NVX-CoV2372, monovalent mRNA and bivalent mRNA vaccines elicit broadly cross-reactive antibodies against emerging SARS-CoV-2 variants Authors: Xuan Ying Poh, Shirley Y. Y. Mah, Aileen Y. Y. Yeoh, Jean-Marc Chavatte, Daniel R. X. Lim, Cheryl C. Y. Tan, Hao Ling Tang, Mihir Gandhi, Nabilah Rahman, Suma Rao, Po Ying Chia, Sapna P. Sadarangani, Yi-Qing Chin, Lisa F. P. Ng, Laurent Renia, Raymond T. P. Lin, David C. Lye, Lin-Fa Wang, Chee Wah Tan, Barnaby E. Young Published: 2026 ABSTRACT ABSTRACTRandomized clinical trials comparing the breadth and long-term persistence of immunity between different COVID-19 vaccine types and between ancestral and Omicron-targeted vaccines are limited. The PRIBIVAC study (Phase D) is a randomized clinical trial comparing the immunogenicity of monovalent mRNA vs bivalent mRNA vs protein-based NVX-CoV2373 administered as second booster in 176 triple mRNA-vaccinated adults. Primary objective was neutralizing antibody levels against Omicron subvariants at day 28. A 4th vaccine dose significantly boosted 50% neutralization titers against emerging strain XBB.1.16 by 3.2-, 4.1- and 1.6-fold in monovalent mRNA, bivalent mRNA and NVX-CoV2373 group respectively at day 28. The largest absolute increase in inhibition level at day 28 post-booster was observed against the KP.2 subvariant, with bivalent mRNA vaccines exhibiting the highest neutralization level (91.7%) compared with monovalent mRNA (84.4%; p = .027) and NVX-CoV2373 (81.4%; p < .0001). While bivalent mRNA vaccines elicited the highest early immunogenicity, neutralization levels against all Omicron variants tested waned to similar levels between groups by 12 months post-vaccination. Although NVX-CoV2373 induced a lower peak anti-S antibody response, anti-S decay rate was slower in NVX-CoV2373 compared with mRNA vaccines. The geometric mean anti-S fold change (D360/D28) in NVX-CoV2373 group was higher (0.51) relative to both mRNA vaccines (monovalent: 0.31, p = .010 and bivalent: 0.35, p = .017). Improved neutralizing antibody responses against diverse SARS-CoV-2 variants by the ancestral or variant vaccine highlight the immunological benefits of COVID-19 vaccine boosters regardless of the latest variant-based vaccine. Further studies to determine if different vaccine combinations translate to differing protection against infection remain necessary. Introduction Following its emergence in November 2021, the SARS-CoV-2 Omicron variant displaced other lineages, and its subvariants continue to cause repeated COVID-19 surges worldwide. Reflecting SARS-CoV-2 adaption as a human pathogen, the Omicron subvariants are both less virulent and more transmissible than previous variants, while being adept at evolving to evade humoral immunity induced by prior vaccination or infection.1,2 BA.1 and BA.2 lineages predominated in early 2022, reaching a peak prevalence of 75.7% in February and 73.4% in April, respectively.3 By mid–late 2022, BA.5 first replaced BA.2 as the dominating lineage (peak of 86.2% in August 2022) followed by BQ.1 (22.1%) and BQ.1.1 (37.2%), both descendants of BA.5, by December. Reflecting the changing variant landscape, in August 2022 the US Food and Drug Administration (FDA) granted emergency use authorizations to bivalent formulations of the Moderna and Pfizer-BioNtech mRNA vaccines that target both the ancestral SARS-CoV-2 strain and the Omicron BA.4/5 lineages.4 In Singapore, the bivalent mRNA-1273 (ancestral/BA.1, Moderna) and bivalent BNT162b2 (ancestral/BA.4/5, Pfizer) vaccines were added to the National Vaccination Programme from October 2022 and were recommended in place of the original monovalent vaccines. The variant landscape during 2023–24 was characterized by XBB descendants, JN.1 and the rise of new recombinants. The XBB lineage (recombinant of two BA.2 descendant lineages) first emerged in late fall 2022 and its descendants dominated the landscape through much of 2023.5 XBB.1.5 attained predominance (>50% prevalence) between Feb–Jun 2023 (61.5%), and the prevalence of other commonly circulating XBB lineage groups peaked at 21.2% for XBB.1.16-like lineages by July and 26.2% for EG.5-like lineages by September.3,5 First discovered in France in Aug 2023, JN.1 (descendant of BA.2.86) spread rapidly across regions and supplanted XBB.1.5 and other strains as the predominant variant globally (exceeding 60% prevalence) by the end of Jan 2024.6 JN.1 predominance remained till May 2024 when its descendants KP.2 and KP.3 reached global proportions of 9% and 20%, respectively. By December 2024, XEC, a recombinant variant derived from JN.1 descendant lineages, was reported by WHO to represent 36.8% of global sequences.7 As novel Omicron variants continue to emerge and updated COVID-19 vaccines are licensed in an effort to combat them, assessment of long-term immunogenicity and effectiveness of these vaccines against circulating SARS-CoV-2 variants is crucial to inform evidence-based vaccination policies. Studies have consistently shown individuals with hybrid immunity to develop greater antibody breadth than those with vaccination-only or infection-only immunity.8–10 Although the variant-containing boosters raised antibodies and induced protection rapidly, the durability and robustness of this immune response to pre-emergent viruses is limited.11–14 Several studies on bivalent mRNA vaccines have reported waning vaccine effectiveness and neutralizing antibody titers within months,12,13 but the duration of protection remains unclear. Longer follow-up to 12 months post-vaccination is needed to establish if, and for how long, bivalent mRNA vaccines can maintain these levels of immune responses. In addition, head-to-head randomized clinical trials comparing the breadth and magnitude of immune response against future SARS-CoV-2 strains between variant- and ancestral-based COVID-19 vaccines can inform the need for updated vaccines. Compared with traditional vaccine designs, the SARS-CoV-2 mRNA vaccines have the advantage of allowing for rapid modification for emerging variants and induce potent antibody responses although they are more reactogenic and antibody waning occurs rapidly by 6 months.1,15 In Singapore, the only licensed alternative Omicron targeted vaccine for individuals who are unable or do not wish to receive mRNA COVID-19 vaccines is the protein subunit vaccine NVX-CoV2373 (Novavax). In two phase 3 trials conducted during a time when ancestral-like and Alpha variants dominated the infections, NVX-CoV2373 demonstrated 90% and 100% efficacy against symptomatic and severe COVID-19, respectively.16,17 A booster vaccination (three doses) with NVX-CoV2373 was demonstrated to increase neutralizing antibody titers against Omicron BA.1 and BA.4/5 to similar levels as triple mRNA