Calibrating the Gate: Criteria for Assigning Marketed Drug Products to a Graduated-Access Architecture in the United States

1. Introduction

The categorical separation of medicines into prescription and nonprescription classes is a twentieth-century legal construction rather than a feature of medicine. Temin [1] traces its origin to the Federal Food, Drug, and Cosmetic Act of 1938 and decisively to the 1951 Durham–Humphrey Amendment, which drew the line in its modern form [2]. For most of American pharmaceutical history, the consumer, not a licensed intermediary, decided.

The arrangement that resulted has costs that are now well documented on both sides of the gate. Watanabe and colleagues estimate that prescription drug-related morbidity and mortality accounts for over 275,000 deaths annually when harm from drugs prescribed but not taken is counted alongside harm from drugs prescribed and taken [3]. Jiang and colleagues estimate that 1.3 million Americans ration insulin on cost grounds [4]. American prices average roughly three times those in 33 peer countries [5]. Emergency department surveillance continues to document a substantial burden of medication harm inside the regulated channel [6].

Peer democracies have operated graduated alternatives for decades. Canada’s national drug schedules [7], the United Kingdom’s prescription-only, pharmacy, and general-sale classes [8,9], and Australia’s poisons schedules [10] each replace the binary with a cascade. A pharmacist-mediated middle tier has been proposed in the United States since the 1984 ibuprofen switch debate, and the Government Accountability Office examined it twice without endorsement [11,12].

The 2024 additional-condition rule is the first federal instrument to create anything resembling a conditioned middle category [13], and the agency has since sought comment on increasing nonprescription access more generally [14,15]. The published commentary on the rule is thin and treats it primarily as an access question [16].

The gap this paper addresses is not the case for graduated access, which will be argued elsewhere on empirical, economic, constitutional, and institutional grounds [17], nor the governance consequences of the additional-condition pathway specifically [18]. It is the missing operational instrument. A tiered architecture is only as good as the procedure that populates it, and no proposal in the American literature specifies how the existing pharmacopoeia would be reviewed, on what criteria, by whom, with what burden of proof, or in what order. This paper constructs that procedure and applies it.

2. Materials and Methods

This is a document-based policy analysis. Primary sources comprise the United States statutory and regulatory record governing drug classification [2,13]; the published classification criteria of three peer regulators and the European Union [10,19,20,21]; the two Government Accountability Office evaluations of a behind-the-counter class [11,12]; the record of United States and United Kingdom reclassification decisions; and the clinical and health-services literature bearing on self-assessment, self-monitoring, and cost-related nonadherence.

Three conventions govern the presentation. First, criteria are distinguished from applications of criteria, and the distinction is marked in the text. Second, every tier assignment proposed in Section 3.8 is stated with the factor that produces it, so that a reader who rejects the assignment can identify which criterion is in dispute. Third, claims resting on a single secondary source, or on a regulatory precedent whose date has not been verified against the primary record, are identified as requiring verification.

The analysis makes no claim about the motives of any participant. It concerns the functions a classification criterion performs and the incentives its structure creates, which are observable independently of intention.

3. Results

3.1. The Threshold Problem

A review of the marketed pharmacopoeia has no baseline to inherit. The current prescription class was not assembled by applying criteria drug by drug. The Durham–Humphrey test contains a prong limiting a drug to prescription use if the terms of its approved application so provide [2], which means the great majority of the class carries prescription status because its application said so at approval, not because any adjudicator asked whether the substantive test was satisfied.

The review is therefore a first application of criteria to a category populated largely by default, not a re-litigation of prior determinations. That framing governs burden allocation throughout.

3.2. Four Design Decisions That Precede Any Substantive Criterion

3.2.1. Unit of Classification

Assignment should attach not to the active pharmaceutical ingredient (API) solely but to API × route × strength band × indication, with the API inheriting the most restrictive assignment among its unclassified variants. European practice already recognizes that maximum single dose, strength, pharmaceutical form, and packaging can themselves make supply without prescription appropriate for a substance that would otherwise meet the prescription criteria [20]. Without a disaggregated unit, low-dose oral formulations are held hostage to high-dose parenteral ones and the review collapses into the binary it exists to replace.

3.2.2. Direction of the Default

Every criterion is written as an affirmative showing required to impose a restriction, never as a showing required to remove one. Absence of evidence resolves downward. This is the most consequential choice in the framework, because in a review of the whole pharmacopoeia the modal finding will be that no relevant evidence exists in either direction.

