How Many U.S. Traffic Deaths Are Really Linked to Cannabis?

Driving under the influence of any impairing substance is dangerous, and cannabis is no exception. As more states legalize medical and adult-use cannabis, understanding how cannabis affects road safety has become an increasingly important public health priority. But one question has remained surprisingly difficult to answer: How many U.S. traffic deaths are actually caused by cannabis? 

A 2026 peer-reviewed commentary¹ is challenging one of the most widely cited estimates in cannabis research that claims approximately 5,000 U.S. traffic deaths each year are attributed to cannabis. Rather than dismissing concerns about impaired driving, the author argues that the statistical methods behind that estimate likely overstate cannabis-only risk. After re-examining the calculations, the analysis arrives at a much lower estimate while also emphasizing that the available evidence remains surprisingly uncertain.

The debate is about much more than a single number. It illustrates why accurate science is one of the most important tools in harm reduction. Let’s dive into the numbers and why risk estimates need to be as precise as possible.

Why Measuring Cannabis Risk Is Challenging

Unlike alcohol, cannabis presents unique challenges for researchers trying to measure impaired driving. The primary psychoactive compound in cannabis, delta-9 tetrahydrocannabinol (THC), can remain detectable in the body long after noticeable impairment has faded. A driver may test positive for THC days, or sometimes even longer, after consumption, depending on numerous factors including frequency of use, metabolism, and testing methods. Because of this, a positive THC test does not automatically mean cannabis caused a crash or that the driver was impaired at the time of the collision.

Researchers instead rely on statistical models that estimate how much recent cannabis use increases crash risk across large populations. Those estimates depend heavily on the assumptions built into the calculations, meaning relatively small methodological differences can produce dramatically different conclusions.

Revisiting the 5,000 Cannabis Fatality Estimate

The commentary focuses on a previous peer-reviewed study that estimated roughly 5,000 annual U.S. traffic fatalities could be attributed to cannabis. Rather than collecting new crash data, the author carefully reexamined the mathematical framework used to produce that estimate.

The first issue involved the fatality denominator. The earlier analysis used National Safety Council motor vehicle fatality totals, which include a broader range of motor vehicle deaths than the National Highway Traffic Safety Administration’s Fatality Analysis Reporting System (FARS). Because FARS specifically tracks traffic fatalities occurring on public roads, the commentary argues that it represents the more appropriate dataset for estimating traffic deaths attributable to cannabis.

The second issue involved the measure of risk itself. The previous analysis treated an odds ratio as though it were a relative risk. While the two statistics are related, they are not interchangeable. Converting the published odds ratio into the corresponding relative risk reduced the estimated contribution of cannabis to fatal crashes.

Correcting these two methodological issues alone reduced the estimated number of annual cannabis-attributable traffic fatalities from approximately 5,000 to about 3,550.

The Role of Alcohol Co-Use

The commentary then examined another important source of potential overestimation: alcohol. Many fatal crashes involving THC-positive drivers also involve alcohol, particularly blood alcohol concentrations of 0.08% or higher. Alcohol at these levels is already known to substantially increase crash risk.

The author argues that simply counting every THC-positive fatality as evidence of cannabis contribution may inflate cannabis-only estimates because many of those crashes are better explained by alcohol impairment.

To address this issue, the commentary separates THC-positive cases involving high blood alcohol concentrations from cases involving cannabis without significant alcohol impairment. Using this approach, the preferred estimate falls further to approximately 1,787 cannabis-attributable traffic fatalities annually. While, this lower estimate is not presented as a definitive answer, it reflects what the author believes are more appropriate assumptions based on the currently available evidence.

Why the Evidence Remains Uncertain

Perhaps the most important conclusion is not the revised estimate itself but the uncertainty surrounding all current estimates. The commentary repeatedly acknowledges that measuring cannabis-impaired driving remains difficult because researchers cannot simply count THC-positive drivers and determine whether cannabis caused a crash. Reliable estimates depend on several variables, including the prevalence of recent cannabis use among drivers, the prevalence of THC among fatal crash victims, the relative risk associated with cannabis use alone, and how alcohol co-use is handled in the analysis.

The commentary goes so far as to state that, given current evidence, the true number of annual cannabis-attributable traffic fatalities could potentially be much lower than 1,787 and may even approach zero under certain assumptions. That statement reflects the author’s interpretation of the available data rather than scientific consensus, but it underscores just how contested this area of research remains. As researchers are working toward a more precise understanding of cannabis-impaired driving, today’s estimates should be viewed as evolving rather than final.

Why Accurate Numbers Matter

Whether the final estimate is closer to 5,000, 1,787, or somewhere in between, obtaining the most accurate number possible is essential, as effective harm reduction depends on credible evidence. If public health agencies overestimate risk, they risk losing public confidence when later studies revise those estimates. Conversely, underestimating risk could weaken education campaigns and leave drivers unaware of genuine dangers, and neither outcome benefits road safety.

Accurate estimates also allow policymakers to prioritize resources more effectively. Researchers can help to identify knowledge gaps, healthcare professionals can provide better guidance, and educators will be better able to develop prevention campaigns based on the best available science rather than assumptions. As cannabis legalization expands, improving the quality of impaired-driving research will become increasingly important.

Final Thoughts

One of the greatest strengths of science is its willingness to question previous conclusions when new evidence or improved methods become available. This commentary demonstrates exactly that process, asking whether current estimates accurately isolate cannabis-only risk after accounting for statistical methodology and alcohol co-use.

We all know driving while impaired should absolutely be avoided. Cannabis, as with other substances, can slow reaction time, reduce attention, and affect decision-making behind the wheel. However, for prevention, harm reduction efforts need to be built on the most accurate evidence available to be successful, which is exactly what makes this research so important.

As researchers continue refining methods for measuring cannabis impairment and crash risk, future estimates will likely become more precise. Until then, the latest commentary serves as an important reminder that numbers influencing public policy should be continuously tested, challenged, and improved. Better science ultimately leads to better public health and better harm reduction for everyone on the road.

References:

^{1. Michael A White, How many traffic fatalities can be attributed to cannabis each year in the US?, International Journal of Drug Policy, Volume 155, 2026, 105406, ISSN 0955-3959, https://doi.org/10.1016/j.drugpo.2026.105406}