Unqualified: How the mRNA Vaccines Failed the EUA Standard

Corrected Efficacy Curves, Suppressed Risks, and Policy Recommendations. IPAK-EDU White Paper 2025-ENDEUA

Executive Summary

In December 2020, the U.S. Food and Drug Administration (FDA) granted Emergency Use Authorizations (EUAs) for the first SARS-CoV-2 mRNA vaccines—Pfizer-BioNTech and Moderna—based on interim data from large-scale Phase III trials. The legal standard under §564 of the Federal Food, Drug, and Cosmetic Act requires that a product may be authorized for emergency use only if it is “reasonable to believe” that the product “may be effective” against a life-threatening condition, that the known and potential benefits outweigh the known and potential risks, and that no adequate alternatives are available. This report demonstrates, through a rigorous retrospective analysis, that these criteria were never met.

Key Findings

1. Overstated Efficacy from the Outset

   The widely publicized efficacy claims—95% for Pfizer and 94.1% for Moderna—were derived from case-counting methods that excluded entire categories of trial participants. These included individuals who developed symptoms shortly after vaccination but prior to a defined post-vaccination window (7 days for Pfizer, 14 days for Moderna) and symptomatic cases that were not PCR-confirmed. When these participants are included—following standard first-in-human trial principles—efficacy drops substantially. Lyons-Weiler’s dropout-inclusive recalculation yields a corrected VE of approximately 75%. Doshi’s analysis suggests it may have fallen as low as 19% under full reintegration of excluded cases.

2. Early Risk Period Excluded from Analysis

   The period from first dose to the start of the official efficacy window—when the immune system is unprimed and biologically vulnerable—was excluded from primary analysis. This design decision effectively concealed potential early harms, including those arising from mechanisms such as antibody-dependent enhancement (ADE) or immune suppression, which are most likely to manifest during immune immaturity.

3. Rapid and Predictable Waning of Protection

   Real-world data from 2021 and 2022 confirm that VE against infection declines rapidly. Israeli and UK datasets show efficacy against symptomatic infection falling below 40% within five months and below 20% for Omicron, even after boosters. Protection against severe disease persisted longer but also declined significantly, especially in older or immunocompromised populations. This collapse was anticipated by modeling using a 7.5% monthly exponential decay rate.

4. Dropout-Inclusive Models Fail to Reach EUA Thresholds

   When efficacy is recalculated from a corrected base of 75%, and further adjusted for waning, variant escape, population risk factors, and untested mechanistic hazards (e.g., ADE, molecular mimicry), the resulting population-level VE never exceeds 50%. These findings are not speculative. They are consistent with observed empirical VE trajectories for Pfizer and Moderna, which, once similarly adjusted, fall below EUA-relevant efficacy thresholds within 2–4 months of full vaccination.

5. Suppression of Risk and Immunopathology Evidence

   No ferret challenge studies were conducted to rule out ADE. No epitope exclusion or autoantibody screening was required prior to rollout. Peer-reviewed bioinformatics work published before vaccine deployment warned of widespread homology between SARS-CoV-2 proteins and human tissues. These findings were confirmed by later clinical studies showing cross-reactivity and the presence of autoantibodies in vaccine recipients. Yet these signals were not considered in the EUA process. Passive surveillance systems later revealed increases in myocarditis, thrombotic events, neurological symptoms, and autoimmune conditions—often in age groups for which the baseline risk of COVID-19 was low.

6. Structural Bias and Misclassification in Observational Data

   Post-market surveillance systems, including the UK Office for National Statistics, misclassified vaccinated individuals as “unvaccinated” if infection occurred before the post-vaccination case-counting threshold. Fenton and Neil identified this as "case counting window bias"—a statistical artifact that artificially inflated VE estimates. These findings corroborate the structural distortions originally revealed in the Phase III trial design.

7. Legal and Regulatory Breakdown

   The statutory threshold that a product “may be effective” was never fulfilled under a corrected efficacy assessment. Moreover, the failure to reevaluate EUA eligibility as VE collapsed in real-world conditions constitutes a breakdown in regulatory duty. The failure to revoke or re-review the EUA as Omicron rendered the original formulation increasingly ineffective reflects a lack of accountability and scientific integrity.

Conclusion

The evidence presented in this report shows that the SARS-CoV-2 mRNA vaccines did not meet the minimum statutory requirements for Emergency Use Authorization when evaluated through corrected dropout-inclusive models, adjusted real-world decay trajectories, and documented immunopathological risks. The EUA process failed—not because of a single miscalculation, but because of a cascade of exclusions, omissions, and optimistic assumptions that compounded into systemic error.

The purpose of this white paper is not simply to audit a decision. It is to confront the institutional vulnerabilities that allowed the decision to persist uncorrected despite mounting evidence of failure. What follows is a set of urgent and concrete reforms designed to restore integrity to the emergency use framework, safeguard public trust, and prevent this class of error from recurring.

Introduction

The issuance of Emergency Use Authorizations (EUAs) for the SARS-CoV-2 mRNA vaccines in December 2020 marked one of the most consequential decisions in the history of U.S. public health regulation. Under §564 of the Federal Food, Drug, and Cosmetic Act (FD&C Act), an EUA may be granted to a product that has not yet met the criteria for full approval, provided five statutory conditions are met. Chief among these are: 1) that it is “reasonable to believe” the product “may be effective” in preventing or treating a serious or life-threatening disease; and 2) that the known and potential benefits of the product outweigh the known and potential risks, considering the public health emergency context.

