With decades of research, the link between alcohol and motor vehicle crashes is clear. But the research on cannabis and driving is mixed.

Last month a high school senior in Ohio drove off the road on the way home from prom, striking two utility poles and killing his passenger, 17-year-old Lindsey Rotuno.

The Ohio Highway Patrol eventually confirmed that driver Chase Johnson, 18, had marijuana in his system. Rotuno was not wearing a seatbelt.

Police are still investigating the crash, and the Lorain County prosecutor’s office has yet to determine if Johnson will face charges.

But some opponents of marijuana legalization are concerned that these types of crashes will only increase as more U.S. states relax restrictions on marijuana use.

While decades of research have shown that alcohol increases the risk of vehicle crashes, marijuana research is mixed. Still, it’s clear enough that some researchers urge caution.

“Alcohol is still the main contributor to [vehicle crash] risk. It doesn’t mean that marijuana or cannabis don’t contribute. To me, it’s very clear that if you are stoned, you are at risk. You shouldn’t drive at all,” Eduardo Romano, PhD, a senior research scientist at the Pacific Institute for Research and Evaluation, told Healthline.

Read more: The experts weigh in on marijuana legalization »

Currently, eight U.S. states and the District of Columbia have legalized adult-use recreational marijuana. Twenty-nine states and D.C. have legalized medicinal marijuana.

Between 2007 and 2014, the percentage of drivers in the United States who tested positive for THC — the main psychoactive chemical in cannabis — increased from 8.6 percent to 12.6 percent, according to a report by the National Highway Traffic Safety Administration (NHTSA).

Another NHTSA report found a similar increase in THC-positive drivers in the state of Washington after marijuana for recreational use was legalized in July 2014.

However, the only statistically significant change was in the percentage of drivers who tested positive for marijuana during the daytime — increasing from 7.8 percent just before the law went into effect, to 18.9 percent one year later.

Marijuana and alcohol were the most common substances in drivers’ systems. Drivers who tested positive for alcohol ranged from 4.4 percent to 6 percent.

The Washington State study used different methods than the countrywide study, so it’s not possible to directly compare the results.

A recent study by the Highway Loss Data Institute also estimated that vehicle collisions are about 3 percent higher in Colorado, Oregon, and Washington than if these states hadn’t legalized marijuana.

The study, though, cannot tell whether the increase in crashes is directly caused by drivers who were high.

But another study last year found that fatal vehicle crashes involving people who’d recently used marijuana before driving doubled after Washington State legalized the drug.

In this study, researchers from the AAA Foundation for Traffic Safety looked at crash records and drug tests done on the drivers.

However, even with the results of this study, researchers did not know when the driver had last used marijuana or if they were impaired by it.

Read more: Do medical marijuana laws encourage adults to misuse the drug? »

Although marijuana research is mixed, it points toward some effect of the drug on driving ability and crash risk.

“The evidence [for marijuana] is leaning toward something going on. But even with the studies that show an increased risk of crash involvement, the effect sizes are relatively low to moderate,” Mark Johnson, PhD, center director and senior research scientist at the Pacific Institute for Research and Evaluation, told Healthline.

Certain types of studies are more consistent.

“Experimental lab studies show with somewhat consistency that dosing people with cannabis does appear to impair performance on driving-related skills,” said Johnson.

In driving simulator experiments, marijuana increased drivers’ reaction times.

Also, drivers who had used marijuana displayed increased following distance and lane weaving.

The size of the effects varied from study to study, and depended on the THC dose and whether the drivers regularly used marijuana.

But does this translate to the real world?

“Just because you have a slow reaction time on some test, or you deviate within your lane 3 centimeters more than normal, it’s hard to say to what extent that means you’re more likely to crash,” said Johnson. “It says something, but it’s hard to piece it all together.”

With studies that look at actual crashes and whether marijuana was a contributing factor, “there’s a lot more inconsistency,” said Johnson.

A 2012 review in the BMJ of nine previous studies found that use of cannabis almost doubled the risk of a vehicle crash — although this increase ranged from 35 to 173 percent. Risk also varied based on the type of study and whether researchers looked at fatal or nonfatal crashes.

A later 2016 re-review of the same nine studies, published in the journal Addiction, estimated that cannabis increased the risk of a crash by 10 to 61 percent — what the researchers called a “low to medium” increase.

In this later study, researchers took into account factors that affect marijuana use. For example, young adults and males are more likely to drive under the influence of marijuana.

In comparison, alcohol has larger — and more consistent — effects on driving performance and crash risk.

But that doesn’t mean something isn’t going on with marijuana.

