Can a Teen Driver Deny Cannabis Use Yet Test THC-Positive? What 718 Roadside Surveys Show

Answer: Yes. In a 2013–2014 U.S. roadside survey of 718 drivers aged 16–19 years, 28 participants who denied using marijuana during the previous year nonetheless had a positive laboratory test for THC-related analytes. That disagreement matters when assessing a statement about drug use, but it does not, by itself, establish dishonesty, the time or dose of consumption, driving impairment, or responsibility for a collision. The study also compared oral-fluid testing with blood testing in a subset, finding 17 disagreements among 305 paired samples. A positive test and a disputed account answer different forensic questions.

The original research by Pressley, Arora, and Sarmah was published in Injury Epidemiology in 2019: Marijuana use in U.S. teen drivers: a comparison of a road-side survey of reported use and fluid tests for tetrahydrocannabinol (THC). Its value lies in directly comparing anonymous self-reports with laboratory results while showing why sample selection, different biological matrices, and the wording of a recall question limit individual-case conclusions.

What did the investigators actually measure?

The investigators used the 2013–2014 National Roadside Survey, which recruited voluntary, compensated, noncommercial drivers at roadside locations across 24 states. The larger survey included 11,100 eligible drivers aged 16 years or older, of whom 8,825 completed survey procedures. The teen analysis concerned 718 drivers aged 16–19, not 8,825 teen drivers and not a national census of adolescent driving.

Of those 718 teenagers, 666 supplied at least one usable oral-fluid or blood test. Ninety-one of the 666 tested positive, yielding the reported 13.7% sample positivity. The remaining 52 teens had no qualifying biological result; they cannot simply be categorized as laboratory-negative. The authors presented unweighted analyses. Even though the underlying roadside program used a multistage sampling design, these published percentages are results for the analyzed participants, not population-weighted estimates for all U.S. teen drivers.

Participants answered questions about how recently they had used marijuana, tobacco, and alcohol. Those statements were compared with positive or negative cannabinoid results at the roadside stop. Investigators also assessed associations with age, sex, season, household income, passenger carriage, and other characteristics. The study did not administer field sobriety tests, measure lane keeping, reconstruct crashes, or observe a particular driver at the time of an alleged offense.

Why do the denominators change from 718 to 666 to 305?

The three denominators refer to different groups, not contradictory counts of independent subjects. The 718 were teen survey participants; 666 had at least one fluid result; and 305 supplied both oral fluid and blood that could be compared as paired results. A denominator suitable for one question cannot be borrowed for another. Using 91/718 as though every participant was tested would silently treat missing specimens as negative and understate measured positivity among those actually tested.

The paired 305 are especially important because agreement between two matrices can only be evaluated when both were collected from the same person. The authors reported 17 discordant pairs, or 5.6%, and Cohen’s kappa of 0.78 with a 95% confidence interval of 0.69–0.88. Neither statistic supplies a complete directional false-positive/false-negative table. It would be incorrect to label all 17 discordances oral-fluid false negatives or to calculate a diagnostic sensitivity that the paper did not demonstrate.

What exactly did the oral-fluid and blood procedures identify?

Laboratory methods included an ELISA screening step followed by gas chromatography–mass spectrometry or liquid chromatography–mass spectrometry confirmation. The study reported screening thresholds of 2 ng/mL for oral fluid and 1 ng/mL for blood. These are matrix-specific analytical decisions. They are not interchangeable concentrations, equivalent systemic exposures, impairment cutoffs, or proof of a particular smoked or ingested dose.

Ages 16–17 were eligible for oral-fluid collection, but blood was not collected from those younger than 18. Ages 18–19 could provide both matrices. The study classified a teenager as positive when either available test was positive. This structure creates an age-related difference in the opportunity to detect a positive result. It is a methodological feature of the evidence, not proof that the older group necessarily had a greater biologically comparable exposure.

The paper refers to laboratory THC-related detection in its roadside protocol. It does not authorize converting an oral-fluid result into an individual forensic venous whole-blood THC concentration, substituting serum or plasma measurements, or inferring an unmeasured urinary THC-COOH concentration. The distinction among biological matrices is therefore central: the chemical information a specimen can support depends on where and how it was collected, what was actually assayed, and the laboratory’s analytical method.

Did a positive result prove that the teenager lied?

No. The source describes 28 teenagers who denied marijuana use during the previous year and nevertheless tested positive. That is a genuine discordance between a categorical survey answer and a laboratory result. The investigators did not establish why any one answer conflicted with testing. A refusal to disclose, uncertainty about the reference period, recall error, differences in the meaning of “use,” misunderstanding of the question, or a test-related limitation cannot be distinguished for each individual from the published group data alone.

The issue is even clearer with shorter recall windows. Among THC-positive teenagers, the authors reported that nearly half did not report using marijuana within the previous month. A statement about “last month” should not be silently rewritten as “never used,” and a laboratory-positive result should not automatically be presented as a contemporaneous confession. Whether an account is credible is a separate evidence assessment involving the exact question asked, verbatim response, chain of custody, confirmatory chemistry, and all available case circumstances.