At face value, the Pfizer-BioNTech and Moderna Phase III trial data appeared to meet these criteria, reporting 94.1–95% efficacy against symptomatic COVID-19 and no major safety signals at the time of EUA issuance. The U.S. Food and Drug Administration (FDA), operating under extraordinary pressure in late 2020, relied on two-month median follow-up data and designated this level of interim evidence as sufficient. However, a growing body of independent reanalysis and forensic examination reveals that both the efficacy and safety profiles of the mRNA vaccines were likely misrepresented—whether by omission, by methodological sleight, or by structural bias in trial design and reporting.

This white paper conducts a post-hoc forensic evaluation of the mRNA vaccine EUA authorizations in light of:

• Revised estimates of vaccine efficacy when previously excluded cases are reinstated (Lyons-Weiler);

• Clear evidence of rapid waning efficacy within months;

• Evidence of widespread immunopathology, including molecular mimicry and pathogenic priming (Lyons-Weiler, Vojdani, Kanduc et al.);

• Statutory inconsistencies in the EUA issuance process;

• The cultural and structural silencing of scientific dissent that prevented timely course correction.

It is now possible to reasonably question whether the key conditions required for EUA were ever fully satisfied.The risk/benefit analysis, when accurately conducted, tips unfavorably. The legal requirement of “no adequate alternative” is cast into doubt by the suppression of early treatment protocols and known therapeutics.

The assumption that the mRNA vaccines would provide meaningful and sustained efficacy appears unsupported in light of Omicron-era real-world data.

The goal of this paper is not merely retrospective. The regulatory and epistemic failures surrounding the EUA of SARS-CoV-2 mRNA vaccines demand structural reform—both to restore public trust in the scientific process and to protect against similar lapses in future emergency authorizations.

Claimed vs. Actual Efficacy

The Emergency Use Authorizations (EUAs) for the Pfizer-BioNTech and Moderna mRNA vaccines were issued in December 2020 based on interim analyses from their respective Phase III trials. These analyses reported vaccine efficacy (VE) of 95% for Pfizer and 94.1% for Moderna against symptomatic, PCR-confirmed COVID-19. These striking estimates were widely disseminated in press releases, news media, and regulatory briefings, becoming the foundation of global policy decisions. However, closer examination of how those efficacy numbers were derived—and what was deliberately excluded—reveals a deeply flawed picture. When viewed with scientific rigor and full transparency, it becomes clear that the claims of near-perfect protection were built on exclusionary criteria, case selection bias, and narrow post-hoc analysis windows that fail to reflect real-world efficacy.

The first and most foundational critique of the reported efficacy figures came from Dr. James Lyons-Weiler, who introduced a dropout-inclusive method of recalculating VE in early 2021. Lyons-Weiler noted that the official efficacy estimates excluded entire categories of trial participants—those who had experienced COVID-19-like symptoms but were not PCR-confirmed, those who were removed due to protocol deviations, and most importantly, those who contracted COVID-19 or experienced adverse outcomes in the critical period immediately following vaccination. This early post-vaccination period, spanning from the first dose to 7 or 14 days after the second dose (Day 28 for Pfizer and Day 42 for Moderna), was not included in the primary efficacy analysis. This was precisely the window during which immunologic instability—including suppressed immune responses and antibody-dependent enhancement—would be most likely to occur.

Lyons-Weiler’s recalculated estimate, which included these excluded participants and periods, yielded a substantially lower vaccine efficacy: approximately 75%. This represented a deliberate rigorous methodological step to apply first-in-human trial standards, where the default assumption is to include all enrolled participants in safety and efficacy assessments unless compelling, documented reasons exist for exclusion. His recalibration demonstrated that the headline efficacy claims had been made conditional on the removal of precisely those data that would tend to reduce the appearance of benefit.

Building upon and validating this analytical challenge, Dr. Peter Doshi later published a related critique drawing attention to the fact that Pfizer’s trial had excluded 1,594 cases of “suspected” COVID-19—symptomatic cases that were not confirmed by PCR testing. These were almost evenly distributed between the vaccine and placebo groups, with 1,816 cases in the placebo arm and 1,778 in the vaccine arm. Had even a modest fraction of these cases been confirmed, the vaccine’s reported efficacy would have dropped dramatically. In fact, Doshi calculated that if all suspected cases were counted, the Pfizer vaccine's efficacy might fall to just 19%. This analysis did not negate the possibility of short-term benefit. But it did expose how the high efficacy narrative rested on analytical choices that excluded data which, if considered, would have substantially lowered the reported benefit.

Further generalization of the case exclusion problem emerged in 2023 through the work of Fenton and Neil, who described a broader statistical manipulation they termed “case counting window bias.” They demonstrated how observational datasets, such as those from the UK Office for National Statistics (ONS), frequently misclassified early post-vaccination infections, hospitalizations, and deaths as occurring in “unvaccinated” individuals, artificially deflating case numbers in the vaccinated group. While their work focused more on post-market surveillance than on the trials themselves, it contributed a broader conceptual framework that aligned with and extended earlier critiques of what Lyons-Weiler and Doshi had already identified in the Phase III data: that vaccine efficacy figures were not only inflated, but structurally rigged by selectively counting cases only after the immune system was presumed to have achieved maximal antibody production—while discarding all preceding immunological events.