3.2.3. Combination Rule

Cascade, not score. A product is assessed against the Tier 1 factors first. If none applies, then against Tier 2. If none applies, it falls to Tier 3. This is the structure Canada and Australia already use [10,19]. A weighted index is the wrong instrument as it permits a strong showing on one dimension to offset a catastrophic showing on another, invites gaming by whoever assembles the dossier, and produces a false precision that would not survive the first contested assignment.

3.2.4. Who Produces the Evidence

Under the present arrangement, classification review is manufacturer-initiated, which is why the direction of motion has been asymmetric [22,23]. Standing to petition should extend to the agency on its own motion, to pharmacy and physician organisations, to payers, and to patient organisations, with agency-funded evidence generation for off-patent molecules that no sponsor has commercial reason to move.

3.3. The Substantive Cascade

A working template exists. The Canadian national scheduling factors turn Schedule I on: (1) whether the need for the drug is identifiable only by the prescribing practitioner; (2) whether use requires adjunctive therapy or evaluation, whether appropriate use may produce dependency; (3) whether serious adverse reactions occur at normal therapeutic dosage; (4) whether the margin between therapeutic and toxic dose is narrow; (5) whether serious interactions are known; (6) whether use contributes to resistant organisms; and (7) whether the ingredient is new with the consequences of widespread use not established [19]. Those factors were written to sort a category rather than to discipline a burden of proof, and are adapted rather than adopted. Table 1 states the adapted Tier 1 set where any one factor is sufficient, and each requires affirmative evidence.

The immaturity factor does substantial work. It places every new molecule at Tier 1 by default, to remain there until it earns its way down, which answers the public-health objection to graduated access without conceding anything about molecules carrying 40 years of exposure data.

Tier 2 is reached only when no Tier 1 factor applies. Adapted from the Canadian Schedule II and III sets [19], the triggers are: (1) the condition is new to patient self-assessment or generally not amenable to it; (2) use may mask or delay recognition of serious disease; interactions or adverse reactions exist that labelling alone cannot adequately convey; (3) there is meaningful potential for misuse short of the Tier 1 threshold; or (4) initial need requires professional confirmation while continuation does not. Tier 3 is the residual, requiring no affirmative showing and subject to mandatory labelling and post-market surveillance.

3.4. Selecting the Tier 2 Condition

Tier 2 is not a single arrangement. The available instruments, such as pharmacist consultation, structured questionnaire, verified laboratory input, home-test linkage, pack-size limit, strength cap, and registry enrollment, differ substantially in cost, access burden, and the information they generate. Three rules discipline the choice. Each condition imposed must be tied in the classification decision to a specific identified failure mode, and conditions imposed for general caution are not permitted. Where two conditions address the same hazard, the one imposing the lower access cost is required. Pack size is systematically underused, and European guidance already recognizes small pack size as a safeguard against overdose or delayed care-seeking [21]. And conditions are justified individually rather than as a bundle, which is the rule that prevents Tier 2 from becoming Tier 1 with additional steps.

The governance requirements developed for the additional-condition pathway [18] attach here as conditions of assignment, that is, portability of the information a condition collects, aggregate exclusion reporting, referral neutrality, and purpose limitation. Enforcement experience with consumer-facing health companies operating outside the Health Insurance Portability and Accountability Act indicates that these cannot be left to privacy policy [24].

3.5. The Access-Cost Criterion

No existing scheduling system scores the harm of withholding, and the omission is the structural defect this paper identifies. The four comparator frameworks all ask what harm the drug may do, but none asks what harm the restriction does. The European test is explicitly one-sided, like danger when used correctly without supervision, frequent incorrect use, activity or adverse effects requiring further investigation, or normal administration by injection [20]. Nothing in it registers the patient who never obtains the drug.

This paper proposes a mandatory finding recorded for every product at every tier rather than a factor competing inside the cascade. The recorded elements are the consequence of nontreatment or interrupted treatment. Documented cost-related nonadherence in the relevant population; prescriber availability in the relevant geography; time-criticality, meaning whether delay itself causes harm; and encounter cost as a fraction of total therapy cost. Two uses follow. The finding becomes the tiebreaker where the cascade is indeterminate, and it is published alongside every Tier 1 and Tier 2 assignment, so that the cost of each retained restriction is on the record rather than assumed to be zero. The microeconomic literature on prescription-to-nonprescription switching supplies the analytic form such a finding should take [25].

3.6. Considerations Excluded by Rule

Six considerations are barred explicitly, because each will otherwise enter indirectly: (1) prescription status as of 1951 or at original approval; (2) effect on prescriber, pharmacy, or manufacturer revenue; (3) effect on insurance coverage or formulary placement, which is a real consequence but one to be addressed in benefit design rather than absorbed into a safety determination; (4) sponsor preference for tier assignment, which reflects channel strategy rather than risk; (5) moral evaluation of the indication; and (6) generalized caution unaccompanied by a named hazard.