vaccination at one-month post-booster.18 However, first and second booster studies comparing the durability of antibody responses between NVX-CoV2373 and ancestral mRNA-based vaccines have reported a slower antibody decay rate for NVX-CoV2373 compared with mRNA vaccines.19,20 How a fourth-dose NVX-CoV2372 vaccine compared with an mRNA vaccine, including newer generation variant-based vaccines, in terms of antibody breadth and durability remains unknown. This question remains important in the context of emerging SARS-CoV-2 variants and hybrid immunity to inform future booster vaccination recommendations. PRIBIVAC is an adaptive, subject-blinded, randomized phase 4 trial that aims to fill some of these knowledge gaps on SARS-CoV-2 booster vaccination.21 Results from early phases of the study (Phases A and B) that examined the immunogenicity following a first SARS-CoV-2 vaccine booster (i.e. 3rd dose) were reported previously.22–24 The last phase (Phase D) of the trial compared the humoral immune response of different SARS-CoV-2 vaccines administered as second booster (4th dose) in individuals who had received three doses of monovalent mRNA vaccine. The primary immunogenicity endpoint was neutralizing antibody levels against wild-type SARS-CoV-2 and Omicron subvariants at day 28 between monovalent versus bivalent mRNA, monovalent mRNA versus NVX-CoV02373 and bivalent mRNA versus NVX-CoV-2373. This was measured using pseudovirus neutralization test and surrogate virus neutralization test. Durability of humoral immunity was evaluated up to day 360. We hypothesize that a fourth-dose NVX-CoV2373 will induce lower peak titers but with a slower antibody decay and broaden the immune response compared with both ancestral- and variant-based mRNA vaccines. Materials and methods PRIBIVAC is a subject-blinded, randomized-controlled trial to assess the immunogenicity of different SARS-CoV-2 booster vaccine combinations. The last phase of the study (Phase D) examines the immune response following a second SARS-CoV-2 vaccine booster (i.e. fourth dose). Participants were enrolled at the National Centre for Infectious Diseases (NCID), Singapore. Enrollment and randomization From August 2022 to April 2023, we enrolled 176 adults who had received three doses of an ancestral mRNA COVID-19 vaccine (BNT162b2 or mRNA-1273), with the third dose at least 5 months earlier. Key exclusion criterion was an immunocompromising medical condition. Participants were randomized to receive one of the three vaccine categories as second booster: BNT162b2 (monovalent or bivalent (Ancestral/Omicron BA.4/5)), mRNA-1273 (monovalent or bivalent (Ancestral/Omicron BA.1)), or NVX-CoV2373. Randomization was stratified by age (<60 y, ≥60 y), primary vaccine series (ancestral BNT162b2 or mRNA-1273), and first booster vaccine received (ancestral BNT162b2 or mRNA-1273). Participant eligibility for the study followed the licensed indication for vaccination and changed over the course of the study, both as groups recommended by the Singapore Ministry of Health to receive 4th booster changed and as the bivalent mRNA and NVX-CoV237 vaccines became available for clinical use. Older adults were advised to receive their second booster vaccination at the time of study initiation in August 2022, when only the monovalent mRNA vaccines and NVX-CoV2373 were available. This age limit was reduced to include all individuals aged 18 y and above in November 2022 when bivalent mRNA vaccines became available under the National Vaccination Programme – bivalent mRNA-1273 being available from 14 October 2022 and bivalent BNT162b2 available from 12 December 2022. A web-based randomization system hosted by the Singapore Clinical Research Institute was used to randomize study participants.21–23 Randomization was performed on the day of study enrollment by study team from the Singapore Infectious Disease Clinical Research Network. The randomization list with randomized permuted block size of 3 was generated by the trial statistician. Block size was revealed only after study completion. Blood samples were obtained from participants for immunogenicity assessment pre-booster (day −28 to day 0) and at four timepoints post-booster: day 7 (±2 d), day 28 (±7 d), day 180 (±14 d) and day 360 (±14 d).21–23 Participants self-report local and general symptoms (both solicited and unsolicited) experienced in the first 7 d after vaccination on a symptom diary card. Surveillance for COVID-19 infection throughout the study period was performed by measuring anti-nucleocapsid (anti-N) antibody levels using the Elecsys® (Roche, Basel, Switzerland) immunoassays, and by participants’ self-reported positive antigen rapid test (ART) and/or PCR results.22 Participants who had anti-N seroconversion between visits and/or reported positive ART and/or PCR results were considered to have COVID-19 breakthrough infections. Primary endpoint The primary objective for Phase D of PRIBIVAC was to compare the humoral immunity (neutralizing antibody levels against wildtype SARS-CoV-2 and/or VOCs at day 28) between monovalent mRNA, bivalent mRNA and protein vaccine administered as 2nd booster. This was measured using pVNT and sVNT. Sample size calculation The sample size calculation for Phase D is based on estimating the primary endpoint of between arms mean difference in log10-transformed anti-S titer at day 28 with a pre-specified precision. The approach was selected as Phase D aimed to estimate the mean difference in humoral immunity at day 28 between Pfizer vs Moderna, Pfizer vs Nuvaxovid and Moderna vs Nuvaxovid when received as 2nd booster (i.e. three mean difference estimations), without hypothesis testing. A total of sample size of 150 participants (50 in each arm with equal allocation ratio) will achieve a precision width of ±0.13 with 95% probability (i.e. half-width of 95% confidence interval) for between arms mean difference in log10-transformed anti-S titer at day 28, while assuming standard deviation of 0.31 and allowing 5% discontinuation rate. The pooled standard deviation of 0.31 in day 28 log10-transformed anti-S titer from Phase A study participants aged 60 y and above was assumed to apply to responses to the second booster. Phase A aimed to determine whether heterologous vaccine booster led to a non-inferior humoral immune response against wild-type SARS-CoV-2 and VOC at day 28 compared with homologous vaccine booster. The age restriction followed MOH advisory then for those aged 60 y and above to receive a second COVID-19 booster. PASS 21.0 software was used for sample size calculation. Multiplex surrogate virus neutralization test (sVNT) Neutralizing antibody levels in serum samples were tested at a single dilution with a multiplex surrogate virus neutralization test (sVNT) using the Luminex platform as previously described.22,24–26 Briefly, AviTag-biotinylated RBD proteins from wild-type SARS-CoV-2 and Omicron subvariants (BA.1, BA.5, XBB, XBB.1.16, EG.5, JN.1, KP.2, XEC and XDV) were coated on a MagPlex Avidin microsphere (Luminex). RBD-coated microspheres were pre-incubated with serum at a minimum final concentration of 1:20 for 15 min. After 15 min incubation, 50 uL of phycoerythrin (PE)-conjugated hACE2 (GenScript 2ug/mL) were added to the well and incubated for 15 min, followed by two washes with PBS-1% BSA. All incubations were done at 37°C. The final readings were acquired using the MAGPIX system. Pseudovirus neutralization test (pVNT) SARS-CoV-2 wild-type, Omicron BA.1, Omicron BA.5 and XBB1.16 full-length spike pseudotyped viruses were produced and packaged as previously described.25–28 Briefly, 5 million HEK293T cells were transfected with 20 µg of pCAGGS spike plasmid using FuGENE6 (Promega) for 24 h before being incubated with vesicular stomatitis virus (VSV) ∆G luc seed virus (at multiplicity of infection of 5) for 1 h. After two PBS washes, the infected cells were replenished with complete growth media supplemented with 1:5000 diluted anti-VSV-glycoprotein monoclonal antibody (clone 8GF11, Kerafast) and spike pseudoviruses were harvested at 24 h post-infection. For pVNT assay, 2 × 106 RLU of spike pseudotyped viruses were pre-incubated with four-fold serial-diluted serum samples (1:20 to 1:4320 dilution) in a final volume of 50 µL for 1 h at 37°C, followed by infection of huACE2 stably-expressing A549 cells. At 20–24 h post-infection, an equal volume of ONE-Glo luciferase substrate (Promega) was added, and the Cytation 5 microplate reader (BioTek) was used to measure the luminescence signal. Results were interpreted with reference to the nominal “seronegative” threshold of 1:100, which was determined by testing pre-pandemic sera against these pseudoviruses. 50% neutralization titers (NT50) were calculated with non-linear regression, i.e. log (inhibitor) vs normalized response – variable slope, using Graphpad Prism 10.4.0. Statistical methods Demographic and baseline characteristics were summarized by vaccine groups. Categorical variables were analyzed as absolute numbers and relative frequencies and compared using Chi-square test or Fisher’s exact test when the value in any of the cells was below five.22 Continuous variables were analyzed as median and interquartile range (IQR) or mean and SD and compared using Mann-Whitney U test or independent t test as appropriate based on the results of assessing normality using GraphPad Prism version 10.4.0. Homogeneity of variance was assessed using F test prior to conducting the independent t test. Student’s t test was used when equal variance was assumed, and Welch’s t test for unequal variance.22 One-way ANOVA with Holm-Sidak’s multiple comparisons test or Kruskal–Wallis test with Dunn’s multiple comparisons test was used when comparing three or more groups. Paired pVNT levels in each vaccine group were compared using Wilcoxon matched-pairs signed rank test, paired anti-S antibody titers across timepoints were compared using mixed-effects model with Holm-Sidak’s multiple comparisons test. Anti-S antibody titers were log10-transformed for all statistical analysis. Because a large number of participants had breakthrough infections during the study period, a sensitivity analysis was conducted to investigate the impact of breakthrough infections on D360 immunogenicity analyses. In this sensitivity analysis, we included all self-reported or laboratory confirmed (by anti-N) COVID-19 infection. For the post-hoc analysis, anti-S antibody decay modeling was performed to evaluate the decay rates of mRNA vaccines relative to NVX-CoV2373. The geometric mean ratios (GMR) and 95% confidence interval (CI) of the anti-S titers were estimated at D28, D180 and D360 for vaccine arms compared with NVX-CoV2373 as the reference. If the GMRs of a vaccine decreased between D28 and D360, this meant the decay rate of this vaccine’s anti-S titer was faster than NVX-CoV2373 within 360 d post-booster. To test the difference of decay rates between arms, we also present the anti-S fold change between D360 and D28 (D360-to-D28 ratio) for each participant and the geometric mean of D360-to-D28 ratio for each vaccine arm, where a lower ratio indicates a faster decay. The GMRs of the D360-to-D28 ratio (i.e. a ratio of ratios) to NVX-CoV2373 arm were also presented. If a GMR of the D360-to-D28 ratio is less than one, this means the decay is faster than in the NVX-CoV2373 reference arm. The GMRs and 95% CIs were estimated using a multiple linear regression model for each time point separately. The log10-transformed anti-S titers or D360-to-D28 ratios were the dependent variable. All models adjusted for age (<60 y; ≥60 y), baseline immunogenicity, primary vaccination series (mRNA-1273; BNT162b2), first booster (mRNA-1273; BNT162b2), sex, prior SARS-CoV-2 infection and time since first booster (≤1 y, >1 y). The GMR was calculated as the antilogarithm of the adjusted difference between arms in the model. We employed multiple linear regression to explore the effect of booster vaccine (monovalent mRNA or bivalent mRNA or NVX-CoV2373) on post-vaccination anti-S fold change and day 360 anti-S titers. Multiple logistic regression was employed to investigate the effect of vaccine response factors on COVID-19 breakthrough infection. All models controlled for baseline values including randomization stratification factors [age (<60 y; ≥60 y), primary vaccination series (mRNA-1273; BNT162b2) and first booster (mRNA-1273; BNT162b2)], sex, Charlson’s comorbidity index (0; ≥1), smoking status (nonsmoker; smoker; former smoker), prior SARS-CoV-2 infection and time since first booster (≤1 y, >1 y). A multiple linear regression model of day 360 anti-S antibody titers was constructed that included day 28 anti-S titers in addition to baseline factors mentioned above. The multiple linear regression model of post-vaccination anti-S fold change included log10-transformed D28/D0 anti-S antibody ratio as the dependent variable. Normality and homogeneity of variance of residuals were assessed using graphical methods via QQ plot (actual residuals on x-axis vs predicted residuals on y-axis) and residual plot (predicted value on x-axis vs residuals on y-axis), respectively. A multiple logistic regression model of COVID-19 breakthrough infection was constructed that included second booster vaccine group and anti-S antibody titers at day 28 post-booster. Five subjects who had anti-N seroconversion between screening and day 28 were removed from this analysis as they were likely to be infected around the time of booster vaccination. The data