On March 29, 2021, the Centers for Disease Control and Prevention released a study of 3,950 health care workers, first responders and other essential workers in six states. The press release carried the finding near the top: two or more weeks after the second dose, the risk of infection fell by 90 percent. Director Rochelle Walensky’s statement followed it. “This study shows that our national vaccination efforts are working.”
considered excluded at-risk person-time because immunity was considered to be indeterminate
The study itself, published in the Morbidity and Mortality Weekly Report, contains a passage the release leaves out. The 13 days after vaccination, the authors wrote, were “considered excluded at-risk person-time because immunity was considered to be indeterminate.” Inside that window, across 67,483 person-days of follow-up, 33 PCR-confirmed infections turned up, and the report records that they were “identified and excluded from the outcome.” The press release, which led with the 90 percent and explained the two-week rule to the public, does not mention them.
Nothing in the record suggests the exclusion was concealed. It sits in the results section, in plain view of anyone who read past the headline. The question the documents raise is what that rule did to the number.
The clock only one group wears
Under the report’s definitions, a participant counted as “fully immunized” only from 14 days after the second dose and “partially immunized” from 14 days after the first dose until the second, which produced effectiveness estimates of 90 percent and 80 percent respectively. The days right after the first needle belonged to neither box. Infections that landed there were not scored as vaccine failures, and they were not scored as anything else either.
The agency applied the same logic to its national surveillance. Its May 2021 breakthrough report defined a vaccine breakthrough as a positive specimen collected 14 days or more after receipt of all recommended doses. Take a hypothetical nurse who gets her second shot on a Monday and tests positive that Saturday. Under that definition hers is not a breakthrough, so it never reaches the breakthrough ledger at all.
A useless vaccine, 48 percent effective
In February 2024 the Journal of Evaluation in Clinical Practice published a paper by Kaiser Fung, Mark Jones and Peter Doshi, “Sources of bias in observational studies of covid-19 vaccine effectiveness.” They call the problem case-counting window bias. In the randomized trials, both arms received injections, so the waiting period before cases counted applied to vaccine and placebo groups alike and washed out. Observational studies have no placebo arm. The unvaccinated never get a shot and never start a clock, so their infections count from day one, while the vaccinated group’s earliest infections are set aside.
The authors built a worked example on the timing of Pfizer’s trial, using a vaccine that does nothing at all: the same infections in both arms, and no benefit when the data are analyzed evenly. Drop the cases from the first 36 days in the vaccinated arm only, and the inert product registers as 48 percent effective. The paper does not claim the vaccines did nothing. What it shows is how far the arithmetic alone can move an effectiveness estimate with no biology behind the movement, and the CDC’s March 2021 study, which set aside vaccinated person-time while counting unvaccinated person-time from enrollment, was built on that same structure.
The window was not empty, either. Researchers at Italy’s national health institute, the Istituto Superiore di Sanità, followed over 7 million vaccine recipients from 27 December 2020 to mid-April 2021 and published the results in Eurosurveillance. They used the first 14 days after the first dose as their reference period, and against that yardstick infection risk fell gradually until days 42 to 49, where it reached a 78 percent reduction. When they reran the analysis using only the first 7 days as the reference, the risk of infection in days 7 to 14 came out higher. The authors offered two explanations: that “vaccinated individuals may be less likely to respect social-distancing rules in the immediate days after vaccination,” and that immunity is incomplete during that stretch.
| Measure | Share |
|---|---|
| of 10,262 reported breakthrough infections | 27 percent |
The Italian figures compare vaccinated people with themselves at different points in time, not with the unvaccinated, so they cannot say whether that second-week rise exceeded the background rate. They do establish, in a cohort of millions, that infection risk climbed in the week after the first shot, which falls inside the window the American study classified as indeterminate and set aside. Maryanne Demasi, whose MD Reports brought the pattern back into view this month, argues that “Covid infections spiked in the days after vaccination” and that the CDC, having excluded those early cases from its effectiveness math, told people their post-shot Covid “was just a coincidence.”
The ledger after May 1
The second narrowing is dated. Through April 30, 2021, the CDC had received reports of 10,262 breakthrough infections from 46 states and territories, of which 27 percent were asymptomatic; 995 patients were hospitalized and 160 died. The agency’s own report described that count as “likely a substantial undercount” because surveillance relied on passive, voluntary reporting and many people with mild or no symptoms never sought a test. In the same document it stated: “Beginning May 1, 2021, CDC transitioned from monitoring all reported COVID-19 vaccine breakthrough infections,” and from then on it investigated only cases in patients who were hospitalized or died. Having called its count too low, the agency responded by counting less.
Put the two decisions side by side. Infections in the first two weeks after a shot sat outside the effectiveness study by design, and after May 1 mild and asymptomatic infections in the fully vaccinated sat outside the CDC’s breakthrough monitoring by policy. The agency documented both decisions, and each one shrank what could be tallied against the vaccine.
The documents do not show that the shots caused the infections that followed them, and they do not show what the 90 percent would have been with the excluded person-time counted back in. They show an effectiveness figure announced with a director’s endorsement and no mention of the 33, a peer-reviewed demonstration that the same counting structure can turn nothing into 48 percent, and a surveillance system cut back weeks after the agency called its own count an undercount.
Outside researchers have the published tables and a worked example. The CDC has the participant-level dose dates and test dates from the March 2021 study, along with the state reports behind the mild breakthrough cases it stopped investigating after April 30. Nobody outside the agency can rerun the 90 percent with those 13 days counted back in. The person-time, and the decision to release it, sit with the CDC.
Sources
- CDC MMWR, Thompson et al., interim mRNA vaccine effectiveness among frontline workers (April 2, 2021)
- CDC press release, “CDC Real-World Study Confirms Protective Benefits of mRNA COVID-19 Vaccines” (March 29, 2021)
- Fung, Jones & Doshi, “Sources of bias in observational studies of covid-19 vaccine effectiveness,” Journal of Evaluation in Clinical Practice 2024;30(1):30-36
- CDC MMWR, COVID-19 vaccine breakthrough infections reported to CDC (vol. 70, no. 21)
- Mateo-Urdiales et al., risk of SARS-CoV-2 infection at intervals since first dose, Italy, Eurosurveillance 2021;26(25)
- Maryanne Demasi, MD Reports, “Caught Covid soon after your jab? You were not alone”