THC and Crash Risk: What NHTSA’s 2016 Case-Control Study Actually Found

A positive THC result and a scientifically demonstrated increase in crash risk are not the same conclusion.

That distinction is central to the National Highway Traffic Safety Administration's 2016 report, Drug and Alcohol Crash Risk: A Case-Control Study. The study compared more than 3,000 crash-involved drivers with more than 6,000 control drivers in Virginia Beach. It collected 10,221 breath samples, 9,285 oral-fluid samples, and 1,764 blood samples.

The study initially found a modest statistical association between THC positivity and crash involvement. But after the researchers accounted for major differences between the groups, the association disappeared. The adjusted THC odds ratio was 1.00, with a 95 percent confidence interval from 0.83 to 1.22.

That does not mean cannabis can never affect driving. It means this study did not establish that a THC-positive test, by itself, increased crash risk after the measured confounding variables were taken into account.

Why the case-control design matters

The researchers did not simply compare drug-positive crash drivers with the general population. For each crash, they returned one week later and recruited control drivers from the same location, direction of travel, day of the week, and time of day. The purpose was to make the crash and non-crash groups more comparable.

This design matters because crash risk is influenced by many variables unrelated to a toxicology result. Age, gender, driving exposure, time, location, alcohol use, and other factors can differ between people who test positive and those who do not.

The study used statistical models to estimate whether alcohol or drug positivity remained associated with crash involvement after adjustment. The difference between an unadjusted and adjusted estimate is not a technical footnote. It can change the conclusion.

The THC result before and after adjustment

The unadjusted odds ratio for THC was 1.25. Standing alone, that number suggested that THC-positive drivers were more common among crash cases than among matched controls.

The investigators then adjusted for gender, age, race or ethnicity, and alcohol. The adjusted THC odds ratio became 1.00, with a 95 percent confidence interval of 0.83 to 1.22.

An odds ratio of 1.00 indicates no measured increase or decrease in the odds of crash involvement in that model. The confidence interval crossed 1.00, so the adjusted association was not statistically significant.

The report also found no significant adjusted crash-risk contribution for drugs as an overall category. Its adjusted estimate for all drugs was 0.99, with a 95 percent confidence interval of 0.84 to 1.18. Prescription and over-the-counter medications had an adjusted estimate of 1.02, with a 95 percent confidence interval of 0.83 to 1.26.

These results show why a raw comparison can be misleading. The THC-positive population differed from the THC-negative population in ways that were independently related to crash risk.

Alcohol produced a different pattern

Alcohol was the largest contributor to crash risk in the study. The relationship remained after adjustment for age and gender and increased with concentration.

At a breath alcohol concentration of 0.05 grams per 210 liters, the adjusted crash-risk estimate was 2.07. At 0.08 grams per 210 liters, the adjusted estimate was 3.93.

The contrast matters. The study detected a concentration-related alcohol effect after adjustment, while the adjusted THC-positive estimate was 1.00. Treating those findings as analytically equivalent would erase the central result of the report.

Drug presence is not a concentration-response finding

The THC analysis was based on a positive or negative toxicology classification. It was not a validated rule establishing that a particular blood THC concentration corresponds to a particular degree of impairment or crash risk.

THC moves rapidly from blood into tissues. Concentrations are affected by the route and timing of use, specimen type, collection delay, frequency of use, analytical cutoff, and individual pharmacokinetics. A detected concentration can answer an analytical question about what was present in the specimen. It does not automatically answer the separate questions of when the drug was used, whether the driver was impaired, or whether it caused a crash.

For the analytical side of that distinction, see Drug Analysis, Confirmatory Drug Testing, and Drug Metabolites in Toxicology.

What the study does not prove

The report should not be stretched beyond its design. It does not prove that cannabis has no effect on driving. It does not establish safety at any THC concentration. It does not validate a universal per se THC threshold. It does not decide whether a particular driver was impaired.

It also does not eliminate the possibility of effects in a subgroup or under circumstances the study was not powered to isolate. The adjusted estimate describes the data and model used in this population. It is not a biological law.

The scientifically supportable conclusion is narrower: after the study adjusted for age, gender, race or ethnicity, and alcohol, THC positivity was not associated with a statistically significant increase in crash risk.

Questions for a case-specific review

A case review should separate the population study from the individual evidence. Relevant questions include:

1. What specimen was tested, and when was it collected relative to driving and the crash?
2. Was delta-9 THC measured, a metabolite measured, or both?
3. What were the method's cutoff, limit of quantitation, uncertainty, and identification criteria?
4. Was alcohol present, and was the alcohol measurement obtained from the same relevant time period?
5. Were other drugs or medications detected?
6. What driving behavior was actually documented before the collision?
7. Are observations consistent across witnesses, video, medical records, and the toxicology timeline?
8. Does an opinion rely on drug presence, concentration, observed impairment, crash causation, or an unsupported mixture of those concepts?

The laboratory result should also be evaluated through the actual batch records, quality controls, calibrators, chromatograms, chain of custody, and method validation. See Chain of Custody for the evidence trail that connects the collected specimen to the reported result.

The bottom line

NHTSA's 2016 case-control study is important because it demonstrates what happens when confounding variables are addressed. The unadjusted THC association did not survive adjustment. The adjusted odds ratio was 1.00, and its confidence interval included no effect.

The finding does not declare cannabis harmless. It does reject a shortcut: a THC-positive result cannot be converted into increased crash risk merely by pointing to the unadjusted comparison.

In a forensic case, the question is not whether a drug name appears on a report. The question is what the validated analytical result, the timing, the observations, the other substances, and the appropriate scientific literature support together.

Primary source

Lacey JH, Kelley-Baker T, Berning A, Romano E, Ramirez A, Yao J, Moore C, Brainard K, Carr K, Pell K, Compton R. Drug and Alcohol Crash Risk: A Case-Control Study. Report No. DOT HS 812 355. National Highway Traffic Safety Administration; December 2016. Official NHTSA report page. Official report PDF.

For a broader framework for interpreting cannabinoid evidence, see Cannabis & THC Forensic Toxicology.

Editorial note: This article paraphrases the government report. Do not upload or reproduce the source PDF, publisher formatting, tables, or figures.