"Conducting case-control studies to estimate the risk of crash involvement from drug use presents many difficulties. The first challenge has been getting reliable and accurate estimates of drug use. Many studies rely on self-report (which have obvious inherent problems) rather than actual measurement of THC in blood or oral fluid.
"Also, the method of selecting truly comparable control drivers in an unbiased fashion for the crash involved drivers has varied considerably. The more carefully controlled studies, that actually measured marijuana (THC) use by drivers rather than relying on self-report, and that had more actual control of covariates that could bias the results, generally show reduced risk estimates or no risk associated with marijuana use (Elvik, 2013)."

Source

Compton, R. P. & Berning, A. (2015, February). Drug and alcohol crash risk. Traffic Safety Facts Research Note, Report No. DOT HS 812 117. Washington, DC: National Highway Traffic Safety Administration.