Correlates of Protection, Thresholds of Protection, and Immunobridging among Persons with SARS-CoV-2 Infection

Abstract

Several studies have shown that neutralizing antibody levels correlate with immune protection from COVID-19 and have estimated the relationship between neutralizing antibodies and protection. However, results of these studies vary in terms of estimates of the level of neutralizing antibodies required for protection. By normalizing antibody titers, we found that study results converge on a consistent relationship between antibody levels and protection from COVID-19. This finding can be useful for planning future vaccine use, determining population immunity, and reducing the global effects of the COVID-19 pandemic.

Keywords: 2019 novel coronavirus disease; COVID-19; SARS-CoV-2; coronavirus disease; correlates of protection; neutralizing antibody; respiratory infections; severe acute respiratory syndrome coronavirus 2; standardization; vaccine efficacy; viruses; zoonoses.

Figure 1

Predicting protection from symptomatic SARS-CoV-2 infection by using approaches to elucidate the relationship between neutralizing antibody titers and protection from COVID-19 (the protection curve): the vaccine-comparison (A–C) and breakthrough-infection (D–F) approaches. The 2 approaches are illustrated schematically: data used (A, D); model fit to data (B, E); and estimated protection (C, F) The vaccine-comparison approach used data on mean neutralization titer from phase 1/2 vaccine trials (normalized to convalescing persons in the same study; x-axis) and observed vaccine efficacy against symptomatic SARS-CoV-2 infection in phase 3 trials (y-axis; n = 7 vaccine trials plus 1 study of infection risk in convalescing persons) (A, B). Using the observed distribution in neutralization titers for a given vaccine and the protection curve, we sum over the whole population to predict the proportion of susceptible (red) or protected (blue) persons for a given vaccine and to estimate protective efficacy for different neutralizing antibody levels (C). Fitting across all vaccines and convalescent persons simultaneously derives the protection curve that best fits the neutralization and protection data (B). The breakthrough-infection model uses neutralization titers of persons with symptomatic breakthrough infections (n = 36 for mRNA-1273 [Moderna, https://www.modernatx.com] and n = 47 for ChAdOx1 [AstraZeneca, https://www.astrazeneca.com]) and uninfected persons (n = 1,005 for mRNA-1273 and n = 828 for ChAdOx1) (3,4). This method’s underlying risk model adjusts for demographic risk factors and for the probability of being sampled in the study to remove these potential sources of bias (E). The protection curve reflects an estimate of the vaccine efficacy in subgroups of persons with specific neutralization titers after the 2-phase sampling design was adjusted for (F). Data and model relationship in panels A and B are from (2).

Figure 2

Comparisons of the estimated curves for protection from SARS-CoV-2 infection from 2 vaccines: A) mRNA-1273 (Moderna, https://www.modernatx.com) (4); B) ChAdOx1 (AstraZeneca, https://www.astrazeneca.com) (3). The relationships between vaccine efficacy against COVID-19 infection (y-axis) and neutralization titers (protection curve) that were estimated in each study (–4) are shown. The protection curve derived from the vaccine-comparison model (red dashed line) is compared with the modeled protection curves estimated from breakthrough-infection studies by Gilbert et al. (4) (dark brown for the results from the ID₅₀ and teal lines for the results from the ID₈₀ neutralization titer in in vitro pseudovirus neutralization assays) (A) and Feng et al. (3) (purple for the results from in vitro native (live) SARS-CoV-2 virus and light brown for the pseudovirus neutralization assays) (B). Shaded areas indicate 95% CIs from each model. These curves were extracted from the cited studies (Appendix, https://wwwnc.cdc.gov/EID/article/29/2/22-1422-App1.pdf), and differences between assays were controlled for by normalizing the curve from each study by the mean neutralization titer of the uninfected vaccinees in each study. The normalized curves were then represented on a fold-of-convalescent scale by multiplying by the mean neutralization titer of vaccinees compared with convalescing persons as reported in the phase 1/2 trials (9,10). The vaccine-comparison model agrees closely with the breakthrough-infection models in the neutralization titer ranges where data were most abundant (vertical gray lines indicate 10th to 90th percentiles of the data available in each study). ID₅₀, 50% infectious dose; ID₈₀, 80% infectious dose.

Figure 3

Breakthrough-infection data and protection from SARS-CoV-2 infection showing the association between neutralizing antibody titer and protection from symptomatic SARS-CoV-2 infection for an individual person. A) BNT162b2 (Pfizer-BioNTech, https://www.pfizer.com) (5); B) mRNA-1273 (Moderna, https://www.modernatx.com), pseudovirus ID₅₀ (4); C) ChAdOx1 (AstraZeneca, https://www.astrazeneca.com), live virus (3); D) ChAdOx 1, pseudovirus ID₅₀ (3). The protection curve derived from the vaccine-comparison model (red dashed line and shading 95% CIs) is compared with the observed normalized frequencies of neutralization level (calculations in Appendix) of breakthrough infections reported in 3 studies (gray/black dots). Data from 2 mRNA vaccine studies of mRNA-1273 (A) and BNT162b2 (B), and the adenoviral vector vaccine ChAdOx1 nCoV19 (C, D) are shown. Lower opacity dots indicate fewer persons with neutralization titers in that range. Also shown in each panel are modelled protection curves showing the relationship between individual neutralizing antibodies and protection estimated in each breakthrough-infection study. Note: Breakthrough-infection data of BNT162b2 vaccinees were generously supplied by the authors of reference (5). The data were unavailable for the other 2 studies and were extracted from the original manuscripts; extraction of data from Gilbert et al. (4) was conducted manually and may be less reliable than that of the other studies (Appendix). ID₅₀, 50% infectious dose; ID₈₀, 80% infectious dose.