There is increasing concern about the potential traffic safety risks associated with driving under the influence of drugs (DUID). This concern corresponds with the rising prevalence of drug impaired driving in Canada and other nations (Grotenhermen, Leson, Berghaus, Drummer, Kruger, Longo, Moskowitz, Perrine, Ramaekers, Smiley, and Tunbridge 2007). The Canadian government has recently passed new legislation to strengthen the laws against DUID. On 2 July 2008, Bill C-2 came into force. This legislation amends the impaired driving provision of the Criminal Code, by requiring drivers suspected of being under the influence of drugs to submit to a series of tests to determine impairment. Prior to Bill C-2, drivers suspected of DUID had the right to refuse to be examined for drug impairment. While amendments to the Criminal Code have now taken effect, there is still debate in Canada over (1) at what level drugs impair the functions necessary for safe driving, (2) whether drivers under the influence of drugs are at an increased risk of motor collision and injury, and (3) the effectiveness of the different techniques used to identify drug impairment in motorists (Asbridge 2006).
Cannabis is the second most commonly used intoxicant in Canada, next to alcohol (Adlaf, Begin, and Sawka 2005), and research indicates that the prevalence of driving after consuming cannabis is increasing (Asbridge 2006). Over the past 50 years, evidence has established a direct relationship between increasing levels of blood alcohol concentrations (BAC) and increased risk of motor vehicle accidents (Walsh, Verstraete, Christophersen, Mercier-Guyon, Kintz, Oliver, Moeller, Compton, Sweedler, Potter, and de Gier 2000). As a result, appropriate instruments have been developed to enable police to identify drivers impaired by alcohol and to secure convictions for driving under the influence (DUI). Devices used to estimate BAC from a breath sample, such as the breathalyzer, have become increasingly sophisticated and reliable and are thus widely accepted by the courts. However, no such limits or devices exist for screening drivers for cannabis impairment. Thus, there is a need to develop a reliable method of determining cannabis impairment in motorists that will stand up in court. This paper will attempt to address the third point in this debate, effectiveness, through a review of scientific studies that have evaluated the following methods, used to detect cannabis impairment in drivers: the Drug Evaluation Classification Program (DEC) / Drug Recognition Expert (DRE), on-site oral fluid screening devices, and on-site urine screening devices.
A literature search was conducted with regard to the use and evaluation of techniques and devices used to screen on-site for cannabis impairment in motorists. As indicated above, the studies were then broken down into three categories. Only studies which reported the necessary measures of accuracy (sensitivity, specificity, and accuracy), or in which sufficient data was available to calculate these measures, were included in the present review.
Measures of accuracy
The number of cannabis positive cases identified by the given technique or device are known as true positive (TP). Factual negative results are classified true negative (TN). Discrepancies between the method of detection (i.e., DEC, oral fluid testing device) and the toxicology confirmation results are labelled either false positive (FP) if impairment is incorrectly identified or false negative (FN) if impairment is missed by the technique. The first measure of accuracy used in this review is sensitivity or the hit rate. This measure addresses the likelihood that a driver who has consumed cannabis will be detected by the screening method. Sensitivity is calculated by dividing the number of cannabis impaired cases that the method or device identifies (TP) by the number of cannabis positive cases identified by the toxicology (TP+FN). This measure is important, as tests with high sensitivity minimize the number of false negatives (FN); that is, drivers who have consumed cannabis but who go undetected.
The second measure of accuracy is specificity, which refers to the number of correctly identified cannabis negative cases. Specificity is determined by dividing the number of TN cases by the total number of drug negative cases identified by the toxicology (TN+FP). Tests with high sensitivity minimize the number of drivers who are incorrectly identified as having consumed cannabis. The final measure used in this review is accuracy. Accuracy represents the proportion of cases that have been correctly identified as either hits or rejections. Accuracy is calculated as follows: (TP+TN)/(TP+FP+TN+FN). This measure captures the overall performance of a given procedure (Beirness, Beasley, and LeCavalier 2008).
