BNT162b2-induced protection against infection builds rapidly after the first dose, peaks in the first month after the second dose, and then gradually wanes in subsequent months. The waning appears to accelerate after the fourth month, to reach a low level of approximately 20% in subsequent months. Although the protection against asymptomatic infection diminished more quickly than that against symptomatic infection, as would be expected in a vaccine that prevents symptoms given infection,31,32 no evidence was found for an appreciable waning of protection against hospitalization and death, which remained robust — generally at 90% or higher — for 6 months after the second dose. Implications of these findings on infection transmission remain to be clarified, but vaccine breakthrough infections were found recently, in this same population, to be less infectious than primary infections in unvaccinated persons.33
Because the immunization campaign prioritized vaccination of persons with severe or multiple chronic conditions and prioritized vaccination according to age group, this pattern of waning of protection could theoretically be confounded by effects of age and coexisting conditions. However, this possibility was not supported by our results, because a similar pattern of waning of protection was observed for all ages. Old age may (partially) serve as a proxy for coexisting conditions, and the number of persons with severe or multiple chronic conditions is small among the young, working-age population of Qatar.17,28 The national list of vaccine prioritization included only 19,800 persons of all age groups with serious coexisting conditions to be prioritized in the first phase of vaccine rollout.
Infection incidence was driven by different variants over time; thus, it is possible that waning of protection could be confounded by exposure to different variants at different time points. However, this seems unlikely. By far the dominant variant during the study was B.1.351,2,4,8-10 and a similar pattern of waning of protection was observed for B.1.1.7, B.1.351, and B.1.617.2.
Vaccinated persons presumably have a higher rate of social contact than unvaccinated persons and may also have lower adherence to safety measures.34-36 This behavior could reduce real-world effectiveness of the vaccine as compared with its biologic effectiveness, possibly explaining the waning of protection. Public health restrictions have been easing gradually in Qatar but differently for vaccinated and unvaccinated persons. Many social, work, and travel activities now require evidence of vaccination (a “health pass”) that is administered through a mandatory mobile app (the Ehteraz app). Risk compensation may be even higher with increasing time since receipt of the second dose — that is, there could be a progressive normalization of behavior.35-37 However, risk compensation is perhaps more likely to affect the overall level of estimated effectiveness than the observed rapid waning of protection over time, unless such risk compensation increases rapidly with time after the second dose.
PCR testing in Qatar is done on a mass scale, with approximately 5% of the population being tested every week.5 Approximately 75% of those who receive a diagnosis of SARS-CoV-2 infection at present do so not because of the appearance of symptoms but because of routine testing. It is possible that many asymptomatic infections were diagnosed among vaccinated participants that otherwise would have been missed. The higher ascertainment of infection may have lowered the effectiveness estimates. This idea is supported by the observed lower effectiveness against asymptomatic infection.
Emerging evidence supports the findings of this study. An increasing number of studies suggest substantial waning of BNT162b2 effectiveness.38-42 The findings are also supported by recent reports from Israel and the United States that indicate declining BNT162b2 effectiveness against infection with elapsed time and according to calendar month.42-46 Our findings, along with the greater immunogenicity of a schedule with a longer dose interval,47 may also explain the observed low effectiveness against B.1.617.2 in countries where the second dose was implemented 3 weeks after the first dose, such as in Israel,43 Qatar,30 and the United States,46 where B.1.617.2 has been dominant at a time when a nonnegligible proportion of the population had their second dose in January or February of 2021. However, higher effectiveness against B.1.617.2 has been observed in countries where a delayed interval schedule has been implemented, such as in Canada15 and the United Kingdom,13,14 where B.1.617.2 became dominant at a time when a negligible proportion of the population had their second dose in January or February of 2021.
This study has limitations. Individual-level data on coexisting conditions were not available; therefore, they could not be explicitly factored into our analysis. However, adjusting for age may have served, in part, as a proxy. With the young population of Qatar,17,28 only a small proportion of the study population may have had serious coexisting conditions. Only 9% of the population are 50 years of age or older,17,28 and 60% are young, expatriate craft and manual workers involved in mega-development projects.18,19,48 Our findings may not be generalizable to other countries where elderly persons constitute a sizable proportion of the total population.
Effectiveness was assessed with the use of an observational, test-negative, case–control study design,11,12 rather than a randomized, clinical trial design, in which cohorts of vaccinated and unvaccinated persons were followed. We were unable to use a cohort study design owing to depletion of the unvaccinated cohorts by the high vaccine coverage. However, the cohort study design that was applied earlier to the same population of Qatar yielded findings similar to those reported for the test-negative, case–control design,2,4 which supports the validity of this standard approach in assessing vaccine effectiveness for respiratory tract infections.2,4,11-15 The results of this study are also consistent with our previous estimates of vaccine effectiveness immediately after the first and second doses.2,29 We note that the earlier estimates involved (mostly) symptomatic infections with low PCR cycle threshold values, whereas the present study estimates involve (mostly) asymptomatic infections of both high and low PCR cycle threshold values.
Nonetheless, one cannot rule out the possibility that in real-world data, bias could arise in unexpected ways or from unknown sources, such as subtle differences in test-seeking behavior or changes in the pattern of testing with the introduction of other testing approaches, such as rapid antigen testing. For example, inclusion of PCR testing before travel or at port of entry was found to introduce a negative bias — that is, lowering the effectiveness estimates (Table S10) — perhaps because of different test-seeking behaviors of those vaccinated as compared with those unvaccinated, as a consequence of the travel privileges granted only to vaccinated persons.49
Vaccine effectiveness for participants at 0 to 13 days after the first dose was just below zero, possibly suggesting a negative bias. However, this has also been observed elsewhere for both Covid-19 vaccines50-52 and other vaccines.53 This effect may reflect differences in social behavior at or after vaccination or an immunologic effect.53
Notwithstanding these limitations, consistent findings of this study were reached that indicated a large effect size for the waning of vaccine protection over time, regardless of the reason for PCR testing and whether there were symptoms. Moreover, with the mass scale of PCR testing in Qatar,5 the likelihood of bias is perhaps minimized. Indeed, the different sensitivity and additional analyses that were conducted to investigate effects of potential bias, such as by modifying the inclusion and exclusion criteria, all yielded findings that indicated a rapid waning of vaccine protection.
In this study, we found that BNT162b2-induced protection against infection peaked in the first month after the second dose and then gradually waned month by month, before reaching low levels 5 to 7 months after the second dose. Meanwhile, BNT162b2-induced protection against hospitalization and death persisted with hardly any waning for 6 months after the second dose. These findings suggest that a large proportion of the vaccinated population could lose its protection against infection in the coming months, perhaps increasing the potential for new epidemic waves.