Immunogenicity subcohort, case-cohort set, and RSV endpoints

An immune correlate analysis was conducted based on participants in the case-cohort set (Fig. S2) who had RSV nAb and IgG bAb data assessed at Day 29 during the window (15–43) postinjection (vaccine or placebo), and had the RSV endpoint onset or censored more than 7 days after Day 29 (accounting for potential alteration of Day 29 antibody marker by likely natural RSV infection on the occurrence of RSV endpoints). The case-cohort set was comprised of a stratified random subcohort of participants (immunogenicity subcohort), plus all postinjection RSV-ARD cases (including early cases before 14 days postinjection)¹⁹, and participants for the correlates analyses were denoted as the Day 29 case-cohort set (Fig. S3). Information about the case-cohort sampling design is provided in the supplement.

RSV endpoints (RSV-LRTD-2+, RSV-LRTD-3+, and RSV-ARD) were analyzed in the correlates analyses by the data cutoff (April 30, 2023), the same RSV endpoints studied in the additional analysis¹⁶. In the correlates analyses, per each RSV endpoint, cases were defined as participants in the Day 29 case-cohort set, with corresponding RSV endpoint onset more than 7 days after Day 29; non-cases were defined as participants in Day 29 case-cohort set with no evidence of the corresponding RSV endpoint onset up to the data cutoff. Overall, the correlates analyses included 2059 participants for analyzing RSV-LRTD-2+ (44 breakthrough cases in vaccine vs. 114 cases in placebo); 2071 participants for analyzing RSV-LRTD-3+ (19 breakthrough cases in vaccine vs. 49 cases in placebo); and 2045 participants for analyzing RSV-ARD (79 breakthrough cases in vaccine vs. 160 cases in placebo) (Tables S1 andS2), respectively. The maximum event time of RSV endpoints following Day 29 was 345 days (the study period), which was used to estimate cumulative incidence and VE against RSV endpoints following Day 29.

Participant demographics

Participant characteristics in the per-protocol immunogenicity subcohort (vaccine, n = 1489; placebo, n = 327) are displayed in Table S3, which consisted of all participants in the immunogenicity subcohort who had the RSV-ARD endpoint onset or censored more than 7 days after Day 29. Of 1816 participants in the subcohort, 45% were females, 45.5% were aged ≥75 years, 39.7% had a LRTD risk factor (COPD or chronic heart failure [CHF]) present at baseline (Day 1), 57.4% had ≥1 pre-existing comorbidities of interest (COPD, asthma, chronic respiratory disease, diabetes, CHF, advanced liver disease or renal disease), and 13.4% had a history of COVID-19. Overall, 56.5% were from the Northern hemisphere, 56.2% lived in high-income countries, and 45.9% were Hispanic or Latino. Participants in the per-protocol immunogenicity subcohort were well balanced in the vaccine and placebo groups according to these key baseline characteristics, but were not fully representative of the study cohort for the immune correlate analysis. Inverse probability of sampling weight (IPS-weight) was calculated based on the sampling design and applied to adjust the Day 29 case-cohort set in the correlates analyses (see Supplementary Information).

Both Day 29 and baseline antibody levels were lower in RSV breakthrough cases versus non-cases

For all Day 29 and baseline antibody markers, 100% of vaccine and placebo recipients had antibody levels above the assay detection limits (Table S4, Fig. 2, Figs. S4–S6). Furthermore, the ratio of geometric mean (GM) values was approximately 0.5 to 0.8 comparing RSV-LRTD-2+ cases to non-cases for all Day 29 and baseline antibody markers in both vaccine and placebo groups (Table 1, Table S5). In the vaccine group, Day 29 RSV-A and RSV-B nAb geometric mean titers (GMTs) in RSV-LRTD-2+ cases were lower than those observed among non-cases; a similar trend was observed for Day 29 preF IgG and postF IgG geometric mean concentrations (GMCs) (Table 1, Table S5).

