By Eyal Shahar at Brownstone dot org.

Dear Editor:

In an ideal world of uncensored Covid science, I would have submitted this letter through the online submission website. However, my experience in 2021 and 2022 and more recently has taught me that there was zero chance that you would have published this text.It has been over four years since the following letter to the editor was published in your journal, but I discovered it only last month. I think there is no expiration date for the search for the truth, and I hope you agree.

Relying on data from 280 nursing homes across 21 states, the authors concluded: "These findings show the real-world effectiveness of the mRNA vaccines in reducing the incidence of asymptomatic and symptomatic SARS-CoV-2 infections in a vulnerable nursing home population."

That is far from the truth.

First, they did not report a single estimate of effect, such as a risk (probability) ratio. That the authors conclude "real-world effectiveness" without showing any estimate is astonishing. It is also astonishing that peer-reviewers or the editorial board have allowed that to happen.

Second, in every nursing home an unvaccinated resident was followed at least three weeks longer than a fully vaccinated resident, so their risk (probability) of infection was higher. Time at risk was neither reported nor considered.

Third, a key risk ratio I will shortly compute from the data is confounded by time trends in the background risk of an infection.

Fourth, comparing the risk ratio of infection (mucosal immunity) with the risk ratio of symptoms if infected (systemic immunity), we observe implausible results.

Lastly, a rudimentary correction suggests near-zero effectiveness of two doses of an mRNA vaccine in this population.

To set the record straight, I offer a peer review of the study and show several risk ratios.

The first dose of an mRNA vaccine was administered on December 18, 2020. Follow-up of nursing home residents who received two doses began at least 21 days later, on January 8, and lasted till March 31. The timeline is shown in the figure along with the epidemic curve.

Unvaccinated residents "were present at their facility on the day of the first vaccination clinic" (i.e., at the time of the first dose, if administered by February 15) and were not vaccinated by March 31. Therefore, in every facility, the follow-up time of unvaccinated residents was three weeks longer if the second dose was the Pfizer vaccine and four weeks longer if it was Moderna.

Moreover, follow-up of unvaccinated residents in some nursing homes started between December 18 and January 8. Not only was it earlier, but that was a period of high risk of infection just before the peak of the winter wave (see figure). All two-dose recipients were spared that early, high-risk exposure time. This bias-confounding by time trends in the background risk-has operated in other "real-world" studies from that time.

The bias is worse if the follow-up is delayed until 14 days after the second dose (to allow full immunity). In this case, follow-up of recipients of two doses began on January 22, ten days after the peak.

Using data from Table 1 in the letter, I computed three risk ratios (RR). In every facility, day 0 for the unvaccinated was 3-4 weeks earlier than day 0 for two-dose recipients.

The key number is the risk ratio of symptomatic infection. It is 0.1 (90% vaccine effectiveness). Surprisingly, the mRNA vaccines seem to have offered fragile residents of nursing homes with a weakened immune response almost the same level of protection that was reported for younger, healthy populations. Remarkable if true, or difficult to believe.

The risk ratio of symptomatic infection, which I questioned, is the product of two risk ratios: the risk ratio of infection (0.19) times the risk ratio of symptoms if infected (0.52).

The first estimate is unquestionably implausible. Upper respiratory infection is primarily prevented by secretory IgA antibodies on ...