Predicting vaccine effectiveness for mpox

Abstract

The Modified Vaccinia Ankara vaccine developed by Bavarian Nordic (MVA-BN) was widely deployed to prevent mpox during the 2022 global outbreak. This vaccine was initially approved for mpox based on its reported immunogenicity (from phase I/II trials) and effectiveness in animal models, rather than evidence of clinical efficacy. However, no validated correlate of protection after vaccination has been identified. Here we performed a systematic search and meta-analysis of the available data to test whether vaccinia-binding ELISA endpoint titer is predictive of vaccine effectiveness against mpox. We observe a significant correlation between vaccine effectiveness and vaccinia-binding antibody titers, consistent with the existing assumption that antibody levels may be a correlate of protection. Combining this data with analysis of antibody kinetics after vaccination, we predict the durability of protection after vaccination and the impact of dose spacing. We find that delaying the second dose of MVA-BN vaccination will provide more durable protection and may be optimal in an outbreak with limited vaccine stock. Although further work is required to validate this correlate, this study provides a quantitative evidence-based approach for using antibody measurements to predict the effectiveness of mpox vaccination.

Fig. 1. The estimated effectiveness of different mpox vaccine regimens.

A The estimated effectiveness of first-generation vaccines administered years earlier (1st Gen, n = 5), as well as recent vaccination with a single dose of MVA-BN (n = 6), and two doses of MVA-BN (n = 4) are shown. The coloured points show the estimated effectiveness (median of the posterior distribution) for each regimen and study, the error bars indicate the corresponding 95% credible intervals. Where two regimens were compared in the same study, the effectiveness of the two regimens is joined by a line. The combined estimates (median of the posterior distribution) and 95% credible intervals for each regimen across all studies are shown in black. B The additional protection (odds ratio) provided by a two-dose regimen compared to a one-dose regimen (n = 4). The combined estimates (black) are the medians of the posterior distribution (circle) with the 95% credible intervals (error bars).

Fig. 2. Comparison of the reported geometric mean vaccinia-binding titers induced by vaccination with MVA-BN (n = 12) and historic first-generation vaccines (n = 3).

The GMTs from the freeze-dried MVA-BN formulation (n = 5, light-color) are higher than the liquid frozen formulation (n = 8, dark-color). That is, the ratio of titers in the freeze-dried and liquid frozen formulations (median of the posterior distribution) is 1.32-fold (95% credible interval:1.20−1.48). The points and error bars indicate the GMT reported in each study, along with 95% confidence bands extracted from each study respectively. Horizontal lines indicate the combined estimate (median of the posterior distribution) for each vaccination and formulation (shaded regions are the 95% credible intervals). The different colours represent the different vaccines with one and two doses of MVA-BN coloured differently.

Fig. 3. Relationship between vaccine effectiveness and the vaccinia-binding GMT.

The contour lines, represent the 20%, 40%, 60% and 80% highest density regions of the joint-posterior distribution (i.e., smallest areas that contain x% of the posterior samples) for the different vaccines (indicated by colour). The association between antibody titers and effectiveness (solid black line) is fitted using all of the underlying data (accounting for the interstudy heterogeneity using a hierarchical model structure) (Table S5). The solid black line indicates the best estimate (median of posterior), and shaded region show the 95% credible intervals of the predicted effectiveness at different GMTs.

Fig. 4. Predicting the durability of protection.

A The decay in GMT in the different immunogenicity trials as fitted using a two-phase decay model across all immunogenicity studies (n = 13). The estimated GMT (median of the posterior distribution, solid lines) and 95% credible intervals (shaded) are shown over the two-year period for which immunogenicity data was measured in the one and two dose schedules (vaccine schedule indicated by colour). This is compared to the GMT (points) and standard deviation (error bars) reported in each study. B The effect of delayed dosing as estimated from fitting all the studies (n = 13) in (A). The median GMT (points) and 95% credible intervals (error bars) at the approximate peak (2 weeks after the final dose or 28 days after the first dose, whichever is later) is shown for different MVA-BN vaccination regimens. C The predicted GMT one-year post-vaccination (points) and 95% credible intervals (error bars), accounting for the early fast-decay and the late slow-decay of antibodies using the model fitted in (A) (Table S8). Predicted GMTs for regimens without a datapoint later than 5 months post-vaccination are shown with reduced opacity. D The predicted vaccine effectiveness (solid line) and 95% credible intervals (shaded area) over a 10-year period for the different vaccination schedules. The grey region highlights the prediction extrapolated beyond the available time course of immunogenicity data. The three-dose schedule involves vaccination on day 28 and two years after the initial dose (3rd dose delivered on day 730).