Drug Evaluation Classification Program / Drug Recognition Expert (2)
The first method of detecting DUID to be examined is the DEC program and the associated DRE police officers. DREs are police officers who have been specially trained and certified to identify drug impairment in suspected motorists. The DREs utilize a procedure that relies on the observation of the suspected driver's socio-behavioural cues, biological and vital signs, and direct questioning (Asbridge 2006). Based on the information gathered, the DRE forms an opinion as to whether the suspected motorist is impaired, and if so, by what class of drug (i.e., CNS depressants, hallucinogens, cannabis). First, the suspected motorist is administered a standardized field sobriety test (SFST), much like the test for alcohol. If the motorist fails the SFST and is deemed by the police officer not to be impaired by alcohol, s/he is evaluated by a DRE. The DRE evaluation can take place either at the roadside or at the police station. If upon examination, the DRE believes the motorist is impaired by a drug, then a bodily fluid sample (i.e., blood) is taken and submitted to a laboratory for toxicology testing. The DEC program, developed in the late 1970s by the Los Angeles Police Department, is currently in use across the United States, as well as in Europe, Australia, and Canada. Included in the recent measures taken to tackle DUID was funding to increase the number of certified DREs in Canada to help enforce the newly strengthened impaired driving laws.
The DRE studies included in this review have been separated into two categories--laboratory studies and field studies (enforcement studies). Both methods have strengths and weaknesses. Laboratory studies involve systematic investigations, conducted in a highly controlled environment, with volunteer research participants who are administered measured doses of cannabis. Field studies, on the other hand, involve the examination of information gathered in enforcement settings, providing a more realistic research environment. The purpose of both types of studies is to evaluate the ability of DREs to detect signs of drug impairment.
Bigelow, Bickel, Roache, Liebson, and Nowowieski 1985
The first laboratory evaluation of the DEC program was conducted by Bigelow et al. (1985) at the John Hopkins University School of Medicine. In this evaluation, 80 volunteers were randomly assigned to one of eight categories of drugs (d-amphetamine, 15 mg; d-amphetamine, 30 mg; marijuana, 12 puffs of 1.3% THC; marijuana, 12 puffs of 2.8% THC; diazepam, 15 mg; diazepam, 30 mg; secobarbital, 300 mg; or a placebo). Each volunteer was then examined by four DREs from the LAPD, for a total of 320 assessments (80 research participants examined by 4 DREs). The DREs were notified that some subjects would receive a placebo (control) and that no subjects would be administered alcohol, PCP, LSD, or any combination of drugs. The results for cannabis, as found by Bigelow et al. (1985) are reported in Table 1. It is evident from the results that the DREs in this study were able to identify research subjects who had been administered cannabis about half the time (48.8%). The DREs were much better at determining that a research subject had not consumed cannabis (92.7%), yielding an accuracy rate of (63.6%). Thus, while the DREs were quite capable of identifying participants who had not been administered cannabis, over half of those who had been were not detected by the DREs.
Heishman, Singleton, and Crouch 1996
In a subsequent laboratory study, conducted by Heishman et al. (1996), 18 drug-using volunteers were recruited to participate in nine experimental sessions. In each session, the participants received either a placebo, or a high or low dose of ethanol, cocaine, or cannabis. A total of 162 experimental studies were conducted, of which 4 were excluded because cannabis could not be detected in the confirmatory toxicology samples. Twenty-nine certified DREs were recruited to evaluate participants. There was no interview component, as in the actual DEC, but the DREs were informed that the participants might have been administered ethanol, and/or CNS depressants, CNS stimulants, phencyclidine, narcotic analgesic, cannabis, or a placebo. In actuality, with the exception of the placebo, only one drug was administered to the research subjects in each session. The results of the study are presented in Table 1. Again, just over half (53.1%) of the cannabis cases were correctly identified by the DREs, meaning that almost half of those who had consumed cannabis were not detected by the DREs. In this case, the DREs were less able to rule out cannabis consumption in the research subjects.
Heishman, Singleton, and Crouch 1998
A second study by Heishman et al. (1998) evaluated the accuracy of the DEC in identifying four types of drug use. The 12 research participants were administered a dose of a CNS depressant, a CNS stimulant, a narcotic analgesic, or cannabis. In each session, the participants were given either a placebo, a low dose, or a...