Prohibition and Potency: the Analogue Treadmill

This page is analysis and advocacy — it is not legal advice

Read this before you read anything else on the page. What follows is a research and policy brief. It is written to be used by regulators, legislators, counsel and industry researchers as an argument, and the argument is stated at full strength because a hedged brief is useless to the people who need it. It is not a legal opinion, it is not a clearance, and it is not advice to you. Three things make that more than a formality. First, cannabinoid law varies enormously by state: a compound that is lawful under federal law can be a felony in the jurisdiction where you live, ship, store or sell, and state analogue statutes, total-intoxicant caps and compound-by-name bans operate independently of the federal scheme described here. Second, this area of law is changing quickly — the federal definition of hemp itself was rewritten in late 2025 with a delayed effective date, state legislatures have moved in both directions in every recent session, and agency positions have shifted. Anything written here can be overtaken by an amendment, a rule, or a decision. Third, an argument that a substance is not a controlled substance analogue is a defence, not a permission: it is something a defendant says after being charged, and the fact that it is a good argument does not mean nobody will be charged. Do not make a business decision, a product decision or a personal decision on the basis of this page. Get your own lawyer, licensed in your own jurisdiction, and give them the actual facts. This page is a policy and historical argument built on the scheduling record. It is not a claim about what any compound's current legal status is, in any jurisdiction, and it is certainly not a suggestion that anyone act on the market dynamics it describes.

• Not legal advice, and no lawyer-client relationship arises from reading it.
• Federal lawfulness does not imply state lawfulness — check your own state, and the states you ship to.
• The law described here is in motion; a delayed federal effective date sits in the middle of it.
• A strong non-analogue argument is a defence to a charge, not immunity from one.
• Nothing here is medical advice either, and nothing here is an endorsement of using any compound.

The claim, and why the Federal Register is the right evidence for it historical / ethnographic

The claim is narrow enough to be tested and general enough to matter: where a market for an intoxicant is subject to compound-by-compound prohibition, each prohibition of a named compound displaces the market onto a successor with more effect per unit of detectable mass, and the successor is by construction the one with less human toxicology, no reference standard and no case literature. The market moves up the potency and danger curve rather than out of existence. This is the dynamic Richard Cowan named the iron law of prohibition in 1986, writing about crack, and that Beletsky and Davis restated for the fentanyl era in 2017: interdiction rewards concealability, concealability is potency per unit volume, and potency per unit volume is exactly the property that makes dosing lethal. In the novel-psychoactive-substance literature the same thing is usually called the analogue treadmill, and it is treated as an operational nuisance rather than as a harm mechanism. It is a harm mechanism. What makes this argument worth making in a research library rather than in a pamphlet is that the evidence for it is not advocacy material, an industry estimate or a modelled projection. It is the Federal Register: a dated, official, citable record in which the government documents each compound it schedules and, by the same act, names the generation that followed the last one. The record is not being read against the government's intentions — the orders say plainly that each new set of compounds appeared in commerce after the previous set was controlled. Every volume-and-page citation on this page was resolved against the Federal Register API before it was written down; nothing here is reconstructed from memory or from secondary reporting.

• The claim is about the direction of compound substitution, not about the size of the demand effect.
• Mechanism: interdiction selects for effect per unit of detectable mass, which is also what makes a dose hard to control.
• Evidence: the scheduling record itself, in date order, from the agency that did the scheduling.
• Named twice in the literature: the iron law of prohibition (Cowan), the analogue treadmill (the NPS literature).

