This is the strongest part of the piece. The list of infections with a long tail is not speculative. Group A strep causes rheumatic heart disease. Polio causes post polio syndrome. Tuberculosis leaves scarred lungs. Measles is the cleanest case. It wipes out 11% to 73% of a child's existing antibody repertoire, a change that MMR vaccination does not cause (Mina, Science 2019). Measles epidemics track non measles childhood deaths for two to three years afterward (Mina, Science 2015). A disease that erases immune memory is one where vaccine benefit has to be counted in years.

The figure of "at least 400 million" people with long COVID is a modeled cumulative estimate built on assumed incidence rates, with wide uncertainty (Al‑Aly, Nat Med 2024). The cardiovascular risk numbers after COVID come largely from the Veterans Affairs cohort (Xie, Nat Med 2022), which is mostly older white men, whose controls include people with undetected infections, and whose tested patients have more contact with the health system. Later studies in broader populations generally found smaller excess risks.

Epigenetic clocks run fast in people sick with COVID, more so with severe disease (Cao, Nat Commun 2022). Flu drives cellular senescence in mouse lungs, and clearing senescent cells improves repair. In mice, prior severe viral pneumonia leaves the lung full of tumor friendly neutrophils with lasting chromatin changes, and tumors grow faster. Vaccination blunted that effect (Qian, Cell 2026). The authors say plainly that definitive evidence is lacking.

The clocks in Cao were measured during or near acute illness, when inflammation and shifts in blood cell mix move methylation, and some of the acceleration reversed later. That reading reflects acute illness more than an older body. The human half of the cancer claim is an electronic record association (hazard ratio about 1.24 for lung cancer after COVID hospitalization) adjusted only for age, sex and smoking. People hospitalized with COVID get more chest CT scans, and an early cancer can make COVID more severe rather than the other way around. The Cell abstract still says the findings "establish a causal link." And the Cell paper comes from the senior author's own lab, with Cheon as a coauthor (see below).

This has randomized evidence behind it, which is rare for an "off target" claim. The meta analysis the authors cite pools six randomized trials (9,001 patients) and finds major cardiovascular events fell from 5.4% to 3.6%, a relative risk of 0.66 (Behrouzi, JAMA Netw Open 2022). In IAMI, a double blind trial of flu vaccine given right after a heart attack, death, heart attack or stent thrombosis fell from 7.2% to 5.3% at one year (hazard ratio 0.72), and all cause death fell too.

The benefit sits almost entirely in people with a recent acute coronary syndrome (relative risk 0.55). Without one, it disappears (1.00), and cardiovascular death was not significantly reduced overall. On the high dose question the editorial highlights, DANFLU‑2 missed its primary endpoint (flu or pneumonia hospitalization, 0.68% versus 0.73%). The heart failure finding is a secondary, exploratory outcome from an open label trial funded by the vaccine maker. The absolute difference was 0.11% versus 0.15%, roughly 2,500 people given high dose vaccine for each heart failure admission avoided. INVESTED, an earlier blinded trial of high versus standard dose in 5,260 high risk cardiac patients, found nothing (hazard ratio 1.06).

The French national health data study is enormous and well run. It compared 22.8 million vaccinated adults aged 18 to 59 with 5.9 million unvaccinated adults for a median of 45 months. Vaccinated people had 25% lower all cause mortality (weighted hazard ratio 0.75) and far fewer severe COVID deaths (0.26). The authors weighted for 41 comorbidities and started the clock six months in to avoid a known bias. As a safety finding it is reassuring. Four years after millions of mRNA doses, there is no excess death signal.

The editorial says this result "strongly" suggests a long term health benefit. Look at the paper's own cause of death table. Vaccinated people were also 27% less likely to drown (0.73), 26% less likely to die in a traffic crash (0.74) and 20% less likely to die in a fall (0.80). An mRNA vaccine does not teach anyone to swim. Those numbers measure the difference between people who show up for preventive care and people who do not, the healthy vaccinee effect. A drowning gap of about the same size as the all cause gap means the all cause gap cannot be read as a vaccine effect. This newsletter made the same point in August about a shingles vaccine poster that reported a two thirds cut in all cause death in one year (The Number That Gives It Away). The rule does not change because the conclusion is one I like.

The direction is consistent across studies and the mechanism is plausible, since a milder acute illness should leave less damage behind. In the Veterans Affairs data, one year long COVID incidence in the Omicron era was 3.5 per 100 in vaccinated people versus 7.8 per 100 in unvaccinated people, and the authors attribute most of the decline in long COVID over the pandemic to vaccination (Al‑Aly, NEJM 2024).

Every study is observational. A systematic review of 16 of them found odds ratios ranging from 0.22 to about 1.0, too inconsistent to pool, and rated the certainty low because of confounding (Byambasuren, BMJ Med 2023). Long COVID itself has no agreed definition, which is one of the editorial's own complaints.

This is the right prescription, and the best proof that the second axis is real comes from studies the editorial does not cite. In Wales and in Australia, shingles vaccine eligibility was set by date of birth, which created natural experiments where people a week apart in age had very different access. In Wales, eligibility cut new dementia diagnoses by 3.5 percentage points over seven years, about a 20% relative reduction (Eyting, Nature 2025). Australia found 1.8 points over 7.4 years, with no change in other preventive care that would signal healthier people self selecting (Pomirchy, JAMA 2025). And the oldest version of the argument is not controversial at all. HPV vaccination given before age 17 cut invasive cervical cancer by about 88% in a Swedish national cohort (Lei, NEJM 2020). Hepatitis B vaccine prevents liver cancer. Nobody calls those off target effects. We call them the reason for the vaccine.

Registries capture what gets coded, so "biological ageing" and "health span" will be measured by proxies for a long time. And a second axis has to count both columns. The same follow up that finds fewer heart failure admissions will find the rare long tail harms. For mRNA myocarditis the long tail is already measured. In France, vaccine associated myocarditis had fewer cardiovascular complications at 18 months than ordinary myocarditis (weighted hazard ratio 0.55), while myocarditis after COVID infection itself did no better than ordinary myocarditis (1.04). That is how the honest ledger looks. Small harms, counted, and set against benefits, counted.