The CDC’s ‘1 Percent’ Cannabis Overdose Stat: What the Data Actually Shows
Mid Hudson News ran a piece last week headlined “CDC study links marijuana uses to health risks,” reporting that “marijuana was listed as a cause of death in 1 percent of the 209,166 overdose deaths reviewed over the past five years.” That single sentence is now bouncing around local news feeds and prevention-council press releases as evidence that cannabis is quietly killing people. So I went and read the actual CDC report the story is based on. What it says, and what a one-percent headline implies, are not the same thing.
The Report Behind the Headline
The source is a short CDC “Notes from the Field” bulletin titled “Cannabis Detection Among Overdose Deaths — United States, 2021–2025,” published in the Morbidity and Mortality Weekly Report on September 3, 2026, authored by Lauren Tanz and colleagues. It draws on the State Unintentional Drug Overdose Reporting System, known as SUDORS, which pools death certificates, medical examiner and coroner reports, and postmortem toxicology data from 31 states plus D.C. The dataset covers 209,166 unintentional or undetermined-intent overdose deaths between January 2021 and June 2025.
Here’s the part that gets lost in translation every time this study gets summarized: SUDORS records three genuinely different measurements, and they answer three genuinely different questions.
|
Measure |
Count |
Share of 209,166 deaths |
|---|---|---|
|
Cannabis detected on postmortem toxicology |
43,880 |
21.0% |
|
Cannabis listed by the medical examiner/coroner as involved in causing death |
1,687 |
0.8% |
|
Cannabis the only drug listed as involved |
9 |
0.004% |
The “1 percent” figure circulating in the press is the middle row, rounded up from 0.8 percent. That’s cannabis being involved in a death, per a medical examiner’s determination, almost always alongside other substances. It is not cannabis being the cause, and it is absolutely not cannabis acting alone. That distinction isn’t semantic hairsplitting. It’s the entire story.
Detected, Involved, and Alone Are Three Different Claims
SUDORS captures every drug that shows up on postmortem toxicology, whether or not a medical examiner ultimately determined that drug caused the death. That’s why cannabis shows up in the toxicology of 21 percent of all overdose deaths in this dataset, a huge number that simply reflects how common cannabis use is, including among people who die of something else entirely, like a fentanyl overdose, while cannabis happens to also be in their system.
The medical examiner’s “involved” determination narrows that down to 1,687 deaths, 0.8 percent, where cannabis was judged to have contributed to the death. And the report’s own authors are explicit about what that category actually represents. As they put it, cannabis detection is “mainly a marker of polysubstance use” rather than an independent driver of fatal overdose. Among those 1,687 cannabis-involved deaths, 73.5 percent also involved illegally manufactured fentanyls, and 60.4 percent also involved stimulants. In other words, the overwhelming majority of the “cannabis-involved” deaths were fentanyl or stimulant deaths where cannabis was present as a co-factor, not the driving cause.
That leaves nine deaths, nine out of 209,166, where cannabis was the only substance listed as involved. Nine. That’s 0.004 percent, effectively a rounding error, and even that number comes with a giant asterisk I’ll get to below.
You Cannot Go Verify the Nine Cases Yourself
Here’s something worth knowing before anyone cites this study with confidence in either direction: the underlying SUDORS case files, the death certificates, the ME and coroner narratives, the full toxicology reports, are not public. They’re restricted, accessible only through CDC’s Research Data Center to researchers with an approved proposal. They are not posted online, not available through a FOIA request as named case files, and not described case by case in the MMWR bulletin itself.
That means nobody outside that restricted system can currently pull up the ages, jurisdictions, THC concentrations, comorbidities, or circumstances behind those nine cannabis-only deaths. We don’t know if they involved edibles, synthetic cannabinoids mislabeled as cannabis, an undiagnosed cardiac condition, a coding quirk, or something else entirely. Later commentary, including from NORML, has flagged this exact gap and suggested those nine cases are just as plausibly explained by testing artifacts, coding decisions, or unlisted co-factors as by a genuine, isolated cannabis toxicity. Nobody can currently prove it either way, and any article, mine included, that claims certainty about what killed those nine people is overstating what’s actually knowable from the public record.
The Number That Actually Deserves the Headline
If Mid Hudson News wanted a genuinely alarming, well-supported statistic from this same report, they had one sitting right there and mostly buried it: cannabis detection was highest, by a wide margin, among decedents aged 12 to 17, showing up in 42.7 percent of the 1,028 adolescent overdose deaths in the dataset, the highest rate of any age group and roughly double the overall 21 percent. And among those adolescent cases where cannabis was detected, fentanyl was also present in 88.6 percent of them, with stimulants in 69.0 percent.
That’s a real, well-documented finding, and it’s a genuinely useful one for parents, schools, and prevention programs like the one quoted in the original article. It says something specific and actionable: kids who are overdosing are very often mixing substances, cannabis included, with illicit fentanyl, and cannabis use in that age group correlates strongly with exposure to the drug actually driving the overdose crisis. That’s a story worth telling accurately. “Marijuana caused 1 percent of overdose deaths” is not that story. It’s a flattened, decontextualized number that turns a polysubstance-use warning into a cannabis scare headline.
Why This Matters Beyond One Local Article
It’s also worth noting that screening for cannabinoids isn’t universal across every jurisdiction in SUDORS, so even that 21 percent detection figure likely undercounts actual cannabis presence. That cuts against, not toward, any narrative that this dataset is somehow inflating cannabis’s apparent role.
And if you actually want to study how cannabis appears on U.S. death certificates more broadly, using the ICD-10 code T40.7 as a contributing cause, that’s a separate CDC dataset entirely, the National Center for Health Statistics’ multiple-cause-of-death mortality data, accessible through the WONDER database. It covers a different set of deaths, uses a different methodology, and comes with the same lack of public case narratives. It is not interchangeable with the SUDORS numbers being cited here, and conflating the two datasets is its own separate way to get this story wrong.
The CDC’s own bottom line, buried under the headline everyone’s quoting, is that fatal cannabis-only overdose is rare, full stop, and that the real public health signal in this data is co-use, especially among adolescents mixing cannabis with fentanyl and stimulants. That’s a legitimate, important finding. It’s just not the one-percent scare stat making the rounds in local news, and the difference between the two is exactly the kind of thing that should matter to anyone using this study to argue for policy, whether that’s tighter regulation or renewed prohibition.
CAN YOU OVERDOSE ON WEED, READ ON…