3.7. The Continuation Exemption

Applying the cascade to marketed products requires one addition to the unit of classification, which is the patient’s state. A product may satisfy a Tier 1 factor at initiation and none at continuation, and for a substantial share of the pharmacopoeia that is the correct result. The proposed qualifying criteria are documented continuous dispensing at a stable dose for a defined period (minimum or 12 months); no interruption exceeding a lapse threshold; no intervening event, meaning hospitalisation attributable to the drug, a new contraindicating diagnosis, pregnancy, or a dose change within the window; and, where the Tier 1 factor was monitoring dependency, evidence that the monitored parameter remained within range.

The verification problem that dominates the additional-condition pathway does not arise here. The qualifying fact — prior dispensing at a stable dose — is already held in verified form by the entity making the sale. No new infrastructure, no self-report, and no sponsor-operated instrument is required. The continuation exemption is consequently a pharmacy arrangement and cannot be operated from a self-checkout terminal, which distinguishes it from the additional-condition architecture and answers the objection that graduated access is that architecture with a wider footprint.

One exclusion must be written in rather than left to judgement. The exemption must not apply where duration is itself the hazard, for example, long-term PPIs, NSAIDs, bisphosphonates, systemic corticosteroids, and benzodiazepines. For these, uneventful long use strengthens the case for review rather than for release.

3.8. Sequencing the Review

Reviewing roughly 1,900 active pharmaceutical ingredients requires multiple steps. The cohorts are ordered by six factors: (1) population reach weighted by chronicity; (2) nonprescription precedent among peer regulators, indicating that supporting evidence already exists; (3) availability of inexpensive patient self-monitoring; (4) documented cost-related nonadherence; (5) a partial United States switch, indicating that dose and duration boundaries are already established; and (6) the absence of third-party externalities.

3.8.1. Cohort 1

Two entries carry weight disproportionate to their length. The insulin row is the strongest domestic evidence available: recombinant human insulin is dispensed without a prescription behind the pharmacy counter in 49 states, an unlegislated second tier that the FDA confirmed would survive the transition of insulin to biological product regulation [29]. The Government Accountability Office asked whether a behind-the-counter class had produced demonstrable benefits abroad [11,12] and did not examine the domestic instance. The antihypertensive row demonstrates the monitoring criterion discriminating rather than serving as a blanket exclusion: blood pressure is measurable at home for a trivial sum, which is why amlodipine descends and warfarin does not.

The diabetes entry carries a methodological point best met where it arises. Diabetes does not survive the rubric as a class. Human insulin and metformin sit in Cohort 1, oral semaglutide for type 2 diabetes sits in Cohort 2, and SGLT2 inhibitors and the GLP-1 agonists for weight management sit among the delayed classes [30]. That dispersal is the API × route × indication unit doing what Section 3.2.1 predicted, and it is the clearest demonstration that therapeutic class is a convenience of presentation rather than a unit of assignment. Over-the-counter continuous glucose monitoring [31,32] strengthens the case for the products that remain, and its indication limits weaken it for those that do not.

Table 2. Cohort 1: classes for first review. Tier at initiation is stated separately from the continuation assignment.

Class	Molecules reachable on continuation	Anchor	Tier at initiation
Statins and ezetimibe	atorvastatin, rosuvastatin, simvastatin, pravastatin, ezetimibe	UK simvastatin 10 mg as a pharmacy medicine from 2004 [26]; four prior US switch attempts; a technology-assisted self-selection trial reporting high agreement with clinical assessment	2 (risk estimation)
Antihypertensives	amlodipine; lisinopril, losartan, hydrochlorothiazide on continuation	Home blood pressure measurement is cheap, ubiquitous, validated, and available at no cost at many access sites — the rare case in which the monitoring parameter is patient-obtainable	2
Inhaled asthma maintenance	albuterol; budesonide, fluticasone; ICS–LABA combinations	Non-prescription in several peer markets; symptom-based self-monitoring; documented mortality from access interruption	2
Diabetes	human insulin (Novolin/Humulin R, N, 70/30), already dispensed without prescription; metformin; analogue insulins and sulfonylureas on continuation. Excludes SGLT2 inhibitors and GLP-1 receptor agonists, held to Table 4	The United States already operates an unlegislated second tier here: human insulin is sold behind the pharmacy counter without a prescription in every state but Indiana. Over-the-counter continuous glucose monitoring, cleared from 2024, makes glucose the second patient-obtainable monitoring parameter after blood pressure, though those clearances are indicated for people not using insulin	2 (metformin: renal function input)
Acid suppression	omeprazole, esomeprazole, lansoprazole at prescription dose and duration; H2 antagonists	Three PPIs already switched at 15–20 mg for 14 days; the line-drawing work is done	2, with alarm-symptom screen and duration cap
Hormonal contraception	combined oral contraceptives; progestin-only pills	Norgestrel switched in 2023 [27]; the most complete self-screening validation literature of any class	2
Migraine abortives [28]	sumatriptan, rizatriptan, naratriptan	UK pharmacy status from 2006; established diagnosis, self-limited use, pack-size control	2 at initiation, 3 thereafter
Residual antihistamines and intranasal agents	remaining prescription second-generation agents, azelastine	Class largely switched already	3

