Abstract
The emergence of the Omicron variant (B.1.1.529) of SARS-CoV-2 has raised concerns about how mutations in the spike protein might influence immune escape and vaccine protection against infection and disease, COVID-19. Initial estimates of immune escape measure neutralizing antibody titers, which have been shown to be a correlate of protection for COVID-19, but vary among studies. However, no studies have examined variation in vaccine effectiveness (VE) using estimated reductions in neutralizing antibody titers across virus variants. We quantified consistency in relative neutralizing antibody titers across studies. We then examined relationships between variant-specific reductions in neutralizing antibodies and protection against documented infection, symptomatic disease, and hospitalizations across variants and vaccines. We found considerable variation in variant-specific neutralizing antibody titers between studies, but within-study comparisons across variants were far more robust. There was insufficient data to estimate VE for a single vaccine across variants, especially for higher levels of immune evasion (>7-fold reductions in neutralizing antibody titers) observed with the Omicron variant (40-fold). Instead, we leveraged variation among both vaccines and virus variants to estimate VE -neutralizing antibody titer relationships across a 30 to 100-fold range of neutralizing antibody titers reduction. Omicron increased the risk of hospitalization four to five-fold and increased the risk of symptomatic disease seven to ten-fold for mRNA vaccinees, with similar relative effects for recently vaccinated, or individuals with waned antibody titers. Third doses restored titers and protection to levels similar to waned immunity against Delta. Overall, these analyses indicate that vaccine effectiveness against severe disease is significantly diminished for waned individuals, and protection against infection, symptomatic disease and transmission is nearly eliminated. However, third doses significantly ameliorate these reductions but only restore protection to levels equivalent to waned protection against the Delta variant. The invasion of Omicron is likely to result in widespread infection, and substantial hospitalizations unless widespread boosting of immunity occurs.
Funding California Department of Health, National Science Foundation
Introduction
The evolution of SARS-CoV-2 in populations with high immunity from infection or vaccination is likely to lead to new variants with substantial immune escape (1, 2), including the Omicron variant (B.1.1.529) in November 2021. Determining the extent of immune evasion on the effectiveness of vaccines is critical both for assessing the need for new and variant-specific vaccines and additional non-pharmaceutical interventions to limit spread(3). Traditional vaccine effectiveness studies can be costly and can only be performed when there is significant transmission of a new variant in a partly vaccinated population. Delays in implementing interventions until these studies can be performed can result in rapid growth of a new variant, stress on health-care systems and preventable disease and death (3).
A much quicker way to estimate VE against a new variant is to use a surrogate of protection, such as neutralizing antibody titers, which can often be measured much more quickly than performing a traditional VE study. Variation in neutralizing antibody titers across vaccines are strongly correlated with protection against symptomatic disease, all infections, and transmission (4-7). Here, we extend this approach and develop relationships between VE against documented infection, symptomatic disease, and hospitalization and neutralizing antibody titers. We use these relationships to estimate the impact of immune evasion, specifically, reductions in neutralizing antibody titers on VE, including for the Omicron variant.
Methods
Reductions in neutralizing antibody titers by variant
We collected data from a recent review (8) on neutralizing antibody titers for four SARS-CoV-2 variants, Alpha, Beta, Gamma, and Delta (Table S1), and for each vaccine, normalized by convalescent sera (4). We excluded studies that did not include values for at least 3 variants because between-study variation was very high (see below). We divided each estimate by the neutralizing antibody titer for wild-type (WT) SARS-CoV-2 (Wuhan-Hu-1, US-WA1/2020, B, B.1) from each study, to produce a ratio representing the reduction (expressed as an x-fold difference) between a variant and WT virus. We estimated the mean differences among variants using a linear model with variant as a fixed effect (a mixed model with study as a random effect had weaker support; AIC: 0.7). We used these variant-specific reduction in neutralizing antibody levels (relative to WT) and the neutralizing antibody levels for each vaccine relative to convalescent sera (4) to estimate a vaccine-variant specific relative neutralizing antibody titer ratio, NATR.
VE against documented infection, symptomatic disease, and hospitalization
We collected data from the literature (including a systematic living review: (9)) on VE for COVID-19 and categorized each study by variant type (Alpha, Beta, Gamma, Delta) and endpoint (hospitalization, symptomatic disease, and documented infection). We excluded estimates where a dominant variant for the study period was not reported, or where two variants were circulating and no variant-specific estimate was provided.
