Predicting vaccine effectiveness against severe COVID-19 over time and against variants: a meta-analysis

Abstract

Vaccine protection from symptomatic SARS-CoV-2 infection has been shown to be strongly correlated with neutralising antibody titres; however, this has not yet been demonstrated for severe COVID-19. To explore whether this relationship also holds for severe COVID-19, we performed a systematic search for studies reporting on protection against different SARS-CoV-2 clinical endpoints and extracted data from 15 studies. Since matched neutralising antibody titres were not available, we used the vaccine regimen, time since vaccination and variant of concern to predict corresponding neutralising antibody titres. We then compared the observed vaccine effectiveness reported in these studies to the protection predicted by a previously published model of the relationship between neutralising antibody titre and vaccine effectiveness against severe COVID-19. We find that predicted neutralising antibody titres are strongly correlated with observed vaccine effectiveness against symptomatic (Spearman [Formula: see text] = 0.95, p < 0.001) and severe (Spearman [Formula: see text] = 0.72, p < 0.001 for both) COVID-19 and that the loss of neutralising antibodies over time and to new variants are strongly predictive of observed vaccine protection against severe COVID-19.

Fig. 1. Summary of previously available data linking neutralising antibodies and vaccine effectiveness and data contributing to this analysis.

A Previously reported relationship between neutralising antibody titre and vaccine efficacy in the prevention of symptomatic (dark red) and severe (orange) COVID-19 (reproduced from Khoury et al.). Solid lines indicate best-fit model and shaded areas indicate 95% confidence intervals. Neutralisation titre and efficacy data used to parameterise the model are indicated as dots (95% CI indicated as whiskers). B–D Summary of the clinical studies used in this analysis (Table 1).

Fig. 2. Results of linear mixed effects regression model fit to vaccine effectiveness data against severe COVID-19 extracted from our systematic review.

Vaccine effectiveness data against severe COVID-19 (points and whiskers for 95% CI) are shown for pre-Delta (top row, panels (A)–(D), Delta (middle row, panels (E)–(H)) and Omicron (bottom row, panels (I)–(L)) variants. Note that for panels (D) and (I), no effectiveness data was available. Opacity indicates the degree of confidence in the data as determined by the width of the confidence interval. Predicted vaccine effectiveness using a linear mixed effects regression model (solid lines) and 95% confidence intervals (shaded area) are overlaid. The figure shows effectiveness following mRNA-1273 (red), BNT162b2 (blue), any mRNA (purple) and ChAdOx-nCov-1 (pink) vaccination.

Fig. 3. Correlation between estimated neutralising antibody titres and vaccine effectiveness.

Correlation between estimated neutralising antibody titres (accounting for vaccine used, variant studied and time since vaccination) and clinical data for A vaccine effectiveness against symptomatic SARS-CoV-2 infection, B vaccine effectiveness against severe COVID-19. C Correlation between vaccine effectiveness against symptomatic and severe COVID-19. Solid lines indicate the predicted relationship taken from ref. , and shading indicates 95% CI of the model estimates. X-axis confidence intervals in (A) and (B) represent the degree of confidence in the estimate neutralising antibody titre. A breakdown of the relationship shown in panels (A) and (B) by variant and study type is shown in Supplementary Fig. S1. Figure shows effectiveness following mRNA-1273 (red), BNT162b2 (blue), any mRNA (purple) and ChAdOx-nCov-1 (pink) vaccination against pre-Delta (pink background) Delta (blue background) and Omicron (green background) variants.