Table 2. Cohort 1: classes for first review. Tier at initiation is stated separately from the continuation assignment.

Class	Molecules reachable on continuation	Anchor	Tier at initiation
Statins and ezetimibe	atorvastatin, rosuvastatin, simvastatin, pravastatin, ezetimibe	UK simvastatin 10 mg as a pharmacy medicine from 2004 [26]; four prior US switch attempts; a technology-assisted self-selection trial reporting high agreement with clinical assessment	2 (risk estimation)
Antihypertensives	amlodipine; lisinopril, losartan, hydrochlorothiazide on continuation	Home blood pressure measurement is cheap, ubiquitous, validated, and available at no cost at many access sites — the rare case in which the monitoring parameter is patient-obtainable	2
Inhaled asthma maintenance	albuterol; budesonide, fluticasone; ICS–LABA combinations	Non-prescription in several peer markets; symptom-based self-monitoring; documented mortality from access interruption	2
Diabetes	human insulin (Novolin/Humulin R, N, 70/30), already dispensed without prescription; metformin; analogue insulins and sulfonylureas on continuation. Excludes SGLT2 inhibitors and GLP-1 receptor agonists, held to Table 4	The United States already operates an unlegislated second tier here: human insulin is sold behind the pharmacy counter without a prescription in every state but Indiana. Over-the-counter continuous glucose monitoring, cleared from 2024, makes glucose the second patient-obtainable monitoring parameter after blood pressure, though those clearances are indicated for people not using insulin	2 (metformin: renal function input)
Acid suppression	omeprazole, esomeprazole, lansoprazole at prescription dose and duration; H2 antagonists	Three PPIs already switched at 15–20 mg for 14 days; the line-drawing work is done	2, with alarm-symptom screen and duration cap
Hormonal contraception	combined oral contraceptives; progestin-only pills	Norgestrel switched in 2023 [27]; the most complete self-screening validation literature of any class	2
Migraine abortives [28]	sumatriptan, rizatriptan, naratriptan	UK pharmacy status from 2006; established diagnosis, self-limited use, pack-size control	2 at initiation, 3 thereafter
Residual antihistamines and intranasal agents	remaining prescription second-generation agents, azelastine	Class largely switched already	3

3.8.2. Cohort 2

The alpha-blocker row illustrates a distinction the framework must make explicit and no existing system does. The United Kingdom model permits a pharmacist to initiate therapy and requires a physician to authorize its continuation [37]. The continuation exemption proposed here permits continuation without a physician after an initiation that required one. Both can be correct, because they answer different hazards. In benign prostatic hyperplasia, the risk is a malignancy that declares itself over time, so the referral belongs downstream of first supply. In hypothyroidism, the risk sits at titration, so the professional encounter belongs upstream and the stable-dose patient needs nothing further. Every Tier 2 assignment must therefore specify whether its condition points forward or backward in the course of therapy.

Table 3. Cohort 2: classes where the case is strong but a required condition remains to be specified.

