Introduction

The effectiveness of available Coronavirus Disease 2019 (COVID-19) vaccines may differ by variants of concern (VOCs) and by waning immunity. Before the emergence of the Omicron VOC, real-world vaccine effectiveness (VE) studies had reported substantial protection against symptomatic infection and severe outcomes (i.e., hospitalization and death) [1–3]. However, the protection offered by COVID-19 vaccines has been shown to wane over time [4,5], prompting many countries to provide booster doses [6]. With Omicron’s emergence and rapid spread, the booster program was expedited and expanded in several countries. A few studies have evaluated the protection offered after the booster, although with conflicting results [7–9].

The available, still limited, body of evidence indicates a rapid waning of protection against symptomatic infection offered by an mRNA booster after a homologous primary series [7,8,10,11]. Findings are more conflicting concerning the potential waning of protection against severe outcomes. After three doses of mRNA vaccine, sustained effectiveness against hospital or intensive care unit (ICU) admission was reported in different studies, including young and elderly individuals within 2 months after booster [8,9,12]. At the same time, another study has demonstrated a significant waning of protection against emergency department visits and hospitalizations 4 months after the third dose [7]. These studies need more data about the medium- and longer-term effectiveness of heterologous schemes and provide limited insights about severe events across age groups. Data on the duration of protection against severe outcomes in boosted individuals by age group are crucial to guide health policies about vaccination programs.

Brazil and Scotland present similarities in vaccination programs (vaccine type used for primary series and booster) and the speed of Omicron spread. Both countries have been offering BNT16b2 or ChAdOx1 as a primary series to all adults and an mRNA booster dose, i.e., BNT16b2 in Brazil and either BNT16b2 or mRNA-1273 in Scotland. Additionally, these countries have reported a rapid surge and dominance of the Omicron variant [13]. The similarities in vaccine administration between Brazil and Scotland, coupled with essential differences in several potential confounders (such as age structure, the timing of delivery to different age groups, and healthcare characteristics such as access to hospitals), offer an opportunity to undertake robust national analyses on the duration of VE during the Omicron era. We aimed to assess the extent and duration of protection against Omicron-associated symptomatic infection and severe outcomes (i.e., COVID-19 hospitalization and death) after an mRNA booster dose in individuals of different age groups who received either BNT16b2 or ChAdOx1 for their primary vaccination series.

Methods

Results

From January 01, 2022, to April 23, 2022, 5,832,210 individuals (Brazil: 5,276,385; Scotland: 555,825) were analyzed, and the median age and sex ratio were similar among cases and controls (Table 1 and Table A in S1 Appendix). Most of the tests performed in Brazil during the study period were positive (3,009,052; 57.0%), slightly elevated in unvaccinated individuals (203,964; 61.7%) and vaccinated (2,805,088; 56.7%). In Scotland, 352,015 (63.3%) tests were positive, more frequent in unvaccinated individuals (38,988; 74.6%) than in those vaccinated (313,027; 62.2%) (Table 1 and Tables A, C, and D in S1 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of individuals tested for SARS-CoV-2 in Brazil and Scotland.

https://doi.org/10.1371/journal.pmed.1004156.t001

Compared to unvaccinated individuals

In Brazil, a primary course of ChAdOx1 and a BNT16b2 booster (ChAdOx1-BNT162b2) dose led to an estimated VE against symptomatic infection of 51.6% (95% CI [51.0, 52.2], p < 0.001) after 14 to 29 days, waning to 4.2% (95% CI [0.7, 7.6], p = 0.02) at ≥4 months after the booster dose. The VE of a primary course of BNT162b2 and a BNT162b2 booster (BNT162b2-BNT162b2) was 44.6% (95% CI [43.4, 45.8], p < 0.001) at 14 to 29 days, waning to −11.8% (95% CI [−35.9, 8.0], p = 0.26) at ≥4 months past the booster dose. (Fig 1 and Table E in S1 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Estimated VE against symptomatic SARS-CoV-2 infection in Brazil and Scotland, overall and stratified by age group.

The square represents the estimated VE of the booster dose after the primary series with ChAdOx1, and the circle represents the booster dose after the primary series with BNT162b2. Green represents the booster dose with BNT162b2 and blue with mRNA-1273. Error bars represent the 95% Wald CI. CI, confidence interval; SARS-CoV-2, Severe Acute Respiratory Syndrome Coronavirus 2; VE, vaccine effectiveness.

https://doi.org/10.1371/journal.pmed.1004156.g001

The estimates of VE against severe outcomes of ChAdOx1-BNT162b2 peaked in Brazil at 93.5% (95% CI [93.0, 94.0], p < 0.001) at 14 to 29 days and 82.3% (95% CI [79.7, 84.7], p < 0.001) after ≥4 months. The VE of a BNT162b2-BNT162b2 was 92.7% (95% CI [91.0, 94.0], p < 0.001) at 14 to 29 days and 74.1% (95% CI [9.1, 92.6], p = 0.03) at ≥4 months (Fig 2 and Table F in S1 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Estimated VE against severe COVID-19 (hospitalization or death) in Brazil and Scotland, stratified by age groups.

