A 24% Drop in Overdose Deaths Is Real. So Is the Racial Gap That Didn't Shrink.

On June 2, 2026, researchers at UC San Diego, UCSF, NYU, and UCLA published the most comprehensive analysis yet of where the overdose crisis stands. The study, in the journal Addiction, traced every recorded U.S. overdose death from 1999 to 2024 using the CDC WONDER database. The headline number — a 24.4% decline in the age-adjusted drug overdose death rate from 2023 to 2024, from 31.3 to 23.7 deaths per 100,000 standard population — is, by any measure, historic. It is the steepest single-year drop in the recorded history of the American overdose crisis, and the first across all four waves simultaneously.

That is unambiguously good news. But the same study, read carefully, contains a set of facts that complicate any clean narrative about the crisis abating.

The paper: “Why have overdose deaths decreased?” by Joseph Friedman, MD, PhD, MPH (UC San Diego, first author), Steffanie Strathdee, PhD (UC San Diego, senior author), Annick Borquez, Daniel Ciccarone (UCSF), Chelsea Shover (UCLA), Joseph Palamar (NYU), and collaborators. Published June 2, 2026, in Addiction.

What the data show:

Fentanyl deaths — the driver of the fourth wave — fell sharply. Deaths involving synthetic opioids other than methadone dropped 35.6%, from 31,193 to 19,673 in the fentanyl-alone category, and from 41,583 to 28,062 for fentanyl in combination with stimulants. These are enormous declines in absolute numbers: roughly 25,000 fewer fentanyl-involved deaths in a single year. Total overdose deaths in the U.S. in 2024 are estimated at approximately 80,000 — still catastrophic by any historical comparison but a significant reduction from the 107,000+ peak.

Friedman’s team proposes two primary mechanisms. First, widespread market saturation: fentanyl has penetrated the drug supply so thoroughly that tolerance among people who use drugs regularly has increased at a population level, raising the effective dose threshold for fatal overdose. Second, some evidence of reduced use in high-risk populations, potentially driven by expanded naloxone access, treatment availability, and harm reduction infrastructure built during the crisis years. STAT News had separately reported in January 2026 that fentanyl potency on the street may also be declining — a supply-side factor that could account for part of the improvement.

What the data also show:

Stimulant-only deaths — overdoses involving cocaine or methamphetamine without fentanyl — rose. They grew from 18,142 in 2023 to 18,907 in 2024, increasing their share of total overdose fatalities from 17.3% to 23.8%. As the fentanyl toll falls, the stimulant toll becomes a larger fraction of a smaller total — and unlike fentanyl overdoses, stimulant overdoses have no established pharmacological reversal agent. Naloxone doesn’t help. There is no “Narcan for meth.” The expanding contingency management evidence base is the most promising treatment direction, but it is not a crisis intervention tool in the way that naloxone is.

Racial disparities narrowed in some categories and did not narrow enough in others. Non-Hispanic Black Americans saw a 29.3% decline in overdose death rates — a meaningful improvement, though their rate remains 1.5 times the national average. American Indian and Alaska Native people face an overdose death rate of 50.8 per 100,000 population, more than double the national figure. The improvement in those communities was not sufficient to close the gap.

Friedman described the findings as “a historic shift” while flagging the stimulant and disparity trends as requiring specific policy responses. The word “historic” is earned. So is the caveat.

Why it matters for treatment access:

The 24% decline is partly a story about what worked. Expanded naloxone distribution, buprenorphine deregulation, the growth of low-barrier treatment access — these interventions scaled over the past several years, and the data are consistent with their having contributed to the reduction. For Rize, that’s an argument for the model: matching people to treatment faster, with less friction, produces outcomes that show up in population-level data.

The stimulant mortality picture is a direct challenge to that model, because the stimulant treatment landscape lacks the same pharmacological scaffolding as opioid treatment. There is no buprenorphine for methamphetamine. Contingency management — paying people for negative urine screens — works, but federal caps on payment amounts have constrained its rollout. The people dying of stimulant overdoses are dying without the pharmacological safety net that has driven the fentanyl progress.

Contingency management — paying people for negative urine screens — works, but federal caps on payment amounts have constrained its rollout.

And the racial disparities are a challenge to any framing of this as a solved or solving problem. American Indian and Alaska Native communities dying at 50.8 per 100,000 is not a rounding error in an otherwise good news story. It is a separate crisis, running in parallel, that requires specific and targeted responses — not a general improvement in the average.

The study is the best data available on where the crisis stands. The headline is real. The asterisks are real too.