Class	Molecules reachable on continuation	Anchor	Tier at initiation
Thyroid replacement	levothyroxine	The single best test of the continuation exemption: tens of millions at a stable dose, an annual encounter that transfers no information, and a monitoring parameter the patient cannot self-obtain	1 (2 on continuation, with periodic thyrotropin)
Oral GLP-1 receptor agonists, type 2 diabetes only	oral semaglutide at a stable maintenance dose	Escalation runs against gastrointestinal tolerability rather than a laboratory value, so the titration factor does not fire; hypoglycaemia risk is low as monotherapy and the thyroid C-cell contraindication is a family-history screen	2 on continuation (1 at initiation; the weight-management indication is excluded to Table 4)
PDE5 inhibitors	sildenafil, tadalafil	UK pharmacy status from 2018 [33]; nitrate co-use is the single hazard and is questionnaire-detectable	2
Alpha-blockers for benign prostatic hyperplasia	tamsulosin, alfuzosin	UK reclassification announced in 2009 and marketed from 2010; a symptoms-check questionnaire was a condition of the licence and supply is capped before physician review	2, with mandatory referral trigger
Topical dermatology	adapalene, tretinoin, mid-potency corticosteroids, azelaic acid	Adapalene 0.1% already switched in the United States in 2016 [34]; negligible systemic absorption; the hazard is masked infection and skin atrophy rather than systemic toxicity	3 (adapalene), 2 (tretinoin, mid-potency steroids)
SSRIs and SNRIs, continuation only	sertraline, escitalopram, fluoxetine, venlafaxine	The access failure is abrupt discontinuation, a harm the present arrangement causes rather than prevents; initiation carries diagnosis and the under-25 suicidality signal	1 (2 on continuation; exclude under 25)
Urate-lowering therapy	allopurinol	Stable-dose continuation is uneventful; initiation carries hypersensitivity risk and renal dose adjustment	1 (3 on continuation)
Overactive bladder	oxybutynin transdermal, mirabegron	Transdermal oxybutynin already switched in the United States in 2013 [35]; the hazard is anticholinergic burden and masked urinary tract pathology	2, with duration cap
Smoking cessation	varenicline, bupropion SR	Nicotine replacement is fully switched; the neuropsychiatric boxed warning on varenicline was removed following a randomized safety trial, retiring the stated basis for its prescription status [36]	2

Table 3. Cohort 2: classes where the case is strong but a required condition remains to be specified.

Class	Molecules reachable on continuation	Anchor	Tier at initiation
Thyroid replacement	levothyroxine	The single best test of the continuation exemption: tens of millions at a stable dose, an annual encounter that transfers no information, and a monitoring parameter the patient cannot self-obtain	1 (2 on continuation, with periodic thyrotropin)
Oral GLP-1 receptor agonists, type 2 diabetes only	oral semaglutide at a stable maintenance dose	Escalation runs against gastrointestinal tolerability rather than a laboratory value, so the titration factor does not fire; hypoglycaemia risk is low as monotherapy and the thyroid C-cell contraindication is a family-history screen	2 on continuation (1 at initiation; the weight-management indication is excluded to Table 4)
PDE5 inhibitors	sildenafil, tadalafil	UK pharmacy status from 2018 [33]; nitrate co-use is the single hazard and is questionnaire-detectable	2
Alpha-blockers for benign prostatic hyperplasia	tamsulosin, alfuzosin	UK reclassification announced in 2009 and marketed from 2010; a symptoms-check questionnaire was a condition of the licence and supply is capped before physician review	2, with mandatory referral trigger
Topical dermatology	adapalene, tretinoin, mid-potency corticosteroids, azelaic acid	Adapalene 0.1% already switched in the United States in 2016 [34]; negligible systemic absorption; the hazard is masked infection and skin atrophy rather than systemic toxicity	3 (adapalene), 2 (tretinoin, mid-potency steroids)
SSRIs and SNRIs, continuation only	sertraline, escitalopram, fluoxetine, venlafaxine	The access failure is abrupt discontinuation, a harm the present arrangement causes rather than prevents; initiation carries diagnosis and the under-25 suicidality signal	1 (2 on continuation; exclude under 25)
Urate-lowering therapy	allopurinol	Stable-dose continuation is uneventful; initiation carries hypersensitivity risk and renal dose adjustment	1 (3 on continuation)
Overactive bladder	oxybutynin transdermal, mirabegron	Transdermal oxybutynin already switched in the United States in 2013 [35]; the hazard is anticholinergic burden and masked urinary tract pathology	2, with duration cap
Smoking cessation	varenicline, bupropion SR	Nicotine replacement is fully switched; the neuropsychiatric boxed warning on varenicline was removed following a randomized safety trial, retiring the stated basis for its prescription status [36]	2

3.8.3. Delayed Classes

Delay is not a single judgement, and the distinction matters more here than anywhere else. Some classes are delayed because they satisfy Tier 1 factors on the merits and would do so under any defensible rubric. Others are delayed because a necessary instrument has not been designed. Conflating the two concedes ground the case for graduated access has already taken, so Table 4 states what would have to change.

Table 4. Delayed classes, with the basis for delay distinguished from the condition that would reopen the assignment.