were analyzed for all outcomes in a modified intention-to-treat (mITT) population, which included all randomly assigned participants except those who never received the booster vaccination (n = 7), withdrew from the study prior to day 28 (n = 1), had anti-N seroconversion between screening and day 28 (n = 5), or had COVID-19 breakthrough infections (n = 43). Statistical significance was defined as p < .05. Analyses (including NT50 calculation) were performed and figures generated using GraphPad Prism version 10.4.0. Ethics statement and data availability Written informed consent was obtained from all study participants (Domain Specific Review Board ref no: 2021/00821). The study was registered with ClinicalTrials.gov (NCT05142319). All data sharing requests should be addressed to the corresponding authors. Results Participants Among 176 participants who received three doses of mRNA vaccine (BNT162b2 or mRNA-1273), 38 were randomized to receive the monovalent mRNA vaccine (17 BNT162b2 and 21 mRNA-1273), 82 to bivalent mRNA (40 BNT162b2 and 42 mRNA-1273) and 56 to NVX-CoV2373 from August 2022 to April 2023. Monovalent mRNA arm was smaller due to policy changes in late 2022, i.e. the replacement of monovalent mRNA vaccines by the updated bivalent mRNA vaccines. A total of eight participants withdrew from the study − seven withdrew prior to booster vaccination and one withdrew before day 28 follow-up visit. About a quarter (43/163 [26%]) of participants had breakthrough infections by the end of study, of which 30% (13/43) were asymptomatic and detected by anti-N seroconversion. The proportion of breakthrough infections between vaccine groups were similar (monovalent mRNA 23.5% vs bivalent mRNA 24.0% vs NVX-CoV2373 31.5%, p = .58, Chi-square) (Supplementary Figure 1A). In a post-hoc analysis, higher anti-S antibody titers at D28 post-booster (p = .022) and prior COVID-19 infection (p < .0001) were associated with lower odds of breakthrough infection (OR = 0.13; 95 % CI, 0.022–0.69 and OR = 0.089; 95 % CI, 0.029–0.25, respectively), but not affected by type of vaccine booster received (Supplementary Figure 1B). To avoid biases in examining the durability of vaccine-induced antibody responses, participants who contracted SARS-CoV-2 infection during the study were excluded from the immunogenicity analyses – resulting in a final analysis sample size of 26, 57 and 37 for monovalent mRNA (15 BNT162b2; 11 mRNA-1273), bivalent mRNA (26 BNT162b2; 31 mRNA-1273) and NVX-CoV2373 groups, respectively (Figure 1). Figure 1.Consort flow diagram. Consort flow diagram. Baseline characteristics of the cohort (n = 120) stratified by vaccine type are shown in Table 1. Reflecting changes in vaccine availability/recommendations over the course of the study, participants who received the monovalent mRNA vaccines were substantially older than those who received bivalent mRNA vaccines. The dosing interval between the first and second booster doses was also considerably shorter in the monovalent mRNA group compared with bivalent mRNA and NVX-CoV2373.Table 1.Demographics and baseline characteristics of the cohort stratified by vaccine type. Monovalent mRNABivalent mRNANVX-CoV2373n265737Age in years, median (IQR)54 (39–62)36 (26–49)37 (29–60)Male, no. (%)14 (53.8)22 (38.6)22 (59.5)Chinese, no. (%)24 (92.3)50 (87.7)33 (89.2)Charlson comorbidity index, median (IQR)0 (0–0)0 (0–0)0 (0–0)Days since first booster, median (IQR)359 (304–378)420 (386–464)391 (355–459)Current smoker, no.031Prior COVID-19 infectiona no. (%)20 (76.9)53 (93.0)37 (100)Sample size at each timepoint, n – D7– D28– D180– D360– 26– 26– 26– 25– 56– 57– 53– 54– 36– 37– 35– 35Duration of follow-up at each timepoint, median (IQR) – D7– D28– D180– D360– 7 (7–8)– 23 (21–28)– 176 (171–183)– 351 (349–354)– 7 (6–8)– 25 (23–28)– 182 (173–189)– 356 (349–360)– 7 (7–8)– 24 (21–28)– 178 (169–186)– 354 (350–358)aDetermined by RT-PCR or rapid antigen testing and/or positive anti-nucleocapsid antibody at baseline (day 0). Table 1.Demographics and baseline characteristics of the cohort stratified by vaccine type. Monovalent mRNABivalent mRNANVX-CoV2373n265737Age in years, median (IQR)54 (39–62)36 (26–49)37 (29–60)Male, no. (%)14 (53.8)22 (38.6)22 (59.5)Chinese, no. (%)24 (92.3)50 (87.7)33 (89.2)Charlson comorbidity index, median (IQR)0 (0–0)0 (0–0)0 (0–0)Days since first booster, median (IQR)359 (304–378)420 (386–464)391 (355–459)Current smoker, no.031Prior COVID-19 infectiona no. (%)20 (76.9)53 (93.0)37 (100)Sample size at each timepoint, n – D7– D28– D180– D360– 26– 26– 26– 25– 56– 57– 53– 54– 36– 37– 35– 35Duration of follow-up at each timepoint, median (IQR) – D7– D28– D180– D360– 7 (7–8)– 23 (21–28)– 176 (171–183)– 351 (349–354)– 7 (6–8)– 25 (23–28)– 182 (173–189)– 356 (349–360)– 7 (7–8)– 24 (21–28)– 178 (169–186)– 354 (350–358)aDetermined by RT-PCR or rapid antigen testing and/or positive anti-nucleocapsid antibody at baseline (day 0). Demographics and baseline characteristics of the cohort stratified by vaccine type. – D7 – D7 – D28 – D28 – D180 – D180 – D360 – D360 – D7 – D7 – D28 – D28 – D180 – D180 – D360 – D360 aDetermined by RT-PCR or rapid antigen testing and/or positive anti-nucleocapsid antibody at baseline (day 0). At enrollment, 91.7% (110/120) of participants had prior confirmed COVID-19 infection, and a substantially lower proportion was found in monovalent mRNA group (Table 1). The median duration of follow-up after the booster dose was similar between intervention arms: 351 d (IQR 349–354) in monovalent mRNA group, 356 d in bivalent mRNA (IQR 349–360) and 354 d in NVX-CoV2373 (IQR 350–358). Level of anti-Spike antibodies against ancestral SARS-CoV-2 The levels of anti-Spike (S) antibodies against ancestral SARS-CoV-2 were measured in serum samples collected pre-booster (day−28 to day 0) and at days 7, 28, 180 and 360 after the booster dose. Before booster vaccination, the mean anti-S antibody titer across all participants was 21,577 IU/mL (95% confidence interval [CI], 18,323 to 25,410) and similar between intervention groups and age groups (Supplementary Table 1). On multiple regression, baseline anti-S titers were significantly higher among those with prior COVID-19 infection (p < .0001), whereas current smokers showed a trend toward lower titers (p = .052) (Supplementary Table 2). After the booster dose, anti-S titers in mRNA-vaccinated individuals were boosted to high levels within a week and remained significantly higher than NVX-CoV2373 recipients up to day 28 post-booster (Figure 2(a)). Unlike mRNA vaccines which induced a peak antibody response at day 7, anti-S titers among NVX-CoV2373 vaccinees continued to increase from day 7 to day 28 (Supplementary Figure 2). By days 180 and 360, anti-S titers were comparable between vaccine