A Urine Cannabis Recency Model Performed Differently After Edibles and Inhalation
A urine model that performed well after smoked or vaporized cannabis performed poorly after oral cannabis in a controlled study. Validation for one exposure route did not establish that the same interpretation worked for edibles.
A controlled comparison across routes
Eleven frequent and nine occasional cannabis users completed a randomized within-person study involving smoked, vaporized or oral cannabis with a nominal THC content of 50.6 mg per active session. Researchers analyzed 2,232 urine specimens for free cannabinoids and metabolites, including glucuronides.
Participants stayed on a secure unit during the collection periods. Frequent-user, occasional-user and optional oral-session follow-up periods differed, and some markers remained detectable when observation ended.
Recent use and new use were separate questions
One published model evaluated recent use from sequential creatinine-normalized urinary THC-glucuronide concentrations. Applied to these data, the authors reported good predictability after smoking or vaporization—88.9% for occasional users and 90.9% for frequent users—but much lower performance after oral exposure, at 28.6% and 37.5%.
Oral dosing produced later peaks and some late specimen pairs met the model’s recent-use criteria. Residual THC-glucuronide also met criteria during placebo sessions in two frequent users. A positive mathematical classification therefore required attention to route and prior exposure.
Do not combine different models into one accuracy claim
A separate model addressed whether new use occurred between specimens using creatinine-normalized THC-COOH-related measurements and collection intervals. It correctly classified many pairs without new use but also missed some known new-use pairs.
Hundreds of specimen pairs did not represent hundreds of independent participants. Multiple pairs came from the same small group, and some errors clustered within individuals. Pair-level performance cannot simply be presented as a population-wide probability.
What a case-specific interpretation requires
Identify the exact model, metabolite, normalization method, collection intervals, use history and exposure route. Confirm whether the case falls within the conditions under which the model was evaluated.
The study supplied urinary exposure and elimination data. It did not establish a urine threshold for driving impairment. Its central lesson is that a model’s intended question and validation domain must travel with the result.
Related: Interpreting THC concentration and impairment.
Primary research
Free and Glucuronide Urine Cannabinoids after Controlled Smoked, Vaporized and Oral Cannabis Administration in Frequent and Occasional Cannabis Users. DOI: 10.1093/jat/bkaa046.
