The SARS-CoV-2 omicron (B.1.1.529) variant was originally identified in Botswana and South Africa in November, 2021. Compared to the original SARS-CoV-2 strain (Wuhan-Hu-1), omicron (including BA lineages of B.1.1.529) has more than 30 mutations in its spike (S) protein. Many of these mutations are in the receptor-binding domain and have been shown to correspond with a high degree of transmissibility and neutralising antibody escape.

Subsequently, omicron was declared a variant of concern by WHO on Nov 26, 2021, and has since rapidly become the dominant SARS-CoV-2 variant globally.

Preliminary studies have shown that two doses of the BNT162b2 (Pfizer–BioNTech) mRNA vaccine do not sufficiently neutralise the omicron variant. Additionally, although levels of neutralisation against omicron after three doses of BNT162b2 are substantially higher than those seen after two doses, they are lower than those observed for the wild-type strain and for the delta (B.1.617.2) variant.

^, 

^, 

^, 

^, 

However, early data suggest that T-cell responses induced by previous SARS-CoV-2 infection and by the BNT162b2 vaccine are not significantly altered against omicron.

^, 

^, 

^, 

^, 

Although these findings suggest that protection against severe SARS-CoV-2 infection with omicron is likely to be better preserved than that against less symptomatic infection in most individuals, the extent of the impact on vaccine effectiveness and potential waning is unclear.

Estimates of the real-world effectiveness of BNT162b2 (and other COVID-19 vaccines) against infection caused by the omicron variant are limited to data from South Africa

and the US Centers for Disease Control and Prevention (CDC),

^, 

^, 

and preprint reports from the UK,

the USA,

Canada,

Qatar,

and Denmark.

Generally speaking, these data suggest that two doses of the mRNA vaccine are likely to provide only limited and short-lived protection against infection caused by the omicron variant and symptomatic COVID-19,

^, 

^, 

^, 

^, 

^, 

^, 

^, 

^, 

and two doses have reduced effectiveness against hospital admission due to omicron compared to that seen for previous SARS-CoV-2 variants.

^, 

^, 

Additionally, preliminary data suggest that three doses of the mRNA vaccine provide increased but modest protection against SARS-CoV-2 infection

^, 

^, 

^, 

that wanes quickly,

as well as increased protection against hospital admission in the first few months.

^, 

The effectiveness of a third (booster) dose of an mRNA vaccine against infection with the omicron variant or symptomatic COVID-19 has been estimated to be higher than that achieved after two doses,

^, 

^, 

^, 

but with waning observed in the first few months.

^, 

Only two published reports from the CDC

^, 

and one non-peer-reviewed report

have estimated the effectiveness of a third dose against hospital admission due to omicron. All studies reported vaccine effectiveness of roughly 90% soon after receipt of the third dose of an mRNA vaccine; however, none reported vaccine-specific estimates

^, 

^, 

and only one provided long-term follow-up after receipt of a booster dose.

Thus, studies evaluating the effectiveness of BNT162b2 against omicron in other robust health systems across the world, especially for severe outcomes and with sufficient follow-up, are urgently needed to inform public-health decision making about the need for (and timing of) additional boosters, improved COVID-19 vaccines tailored to target the omicron variant, or other effective COVID-19 vaccines. We aimed to evaluate the age-specific effectiveness and durability of two and three doses of BNT162b2 against hospital and emergency department admissions due to the omicron variant in a large integrated health system in the USA.

Between Dec 1, 2021, and Feb 6, 2022, there were 16 063 hospital admissions and 19 699 emergency department admissions across the KPSC health-care system with a documented SARS-CoV-2 PCR test. 3716 (23%) of 16 063 hospital admissions and 9367 (48%) of 19 699 emergency department admissions were for acute respiratory infection (figure 1). Of these, the final study population consisted of 11 123 encounters where patients were either SARS-CoV-2 negative, tested via the ThermoFisher TaqPath COVID-19 Combo Kit (and thus SGTF could be determined), or were assigned a variant lineage based on the time when the infection occurred (ie, when ≥95% of all infections were due to the delta or omicron variant). Among the final study population, 7361 (66%) of 11 123 patients with hospital encounters had a positive SARS-CoV-2 test, with 4523 (61%) of 7361 identified as omicron, and 2838 (39%) of 7361 identified as delta. Overall, 5165 (46%) of 11 123 patients were unvaccinated, 3981 (36%) of 11 123 were vaccinated with two doses only, and 1977 (18%) of 11 123 were vaccinated with three doses. The median age of the study population was 51 years (IQR 35–69). Compared to those who tested positive for SARS-CoV-2, those who tested negative tended to be older and White and more likely to have comorbidities and evidence of previous SARS-CoV-2 infection (table 1). Compared to patients who received three doses, unvaccinated patients were younger, more likely to be Black or Hispanic, less likely to have comorbidities, and less likely to have previously received influenza or pneumococcal vaccination (table 2).