The trial designs not only omitted the period most susceptible to immune dysfunction but also relied exclusively on relative risk reduction (RRR) while avoiding disclosure of absolute risk reduction (ARR), which in the case of the Pfizer trial was a mere 0.84%. The ARR provides a more realistic measure of the likelihood that a vaccinated individual avoids illness compared to an unvaccinated individual. Focusing solely on RRR, without contextualizing it with ARR, obscures the actual benefit, especially when the background incidence of disease is low.

The artificial nature of the efficacy estimates quickly unraveled once the vaccines entered the real world. In early 2021, data from Israel showed that vaccine effectiveness against infection had dropped to 39% by July. The United Kingdom reported in The Lancet that by 20 weeks post-vaccination, efficacy against Delta had declined to below 60%. In the United States, CDC reports during the Omicron wave showed that vaccine effectiveness against infection had fallen to under 20%—even with boosters. Although protection against hospitalization and death remained higher for a time, even this benefit diminished with time and age. The elderly, immunocompromised, and chronically ill—all underrepresented in the original trials—began to experience breakthrough infections, severe disease, and death at rates that contradicted the initial efficacy claims. Booster doses provided a temporary restoration of antibody titers, but even that protection waned rapidly, often within three to four months.

When the full context is included—dropouts, suspected cases, early post-vaccination events, and real-world variant-driven escape—the early efficacy claims can no longer stand as valid representations of what the vaccines actually delivered. What had been framed as a 95% reduction in symptomatic COVID-19 was, in practice, a transient benefit that rapidly eroded. The trial numbers were not wrong per se; they were simply the product of a highly restricted analytical frame that excluded all of the data that would have made the vaccines appear less effective.

Taken together, Lyons-Weiler’s dropout-inclusive recalculation, Doshi’s confirmation of excluded cases, and the Fenton/Neil window bias framework demonstrate that the reported efficacy was more a function of when and whom the trials decided to count than of any stable biological effect. Once those constraints are removed, the case for meeting the EUA’s "may be effective" standard begins to fall apart.

Systemic Underestimation of Risk

While the efficacy claims of the SARS-CoV-2 mRNA vaccines have proven to be conditional, transient, and in some cases illusory, the story of how risk was assessed—or more accurately, how risk was precluded from being fully assessed—reveals an equally serious breakdown in scientific integrity. The Emergency Use Authorization process demands not only a favorable risk-benefit profile but also that potential risks be rigorously evaluated and transparently reported. In the case of the mRNA vaccines, this standard was not met. Instead, structural and procedural choices at every level of development and review systematically minimized, obscured, or delayed detection of potential harms. These decisions reflected systemic gaps in oversight and protocol design that had downstream regulatory implications.

To begin with, the Phase III trials for Pfizer and Moderna both employed designs that sharply curtailed the detection of early or rare adverse events. Participants were followed for a median of just two months prior to the EUA application submission, far shorter than the six to twelve months typical for biologics intended for healthy populations. The control groups—initially randomized and blinded—were rapidly compromised. After the EUA was issued, placebo recipients were offered the vaccine, effectively destroying the blinded control arm. This prevented long-term, randomized comparisons from being made and eliminated the very conditions necessary to detect delayed onset events, such as autoimmune conditions, cancers, or neurodegenerative sequelae.

Moreover, the safety evaluations performed during this truncated window focused largely on solicited adverse events such as injection-site pain, headache, or fever. Serious adverse events (SAEs) were monitored but were exceedingly rare given the small sample sizes and short timelines. The trials were underpowered to detect events that might occur in 1 in 10,000 or 1 in 100,000 individuals—rates that, while statistically small, become critically important when billions of doses are administered. Furthermore, important categories of participants—such as pregnant women, individuals with autoimmune diseases, and the elderly with comorbidities—were either excluded from or underrepresented in the trials, thereby preventing reliable inferences about safety in those high-risk populations.

One of the most profound omissions in the safety review was the absence of any testing for pathogenic priming or molecular mimicry—mechanisms known to underlie autoimmune pathology following viral infection or vaccination. Previous coronavirus vaccine efforts, particularly for SARS and MERS, had already revealed that spike-protein-based vaccines could induce a form of immune pathology upon subsequent viral exposure. In animal models, vaccinated subjects developed eosinophilic infiltrates in lung tissue and liver inflammation following challenge. These events were indicative of what was euphemistically called “immune enhancement,” but more precisely described as pathogenic priming: a condition in which the immune system, having been sensitized to a viral protein, mistakenly targets homologous human proteins and initiates an autoimmune response.

Despite these established risks, no epitope exclusion screening was conducted for the mRNA vaccine antigens, and no animal studies were performed to assess pathogenic priming risk using the actual mRNA formulations authorized for human use. Instead, reliance was placed on generic preclinical assessments in macaques, even though previous research had demonstrated that ferrets—more sensitive to pulmonary immunopathology—were the more appropriate model for this type of vaccine. The decision to skip ferret studies and to not examine potential epitope homologies between the spike protein and human proteins constitutes a glaring breach of first-in-human trial safety expectations.