Class	Representative agents	Basis for delay	What would reopen the case
Vitamin K antagonists	warfarin	Tier 1 on the merits: narrow margin, bleeding as the failure mode, harm not rescuable at the point of sale, and titration against a laboratory value	Validated patient self-testing and self-management, already accepted in several European systems, would convert the monitoring factor into a patient-obtainable measurement
Direct oral anticoagulants	apixaban, rivaroxaban, edoxaban, dabigatran	Tier 1 at initiation on different grounds from warfarin: diagnostic dependency is absolute, dose selection turns on renal function, age, and weight, and bleeding is reversible only in hospital. Routine coagulation monitoring is not required, so the factor that places warfarin here does not apply	A patient-obtainable measure of renal function. The continuation case is the strongest in this table: short half-lives mean an interrupted supply loses protection within days, and the consequence is stroke. Consumer electrocardiography detects atrial fibrillation but does not supply the decision to anticoagulate
Antiarrhythmics	amiodarone, flecainide, sotalol	Tier 1 on the merits: proarrhythmia, and for amiodarone cumulative thyroid, pulmonary, hepatic, and ocular toxicity requiring serial surveillance	Nothing foreseeable. Amiodarone additionally triggers the duration-as-hazard exclusion from any continuation exemption
Oncology agents	cytotoxics, targeted agents, hormonal agents in active treatment	Tier 1 on the merits: diagnostic dependency is absolute, and administration and monitoring requirements compound it	Nothing within this framework. Failures here, such as fluoropyrimidine toxicity in undetected enzyme deficiency, are failures of the gatekeeper rather than arguments for removing the gate
Immunosuppressants and biologics	tacrolimus, methotrexate, TNF inhibitors	Tier 1 on the merits: trough concentration monitoring, infection surveillance, and administration requirements	Partial. Stable-dose self-injected biologics on continuation already operate as home administration and are the nearest candidate for a narrow exemption
Antipsychotics at initiation	aripiprazole, risperidone, olanzapine, quetiapine	Tier 1 at initiation: diagnosis, plus metabolic and movement-disorder monitoring	A continuation case exists and resembles the SSRI row in Table 3 more than it resembles oncology
SGLT2 inhibitors	empagliflozin, dapagliflozin, canagliflozin	Euglycaemic ketoacidosis presents at normal or near-normal glucose, defeating the self-monitored parameter that qualifies the rest of the diabetes class; estimated glomerular filtration rate is required and is not patient-obtainable; sick-day withholding is a conditional action that labelling conveys poorly	A validated home ketone measurement paired with a sick-day protocol whose comprehension has been tested. Blood beta-hydroxybutyrate meters already exist; the protocol has not been validated
Oral GLP-1 receptor agonists, weight management	oral semaglutide, orforglipron	Evidentiary immaturity. Orforglipron was approved in April 2026, fifty days after filing under a priority voucher pilot [38]; oral semaglutide received its weight-loss indication in December 2025. No validated point-of-sale instrument screens for disordered eating	A stabilized indication set with post-market exposure at scale, and a validated eating-disorder screen. Manufacturer-direct distribution is already operating for this class, so sponsor support for a lower tier reflects channel strategy, which Section 3.6 excludes
Systemic antimicrobials	most oral antibiotic classes	Not a safety deferral. The externality criterion is satisfied and no stewardship instrument yet exists to accompany a lower tier	A designed stewardship mechanism: quantity limits, mandatory pharmacist counselling, and cross-supply tracking. The class is delayed for want of that design, not because patients cannot be trusted with it
Anticonvulsants	levetiracetam, lamotrigine, valproate, carbamazepine	The genuine edge. Narrow therapeutic index, interaction burden, and valproate teratogenicity argue Tier 1; seizure and status epilepticus on interruption argue that access failure is the more probable harm	Resolution turns on how heavily the access-cost criterion is weighted against the conventional factors. This is the framework’s test case rather than a settled exclusion
Controlled substances	opioids, benzodiazepines, stimulants	Outside the framework. Scheduling under the Controlled Substances Act is a separate statutory regime	Not a deferral on the merits. Revision of the prescription requirement need not disturb the schedules, though benzodiazepines separately trigger the duration-as-hazard exclusion

Table 4. Delayed classes, with the basis for delay distinguished from the condition that would reopen the assignment.