Among the study population, 3275 (29%) of 11 123 patients were admitted to the hospital and 7848 (71%) of 11 123 were admitted to the emergency department (without a subsequent hospital admission). Of those admitted to the hospital with a positive SARS-CoV-2 test, 647 (33%) of 1949 were infected with the delta variant and 1302 (67%) of 1949 were infected with the omicron variant (figure 1). Of those admitted to the emergency department only, 2191 (40%) of 5412 were infected with the delta variant and 3221 (60%) of 5412 were infected with the omicron variant.

In adjusted vaccine effectiveness analyses, three doses of BNT162b2 restored immunity against hospital admission and emergency department admission (without a subsequent hospital admission) due to the delta variant to the level seen 3 months after the second dose, which waned 6 months after receipt of only two doses (figure 2; appendix p 5). Vaccine effectiveness against hospital admission due to the delta variant after two doses remained relatively consistent, at 84% (95% CI 74–90) less than 6 months after two doses and 73% (63–80) at least 6 months after two doses. The effectiveness of three doses against hospital admission due to the delta variant appeared to fall from 89% (95% CI 83–93) less than 3 months after three doses to 71% (40–86) from 3 months onwards, although the confidence intervals overlapped and were wide around the latter estimate. Against emergency department admission due to the delta variant, the effectiveness of two doses of BNT162b2 declined from 78% (95% CI 69–85) less than 3 months after the second dose to 57% (45–66) from 9 months onwards. Vaccine effectiveness against emergency department admission due to the delta variant was 84% (95% CI 80–87) less than 3 months after the third dose. Overall vaccine effectiveness after a third dose was 87% (95% CI 81–92) against hospital admission due to the delta variant and 83% (79–86) against emergency department admission due to the delta variant (appendix p 5).

Immunity conferred by two or three doses of BNT162b2 was lower for the omicron variant than for the delta variant (figure 2; appendix p 5). The effectiveness of two doses against hospital admission due to the omicron variant was 68% (95% CI 48–80) at less than 3 months following a second dose and 41% (21–55) at 9 months onwards. The effectiveness of three doses against hospital admission due to the omicron variant was 85% (95% CI 80–89) at less than 3 months following the third dose but fell to 55% (28–71) from 3 months onwards, although the confidence intervals were wide for this latter estimate. Protection against emergency department admission due to the omicron variant also appeared to wane after two doses, falling from 64% (95% CI 51–73) at less than 3 months after the second dose to 51% (43–59) at 6–9 months after the second dose and 31% (16–43) at 9 months onwards. The effectiveness of three doses against emergency department admission due to the omicron variant also waned, declining from 77% (95% CI 72–81) at less than 3 months following a third dose to 53% (36–66) at 3 months onwards (appendix p 5).

Overall, stratification of vaccine effectiveness by age group showed similar trends as with all ages, but revealed slightly higher vaccine effectiveness estimates among individuals aged 65 years and older versus those aged 18–64 years (appendix pp 6–8).

Our estimates of the effectiveness of three doses against hospital admission due to the omicron variant less than 3 months after the third dose (85% [95% CI 80–89]) were generally similar to those seen for hospital admission due to the delta variant (89% [83–93]) in the first 3 months following the third dose. Our vaccine effectiveness estimates against hospital admission due to the omicron variant less than 3 months after a third dose were also similar to those reported by the CDC (which reported effectiveness of around 90% for both mRNA vaccines)

^, 

and the UK Heath Security Agency (which reported effectiveness of approximately 85–90% for all vaccines used in the UK combined).

Our study, to the best of our knowledge, is one of the first to report long-term follow-up outcomes after receipt of a booster dose, stratified by age. Similar to a recent CDC report showing waning effectiveness of mRNA COVID-19 vaccines after a third dose,

we also detected early signs of waning against severe outcomes caused by the omicron variant from 3 months onwards after the receipt of a booster dose. More long-term follow-up data are needed, however, to fully understand the true magnitude of this decline.

Although waning was observed, effectiveness of two and three doses against hospital admission was generally higher than that seen for the less severe endpoint of emergency department admission across all timepoints. Previous reports have indicated that T-cell immunity might have a role in the prevention of severe SARS-CoV-2 infection,

and laboratory results have suggested that T-cell epitopes have remained mostly unaltered for the omicron variant.

^, 

^, 

^, 

^, 

The relationship between cell-mediated immunity and protection against severe infection, however, needs more research. It is also likely that trained innate immunity has an important role in durable protection, particularly against severe infection. Unrelated live attenuated vaccines and adjuvanted vaccines can provide sustained protection against heterologous infections and severe outcomes, including COVID-19, possibly through induction of trained innate immunity, which mRNA vaccines also induce.

^, 

^, 

Our results showed waning effectiveness of two doses against emergency department admission for both the delta and omicron variants, consistent with previous evidence of waning effectiveness of BNT162b2 against the delta variant over time

and with preliminary studies showing waning effectiveness against the omicron variant and symptomatic SARS-CoV-2 infection.

^, 

^, 

^, 

After a third dose, effectiveness against emergency department admission increased to 84% (95% CI 80–87) for the delta variant and 77% (72–81) for the omicron variant. Waning of immunity against emergency department admission after three doses was moderate for the delta variant but was more pronounced for the omicron variant, falling to 53% (36–66) at 3 months or more after the third dose. This finding is consistent with early reports from the UK Health Security Agency suggesting that even after three doses, vaccine effectiveness against symptomatic SARS-CoV-2 infection due to the omicron variant wanes rapidly.