Sources: Cowan RC 1986* · Beletsky L 2017 · Andrews R 2022 · Banister SD 2018

Case one: the synthetic cannabinoids, turn by turn human data

This is the case the safety shelf reconstructs from the clinical side, and it is the cleanest of the three because the generations are structurally distinct and the clinical severity signal is unambiguous. On 1 March 2011 the Drug Enforcement Administration issued a final order temporarily placing five compounds in schedule I — JWH-018, JWH-073, JWH-200, CP-47,497 and the CP-47,497 C8 homologue. Those same five were placed in schedule I permanently on 9 July 2012, by statute rather than by rule: the Synthetic Drug Abuse Prevention Act of 2012, title XI subtitle D section 1152 of Public Law 112-144, which also wrote a structural-class definition of cannabimimetic agents into the Controlled Substances Act. The agency's implementing rule of 4 January 2013 records that the Act superseded its own pending permanent-scheduling rulemaking for those five. What followed is the whole argument. By 16 May 2013 the agency was issuing a temporary order for UR-144, 5-fluoro-UR-144 (XLR11) and APINACA (AKB48) — a different acyl group on the indole core. By 10 February 2014 it was issuing one for PB-22, 5F-PB-22, AB-FUBINACA and ADB-PINACA, which is the arrival of the indazole carboxamides, made permanent in September 2016. AB-FUBINACA is the compound behind the Brooklyn mass-casualty event of 12 July 2016, characterised in the New England Journal of Medicine: thirty-three people in one neighbourhood presenting simultaneously with a strikingly uniform profound-sedation picture from one packaged herbal product. That event happened two and a half years after the order that named its parent compound, and in a generation of molecules that existed in commerce because the previous generation had been scheduled. The cycle has not closed. On 12 December 2023 the agency temporarily scheduled MDMB-4en-PINACA, 4F-MDMB-BUTICA, ADB-4en-PINACA, CUMYL-PEGACLONE, 5F-EDMB-PICA and MMB-FUBICA; in October 2025 it proposed permanent placement for the first of those. Fifteen years, at least four generations, each one less characterised than the last, and at no point in the record does the market disappear.

Sources: Drug Enforcement Administration 2011 · United States Congress 2012 · Drug Enforcement Administration 2013 · Drug Enforcement Administration 2013 · Drug Enforcement Administration 2014 · Drug Enforcement Administration 2016 · Drug Enforcement Administration 2023 · Drug Enforcement Administration 2025 · Adams AJ 2017 · Trecki J 2015 · Andrews R 2022 · Banister SD 2018

What changed between the cannabinoid generations — and what must not be claimed about it contested in vitro

The selection argument needs a statement about how the generations differed, and that statement is easy to overstate into something false, so it is worth stating carefully once. The defensible version: the early JWH-era products had an effect profile and a case-report severity markedly different from the later indazole carboxamides. That is not the proposition that the early compounds were relatively safe, and this shelf does not advance the second. JWH-018 is a full agonist at CB1 with no ceiling effect; it produced seizures, tachyarrhythmias, myocardial infarction in young people without coronary disease, acute kidney injury and deaths, from the beginning. What changed was degree and shape. Degree: the indazole carboxamides are substantially more potent again on in-vitro functional measures, which compressed the margin between an active and a harmful quantity and reduced the mass needed for a severe outcome below what any distribution method in actual use could control. Shape: the characteristic event of the later era is the burst — a cluster of severe, uniform presentations from one product, in one place, at one time — which is what happens when a compound is potent enough that ordinary batch heterogeneity becomes lethal heterogeneity. The conclusion the evidence supports is therefore not that prohibition made a safe thing dangerous. It is that prohibition made a dangerous thing more dangerous, repeatedly, in a documented direction, while never removing it from the market. That is a smaller claim than the rhetoric usually reaches for and it is the one that survives contact with the clinical literature.

• NOT claimed: that the JWH-era products were safe. They were full CB1 agonists with a documented record of seizures, cardiac events and deaths.
• Claimed: a marked difference in effect profile and case-report severity between the generations.
• Degree — later generations more potent on in-vitro functional measures, compressing the active-to-harmful margin.
• Shape — mass simultaneous events, the fingerprint of lethal heterogeneity inside a single batch.
• Supportable conclusion: prohibition made a dangerous thing more dangerous, repeatedly, without removing it.

Contested — caveat. FLAGGED. Any assertion that the early JWH-era products were "relatively safe" is unsupported and is not made here. The potency comparison between generations rests on in-vitro functional and binding assays, since no human dose-response data exist for most of these compounds. The generational contrast in case severity is drawn from case series and surveillance collected under different reporting regimes in different years, which is not a controlled comparison, and the later generations were also present in a market with more testing and more surveillance attention, which affects what got reported.

