Prediction models for neutralization activity against emerging SARS-CoV-2 variants: A cross-sectional study.

Objective: Despite extensive vaccination campaigns to combat the coronavirus disease (COVID-19) pandemic, variants of concern, particularly the Omicron variant (B.1.1.529 or BA.1), may escape the antibodies elicited by vaccination against SARS-CoV-2. Therefore, this study aimed to evaluate 50% neutralizing activity (NT50) against SARS-CoV-2 D614G, Delta, Omicron BA.1, and Omicron BA.2 and to develop prediction models to predict the risk of infection in a general population in Japan. Methods: We used a random 10% of samples from 1,277 participants in a population-based cross-sectional survey conducted in January and February 2022 in Yokohama City, the most populous municipality in Japan. We measured NT50 against D614G as a reference and three variants (Delta, Omicron BA.1, and BA.2) and immunoglobulin G against SARS-CoV-2 spike protein (SP-IgG). Results: Among 123 participants aged 20-74, 93% had received two doses of SARS-CoV-2 vaccine. The geometric means (95% confidence intervals) of NT50 were 65.5 (51.8-82.8) for D614G, 34.3 (27.1-43.4) for Delta, 14.9 (12.2-18.0) for Omicron BA.1, and 12.9 (11.3-14.7) for Omicron BA.2. The prediction model with SP-IgG titers for Omicron BA.1 performed better than the model for Omicron BA.2 (bias-corrected R 2 with bootstrapping: 0.721 vs. 0.588). The models also performed better for BA.1 than for BA.2 (R 2 = 0.850 vs. 0.150) in a validation study with 20 independent samples. Conclusion: In a general Japanese population with 93% of the population vaccinated with two doses of SARS-CoV-2 vaccine, neutralizing activity against Omicron BA.1 and BA.2 were substantially lower than those against D614G or the Delta variant. The prediction models for Omicron BA.1 and BA.2 showed moderate predictive ability and the model for BA.1 performed well in validation data.