Real-world data have since validated many of these early concerns. Adverse event reporting systems such as VAERS, EudraVigilance, and Yellow Card have shown marked increases in reports of myocarditis, pericarditis, menstrual irregularities, neurological complications, and autoimmune disease following mRNA vaccination—often at rates higher than those observed for other licensed vaccines. While passive surveillance systems cannot confirm causality, they do provide signal detection, and the sheer volume and severity of these signals warranted more urgent and transparent follow-up than what occurred.

Additionally, rare but serious risks such as post-vaccination myocarditis in young males, vaccine-induced thrombotic thrombocytopenia (VITT), and post-acute COVID vaccine syndromes (PACVS) emerged only after mass deployment. These risks were either absent from trial reports or not detected due to low sample size and follow-up duration. Moreover, several immunological studies have now shown cross-reactivity between anti-spike antibodies and human tissue antigens, as well as the presence of autoantibodies following vaccination. These findings, consistent with predictions from bioinformatic analyses such as those by Lyons-Weiler and others, support the hypothesis that molecular mimicry and immune dysregulation were not only plausible risks but real and measurable outcomes.

Finally, the most systematic form of risk suppression came not through design flaws but through narrative control. Public health agencies and regulatory authorities downplayed or dismissed adverse event signals, discouraged open debate, and often framed post-market injuries as coincidental or anecdotal. Scientists raising concerns about immune pathology, waning efficacy, or trial design were marginalized, censored, or publicly maligned. This suppression of dissent ensured that the risk side of the EUA risk-benefit calculus remained artificially low, both in perception and in policy.

In total, the systemic underestimation of risk in the development, authorization, and deployment of mRNA vaccines was not a single failure but a cascade of omissions: of long-term data, of key populations, of established autoimmune risk screening, of appropriate animal testing, of meaningful post-market pharmacovigilance, and of open scientific discourse. These failures—individually serious—together form a case that the EUA process for the mRNA vaccines did not meet the standard of “known and potential benefits outweighing known and potential risks,” as required by law. The next section will examine one of the most underappreciated components of this risk landscape: the immunological and molecular mechanisms by which vaccine-induced autoimmunity and immune suppression may have occurred.

Immune Suppression and Autoimmunity via Pathogenic Priming and Molecular Mimicry

One of the most neglected yet crucial domains of risk in the development and authorization of the SARS-CoV-2 mRNA vaccines is the possibility of immune dysregulation through mechanisms long known to follow viral exposure: namely, pathogenic priming and molecular mimicry. Far from hypothetical, these risks were foreseeable, measurable, and, in many cases, supported by the scientific literature prior to vaccine rollout—yet they were systematically excluded from pre-authorization safety assessments. The failure to screen for these immunopathological mechanisms represents not just a scientific oversight, but a breach of ethical and regulatory due diligence.

In early 2020, Dr. James Lyons-Weiler published the first peer-reviewed analysis identifying SARS-CoV-2 vaccine-induced autoimmunity as a predictable risk due to shared peptide homology between viral and human proteins. His paper introduced the term “pathogenic priming” to describe a specific kind of immune sensitization—one in which prior exposure to viral antigens (via infection or injection) could prime the immune system to attack host tissues when later challenged. The study used bioinformatic tools to compare all immunogenic epitopes of SARS-CoV-2 with human proteins, revealing extensive local homology. Of the 37 SARS-CoV-2 proteins evaluated, 36 contained epitopes with high degrees of identity to human proteins, including those critical to antigen presentation (MHC Class I and II), PD-1 signaling, and other immune regulatory pathways.

This foundational work laid the groundwork for dozens of later studies, many of which cited Lyons-Weiler’s analysis as the first to systematically warn of these mechanisms in the context of COVID-19. Vojdani et al. confirmed that antibodies generated against SARS-CoV-2 proteins exhibited cross-reactivity with more than two dozen human tissue antigens, including those in the thyroid, pancreas, brain, and gut. Similar findings were reported by Lerner et al., who found that SARS-CoV-2-specific antibodies cross-reacted with enteric tissues, suggesting a mechanism for vaccine-related gastrointestinal autoimmune symptoms.

Multiple case series and mechanistic reviews have since linked post-vaccination syndromes—including myocarditis, immune thrombocytopenia, neurological disorders, and autoimmune hepatitis—to molecular mimicry between SARS-CoV-2 spike protein and host tissues. Moreover, the immune dysregulation observed in many cases of “Long COVID” and Post-Acute COVID-19 Vaccine Syndrome (PACVS) shows patterns of multisystem inflammation and autoantibody production, implicating mimicry-related autoimmunity as a shared pathophysiological thread.

Lyons-Weiler’s model provided not just correlation, but causally plausible targets. Homologies were found in peptides expressed in immune tissues such as B-cells and T-cells, as well as in neural, cardiac, and gastrointestinal proteins. In animal studies of SARS and MERS, similar spike-protein-based vaccines had previously triggered lethal eosinophilic immune reactions, particularly upon viral challenge. Despite this precedent, the mRNA vaccine developers skipped ferret models—known for their sensitivity to such pathology—and instead used non-challenged macaques, thus eliding the most relevant safety signal.