groups. The same trend was observed between the five vaccine groups (Supplementary Figure 3). Figure 2.Longitudinal measurements and kinetics of SARS-CoV-2 anti-spike (S) antibody over a 1-y study period. (a) Level of anti-S antibody between different vaccine groups. Data was analysed using one-way ANOVA with Holm-Sidak’s multiple comparisons test. Bars depict mean and 95% CI. (b) Kinetics of anti-S antibody from D28 post-booster. Data presented are geometric mean concentrations and 95% CI. (c) Summary data. (d) Comparison of anti-S titres at D28, D180 and D360, and D360-to-D28 ratio. One multiple linear regression model was fitted for each time point, adjusting for baseline anti-S titres, age group (<60 y; ≥60 y), the interval between first and second boosters (≤1 y; >1 y), primary vaccination series (mRNA-1273; BNT162b2), first booster (mRNA-1273; BNT162b2), sex and prior SARS-CoV-2 infection. GM: geometric mean; GMR: geometric mean ratio. ***p < .001; ****p < .0001. Longitudinal measurements and kinetics of SARS-CoV-2 anti-spike (S) antibody over a 1-y study period. (a) Level of anti-S antibody between different vaccine groups. Data was analysed using one-way ANOVA with Holm-Sidak’s multiple comparisons test. Bars depict mean and 95% CI. (b) Kinetics of anti-S antibody from D28 post-booster. Data presented are geometric mean concentrations and 95% CI. (c) Summary data. (d) Comparison of anti-S titres at D28, D180 and D360, and D360-to-D28 ratio. One multiple linear regression model was fitted for each time point, adjusting for baseline anti-S titres, age group (<60 y; ≥60 y), the interval between first and second boosters (≤1 y; >1 y), primary vaccination series (mRNA-1273; BNT162b2), first booster (mRNA-1273; BNT162b2), sex and prior SARS-CoV-2 infection. GM: geometric mean; GMR: geometric mean ratio. ***p < .001; ****p < .0001. A difference in anti-S decay rate was observed between vaccine arms. The decay rate of NVX-CoV2373 was slower than that of mRNA vaccines, as shown by the convergence of the three kinetic curves (Figure 2(b)) and a significantly higher D360-to-D28 ratio in NVX-CoV2373 (0.51, 95 % CI, 0.42–0.61) relative to both mRNA vaccines (monovalent: 0.31, 95 % CI, 0.23–0.40, p = .010 and bivalent: 0.35, 95 % CI, 0.31–0.41, p = .017) (Figure 2(c)). The GMRs of both monovalent and bivalent mRNA vaccines compared to NVX-CoV2373 decreased between days 28 and 360, further supporting the notion that the decay rate of mRNA vaccines was faster than NVX-CoV2373 within 360 d post-booster. Absolute anti-S titer waning per year was 16,037 (95 % CI, 13,927–17,908), 41,088 (95 % CI, 37,859–43,733) and 36,777 (95 % CI, 34,714–38,614) IU/mL for NVX-CoV2373, monovalent mRNA and bivalent mRNA groups, respectively. Post-hoc multiple linear regression (Supplementary Figure 4) and sensitivity analysis including all breakthrough infections (Supplementary Figure 5) showed similar results. While post-vaccination anti-S fold change at day 28 was greater after bivalent mRNA vaccination compared with NVX-CoV2373 (monovalent mRNA showed the same trend; p = .075) (Supplementary Figure 4a), day 360 anti-S titers were higher in NVX-CoV2373 vaccines compared with monovalent or bivalent mRNA (Supplementary Figure 4b). Breadth of antibody neutralization against vaccine strains, emerging and future Omicron subvariants To determine the change in neutralization titers against vaccine strains and the emerging Omicron subvariant XBB.1.16 at 28 d post-booster, a pseudovirus neutralization test (pVNT) was used. At baseline, geometric mean 50% neutralization titer (NT50) against ancestral SARS-CoV-2 was 7133 (95 % CI, 5,970–8,522), BA.1 1768 (95 % CI, 1,416–2,209), BA.5 1130 (95 % CI, 915–1396) and XBB.1.16 216 (IQR 175–267) across all participants and were similar between vaccine groups. After the booster dose, NT50 against ancestral, BA.1, BA.5 and XBB.1.16 increased in all vaccine groups by day 28 except ancestral SARS-CoV-2 in NVX-CoV2373 vaccines (Figure 3(a)). Geometric mean NT50 against BA.5 in NVX-CoV2373 group increased from baseline 1399 (95 % CI, 1029–1902) to 1934 (95 % CI, 1491–2507) at day 28, but this did not achieve statistical significance (p = .094). Across all participants, geometric mean NT50 was lower against newer Omicron subvariants (BA.1 > BA.5 > XBB.1.16). Viral neutralization against BA.1, BA.5 and XBB.1.16 were 2.3-, 4.1- and 16.5-fold lower compared to wildtype. Importantly, the 4th vaccine dose significantly increased the proportion of participants seropositive for XBB.1.16 (NT50 > 100) from 80% (96/120) pre-booster to 96.7% (116/120) one-month post-booster (p < .0001). 100% of bivalent mRNA vaccines seroconverted. Among the four participants who remained seronegative for XBB1.15, three had received NVX-CoV2373 and one monovalent mRNA. Figure 3.50% neutralization titers (NT50) against full-length spike protein of vaccine strains and emerging Omicron XBB.1.16 subvariant in (a) each vaccine group, (b) between vaccine groups at D28, (c) fold change relative to pre-boost titres, and (d) summary data. Level of percent inhibition was determined using pseudovirus neutralisation test (pVNT). (a) Lines connect data points that belong to the same individual. Pre- (D0) and D28 post-booster titers were compared using paired Wilcoxon signed rank test. Red dotted line indicates the nominal “seronegative” threshold of 1:100. (b–c) NT50 between vaccine groups were compared using Kruskal–Wallis test with Dunn’s multiple comparisons test. Bars depict geometric mean and 95% CI. (d) Geometric mean titres with 95% CI are shown. *p < .05; **p < .01; ***p < .001; ****p < .0001. 50% neutralization titers (NT50) against full-length spike protein of vaccine strains and emerging Omicron XBB.1.16 subvariant in (a) each vaccine group, (b) between vaccine groups at D28, (c) fold change relative to pre-boost titres, and (d) summary data. Level of percent inhibition was determined using pseudovirus neutralisation test (pVNT). (a) Lines connect data points that belong to the same individual. Pre- (D0) and D28 post-booster titers were compared using paired Wilcoxon signed rank test. Red dotted line indicates the nominal “seronegative” threshold of 1:100. (b–c) NT50 between vaccine groups were compared using Kruskal–Wallis test with Dunn’s multiple comparisons test. Bars depict geometric mean and 95% CI. (d) Geometric mean titres with 95% CI are shown. *p < .05; **p < .01; ***p < .001; ****p < .0001. Between vaccine groups, neutralization titers against vaccine strains (ancestral, BA.1 and BA.5) were comparable between monovalent vs bivalent mRNA and monovalent vs NVX-CoV2373 at day 28 post-booster (Figure 3(b)). NT50 against XBB.1.16 was highest in bivalent mRNA group at 1.8- and 2.3-fold higher