Relationships between VE and neutralizing antibody titers by vaccine and variant
We used generalized linear models with a binomial distribution with a logit link to model the relationship between variant-specific reductions in neutralizing antibody titers and VE for each endpoint across all vaccines and variants for each endpoint. We used these relationships to estimate VE for three endpoints, documented infection, symptomatic disease, and hospitalization, for the Omicron variant, using a ∼40-fold reduction in neutralizing antibody titers relative to WT (10-12). We also estimated VE for all infections and transmission given infection using previously identified relationships between VE and neutralizing antibody titers (5).
All code and data to replicate the results can be found at https://github.com/marmkilpatrick/New-Variant-VE.
Results
There was enormous variation in the ratios of neutralizing antibody titers to WT (5-10-fold), but variation among variants in normalized titers was much smaller (Fig 1). Neutralizing antibody titers relative to WT were1.86-fold (95% CI: 1.3-2.4) lower for Alpha, 6.7-fold (95% CI: 5.0-9.1) lower for Beta, 2.3-fold (95% CI: 5.0-9.1) for Gamma, and 2.7-fold (95% CI: 2.7-5.1) lower for Delta (Fig 1).
Reduction in neutralizing antibody titers plotted against SARS-CoV-2 variant. Each point and color show ratio from an individual study, and points from the same study are connected by lines. Black squares and error bars show the mean and 95% confidence intervals for each variant. Means and 95% confidence intervals are 1.73 (95% CI 1.27-2.36) for Alpha, 6.50 (95% CI: 4.77-8.85) for Beta, 2.74 (95% CI: 1.44-5.22) for Delta, and 2.30 (1.69-3.13) for Gamma.
There were VE estimates for five vaccines against 2 to 4 variants for hospitalization (Fig 2), symptomatic disease (Fig 3), and documented infection (Fig 4). In some cases there was evidence for reductions in VE with decreasing neutralizing antibody titers, but for hospitalization and symptomatic disease endpoints the VE estimates for the variant with the largest reduction in neutralizing antibody titers (Beta) were highly uncertain, resulting in a vary narrow range of neutralizing antibody titers, and insufficient information to estimate the VE for Omicron based solely on comparisons within a vaccine across variants.
VE against hospitalization plotted against variant-specific reductions in neutralizing antibody titers by vaccine. Each point (and 95% CI) represents a single estimate of VE for a single vaccine & virus variant from observational studies. Points have been jittered along the x-axis to facilitate presentation.
VE against symptomatic disease plotted against variant-specific reductions in neutralizing antibody titers by vaccine. Each point (and 95% CI) represents a single estimate of VE for a single vaccine & virus variant from observational studies. Points have been jittered along the x-axis to facilitate presentation.
VE against documented infection plotted against variant-specific reductions in neutralizing antibody titers by vaccine. Each point (and 95% CI) represents a single estimate of VE for a single vaccine & virus variant from observational studies. Points have been jittered along the x-axis to facilitate presentation.
When we analyzed variation in VE with neutralizing antibody titer ratios across both variants and vaccines there were strong relationships across a 30 to 110-fold range of neutralizing antibody titers for the three endpoints (Fig 5A-C). Estimated VE against hospitalization for Omicron was substantially lower than Delta, with relative risk (1-VE) increasing four to five-fold for both Pfizer and Moderna across all vaccination states (Table S3; Fig 6). For example, VE was estimated to be 96.3% (95% CI: 96.1-96.6) against the Delta variant for recently vaccinated Pfizer recipients but 84.9% (95% CI: 83.0-86.6) against the Omicron variant (Table S3; Fig. 6). Waned VE was 88.8% (95% CI: 87.7-89.8) for Delta but dropped to an extrapolated 63.1% (95% CI: 56.9-68.9) w/ Omicron. Fortunately, a third Pfizer dose reversed much of the loss in protection and increased VE to 91.7% (95% CI: 91.0-92.2) against Omicron, which was slightly higher than waned Pfizer VE against Delta (Fig 6). Relative patterns were similar for Moderna vaccinees, except VE was a bit higher than Pfizer for recently vaccinated and waned, but boosting was less effective, resulting in VE being similar for both Omicron and Delta for Pfizer and Moderna vaccinees (Fig 6).
