Estimating the time of last cannabis use is important in assessing possible impairment of drivers involved in accidents, in verifying accuracy of court testimony and in the future, helpful in therapeutic monitoring of cannabis agonists. In 1992, Huestis et al developed model 1, based on plasma Δ9-tetrahydrocannabinol (THC) concentrations, and model 2, on plasma 11-nor-9-carboxy-Δ9-tetrahydrocannbinol/THC ratios, that predicted 95% confidence intervals for time of last cannabis use. These models seemed to be valuable when applied to the small amount of data from published studies of oral ingestion, a route of administration more popular with the advent of cannabis therapies. A study was designed to further validate the models after oral ingestion of THC, and to determine whether they could predict last usage after multiple oral doses. Eighteen subjects in IRB-approved studies participated after providing informed consent. Each of 12 subjects in one group received a single 10 mg oral dose of dronabinol (synthetic THC). In another protocol, 6 subjects received 4 different oral daily doses, divided into thirds and administered with meals for 5 consecutive days. There was a 10-day washout period between each dosing regimen. Daily doses were 0.39, 0.47, and 14.8 mg THC in hemp oil and 7.5 mg dronabinol. Blood specimens were collected throughout the study and analyzed for plasma THC and 11-nor-9-carboxy-Δ9-tetrahydrocannbinol by gas chromatography/mass spectrometry with limits of quantification (LOQs) of 0.5 and 1.0 ng/mL, respectively. Actual times between ingestion of THC and blood collection spanned 0.5 to 16 hours. All plasma specimens with analyte concentrations >LOQ (n=90) were evaluated. Models 1 and 2 correctly predicted time of last THC ingestion for 74.4% and 90.0% of plasma specimens, respectively. 96.7% of predicted times were correct with one overestimate and 2 underestimates using the time interval defined by the lowest and highest 95% confidence limit of both models. These results provide further evidence of the usefulness of the predictive models in estimating the time of last oral THC ingestion after single or multiple doses.