Draft 2 archive: This page preserves the public review version and its earlier framing. The revised paper is published as Approval Is Not Exposure: The COVID-19 Category Error.

Approval Is Not Exposure

The COVID-19 Category Error and the Point at Which Informed Consent Failed

Sid J.A. Hubbard

Draft 2 — July 2026

Copyright (c) 2026 Sid J.A. Hubbard. All rights reserved.

Abstract

Regulatory approval and accumulated human exposure answer different questions. Approval records an agency judgment about a defined product, indication, population, regimen, and evidence package. Exposure records opportunities to observe what happens when a substance is used in human beings across time and variation. During COVID-19, those categories were repeatedly collapsed. Established medicines entered the emergency without a COVID-specific indication because the disease was new, while novel products entered with sponsors capable of producing indication-specific evidence and obtaining authorization. The resulting regulatory statuses were then treated as a comparative biological ranking.

Anthony Fauci's March 2020 diary captures the distinction before it disappeared from public decision-making. Fauci contrasted the “tons of experience” with hydroxychloroquine against his greater concern about the potential toxicities of investigational remdesivir. The following day he recorded that he had convinced officials not to make the drugs broadly available through compassionate use. The episode does not establish hydroxychloroquine's efficacy. It establishes that accumulated exposure was recognized as safety evidence while access to a familiar candidate was made subordinate to the preferred evidence-production process.

The same category error later operated in reverse. Authorization and approval of novel mRNA vaccines were often treated as though they completed the safety comparison for every recipient, despite necessarily limited duration of exposure and unresolved population-specific risks. Informed consent failed at the precise point where a regulator's population-level judgment was allowed to substitute for a person's comparison of indication-specific evidence, known toxicity, accumulated exposure, residual uncertainty, and available alternatives.

1. Two Kinds of Evidence

FDA approval means that a defined product satisfied a regulatory standard for a defined use. It does not mean that the product has no hazards. The FDA describes approval as a determination that benefits outweigh potential risks for the intended use.¹

Accumulated exposure is different. It records how many people have used a substance, in what formulations and doses, for how long, and with how much opportunity for common, rare, delayed, or population-specific effects to appear.

The two forms of evidence overlap but are not interchangeable:

approval
    product- and indication-specific evidence
    manufacturing and labeling review
    a regulatory benefit-risk judgment

accumulated exposure
    observed use across people and time
    characterized toxicities under known conditions
    opportunity for unexpected effects to emerge

A familiar drug used for a new disease may have extensive exposure evidence and weak indication-specific evidence. A novel product may have strong indication-specific trial evidence and little accumulated exposure. Neither fact erases the other.

When SARS-CoV-2 appeared, no established medicine could already be approved for COVID-19. Nonapproval was the necessary regulatory starting condition of every repurposed drug, not the result of a completed comparative safety evaluation. It did not reveal whether a candidate had failed, had not yet been studied, or lacked a sponsor positioned to finance a new indication.

Treating “not approved for COVID-19” as synonymous with “more dangerous” or “medically illegitimate” converted the novelty of the disease into a false biological verdict.

2. The Distinction Was Visible Inside the Response

On March 18, 2020, the White House was considering broader access to hydroxychloroquine and remdesivir. Fauci objected to treating the candidates as though they presented the same uncertainty.

In a follow-up email reproduced with his diary, he described hydroxychloroquine as an approved drug with adverse effects but “tons of experience with it.” Remdesivir was unapproved and investigational. Fauci wrote that he had “more concern about the potential toxicities” of remdesivir and noted its lack of clear benefit in the Ebola experience. He proposed leading with hydroxychloroquine in a flexible distribution protocol while accelerating the remdesivir trial.²

That comparison was technically sound. It did not claim that hydroxychloroquine treated COVID-19. It kept efficacy evidence separate from exposure evidence.

The next day's diary entry records a consequential change. Fauci wrote that the administration had intended to make the drugs broadly available through compassionate use: “I convinced them not to.” On March 22, he called an online system intended to track the medicines and make them available through physicians a “Bad idea,” acceptable only if it did not interfere with randomized trials.³

Randomized trials were necessary, and later evidence could properly reject a candidate for a particular regimen or population. But preserving trial quality and preserving monitored clinical access were not mutually exclusive objectives. The diary shows that the distinction between the candidates was known, yet the response did not build an access-and-learning system around that distinction. It narrowed access in order to protect one method of learning.

The validity of later findings does not change the category error. Later evidence may show that a drug does not benefit a defined population or that its risks outweigh its benefits. It cannot retroactively turn the drug's initial lack of a COVID-specific indication into evidence that its historical human exposure did not exist.

3. The Exact Mistake

Four questions were compressed into a single regulatory label:

1. Does the intervention work for this disease, regimen, and population?
2. What toxicities are already known?
3. How much human exposure supports that knowledge?
4. What remains unknown because the intervention, formulation, route, or use is new?

Approval answers none of these questions by itself. It reports the agency's judgment after evaluating evidence relevant to all four. That judgment remains bounded by the evidence available at the time.

The mistake can be stated precisely:

Valid:
FDA approved product P for use I after finding its demonstrated benefits
outweighed its known and potential risks for the specified population.

Invalid:
FDA approval proves P is safer for person X than every unapproved option,
resolves risks that limited exposure could not reveal, and supplies X's
informed consent.

Hydroxychloroquine and ivermectin had long histories of human use before COVID-19. Those histories did not prove efficacy against SARS-CoV-2, did not validate every proposed dose or combination, and did not make the drugs interchangeable with vaccination. They nevertheless contained real safety information.

