Introduction

Cannabis is the most commonly used illicit drug worldwide, with 2.8–5.0 % of 15–64 year-olds (129–230 million people) consuming cannabis at least once in 2011 [1]. Δ9-Tetrahydrocannabinol (THC), the main psychoactive component of cannabis, was the most prevalent illicit drug detected in injured drivers in Victoria, Australia (9.8 %) [2]. Cannabinoids were found in 8.6 % of nighttime drivers’ blood and/or oral fluid (OF) in the 2007 US Roadside Survey [3].

With changes to drug policy for recreational and medicinal cannabis intake, increased recreational and medicinal cannabis use [1, 4, 5], and reduced perceived risk [3], rapid and accurate on-site cannabinoid testing is required. OF is increasingly used for on-site drug testing in drug treatment, workplace, pain management, and driving under the influence of drugs (DUID) testing. OF advantages include non-invasiveness, observed specimen collection, and lower potential for adulteration compared with urine. Optimum OF on-site devices require rapid analysis times, OF collection volume indicators, objective result interpretation, and quality-control information, for example adequacy of lateral flow. However, different device formulation, including elution buffers, sample collection volume, buffer dilution, cannabinoid antibodies, and detection mechanisms make comparison of performance among devices difficult. Furthermore, manufacturers make frequent device reformulations to improve cannabinoid sensitivity. Many previously available on-site devices failed to meet the 80 % diagnostic sensitivity, specificity, and efficiency criteria proposed by the driving under the influence of drugs, alcohol and medicines (DRUID) project for OF drug detection [6–15]. On-site OF cannabinoid efficiency was 80 % for OraLine IV s.a.t. [8], 57.5–92.8 % for Drugwipe 5 [10–13, 16], 60–94.3 % for Cozart DDSV [14, 12, 17], 66–93.0 % for RapidStat [12, 13, 15–17], 94.3 % for A-Concateno DDS [16], 68–96.1 % for DrugTest 5000 [12, 13, 16, 17], and 78–90.9 % for OrAlert [12, 17]. Although the OraLine IV s.a.t. and A-Concateno DDS devices achieved ≥80 % efficiency, sensitivity was only 69.2 % and 37.8 %, respectively. Overall, the authors concluded that the DrugTest 5000 achieved the best diagnostic efficiency for detection of OF cannabinoids [13, 16–18].

We recently documented high diagnostic sensitivity (90.7 %), specificity (75.0 %), and efficiency (87.9 %) for the DrugTest 5000 with the manufacturer’s 5 μg L−1 screening cutoff and a 2 μg L−1 THC confirmation cutoff [19]. However, that study included only frequent smokers and monitored cannabinoids for only 22 h after smoking, leading to overall high cannabinoid concentrations and too few true negatives to evaluate specificity. Others also evaluated this newest DrugTest 5000 version (5 μg L−1 THC screening cutoff) and reported diagnostic sensitivity, specificity, and efficiency of 53–93, 71–99, and 84–94.3 %, respectively, with OF THC confirmation cutoffs of 1–10 μg L−1 [12, 13, 16, 17]. Recency of smoking was shown to affect diagnostic sensitivity and efficiency [12, 18].

As far as we are aware, no studies have evaluated differences among OF collection devices after controlled smoking and with authentic OF samples, although significantly higher concentrations were found in expectorated OF from Dutch “Coffee Shop” patrons than in samples consecutively collected from the same individuals with the StatSure device [20]. Furthermore, OF THC concentrations in consecutively collected duplicate expectorated samples were much more variable than samples collected with the StatSure device. No significant differences between OF samples simultaneously collected with the Quantisal [21] or Intercept devices [22] were noted.

