key: cord-0293951-zar31gy1 authors: Netzl, Antonia; Tureli, Sina; LeGresley, Eric; Mühlemann, Barbara; Wilks, Samuel H.; Smith, Derek J. title: Analysis of SARS-CoV-2 Omicron Neutralization Data up to 2021-12-22 date: 2022-01-07 journal: bioRxiv DOI: 10.1101/2021.12.31.474032 sha: 770cf8743a04f8291416c6601b8607c378a2503e doc_id: 293951 cord_uid: zar31gy1 The rapid spread of the Omicron SARS-CoV-2 variant (B.1.1.529) resulted in international efforts to quickly assess its escape from immunity generated by vaccines and previous infections. Numerous laboratories published Omicron neutralization data as preprints and reports. The understandable limitations and variability in such rapid reporting of early results however made it difficult to make definitive statements about the data. Here, we aggregate and analyze Omicron neutralization data from 23 reporting laboratories up to 2021-12-22. There are enough data to identify multiple trends and make two definitive points. First, in twice-vaccinated individuals, titer fold drop of Omicron relative to wild type is more than 19x, likely substantially more given the number of measurements below the limit of detection of the assay. Second, out to one month post third vaccination with an mRNA vaccine, or twice vaccinated after an earlier infection, the titer fold drop to Omicron is substantially less at approximately 7x. This substantially lower fold drop and somewhat higher titers after 3rd vaccination are strong early evidence for the utility of booster vaccination. The Omicron variant (B.1.1.529) was first reported to WHO on November 24, 2021 and has been spreading quickly. To prepare for a new COVID-19 wave caused by Omicron, it is important to get an early read on its ability to escape immunity acquired through vaccination, factoring in different vaccine types and vaccination strategies. Multiple groups have quickly produced data with serum sets and variants to hand and released them, mostly as preprints or preliminary reports, for public use. In this manuscript we analyze the available data to identify trends in Omicron's escape across laboratories and assays which might be explained by infection and vaccination history. We present our results as forest-plot based visual analysis to facilitate their joint interpretation. The aggregated data as basis for our analysis is available as a publicly accessible google sheet document 1 . We have analyzed Omicron virus neutralization data from 23 laboratories which are either in preprint form or are otherwise in the public domain. These data include neutralization of Omicron by different vaccine sera and sera of individuals infected with the wild type, Alpha, Beta or Delta variant. A variety of neutralization assays and cell types were used by the different laboratories (Table 1) . Fold drops in neutralization of Omicron compared to wild type in different vaccine sera are shown in Figure 1 , grouped by serum type and ordered by decreasing fold drop. We categorize the data across studies into serum groups by their infection or vaccination history. In the "2x Vax" group (n=44) we include double vaccinated individuals, independent of vaccine type, and single dose Johnson & Johnson (J&J) vaccinated individuals as a single J&J dose is the recommended vaccination regime. The convalescent sera ("Conv") group (n=21) contains sera from individuals that were infected with a SARS-CoV-2 variant, the "Inf + 2 x Vax" (n=9) sera from individuals that were vaccinated after being infected. Similarly, the "2x Vax + Inf" group (n=4) contains sera from individuals that were infected subsequently to their vaccinations. FInally, the "3x Vax" group (n=19) consists of triple vaccinated sera. The largest serum group consists of double vaccinated or J&J single dose vaccinated individuals. This twice-vaccinated regroup has the widest spread and largest uncertainty in fold drop of Omicron neutralization compared to wild type. We find an approximately 19x fold drop in the double vaccinated group. However, the majority of fold drops in double vaccinees are likely greater than the point estimate given here due to many Omicron titers being below the limit of detection for the assay. Further, fold drops less than 2 are often the result of low titers against the reference antigen, as in Balazs 19 J&J recently vaccinated. The average fold drop in the double vaccinated group is thus likely substantially greater than the mean of 19x seen here. There is a strikingly different pattern in the group of triple vaccinated individuals. Here, almost all Omicron titers were detectable and fold drops lower and more narrowly distributed than in double vaccinees. The average fold drop from wild type (WT) is 7x. The majority of sera from three times vaccinated individuals were, however, taken within one month of the last vaccination and thus do not provide information of how these titers will develop over time. The infected + double vaccinated cohort is closest to the triple