fb-pixelMGH claims breakthrough in detecting marijuana impairment - The Boston Globe Skip to main content

MGH claims breakthrough in detecting marijuana impairment

Study also highlights flaws of existing “drug recognition expert” system

Researchers at Massachusetts General Hospital, seen here in a 2019 file photo, say they have developed a novel approach for detecting cannabis impairment.David L. Ryan/Globe Staff

Boston researchers say they’ve developed a new, noninvasive technique for detecting marijuana highs that can reliably tell the difference between people who are truly impaired by the drug and those who merely used it recently.

The claimed breakthrough by scientists at Massachusetts General Hospital comes amid intensified debate in the state over how to police stoned driving in the wake of cannabis legalization, with Governor Charlie Baker recently calling for a crackdown even as critics question the extent of the problem and warn that existing impairment tests are flawed.

The MGH researchers eschewed older methods that attempt to infer functional impairment from the amount of marijuana compounds in someone’s blood or saliva in favor of a more direct approach: Peering into the brain itself with light-based imaging.


While further development and validation is needed, the team behind the work is hopeful it could lead to a roadside device that would allow police to catch dangerously stoned drivers without sweeping up law-abiding cannabis consumers and medical marijuana patients who have THC in their system but are not actively impaired.

“For so long, our model has been alcohol, so there’s been a lot of focus on breath and blood levels,” Dr. Jodi Gilman, who led the research, said. “Our thought was, ‘What about looking directly at the brain?’ ”

In their study, published in January in the journal Neuropsychopharmacology, the MGH researchers first measured levels of oxygenated hemoglobin in the brains of 169 sober volunteers using functional near-infrared spectroscopy, or fNIRS.

Unlike the massive, electricity-hungry MRI scanning machines employed by hospitals, fNIRS is relatively portable; it measures the photon reflections from low-power LED bulbs mounted on a skullcap and shined into the skull. Similar technology is already widely used in smartwatches and other fitness gadgets to measure users’ heart rates and blood oxygenation.


After giving some of their volunteers THC capsules and others a placebo, the MGH team classified each person as impaired or not impaired based on self-reporting by the study subjects and the consensus of multiple clinicians who were unaware of which subjects had eaten the “real” edible and had made thorough before-and-after observations of their behavior (an impossibility on the roadside).

The researchers then conducted a second round of brain scans and found that the people classified as impaired had significantly higher levels of oxygenated hemoglobin than subjects who ate the placebo or didn’t get too high from the THC-infused edible.

“Essentially, the [impaired] brain becomes more active but less efficient at processing, so the body gives it priority and sends more oxygen,” said Dr. A. Eden Evins, who directs the MGH Center for Addiction Medicine and helped oversee the research.

Next, the scientists trained a computer algorithm to spot the differences in oxygenated hemoglobin between those who were deemed high and those who were not. Later, analyzing only the “after” scans, the software yielded false positives in just 10 percent of subjects and correctly guessed which were impaired about 76 percent of the time, a significant improvement over existing techniques and a figure the researchers believe they can boost substantially with further refinements.

Crucially, the system rarely indicated impairment in test subjects who consumed the THC-infused edibles yet were not deemed functionally impaired. That puts it far ahead of older methods that automatically designate anyone with a high enough level of marijuana metabolites in their system as impaired, regardless of how well they’re actually functioning.


Globe Staff

The scan results lend further credence to a wave of recent studies indicating (as marijuana consumers have long insisted) that there is little if any connection between a given dose of THC, the level of marijuana metabolites in blood or saliva at a given time after use, and a particular level or even likelihood of impairment. Reactions to the drug simply vary too drastically between individuals, as do methods of ingesting cannabis and the rates at which people metabolize it. Meanwhile, heavy consumers appear to at least partially adapt, and some medical marijuana patients even show cognitive improvements following use.

Taken together, those confounding factors mean traditional blood and saliva tests are barely more accurate than a coin toss at detecting impairment, according to Dr. Thomas Arkell of the University of Sydney’s Lambert Initiative cannabis research center.

“People wanted an easy cutoff like the blood alcohol limit, but it turns out cannabis doesn’t really work that way at all,” said Arkell, who conducted a simulated driving study last year in which common limits on THC concentrations in blood failed to accurately sort out which test subjects were actually stoned. “There is a need for a new model.”

The MGH study also raises troubling questions about the validity of another existing technique for measuring pot impairment: examinations by so-called drug recognition experts, police officers who are trained to detect impairment through a series of observations and simple physical tests like those used to assess suspected drunk drivers.


The researchers had initially planned to compare their experimental results to verdicts rendered by DREs, which proponents hail as the gold standard for assessing drug impairment. Officers across the country, including dozens in Massachusetts, have been using the system for years to evaluate suspected drugged drivers and present evidence against them in court.

But the MGH scientists quickly discarded the DRE protocol as a reference after their experiments showed it produced false positives in a staggering 34 percent of the subjects who were not deemed impaired by the clinical and self-assessments. The DRE-trained observers even incorrectly flagged 20 percent of the volunteers who had eaten the placebo and were verifiably sober as stoned.

Baker last year proposed a bill that would significantly expand the deployment of DRE-trained officers in Massachusetts and require courts to accept their testimony as experts. Critics have warned the proposed law, under which drivers who refuse a blood test for cannabis metabolites would lose their licenses, is subjective, prone to officer bias, and would result in the arrests of innocent marijuana consumers. Earlier this week, a key State House committee sent the proposal to study, likely ending its hopes of passage during the current legislative session.

Evins and Gilman hope their novel, more objective approach will prove it can sidestep those limitations, though they stressed it isn’t ready to be deployed. One critical next step will be testing the system on a larger group of volunteers, in part to ensure that other drugs or health conditions don’t produce scan results that mimic those of stoned people. Researchers will also need to develop a slightly smaller and more rugged fNIRS device suitable for roadside use, perhaps one that sends its readings over the cellphone data network to a remote computer for analysis.


“Officers need a better tool,” Evins said. “A big part of what got us going was the potential for bias in the system that exists now. There’s a real urgency to develop a reliable and objective way to identify marijuana impairment and make our roads safer, and we’re delighted to contribute to that.”

Other researchers in the field praised MGH’s innovation as intriguing and worthy of further study, but also pointed to potential limitations.

“A lot of factors can potentially influence blood flow in that part of the brain,” Arkell said. “Without having an individual baseline, I’m not sure how effective this would be as a roadside strategy.”

Dr. Timothy Naimi, a Boston Medical Center physician and public health researcher at the Boston University Schools of Medicine and Public Health, said MGH’s approach was promising but would need to be extensively validated in future studies that included driving simulations instead of leaning on self-assessments of impairment.

“There could be a host of legal issues about using this in practice,” he said. “The reliability would have to be really good to use as the basis of a [court] proceeding.”

Dan Adams can be reached at daniel.adams@globe.com. Follow him on Twitter @Dan_Adams86.