To address these knowledge gaps, we evaluated performance characteristics and OF detection windows for the on-site DrugTest 5000 and the StatSure and Oral-Eze devices for occasional and frequent cannabis users after smoking of a 6.8 % THC cigarette. THC, 11-nor-9-carboxy-THC (THCCOOH), 11-hydroxy-THC (11-OH-THC), cannabidiol (CBD), and cannabinol (CBN) were quantified by two-dimensional gas chromatography–mass spectrometry (2D-GC–MS) [23] to investigate different cannabinoid markers and windows of cannabinoid detection to meet the objectives of diverse drug-testing programs.

Methods

Participants

Healthy male and female cannabis smokers provided written informed consent to participate in this National Institute on Drug Abuse Intramural Research Program Institutional Review Board-approved study. Individuals were recruited by use of television, radio, and newspaper advertisements, flyers, and participant referrals. Participants received a comprehensive medical and psychological evaluation to verify compliance with eligibility criteria. Inclusion criteria were:

  • ages 18 to 45 years; and

  • self-reported average frequency of cannabis smoked of less than twice per week (occasional smoker) or at least four times per week (frequent smoker) in the past three months.

History of cannabis use was confirmed by a positive urine cannabinoid test for chronic frequent smokers. Exclusion criteria included:

  • breastfeeding or pregnant women;

  • current clinically significant medical condition or history of neurological illness;

  • history of a clinically significant adverse event associated with cannabis intoxication;

  • >450 mL blood donation within 30 days of drug administration;

  • clinically significant anemia;

  • increased systolic or diastolic blood pressure or heart rate >100 bpm after 5 min rest;

  • clinically significant electrocardiogram abnormality; or

  • interest in drug abuse treatment within 60 days of study screening.

Pregnancy tests were administered at screening and on study admission to women with reproductive potential.

Study design

Participants entered the secure research unit approximately 19 h before smoking, to preclude intoxication at the time of cannabis dosing. Participants smoked one (average ± SD) 6.8 ± 0.2 % (54 mg) THC, 0.25 ± 0.08 CBD (2 mg), and 0.21 ± 0.02 % CBN (1.6 mg) cannabis cigarette ad libitum for up to 10 min. OF was collected with the StatSure Saliva Sampler (StatSure Diagnostics Systems, Framingham, MA, USA), the Oral-Eze (Quest Diagnostics, Madison, NJ, USA), and the DrugTest 5000 (Draeger Safety Diagnostics, Lübeck, Germany), in that sequence. The StatSure device consists of an absorptive cellulose pad placed below the tongue, a volume adequacy indicator that turns blue on collection of 1.0 mL OF, and a polypropylene tube containing 1 mL elution/stabilizing buffer, yielding 1/2 OF dilution. The Oral-Eze device has an absorptive cotton pad positioned between the lower cheek and gum (with plastic shield against the cheek), a volume adequacy indicator that turns blue on collection of 1.0 mL OF, and a plastic tube containing 2 mL stabilizing buffer, yielding a 1/3 OF dilution. The DrugTest 5000 test cassette is equipped with a polymeric non-compressible pad for OF collection. OF was collected by swiping the test cassette on the tongue and side of the cheeks. The test cassette collects 270 μL ± 15 % OF, as indicated by the volume adequacy indicator. Oral intake (eating, drinking, cigarette smoking) was prohibited 10 min before OF collection. Samples were collected on admission, 1 h before, and 0.5, 1, 2, 3, 4, 5, 6, 8, 10.5, 13.5, 21, 24, 26, 28, and 30 h after the start of smoking. OF was collected until the volume indicator turned blue or for a maximum of 10 min. StatSure samples were stored upright at 4 °C and Oral-Eze samples were stored horizontally at room temperature in accordance with the manufacturers’ recommendations; all (except 5 Oral-Eze samples analyzed within 96 h) were analyzed within 24 h of collection. Participants remained at the secure residential unit until the end of the study.

