Résumé : Oral fluid (collected with the Intercept® device) and plasma samples were obtained from 139 individuals suspected of driving under the influence of drugs and analyzed for Δ9-tetrahydrocannabinol (THC), the major psychoactive constituent of cannabis, using a validated quantitative LC-MS-MS method. The first aim of the study was to investigate the correlation between the analytical data obtained in the plasma and oral fluid samples, to evaluate the use of oral fluid as a 'predictor' of actual cannabis influence. The results of the study indicated a good accuracy when comparing THC detection in oral fluid and plasma (84.9-95.7% depending on the cut-off used for plasma analysis). ROC curve analysis was subsequently used to determine the optimal cut-off value for THC in oral fluid with plasma as reference sample, in order to 'predict' a positive plasma result for THC. When using the LOQ of the method for plasma (0.5 ng/mL), the optimal cut-off was 1.2 ng/mL THC in oral fluid (sensitivity, 94.7%; specificity, 92.0%). When using the legal cut-off in Belgium for driving under the influence in plasma (2 ng/mL), an optimal cut-off value of 5.2 ng/mL THC in oral fluid (sensitivity, 91.6%; specificity, 88.6%) was observed. In the second part of the study, the performance of the on-site Dräger DrugTest® for the screening of THC in oral fluid during roadside controls was assessed by comparison with the corresponding LC-MS-MS results in plasma and oral fluid. Since the accuracy was always less than 66%, we do not recommend this Dräger DrugTest® system for the on-site screening of THC in oral fluid. © 2006 Elsevier Ireland Ltd. All rights reserved.