a RSV-A nAb. b RSV-B nAb. c RSV preF IgG bAb. d RSV postF IgG bAb. The violin-box plot is composed of an interior box plot and rotated probability density plots (estimated by a default Gaussian kernel density estimator) of the antibody marker data on each side. In the box plot, the middle line and the lower and upper horizonal edges represent the 50th, 25th, and 75th percentile of antibody titers or concentrations, and the vertical whiskers represent the distance from the 25th (or 75th) percentiles of antibody titers or concentrations and the minimum (or maximum) antibody titers or concentrations within the 25th (or 75th) percentile of antibody level minus (or plus) 1.5 times the interquartile range. The GMT or GMC level and the corresponding 95% confidence interval are adjusted by the IPS-weight. ARD acute respiratory disease, bAb binding antibody, GMC geometric mean concentration, GMT geometric mean titer, IgG immunoglobulin G, IPS inverse probability of sampling, LLOQ lower limit of quantification, LRTD lower respiratory tract disease, nAb neutralizing antibody, postF postfusion, preF prefusion, RSV respiratory syncytial virus, ULOQ upper limit of quantification.

For Day 29 RSV-A nAb and preF IgG markers, the GM value was approximately 8–11 times higher for vaccine than placebo recipients by RSV-LRTD-2+ case status; for Day 29 RSV-B nAb and postF IgG markers, the respective GM value was about 6 times higher for vaccine than placebo recipients. Each pair of the 4 Day 29 antibody markers was highly correlated (e.g., preF bAb and RSV-A nAb, Spearman correlation r = 0.88; preF IgG and RSV-B nAb, \(r=0.79\); and RSV-A nAb and RSV-B nAb, r = 0.76; Figs. S7–S9), while baseline antibody markers were moderately correlated with each other (Figs. S10–S12). In addition, concordance analysis showed high concordance rates between RSV nAb and preF IgG markers by pooling baseline and Day 29 markers together (Fig. S13). For all Day 29 and baseline antibody markers, the GM value and ratio of GM (case vs. non-case; vaccine vs. placebo) by RSV-LRTD-3+ and RSV-ARD case status are shown in Tables S6–S7. Reverse cumulative distribution function curves for each Day 29 antibody marker are displayed in Figs. S14–S16.

Day 29 antibody marker level inversely correlates with the risk of RSV endpoints

All Day 29 antibody markers were significantly inversely correlated with RSV-LRTD-2+ and RSV-ARD risk, and were consistently inversely correlated with RSV-LRTD-3+ risk (Table 2, Table S8). The covariate-adjusted hazard ratios (HR) of RSV-LRTD-2+ and RSV-ARD per each 10-fold increase in marker levels were similar (0.40–0.55). Given the small number of RSV-LRTD-3+ breakthrough cases (n = 19) in vaccine recipients, only RSV-A nAb showed significant inverse correlation with RSV-LRTD-3+ risk; the other three markers showed consistent inverse correlations, albeit this did not reach statistical significance (p-values > 0.05). Table S9 shows that the correlation of each Day 29 antibody marker and each RSV endpoint was not statistically different between vaccine and placebo.

The estimated cumulative incidence of each RSV endpoint by placebo group and vaccine tertile subgroups (defined by Day 29 antibody marker tertiles in vaccine recipients) indicated that RSV risk decreased from placebo through low, medium, and high vaccine tertile subgroups for each marker (except postF IgG) (Figs. S17–S19). The estimated instantaneous hazard rate of each RSV endpoint by placebo and vaccine tertile subgroups for each Day 29 antibody marker is shown in Figs. S20–S22, depicting coherent trends to support RSV risk inversely correlating with increased vaccine tertiles. Covariate-adjusted HR of RSV-LRTD-2+ for vaccine tertile subgroups versus placebo showed decreased risk with increasing vaccine tertile subgroups for RSV-A nAb, RSV-B nAb, and preF IgG, but not for postF IgG (Fig. 3). For RSV-LRTD-2+, p-values and family-wise error rate (FWER)-adjusted p-values for all vaccine tertiles by each Day 29 antibody markers were significant (p < 0.05). Covariate-adjusted HRs of RSV-LRTD-3+ and RSV-ARD in vaccine tertile subgroups are shown in Table S10, demonstrating that RSV risk decreased with increments of RSV nAb and preF IgG levels.

