Analytical reliability of four oral fluid point-of-collection testing devices for drug detection in drivers.

BACKGROUND

Point-of-collection testing (POCT) devices for psychoactive substance detection through oral fluid samples are used in several countries for traffic enforcement. However, the reported reliability of such devices is quite heterogeneous among studies, and evaluating and comparing their analytical performance is of paramount importance to guide enforcement policies.

AIM: To evaluate the analytical reliability of four POCT devices for the detection of cocaine and cannabinoids using oral fluid samples of Brazilian drivers.

METHOD: A total of 168 drivers were recruited during standard roadblockfI procedures in Southern Brazil. Subjects were screened using one of the following POCT devices: the DDS2™, the DOA MultiScreen™, the Dräger Drug Test 5000™ and the Multi-Drug Multi-Line Twist Screen Device™ (MDML). Results of the screening tests were compared with chromatographic analyses in order to obtain the reliability parameters.

RESULTS

The prevalence of confirmed positive samples for cocaine and cannabinoids were 9 % and 4.4 %, respectively. For cocaine, three POCT devices (MDML™, Dräger DrugTest 5000™, DOA MultiScreen™) showed good reliability, greater than 80 % of performance measures, using guidelines for research on drugged driving published by Walsh et al. (cutoff 10ng/mL). However, for cannabinoids, the devices had low reliability-only Dräger DrugTest 5000™ had good performance using cut-offs proposed by Walsh et al. (cutoff 2ng/mL).

CONCLUSION

We observed a high prevalence of drivers testing positive for cocaine and cannabinoids. Most devices achieved good reliability performance for cocaine detection using cutoffs proposed by Walsh et al. or using the device's own cutoff. Instead, the reliability for cannabinoid detection obtained the desired parameters in just one device using cut-offs proposed by Walsh et al. and its own cutoff. Difficulties in detecting cannabinoids at the roadside should be better evaluated before the implementation of such tests.