Alcohol Use Disorder Is Finally Falling. Cannabis Use Disorder Is Climbing to Take Its Place.

Published: JAMA Psychiatry, August 19, 2026 | Study type: National survey analysis (NSDUH, 2021–2024) | Lead author: Beth Han, using federal data on 186,823 U.S. adults 18 and older

The finding

For years, the addiction field’s two biggest legal-substance stories ran on separate tracks: alcohol use disorder as the quiet, under-discussed epidemic; cannabis as the substance whose risks were shrinking in public perception as legalization spread. A new JAMA Psychiatry analysis says those two tracks have started running toward each other.

Alcohol use disorder is genuinely declining. Cannabis use disorder is rising just as fast in the other direction — and this isn’t two separate trends, it’s one substitution pattern.

Between 2021 and 2024, the number of U.S. adults with alcohol use disorder fell from 28.8 million to 27.1 million. Among men aged 21 to 34, prevalence dropped from 20.0% to 17.0%; women 21 to 49 saw the largest declines of any group. In the same window, adults with cannabis use disorder rose from 15.3 million to 19.4 million — men from 7.3% to 9.3%, women from 4.5% to 5.6%.

Where it lands hardest

The headline numbers undersell what’s happening among women specifically, and this is the detail worth sitting with rather than skimming past: among women who have cannabis use disorder, the share with moderate-to-severe cases jumped from 37.0% to 45.0% of all cases in just three years. Among women 50 and older, moderate-to-severe cannabis problems rose from 0.1% to 0.7% of the population — a sevenfold increase in a demographic almost nobody was screening for cannabis problems three years ago, because almost nobody expected them there.

Dr. Benjamin Caplan, chief medical officer of CED Clinic, told Forbes the drivers are the ones you’d guess if you’ve watched a dispensary menu update over the past five years: rising product potency, wider legal accessibility, and a steady erosion of the sense that cannabis carries real risk. That erosion is measurable in a second, separate study published the same week: NYU Langone researchers, led by Dr. Omar El-Shahawy, analyzed 17,732 adults from the same NSDUH dataset and found that 14% of past-month cannabis users are “polymodal” — smoking, vaping, dabbing, and eating edibles rather than sticking to one method. Half of polymodal users consume daily, compared to 28% of people who only smoke, and 67% of polymodal users meet criteria for cannabis use disorder, compared to 36% of single-method smokers. In plain terms: the more ways someone has of getting high, the more likely daily use has already become dependence, whether or not they’d describe it that way themselves.

What’s missing, and why the gap matters

Here’s the asymmetry that should worry a treatment provider more than the raw numbers: alcohol use disorder has three FDA-approved medications — naltrexone, acamprosate, disulfiram — plus a fast-growing body of evidence for GLP-1 drugs like semaglutide as an adjunct. Cannabis use disorder has zero approved medications. The entire treatment toolkit for a fast-rising, moderate-to-severe-skewing disorder is behavioral: cognitive behavioral therapy and motivational enhancement therapy, both real and evidence-based, neither backed by anything a doctor can prescribe. That gap was tolerable when cannabis use disorder was a smaller, slower-growing problem. It’s a bigger liability now that the prevalence curve is bending the way alcohol’s used to.

The timing compounds the gap. DEA’s final rescheduling briefs, filed August 17 arguing marijuana belongs in Schedule III, are moving through the system at the exact moment this data shows the clinical picture getting more serious, not less. Rescheduling would ease research restrictions that have slowed medication development for decades — which means the DEA’s docket and the JAMA Psychiatry dataset are, without either side necessarily intending it, arguing for the same next step from opposite directions.

The limits, stated plainly

This is NSDUH self-report data, not diagnostic chart review, and self-reported substance use surveys undercount in ways that are well documented but hard to correct for precisely. The study also can’t establish that people are causally substituting cannabis for alcohol rather than the two trends simply moving independently in the same period — a real limitation the authors themselves would be the first to name. What it can do is document, with a large and nationally representative sample, that the substance driving the most new moderate-to-severe cases in America right now is not the one most of the field has spent the last decade worrying about.

This is NSDUH self-report data, not diagnostic chart review, and self-reported substance use surveys undercount in ways that are well documented but hard to correct for precisely.