Overdose Deaths Are Falling Fast — Because the Supply Broke, Not Because We Fixed Anything

A woman in Kensington is counting differently than the CDC is

Elena tracks overdoses the way people who actually use fentanyl track weather — by what she sees on her block, not by what an agency publishes months later. When NPR spent time with her in Philadelphia’s Kensington neighborhood, one of the most concentrated open-air drug markets in the country, she confirmed what the national numbers were starting to show: “I can tell, you know, the numbers dropped a lot.” She credited part of it to people switching from injecting fentanyl to smoking it, and part of it to naloxone — the opioid-overdose-reversal medication, sold as Narcan — which friends have used on her more than once. Her partner Vadim had a colder theory: people adjusted their doses to a drug that used to catch them off guard. “When fentanyl first came out, people were used to shooting heroin. That’s why they were dying,” he said. They’ve both survived multiple overdoses since.

Neither of them credited a task force. Neither mentioned the DEA. What they described was a population that had, through trial, death, and word of mouth, partly figured out how to survive a drug supply that had briefly stopped moving the target on them. That is a meaningfully different story than the one arriving from Washington this week, timed to Overdose and Drug Poisoning Awareness Week and the August 31 observance of International Overdose Awareness Day. It’s also, underneath the noise, closer to the truth.

Rize’s take: the overdose decline is real and it is saving real lives — but it is a supply shock being sold to the public as a policy win, and supply shocks have a documented history of getting more dangerous before they get safer.

The drop is real — read the actual numbers before anyone spins them

Start with what isn’t in dispute. The CDC’s National Center for Health Statistics now projects 68,641 drug overdose deaths for the 12-month period ending in February 2026 — a 12.1% decline from the same period a year earlier. That number is “provisional,” which is a specific technical word worth translating: it means the CDC is still waiting on death certificates to finish working their way through medical examiners and state registries, so the agency statistically adjusts for the ones it knows are missing. The CDC’s own guidance is blunt about the limits of that math — “true declines or plateaus in the numbers of drug overdose deaths across the U.S. cannot be ascertained until final data become available.” Provisional numbers have moved before. They can move again.

Layered on top of that baseline, other tallies are circulating with bigger, punchier percentages. VIN News reports fentanyl deaths down 54% from their 2023 peak, total 2025 deaths at 69,973 (about a 14% one-year drop), and a 35% reduction since the all-time high. The White House’s own message for Overdose Awareness Week cites yet another pairing — fentanyl deaths down 22%, overall overdose deaths down 14% — attributed almost entirely to enforcement, cartel designations, and the new HALT Fentanyl Act. Notice that none of these three headline numbers describe the identical twelve months against the identical baseline. That’s not necessarily dishonest — different agencies measure different windows — but it means every one of these figures was picked, from a menu of true numbers, for how it sounds. The number Washington keeps repeating this week is the one that credits people with badges, not chemistry.

The trend underneath all three versions, though, is the same trend, and it’s a genuinely large one. Deaths have now fallen for three consecutive years after peaking near 114,000 in 2023. Nobody serious disputes that fewer Americans are dying of overdose today than at any point since 2020. The argument worth having isn’t whether the decline is happening. It’s about who actually gets credit for it — and what that credit-taking obscures about how fragile the win might be.

Enforcement is taking a bow for a trend it mostly followed

The administration’s evidence for its own role is genuinely large in scale. The DEA’s Operation Fentanyl Free America reported that in one 30-day surge alone — January 12 to February 10, 2026 — agents seized more than 4.7 million counterfeit fentanyl pills and nearly 2,396 pounds of fentanyl powder, made over 3,000 arrests, and took guns, cash, and other drugs off the street in the same sweep. Those are real seizures with a real, if hard to quantify, effect on supply in specific cities during specific weeks.

Those are real seizures with a real, if hard to quantify, effect on supply in specific cities during specific weeks.