compared with monovalent mRNA (p = .034) and NVX-CoV2373 (p < .0001), respectively. Analysis of XBB.1.16 and the closely related XBB by a surrogate virus neutralization test (sVNT) demonstrated the same finding (Supplementary Figure 6). Although XBB.1.16 sVNT inhibition level at day 28 reached saturation, there exists a clear positive correlation between NT50 and level of inhibition (Supplementary Figure 6c). When normalized to pre-booster titers, mRNA vaccines elicited a higher NT50 fold change than NVX-CoV2373 against BA.5 and XBB.1.16 (Figure 3(c)). To assess the breadth and longevity of antibody neutralization induced by the different vaccine types, a multiplex sVNT was conducted. At day 180 post-booster, inhibition level against BA.1 and BA.5 were highest in bivalent mRNA vaccines (Supplementary Figure 7). By day 360, BA.1 and BA.5 inhibition levels were comparable between groups and remained higher compared to pre-booster. Although the median WT SARS-CoV-2 inhibition level was modestly different between groups at day 360 likely due to saturation, monovalent mRNA group had the highest inhibition level (Supplementary Table 3). Analysis by the five vaccine arms demonstrated the same finding (Supplementary Figure 8). While inhibition levels against earlier Omicron subvariants are likely underestimated due to saturation, they provide qualitative information about the level of neutralizing antibodies between vaccine arms (Supplementary Figures 6–8). Apart from vaccine strains BA.1 and BA.5, other Omicron subvariants tested include EG.5, JN.1, KP.2, XEC and XDV. During the conduct of the trial between late Aug 2022 to early May 2024, the dominating circulating variants in Singapore included XBB, XBB.1.16, EG.5 and JN.1 according to a national surveillance program (data not shown). These variants are classified as “emerging variants” in this report. KP.2 emerged toward the end of the trial in Mar 2024, by which only 15% (26/168) of participants had yet to return for their last study visit. Recombinant variants XEC and XDV emerged after trial completion. Collectively, these Omicron subvariants that emerged close to or after trial completion are termed “future variants.” The level of neutralizing antibodies increased by day 28 post-boost in all vaccine groups for all Omicron subvariants tested, including future variants (Supplementary Figure 9). Across all participants, the largest absolute increase in inhibition level at day 28 post-booster was observed against the KP.2 subvariant (pre-booster 72.6%, IQR 61.0%−83.8% vs Day 28 post-booster 87.2%, IQR 77.7%−93.2%; p < .0001; Wilcoxon signed rank test). Comparing intervention groups, bivalent mRNA vaccines had the highest level of neutralizing antibodies against circulating strains EG.5 and JN.1, both of which were yet to emerge at the time of sample collection at day 28 (Figure 4(a), Supplementary Table 4). Neutralizing antibody levels against future variant KP.2 was also highest in bivalent mRNA vaccines at day 28, with a similar trend observed for XEC and XDV (Figure 4(b), Supplementary Table 5). By 1-y post-booster, the level of neutralizing antibodies against all Omicron subvariants tested were similar regardless of vaccine received. Sensitivity analysis demonstrated the same finding (Supplementary Figure 10). Similar to the kinetics of anti-S antibody, while NVX-CoV2373 induce a smaller magnitude of increase in neutralizing antibody levels at day 28 post-booster, the rate of antibody decline is slower compared to monovalent and bivalent mRNA vaccines (Figure 4(c)). Figure 4.Longitudinal measurements of antibody neutralization against Omicron subvariants determined using a multiplex surrogate virus neutralization test (sVNT) as previously described. Omicron subvariants tested are grouped as (a) circulating strains and (b) future variants. Red dotted line indicates inhibition of 30% (nominal “seronegative” threshold). Comparisons between vaccine groups are made using Kruskal–Wallis test with Dunn’s multiple comparisons test. Summary data for (a) and (b) is presented in Supplementary Tables 3 and 4 respectively. (c) Graphical representation of the waning of neutralising antibodies between vaccine groups over the 1-y study period. Each point represents the median neutralisation level with IQR for that timepoint. *p < .05; **p < .01; ***p < .001; ****p < .0001. Longitudinal measurements of antibody neutralization against Omicron subvariants determined using a multiplex surrogate virus neutralization test (sVNT) as previously described. Omicron subvariants tested are grouped as (a) circulating strains and (b) future variants. Red dotted line indicates inhibition of 30% (nominal “seronegative” threshold). Comparisons between vaccine groups are made using Kruskal–Wallis test with Dunn’s multiple comparisons test. Summary data for (a) and (b) is presented in Supplementary Tables 3 and 4 respectively. (c) Graphical representation of the waning of neutralising antibodies between vaccine groups over the 1-y study period. Each point represents the median neutralisation level with IQR for that timepoint. *p < .05; **p < .01; ***p < .001; ****p < .0001. Discussion As divergent Omicron immune escape subvariants continue to emerge, booster vaccinations can augment protection by increasing humoral responses.29 However, the rapid waning of protection against infection with pre-emergent Omicron sub-variants12,30 has raised questions about the optimal long-term vaccination strategy for the general population. Not only was vaccine effectiveness of the updated XBB.1.5 mRNA vaccines lower against pre-emergent JN.1 infection than XBB sublineages, it also waned faster.30 Similarly, a more rapid waning of vaccine effectiveness was observed in BA.4/BA.5 bivalent vaccine against XBB sublineages compared to the vaccine strains BA.4/BA.5.12 Another durability study of BA.4/BA.5 bivalent booster by Lin et al. also reported rapid waning of vaccine effectiveness against vaccine strains from its peak of 25.5% at 2 weeks to only 4.2% by week 15 post-vaccination.31 Collectively, these studies demonstrate lower clinical protection of the updated vaccines against antigenically advanced strains and rapid fading of vaccine effectiveness within months, but durability of neutralization levels against novel Omicron subvariants remain unknown. Randomized clinical trials providing head-to-head comparisons of binding and functional antibody responses between different vaccine platforms and between ancestral and variant-based vaccines are limited. In this study, we compared the 1-y duration of humoral response between monovalent mRNA vs bivalent mRNA vs protein-based NVX-CoV2373 booster in adults who received three doses of monovalent mRNA vaccine. Here, we found