Covariate-adjusted instantaneous hazard rate and cumulative incidence of RSV-LRTD-2+ by placebo and by low, medium, and high tertile of Day 29 RSV nAb or IgG bAb marker level in vaccine recipients. (a) and (c) for Day 29 RSV-A nAb. (b) and (d) for Day 29 RSV preF IgG bAb. e Day 29 RSV-A and RSV-B nAb and RSV preF and postF IgG bAb. Baseline risk factors are adjusted in the univariable (qualitative) IPS-weighted Cox PH regression model, including the actual stratification factors age and LRTD at-risk and baseline risk score. bAb binding antibody, FWER family-wise error rate, IgG immunoglobulin G, IPS inverse probability of sampling, LRTD lower respiratory tract disease, nAb neutralizing antibody, PH proportional hazard, point est. point estimate, postF postfusion, preF prefusion, RSV respiratory syncytial virus.

Day 29 antibody marker level positively correlates with VE for RSV endpoints

Fig. S23 shows that the marginal treatment effect for each RSV endpoint was consistent with clinical VE, and that the treatment effect conditional on each Day 29 bAb and nAb marker (except postF IgG) was not significant and was nearly mediated (i.e., estimated HR close to 1; although to a lesser extent for RSV-LRTD-3+ due to a smaller number of breakthrough cases [n = 19] in the vaccine group), indicating the risk of getting each RSV endpoint can be predicted by Day 29 antibody marker independently of treatment (vaccine or placebo). By the Prentice surrogate endpoint criteria²⁰, all RSV nAb and preF IgG markers measured at Day 29 were supported as surrogate endpoints (i.e., CoPs) for all RSV endpoints; RSV-A nAb and preF IgG showed the strongest evidence (small HRs of RSV endpoints per 10-fold increase marker levels and large p-values for nearly mediated conditional treatment effect).

Figure 4 displays further correlates analysis specifically for RSV-LRTD-2+ and shows the estimated (1) cumulative incidence of RSV-LRTD-2+ during the study period across a range of assigned marker levels by vaccine and placebo; (2) controlled VE against RSV-LRTD-2+ during the study period across a range of assigned marker levels; and (3) cumulative incidence of RSV-LRTD-2+ during the study period above a range of assigned marker levels (thresholds) by Day 29 RSV-A nAb and RSV preF IgG markers of vaccinees (see Supplementary Information), respectively. Specifically, for both Day 29 RSV-A nAb and preF IgG, the estimated cumulative incidence of RSV-LRTD-2+ by vaccine and placebo group was similar and overlapped by bootstrap pointwise 95% CIs; estimated VE increased as the antibody marker level increased. Additionally, we conducted a threshold analysis for each Day 29 antibody marker by analyzing subgroups with antibody levels greater than or equal to a certain value, supporting that the higher postvaccination immune response of subgroups correlated with decreased risk of RSV disease. Further CoR and CoP analyses for each RSV endpoint by Day 29 antibody marker are shown in Figs. S24–S29.

a and d Solid red and blue curves demonstrate point estimates of the covariate-adjusted cumulative incidence of RSV-LRTD-2+ during the study period for vaccine and placebo recipients across a range of assigned antibody titers or concentration levels (within 0.5th–99.5th percentiles of observed antibody values in vaccine and placebo groups). Dashed red and blue curves, along with the shades, represent the bootstrap pointwise 95% CIs. Solid and dashed horizontal gray lines represent the point estimates and 95% CIs of the average covariate-adjusted cumulative incidence of RSV-LRTD-2+ in vaccine and placebo recipients. b and e Solid black curve shows the point estimate of controlled VE across a range of assigned antibody titers or concentration levels (within 0.5th–99.5th percentiles of observed antibody values in vaccine and placebo groups), dashed black curves demonstrate the bootstrap pointwise 95% CIs. Rug lines below and above represent RSV-LRTD-2+ cases and non-cases by vaccination status, respectively. The shaded gray area between dashed curves highlights the VE above the median of the antibody level in placebo recipients. Solid and dashed horizontal gray lines are point estimates and 95% CIs of clinical VE in the additional analysis. c and f Red curve along the blue area represents the reverse cumulative density function values for the observed antibody marker values in vaccine recipients. The black curve is the covariate-adjusted cumulative incidence of RSV-LRTD-2+ during the study period above a range of antibody marker levels (below the 97.5th percentile of observed antibody values in the vaccine group). The shadowed gray area is the bootstrap pointwise 95% CIs. The stacked histogram of the observed antibody marker titers or concentration levels by vaccination status overlayed on the bottom of cumulative incidence plots (a, d) and VE plots (b, e). Baseline covariates of age, LRTD at-Risk, and baseline risk score are adjusted in the IPS-weighted Cox regression model. bAb binding antibody, CDF cumulative distribution function, CI confidence interval, CoP correlate of protection, CoR correlate of risk, IgG immunoglobulin G, IPS inverse probability of sampling, LRTD lower respiratory tract disease, nAb neutralizing antibody, postF postfusion, preF prefusion, RSV respiratory syncytial virus, VE vaccine efficacy.

