The one-shot lobby has received a booster shot from the U.K. government, which declared that half a dose of the coronavirus vaccine is sufficient to protect its citizens (link). (I deliberately wrote **half a dose** because in the AstraZeneca/Oxford trial, it is claimed that the participants who received an accidental half dose in the first shot unexpectedly attained 90% efficacy, much higher than those who received the full dose.) As a result, there is renewed pressure on the FDA to approve a single-dose vaccine that has not been tested. On Twitter, I compared this request to **students who hand in half the homework and demand an A grade anyway**.

Having anticipated this controversy, I previously provided the basic scientific reasons why no statistician is likely to endorse such a policy (link).

Let me now present a list of questions to those who support the one-shot program. These questions address some **assumptions** they must have made in order to justify single doses.

- What is the
**primary goal**of your vaccination policy? Is it to minimize infections, cases, hospitalizations, ICU usage or deaths? Is it to maximize the number of vaccine doses administered? To achieve "herd immunity"? (Pick one)

- Do you agree that any
**extrapolation**of the clinical trial results (Pfizer/Biontech, Moderna, AstraZeneca/Oxford) to a single-dose treatment regime requires assumptions that are not investigated in these trials? (Yes or No)

- What is your expectation of the single dose's vaccine efficacy in the first six months? Draw your line on the following chart (which shows Pfizer results).

- In estimating the VE curve in question 3, how are you using the outcomes from the three different trials? If you selected the values from a single trial, explain your choice. If you averaged the results, describe your methodology regarding aligning test populations, dosing schedules, and case definitions.

- In estimating the VE curve of a single dose in question 3, you must have subtracted an expected impact of the second dose for the period of time after the second dose was administered during the vaccine trials. What is your assumption of the second dose's effect on VE? (If you did not adjust the VE curve from the trial, you should answer 0.)

- Are you assuming a
**withering**of vaccine efficacy over time? (Yes or no) If yes, when do you expect single-dose VE to start declining? And what is the rate of decline?

- What proportion of Covid-19 cases do you assume to be due to
**asymptomatic**spread? Do you assume the vaccines to have any effect on reducing asymptomatic spread? If yes, what is the basis for your assumption?

- In your mathematical model of the spread of the novel coronavirus, which of the following scenarios result in fewer infections (substitute the goal you expressed in question 1):

a) half the population immunized with a vaccine with 90% efficacy

b) the entire population immunized with a vaccine with 50% efficacy

- In your model of disease spread, what is your assumption of mask-wearing and social distancing after vaccination? What level of compliance are you assuming?

***

It's not that it's never permissible to extrapolate experimental findings. But any extrapolation is built upon assumptions. We can't judge the conclusion unless we know what those assumptions are.

When students nag their professors for an A despite not completing all the course requirements, some kind professors will request additional work to justify the grade change. That's what the FDA should be requesting from the one-dose lobby.

## Comments

You can follow this conversation by subscribing to the comment feed for this post.