that a fourth dose of the bivalent mRNA vaccine elicited more immunogenic responses than monovalent mRNA and NVX-CoV2373 vaccine in the first month post-booster. Our results are consistent with, and extend findings from, previous trials that compared either the monovalent and bivalent mRNA vaccines,32 or the bivalent mRNA to protein vaccine.1 However, unlike the PICOBOO fourth dose study that reported minimal neutralizing antibody against Omicron subvariants using ancestral-based vaccines,19 here we show that all vaccine types, regardless of whether they are ancestral- or variant-based, elicited robust neutralizing antibody responses against newer Omicron subvariants such as EG.5, JN.1 and KP.2 that were yet to emerge at the time of trial. This distinction may reflect differences in the rate of prior COVID-19 infections between the two studies which in PICOBOO was only 21% versus 92% in PRIBIVAC. This suggests that in the context of hybrid immunity, these vaccines offer not only ‘back-boosting’ of immune memory but drive continued maturation of the immune response resulting in increased breadth of neutralization. A molecular-level investigation of polyclonal IgG anti-S antibodies elicited from infection-only, vaccination-only, or hybrid immunity has revealed distinct immunological imprinting between infection and vaccination.10 While the primary IgG response is imprinted predominantly against epitopes residing outside the RBD upon infection, vaccination mainly induces anti-RBD antibodies. Upon secondary exposures, over 60% of the original IgG pool is recalled, and hybrid immune IgG monoclonal antibodies can increase neutralization breadth, potency and binding affinity against SARS-CoV-2 variants. In addition, studies have shown individuals with hybrid immunity to have considerably greater antibody half-life (i.e. longer antibody durability) than individuals with only vaccine-induced immunity.33,34 Collectively, these data support the national policy recommendation for individuals at increased risk of severe COVID-19 (individuals aged 60 and above, medically vulnerable individuals and residents of aged care facilities) to receive the updated COVID-19 vaccine even if they have had prior COVID-19. This is because the benefit is greatest for these groups. Immunity wanes over time and variants continuously evolve to evade existing immunity. Subsequent exposures to the latest strain-matched vaccine can increase antibody potency and binding affinity of the recalled IgG responses, thereby increasing protection against severe disease. Bivalent mRNA vaccines were substantially more immunogenic than the monovalent mRNA and NVX-CoV2373 vaccines at 28 d post-vaccination, with a higher level of neutralizing antibodies against XBB, XBB.1.16, EG.5, JN.1 and KP.2. Although neutralizing antibodies have been associated as correlates of protection against SARS-CoV-2 infection,35 specific protective thresholds against emerging variants of interest have not been identified and how these translate into clinical protection against emerging SARS-CoV-2 variants will require further real-world vaccine effectiveness studies, particularly in the context of an antigenic mismatch between circulating virus and vaccine strain. Furthermore, the role of T cells in vaccine-induced protection against SARS-CoV-2 infection and disease is increasingly being recognized. Unlike the negative effect of immune imprinting on neutralizing antibodies, a recent pre-clinical study found little immune imprinting on spike-specific T cells in mice receiving prototype mRNA vaccine as primary series followed by bivalent mRNA vaccine booster.14 Cellular immunity is not reported here, but ongoing analysis will examine T cell responses using high-dimensional flow cytometry to provide insights into potential long-term protection. With increased public hesitancy surrounding mRNA vaccines,36 the immunogenicity and effectiveness of alternative vaccine designs remains important to assess, however, longitudinal studies comparing NVX-CoV2373 with mRNA vaccines past 6 months have been very limited. Several studies have shown monovalent mRNA vaccines to induce higher initial immunological responses than NVX-CoV2373 vaccine, but NVX-CoV2373 had a slower antibody decay rate.19,20 Our study is consistent with and extend previous findings to day 360 post-second booster. Although bivalent mRNA-boosted group elicited the highest neutralization levels against several Omicron subvariants at day 28, the three vaccine groups had comparable neutralizing antibodies by month 12, including against vaccine strains BA.1 and BA.5 that reached the upper limit of saturation. The significantly higher D360-to-D28 anti-S ratio in NVX-CoV2373 relative to mRNA vaccine groups (monovalent and bivalent), as well as the post-hoc linear regression analysis which found NVX-CoV2373 group to have significantly higher anti-S titers than mRNA groups at day 360 provide evidence that protein vaccine has a slower antibody decay rate compared with mRNA vaccines. Sensitivity analysis showed that our immunogenicity results and study conclusions on durability of vaccine-induced antibody responses were robust. The enhanced durability of immune response may be attributed to the presence of adjuvant (Matrix-M) in NVX-CoV2373, which aids in the formation of germinal center reactions and bone marrow plasma cells that are critical for ongoing antibody secretion and memory.37 This is supported by the preclinical study in rhesus macaques which found three immunizations with NVX-CoV2373 to induce a stable pool of bone marrow-resident plasma cells.38 Concerns have emerged regarding immune escape with newer Omicron subvariants. Across all participants, we observed a 2.3-fold decrease in viral neutralization against BA.1, 4.1-fold decrease against BA.5 and 16.5-fold decrease against XBB.1.16 in comparison to wildtype. Successive waves of COVID-19 driven by immune-evasive Omicron subvariants has led the WHO and other regulatory authorities to recommend the updated variant-based vaccines for booster programs. However, there is limited real-world data comparing the immunogenicity against current and emerging variants between vaccine types (mRNA and protein), including updated vaccines as booster doses. Similar to other NVX-CoV2373 booster vaccine studies,18,40,41 our results demonstrate that NVX-CoV2373 vaccination, despite being an ancestral vaccine, can elicit potent neutralizing antibody responses against different Omicron subvariants including the newly emerging XBB.1.16, KP.2, XEC and XDV. The ability of NVX-CoV2373 to induce broadly cross-reactive antibodies may be driven by the presentation of epitopes that are conserved across variants on the full-length S protein42 and may possibly