Study population and setting
This study used in vitro SARS-CoV-2 antibody neutralization data from published vaccine studies and related these data to vaccine efficacy data (against symptomatic SARS-CoV-2 infection) from phase 3 clinical trials. The authors considered seven vaccines: 1) Moderna (mRNA-1273), 2) Novavax (NVX-CoV2373), 3) Pfizer (BNT162b2), 4) Sputnik (rAd26-S+rAd5-S), 5) AstraZeneca (ChAdOx1 nCoV-19), 6) Johnson & Johnson (Ad26.COV2.S), and 7) Sinovac (CoronaVac). The authors used the mean and standard deviation of the neutralization titer from published studies of the seven vaccines; these were normalized to the mean convalescent titer from the same assay in each study. Then, to estimate a “50% protective neutralization level,” the authors fit a logistic model that best fit the observed protection in each of the seven studies, expressing the estimated 50% protective neutralization titer in relation to the mean convalescent titer across studies. There were several ancillary analyses, including estimation of decline in neutralization titers and estimation of the 50% protective neutralization titer against severe infection.
Summary of Main Findings
There was a strong positive, non-linear relationship between neutralization titers and the protective efficacy of the seven vaccines (Spearman r = 0.905; p<0.01). From a logistic model, the estimated protective neutralization level providing 50% protection against infection was 20.2% (95% CI: 14.4% to 28.4%) of the mean convalescent level. Using a non-parametric method, the estimated protective threshold (which estimates a binary cutoff, above which full protection is assumed) was 28.6% (19.2% to 29.2%) of the mean convalescent level. Models fit to neutralization titers measured between 26 and 115 days after the time origin (either mRNA-based vaccination or symptom onset from infection) found similar half-lives between vaccinated (65 days) and convalescent sera (58 days, p=0.88). The estimated 50% protective neutralization titer against severe infection was 3% (0.7% to 15%) of the mean convalescent level.
Normalizing antibody titers to the mean convalescent titer from the same assay in each study allowed the authors to compare results across studies, despite the fact that the seven studies used widely different neutralization assays. The authors were explicit about the assumptions underlying their analyses.
This analysis aggregated data from studies that used different neutralization assays, definitions of convalescent plasma, case definitions for symptomatic infection, case definitions for severe infection, and time periods under consideration. Although standardizing by mean convalescent titer addressed the differences in neutralization assays, other considerable disagreements in classification across studies could have affected the results in unpredictable ways. The study was unable to assess the role of age or other individual factors in immune protection (e.g., how neutralizing antibody responses vary by age) or how other immune system components (e.g., B cells or T cells) may impact the host response to SARS-CoV-2. Importantly, finding a correlate of vaccine protection does not necessarily mean the correlate is causally related to protection, even though there are reasons to suspect a causal relationship in this case. Results from this study should not be used to establish an absolute antibody threshold of protection; doing so requires further study with data at an individual level.
Along with a separate study published at the same time that arrived at similar conclusions, this study is an extremely important demonstration that neutralizing antibody response appears to be a strong correlate of vaccine protection against SARS-CoV-2 infection.
This review was posted on: 21 September 2021