Sources: Banister SD 2018 · Andrews R 2022 · Adams AJ 2017 · Trecki J 2015 · Drug Enforcement Administration 2014

Case two: the phenethylamines, and the NBOMe series as the endpoint contested human data

The cannabinoid case could be dismissed as a quirk of one chemical class, so the second case matters more than the first for establishing that this is a general law. The substituted phenethylamines were controlled one compound at a time over nearly twenty years, and each step is on the record. 2C-B was temporarily placed in schedule I on 6 January 1994 and permanently on 2 June 1995. 2C-T-7 was temporarily placed on 20 September 2002 and permanently on 18 March 2004. Alongside them the tryptamines moved: alpha-methyltryptamine and 5-MeO-DIPT were temporarily placed on 4 April 2003 and permanently on 29 September 2004. Then, on 9 July 2012, the same statute that permanently scheduled the five synthetic cannabinoids placed nine 2C compounds in schedule I in one action — 2C-E, 2C-D, 2C-C, 2C-I, 2C-T-2, 2C-T-4, 2C-H, 2C-N and 2C-P — which the agency's own implementing rule enumerates. That coincidence is worth pausing on, because it is the same instrument, on the same day, doing the same thing to two unrelated chemical classes. And sixteen months later, on 15 November 2013, the agency temporarily scheduled 25I-NBOMe, 25C-NBOMe and 25B-NBOMe: the N-(2-methoxybenzyl) derivatives of 2C-I, 2C-C and 2C-B respectively, made permanent in September 2016. The NBOMe compounds are the point of the case. They are derivatives of the very compounds the preceding actions had controlled, they are active in the low hundreds of micrograms where their 2C predecessors are active in the tens of milligrams, and the published clinical review literature documents fatalities and severe toxicity — seizures, hyperthermia, agitated delirium, rhabdomyolysis — at doses at which the predecessors would not have produced those outcomes. They also reached users on blotter, a presentation indistinguishable from LSD, which removed the last defence a user had: knowing which compound they had. Potency per unit rose at every step in this sequence, and the step with the largest potency increase is the one that came after the largest single scheduling action.

Contested — caveat. The scheduling dates and compound identities are verified against the Federal Register. The causal reading — that each action produced the next generation rather than merely preceding it — is an inference from sequence, and the NBOMe series had other drivers too, including online retail and the blotter presentation. The potency comparison between the 2C and NBOMe series, and the claim that NBOMe fatalities occurred at doses at which the predecessors would not have, come from a review of case reports (Suzuki et al. 2015) rather than from controlled dose-response work, and case-report literature systematically over-represents severe outcomes.

Sources: Drug Enforcement Administration 1994 · Drug Enforcement Administration 1995 · Drug Enforcement Administration 2002 · Drug Enforcement Administration 2004 · Drug Enforcement Administration 2003 · Drug Enforcement Administration 2004 · United States Congress 2012 · Drug Enforcement Administration 2013 · Drug Enforcement Administration 2013 · Drug Enforcement Administration 2016 · Suzuki J 2015

Case three: the opioids, where the mechanism is least deniable contested human data

The third case is the one nobody disputes the direction of, which is why it is the most useful. The sequence runs opium to morphine to heroin under supply and enforcement pressure, then heroin to fentanyl, then fentanyl to the fentanyl analogues, then to the 2-benzylbenzimidazole opioids known as nitazenes. Each step is a large increase in potency per unit of mass, and potency per unit of mass is the property that decides whether a dose can be measured at all outside a laboratory. The scheduling record for the last two steps is precise. Through 2017 and into January 2018 the agency issued temporary orders for individual fentanyl analogues one at a time — acryl fentanyl in July 2017, ortho-fluorofentanyl, tetrahydrofuranyl fentanyl and methoxyacetyl fentanyl in October 2017, cyclopropyl fentanyl in January 2018 — and then, on 6 February 2018, abandoned the compound-by-compound approach and issued a class-wide temporary order covering fentanyl-related substances as a structural class. That order is itself an admission that compound-by-compound scheduling had failed to keep pace. What arrived after it was a structurally unrelated class: isotonitazene, temporarily scheduled on 20 August 2020, followed by seven further nitazenes on 12 April 2022. The nitazenes are not fentanyl analogues and were therefore outside the class order; the primary pharmacology, characterised by Vandeputte and colleagues in 2021, reports µ-opioid receptor activity for several of them in the range of, and for some exceeding, fentanyl. Beletsky and Davis made precisely this argument in 2017, before the nitazenes had arrived, on the basis of the iron law: interdiction selects for potency per unit volume, and the prediction it generates is a supply that becomes progressively harder to dose. The opioid case is also the one where the consequence is measured in a mortality series rather than in case reports, which is why it carries more evidential weight than the other two even though the compounds are further from this shelf's subject matter.

