THC Oil and Oral-Fluid Tests: Cutoffs and Sensitivity

The practical question: What exactly did this oral-fluid test classify, at what cutoff, and against which laboratory comparison? Those questions should come before any claim that a device is simply accurate or inaccurate.

AI-assisted research and evidence-review guide, September 27, 2026. This article does not represent personal review or adoption by Okorie Okorocha and is not a case-specific opinion, medical advice, or a driving-safety recommendation.

What the controlled oil study found

Suraev and colleagues studied 20 adults with insomnia in a randomized, double-blind crossover experiment. Each received a single oil dose containing 10 mg THC plus 200 mg CBD, or placebo. Oral fluid was checked at baseline and 0.5, 10, and 18 hours. The tested DrugWipe 5S and DrugTest 5000 configurations used 10 and 25 ng/mL cutoffs.

At 0.5 hours, matched-cutoff comparisons yielded sensitivity of 2/8 (25%) and 2/4 (50%). These denominators count reference-positive comparisons, not all 20 participants. At 10 and 18 hours, available samples were below those device cutoffs and screens were negative. However, 28 laboratory samples were unavailable because of instrument failure. Single-dose results in this selected population cannot establish performance after repeated dosing. The substudy does not supply a universal driving-safety waiting time. Source: Suraev et al., Drug Testing and Analysis, 2024; PubMed record.

Accuracy and sensitivity answer different questions

The following is an invented arithmetic example, not data from the oil study or a performance claim about either named device. Suppose a dataset contains 100 comparisons. The reference method classifies 10 as positive and 90 as negative. The screening test detects six of the positives, misses four, and correctly classifies all 90 negatives.

• Overall accuracy: 96 correct classifications out of 100, or 96%.
• Sensitivity: six detections out of 10 reference-positive comparisons, or 60%.
• Missed positives: four of the 10 reference-positive comparisons.

Both percentages are mathematically correct. They describe different denominators. The large group of correct negatives raises overall accuracy without repairing the four missed positives. A reader who sees only the 96% figure does not receive the full picture.

For an actual study, reconstruct the classification table before comparing percentages. Keep the number of people separate from the number of specimens and test occasions. Do not turn a small set of repeated measurements into an apparent study of hundreds of independent individuals.

The reference cutoff is part of the question

Consider a second hypothetical example. A laboratory reports 12 ng/mL, and a screening device returns positive. If the comparison rule defines a reference-positive result as at least 10 ng/mL, that pair is classified as a true positive. If the comparison instead requires at least 25 ng/mL, the same pair is classified as a false positive. This example assumes an accepted laboratory measurement and illustrates classification rules only.

The specimen and device result did not change. The definition of the target condition changed. A reviewer should therefore write down both the screening threshold and the reference threshold before using terms such as false positive, false negative, or sensitivity. A threshold used in a research analysis must not silently become an impairment threshold or a statement of current law.

How to report the comparison without overstating it

FDA’s 2007 statistical guidance recommends defining the target condition, intended-use population, study population, and comparative benchmark. It distinguishes accuracy against a reference standard from agreement with another method. It also recommends reporting fractions, percentages, confidence intervals, and an accounting of omitted or ambiguous results. These are useful reporting principles here, not a claim that FDA has approved either device or established a roadside-testing rule. The document contains nonbinding recommendations for diagnostic-device submissions. Read the FDA statistical guidance.

In practical terms, ask for the numerator and denominator behind a quoted percentage. Ask whether the comparator answers the question being claimed. If the claim concerns driving performance but the comparison classifies a chemical concentration, identify that gap explicitly. Additional evidence is needed before moving from one question to the other.

A records checklist for an oil-related result

1. Product identification: obtain the product name, formulation, lot, available composition evidence, and reported route of use. Record what is documented separately from what is assumed.
2. A specimen timeline: record the reported administration time, each collection time, screening time, and laboratory analysis time. Mark uncertain times as uncertain.
3. The actual device configuration: request the device model, applicable instructions, selected cutoff, lot information, and original result printout or contemporaneous record.
4. The laboratory comparison: identify the specimen tested, method, reported concentration, units, reporting threshold, and the records supporting identification and quantification.
5. Missing and invalid results: account for unsuccessful collections, invalid device outputs, omitted specimens, and whether a replacement sample was tested. Do not silently treat missing results as negatives.
6. The claimed conclusion: state whether the question concerns chemical detection, a decision threshold, time of use, or observed performance. Then identify which evidence addresses that specific question.

This checklist is an original review aid, not a finding about a particular laboratory, product, driver, or case. A missing record is a question to resolve; it is not automatically proof that a reported result is wrong.

Keep the related questions separate

Use the oral-fluid toxicology overview for the specimen context. For a different research question, see the residual-THC evidence guide. If the disputed inference concerns impairment or crash causation, the four-question cannabis evidence framework explains why those issues should be assessed separately rather than collapsed into one test result.

Review scope: the cited study’s main-paper methods, results, limitations, and tables were consulted, including visual inspection of Table 3. Supplementary tables were not independently reviewed. No source PDF or table is reproduced here.