Sample analysis

THC, 11-OH-THC, THCCOOH, CBN, and CBD were quantified by 2D-GC–MS in accordance with a method reported elsewhere [23], with minor modifications. For the StatSure, calibrators and quality controls were prepared in 0.25 mL blank OF and 0.25 mL StatSure buffer to take into account OF dilution. For the Oral-Eze, calibrators and quality controls were prepared with 0.25 mL blank OF and 0.5 mL Oral-Eze buffer. The GC column configuration for neutral cannabinoid analysis was changed, with the DB-1MS (Agilent Technologies, Wilmington, DE, USA) column as the primary column and the ZB-50 (Phenomenex, Torrance, CA, USA) as the secondary column. Before loading the first elution solvent, 0.4 mL methanol (StatSure) or hexane (Oral-Eze) was added to the solid-phase extraction columns. Limits of quantification (LOQ) were 0.5 μg L−1 for THC, 11-OH-THC, and CBD, 0.5 μg L−1 (StatSure) or 1 μg L−1 (Oral-Eze) for CBN, and 15 ng L−1 for THCCOOH. For StatSure, the linear range was 0.5–50 μg L−1 (THC, CBD, CBN, and 11-OH-THC) or 10–500 ng L−1 (THCCOOH), and for the Oral-Eze, 0.5–50 μg L−1 (THC, CBD, and 11-OH-THC), 1–50 μg L−1 (CBN), and 15–500 ng L−1 (THCCOOH). Intra-assay imprecision were 1.0–2.7 % (n = 6) and 1.0–4.7 % (n = 6) for the StatSure and Oral-Eze devices, respectively; inter-assay imprecision was <7.6 % for both. OF specimens were diluted with drug-free OF–buffer mixture if analyte concentrations exceeded the upper LOQ.

Data analysis

Qualitative OF DrugTest 5000 cannabinoid results at the preprogrammed 5 μg L−1 THC cutoff were evaluated against quantitative StatSure and Oral-Eze OF 2D-GC–MS results. True positive (TP, DrugTest 5000 and GC–MS positive), true negative (TN, DrugTest 5000 and GC–MS negative), false positive (FP, positive DrugTest 5000, but negative GC–MS) and false negative (FN, negative DrugTest 5000, but positive GC–MS) results were calculated at DrugTest 5000 screening cutoffs of 5 μg L−1 THC and GC–MS THC confirmation cutoffs of 1 μg L−1 (DRUID), and 2 μg L−1 (Substance Abuse and Mental Health Services Administration, SAMHSA), and confirmation cutoffs of THC and/or THCCOOH (20 ng L−1). Sensitivity (100 × (TP/[TP + FN])), specificity (100 × (TN/[TN + FP])), and efficiency (100 × ([TP + TN]/[TP + TN + FP + FN])) were calculated at multiple confirmation cutoffs. Amounts detected and windows of detection were evaluated with the DrugTest 5000 and different confirmation analytes and cutoffs.

The Mann–Whitney U test was used to compare time of last detection between occasional and frequent smokers for different screening and confirmation cutoffs; detection ≥30 h was assigned as 30 h for statistical purposes. The Wilcoxon signed rank test was used to compare Oral-Eze and StatSure concentrations and times of last detection (t last) between devices; a 1 μg L−1 cutoff was used for CBN to keep consistency between devices. Any pair for which the StatSure or Oral-Eze volume adequacy indicator did not turn blue was excluded. All analysis was performed with SPSS Version 20 (IBM, Armonk, NY, USA), with two-tailed p < 0.05 considered significant.

Results

Human participants

Fourteen healthy frequent and 10 occasional smokers (17 men, 7 women), ages 19–41 years, participated in the study (Table 1). Frequent smokers were significantly younger, had smoked for a significantly shorter period of their lifetime, and smoked significantly more recently, more joints or joint-equivalents (empirically-normalized joint consumption, to take into account different smoking methods, i.e. bowl, pipe, blunt) in the last 14 days, and more frequently than occasional smokers. Two participants (M and N) were originally classified as occasional smokers by self-report, but after analysis of multiple biological (blood, urine, and oral fluid) samples were reclassified as chronic frequent smokers.

