Booze is behind an estimated 2.1 million car accidents each year in the U.S.—which cause almost 11,000 traffic fatalities annually. But many drug users have claimed that a few puffs of pot before getting behind the wheel are perfectly harmless. A new study, however, shows that drivers who smoke marijuana within a few hours of hitting the road are almost twice as likely as stone-sober motorists to be in a crash that results in serious injury or death.

Authors of the new paper, published online Thursday in the British Medical Journal (BMJ), sifted through nine previous studies to develop a clearer picture of the risks to users who light up before revving up. Previous studies have left the effects of marijuana on its own—when not combined with alcohol or other drugs—a little hazy.

But the researchers' findings make sense to others in the field. "Their results are consistent with experimental evidence that cannabis use leads to dose related impairments in simulated driving, psychomotor skills and on-road driving," Wayne Hall, of the University of Queensland's Center for Clinical Research who was not involved in the new research, wrote in a related essay in BMJ.

In addition to the finding that drivers who had recently smoked pot were substantially more likely to be involved in a serious accident, the researchers found that those who had died in these crashes had higher amounts of the drug's compound tetrahydrocannabinol than those who survived. But there was not enough data to link concentrations of the compound to various outcomes in order to suggest a threshold for dangerous intoxication, noted the researchers, who were led by Mark Asbridge, of Dalhousie University's Department of Community Health and Epidemiology.

Driving while stoned has become a hot topic as more states allow for medical use of marijuana. The 2009 National Survey on Drug Use and Health found that more than 10 million people admitted to having driven while on at least one illegal drug—with pot being the most common. More than a dozen states currently have roadside drug tests for cannabis that sample drivers' saliva for traces of tetrahydrocannabinol. But, as Hall noted, ascertaining a dangerous level—as is currently used in a breath-based test for alcohol (0.05 percent)—is less clear cut. So far many governments are using a zero-tolerance rule, but, as Hall pointed out, "researchers have proposed a concentration of tetrahydrocannabinol below which driving is not impaired."

Assessing the definitive risk for actually being in a car crash because of marijuana use is also tricky because studies haven't always looked at drivers who were not determined to be "at fault"—or passengers of vehicles or people involved in minor accidents.

Nevertheless, Asbridge and his colleagues noted, "This information could be used as the basis for campaigns against drug impaired driving." But the roadside testing has not been as widely publicized as rapid alcohol breath tests have been, so the chance of getting "caught" with pot in one's system doesn't seem to have scared very many people into not smoking before driving. As Hall wrote, the idea that roadside tests for pot will reduce traffic fatalities as drastically and as rapidly as breathalyzers did for alcohol "is probably too optimistic." But that doesn't mean that attempts to stub out the dangerous habit should be written off just yet. "Better evidence is essential," Hall said of the attempts to fight pot-impaired driving with more roadside testing.