The square represents the estimated VE of the booster dose after the primary series with ChAdOx1, and the circle represents the booster dose after the primary series with BNT162b2. Green represents the booster dose with BNT162b2 and blue with mRNA-1273. The error bar represents the 95% Wald CI. CI, confidence interval; COVID-19, Coronavirus Disease 2019; VE, vaccine effectiveness.

https://doi.org/10.1371/journal.pmed.1004156.g002

Scotland had a similar pattern for both booster vaccines (BNT162b2 or mRNA-1273). After the primary series of ChAdOx1, the VE peak at 98.3% (95% CI [87.3, 99.8], p < 0.001) 14 to 29 days past booster dose for ChAdOx1-BNT162b2 and 94.4% (95% CI [87.7, 97.5], p < 0.001) for ChAdOx1-mRNA-1273 in the same period, declining to 77.8% (95% CI [51.4, 89.9], p < 0.001) and 76.0% (95% CI [59.3, 85.9], p < 0.001) at ≥4 months, respectively. (Fig 2 and Table F in S1 Appendix).

Discussion

Following a homologous primary series of BNT16b2 or ChAdOx1 vaccine, mRNA boosters (BNT16b2 or mRNA-1273) provided substantial protection against severe COVID-19 cases during the predominance of Omicron variant for at least 3 months. However, there was only moderate protection against symptomatic infection at 14 to 29 days after the booster dose administration, which sharply decreased by ≥4 months.

Our results are comparable to previous observational studies during the Omicron period, which reported the waning effectiveness of mRNA boosters against symptomatic infection. Protection against infection is modest following the booster dose and quickly decreases 4 weeks post-booster [10,11]. Older individuals experienced a faster waning than younger people [4,17]. However, we found a more pronounced waning against symptomatic infection in the younger age groups in Brazil and a comparable waning in all age groups in Scotland.

The finding of temporary protection against symptomatic infection is consistent with neutralization data, suggesting the need for a third dose to elicit antibodies with neutralizing activity against Omicron and decaying titers over time [18,19]. Neutralizing antibody titers seem to increase again early after a fourth dose, but the duration remains unknown [20]. Observational studies evaluating the effect of vaccination on previously infected individuals reported that antibody levels peaked after three immune stimuli, either by vaccine or infection, without any significant increment after a fourth stimulus [21,22]. On the other hand, cellular immunity seems to stay robust against the Omicron variant after 3 months [23]. It plays a significant role in protecting against severe disease [24,25]. It likely will stay highly effective against variants of the SARS-CoV-2 virus due to the capacity of T cells to still recognize mutated epitopes from SARS-CoV-2 [23,25,26]. Consistent with data on cellular immunity, we observed a slight waning of protection against severe outcomes in the Omicron period. A fourth dose seems to increase the protection against severe illness but not against infection in individuals aged ≥60 years [27]. We found similar levels of protection against symptomatic infection in individuals with one and two booster doses. Together, these data suggest that while humoral response and VE against infection seem to be highly affected by the Omicron variant, a limited impact is observed on cellular immunity and protection against severe disease.

We analyzed the waning by age group due to the influence of immunosenescence in VE. We observe sustained protection against severe outcomes and a slight decline in protection in individuals aged ≥65 years, represented by the decrease in the last period in both comparisons: against unvaccinated individuals and individuals with only two doses of vaccine. Different age structures may have contributed to the apparently contradictory results observed in some studies that have addressed the duration of protection of mRNA booster dose against severe disease. For example, there was high and sustained protection (over 80% VE) at ≥7 weeks after booster and no evidence of waning in Qatar, a country with only 9% of the population ≥50 years old [8]. In contrast, a study analyzing data from 10 states of the United States of America found evidence of waning protection against hospitalization, with VE dropping from 91% at 2 months to 78% at 4 months of booster dose [7]. In addition to age, follow-up time seems essential in the analysis of waning. In a study performed in Finland only on individuals >70 years old, sustained protection was observed against admission to hospital and ICUs. Still, follow-up after booster was up to 2 months [9], which may at least partially explain the differences between their observed declines and what we found in VE in the elderly. Indeed, we observed the lowest VE against severity for the elderly past 4 months. Those previous studies provided no data about VE across age groups. Thus, differences in age groups and the follow-up time seem to be involved in the differences observed in VE after mRNA booster dose across the studies.