Table 1 Demographic characteristics and smoking histories of 14 frequent and 10 occasional cannabis smokers
Full size table

Sample analysis

Participant B withdrew from the study after the 26 h sample collection; 28 and 30 h OF samples were not collected. Additional Oral-Eze and DrugTest 5000 samples were missed for participant C at 0.5 h because of an adverse event, and for participant E at 30 h because of a broken Oral-Eze tube. Four-hundred and four DrugTest 5000-StatSure sample pairs and 403 DrugTest 5000-Oral-Eze sample pairs were obtained. Nine DrugTest 5000 samples (2.2 %) produced invalid results, yielding 395 OF pairs for comparison with StatSure and 394 OF pairs for comparison with Oral-Eze. Quantitative cannabinoid disposition in OF collected with the StatSure and Oral-Eze devices has been reported elsewhere [24, 25].

StatSure and Oral-Eze cannabinoid concentration comparison

StatSure and Oral-Eze THC and CBD concentrations were not significantly different between devices (Z = −1.500, p = 0.134 for THC and Z = −1.551, p = 0.121 for CBD) in consecutively collected samples. However, StatSure and Oral-Eze THCCOOH and CBN concentrations differed significantly by device (Z = −11.439, p < 0.001 for THCCOOH and Z = −2.520, p < 0.05 for CBN), with THCCOOH and CBN concentrations usually being higher in Oral-Eze samples. Given the low THCCOOH detection rate among occasional smokers [24, 25], we also compared THCCOOH concentrations between devices for frequent smokers only; concentrations remained significantly different (Z = −10.380, p < 0.001).

Detection rates and windows of detection

StatSure and Oral-Eze OF detection rates, with and without DrugTest 5000 results, for THC ≥1 and ≥2 μg L−1, THCCOOH ≥20 ng L−1, THC ≥2 μg L−1 or THCCOOH ≥20 ng L−1, and THC ≥2 μg L−1 and THCCOOH ≥20 ng L−1 are shown in Figs. 1 and 2. All occasional smokers were negative on admission for all analytes and cutoffs, whereas 85.7–100 % of frequent smokers were positive on admission, depending on the collection device and cutoff. Adding the DrugTest 5000 results generally reduced detection rates (Figs. 1 and 2).

Fig. 1
figure 1

Cannabinoid detection rates for Oral-Eze and StatSure oral fluid collection devices and occasional and frequent smokers after controlled smoking of a 6.8 % ∆9-tetrahydrocannabinol (THC) cannabis cigarette

Full size image
Fig. 2
figure 2

Cannabinoid detection rates for the DrugTest 5000, with Oral-Eze or StatSure oral fluid confirmation, for occasional and frequent smokers after controlled smoking of a 6.8 % ∆9-tetrahydrocannabinol (THC) cannabis cigarette

Full size image

Detection windows were usually shorter for occasional smokers than for frequent smokers, especially when THCCOOH ≥20 ng L−1 was included in the confirmation criteria (Table 2). Three frequent smokers were still DrugTest 5000 positive at 30 h, but had 1, 5, and 5 negative samples before 30 h. Cannabinoid t last values were not significantly different between devices, except for THC ≥2 μg L−1 (Table 3).