This regulatory negligence was further compounded by the decision not to conduct human autoantibody screens prior to rollout. Epitope exclusion—long considered best practice in autoimmune-aware vaccine design—was never implemented. Thus, not only were known mechanisms ignored, but no new data were generated to rule out their activation.

What makes this failure especially egregious is that the molecular mimicry profile of SARS-CoV-2 was atypically broad. Later studies would show that even variants such as Omicron 21L generated new, de novo mimicry risks in serotypes such as HLA-A24:02 and HLA-B27:05. These findings confirm that mimicry is not merely a relic of the original Wuhan strain but a moving target—one that evolves with the virus and continues to pose autoimmune risk in both natural infection and vaccine exposure.

Together, these studies confirm Lyons-Weiler’s initial prediction and broaden its clinical relevance. The autoimmune and immunosuppressive consequences of pathogenic priming are no longer theoretical—they are evident in the clinical literature, autopsy findings, and serological data. Yet the EUA process failed to incorporate any of this risk into its calculus.

In sum, the scientific community was warned—clearly, early, and in published peer-reviewed literature—that pathogenic priming posed a real threat. Subsequent literature did not overturn that conclusion; it reinforced it. That the FDA granted EUA to products encoding a protein with known homology to critical human proteins—without requiring epitope exclusion, autoantibody screening, or proper animal models—represents a systemic failure of immunological risk management.

The next section will explore how these risks and omissions violated the core legal and ethical conditions of the Emergency Use Authorization framework and why the FDA's decision cannot be justified under its own statutory guidelines.

Collapse of the EUA Framework: The Vaccines Never Qualified

The Emergency Use Authorization (EUA) statute was not written to accommodate speculative hope. It was crafted for extraordinary measures under extraordinary conditions, with one core obligation: that a product demonstrate, on the totality of available evidence, that it may be effective, that its known and potential benefits outweigh its known and potential risks, and that no adequate, approved, and available alternatives exist.

The mRNA vaccines for SARS-CoV-2 were authorized under these terms. The basis for that authorization was a series of highly publicized, rapidly interpreted efficacy claims—95% for Pfizer, 94.1% for Moderna—derived from data that was selectively counted, contextually constrained, and fundamentally unsuited for long-term or real-world generalization.

When adjusted for broader real-world conditions and inclusive analysis, those efficacy claims are materially weakened. Under such conditions, the original EUA justification becomes difficult to sustain.

The plots below present the real-world behavior of vaccine efficacy (VE) over time, recalculated from first principles with essential corrections applied. The initial VE has been adjusted from 95% to 75%—a recalibration justified by the reintegration of patients who were dropped from the trial analyses, including suspected COVID-19 cases and those who contracted illness before the arbitrary case-counting windows began. This dropout-inclusive estimate, first introduced by Lyons-Weiler and later echoed by Doshi, restores the integrity that first-in-human trials require but the Phase III protocols abandoned.

To this corrected efficacy, further necessary adjustments have been made: decay of immunity over time, reduced effectiveness against emerging variants, increased vulnerability in real-world populations, and a 5% penalty for the untested causal risks such as antibody-dependent enhancement or immune suppression. Wide confidence intervals are included to reflect the uncertainty that regulators ignored.

Figure 1. Dropout-Inclusive VE Against Infection (Adjusted)
Recalculated from 75% base efficacy, applying 7.5% monthly decay, 20% reduction for Delta, 70% for Omicron, a 1.5× population vulnerability scale, and a 5% penalty for causal uncertainty.

Even under the most generous assumptions, efficacy against infection does not remain above 50% for any variant beyond six months. Omicron fails immediately. Delta collapses by the fifth month. Alpha—the original and most favorable context—drops below 60% within a year. These are the expected curves, based on empirically supported rates of decay and variant escape.

They fall short of what would reasonably be required to justify emergency use under the statute

What of the commonly invoked fallback—that even if protection against infection fades, defense against hospitalization or severe disease remains robust? The second plot answers that question with equal clarity.

Figure 2. Dropout-Inclusive VE Against Hospitalization/Severe Disease (Adjusted)
Starting from 90% for Alpha, with slower waning (3%/month), 15% reduction for Delta, 30% for Omicron, and the same vulnerability and uncertainty adjustments applied.

Here too, the fall is real. Omicron efficacy against hospitalization drops below 50% within eight to nine months. Delta declines steadily, reaching similar levels by one year. Only Alpha shows durability, and even it erodes beneath the legal and ethical threshold for emergency authorization when the totality of population risk is considered.

These figures are not exceptions. They are the rule when trial data is subjected to the kind of scrutiny that a first-in-human medical product demands. They show that efficacy was too low, not durable, not variant-proof, and not ethically presented. What the public received was a curated moment of peak performance, frozen at the most favorable point on the curve and offered as if it were the whole trajectory.

In legal terms, the EUA requires more than optimism. It demands evidence. If the FDA had been presented with these corrected efficacy curves—curves reflecting realistic decay, true biological heterogeneity, and the full population spectrum—then the conclusion would have been unavoidable: the vaccines do not and did not meet the “may be effective” threshold. Not against infection. Not against transmission. Not even against hospitalization, once decay and variant escape are acknowledged.

There is no room within the EUA statute for a product that loses meaningful protection within months. Nor is there protection for authorizations built on selectively reported data, or efficacy propped up by assumptions rather than proof. The use of efficacy endpoints that excluded early adverse events—where mechanisms like ADE were most likely to occur—and the disregard for dropout-inclusive recalculation represent a fundamental breach of the evidentiary standards the EUA demands.