Assessment of vaccine waning from observational studies during the COVID-19 pandemic is methodologically challenging due to dynamic changes during the vaccination program. Challenges include prioritizing vaccine delivery to higher-risk groups, e.g., the elderly, individuals with comorbidities, and healthcare professionals, which led to more time after the booster dose than in other individuals. In addition, different intervals between the last dose in the primary vaccination series and the booster dose have been used in different places. All these factors are compounded by the different sublineages of the Omicron variant circulating during the follow-up period, which can introduce potential bias. However, conducting harmonized analyses in two countries at the national level mitigates the possibility of spurious results driven by unmeasured confounders.

Brazil and Scotland present several differences, such as variability in vaccination programs, circulating subvariants of Omicron, and population characteristics such as age structure, testing policy, and vaccination status. For example, up to April, Brazil had vaccinated less than 50% of the eligible population for booster doses, while in Scotland, this index was more than 85%. These aspects may have influenced the lower VE against infection in Brazil compared to Scotland. However, despite all dissimilarities between countries, a similar pattern of quickly VE waning against infection and durable protection against severe disease was observed, reinforcing the robustness of these findings.

To our knowledge, the present study is the largest to investigate the waning of mRNA booster doses against severe outcomes in the Omicron era. In both countries, similar results were obtained using different reference groups to assess vaccine protection: unvaccinated individuals in the main analysis and individuals with only two vaccine doses in the sensitivity analyses. We adjusted for several clinically relevant factors by deterministically linking various national clinical databases. Using a TND case–control study, we mitigated the risk of bias due to differences in health-seeking behavior between vaccinated and unvaccinated groups [14,28].

However, there are some limitations to note. First, we assumed all cases in the study period were associated with the Omicron variant. A few cases may have been due to other variants, including Delta. To mitigate this limitation, we restricted our analyses to the period when Omicron was dominant in both countries. Second, most of the tests performed in Brazil during the study period were lateral flow, which may induce bias in the results due to misclassifying cases as controls. Third, we were unable to discriminate between Omicron lineages in our study. Nevertheless, initial studies have suggested VE against Omicron subvariants to be similar [8,29,30]. Fourth, as in any observational study, residual confounding might exist. However, in previous studies, the adjustment for the chosen cofounders provided demonstrable control for bias. We cannot exclude that bias could arise from the unexpected effects of COVID-19 vaccines in other acute respiratory illnesses, protecting these individuals. Fifth, we have defined COVID-19-associated hospitalization as any admission episode that occurs within 14 days of a positive SARS-CoV-2 test or positive test within 72 hours of hospital admission. Consequently, a subset of hospitalizations may be composed of incidental cases. However, although possible in Scotland, such incidental cases are unlikely to occur in Brazil. The Brazilian hospitalization database used in the present study only includes individuals with severe acute respiratory syndrome symptoms tested for SARS-CoV-2 infection. Sixth, there are a disproportional number of individuals in our sample from Brazil’s first quintile of deprivation (low deprivation). It is likely due to the use of the municipality deprivation index, indicating that cities with lower deprivation maintain a broader testing policy than cities with higher deprivation. Seventh, asymptomatic and mild infections may have been underrecognized if previously infected individuals were not tested. In Brazil, we found negative VE against symptomatic infection in the younger group in the last period past booster. However, in the exploratory analysis using only individuals with previously confirmed SARS-CoV-2 infection, in this scenario, the result was not replicated, indicating a possible bias of under-ascertainment of previous infection in the younger individuals, with more unvaccinated individuals with previous undetected infection.

In summary, our study has shown that older individuals are at the highest risk of experiencing severe outcomes after infection with the Omicron variant, even after receiving a booster dose. The recommendation for a second mRNA booster dose, now being implemented for these groups in several countries, seems sensible for preventing severe forms of COVID-19. Our data indicate that infection prevention and, thus, community protection may not be a realistic target with currently available vaccines. The durability of protection against hospitalization and death remains an open question underscoring the need for studies with extended follow-up periods. New vaccines, mainly those aimed at interrupting transmission, possibly by enhancing mucosal immunity, are necessary to reduce the risk of Omicron infection and transmission.

Supporting information

Acknowledgments

This study is part of the Fiocruz VigiVac program, and the authors acknowledge DATASUS for its diligent work in providing the unidentified Brazilian databases. GLW, MLB, VSB, and MB-N are Brazilian National Research Council research fellows. SVK acknowledges funding from an NRS Senior Clinical Fellowship (SCAF/15/02), the Medical Research Council (MC_UU_00022/2), and the Scottish Government Chief Scientist Office (SPHSU17). ESP acknowledges funding from the Wellcome Trust (213589/Z/18/Z). T.C.-S. is a PhD student at the Post-Graduation Program in Health Sciences-UFBA, which is supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior-Brasil, finance code 001. We thank Dave Kelly from Albasoft (Inverness, UK) for his support with making primary care data available; Wendy Inglis-Humphrey, Vicky Hammersley, and Laura Brook (University of Edinburgh, Edinburgh, UK); and Pam McVeigh, Amanda Burridge, and Afshin Dastafshan (Public Health Scotland, Glasgow, UK) for their support with project management and administration.