The authors also observed that among 45 teenagers reporting use in the previous 24 hours, 71.1% tested positive. This illustrates partial overlap between reported timing and analytical positivity, not a universal detection rate. The study did not administer known doses and sample every participant repeatedly to derive a pharmacokinetic curve; it therefore cannot establish when a particular defendant last consumed cannabis.

Why might oral fluid and blood disagree in the same driver?

Oral fluid and blood are physically distinct specimens collected in different ways and examined at different decision thresholds. A matrix-specific positive result can depend on the analyte, collection conditions, analytical sensitivity, confirmation method, and biological timing. The paper reports the observed 17 discordances, but it does not assign a validated mechanistic explanation to each pair. Reconstructing a dose or exposure history from those disagreements would go beyond what the study measured.

For a forensic report, the proper first steps are to identify which specimens actually exist, the collection times and handling, the screening and confirmation records, the measured concentrations if reported, and any quality-control limitations. Specimen collection and handling can matter independently from any later statistical comparison. The absence of a blood result for a 16- or 17-year-old is a feature of this protocol, not a demonstrated negative blood THC finding.

Study-wide kappa summarizes agreement after the authors’ binary classification choices. It cannot decide which result is chemically correct in an individual discordant pair and cannot establish a false-negative frequency against an independent diagnostic gold standard that was not supplied. Confirmation with a chemically selective method addresses compound identification but does not transform compound presence into observed impairment.

Did older teenagers test positive more often?

In the unadjusted comparison, 15.3% of 18–19-year-olds had a positive result compared with 7.7% of 16–17-year-olds; the paper reported p = .019. That is an observed difference in the study’s composite test outcome. It should be read beside the testing design: only the older teenagers had an opportunity for blood testing in addition to oral fluid, and the authors identified nine older teenagers through blood testing who would otherwise have been missed.

In the multivariable Model 3, the age-group adjusted odds ratio was 1.21 with a 95% confidence interval from 0.57 to 2.57. Because that interval includes one, the adjusted model did not establish a statistically significant age-group association. Likewise, an unadjusted male odds ratio of 1.66 (1.02–2.71) became an adjusted estimate of 1.58 (0.88–2.80), which did not meet that significance criterion. Reporting the unadjusted findings without the adjusted null results would give an incomplete picture.

Which predictors remained associated with a positive test?

The investigators fitted regression models to assess associations, not to determine causes. In their Model 3, self-reported marijuana use in the previous year, tobacco use, reported first alcohol consumption at age 15 or younger, and surveys performed during school months were each associated with higher odds of a positive composite fluid result. Those estimates compare groups after adjustment for specified covariates. They do not show that tobacco caused a THC-positive specimen or that school attendance altered drug elimination.

The table below reproduces selected original estimates without presenting odds ratios as probabilities or individualized diagnostic scores. An odds ratio of 5.28 does not mean that 5.28 out of 10 individuals would test positive. Also, the seasonal comparison describes when the roadside encounter occurred; it cannot establish a biological seasonal mechanism or transfer unchanged to a new jurisdiction.

What about teenagers reporting no cannabis use in the prior month?

The study separately examined drivers who said they had not used marijuana recently, treating the previous month as a relevant survey window. In that restricted analysis, a positive test was associated with survey collection during September–March (adjusted odds ratio 3.04, 95% CI 1.28–7.24), tobacco use in the prior 24 hours (3.49, 1.67–7.27), and a passenger being present (2.42, 1.20–4.89). The reported odds ratios for certain state medical-marijuana-law categories also had wide intervals, including 5.00 (1.20–20.76) for a particular reference comparison.

Those characteristics are correlates of a result within a selected subpopulation. A passenger in the car is not evidence that the driver knowingly misstated drug use. A state-law category is not proof of a policy effect without a design that isolates the policy from other differences. The broad confidence intervals also warn against treating subgroup point estimates as precise individualized probabilities. A survey-defined “nonuser in the past month” group is not interchangeable with a lifetime abstainer group.

Why a toxicology-positive roadside sample is not an impairment determination

A valid chemical identification may establish that a target analyte was detected in a specified specimen subject to the method and laboratory record. It does not, without more, prove psychomotor deficits at the time of a driving event. Pressley and colleagues collected no validated performance battery that could connect individual concentrations to reaction times, lane weaving, divided attention, or collision behavior. The study supplies no measured driving-error denominator, no case-specific crash findings, and no impairment sensitivity or specificity.

Even the 305 paired matrices are an agreement study, not a demonstration of fitness to drive. Confirmatory laboratory methods can reduce some analytical identification concerns, but confirmation of drug identity answers a different question from the effect of a compound on a particular person’s abilities. No serum, plasma, urine, or mathematical blood-equivalent value should be retrofitted into the source to create a missing impairment endpoint.