VE against (A) hospitalization (B) symptomatic disease, and (C) documented infection, plotted against variant- and virus-specific neutralizing antibody titer ratios by vaccine. Each point (and 95% CI) represents a single estimate of VE for a single vaccine & virus variant from observational studies. Points have been jittered along the x-axis to facilitate presentation. Lines (and 95% CI ribbons) show fitted logistic regression models (A: logit(VE_hosp.) = 3.35+0.397*log2(Neut Ab); P<1×10-16; B: logit(VE_hosp.) = 1.74+0.708*log2(Neut Ab); P<1×10-16 ; C: logit(VE_hosp.) = 1.76+0.542*log2(Neut Ab); P<1×10-16).
Estimated relative risk (1-VE) for five endpoints, two vaccines, and three variants for three vaccine statuses (shortly after vaccination(two doses), 6 months (Waned), and shorty after a 3rd dose (Boosted).
Protection against symptomatic disease showed even larger declines. VE against symptomatic disease for waned neutralizing antibody titers (which were ∼8-fold lower than titers in recently vaccinated individuals; (5)) with Pfizer or Moderna against Omicron were far lower than the range of our data and VE was near 0. For recently vaccinated or boosted vaccinees with Pfizer and Moderna, relative risk increased 4 to 10-fold; e.g. VE was 83.1% (95% CI: 82.9-83.2) against Delta for recently vaccinated Pfizer recipients but only 21.4% (95% CI: 17.2-26.4) against Omicron (Fig. 2B). Similar reductions in protection were apparent for documented infections, all infections, and transmission given infection (Table S3).
Discussion
The emergence and rapid growth of Omicron, a new variant of SARS-CoV-2 with numerous known and novel mutations in the spike protein, has created an urgent need to estimate the impact on protection from vaccines. The fastest way to assess immune escape is by measuring neutralizing antibody titers with sera from vaccinated individuals, but translating these data into estimates of VE across variants had yet to be done. We have shown that ratios of normalized neutralizing antibody titers (normalized to WT virus) among virus variants are moderately repeatable, despite single estimates of normalized neutralizing antibody titers vary over an order of magnitude. We also showed that VE vs hospitalization, symptomatic disease, and documented infection were strongly correlated with neutralizing antibody titer ratios across a 30 to 100 fold range of neutralizing antibody titers, but we had to use variation across both vaccines and variants to span a sizeable range of neutralizing antibody titer.
We used these relationships to estimate VE for Omicron for vaccinees under a range of states (recently vaccinated, waned and boosted). The 40-fold estimated reduction in neutralizing antibody titers for Omicron increased the relative risk of hospitalization four to five-fold and increased the risk of symptomatic disease seven to ten-fold for mRNA vaccinees. In contrast, neutralizing antibody titers for individuals that were vaccinated with mRNA vaccines >4 months ago will have waned ∼8-fold (5) and would now be so low for Omicron (∼320-fold lower than recent vaccinees against WT) putting them outside the range of our data, and indicating very substantial reductions in VE for hospitalization, and nearly zero protection against symptomatic disease, and documented infection. Boosting neutralizing antibody titers with a third dose of mRNA vaccines increased titers substantially, especially compared to waned levels (5), and restored protection back to waned levels against the Delta variant.
These results should be interpreted in the context of several important caveats. First, and foremost, neutralizing antibody titers are a surrogate for VE, and despite strong correlations between VE and neutralizing antibody titers (4-7), direct studies of VE are critical to determine the full extent of immune evasion of new variants. Specifically, our analyses assume that differences in disease severity of variants are correlated with differences in neutralizing antibody titers. Whether this will hold across the 40-fold reduction in neutralizing antibody titers is highly uncertain. Second, the extent of Omicron immune evasion is so large relative to previous variants (40-fold vs 7-fold for Beta), and VE estimates for the next most evasive variant, Beta, were so uncertain that we were unable to estimate vaccine-specific models to estimate VE for Omicron. Instead we assumed that differences in neutralizing antibody titers across vaccines and variants are suitable for estimating VE across variants. Third, the 40-fold reduction that we used for Omicron required extrapolation well outside of available data and prevented us from providing reliable VE estimates for Omicron for vaccinees with waned immunity.
The Omicron variant of SARS-CoV-2 has grown rapidly in many countries. We have shown that the approximately 40-fold reduction in neutralizing antibody titers greatly erodes vaccine protection and increases the relative risk of infection and symptomatic disease more than four-fold and hospitalization two-fold. As a result, this variant is likely to spread much more quickly than the Delta variant, especially in highly immunized populations. The most promising tool to mitigate this reduction in protection is third doses.
Data Availability
All data produced in the present work are contained in the manuscript
Supplemental Tables
Footnotes
Additional statistical analyses and figures added.
