Pfizer C4591001 Trial Audit Analysis: What the Records Show
An examination of claims involving participant records, adverse-event classification, deaths, manufacturing processes and trial oversight in the Pfizer/BioNTech COVID-19 vaccine trial.
The report in brief
What the authors are saying—in plain language
The OpenVAET authors reviewed released records from Pfizer/BioNTech’s large COVID-19 vaccine trial and compared participant lists, case files, adverse-event data, death reports and manufacturing records. Their central argument is that parts of the trial record do not line up cleanly and deserve an independent, record-level explanation.
Participant records
The authors found different participant totals between versions of investigator files and gaps in subject-ID sequences. They interpret these as possible record removal; the discrepancies alone do not prove deletion.
Adverse events
They argue that some medical events were absent from adverse-event totals or later treated as possible COVID-19 symptoms. Establishing why requires reviewing each case and its electronic change history.
Deaths and reporting dates
They question whether deaths known to Pfizer appeared consistently in regulatory and published summaries. The answer depends on when each death occurred, when Pfizer learned of it and which cutoff governed each document.
Manufacturing change
Most pivotal-trial doses used Process 1, while commercial production used Process 2. Regulators assessed analytical comparability, but the report argues that the direct Process 2 clinical group was too small to answer many safety questions by itself.
Trial-site conduct
The report incorporates Brook Jackson’s allegations about Ventavia sites, including consent, protocol and data-integrity problems. These remain allegations unless individually supported by inspection, regulatory, court or primary-document findings.
The practical takeaway
The report raises testable questions; it does not, by itself, prove fraud or intent. Audit trails, inspection findings and participant-level timelines are needed to determine what each discrepancy means.
Reading framework
Three layers that should not be collapsed
Documented
Directly in a primary record
Pfizer, FDA, EMA, protocol, trial-registration, case-report or regulatory material that a reader can inspect.
Discrepancy
A difference requiring explanation
A count, identifier, date, classification or methodological inconsistency visible in compared records.
Allegation
A conclusion beyond the record alone
Claims of intent, deletion, suppression, manipulation or misconduct require direct corroborating evidence.
Unresolved
Evidence needed
Audit trails, version histories, inspection findings, correspondence or case-level review that could resolve the question.
01 · Dataset comparison
Participant and investigator discrepancies
Discrepancy reported
Approximately 1,203 fewer randomized participants across 108 sites
What the authors reportA comparison of investigator files dated November 26, 2020 and March 29, 2021 produced different randomized-participant totals. The report also identifies approximately 301 gaps in sequential subject IDs.
What this does not establishA missing ID or changed total does not by itself prove that a participant record was deliberately deleted.
What would resolve itVersion-change documentation, database audit trails, data-cleaning logs, explanations of site exclusions or transfers, duplicates, screen failures and protocol deviations.
The report compares death counts across released records, the EUA review and published summaries. A responsible comparison must keep five dates separate: date of death, date Pfizer learned of it, database cutoff, submission date and publication date.
Unresolved
Different counts require a participant-level reporting timeline
Observed questionWere all deaths known to the sponsor represented consistently in each summary at its stated cutoff?
Possible administrative explanationsReporting lag, follow-up timing, database-lock timing or late source-document confirmation can produce different counts—but should be demonstrated, not presumed.
What would establish concealmentEvidence that a death was known, fell within the applicable reporting scope and was intentionally withheld. Different totals alone do not establish intent.
OpenVAET says gaps in sponsor-assigned AESPID sequences and selected case records indicate that events were omitted, classified differently or re-qualified as possible COVID-19 symptoms. The figures are the authors’ reconstruction and are not treated here as independently confirmed totals.
What a record may showAn event received a particular coding or classification at a particular time.
What it does not automatically showWhy the investigator chose that classification, whether the earlier entry was duplicative or provisional, or whether the purpose was to suppress a safety signal.
Required reviewCase-by-case CRFs, query history, MedDRA coding rules, protocol definitions, sponsor instructions and the electronic audit trail.
Former Ventavia regional director Brook Jackson alleged problems involving informed consent, eligibility, signatures, backdating or record alteration, protocol compliance and data integrity at trial sites operated by the company.
Allegation
These claims should not be converted into established findings. Each requires confirmation through source documents, an inspection, a regulatory determination, sworn evidence or a court finding. The BMJ reported Jackson’s account; publication of an allegation is not itself regulatory substantiation.
The manufacturing process changed for commercial scale
EMA’s assessment documents comparisons between Process 1 and Process 2 active-substance batches. It records a different DNA-template method and analytical comparisons involving sequence, structure, poly(A) tail patterns and expressed protein, while also recording questions and follow-up commitments. EMA concluded that submitted data supported consistent and comparable Process 2 quality.
Small direct clinical cohort
The report identifies 252 Process 2 vaccine recipients
OpenVAET derives 252 treatment recipients from patient-batch and randomization records, consistent with a protocol plan for approximately 250 participants aged 16–55. The underlying protocol and batch listings make the question testable, but the exact 252 count presented here remains attributed to the authors’ reconstruction.
A cohort this size has limited power for uncommon outcomes. The report discusses lymphadenopathy and menorrhagia, but comparisons in a small subgroup need confidence intervals, prespecified methods and multiplicity controls. A process difference does not by itself show that the commercial product was unsafe or clinically different.
The trial did not monitor only immediate reactions. The protocol and statistical plan included:
Solicited local and systemic reactions for seven days after each dose.
Unsolicited adverse events over defined post-dose periods.
Serious adverse events and deaths during longer follow-up.
Continued safety and efficacy follow-up beyond the initial EUA data cutoff.
The exact reporting window depends on trial phase, endpoint and protocol amendment. A short solicited-reaction diary should not be confused with the full safety-surveillance period.
Can be reproduced against identified source datasets.
Different manufacturing processes were used
Strong / documented
Described in EMA and manufacturing records.
Process 2 clinical sample was relatively small
Verifiable
Protocol, batch and randomization records permit counting.
Certain events received different classifications
Potentially verifiable case by case
Requires CRFs, coding and change history.
Pfizer intentionally suppressed adverse events
Not established by discrepancy alone
Requires evidence of intent.
Participant records were deliberately deleted
Not established from missing IDs alone
Requires audit trail and version history.
Trial fraud occurred
High evidentiary threshold
Requires direct documentary, regulatory or legal support.
Research agenda
Questions worth investigating
What explains participant-count differences between the referenced datasets?
Are audit trails available for removed or changed subject records?
When did Pfizer first become aware of each reported death?
Were classification decisions consistent with the protocol and MedDRA rules?
How did regulators evaluate Process 1 / Process 2 comparability?
What analytical and clinical bridging supported Process 2?
Did regulators independently investigate Ventavia allegations?
Were corrections, deviations, inspections, warning letters or database amendments issued?
Bottom line
Test the claims against the records
The OpenVAET report identifies questions that merit examination against Pfizer and regulatory records—particularly participant-dataset discrepancies, adverse-event classification, mortality-reporting timelines and the transition from Process 1 to Process 2 manufacturing.
Some questions are objectively testable from primary documents. A discrepancy, however, does not automatically establish fraud, suppression or intent. Those conclusions require database audit trails, regulatory correspondence, inspection findings, contemporaneous communications or other direct evidence.
The purpose of this page is neither to accept nor dismiss the report wholesale, but to separate what can be documented from what remains disputed.