Contested — caveat. The opioid sequence has drivers beyond interdiction — prescribing patterns, distributor conduct, the economics of transnational shipping, and precursor availability all contribute, and attributing the whole of it to enforcement pressure would overstate the case. What the record supports is that interdiction is one selection pressure among several and that it pushes in the direction of potency per unit volume. The receptor-activity comparison between nitazenes and fentanyl is from in-vitro functional work and should not be read as a human potency ratio.

Sources: Drug Enforcement Administration 2018 · Drug Enforcement Administration 2020 · Drug Enforcement Administration 2022 · Vandeputte MM 2021 · Beletsky L 2017 · Cowan RC 1986*

Enforcement architecture: the Rannazzisi case, and where the pressure was actually aimed contested historical / ethnographic

The three case histories establish that compound-by-compound prohibition selects for worse compounds. This section is about a different and complementary failure, and it is the operator's specific reference: what happens when enforcement is removed from the layer that has the volume and the records, while pressure stays on the layer that is compliant, documented and taxed. Joe Rannazzisi ran the DEA's Office of Diversion Control, the division that regulates and investigates manufacturers and distributors, and led roughly a decade of aggressive enforcement against distribution before losing his responsibilities and resigning in 2015. In the joint 60 Minutes and Washington Post investigation published in October 2017 he set out his account: prosecuting individual prescribers and pharmacists was not stemming the epidemic, so the division moved up the supply chain to the distributors, on the reasoning that the distributors were the choke point — and he names the three largest as Cardinal Health, McKesson and AmerisourceBergen, which on his estimate control something in the region of eighty-five to ninety percent of the volume going downstream. He also opposed the legislation that became the Ensuring Patient Access and Effective Drug Enforcement Act of 2016, warning that it would cripple the agency's ability to act against rogue distributors. On that last point the statute can be read directly rather than taken on his characterisation of it, which is the better way to use this case. Public Law 114-145, signed 19 April 2016, inserted into 21 U.S.C. § 824(d) a definition of "imminent danger to the public health or safety" — the standard for immediately suspending a registration, which is the instrument for stopping suspicious shipments — requiring a substantial likelihood of an immediate threat that death, serious bodily harm or abuse will occur absent immediate suspension. It also added an opportunity for a registrant to submit a corrective action plan before revocation or suspension. Whatever one concludes about the merits, that is a higher and more specific threshold than the undefined phrase it replaced, and it applies to the layer with the shipment records. The analytical point for this shelf is the shape of the thing rather than the opioid policy: pressure was relieved precisely where the volume, the documentation and the tractable enforcement target were, while it continued to fall on prescribers and patients — people whose conduct was documented and who were therefore easy to act against. That is the same misdirection the 0.4 mg per-container threshold repeats in a different market: it regulates the layer that tests, labels, invoices and pays tax, and leaves the harm vector — an untested channel with no analytical layer in it — untouched and, on the substitution argument, enlarged.

• Rannazzisi ran the DEA division that regulates and investigates manufacturers and distributors; he lost his responsibilities and resigned in 2015.
• Three distributors — Cardinal Health, McKesson, AmerisourceBergen — hold the large majority of downstream volume, on his estimate roughly 85 to 90 percent.
• Pub. L. 114-145 (2016) defined "imminent danger" in 21 U.S.C. § 824(d), raising the standard for immediate suspension, and added a corrective-action-plan step.
• Read the statute, not only the characterisation: the text is the primary evidence and it is short.
• The transferable shape: relieve pressure where the volume and the records are, keep it where conduct is documented and easy to act against.
• The 0.4 mg threshold repeats it — regulate the tested, labelled, taxed layer; leave the untested channel alone.

Contested — caveat. Rannazzisi's account, including the 85-to-90-percent distribution estimate and the claim that the 2016 Act was intended to or did cripple enforcement, is his testimony as reported in the joint investigation, not a finding. The distributors disputed the characterisation. The 85-to-90 figure is his estimate, not a measured market share. The Washington Post half of the investigation could not be reached from this environment, so it is recorded as an unverified citation and the CBS record is the one actually confirmed. What is verified independently is the statutory text: Public Law 114-145 § 2 inserted the definition of "imminent danger to the public health or safety" into 21 U.S.C. § 824(d) and added the corrective-action-plan opportunity. Whether that change made enforcement harder in practice is contested and is not settled here.

