Oral fluid has gained increasing interest as a valuable matrix for roadside drug testing. In contrast to urine, the advantages include ease of use, noninvasiveness, observable sample collection, difficulty to adulterate, and demonstration of recent drug use. However, at present many immunoassays for detecting Δ⁹-tetrahydrocannabinol (THC) in oral fluid do not have high reliability (Bosker and Huestis 2009; Verstraete 2005). This is mainly due to THC adsorption to the collection device, which makes recovery from the device difficult. In addition, currently used cutoffs of point-of-collection testing devices, although improving, are too high to reliably detect THC in oral fluid. Furthermore, the time window for detection is unknown, and this makes interpretation of the results of oral fluid collection devices difficult. Therefore, controlled drug administration studies are needed (Bosker and Huestis 2009).

Standardized Field Sobriety Tests (SFST) are currently in use in countries such as the USA and Canada for documenting impairment in drivers suspected of driving under the influence. The SFST have been validated for detection of alcohol impairment (Stuster 2006; Stuster and Burns 1998), but their sensitivity to impairment caused by other drugs is relatively unknown. Placebo-controlled studies assessing the effects of single doses of THC, alcohol, and their combination on SFST performance (Papafotiou et al. 2005a; Stough et al. 2006) have reported that performance in the SFST was significantly related to the administration of THC only, and THC with alcohol. The proportion of individuals classified as impaired doubled after THC combined with alcohol. The authors also reported that the use of SFST resulted in the correct classification of up to 73.9 % of participants as either impaired or not. The one-leg stand test of the SFST was found to be the best predictor of impairment. Performance on the SFST however only weakly predicted driving impairment as assessed in a driving simulator (Papafotiou et al. 2005b). SFST also failed to predict actual driving impairment in occasional and heavy cannabis users who received single doses of the synthetic cannabinoid dronabinol. Single doses of dronabinol 10 and 20 mg produced driving impairment comparable to blood alcohol concentration (BAC) levels >0.8 mg/mL as measured in a standardized on-the-road driving test, but went undetected in SFST assessments (Bosker et al. 2012).

The SFST was mildly sensitive to the effects of cannabis alone. A dose of 400 μg/kg body weight THC significantly increased the percentage of participants displaying impairments in OLS compared to baseline performance from 21 to 50 %. THC also increased percentage of individuals showing impairment on HGN from 0 to 15 %, relative to baseline, but this change only approached statistical significance. WAT and the overall score on SFST did not discriminate between THC and baseline. These findings appear in line with previous studies that have reported a relation between impairment on the SFST and presence of THC in blood. A study that assessed which signs of the Drug Evaluation and Classification evaluations predicted various drug categories (including cannabis) at best showed that OLS contributed significantly to the prediction, but HGN and WAT did not (Porath-Waller et al. 2009). Papafotiou et al. (2005a) assessed SFST performance in 40 healthy participants who received low and high doses of THC in a placebo-controlled study. On average, blood THC concentrations obtained after the highest dose were comparable to serum THC concentrations achieved in the present study after smoking cannabis. Yet, THC significantly affected performance on OLS, HGN, and WAT and appeared to be more prominent as compared to the current study. For example, in that study THC produced impairments on overall SFST performance in up to 50 % of the participants (Papafotiou et al. 2005a) but in only 30 % of the participants of the present study. These differences may be explained in terms of differences in cannabis use history. In the study by Papafotiou et al. (2005a), the reported frequency of cannabis use of the participants varied from once a week to once every 2–6 months. The present study however only included heavy cannabis users, who smoked cannabis on at least four occasions per week. Previous studies demonstrated that heavy cannabis users develop tolerance to the impairing effects of THC on neurocognitive measures (Hart et al. 2001; Ramaekers et al. 2011). It is likely that many of the participants who participated in the present study, in part or in total, developed tolerance to the impairing effects of THC as well. In such a scenario, the failure of the SFST to demonstrate robust effects of THC is not necessarily an indicator of poor sensitivity, but may reflect the chronic cannabis use of the participants.