Day 29 antibody markers mediate the majority of mRNA-1345 VE against RSV endpoints

For all Day 29 antibody markers, the majority of marker levels (>90%) had overlapping distributions for vaccine and placebo recipients; therefore, it was feasible to assess how much VE for each RSV endpoint could be mediated by these markers using an adapted approach from Benkeser et al.²¹ (see Supplementary Information). The estimated proportion of VE mediated through the antibody markers at Day 29 was highest for RSV-A nAb and preF IgG for all RSV endpoints; Day 29 RSV-B nAb showed comparable or moderate VE mediation against each RSV endpoint compared to Day 29 RSV-A nAb; postF IgG showed limited VE mediation against each RSV endpoint compared to the other 3 Day 29 markers (Fig. S30).

CoR analysis of Day 29 antibody markers against RSV endpoints by RSV subtypes A and B

We further studied Day 29 antibody markers as CoRs against each RSV endpoint by RSV-A and -B subtypes (Fig. S31). The results of the covariate-adjusted HR of each RSV endpoint by RSV subtype per 10-fold increase in Day 29 preF IgG demonstrated that this antibody marker was very consistent as a CoR for all RSV endpoints regardless of subtype (Table S11). Day 29 postF IgG showed less consistent evidence as CoRs compared with preF IgG. Overall, RSV-A and RSV-B nAb were the best CoRs for RSV subtype-matched endpoints, but RSV-A nAb more consistently showed inverse correlates with RSV endpoints by both subtypes versus RSV-B nAb (Table S11).

CoR analysis of baseline antibody markers against RSV endpoints

Using the same Cox regression model as above, we also investigated whether baseline antibody markers were CoRs against each RSV endpoint. All baseline antibody markers except postF IgG were significantly inversely correlated with all RSV endpoints (Fig. S32). Importantly, the treatment effect conditional on any individual baseline antibody markers was similar to the marginal treatment effect (estimated HR in vaccine: 0.36 [0.24–0.55] for RSV-LRTD-2+, 0.27 [0.12–0.61] for RSV-LRTD-3+, and 0.52 [0.36–0.76] for RSV-ARD). This result shows that the probability of RSV depends on both baseline antibody markers and treatment, suggesting that pre-vaccination (baseline) antibody markers alone are not useful for predicting VE.

CoR analysis of fold-rise antibody markers against RSV endpoints

Lastly, we studied whether fold-rise antibody markers are CoRs and mediate protection against each RSV endpoint. As opposed to the above baseline and Day 29 antibody marker analyses, only vaccine recipients in the Day 29 case-cohort set were used to study the correlates of fold-rise antibody markers, since the fold-rise in any antibody markers measured at Day 29 compared to Day 1 (baseline) was nearly unchanged (close to 1) in placebo recipients (Tables S12–S14, Figs. S33–S35). For each antibody marker, fold-rise in vaccine recipients decreased as the baseline marker level increased; fold-rise of RSV-A nAb was significantly higher in breakthrough cases versus non-cases in vaccinees for each RSV endpoint. Fold-rise in other antibody markers was not significantly or consistently different in breakthrough versus non-cases in vaccinees for any RSV endpoint. Moreover, Cox-based model results showed that only fold-rise in RSV-A nAb was significant; the other fold-rise markers showed consistent positive correlates with RSV endpoints to a different extent than RSV-A nAb (Fig. S36). Both GM and model results provide consistent evidence to support fold-rise antibody markers positively correlating with the risk of RSV endpoints. A potential explanation of this paradoxical finding is that, due to the negative correlation between fold-rise and baseline marker level, the positive correlation of fold-rise in antibody markers with the risk of RSV endpoints could be largely masked by the inverse correlation of baseline level with the risk of RSV endpoints.