be enhanced by the adjuvant via antibody avidity, affinity maturation, and epitope spreading.43 While the updated XBB.1.5-containing Novavax vaccine (NVX-CoV2601) was shown to elicit superior neutralizing antibody responses against XBB.1.5 compared with ancestral-based NVX-CoV2373 at day 28, the breadth of neutralizing antibodies against antigenically distant SARS-CoV-2 strains was not assessed.44 Nonetheless, our findings together with others support the utility and continued updates to this vaccine platform. Interestingly, while NVX-CoV2373 induced modest but statistically significant increases in neutralizing antibody titers against Omicron subvariants at day 28 post-booster, the titers against ancestral SARS-CoV-2 strain did not increase. This may be a consequence of negative interference from existing high baseline neutralizing titers against ancestral SARS-CoV-2 strain in these triple mRNA-vaccinated individuals, resulting in limited capacity of NVX-CoV2373 to further boost the titers post 3rd dose. In the recently published PREVENT-19 trial where immunogenicity data after receipt of 3rd and 4th doses of NVX-CoV2373 were reported, neutralizing antibody titers against the ancestral strain post 4th dose were comparable to if not less than those post 3rd dose.39 Another possible explanation is the delayed induction of neutralizing antibodies by NVX-CoV2373 past day 28, as neutralizing antibodies against ancestral SARS-CoV-2 were reported to increase up to day 84 post-booster.19 To our knowledge, this is the first long-term immunogenicity report (up to 12 months post-booster) of NVX-CoV2373 administered as a 4th heterologous dose. The significant boosting of neutralizing antibodies against antigenically advanced strains at day 28 and comparable levels at day 360, regardless of vaccine platform and variant type, supports the use of heterologous protein-based booster vaccines as an acceptable alternative to mRNA-based COVID-19 booster vaccines. This provides flexibility in developing pragmatic booster programs, where additional factors like medical ineligibility for mRNA vaccines, vaccine stability and individual preference for specific vaccine types come into play. In addition, the finding that seroresponse rates to XBB.1.16 were significantly different between the groups (100% among bivalent mRNA vaccines vs 87.5% monovalent mRNA vs 50% NVX-CoV2373) provides supporting evidence for the continued updates to the COVID-19 vaccine formulations, in particular the protein vaccine platform due to the lower magnitude and slower boosting of antibody responses. The study has several limitations. Due to the shift in second booster vaccination policy toward increasingly younger age groups in Singapore, and the later addition of bivalent mRNA vaccines into the study, participants who received monovalent mRNA vaccines were older and the interval between booster doses shorter. However, these factors are unlikely to affect our immunogenicity analysis as baseline anti-S antibody titers, anti-S fold change post-vaccination and durability of antibody titers at day 360 were neither influenced by age nor time since first booster dose on post-hoc analysis. The replacement of ancestral-based (monovalent) mRNA vaccines to bivalent mRNA vaccines shortly after study initiation, in addition to the relatively large number of breakthrough infections which were subsequently removed from immunogenicity analysis, resulted in a smaller than intended sample size for monovalent mRNA and NVX-CoV2373 groups. As a result, the study may lack statistical power to compare the immunogenicity data between vaccine groups. Nonetheless, the findings on longevity of humoral immunity still yield important relevance as many adults would have received these prototype vaccines as a fourth dose and have yet to receive the updated COVID-19 vaccine booster. Moreover, the high prevalence of prior infection in our cohort is reflective of the current global COVID-19 situation, with nations across various income brackets exhibiting an overall increase in their infection rates45 especially since the emergence of the highly-transmissible Omicron in late 2021, thereby making our data generalizable to the population at large. Another limitation was that serum samples were tested at a single dilution for sVNT assay, resulting in saturated measurements for earlier SARS-CoV-2 variants, in particular the vaccine strains (ancestral SARS-CoV-2, BA.1 and BA.5), XBB and XBB.1.16. While measurements above the linear range of the sVNT assay (20–90%) risk an underestimation of the true magnitude of difference between vaccine arms and/or over time, they demonstrate a positive correlation with NT50 even at saturation and hence still provide valuable qualitative information between vaccine arms at a particular timepoint and between timepoints. Nonetheless, neutralizing titers against vaccine strains were also measured using pVNT, allowing for the accurate analysis of humoral immunity against vaccine strains and XBB.1.16 at day 28 post-booster. Lastly, our results may not be generalizable to updated protein vaccines such as the Omicron XBB.1.5-based protein vaccine (NVX-CoV2601),44 which were not available at the time of trial conduct. Nevertheless, our study demonstrates that the protein-based NVX-CoV2373 vaccine induced a slower rate of antibody waning compared to mRNA vaccines and is an alternative for heterologous booster vaccination. Conclusion As COVID-19 becomes an endemic disease, national immunization committees will need to take into account the risk of newly emerged SARS-CoV-2 variants, the overall immunity of the population as a whole, as well as differences between vaccines that can be used as boosters. Among a population with high rates of vaccination and infection, we observed broad serological response to booster vaccination, including against antigenically advanced strains. In particular, the results demonstrate the short-term immunogenicity and the long-term antibody durability of mRNA and protein-based vaccines. Despite the early immunogenic advantages of bivalent mRNA vaccination, humoral responses decay at a faster rate following mRNA vaccines than NVX-CoV2373, with populations exhibiting comparable antibody levels after 1 y. This suggests that individuals who are unable or reluctant to receive mRNA COVID-19 vaccines may consider NVX-CoV2373 to boost and maintain their antibody levels. In addition, our study shows that a flexible, platform-induced booster approach is warranted for the highly immunogenic populations. Although antibody levels are at similar levels by 12 months, clinical effectiveness against infection or severe disease requires real-world evaluation. Supplementary Material PRIBIVAC PHD_KHVI supp_final.docx