The mRNA vaccines entered with COVID-specific randomized evidence and manufacturing controls, but without population-scale or long-duration human exposure. Approval could evaluate the evidence that existed. It could not create observations that required more people or more time.

The regulatory comparison was therefore asymmetric:

established candidate
    known human exposure
    known toxicities under established uses
    uncertain benefit for the new indication
    weak commercial incentive for a new indication

novel proprietary candidate
    indication-specific trials
    organized sponsor and regulatory submission
    limited exposure at authorization
    unresolved rare and delayed risks

Calling the second candidate “approved” and the first “unapproved” accurately described regulatory status. Using those words as a complete comparative safety judgment did not.

4. Where Informed Consent Was Mistaken for Granted

Informed consent is not a property of a product. It is a process between a person and a clinician. It requires an accurate account of the intervention's purpose, expected benefits, burdens, risks, alternatives, and the option of forgoing treatment.⁴

The emergency-use statute made the unresolved nature of the decision explicit. Recipients of an EUA product were to be informed of known and potential benefits and risks, the extent to which they were unknown, the option to accept or refuse, the consequences of refusal, and available alternatives.⁵

The exact failure occurred when authorization or approval was treated as if it had already performed that conversation:

regulatory judgment about a population
                 became
presumed consent by an individual

A truthful comparison would have disclosed both forms of uncertainty:

This novel intervention has indication-specific trial evidence, but limited accumulated human exposure. Rare, delayed, and population-specific effects may become visible only after widespread use.

This established medicine has extensive prior exposure and characterized toxicities under its ordinary uses. Its efficacy, dose, timing, and benefit-risk relationship for COVID-19 remain uncertain.

Neither statement dictates a choice. Together they identify the choice.

Instead, approval was commonly allowed to imply that the novel product's safety uncertainty had been completed, while nonapproval implied that the familiar medicine fell outside responsible consideration. Requirements conditioning employment, education, military service, or participation in ordinary institutions then made regulatory status do still more work. The formal ability to refuse an injection is not equivalent to an unburdened choice when refusal carries a serious institutional penalty.

This is where informed consent was mistaken for granted: not when FDA reviewed an application, but when institutions treated that review as a substitute for disclosing comparative uncertainty and obtaining a voluntary, person-specific decision.

5. What Continued Exposure Revealed

The intended mechanism of the mRNA vaccines was to deliver nucleoside-modified messenger RNA in lipid nanoparticles so that human cells would produce a prefusion-stabilized SARS-CoV-2 spike antigen and provoke an immune response.⁶ The induced antigen was part of the intended biological mechanism, not an accidental contaminant.

Experimental studies of SARS-CoV-2 spike and its S1 subunit reported biological effects relevant to cardiovascular and neurological injury, including endothelial or pericyte dysfunction, passage across the blood-brain barrier in mice, glial activation, and neuronal injury in animal models.⁷⁸ These studies established reasons for investigation. They did not, by themselves, establish that vaccine-produced antigen at vaccine exposure levels caused generalized cardiac or neurological toxicity in humans. Conflating those propositions would repeat the same kind of category error.

Human surveillance did establish an increased risk of myocarditis and pericarditis after mRNA vaccination, particularly in adolescent and young adult males.⁹ The Pfizer approval announcement in August 2021 acknowledged that risk, noted that long-term outcome information was unavailable, and required postmarketing studies.¹⁰ In 2025, FDA required updated warnings with newer incidence estimates and cardiac MRI follow-up. FDA reported that persistent MRI findings indicating myocardial injury were common at a median follow-up of about five months among the studied patients, while their clinical and prognostic significance remained unknown.¹¹

This does not establish that authorization or approval was improper, nor does it resolve the vaccines' net benefits. It establishes the narrower and more important point: approval did not complete the safety knowledge. Exposure continued to produce material information after the regulatory decision.

The newly observed risk also varied by age and sex. That variation demonstrates why a population-level authorization cannot supply an individual's consent. A benefit-risk judgment may be favorable across a broad population while remaining materially different for particular people within it.

6. Keep the Search Space Open

A novel emergency should expand the number of plausible interventions entering disciplined comparison. Access need not be confused with endorsement, and investigation need not be confused with approval.

Established medicines proposed for new uses should enter monitored protocols that record formulation, dose, timing, interactions, outcomes, and adverse events. Publicly financed adaptive trials should test inexpensive candidates when diffuse ownership leaves no sponsor able to recover the cost of a new indication. Weak candidates should be rejected quickly because evidence rejects them, not because their regulatory category prevents evidence from being produced.

Every candidate should be represented on separate axes:

indication-specific efficacy
known toxicity
accumulated human exposure
residual novelty
quality of available evidence
reason the evidence package exists or is absent

No single axis should impersonate the others.

Conclusion

The COVID-19 category error was not simply that one intervention was chosen over another. It was that regulatory status was allowed to replace a multidimensional comparison.

Fauci's diary shows that the missing dimension was visible in March 2020. He distinguished the accumulated experience with hydroxychloroquine from the larger toxicological uncertainty surrounding investigational remdesivir. Access was nevertheless narrowed to protect the preferred trial pathway. Later, the direction of the error reversed: authorization and approval of a novel platform were treated as sufficient answers to safety questions that only accumulated exposure could answer.

Approval is evidence of regulatory review for a specified use. Exposure is evidence produced by human bodies over time. Neither proves efficacy, neither eliminates toxicity, and neither can stand in for the other.

Most importantly, neither can consent on behalf of a person.

References