The EUA may have been granted on the basis of incomplete or overly narrow evidence. These analyses and these curves are not speculations or forecasts. They are post-mortems. The authorization of the SARS-CoV-2 mRNA vaccines under EUA was not just premature—it was mathematically and legally unjustifiable from the outset.

Recommendations and Reform: Restoring Integrity to the Emergency Use Process

The collapse of the EUA justification for the SARS-CoV-2 mRNA vaccines, as demonstrated by the corrected efficacy curves, is not merely a retrospective failure. It is a warning. If unheeded, it guarantees repetition. The public was told that the vaccines were highly effective based on selectively framed data and arbitrary case-counting thresholds. Regulators authorized them on the basis of legal criteria that were not, in truth, met. The result was widespread policy built on error—mandates, coercion, public messaging, and clinical interventions deployed in the name of a benefit that did not persist and, in many cases, never existed at all.

To prevent future medical and regulatory disasters of this magnitude, structural reforms are not optional. They are required. The following recommendations are necessary to restore scientific integrity, rebuild public trust, and align emergency public health responses with the ethical standards demanded when the lives of millions are at stake.

6.1 Require Dropout-Inclusive Efficacy Calculations

No product—especially not a first-in-human gene therapy platform—should ever be authorized under EUA or any other mechanism without including all trial participants in the efficacy calculation. The exclusion of suspected cases, early adverse events, and those who did not complete dosing as scheduled distorts the data, inflates efficacy, and conceals early harm. The dropout-inclusive recalculation method used in this report must become a mandatory part of all EUA evaluations.

6.2 Abolish Arbitrary Case Counting Windows

The practice of beginning efficacy measurement at 7 or 14 days after the second dose is indefensible. It creates a statistical void exactly where the most important safety and efficacy data should be gathered—during the immune priming phase, where risks like ADE and immune suppression are most likely to manifest. This design artifact alone was sufficient to obscure short-term harm and overstate benefit. All events post-first dose must be included in both safety and efficacy analyses.

6.3 Mandate Variant-Specific Reassessment and Ongoing Decay Modeling

Vaccines that rely on sequence-specific antigen presentation must be continuously reevaluated against circulating variants. An EUA granted for Alpha cannot be presumed valid for Delta or Omicron without updated decay modeling, real-world performance data, and explicit demonstration that efficacy remains within the statutory range. The absence of such reassessment in the case of the mRNA vaccines represents regulatory negligence.

6.4 Restore and Codify Independent Risk-Benefit Audits

All EUA applications must be subject to independent review by experts without financial ties to the product developer or affiliated policy bodies. These reviews must include population-level modeling of benefit decay, risk stratification, and confidence-weighted projection. The absence of an adversarial review process created an echo chamber in which assumptions replaced data and efficacy thresholds were never rigorously challenged.

6.5 Forbid Expansion of EUA Products Without Durable Efficacy

If protection against infection or severe disease decays below 50% within six months—as shown for the mRNA vaccines in the corrected analyses—then the product must not be expanded to additional populations, especially children, infants, or pregnant women. Expansion under decay is expansion under fraud.

6.6 Require Transparent Raw Data Disclosure

EUA cannot coexist with secrecy. Every dataset used to justify an emergency authorization must be publicly released before authorization is granted. This includes all excluded participants, all adverse events (including those not classified as “serious”), and all dropout reasons. Anything less constitutes a breach of public trust and scientific duty.

6.7 Revoke EUA When Evidence Fails to Meet Thresholds

The failure to revoke the EUA for the mRNA vaccines after waning efficacy became indisputable was a dereliction of regulatory responsibility. The EUA statute contains no provision for permanent authorization in the face of performance collapse. When updated data contradict the basis for an EUA, that EUA must be suspended, not silently carried forward as a political fait accompli.

---

This is not a matter of political alignment or ideological grievance. It is a matter of epistemological collapse. What we witnessed was not merely a misjudgment—it was the institutionalization of shortcuts. If such behavior is allowed to stand, then emergency use becomes a loophole, not a safeguard. Science becomes branding. Risk becomes acceptable so long as it is uncounted.

The analyses presented in this report do not forecast the future. They reconstruct the past—faithfully, empirically, and without compromise. The conclusions they force upon us are uncomfortable, but clarity does not wait for comfort. A product that loses its protective value within months, and was only ever effective in theory, should never have been authorized.

To prevent this from happening again, the reforms above must be adopted. If not, we have learned nothing. And the next collapse will not come as a surprise. It will come as a design.

Appendix A. Modeling Assumptions and Implementation

This appendix details the structure, assumptions, and implementation of the models used to recalculate and plot dropout-inclusive vaccine efficacy (VE) for SARS-CoV-2 mRNA vaccines (Pfizer-BioNTech and Moderna), as shown in Figures 1 and 2. The models simulate real-world VE over time, correcting for limitations in clinical trial design, biological complexity, and population heterogeneity. They provide adjusted efficacy trajectories for infection and hospitalization/severe disease across Alpha, Delta, and Omicron variant contexts.

