Tuesday, June 6, 2023

Long term anti-SARS-CoV-2 antibody kinetics and correlate of protection against Omicron BA.1/BA.2 infection – Nature Communications

This longitudinal antibody study with follow-up times up to 20 months provided the opportunity to understand long-term anti-SARS-CoV-2 antibody dynamics and to evaluate binding antibody levels from a commercial widely available immunoassay as a correlate of protection against infections during the Omicron BA.1/BA.2 era. Anti-S antibodies persisted up to 20 months after the probable date of infection, with decay dynamics determined by infection and vaccination history. The strongest and longest-lasting antibody boosts occurred with vaccine doses following prior infection. Modeled antibody trajectories enabled the evaluation of binding antibody levels as a correlate of protection against Omicron BA.1/BA.2 infections, for which we found an overall three-fold reduction in the hazard of reporting a positive test for antibody levels above 800 IU/mL. Hazard reduction was, however, not observed for non-infected participants, indicating that the validity of anti-S binding antibody levels as correlates of protection for Omicron BA.1/BA.2 depends on infection history.

This study extends our previous work showing that anti-SARS-CoV-2 spike antibodies remain detectable after 22 months past probable infection as measured with the Roche anti-S immunoassay22. Our kinetic modeling results support previous findings indicating that antibody boost is strongest and longest lasting in vaccinees with a history of infection19,23,24. In contrast with previous findings, we found no significant difference in antibody boosting between age groups and slower decay rates in adults 65 years and older20,25. The slower decay rates may be due to age-specific differences in disease severity that we did not account for in these models, thus limiting the comparability of these findings with previous studies due to differences in disease severity profiles. Furthermore, we had a small number of participants over 65 years of age in our study, and these age-stratified results should be interpreted with caution. Finally, our results highlight the strong individual-level variability in antibody dynamics, which has been shown in previous antibody kinetic studies25,26.

Survival analysis results on Omicron BA.1/BA.2 infections are in line with previous findings from vaccine trials targeting the ancestral strain and the Alpha variant, showing that binding antibody levels are an informative correlate of protection against SARS-CoV-2 infection5,12. These trials had found similar effect sizes of around a fivefold reduction in risk of Alpha infections at anti-S antibody levels of 600 IU/mL12 and a halving of hazard by a 10-fold increase in anti-S titers for ancestral strain infections5. Moreover, our results are in line with available studies on Omicron BA.1/BA.2 subvariants, which have also found binding antibody levels to be correlates of protection against infection using in-house immunoassays15,16,17,18. In particular, one study using the same immunoassay found similar effect size estimates in a prospective study design using measured antibody levels, although not differentiating between infection/vaccination statuses and finding large confidence intervals16. On the other hand, we did not find differences in the hazard of having an Omicron BA.1/BA.2 infection with anti-S antibody levels below or above a certain threshold in the non-infected vaccinated group, as opposed to results reported for Delta infections13. Notably, this finding is supported by our recent work on neutralization capacity in the Geneva population2. Using the same immunoassay as in this study and a cell-free Spike trimer-ACE2 binding-based surrogate neutralization assay8, we did not observe any significant correlation between anti-S binding and neutralizing antibody levels against Omicron subvariants in uninfected participants, as opposed to previously infected participants2. These results can be linked to growing evidence that hybrid immunity (infection plus vaccination) provides the strongest protection against Omicron subvariant infections17,18,27. This infection history-specificity thus warrants care in the interpretation of binding antibodies as correlates of protection against Omicron sublineages and could be immunoassay-dependent.

We note that it remains unclear whether these correlate of protection results extend to subsequent Omicron subvariants (BA.4, BA.5, BA.2.75, BQ.1, and others), which have been found, thanks to specific mutations, to have stronger immune evasion capacity than the parent BA.1 strain28,29. Changes in immune evasion capacity may, theoretically if multiple mutations accumulate on the spike protein, impact the level of binding antibodies at which hazard reduction occurs, as well as its effect size. Moreover, our longitudinal serology follow-up was conducted before the circulation of the Omicron lineage in Geneva. The interpretation of anti-S antibody levels measured with this immunoassay following Omicron infections might need to be revisited in light of the evidence of reduced test sensitivity towards antibodies targeting the Omicron Spike protein30.

This study has several limitations. Firstly, we only used the Roche Elecsys assay, which measures total anti-S antibodies (IgA/M/G), whose levels may correlate differently with overall immune function following infection or vaccination; other immunoassays may have different antibody binding characteristics. In particular, changes in the assay’s antibody avidity with time since infection may impact the interpretation of antibody levels prior to infection related to functional immune capacity10. Secondly, analyses in the 65+ subgroup are limited by the small number of participants. Thirdly, our survival analysis to assess correlates of protection was based on modeled antibody trajectories and not on measurements at defined time points as done in studies available from vaccine trails5. Although modeled trajectories matched well with antibody participant-level measurements, the survival analysis results are subject to modeling uncertainty. While sensitivity analysis using the 2.5% and 97.5% prediction quantiles yielded qualitatively similar correlations of protection results, other sources of modeling uncertainty cannot be excluded. A further modeling limitation relates to the immunoassay’s limit of quantification at the 1:10 dilution level used in this study, which for logistical constraints, could not be further diluted. We addressed this issue through censoring within our modeling framework. We further note that we could not account for unreported infections between the last serology for each participant and the exposure period, given that exclusion criteria for the survival analysis relied on reported infections. This may have resulted in the miss-classification of participants as falsely below the antibody threshold, thus leading to an underestimation of the reduction in infection hazard (i.e., bias towards the null). Fourthly, a large proportion of virologically confirmed infections (44%, 101/227) were self-reported as opposed to the other 56%, which were directly extracted from the state COVID-19 test registry (ARGOS). Reassuringly, of the 1083 participants in our longitudinal sample for whom tests in the registry were available, self-reported positive tests with matching dates were reported in 82% (491/599) of cases, thus suggesting a reasonable sensitivity of self-reporting. Moreover, we note that a proportion of Omicron infections during the exposure period were not reported due to subclinical infections, individual (self-)testing, and test reporting practices. As our outcome measure of reported infections is mainly composed of symptomatic infections, it may probably be more representative of more severe BA.1/BA.2 infections than infections across the clinical spectrum of the disease. Finally, both Omicron BA.1 and BA.2 subvariants circulated in the canton of Geneva during the study exposure period, and sequencing information on infection was not available, thus precluding a differential correlation of protection analysis for both subvariants.

Overall, this study extends findings against previous SARS-CoV-2 variants showing that anti-S binding antibody levels measured by a widely distributed immunoassay are a valid correlate of protection against Omicron BA.1/BA.2 infections. Importantly, we found that the validity of antibody levels as a correlate of protection depends on infection history as quantified with the immunoassay used in this study. Our results highlight the imperfect nature of protection after vaccination and/or infection. Even with perfect knowledge of infection and vaccination histories, inference about population-level immunity continues to pose challenges. Future studies may benefit from the modeling framework developed in this study to leverage longitudinal measurements to epidemiological outcomes. Taken together, these conclusions motivate further investigation of how immune landscape and immunoassay characteristics determine the interpretation of serological surveys into population levels of protection to inform public health decisions.

Source link

Related Articles

Leave a Reply

Stay Connected

- Advertisement -spot_img

Latest Articles

%d bloggers like this: