rethinkTHC Search
Menu
Study breakdown

Can Saliva Tests Estimate Blood THC Levels for Roadside Drug Testing?

ObservationalPreliminary evidence
The takeaway

Dividing oral fluid THC concentration by a regression coefficient provided an acceptable estimate of blood THC levels in a population, though accuracy was poor when fewer than 15% of the population had high blood levels.

Read this if you are involved in drug-impaired driving research or policy.

Regression coefficient method best for estimating blood THC from saliva; fails below 15% prevalence

What the researchers found

Researchers compared five methods for estimating how many people in a population have high blood THC concentrations based on oral fluid (saliva) testing.

The methods ranged from simple calculations (dividing saliva concentration by an average ratio) to complex Monte Carlo simulations.

Dividing by the oral fluid/blood regression coefficient gave the best balance of accuracy and precision, making it the recommended method.

Monte Carlo simulations could give better accuracy but required good data on the distribution of oral fluid/blood ratios, which is not always available.

None of the methods worked well when fewer than 15% of the population had high blood drug concentrations.

Why it matters

Roadside saliva testing for drugged driving is more practical than blood testing. If saliva THC levels can reliably estimate blood levels, it could improve drug-impaired driving enforcement.

The numbers in context

311 cannabis users and 197 amphetamine users. Five estimation methods compared. Best method: dividing by regression coefficient. Accuracy acceptable only when 15%+ of population had high blood levels.

How the study worked

Methodological study comparing five estimation methods using data from 311 cannabis users and 197 amphetamine users from the Rosita-2 Project. Methods included simple ratio calculations and Monte Carlo simulations.

What this study cannot tell us

The relationship between oral fluid and blood THC concentrations is highly variable between individuals. No single method was highly accurate. The 15% prevalence threshold limits applicability in low-use populations.

How to read the evidence

Methodological study using existing datasets. Provides practical guidance but with significant accuracy limitations.

When this study was published

Published in 2010. Oral fluid drug testing technology and methodology have improved since then, though individual variability remains a challenge.

The bigger picture

As cannabis legalization expands, practical tools for roadside impairment assessment become increasingly important. This study contributed to the development of saliva-based drug testing protocols for traffic safety.

Questions still open

  • Can newer saliva testing technologies improve accuracy? Should impairment be assessed by THC levels or by functional tests? How does the oral fluid/blood ratio vary with different cannabis consumption methods?

Common questions

Can a saliva test tell if someone is impaired from cannabis?
Saliva tests can detect THC but estimating impairment from saliva levels is challenging due to high individual variability in the oral fluid/blood ratio. This study found population-level estimates are possible but individual-level predictions remain problematic.
Why not just use blood tests?
Blood tests are more accurate but require medical personnel and are impractical for roadside use. Saliva testing is non-invasive and quick, making it more feasible for traffic enforcement, even if less precise.

Read the original research

Can the prevalence of high blood drug concentrations in a population be estimated by analysing oral fluid? A study of tetrahydrocannabinol and amphetamine.

Forensic science international, 195(1-3), 153-9

Citation

Gjerde, Hallvard; Verstraete, Alain. (2010). Can the prevalence of high blood drug concentrations in a population be estimated by analysing oral fluid? A study of tetrahydrocannabinol and amphetamine.. Forensic science international, 195(1-3), 153-9. https://doi.org/10.1016/j.forsciint.2009.12.011

Explore the wider topic