Dimitri is an FPGA developer, mid-forties, no pre-existing condition. After the second mRNA shot he develops myocarditis. The finding comes from the clinic, not from Twitter. What if this were exactly the kind of case we should have talked about institutionally — before 160 million doses had been administered?
This is a report, not a proof. Dimitri is an anecdote, and anecdotes do not establish causality. What the epidemiology says is further down. The reason he stands here: he asks the right question. Not “was the vaccine wrong?” — but: what proof does a harm need in order to be taken seriously institutionally? And does the same bar apply to a promise?
The occasion is not the subject#
Since a study in JAMA Internal Medicine on 15 June 2026, Karl Lauterbach and Kay Klapproth, immunologist and methods commentator on X, have been arguing over whether the COVID booster protects the heart. We worked through that in Part 1. Short summary: the study is methodologically sound, the effect tiny (2 in 10,000), significant almost only among the over-75s. Both overreach — Lauterbach communicatively, Klapproth methodologically.
That is the occasion for this article. Not the subject.
The subject is a pattern older than this study and bigger than this fight: two different burdens of proof. One for the promise of mRNA technology. One for its documented harm.
The promise: 15 years, no standard#
BioNTech was founded in 2008 — not for viral vaccines, but for personalized cancer therapies. Uğur Şahin and Özlem Türeci promised in the years that followed that mRNA would make cancer treatable in the foreseeable future. This was not a fringe thesis — it was the institutional promise that carried investors, EU funding structures and media coverage for more than a decade.
As of mid-2026: no approved mRNA cancer therapy. No established remission induction in advanced tumors as a therapy standard.
What does exist is the following:
Adjuvant studies — no measurable tumor, after surgery: In a phase-2 study (Rojas et al., Nature 2023, cevumeran/BNT122, pancreatic carcinoma, n=16), a personalized mRNA neoantigen vaccine shows a remarkable benefit in recurrence-free survival — hazard ratio 0.08. But: there is no RECIST tumor regression. The patients no longer had a measurable tumor; the vaccine is meant to prevent recurrence, not push back an existing disease. This is recurrence prevention after surgery, not cancer therapy in the clinical sense.
The same applies to KEYNOTE-942 (mRNA-4157/V940, melanoma, n=157): 49 percent fewer recurrences after five years, HR ~0.51 — pure adjuvant prevention in the non-metastatic stage.
Manifest disease — conference data: At the AACR 2024 congress (abstract CT141, Bauman et al., phase-1 trial NCT03313778), 2 complete and 3 partial remissions were reported in HNSCC patients under mRNA-4157 + pembrolizumab — ORR 50 percent. These are real remissions. The finding stands as a conference abstract, not as a peer-reviewed full text. It deserves attention; it justifies no approval and no therapy standard.
The honest formulation is: personalized mRNA neoantigen vaccines show prolonged recurrence-free survival in adjuvant phase-2 studies — but no established remission induction in advanced tumors as a therapy standard. BNT122 plans the phase-3 start for 2026, BNT111 is in phase 2, BNT211 in phase 2/3.
Fifteen years and no standard. The promise stands. The evidentiary requirement for the narrative was low from the start.
What Ganymed was and what it was not: BioNTech spun off Ganymed in 2016 — Astellas paid more than 400 million euros. Zolbetuximab, today FDA-approved as VYLOY, has shown a real survival benefit in two phase-3 studies in gastric cancer with Claudin-18.2 expression: SPOTLIGHT (mOS 18.2 vs. 15.5 months, HR 0.75) and GLOW (mOS 14.4 vs. 12.2 months, HR 0.77). Zolbetuximab is a Claudin-18.2 antibody — not an mRNA therapeutic.
Ganymed was a niche and a strategic exit, no proof of mRNA cancer therapy and no failure. It does not belong in this argument, in either direction.
The harm: peer-review first ignored, then entered#
Myocarditis after mRNA vaccination is not an anecdotal signal. It is a peer-reviewed finding with consistent incidence estimates:
In young men after the second mRNA dose, the incidence lies at 1.3 to 3.1 per 100,000 persons (Deng et al., npj Vaccines 2025; Scott et al., NEJM 2025). The peak among men under 30 lies at about 1 : 16,750 — a rare but not negligible side effect in a cohort that was barely endangered by COVID-19 itself.
The European Medicines Agency (EMA) added myocarditis and pericarditis to the Comirnaty product information — on 8 July 2021 (EMA-EPAR Comirnaty, product-info update). By that point, around 160 million doses had already been administered in the EU.
No accusation against individual persons. It is the institutional reaction speed that is being recorded here.
The asymmetry#
Stefan Oelrich of Bayer spoke publicly in 2021 of mRNA vaccines being essentially gene therapy. This was no slip — he was formulating what the technology claims of itself: cellular instruction via RNA. The technology thereby received, for the promise (cancer, autoimmune diseases, personalized therapy), the institutional trust advance of a novel platform.
For the same novel character of the platform — that is, the fact that mRNA vaccines differ mechanistically from classical inactivated vaccines — the opposite applied on the subject of safety signals. Here, established surveillance structures were invoked, known adjuvant effects, “mostly mild and self-limiting.”
| Promise (cancer) | Harm (myocarditis) | |
|---|---|---|
| Evidence standard for the narrative | Phase-1/2 data; institutional advance from day 1 | Maximum bar despite peer-reviewed finding |
| Time to institutional reaction | 15+ years — narrative stands | ~6–9 months from approval to EMA entry |
| Communication | Cancer breakthrough as mission | Rare side effect, mostly mild |
| Correction | Promise still communicated unchanged | Product info adjusted — after ~160 million doses |
This is not malice. It is an interest structure: whoever profits from a kept promise carries no institutional pressure to correct. Whoever acknowledges a documented harm carries liability risk and communication effort. The burden of proof follows the incentive gradient.
Sideshow#
The heart-protection debate Lauterbach and Klapproth are conducting on X is a symptom of this pattern — not its origin. Whether the 2024/25 COVID booster protects the heart in men under 65 is irrelevant to most vaccination decisions of the last four years. Those decisions were made on the basis of the promise, not on the basis of this study.
The promise was: mRNA is the platform for the next medicine. Cancer. Autoimmune diseases. Personalized therapy. The network of investors, politicians and media-makers that carried this narrative had an institutional head start built with phase-1 data — and one that no product-info entry of 8 July 2021 takes away.
The harm needed peer-reviewed incidence data from several countries and 160 million administered doses — then myocarditis stood in the product information. The promise needed phase-1 data and a press release — then it stood in every German leading medium. The burden of proof was not equal. That was no accident.