“I’m not trying to say that cannabis doesn’t impair performance,” said Johnson. “It may actually contribute to crash risk, but if it does, they are certainly smaller effects than alcohol, and there’s just more uncertainty about it.”

Read more: Changes in marijuana policy unlikely to have real effects »

Scientists have had a hard time pinning down the effects of marijuana on driving and vehicle crashes for many reasons.

“It’s very difficult to disentangle all these nuances and separate the contribution of marijuana from other effects,” said Romano. “So that’s why some of the inconsistencies are there.”

Romano speculates that marijuana may contribute more to certain types of crashes. Lumping these together can contribute to the mixed results.

Also, while alcohol has a strong effect on driving skills in general, marijuana may depend on the situation.

“Alcohol and cannabis tend to affect different types of skills,” said Johnson. “Some driving-related skills are more sensitive to alcohol. Others are more sensitive to cannabis.”

For example, a driver who knows they are high may slow down when they see someone crossing the road. But if the pedestrian stops in the road suddenly, the driver’s slower reaction time may not be enough to avoid a collision.

One study also showed that heavy cannabis users can develop tolerance to some of the effects of THC. At the same time, these people didn’t develop tolerance to the effects of alcohol from using marijuana.

One thing is clear, though — alcohol and cannabis are not a good combination.

“The combination is really lethal,” said Romano.

Read more: If marijuana is medicine, why can’t we buy it in pharmacies? »

The different effects that marijuana has on different people also makes it difficult for police officers to know if someone is impaired by the drug.

Roadside tests used to see if someone is under the influence of alcohol — like heel-toe walking, following a pen with the eyes, or standing on one leg while counting — may not catch everyone who is impaired by marijuana.

One study suggests that the Standardized Field Sobriety test identifies only about 30 percent of people who are high, but 88 percent of people who are drunk.

Police can determine a driver’s THC level from blood or urine samples. These tests, though, are often done hours later, which may not match the driving level.

Tests using saliva are also available, but are less accurate than blood and urine tests. Scientists are working on new saliva tests that produce better results.

Several states — including Colorado and Washington — have threshold limits for THC — similar to Blood Alcohol Content (BAC) limits. Drivers found over this limit are presumed to be guilty of driving under the influence.

But even with a highly accurate, on-the-spot test, a given THC level doesn’t always translate to driving impairment for every driver.

“Let’s suppose you have 5 nanograms of THC, is that impairment?” said Romano. “The law may say it is impairing, but it depends.”

Marijuana can be detected days or even weeks after use, so THC levels don’t always show when a person used the drug, or if the driver was impaired.

This differs sharply from alcohol, which is supported by more consistent scientific studies.

“We know that the higher the BAC, the greater the risk,” said Johnson, “and we know that 0.08 percent [BAC] is certainly associated with a much higher risk of crash involvement.”

The AAA Foundation for Traffic Safety recommends that instead of an arbitrary legal limit for THC, police officers should be trained to recognize marijuana-impaired drivers. A recent positive marijuana test would be used only to back this up.

Read more: Will Trump administration crack down on marijuana laws? »

Because of the inconsistent scientific research, “it’s easy to pick out studies that support your view and just talk about those,” said Johnson.

This may also fuel mixed public beliefs about the dangers of marijuana and driving.

In a 2016 survey by the AAA Foundation for Traffic Safety, 58 percent of more than 6,000 people thought that using marijuana one hour before driving increased the risk of a crash, while 32 percent didn’t know.

Another survey that year of 865 marijuana users in Colorado and Washington, published in the journal PLOS ONE, found that 44 percent of people had driven while high in the past year. Almost 24 percent had used marijuana within an hour of driving five or more times in the past month.

On the flip side, those who thought driving while high is unsafe were less likely to do it. Familiarity with the marijuana DUI laws, though, had no effect on whether people admitted to driving while high.

The researchers suggest that public education programs that focus on the risks of using marijuana while driving may be more effective than telling people about the laws.

One poster from Colorado’s Drugged Driving campaign takes this approach. It features a totaled car that resembles a smoldering joint, with the caption “Hits lead to hits. Don’t drive high.”

Canada’s Pot and Driving campaign has a poster that shows two pilots lighting a joint in an airplane cockpit. The caption: “If it doesn’t make sense here, why does it make sense when you drive?”

These campaigns are all about passing on a simple message, in spite of the confusing science.

“The only thing that is consistent is that alcohol is riskier than marijuana, but that doesn’t mean marijuana is not,” said Romano. “So to me, don’t smoke and drive.”