Class	Representative agents	Basis for delay	What would reopen the case
Vitamin K antagonists	warfarin	Tier 1 on the merits: narrow margin, bleeding as the failure mode, harm not rescuable at the point of sale, and titration against a laboratory value	Validated patient self-testing and self-management, already accepted in several European systems, would convert the monitoring factor into a patient-obtainable measurement
Direct oral anticoagulants	apixaban, rivaroxaban, edoxaban, dabigatran	Tier 1 at initiation on different grounds from warfarin: diagnostic dependency is absolute, dose selection turns on renal function, age, and weight, and bleeding is reversible only in hospital. Routine coagulation monitoring is not required, so the factor that places warfarin here does not apply	A patient-obtainable measure of renal function. The continuation case is the strongest in this table: short half-lives mean an interrupted supply loses protection within days, and the consequence is stroke. Consumer electrocardiography detects atrial fibrillation but does not supply the decision to anticoagulate
Antiarrhythmics	amiodarone, flecainide, sotalol	Tier 1 on the merits: proarrhythmia, and for amiodarone cumulative thyroid, pulmonary, hepatic, and ocular toxicity requiring serial surveillance	Nothing foreseeable. Amiodarone additionally triggers the duration-as-hazard exclusion from any continuation exemption
Oncology agents	cytotoxics, targeted agents, hormonal agents in active treatment	Tier 1 on the merits: diagnostic dependency is absolute, and administration and monitoring requirements compound it	Nothing within this framework. Failures here, such as fluoropyrimidine toxicity in undetected enzyme deficiency, are failures of the gatekeeper rather than arguments for removing the gate
Immunosuppressants and biologics	tacrolimus, methotrexate, TNF inhibitors	Tier 1 on the merits: trough concentration monitoring, infection surveillance, and administration requirements	Partial. Stable-dose self-injected biologics on continuation already operate as home administration and are the nearest candidate for a narrow exemption
Antipsychotics at initiation	aripiprazole, risperidone, olanzapine, quetiapine	Tier 1 at initiation: diagnosis, plus metabolic and movement-disorder monitoring	A continuation case exists and resembles the SSRI row in Table 3 more than it resembles oncology
SGLT2 inhibitors	empagliflozin, dapagliflozin, canagliflozin	Euglycaemic ketoacidosis presents at normal or near-normal glucose, defeating the self-monitored parameter that qualifies the rest of the diabetes class; estimated glomerular filtration rate is required and is not patient-obtainable; sick-day withholding is a conditional action that labelling conveys poorly	A validated home ketone measurement paired with a sick-day protocol whose comprehension has been tested. Blood beta-hydroxybutyrate meters already exist; the protocol has not been validated
Oral GLP-1 receptor agonists, weight management	oral semaglutide, orforglipron	Evidentiary immaturity. Orforglipron was approved in April 2026, fifty days after filing under a priority voucher pilot [38]; oral semaglutide received its weight-loss indication in December 2025. No validated point-of-sale instrument screens for disordered eating	A stabilized indication set with post-market exposure at scale, and a validated eating-disorder screen. Manufacturer-direct distribution is already operating for this class, so sponsor support for a lower tier reflects channel strategy, which Section 3.6 excludes
Systemic antimicrobials	most oral antibiotic classes	Not a safety deferral. The externality criterion is satisfied and no stewardship instrument yet exists to accompany a lower tier	A designed stewardship mechanism: quantity limits, mandatory pharmacist counselling, and cross-supply tracking. The class is delayed for want of that design, not because patients cannot be trusted with it
Anticonvulsants	levetiracetam, lamotrigine, valproate, carbamazepine	The genuine edge. Narrow therapeutic index, interaction burden, and valproate teratogenicity argue Tier 1; seizure and status epilepticus on interruption argue that access failure is the more probable harm	Resolution turns on how heavily the access-cost criterion is weighted against the conventional factors. This is the framework’s test case rather than a settled exclusion
Controlled substances	opioids, benzodiazepines, stimulants	Outside the framework. Scheduling under the Controlled Substances Act is a separate statutory regime	Not a deferral on the merits. Revision of the prescription requirement need not disturb the schedules, though benzodiazepines separately trigger the duration-as-hazard exclusion

Three rows carry argumentative weight. Systemic antimicrobials are the only entry delayed for a reason external to the patient, and the reason should be stated in those terms whenever the class is raised. The collective-action problem is real, the present gatekeeper has demonstrably not solved it [39], and a lower tier awaits a stewardship design rather than a finding that the drugs are unsuited to the people taking them. The SGLT2 row shows a criterion discriminating within a therapeutic class on a precise pharmacological ground. Euglycemic ketoacidosis defeats the self-monitored parameter that qualifies the rest of the class [40]. Anticonvulsants are the case on which the framework should be judged, being the clearest instance in which the access-cost criterion and the conventional factors point in opposite directions.

