Date: 2026-08-03. Builds on Phase 0's overdose-peak/decline anchor (row 1); does not re-derive it.
1. The fentanyl supply transition
Illicitly manufactured fentanyl began displacing heroin in the Northeast/Midwest/South starting ~2013–2014 (CDC MMWR 66(34), 2017); the West Coast lagged ~4–7 years, with fentanyl's rise there dated to 2017–2020 — consistent with Phase 0's finding that the West peaked later (~2023) than the rest of the country (Oct 2022). By 2023, fentanyl-heroin co-occurrence reaches 91–95% in the hardest-hit states (NJ, MI); heroin's independent share of overdose deaths has fallen to **~4%** nationally. Of 2024's 54,743 opioid-involved deaths, 88% involved fentanyl/synthetic opioids (CDC NCHS Data Brief 549).
A live counter-signal worth flagging for §5/§10: DEA's 2025 National Drug Threat Assessment reports fentanyl purity/potency trending down and detection down >12% from its 2022 peak, while heroin co-use surged ~18% nationwide 2023→2024 — the first significant heroin uptick in nearly a decade, plausibly a substitution effect as fentanyl became less available.
2. Polysubstance mortality
Per CDC's most recent MMWR (Tanz et al., Aug 2025, 309,274 deaths across 49 states+DC, Jan 2021–June 2024): opioid+stimulant co-involvement is 43.1% of overdose deaths — larger than "opioid alone" or "stimulant alone" as single categories, making it arguably the modal death type without being an outright majority. Stimulant-any involvement is 59.0%. Cocaine and psychostimulant (mainly meth) age-adjusted death rates roughly doubled 2018–2023 (4.5→8.6/100k and 3.9→10.4/100k respectively), with the largest increases among American Indian/Alaska Native and Black populations — consistent with Phase 0's AZ/NM/CO and Native American mortality-disparity finding.
3. Naloxone
FDA approved OTC naloxone (Narcan 4mg) March 2023, on retail shelves September 2023; a generic and a second branded product (RiVive) followed in July 2023. Distribution scale is genuinely fragmented — no current national total exists across all channels (pharmacy, OTC retail, harm-reduction orgs, state programs). Last comprehensive estimate: ~17M doses distributed in 2021, 45% outside retail pharmacies/healthcare facilities.
Naloxone's causal contribution to the 2023–2025 decline is asserted, not demonstrated. Both quantitative decomposition studies located (Dowell et al. 2025, Lancet Regional Health Americas; Vangelov/Reuter/Humphreys et al. 2026, Science) treat naloxone as a secondary/contributing factor without an isolated counterfactual estimate — the primary quantified drivers in both papers are supply-side and exposure-side, not demand-side interventions.
4. Drug checking / fentanyl test strips
45 states + DC have decriminalized fentanyl test strips as of late 2023/2025. Evidence of behavior change is real (43% report safer use after a positive result) but no national usage-scale figure could be verified.
5. Root-cause explanations for the 2023–2025 decline — the contest, evidence-graded
Two studies offer actual quantitative decomposition, not just narrative plausibility:
- Dowell et al. 2025 (Lancet Regional Health Americas): counterfactual modeling attributes ~110k counterfactual deaths to a shrinking at-risk population (NSDUH-measured OUD prevalence fell 2.1M→1.6M, 2016–2019) and ~260k to the fentanyl market reaching a saturation plateau.
- Vangelov, Reuter, Humphreys et al. 2026 (Science): triangulates mortality data, DEA seizure/potency data, and online drug-market chatter to argue a fentanyl supply shock — most likely from Chinese precursor-chemical restrictions — cut the fentanyl-death rate by "at least one-third" over two years. Notably, Keith Humphreys — often cited as the "natural depletion" skeptic — is a co-author of this supply-side paper, not a standalone depletion advocate.
QUALIFIED 2026-08-10 by the Phase 1 verification pass (verification-log.md). A third analysis now cuts against the supply-shock reading and postdates this workstream: Dasgupta, Sibley, Gildner, Combs, Post, Tobias, Kral & Pacula, Fentanyl Purity and Overdose Decline: A Reexamination of Geographic Trends (medRxiv 2026, PMID 42078391 — a preprint, not peer-reviewed). It replicates the purity analysis with regional data and finds the national purity series did not track overdose increases in most regions, that the later co-decline is equally visible in unrelated macroeconomic indicators sharing the same time pattern, and that DEA purity data carry differential selection bias. §13 already directed that "the national decline should therefore be described as multi-cause and unresolved, not as a one-cause supply-shock result"; that instruction is applied here and to the public pages. "The two studies that actually decompose the cause" also understates the literature — there are now at least three analyses in play.
Both converge on supply-side/exposure-side mechanisms as better-evidenced than demand-side interventions (naloxone, treatment, FTS) for explaining the specific timing and magnitude of the decline — even though naloxone, treatment, and FTS all have independently-established general effectiveness. Genuine unresolved tensions remain, and were not smoothed over by either paper: fentanyl prices reportedly fell (~$100–140/g → $50–80/g) over the same window, which is the wrong direction for a real supply shortage; the West Coast's later peak timing doesn't cleanly fit the mid-2023 China-crackdown story (a named critique from UNC's Nabarun Dasgupta); and 2022→2023 CDC data showed deaths still rising in American Indian/Alaska Native and Black communities even as the national/white-non-Hispanic rate fell — inconsistent with a uniform depletion story.
A negative finding worth carrying into §4: a NEJM study (relayed secondhand, not independently verified — flag before citing) reportedly found no change in the monthly number of buprenorphine patients after the 2023 X-waiver removal, undercutting treatment-expansion as an explanation for the decline's timing specifically, separate from treatment's general effectiveness.
Implications for §10/§11
- The decline is most credibly explained by forces outside any single policy lever (supply shock, shrinking at-risk population) — §10 architectures that claim credit for the decline, or that are scored against a static "if this program existed, X deaths would be averted" baseline, need to account for this counterfactual explicitly, not assume the current trajectory continues on its own.
- The West Coast's later fentanyl arrival and later peak means any national policy evaluation window must be geography-aware — AZ/NM/CO's continued rise is not noise, it's the tail of the same supply transition arriving late.
- The naloxone/FTS legal-access story (largely won already — OTC approval, 45-state FTS decriminalization) is a completed policy fight whose marginal future value is now about last-mile distribution and public-order politics (safe consumption sites), not further deregulation.
Source register
See the full source register in the underlying research pass (30 sources, primary/administrative and peer-reviewed weighted). Two items flagged for independent re-verification before citing in the whitepaper: the NEJM buprenorphine-null-finding claim and the fentanyl price-decline figures — both relayed secondhand via opioiddata.org rather than fetched from primary sources.
Confidence: Moderate
Strong on fentanyl geography and polysubstance composition (primary CDC/DEA/peer-reviewed data). Weaker on naloxone distribution scale (fragmented, no current national total) and on several root-cause sub-claims relayed through a single independent-researcher blog rather than fetched primary sources.