For an actual case, an expert may have to integrate the precise specimen result with collection time, validated method performance, observed driving, scene evidence, clinical observations, concurrent substances, and alternative explanations. This original study cannot replace that analysis. It is informative evidence about discordance in a teen roadside sample, not a diagnosis of one driver’s cognitive state.

What do the results establish about national prevalence and cannabis laws?

The original study’s unweighted 13.7% positivity among tested volunteers is not a representative national teen-driving prevalence estimate. Roadside participation was voluntary and anonymous, site recruitment occurred under particular sampling constraints, and no universal blood specimen was available for younger teens. Small subgroups and missing specimen data limit generalization. The authors considered features of state medical-marijuana laws, but the design was observational, cross-sectional at the roadside, and lacked a pre/post causal policy comparison for these individual drivers.

Accordingly, the finding cannot be used to say that legalization caused 28 teenagers to misreport, that any state policy caused a driver to be THC positive, or that 13.7% of American teenage drivers were impaired. A distinct question about changes over time would require appropriate population denominators, comparable testing protocols, and a causal design. This source does not supply them.

How to analyze this evidence in a deposition or expert report

A disciplined interpretation begins by separating four propositions: what the participant said, what specimen was actually obtained, what the laboratory identified, and what the relevant legal or behavioral issue requires. The first proposition depends on the exact timeframe of the survey question. The second distinguishes oral fluid alone from paired oral fluid and blood. The third requires screening and confirmatory documentation. The fourth may concern impairment, consumption timing, credibility, or causation; the preceding three do not automatically resolve it.

When questioning an expert, ask for the denominator used in any quoted percentage and whether the opinion treats 718 survey participants as 666 tested drivers. Ask whether under-18 participants could give blood, whether a claim of test disagreement is derived from the original paired specimens, and whether the expert has converted a non-significant adjusted association into a firm individual conclusion. Ask what direct behavioral evidence, if any, links the observed chemical finding to the disputed driving event. These are limitations inherent in the experimental record, not alternative findings invented from it.

Most importantly, a discrepancy between an account and a test is evidence to interpret, not a scientifically established motive for that account. The original investigators did not interview the 28 discordant teens to adjudicate intent. Any claim that they deliberately lied or were impaired rests on evidence outside this study and must be identified as such.

Seven frequently asked forensic questions

1. Did 13.7% of all 718 surveyed teenagers test positive?

No. The source reported 91 positive results among 666 teenagers who supplied oral fluid and/or blood. Fifty-two surveyed teenagers did not have a qualifying laboratory result. The 13.7% describes the tested subset, not all 718 or the national teen-driver population.

2. Does denying cannabis use in the previous year and testing positive prove lying?

No. The investigators reported 28 such discrepancies but did not determine intentional deception, the timing of use, or the explanation for each mismatch. Survey wording, recollection, specimen timing, and individual laboratory records would require separate examination.

3. Were oral-fluid and blood THC measurements interchangeable?

No. Among 305 paired specimens, 17 results disagreed and kappa was 0.78 (95% CI 0.69–0.88). Those metrics quantify observed categorical agreement; they do not establish chemical concentration equivalence, a directional false-negative rate, or equal impairment-detection capacity.

4. Why did the 18–19-year-old group appear more often THC-positive?

Its unadjusted positivity was 15.3%, compared with 7.7% among 16–17-year-olds. But older teens could provide blood as well as oral fluid, and the adjusted age association of 1.21 (95% CI 0.57–2.57) was not statistically significant. Both points matter.

5. Does the 71.1% figure establish the time of cannabis use?

No. It is the proportion testing positive among the 45 study participants who self-reported use within 24 hours. It is neither an individualized ingestion-time equation nor a known-dose clearance study, and it should not be used to back-calculate a particular driver’s last use.

6. Does a confirmed THC result mean the teen driver was impaired?

No. The researchers did not measure individual driving performance or establish a validated chemical impairment threshold. Analytical positivity, time of consumption, current behavioral effect, and collision causation remain distinct questions.

7. Can this study show that legalization caused increased teen impaired driving?

No. It is an observational survey of volunteers at specific roadside sites, analyzed without national population weights, with no crash-outcome or individual impairment measurement. Associations with state-law classifications cannot independently establish a causal policy effect.

Bottom line: preserve the difference between disagreement and proof

The principal lesson is not that self-report is always wrong or that fluid testing is always dispositive. The study demonstrates that self-reported cannabis history and biologically detected THC can disagree within a real teen roadside sample, while two specimen matrices can also disagree within the same person. Its most defensible conclusions are the measured counts, assay-specific observations, and the limits of unweighted group associations. It does not establish why an individual gave a particular answer, when a particular drug dose was taken, or whether a particular driver was impaired.

For primary methods, participant definitions, and adjusted regression results, consult the complete original: Pressley, Arora, and Sarmah, Injury Epidemiology (2019), DOI 10.1186/s40621-019-0204-0. Apply its findings only to the scientific question its measurements actually answer.