3.9. Procedure

Four procedural provisions complete the instrument. A fast track admits any product available without prescription in two or more peer regulators for five or more years, with a retrievable pharmacovigilance record, at the corresponding tier by presumption, rebuttable by affirmative showing; this generates as a byproduct the comparative evidence the GAO found absent [11,12]. A default on inaction moves any cohort the review body has not reached within its statutory window down one tier, without which an under-resourced body never reaches the contested cases and the status quo prevails by attrition. Reclassification in either direction must be governed by the same expressly articulated evidentiary standard to prevent a systematic bias toward progressively more restrictive classifications. And the review body requires structurally guaranteed patient membership. The Canadian committee is drawn from pharmacotherapy, drug utilization, toxicology, pharmacy practice, academic research, industry, and regulatory affairs [19], a defensible expert base with no patient seat.

4. Discussion

This exercise yields three findings that the criteria alone do not predict.

Therapeutic class is not a unit of assignment. The dispersal of diabetes across all three cohorts, and the division of anticoagulants on grounds that a single class label conceals, are the clearest results the application produces. Any graduated-access proposal that assigns by class will produce systematically wrong answers in both directions, restricting oral agents because injectable ones are restricted and releasing agents whose hazard is invisible to the monitoring the class ordinarily supports.

The monitoring criterion is technology-dependent and therefore dated. Blood pressure became patient-obtainable decades ago while glucose became patient-obtainable at scale in 2024 [31]. Renal function, thyrotropin, and international normalized ratio have not. A criterion whose application turns on the current state of consumer diagnostics cannot be applied once and left, which is the strongest argument for the periodic review the proposal specifies and against a one-time reclassification statute.

The access-cost finding changes outcomes, not merely rhetoric. Anticonvulsants and direct oral anticoagulants are both delayed on conventional factors, and in both the harm of interruption is more probable than the harm of supply. Recording that finding does not by itself move either class, but it makes the trade-off visible and assignable, which is what the present arrangement never does [3,4].

The framework is exposed to the standard objection that a body constituted to review classifications will be captured by those it reviews [22,41]. Carpenter’s account of regulatory reputation [42] suggests the objection is a complication rather than a refutation. An agency invested in its standing with multiple constituencies is not straightforwardly the creature of one. The provisions most likely to be resisted — the default on inaction, patient membership, and the published access-cost finding — are precisely those designed against capture, and their fate in any legislative vehicle will be informative about the analysis itself.

The relationship to the additional-condition pathway deserves statement. That rule creates a conditioned nonprescription category operated by sponsors through proprietary digital instruments [13], and the generic industry has argued on the record that the pathway may be unusable by the firms that supply price competition [43]. The architecture proposed here differs in the location of the condition. A continuation exemption verified by the dispensing record is operated by a pharmacy against data it already holds, generates no new identified clinical dataset outside health-privacy law, and requires no sponsor-designed instrument. Where the two overlap, they should be reconciled in favour of the arrangement that creates the smaller governance gap.

4.1. Limitations

This analysis is prospective and document-based. No tier assignment proposed here has been adjudicated, and the cohort tables are arguments about how criteria would apply rather than reports of how they did. The evidentiary base for the access-cost finding is thin outside a handful of classes. Documented cost-related nonadherence is concentrated in insulin, cardiovascular, and respiratory therapy [4], and does not exist longitudinally. The framework will be making that finding on inference for most assignments. The immaturity threshold is arbitrary until a number is chosen, and the number determines how much of the pharmacopoeia is reachable in its first decade. Several Tier 1 factors are stated as binary predicates although the underlying properties are continuous, so a published interpretive guidance with worked examples is required rather than a claim that the factors are self-applying. Finally, this paper does not evaluate whether graduated access would improve population health. That question requires outcome data that do not yet exist. What it offers is the instrument by which such a system could be populated, so that the empirical question becomes answerable.

5. Conclusions

The American prescription category was populated by default rather than by adjudication, and the proposals to replace it have never specified how the resulting pharmacopoeia would be reviewed. The instrument set out here is a cascade adapted from four peer regulators, a burden of proof that resolves downward on silence, a mandatory finding of the harm of withholding, an exclusion list of considerations that may not count, and a continuation exemption verified by data the pharmacy already holds and is intended to make that review tractable.

Its application produces a result worth more than any individual assignment. Therapeutic class does not survive contact with the criteria. Diabetes disperses, anticoagulants divide, and the unit of assignment turns out to be the API, the route, the strength, the indication, and the patient’s position within their own therapy. A graduated architecture built on class labels would reproduce in a new form the categorical error it exists to correct. Seventy-five years is long enough for a categorical rule to demonstrate the benefit on which it formally rests [1,44]. What the alternative has lacked is not argument but machinery. This is a first draft of the machinery, offered for criticism.