Preliminary CDC figures record 69,973 overdose deaths in 2025, from all drugs. That is roughly 14% fewer than the year before and about 35% below the final 2022 peak, when the yearly toll reached 107,941.
The first decline could still be read as noise. The second suggested a pattern. A third consecutive fall makes the direction harder to dismiss. It does not erase the scale of the loss, but it changes what the national record now permits us to say: the curve can turn.
Fentanyl remains central to both the crisis and the decline
Deaths involving synthetic opioids, chiefly fentanyl, remain just over half of the total. They also fell fastest, from an estimated 48,913 deaths to 38,084, a decline of roughly 22%.
The drug categories overlap and cannot be added together. Even so, the movement within them matters. In 2024, every major class in the record declined at once. The broad national fall was not confined to a single drug or a single measurement.
A national turn still contains local curves that climb
The progress is real, but uneven. Eight jurisdictions moved in the opposite direction in the most recent state count. New Mexico, Arizona, and Colorado each rose by 10% or more even as most of the country fell.
Four of the increases occurred in states with fewer than 175 predicted deaths, where small changes can create large percentages. The West also peaked later than much of the country. Its curve may be behind the national turn rather than exempt from it. Either way, a falling national total cannot be mistaken for a uniform local experience.
No single cause owns the reversal
Naloxone became easier to carry and easier to reach. Barriers to treatment eased. A disrupted, sometimes less potent fentanyl supply may have played its part. Awareness also deepened in quiet, practical ways: test strips, counterfeit-pill warnings, carrying naloxone, and the discipline of never using alone.
These forces likely overlap. The mortality data do not divide the decline into causal shares, and no single intervention can claim the whole change. The responsible question is not which explanation wins, but which combination is saving lives and can be sustained.
Plausible contributors, not shares of a measured total
These are forces identified in the reporting and public-health record. They should be read as interacting explanations, not a causal decomposition.
None of this makes 69,973 deaths acceptable. It makes the direction meaningful. Public-health progress is often visible first as a line that stops worsening, then as a fall that survives another year.
The crisis is not over.
But the question has changed.
For twenty years we asked whether the line could turn at all. Now we know it can. The harder, more useful question is what is turning it, and how we keep it moving.
Source and method notes
- National figures use the May 13, 2026 release to preserve the article's 69,973 and 38,084 estimates.
- State change views use the later June 2026 VSRR extract. That release shows eight jurisdictions increasing.
- Final national data are residence-based; provisional state changes are occurrence-based. They provide context but are not interchangeable.
- Drug categories overlap. They are discussed independently and must not be summed.
- The final 2022 national peak was 107,941. The May 2026 estimate of 69,973 is 35.2% lower, rounded here to 35%.