But timing is the problem for the enforcement narrative, and it’s a problem you can check yourself. The national curve didn’t start bending down in early 2026, or even in 2025, when this operation launched. It bent in the summer of 2023 — well before the current wave of task forces existed. Peter Reuter, a drug-policy researcher at the University of Maryland, laid out the arithmetic to STAT News: “From about 1980 to 2022, the number of fatal overdoses from all drugs went up about 20-fold. And then in June, July of 2023, suddenly the curve turns down.” Asked what could explain a one-third drop in two years, he didn’t hedge: “The notion that there’s anything on the demand side that would lead to a reduction of one-third in two years, that’s really stretching things. It has to be a supply-side shock.” His own read points less toward American arrests and more toward a crackdown by the Chinese government on the export of precursor chemicals — the raw industrial ingredients traffickers use to synthesize fentanyl — that made the pills reaching the street measurably weaker.

Not every researcher buys the supply-shock story whole. Nabarun Dasgupta, a pharmacoepidemiologist at the University of North Carolina, told STAT he finds the theory “highly compelling” but incomplete — deaths in Western states were already falling before mid-2023, a pattern the China-precursor timeline alone doesn’t explain. What both researchers agree on matters more than where they split: this is a chemistry-and-trafficking story first, a policy story second, and the policy apparatus taking a bow this week is mostly narrating a wave it didn’t generate. Enforcement can accelerate a decline already underway. It has a much weaker claim to having started one.

Afghanistan’s poppy ban is the quieter author of this story — and its clock is running out

The second real supply-side story getting less airtime than DEA seizure totals is happening 7,000 miles away. After the Taliban banned poppy cultivation in 2022, land under opium production in Afghanistan collapsed from 232,000 hectares to roughly 10,200 hectares by 2025, and annual opium output fell from about 6,200 tons to just 296 tons — a production crash of roughly 95%, according to the United Nations Office on Drugs and Crime. Afghanistan supplied roughly 80% of the world’s illicit opium before the ban. That’s the raw material heroin is made from — a plant-based drug, distinct from fully synthetic fentanyl, that still anchors use and treatment patterns for a large population of older users in the U.S.

Here’s the part that should temper anyone’s applause: UNODC’s own assessment is that Afghanistan’s remaining stockpiles will likely carry the global heroin market through the end of 2026 — meaning the ban’s full effect on price and purity hasn’t even reached the street yet. And UNODC’s warning about what happens after the stockpile runs dry is explicit: a tighter heroin supply is likely to accelerate a shift toward synthetic opioids, including fentanyl and newer, largely unstudied compounds called nitazenes, because synthetics are cheaper to make and don’t depend on a poppy field anyone can bomb, ban, or count from a satellite. The policy being celebrated this week — a shrinking opium supply — is, on its own logic, a mechanism for pushing more people toward the drug that killed 70,000-plus Americans a year at its peak. A falling number today can be the input to a rising number in 2027. That’s not cynicism. It’s the same supply chain, read one step further down the line.

A broken supply doesn’t get safer. It gets weirder, then meaner.

This is the pattern with the deepest evidence behind it, and it’s the one missing from the celebratory framing entirely. When a drug supply gets disrupted, it rarely disappears — it gets adulterated, diluted, or replaced by whatever’s cheapest and easiest to move, and the transition period is often more lethal than the stable supply it replaced. That’s already happening in real time. Xylazine — a veterinary sedative, not approved for human use, that causes severe skin wounds and doesn’t respond to naloxone — spread through the fentanyl supply over the past several years. Now it’s being displaced by something worse. Medetomidine, a related animal sedative more than 100 times more potent at the same receptor, went from essentially absent in Philadelphia’s drug supply to present in 87% of tested fentanyl samples by late 2024, according to a study published in Annals of Emergency Medicine. Between September 2024 and January 2025, 165 patients across three Philadelphia hospital systems were admitted for a withdrawal syndrome so severe — dangerously high blood pressure, racing heart, tremors — that the medications doctors normally use for fentanyl and xylazine withdrawal stopped working. Medetomidine has since turned up in drug supplies in at least 18 states.

Medetomidine has since turned up in drug supplies in at least 18 states.

