foreign studies in two countries (two in the United Kingdom, one in Australia, and one in Norway) in our review of drugs in fatally injured drivers, compared to 10 such studies in North America.

Figure 5-2: Mean Percent of Fatally Injured Drivers Tested Positive for Various Drugs - Reviewed North American and Foreign Studies

image - click "d" for longdescd

Figure 5-3 compares the mean percentages of drug-positive fatally injured drivers in North America with the mean percentages of drug-positive non-fatally injured drivers in North America. For most of the drugs, the percentages for non-fatally injured drivers are greater than the percentages for fatally injured drivers. This is opposite the case for alcohol for which the percentage of involvement of fatally-injured drivers is roughly twice that for non-fatally injured drivers.

Figure 5-3: Mean Percent of Drivers Tested Positive for Various Drugs by Type of Injury -
Reviewed North American Studies

image - click "d" for longdescd

We note that all but one of the seven reviewed North American studies of drugs in non-fatally injured drivers involved drivers who had presented at emergency rooms (usually at a trauma center) and had then been admitted to a hospital. The subjects in the other study (Waller, Blow, Maio et al., 1995) were not necessarily admitted to a hospital and were drug-positive only about half as often as those who had been admitted to a hospital. This admissions factor may be related to the very high percentage of drug involvement reported in trauma-center studies.

Few of the reviewed studies (one in Canada and four in other foreign countries) examined the percentages of various drug classes found in non-crash-involved drivers of vehicles of all types who were tested for drugs after being stopped by researchers. Only two drugs were found to be present in more than 1% of the drivers: benzodiazepines (4% in the Canadian study and a mean of 3% in the other foreign studies), and cannabis (5% in the Canadian study). Just one U.S. study dealt with drugs in drivers using the road but not involved in a crash, and its subjects were tractor-trailer truck drivers at one location in Tennessee. That study found that some 30% of the drivers were positive for either marijuana, cocaine, or stimulants.

Drug-crash risk continues to be an unknown quantity. The single recent North American study addressing risk used the responsibility-analysis approach and found no increased fatal-crash risk associated with marijuana or cocaine alone, but a possible association of multiple drug use with increased crash responsibility. An Australian study also using the responsibility analysis approach found that only alcohol had a statistically significant increased risk of fatal-crash responsibility. The relative risk for cannabis (computed as an odds-ratio with p=0.065) was actually less than one, suggesting a beneficial effect of marijuana use. We note also in passing that the percentage of fatally-injured trailer-truck drivers in an eight-state sample who were drug-positive was roughly the same as that found in the Tennessee tractor-trailer truck drivers using the road but not involved in a crash.

In addition, a third, less formal, approach has been used wherein the percentage of a drug from a crash study (or studies) is compared with the percentage of a drug from a non-crash study. Figure 5-4 is a synthesis of these approaches using data from the studies reviewed in this report. The bars labeled "single studies" are for studies using responsibility analysis (the first approach above) and information about drug use (the second approach) to calculate relative risk, and the bars labeled "separate studies" are for risk estimates based on data from separate studies for crash data and for non-crash data (approach three). The risk figures are for all the studies, North American and foreign, having the required data for vehicles and drivers in general, and are averages across studies for any given drug.

Figure 5-4: Relative Crash Risk of Various Drugs by Type of Study Averaged Across Reviewed Studies

image - click "d" for longdescd

First, it is seen that data for the separate studies approach were available only for benzodiazepines and cannabis, and the risk figures for these two drugs were quite close to those obtained from single studies using one of the other two approaches. Second, none of the drugs was associated with very high relative risk, the maximum risk of about 2.0 occurring for benzodiazepines and cannabis, followed closely by narcotics at 1.5. CNS stimulants (including cocaine and amphetamines) were associated with either no increased relative risk (cocaine) or even decreased relative risk (other stimulants).

These figures provide only a rough idea of the magnitude of the drug-crash risk. Case-control studies of the type performed for alcohol-crash risk in the 1960s and 1970s (and one that is now being completed in a NHTSA-sponsored project) are needed for sharper estimates. Such a study for drugs would compare the percent of given drugs in crash-involved drivers with the percent in non-crash involved drivers at the times and places of the crashes. Because of difficulties in obtaining specimens for testing drug presence, such studies have not been conducted to-date, so only rough estimates of the type presented above are the best that can now be provided.

A number of studies have explored the question of drug presence in drivers stopped or arrested for traffic violations. These studies provide another perspective on the drug-crash problem, as such drivers have already been interdicted by police for suspected hazardous driving behavior, usually DWI. Figure 5-5 is a compilation of the results of studies of this type that have been reviewed in this report. The percentages in the figure are averages across studies for the indicated drugs, presented separately for U.S. studies and foreign studies. Except for benzodiazepines, the percentages of drug-positive drivers were about the same in foreign studies as in U.S. studies, ranging from about 13% for barbiturates to 28% for cannabis. Benzodiazepines appeared in an average of 30% of drivers in foreign studies versus 14% in the U.S. studies. Some data were also available for PCP use in the U.S. (more in general use during the earlier studies), indicating an average of 16% for this drug. Only one foreign study (in Switzerland) had data for cocaine use (11%), and the U.S. studies indicated an average of about 16% of the drivers were positive for cocaine. These figures are notably higher than those for non-crashed drivers in general, for whom only cannabis and benzodiazepines were found in percentages exceeding 1%, and neither of those two drugs was found in more than 5% of the drivers.

Figure 5-5: Percent of Drug-Positive Drivers Amoung Drivers Stopped or Arrested for DUID by Country

image - click "d" for longdescd

In sum, recent epidemiologic research indicates that:

Another complicating factor is the role of drugs taken in combination with alcohol. For many drugs, a drug in combination with alcohol accounts for a significant percentage of the occurrences of that drug in crash victims. Waller et al. (1995) found that roughly one-half of the occurrences of drivers positive for marijuana, cocaine, and / or opiates had elevated BACs, and that the crashes of drivers testing positive for drugs alone were very similar to the crashes of drivers testing negative for both alcohol and drugs. This adds further doubts about the role of drugs in the impairment of crash-involved drivers, and suggests that it may be much smaller than had been suspected.