Objectives: To assess the relative risk of causing a fatal accident (or dying without being at fault) for drivers under the influence of narcotics (cannabis, amphetamines, cocaine, opioids) and/or alcohol, the prevalence of such drivers in the general population, and the number of fatalities attributable to these behaviors in France. Method: Population-based case-control study. Eligible subjects are the 10,748 drivers involved in a fatal accident in France from October 2001 to September 2003, for whom levels of drug and alcohol exposure are known. Cases are the 6,766 drivers at fault for their accidents; controls are a selection of 3,006 drivers from among the 3,982 not at fault. Results: A positive alcohol test and, uniquely among the four drug categories, a positive cannabis test are significantly associated with an increased risk of being at fault and of death after adjusting for various confounding factors, including the driver’s age. Dose-response effects are observed and remain significant after adjustment. The estimated prevalence of cannabis use (3.3%) among drivers on the road is lower than that of positive blood alcohol levels (5.3%, of which 2.7% exceed 0.5 g/L). An estimated 230 deaths annually are attributable to cannabis and 2,270 to alcohol. Conclusions: The number of traffic fatalities on French roads attributable to alcohol is much higher than that associated with narcotics. The number attributable to medications remains to be determined. (R.A.)