Dr. Ben Cocchiaro, a Philadelphia physician who treats addiction patients in Kensington, put it to NPR in a phrase worth sitting with: the newer street drugs are “less lethal, but not less dangerous.” Fewer people are dying in a single moment of overdose. More people are being permanently disfigured, hospitalized for withdrawal syndromes with no established treatment, or locked into a physical dependency stranger and harder to detox from than anything the crisis has produced before. A death curve going down doesn’t automatically mean the drug supply got safer. Sometimes it means the drug supply started hurting people in ways that don’t show up as fast, or as cleanly, in a mortality statistic.

History already ran this experiment once, and the sequel wasn’t better

This isn’t a new story shape. In August 2010, Purdue Pharma pulled the original OxyContin and replaced it with a reformulated version engineered to be harder to crush and snort — an “abuse-deterrent” pill, celebrated at the time as a supply-side fix to the prescription opioid crisis. Prescription opioid misuse did stop climbing that month. But heroin overdose deaths started climbing the very next month, and kept climbing fastest in the places that had the most opioid users and the easiest heroin access to begin with. A National Bureau of Economic Research study later found the reformulation didn’t reduce combined opioid-and-heroin deaths at all — it substituted one for the other, roughly one-for-one. The “fix” didn’t shrink the crisis. It changed which drug was killing people, and heroin’s cheap, unregulated unpredictability turned out to be more dangerous than the pills it replaced. Within a few years, illicitly manufactured fentanyl began showing up cut into that heroin supply, and the death toll didn’t just recover — it went on to multiply many times over. A win got declared in 2010. The actual crisis was still fifteen years from its worst chapter.

If you’re in recovery right now, this is the part that’s about you

If you’ve got months or years of not using behind you, the good headlines this week are not really written for you, and it’s worth naming why. Your tolerance is lower than it was. The supply you’d be returning to is not the same supply you left — it’s a supply in the middle of a shock, cut with sedatives your body has never metabolized, moving through a period researchers openly admit they can’t fully predict. A “good month” nationally is not a promise about the bag someone hands you on a bad night. Overdose risk during and right after relapse has always run higher than during steady use, for the boring biological reason that your body forgot how much it used to tolerate — and a shrinking, unstable supply doesn’t make that math safer, it makes the variance worse. The quiet stretches are, historically, when the most dangerous batches arrive, precisely because nobody’s braced for them yet.

None of that is a reason for despair, and it isn’t the whole picture either. Shatterproof, one of the country’s leading addiction advocacy organizations, reviewed this same 2025 CDC data and called the progress “fragile — and it is absolutely at risk,” which is a more honest sentence than anything in this week’s press releases, precisely because it doesn’t pretend the danger has passed. Naloxone is still free at pharmacies and community programs in nearly every state, still federally funded, and it still works on fentanyl even when the numbers around it are shifting under everyone’s feet. That fact doesn’t get undone by a bad batch, a defunded program, or a press conference. It’s still true tonight, whatever the drug supply is doing this week.

Back to Kensington

Elena told NPR she has hope, and then, almost as a footnote to her own sentence, added why: “I have to, you know?” That’s not optimism about a policy. It’s a survival requirement, stated by someone whose read on the death toll came from watching her own block, not from a CDC dashboard four months behind real time. She and Vadim are both still alive because naloxone worked on them more than once, not because a trafficking route into their neighborhood got permanently severed. The supply that nearly killed them repeatedly is the same supply getting credited, this week, with a historic policy win.

The supply that nearly killed them repeatedly is the same supply getting credited, this week, with a historic policy win.

The honest version of this week’s news is that a genuinely large number of Americans are alive right now who would not have been three years ago, and that is worth saying plainly, without qualification, to anyone who has spent this year afraid for someone they love. It is also true that the mechanism behind that number is a fracture in a criminal supply chain — fragile, foreign-sourced, and already showing the adulterated, meaner replacement chemistry that these fractures reliably produce. Both of those sentences are true at the same time. Anyone asking you to pick only the first one, the week before International Overdose Awareness Day, is not being honest with you about which one your safety actually depends on.