A THC Pharmacokinetic Model Cannot Supply an Unmeasured Impairment Curve

forensic toxicology working group members October 28, 2026 Forensic Toxicology

A detailed model of THC concentrations does not automatically validate an impairment threshold. A pharmacometrics study of oral cannabis explicitly depended on assumptions about an impairment window and a minimally impairing dose.

How the model was assembled

The investigators combined concentration-time data from ten published human studies involving oral or intravenous THC. They digitized published mean curves and modeled parent THC together with 11-hydroxy-THC and THC-COOH.

Some source studies measured whole blood. The investigators converted those measurements to plasma using literature-based ratios. Those conversion assumptions belong in the interpretation; the final simulated values were not direct plasma measurements from a newly recruited cohort.

The outcome definition affects the answer

The study simulated doses and patterns of use, then compared concentration cutpoints with assumed impairment conditions. Its classification results depend on where those assumed boundaries were placed.

Calling a simulated detection a ‘false positive’ for impairment therefore requires qualification. It is disagreement with the study’s operational assumptions, not necessarily an observed error against an independently measured driving outcome.

Averages are not individual reconstruction

Because the input curves were reported means, the researchers could not estimate interindividual variability directly. Between-study variation does not fully replace the missing person-level information. This limits use of the model to reconstruct one individual’s last-use time or functional state.

The paper’s discussion of legal thresholds also should not be treated as current jurisdiction-specific legal advice. The scientific contribution is a way to investigate exposure patterns and test assumptions. Evidence directly linking concentration, time and measured driving performance remains necessary before converting that model into a validated impairment decision rule.