Sources: United States Congress 2016 · Whitaker B (correspondent) 2017 · Bernstein L 2017* · United States Congress 2025 · Beletsky L 2017

The predicted consequence for the 0.4 mg threshold: exposure moves to the inputs contested

The operator's extension of the argument is a prediction about what a harsh penalty attached to a finished article does to buyer behaviour, and it follows from the same mechanism as the rest of the page. Criminal exposure is a cost, and buyers minimise costs. If possessing a finished quantity of a controlled compound carries a serious penalty while the uncontrolled inputs to it carry none, then the total expected cost of obtaining an effect is lower for someone who buys inputs and performs the last step themselves than for someone who buys the finished article — not because the inputs are cheaper, but because the legal exposure attached to them is smaller. The prediction is therefore that a threshold of this severity moves demand from finished, labelled, tested product toward input sets and toward end-user preparation, and moves the point at which a mistake can happen from a facility with some analytical capability to a person with none. The enforcement-gap section of the market-consequences page already establishes the input half of this: the plant-derived starting cannabinoids have been lawful commodities in commerce since 2018 and are held in quantity, and the other input classes are ordinary articles of commerce. This shelf states the market and legal-exposure argument and stops there, deliberately and completely. It does not describe what any such input set would consist of, does not name a precursor as something to obtain, does not describe a route, and does not treat any of this as an opportunity. The legitimate content of the point is exhausted by the sentence a regulator needs to hear: a rule whose penalty falls on the finished article and not on the uncontrolled inputs prices legal risk into the tested layer and out of the untested one, which is the opposite of the intended effect. That is the same structural error as the misdirected enforcement in the previous section and the same structural error as compound-by-compound scheduling in the three cases above — an instrument aimed at what is legible rather than at what is harmful.

Contested — caveat. This is a prediction, not a finding. No data are offered here on the size of any such shift, and the three historical cases on this page concern compound substitution within a supply chain, not a shift from finished product to end-user preparation — so they support the mechanism but are not evidence for this particular prediction. A regulator would fairly answer that end-user preparation requires effort, equipment and tolerance for risk that most consumers of a convenience-retail product do not have, and that the shift may therefore be small.

Sources: United States Congress 2025 · Drug Enforcement Administration 2020 · Kiselak TD 2020 · Beletsky L 2017 · United States Congress 2016

What this argument does not establish contested

A page that proves more than its evidence supports is easier to dismiss than one that marks its own limits, so the limits are stated here rather than left for an opponent to find. It does not establish that aggregate consumption rises under prohibition, or even that it stays constant: nothing on this page measures demand, and the honest position is that reduced legal availability probably does reduce aggregate use, particularly among casual and young users, which is the strongest thing a regulator has to say and is set out on the market-consequences page. It does not establish that any particular threshold or ban will fail on its own terms. It does not establish that the compounds that arrive next will always be more dangerous — occasionally a successor is less potent, and the record contains structures that appeared and did not persist. It does not establish causation in the strict sense for any single pair of scheduling action and successor generation; what it establishes is a direction, repeated across three chemically unrelated classes over three decades, which is the kind of evidence a policy argument is entitled to use and the kind a causal claim about one event is not. And it does not follow from any of this that scheduling is never the right instrument. What does follow is narrower and is the whole point: an instrument that names structures selects for uncharacterised structures, and an instrument that regulates the testable layer relieves the untestable one — so where the aim is to reduce harm rather than to register disapproval, the instrument has to be aimed at identification, potency-weighted quantity, labelling and traceability, which are properties of a product rather than names of molecules. That is the same conclusion the market-consequences page reaches from the arithmetic, and it is worth more having been reached twice from different evidence.

• Not established: that aggregate consumption rises, or stays constant, under prohibition.
• Not established: that any particular threshold will fail on its own terms.
• Not established: strict causation for any single scheduling-action-and-successor pair.
• Established: a repeated direction of compound substitution across three unrelated classes over three decades.
• Follows: aim the instrument at identification, potency-weighted quantity, labelling and traceability — properties, not names.

Contested — caveat. This section is the page arguing against itself, deliberately. The counter-argument a regulator would make — that availability is a gradient rather than a switch and that partial displacement can be a net public-health gain — is stated at more length on the market-consequences page and is not answered here.

Sources: Beletsky L 2017 · Andrews R 2022 · Banister SD 2018 · United States Congress 2025 · Congressional Research Service 2025