In Pfizer's Pivotal Trial, a Volunteer's Pericarditis Became a 'Suspected COVID' Case That Counted Against No One

A vaccine’s headline efficacy number is not measured so much as counted. In Pfizer’s pivotal COVID-19 trial, a volunteer’s illness counted against the vaccine only if it was logged as a confirmed, PCR-positive case of COVID-19. Everything else, every fever, every hospital stay, every strange new symptom, lived or died by how a clinician chose to label it in the record. So the single most consequential act in the whole apparatus is not the injection. It is the clerical decision, made weeks later, about what to call whatever happened next.

That decision is where the renewed scrutiny of the trial’s Argentine operation leads. TrialSite News has catalogued what it counts as 52 dosing errors and 20 regulatory findings at the Buenos Aires site, along with financial ties it says deserve a closer look. To see why a single national site can carry that much weight, you have to understand where Argentina sat in the trial’s architecture, and it was a stranger place than the public ever heard.

By the numbers Pfizer and BioNTech published in the New England Journal of Medicine, the trial injected 43,448 people across 152 sites: 130 in the United States, 9 in Turkey, 6 in Germany, 4 in South Africa, 2 in Brazil, and 1 in Argentina. That lone Argentine site was run out of Fundación INFANT and the iTrials-Hospital Militar Central in Buenos Aires by Fernando Polack, who was also the first author of the paper that announced the vaccine a success. So the physician running the trial’s only Argentine site was also the first author of the paper that announced the vaccine a success, and the sponsor that funded it had also designed the trial, conducted it, and analyzed the data. In ordinary times that arrangement alone would raise an eyebrow.

Now bring in the volunteer. According to the case files assembled by Paul Oosterhuis, Augusto Roux, subject 12312982, was a 35-year-old Buenos Aires lawyer who signed up partly to protect a mother with emphysema. He got his second dose on the evening of September 9, 2020, felt short of breath and chest pain within hours, and by mid-September was admitted to Hospital Alemán, where a CT scan showed a pericardial effusion, fluid around the heart, and a discharge note recording an adverse reaction to the vaccine. The trial record then did something odd with it. It first described a bilateral pneumonia unrelated to the injection, and then, on October 7, 2020, updated the adverse event to “suspected COVID-19 disease,” even though Roux’s PCR test had come back negative. The investigators judged the episode unrelated to the shot, and Pfizer concurred. By his account he was unblinded only in October, after he pressed for an answer, and he eventually took the matter to ANMAT, Argentina’s drug regulator, which made the investigators confirm he had received the vaccine rather than a placebo.

The mechanics of how the trial counted cases explain what that relabeling accomplished. A “suspected COVID-19” case with no confirmatory PCR is not a confirmed case under the trial’s own endpoint definition, so it never touched the efficacy tally the world was watching. And because the investigators had judged the event unrelated to the shot, it did not register as a vaccine injury either. Roux’s pericarditis was not refuted so much as reclassified into a space where it counted toward nothing. The released case files show he was not alone: a 66-year-old at the same Hospital Militar site, who got his second dose on September 19, 2020, carries a “Pericarditis” entry graded 2 and marked not related to the vaccine, with a suspected-COVID note logged on October 17 despite three negative tests.

We have seen this before, in the same trial, on another continent. In late 2021 The BMJ published regional director Brook Jackson’s account of Ventavia Research Group, a contractor that ran several of the trial’s sites in Texas, describing falsified data, unblinded patients, poorly trained vaccinators, and adverse events left for days without follow-up. She emailed her concerns to the FDA and was fired the same day. The agency granted the vaccine full approval in August 2021 without having inspected Ventavia’s sites. Two volunteers in Buenos Aires and one whistleblower in Texas are not the same evidence, but they rhyme, and what they rhyme about is the quiet machinery of coding and oversight rather than the chemistry in the vial.

A couple of miscoded cases cannot by themselves overturn an efficacy signal built on tens of thousands of people, and no honest reading claims they do. The worry is structural. This was a trial designed, run, and analyzed by its manufacturer, its only Argentine site directed by the paper’s own first author, policed by regulators who inspect a fraction of sites and, in these instances, inspected the troubled ones late or not at all. When the people grading the exam also wrote it, built it, and paid for it, the burden of proof does not sit with the volunteer in the hospital bed. It sits with the institution.

What is worth watching now is not another round of commentary but the paper trail. The Argentine case files surfacing in litigation and document releases are the first primary records that let outsiders check the coding against the clinical reality, and ANMAT already holds the one ruling that forced a straight answer out of this trial. The next move belongs to the regulators who can still pull the Buenos Aires site’s raw records, and to the inspectors who can still visit four years after the fact. Whether they do will tell us more about how American and Argentine drug oversight actually works than any efficacy press release ever did.

Sources

1. New England Journal of Medicine (PMC) – Polack et al., “Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine” (2020)
2. The BMJ – Thacker, “Covid-19: Researcher blows the whistle on data integrity issues in Pfizer’s vaccine trial” (2021)
3. David Healy – released Buenos Aires trial case files and adverse-event records
4. Paul Oosterhuis – “A case study in fraud?” (account of subject 12312982, Augusto Roux)
5. TrialSite News – “Pfizer’s Argentina COVID-19 Trial: 52 Dosing Errors, 20 Regulatory Findings, and Financial Ties Under Scrutiny”