A.1. Objective and Design Rationale

The modeling objective was to test whether mRNA vaccines met EUA effectiveness thresholds when dropout-inclusive trial estimates, known waning, variant escape, and underexamined risks such as antibody-dependent enhancement (ADE) were accounted for. Unlike trial estimates fixed to narrow endpoints, these analyses simulate VE continuously from first dose onward and are intended to reflect first-in-human standards of rigor.

A.2. Dropout-Inclusive Initial VE (Alpha Baseline)

The starting VE for Alpha was anchored at 75%, reflecting recalculations that include participants excluded from primary analyses due to:

• Illness prior to the case-counting window (7 or 14 days post-second dose),

• Protocol deviations,

• Classification as “suspected” but PCR-unconfirmed cases.

This dropout-inclusive estimate, introduced by Lyons-Weiler and others, is both statistically conservative and ethically necessary for first-in-human evaluations.

A.3. Modeling Pre-Delay VE

To address the immunologically critical period prior to peak antibody generation, the model assigns a weighted average VE for the first month post-dose two, incorporating the pre-delay period. For Days 1–14 post-first dose, VE was set to 50%, rising to 75% thereafter:

VEmonth 1=14⋅0.50+16⋅0.7530≈63%\text{VE}_{\text{month 1}} = \frac{14 \cdot 0.50 + 16 \cdot 0.75}{30} \approx 63\%VEmonth 1​=3014⋅0.50+16⋅0.75​≈63%

This correction acknowledges that both breakthrough infection and ADE are most likely in the early post-vaccination period—a time explicitly excluded from EUA-era efficacy calculations.

A.4. Variant-Specific Reductions

Efficacy against Delta and Omicron was reduced from the Alpha baseline:

• Delta: 20% reduction → VE₀ = 0.75 × 0.80 = 0.60

• Omicron: 70% reduction → VE₀ = 0.75 × 0.30 = 0.225

These values were applied multiplicatively at t=0t = 0t=0 in the waning model. They reflect well-documented reductions in neutralizing antibody effectiveness and breakthrough infection rates.

A.5. Time-Dependent Waning

VE decay was modeled using an exponential function:

Where:

Decay constants used:

• Infection protection: k=0.075 (7.5% loss per month)

• Hospitalization/severe disease: k=0.03 (3% loss per month)

The slower decay for severe disease reflects greater durability of T-cell–mediated immunity and innate immune memory, while still acknowledging eventual decline due to viral evolution and immunosenescence.

A.6. Booster Dose Exclusion

This model excludes the effects of booster doses to isolate the performance of the initial two-dose regimen authorized under the EUA. While real-world data from 2022 and 2023 includes boosted individuals, the EUA was granted in December 2020 based solely on short-term efficacy following the second dose. Boosters, introduced later under separate regulatory authority, do not retroactively validate the original EUA.

A.7. Population Vulnerability Adjustment

To reflect differences between trial and real-world populations, VE was adjusted downward using a 1.5× vulnerability scale, simulating reduced protection in elderly or immunocompromised groups:

This reflects widely observed higher rates of breakthrough infection and severe outcomes in at-risk populations.

A.8. Penalty for Untested Risks

A 5% penalty was applied to all adjusted VE values to account for unmeasured biological risks not addressed in the EUA process, including:

• Antibody-dependent enhancement (ADE),

• Molecular mimicry and pathogenic priming,

• Lack of epitope screening and inadequate animal challenge models (e.g., absence of ferret studies).

This penalty reflects the precautionary principle and avoids the error of assuming safety in the absence of mechanistic evaluation.

A.9. Final Adjustment Equation

The final efficacy value at time t was computed as:

This equation was applied for each variant separately.

A.10. Confidence Intervals

To reflect both parameter and model uncertainty, wide relative confidence intervals (CIs) were plotted:

• Alpha: ±20% of VE

• Delta: ±25%

• Omicron: ±30%

These were selected to encompass plausible variation due to modeling assumptions, data exclusions, demographic diversity, and biological unknowns.

A.11. Scenario Robustness (Optional Extensions)

Though not shown in the main figures, this model framework supports sensitivity analysis by varying:

• Decay rates: k=0.05 to 0.10 for infection, 0.02 to 0.05 for severe disease,

• Population risk factor: scaling from 1.2× to 2×,

• Causal risk penalty: 0% to 10%,

• Pre-delay VE: 25–50%.

These variations can help simulate best- and worst-case outcomes and support robust policy analysis.

A.12. Computational Implementation

The model was implemented in Python using NumPy and Matplotlib. Time was simulated over a continuous 0–12 month period using 100 time steps. Variant-specific VEs were plotted along with shaded CI bands. The computational code is available upon request for replication or audit.

Empirical Reality Check on the Shape of the Curve

The clearest validation of the dropout-inclusive, risk-adjusted vaccine efficacy (VE) model presented in this report lies in the alignment between its predictions and real-world data. The theoretical shape of the VE curve, when modeled correctly, does not stand in opposition to empirical findings—it matches them.

Between March and August 2021, several real-world surveillance efforts reported VE against SARS-CoV-2 infection for the Pfizer-BioNTech and Moderna vaccines. These data showed that Pfizer's VE declined from approximately 85 percent to below 40 percent over a five-month span, while Moderna began near 90 percent and dropped to about 65 percent in the same period. (See publication here; Cohn et al., 2021). These figures, derived from 1 - hazard ratio analyses, represent some of the most robust empirical VE estimates during the post-deployment period of the two-dose regimens.