Alternatively, one might argue that SFST were conducted too late after cannabis administration. SFST were not performed directly after smoking when impairments or THC concentrations can be expected to be maximal. Instead, SFST were performed 2 h after smoking in the present study, when THC impairments are on the decline, as shown in occasional users. It should be noted however that performance impairment has repeatedly been shown to last for 3–4 h after smoking THC (e.g., Ramaekers et al. 2009, 2006a). In the present study, ratings of subjective high were also significantly elevated at 2 h after smoking as reported elsewhere (Ramaekers et al. 2011). Likewise, average THC concentrations during SFST performance were above THC threshold levels above which performance impairments are expected to appear in occasional users (Ramaekers et al. 2006b). In other words, the SFST were conducted well within the established “impairment window” of 3–4 h post-smoking, even though the level of impairment was submaximal. The relative lack of sensitivity of SFST for cannabis effects in the present study thus cannot be explained by a total lack of cannabis intoxication at the time of testing. Still, it cannot be excluded that SFST might have been more sensitive to the effects of THC if conducted right after smoking.

In general, the present data indicate that SFST were mildly sensitive to the effects of THC depending on dose and cannabis use history. It is noteworthy, however, that SFST were unable to discriminate performance impairments produced by dronabinol, a synthetic cannabinoid (Bosker et al. 2012). In this placebo-controlled study, occasional and heavy users received oral doses of dronabinol that produced significant on-the-road driving impairments that were comparable to those observed after BACs >0.8 mg/mL. THC concentrations after oral dronabinol however were much lower (<10 ng/mL) than those achieved after smoking cannabis. Also, THC concentrations during SFST testing were half that in the dronabinol study compared to the present study. This fits with the general conclusion that SFST are mildly sensitive to the impairing effects of THC, but that impairments may go undetected in some individuals, particularly at lower THC concentrations.

THC with alcohol generally increased the number of individuals displaying impairment on HGN, OLS, and total SFST score. Relative to baseline, percentages of impaired individuals increased after both alcohol combinations with THC in a dose-dependent manner. HGN was the only measure that did not reveal any impairment in participants during baseline. It should be remembered here that all participants included were heavy cannabis users who were always positive for THC, also during baseline. On average, baseline THC levels were 7.1 ng/mL. HGN may thus be an interesting parameter to separate residual THC use from recent THC intoxication when OLS and overall SFST performance show impairment. In general, impairments observed after the combination of THC and alcohol are most likely attributable to alcohol since most of the current participants may have developed tolerance to the impairing effects of THC on performance. As such, the present data confirm the sensitivity of the SFST for alcohol-induced impairment.

The point of collection testing devices provided mixed results. The Dräger Drug Test® 5000 generally performed quite well and only produced false negative in 6 % of the measurements during 2.5 h after smoking. Securetec Drugwipe® 5 however performed poorly in detecting THC. The rate of false negatives was already 8 % 15 min after smoking and rapidly increased to about 40–50 % within the hour. The difference in sensitivity between both devices may well be related to differences in cutoff levels which are 5 ng/mL in the Dräger Drug Test® 5000 and 30 ng/mL in the Securetec Drugwipe® 5 according to the manufacturers. It should be noted that in this study we did not determine the rate of false positives for both devices because our participants were never drug free. Even during baseline, low levels of THC were present in all participants. A recent roadside study reported about 2 % false positives for the Dräger Drug Test® 5000 and about 3 % false negatives. The sensitivity, specificity, and accuracy were 93, 71, and 90 %, respectively (Wille et al. 2010). Together, these data indicate that point of collection testing devices for detecting THC in oral fluid are making large improvements relative to some of the previous devices that were on the market (Bosker and Huestis 2009; Ramaekers et al. 2006b). Still it has been argued in the ROSITA-2 project that the sensitivity and specificity of oral fluid devices should be higher than 90 % (Verstraete and Raes 2006). According to these criteria, the Dräger Drug Test® 5000 would still fall short on specificity (Wille et al. 2009). However, compared to the alternative matrix urine, oral fluid drug concentrations have a better correlation with blood concentrations and recent use of cannabis. It is therefore still the preferable matrix for evaluating THC presence in drivers (Bosker and Huestis 2009; Toennes et al. 2005).