Effectiveness of mRNA COVID-19 vaccine booster doses against Omicron severe outcomes

The study addresses how effective monovalent mRNA COVID-19 booster doses (third and fourth doses) are at preventing severe outcomes (hospitalization or death) due to Omicron among community-dwelling adults aged 50 years and older in Ontario, Canada. Contextually, Ontario expanded booster eligibility in late 2021 and began offering fourth doses to older adults in 2022 in response to waning immunity and the emergence of Omicron, a highly transmissible and immune-evasive variant. With subsequent circulation of sublineages BA.1/BA.2 followed by BA.4/BA.5 and the later introduction of bivalent vaccines, quantifying real-world vaccine effectiveness stratified by age, dose number, time since vaccination, and sublineage predominance is critical for informing booster policies and public health measures.

The discussion compares findings with other jurisdictions. Fourth-dose VE estimates in this Ontario study are slightly higher than those in the United States (e.g., ~80% against hospitalization ≥7 days post-dose among adults ≥50 years). Israeli studies indicated slower waning against severe outcomes than infection, and faster waning of marginal effectiveness against infection after fourth versus third doses; limited follow-up constrained severe-outcome assessments. Evidence on VE against BA.4/BA.5 is mixed: some UK and South African data suggest similar VE relative to BA.1/BA.2, whereas studies from Portugal and Kaiser Permanente observed lower VE against BA.4/BA.5. The IVY Network reported lower 3-dose VE during BA.4/BA.5 compared to BA.1/BA.2 soon after vaccination. Potential reasons for lower VE during BA.4/BA.5 include longer intervals since boosting, increased undocumented prior infections, and greater immune evasion by BA.5. Early immunogenicity results on bivalent boosters show higher neutralizing titres against Omicron sublineages for Moderna in a trial, but two observational studies found similar neutralization for Pfizer and Moderna bivalents compared with monovalent ancestral vaccines.

Design: Test-negative design leveraging linked provincial laboratory, vaccination, and health administrative datasets at ICES. Population: Community-dwelling adults aged ≥50 years with at least one RT-PCR test for SARS-CoV-2 between January 2 and October 1, 2022, in Ontario. Exclusions: Immunocompromised individuals (n=1,154); those receiving a bivalent mRNA vaccine (n=2,433); recipients of Ad26.COV2.S or BNT162b2 under specified criteria (n=10,843 as listed); Delta cases identified by WGS or S-gene target positive before Jan 24, 2022 (n=72). Approximately 95% received mRNA-1273 or BNT162b2 for all doses. Exposure classification: Number of doses (2, 3, 4) and time since most recent dose, evaluated up to ≥300 days (2 doses), ≥240 days (3 doses), and ≥120 days (4 doses). For mRNA-1273 boosters, 50 mcg for <70 years and 100 mcg for ≥70 years. Outcome: COVID-19-associated hospitalization or death due to or partially due to COVID-19; excluded nosocomial infections (infection <3 days after admission and flagged nosocomial). Sampling: Cases and controls sampled by week of test; individuals could contribute repeatedly as controls but once as a case. Controls were symptomatic test-negative individuals; index date defined by earliest of specimen collection, hospitalization, or death. Covariates: Sex, public health unit region, area-level SES variables (income, essential worker, persons per dwelling, visible minority quintiles), prior influenza vaccination (2019–2021 seasons), prior SARS-CoV-2 infection >90 days, number of SARS-CoV-2 tests in 3 months prior to Dec 14, 2020 (proxy for healthcare workers), comorbidities, home care services, and week of test. Analysis: Age-stratified (50–59, 60–69, 70–79, ≥80 years) multivariable logistic regression to estimate adjusted odds ratios of vaccination among cases versus test-negative controls; GEE with exchangeable correlation used to account for repeated controls (13% of controls). VE and marginal effectiveness computed as (1 − adjusted OR) × 100. Sublineage analysis: Interaction for period of predominance (BA.1/BA.2: Jan 2–Jul 2, 2022; BA.4/BA.5: Jul 3–Oct 1, 2022). GEE not used in period interaction models due to convergence; non-GEE and GEE estimates were similar where both converged. Sensitivity analysis excluded those prescribed Paxlovid within 14 days prior to index date (n=177). Statistical software: SAS v9.1; two-sided tests with α=0.05; standardized differences ≥0.1 considered clinically relevant.

• Sample: The study period included 11,160 COVID-19-associated severe outcomes and 62,850 symptomatic test-negative controls (among 53,369 individuals). The abstract also notes inclusion of 11,060 cases and 62,880 tested unvaccinated adults.
• Third dose VE: 91–98% within 7–59 days post–third dose; waned to 76–87% by ≥240 days post–third dose.
• Fourth dose VE: Restored to 92–97% within 7–59 days post–fourth dose; waned to about 86–89% by ≥120 days.
• Age gradient: VE generally lower with increasing age; among 70–79 years, VE decreased from 84% (95% CI, 57–94) to 71% (95% CI, 63–78) after >300 days; examples in figures show strong initial VE after boosters with gradual waning across all age groups.
• Sublineage periods: VE was lower and declined faster during BA.4/BA.5 predominance compared with BA.1/BA.2, with differences widening as time since vaccination increased.
• Marginal effectiveness: Third and fourth doses provided significant marginal effectiveness compared with unvaccinated individuals, with restoration after a fourth dose and waning evident by ≥120 days post–fourth dose.

Booster doses of monovalent mRNA vaccines restore strong protection against Omicron-associated severe outcomes among adults aged ≥50 years, but protection wanes over time after each booster. Third doses maintained high protection (approximately mid-80% range) 8 months post-vaccination among those aged 50–69 years, with lower protection among those ≥70 years. Fourth doses counteracted waning from third doses and provided renewed strong protection for several months, especially in older adults. However, VE was consistently lower and waned faster during the BA.4/BA.5 period versus BA.1/BA.2, likely reflecting a combination of longer intervals since boosting, potential bias from undocumented prior infections, and increased immune evasion by BA.4/BA.5 (with immune evasion considered the most plausible main contributor). Comparisons with other countries show broadly similar trends but some heterogeneity by setting, design, and timing. The findings inform booster policy, suggesting continued need for timely boosters and complementary public health measures as newer sublineages emerge.

Booster doses (third and fourth) of monovalent mRNA COVID-19 vaccines initially restore strong protection against community-associated hospitalization and death among community-dwelling adults aged ≥50 years in Ontario. This protection remains strong for at least 3–4 months after a fourth dose but subsequently wanes, with more pronounced declines during BA.4/BA.5 predominance. Given reduced VE against BA.4/BA.5 and the emergence of new sublineages (e.g., BQ.1.1, XBB), ongoing boosters and non-pharmaceutical interventions (masking, ventilation, indoor air filtration) may be required. Continued VE monitoring is important, particularly with the introduction of bivalent vaccines and evolving variants.

• Lack of data on rapid antigen tests (primary testing modality after Dec 31, 2021) limited adjustment for prior infections confirmed only by antigen testing, potentially biasing VE up or down depending on differential prevalence of undocumented prior infection by vaccination status.
• Incomplete whole genome sequencing prevented estimation of VE separately for BA.1, BA.2, BA.4, and BA.5; periods of predominance were used instead.
• Potential residual confounding due to unmeasured differences between vaccinated and unvaccinated individuals despite adjustment for available covariates. Strengths include long follow-up enabling VE estimation ≥4 months after fourth doses and age-stratified analyses providing refined VE estimates for decision-making.