When these empirical curves are viewed alongside the dropout-inclusive model developed here—anchored at 75 percent and adjusted downward for real-world vulnerability, unmeasured causal risks such as ADE, and variant escape—the alignment is clear. Our model, which begins at approximately 47.5 percent after applying a 1.5× population vulnerability adjustment and a 5 percent causal penalty, tracks a smooth exponential decay of 7.5 percent per month. The modeled curve starts lower than the empirical curves, reflecting a more conservative and ethically rigorous correction for exclusions and early risk. Yet it mirrors the trajectory of decline observed in the empirical data.

Figure 3a presents this initial comparison. The empirical Moderna and Pfizer curves appear to perform better in the early months, but their rate of decline is sharp and consistent. The modeled curve, although lower at baseline due to the inclusion of dropped participants and known risk amplifiers, follows a similar trajectory. It does not conflict with the empirical record—it explains it, while correcting for the overly optimistic assumptions embedded in clinical trial design and post-market narrative construction. Note the pathway toward negative efficay is clear in the original figure (show here as 3b).

Figure 3a. Empirical versus Dropout-Inclusive Modeled Vaccine Efficacy (Infection Only)
This figure compares empirical vaccine efficacy against infection for Pfizer-BioNTech and Moderna, reported from March to August 2021, to the dropout-inclusive, risk-adjusted model developed in this report. The empirical curves show steep declines—Pfizer from 85% to below 40%, Moderna from 90% to 65%—while the modeled curve begins at approximately 47.5% and decays at a constant rate of 7.5% per month. The modeled trajectory reflects a conservative correction for participant exclusions, real-world population vulnerability, and unmeasured risks such as antibody-dependent enhancement (ADE). Although starting lower, the modeled curve aligns with the direction and magnitude of observed VE loss, validating the modeled decay structure.

Figure 3b. Empirical Vaccine Efficacy Over Time by Manufacturer (Cohn et al., 2021)
This figure displays 1 minus the hazard ratio for infection over time (March–August 2021) for three COVID-19 vaccines: Janssen (blue), Moderna (orange), and Pfizer-BioNTech (gray). The vertical axis represents estimated vaccine effectiveness against infection compared to the unvaccinated population. Pfizer-BioNTech efficacy declined from approximately 85% to below 50% over five months, while Moderna showed more gradual decline, maintaining roughly 65% efficacy by August. Janssen showed a markedly steeper decline, nearing zero by Month 5. This empirical decay in vaccine performance corroborates the trajectory predicted by dropout-inclusive modeling and further illustrates the failure of the vaccines to sustain EUA-qualifying levels of protection within months of deployment.

When we apply our correction framework—dropout reintegration, vulnerability scaling, and risk penalties—to the empirical curves themselves, the picture becomes more decisive. Pfizer’s adjusted VE falls below 50 percent by the second month. Moderna, which performs more durably, still drops below 50 percent around the fourth month. The dropout-inclusive model remains between the two, affirming its function as a corrected population-level benchmark.

Figure 4 shows this result: the fully adjusted empirical curves and the modeled VE curve converge around the same end point—a collapse of population-level protection within months of full vaccination. Once stripped of trial exclusions and framed in real-world demographic and biological contexts, neither vaccine delivers protection that satisfies the minimum threshold for EUA eligibility.

Figure 4. Fully Adjusted Empirical Vaccine Efficacy versus Dropout-Inclusive Model
This figure presents the same Pfizer-BioNTech and Moderna empirical VE data after adjustment for trial exclusions, population risk (1.5× vulnerability factor), and a 5% penalty for untested immunological hazards. Once adjusted, Pfizer’s VE falls below 50% by the second month post-vaccination, and Moderna’s declines below 50% by the fourth month. The dropout-inclusive modeled curve remains between the two adjusted trajectories throughout, reinforcing its role as a conservative population-level benchmark. All curves converge toward the same conclusion: population-level protection against infection failed to meet EUA-sufficient thresholds within months of deployment.

The empirical shape of the curve, therefore, confirms rather than challenges the modeling work presented in this report. It affirms that VE, when honestly accounted for, was neither robust nor durable. It declined quickly, well before the public was informed, and long before regulatory oversight responded. The EUA was issued based on claims that VE exceeded 90 percent. Those claims depended on narrow timeframes, non-inclusive participant selection, and case counting strategies that excluded exactly the events most likely to signal early risk. Empirical data from 2021 make clear that even the optimistic narrative could not hold beyond the fifth month.

The corrected model never allowed it to begin with. That both paths arrive at the same conclusion—whether by reanalysis or direct observation—underscores the truth: the vaccines, as authorized under EUA, failed to meet the statutory requirement that a product "may be effective" at reducing infection. And the shape of the curve, when made visible without euphemism or omission, shows that this failure was neither sudden nor surprising. It was predictable, measured, and, in the regulatory process that followed, ignored.

This modeling framework demonstrates that, when evaluated fairly and realistically, the mRNA vaccines authorized under EUA did not maintain sufficient efficacy over time—especially against emerging variants—to justify continued emergency use status. The curves generated reflect a corrected, principled, and transparent reconstruction of the benefit profile that regulators failed to compute.