Table 2 Median (range) time of last cannabinoid detection by screening and confirmation tests for 14 frequent and 10 occasional cannabis smokers after controlled smoking of 6.8 % ∆9-tetrahydrocannabinol (THC) cigarette
Full size table
Table 3 Median (range) times of last detection for Oral-Eze and StatSure oral fluid collection devices after controlled smoking of a 6.8 % ∆9-tetrahydrocannabinol (THC) cigarette by 14 frequent and 10 occasional cannabis smokers
Full size table

DrugTest 5000 and confirmation comparison

Diagnostic sensitivity, specificity and efficiency for multiple confirmation cutoffs are shown in Table 4. When confirming for THC only, diagnostic sensitivity was 5.7–11.0 percentage points higher for frequent as compared with occasional smokers, because of longer detection windows and higher true-positive rates (e.g., 70.6 % vs. 59.6 % for frequent and occasional smokers, respectively, at the 1 μg L−1 Oral-Eze cutoff). Similarly, increased diagnostic sensitivity, specificity, and efficiency were usually documented when considering only 6 or 8 h post-smoking, because of higher concentrations and higher true positive rates for occasional and frequent smokers.

Table 4 Performance characteristics for the Draeger DrugTest 5000 on-site test with a 5 μg L−1 Δ9-tetrahydrocannabinol (THC) screening cutoff in oral fluid with different confirmation cutoffs
Full size table

Discussion

As far as we are aware, no other studies have compared cannabinoid concentrations among OF collection devices from authentic OF samples after controlled cannabis smoking. In addition, we evaluated detection windows for one screening and two confirmation OF collection devices for cannabinoids in OF from occasional and frequent smokers.

THCCOOH and CBN concentrations were significantly higher for Oral-Eze devices. StatSure samples were collected immediately before Oral-Eze samples. Perhaps the presence of the StatSure device in the mouth stimulated OF production, but it is unclear how this could have resulted in higher THCCOOH and CBN concentrations with the Oral-Eze device. Although both devices theoretically collect 1 mL, actual OF volume collected might vary. Previous work documented (on the basis of weight) a mean (RSD) of 0.952 mL (11.97 %) OF for OF collected with the StatSure device [26]; no data are available for the Oral-Eze device. Differences in collected OF volume could potentially explain the differences in THCCOOH and CBN concentrations. The low CBD prevalence and high THC variability immediately after smoking may contribute to the lack of significant differences for these analytes. Differences in cannabinoid recovery from the pad do not explain differences in concentrations. For the StatSure device, cannabinoid recoveries from the pad were 65.5–68.1 %, 62.2–65.9 %, 71.3–73.9 %, and 65.1–71.2 % for THC, CBD, CBN, and THCCOOH, respectively. For the Oral-Eze device, cannabinoid recoveries from the pad were 42.5–48.8 %, 33.5–47.7 %, 35.6–58.7 %, and 68.1–86.2 % for THC, CBD, CBN, and THCCOOH, respectively [25]. The lower cannabinoid recoveries for THC, CBD, and CBN from the Oral-Eze pads are inconsistent with higher Oral-Eze concentrations. Other potential explanations for the different concentrations are placement of the device in the mouth (under the tongue for the StatSure and between the gum and the cheek for the Oral-Eze), different pad composition and thickness, different elution/stabilizing buffer, and sequential (rather than simultaneous) OF collections. Concentration differences because of analyte instability are not expected, because all but five Oral-Eze samples were analyzed within 24 h of collection.

As far as we are aware, no other study has evaluated differences in authentic cannabinoid OF concentrations among collection devices after cannabis smoking. Langel et al. evaluated analyte recovery from fortified samples and collected OF volume from eight OF devices; however, concentrations from authentic OF samples were not evaluated [26]. One study compared expectorated and StatSure OF samples and reported high concentration differences between the two collection methods [20]. In general, expectorated THC concentrations were six times higher than StatSure concentrations, and consecutively collected expectorated samples varied much more in concentration than consecutively collected StatSure samples. The authors suggest that THC adsorption by cell debris in the OF could explain higher concentrations in expectorated OF, and that inhomogeneous THC distribution in the mouth could explain the higher variability in the expectorated OF compared with the StatSure (which is always placed under the tongue). Others documented no significant differences between simultaneously collected OF with the Quantisal [21] or Intercept devices [22].

