BACKGROUND: Whole blood, plasma and urine delta-9-tetrahydrocannabinol (THC), 11-hydroxy-THC (11-OH-THC) and delta-9-carboxy-THC (THCCOOH) concentrations document cannabis exposure. Clinical data generally are derived from specimens obtained after single THC administrations. Few data are available from daily cannabis smokers. Moreover, predicting new cannabis use from urinary creatinine-normalized (CN) THCCOOH concentrations in two specimens has, until now, only been possible for occasional users. METHODS: Specimens were collected from chronic daily cannabis smokers enrolled in multiple clinical protocols. Whole blood and plasma were analyzed by two-dimensional gas-chromatography/mass-spectrometry (2DGCMS). Analysis of fortified samples stored at -20degC, 4degC and room temperature up to two weeks evaluated cannabinoid stability. Plasma was collected multiple times daily during eight days of escalating total daily oral THC doses up to 120mg/day to evaluate free and glucuronidated plasma cannabinoids. Models for predicting new cannabis use were developed and validated from CN THCCOOH concentrations in urine (N=2,377) collected during inpatient and outpatient studies. RESULTS: Cannabinoid concentration decreases >20% were observed in fortified samples stored at -20degC, room temperature, but not 4degC. THC intra- and inter-subject whole blood/plasma ratio variances were large (0.14 and 0.09 for THC, 0.14 and 0.28 for THCCOOH). Escalating daily oral THC yielded 11-OH-THC and THCCOOH, but not THC, accumulation in plasma. THC concentrations remained >1ng/mL, 19.5h after admission and 22.5h after the last dose. Models for predicting new cannabis use in daily cannabis smokers (>123,000 comparisons) were developed at four prediction probabilities (80, 90, 95 and 99%). Rules were devised to account for high %predictions, explained by peak concentrations occurring up to 40h after unit entry, and CN THCCOOH concentrations >200ng/mg up to 14d after unit entry in some participants. CONCLUSIONS: 2DGCMS with cryofocusing reliably quantified cannabinoids in whole blood and plasma. It is unknown whether instability (>20% losses) would occur in similarly stored authentic specimens. Plasma cannabinoids during escalating daily oral THC improves interpretation of free and glucuronidated concentrations achieved during daily THC administration. The urinary model will be extensively employed in multiple drug testing programs. In total, these data are valuable for interpreting whole blood, plasma and urine cannabinoid concentrations in chronic, daily cannabis users.