^, 

Age-stratified analyses showed that effectiveness was generally higher in adults aged 65 years and older. However, this finding might reflect enhanced risk-mitigating behaviours, especially among vaccinated people in this age group, or the potential dilution of vaccine effectiveness among adults younger than 65 years.

There are two important implications of our findings. First, booster doses significantly improve protection against the omicron variant. Our data suggest that, in the first few months after their receipt, booster doses of BNT162b2 provide high levels of protection (around 80–90%) against hospital and emergency department admission due to the omicron variant—a level of protection not seen with only two doses. Thus, although booster doses were initially introduced to counteract waning immunity against infection during a period when the delta variant was predominant,

in the omicron era three doses appear to be necessary to achieve high levels of protection against severe COVID-19 outcomes. These findings are consistent with laboratory data reporting that neutralising antibody concentrations not only improved after a third dose (compared to after two doses) but also that the breadth of coverage against variants of concern seemed to increase.

Second, although the initial effectiveness of a booster dose of BNT162b2 against the omicron variant was high, this protection waned after 3 months—even against a severe outcome such as hospital admission. Although more data are needed about the precise magnitude of this effect, this finding suggests that additional doses of current, adapted, or novel COVID-19 vaccines, which are currently in development, might be necessary in the future. Thus far, studies have shown that heterologous boosting with adenoviral vector vaccines does not improve the cellular response or the breadth of protection after priming with mRNA;

^, 

^, 

^, 

^, 

^, 

^, 

however, continual evaluation of booster strategies will be needed in the future. Preliminary data from Israel have shown improved protection against SARS-CoV-2 infection with the omicron variant and severe COVID-19 in individuals who received a fourth dose of BNT162b2 compared to those who had received only three doses at least 4 months ago.

However, data describing the durability of additional booster doses (beyond the third dose) are needed.

Our study is not without limitations. As this was an observational study, there might be residual confounding driven by differences in the likelihood of exposure to SARS-CoV-2 between vaccinated and unvaccinated individuals. Although we controlled for key sociodemographic and clinical characteristics, we were not able to capture data on occupation or behaviours, and sample size limitations precluded inclusion of all potentially confounding variables in adjusted analyses. Moreover, although vaccinated individuals might be more likely to seek care or testing for SARS-CoV-2 (thus biasing vaccine effectiveness results downwards), the test-negative design of the study, particularly in the hospital setting where there is universal testing, helps mitigate against this bias caused by differences in health-care-seeking behaviour—including the propensity to test.

^, 

^, 

^, 

Furthermore, although we attempted to control for previous SARS-CoV-2 infection, we were not able to assess previous infections among those who do not seek testing at KPSC. If undocumented previous infection was more likely in unvaccinated individuals, for example, this could contribute to underestimation of vaccine effectiveness, which might be more likely in the omicron era. In addition to using a time window where the omicron variant accounted for at least 95% of all new incident cases, we identified omicron cases based on SARS-CoV-2 samples with SGTF on the Thermo Fisher TaqPath assay. This approach could have led to misclassification of some samples. Misclassification, however, occurred in less than 1% of samples tested on ThermoFisher in our small validation study that compared PCR results to those from whole-genome sequencing. It is possible that some emergency department or hospital admissions were “with COVID-19” rather than “for COVID-19.” This could lead to spurious findings of waning against severe outcomes if in fact vaccine breakthroughs were only incidental infections among patients admitted to hospital for something other than COVID-19. This could explain why we found higher effectiveness estimates in adults aged 65 years and older (who are likely to be admitted to hospital for and have more severe COVID-19) compared to those aged 18–64 years who might be more likely to have incidental positivity for the omicron variant. By restricting our analysis to patients with a diagnosis of an acute respiratory infection, we attempted to mitigate this potential bias. However, it remains possible (especially in the emergency department setting) that milder COVID-19 or incidental SARS-CoV-2 infection was included in our analysis, which is likely to bias vaccine effectiveness estimates downwards, especially for the omicron variant given its milder clinical presentation and higher levels of antibody escape. Moreover, it is possible that hospital admission practices changed during our study period. For example, if larger numbers of less severe patients were admitted during the omicron wave, as was reported by the CDC,

this could also bias vaccine effectiveness estimates for the omicron variant downwards. Finally, given the study period, effectiveness estimates 3 months or more after the booster dose are likely to represent findings for individuals who first received booster doses, and additional follow-up will be needed to understand the effectiveness of booster doses over time in the general population and at different stages of the pandemic. Due to limitations in sample size, we were unable to assess long-term durability of the third dose or to provide vaccine effectiveness estimates specific to age groups or high-risk groups. Furthermore, due to the rapid influx of at-home antigen tests and increasing variation in clinic-based testing practices, we were not able to assess vaccine effectiveness against symptomatic or asymptomatic SARS-CoV-2 infection not requiring hospital or emergency department admission.