Given the higher THCCOOH concentrations in the Oral-Eze samples, it is not surprising that THCCOOH detection rates are higher with this device (Figs. 1 and 2). THC ≥1 μg L−1 or THCCOOH ≥20 ng L−1 provided the longest detection times in occasional and frequent smokers (Oral-Eze and StatSure). Because of low THCCOOH prevalence in occasional smokers, requiring a positive THCCOOH ≥20 ng L−1 (e.g. THC ≥1 μg L−1 and THCCOOH ≥20 ng L−1) greatly reduced detection rates in this cohort for the Oral-Eze and StatSure devices (Fig. 1). We previously reported that higher THCCOOH OF concentrations in frequent smokers were because of their frequent past self-administered cannabis intake and high THC body burden [24].

No differences between t last were observed between collection devices, except for THC ≥2 μg L−1. In this case, StatSure OF samples had a later t last. It is unclear why there was a significant difference for this cutoff, because Oral-Eze concentrations were not significantly different. However, for one participant (participant K), THC concentrations were just above the 2 μg L−1 cutoff with the StatSure device but just below the 2 μg L−1 cutoff with the Oral-Eze between 8 and 21 h post smoking, possibly explaining the significantly later t last for StatSure samples. Frequent smokers’ t last were generally later; had we captured the true t last and not had to assign ≥30 h assigned as 30 h for statistical purposes, group differences could have varied from those reported here.

When considering screening and THC confirmation tests, the DrugTest 5000 was usually the limiting factor in the positivity rate because of its higher 5 μg L−1 cutoff compared with GC–MS 1 and 2 μg L−1 cutoffs (Fig. 2). The only exception was for occasional smokers when THCCOOH ≥20 ng L−1 was required for positive confirmation, because of low THCCOOH OF prevalence in this group. These data are consistent with our previously reported evaluation of the DrugTest 5000 for chronic frequent cannabis smokers [19]. Similarly, detection times were reduced with the DrugTest 5000 (Table 2), and differences between occasional and frequent smokers’ t last were reduced except when THCCOOH was a required confirmation analyte.

Diagnostic sensitivity and efficiency after controlled smoking of cannabis were slightly lower than previously reported; however, diagnostic specificity was higher [19]. This can be explained by the higher proportion of true negatives in this study compared with our previous study. Others reported diagnostic sensitivity, specificity, and efficiency of 53–93, 71–99, and 84–94.3 %, respectively (confirmation cutoffs of 1–10 μg L−1 OF THC) [12, 13, 16, 17]. Our data are consistent with our previously published data, indicating that the highest diagnostic efficiency was achieved with a 2 μg L−1 THC confirmation cutoff. Sensitivity was lower for occasional smokers than for frequent smokers (when considering only THC for confirmation), because of the generally higher concentrations in frequent smokers’ OF at the later time points. This resulted in a lower incidence of true positives for occasional smokers. Similarly, diagnostic sensitivity, specificity, and efficiency improved for occasional and frequent smokers when considering samples collected up to 6 or 8 h post dose, because of the higher incidence of true positive tests. This is consistent with two previous studies that documented higher sensitivity for samples collected recently after smoking [18, 12].

Requiring THCCOOH (THC and THCCOOH statement) for confirmation generally reduced diagnostic specificity and efficiency, as it increased the number of DrugTest 5000 false positives, especially for frequent smokers. Because of the low LOQ for THCCOOH and high cannabinoid body burden in frequent smokers, THCCOOH remained positive when THC was no longer detected in OF, increasing the detection rate and window of detection [27].

Conclusion

We document, for the first time, OF cannabinoid disposition in occasional and chronic frequent cannabis smokers with an on-site screening device and with two OF collection devices after controlled smoking of cannabis. We report significantly different THCCOOH and cannabinol, but not THC, concentrations between OF collection devices, which may affect OF data interpretation. The DrugTest 5000 on-site device has high diagnostic sensitivity, specificity, and efficiency for OF cannabinoid detection.