vaccinated in spread and mean fold drop, with the majority of Omicron titers being above the detection threshold and an average fold drop from wild type of 12x. Data from studies that examined the neutralization reduction in vaccinated and later infected individuals suggest a similar pattern. Noteworthy, the study by Figure 1 : Omicron fold drops relative to wild type. Arrows indicate uncertainties in the point estimate due to titers below the limit of detection (LOD) of the assay. A short arrow marks measurements with less than half of Omicron titers below the assay's limit of detection (LOD), or conversely reference antigen titers at or higher than the LOD. Long arrows mark measurements with more than approximately 80% of Omicron titers below the LOD. Light blue dots show NIH SAVE laboratories, gray dots mark data points for which the reference antigen was not stated in the manuscript and is here assumed to be Wu-1. The solid vertical line marks no fold change. In addition to fold drops relative to wild type titers, we report Omicron titers obtained by applying the fold drop shown in Figure 1 to the wild type GMTs. The point estimates grouped by serum type and ordered by decreasing wild type titer are presented in Figure 2 . It might be that the substantially lower fold drop in the three times vaccinees compared to two times vaccinees is because higher titers in three times vaccinees are being underreported, either by laboratories not titrating to the endpoint, or because of a high-titer non-linearity in the assay something discussed in multiple forums. It can be clearly seen however from the data as presented in Figure 2 that in the triple vaccinated cohort, the fold drop from wild type to Omicron is independent of titer magnitude against wild type as evident by horizontal bars of similar length between WT and Omicron point estimates. Thus the substantially lower fold-drop in three times vaccinees is likely real. On average, the highest titers against both WT and Omicron are recorded in the infected + vaccinated serum group (estimated WT GMT: 5674, Omicron GMT: 464). In the triple vaccinated group, mean titer estimates against WT and Omicron are at 40% and 70% of the mean values in the Inf + 2x Vax cohort, respectively (WT GMT: 2238, Omicron GMT: 332). In both serum groups, the majority of Omicron titers are above the assay detection threshold and hence Omicron titer estimates are largely reliable. In contrast, due to many Omicron titers below the detection threshold in the remaining groups, the point estimates and thus also the mean Omicron titers are likely overestimated. Although WT titers in the group of vaccinated + infected sera were well detectable (WT GMT: 1583), due to below threshold titers the average Omicron titer of 57 is likely higher than the real value. Similar patterns of high WT titers but barely detectable titers against Omicron occur in the convalescent (WT GMT: 647, Omicron GMT: 28) and 2x Vax chorts (WT GMT: 491, Omicron GMT: 26). In the 2x Vax cohort, WT titers after a single J&J dose or double SputnikV, CoronaVac tend to be lower than after two vaccinations with AstraZeneca. The mRNA vaccine sera generally exhibit the highest WT titers in this group. Further, we see that Liu 14 's ICU samples (2x Vax + Infected Omicron titers and small blue dots indicate wild type titers. Omicron titers were obtained by applying the fold drop given in Figure 1 to wild type titers, corresponding to the horizontal bar connecting wild type and Omicron point estimates. Arrows indicate uncertainties in the point estimate. A short arrow marks measurements with less than half of Omicron titers below the assay's limit of detection (LOD), or conversely reference antigen titers at or higher than the LOD. Long arrows mark measurements with more than approximately 80% of Omicron titers below the LOD. Dashed lines mark thresholds of protection against symptomatic disease after vaccination with two doses of Moderna (a,d,f,g) 28 or AstraZeneca (b,c,e,f) 27 To estimate the protection against symptomatic disease, we compare the titers here against correlates of protection against symptomatic disease as determined by pseudovirus neutralization studies after two doses of AstraZeneca 27 or Moderna 28 . It must be noted however that it is not yet clear that VE estimates from these studies will apply to Omicron. Figure 3 shows titer estimates ordered by decreasing Omicron titers. The mean Omicron titer estimate in the Inf + 2x Vax group suggests protection against symptomatic disease is >80% for Omicron as reported after two doses of AstraZeneca 27 (80% VE at ID50=185) or Moderna 28 (91% VE at ID50=100). In this group, the lowest Omicron titer would still confer 70% protection. In the triple vaccinated group, titer estimates within 1 month of the third dose for both WT and Omicron are high enough to suggest >80% protection against symptomatic disease. All Omicron titers, except for the Chen 10 3xCoronaVac (BBIBP-CorV) estimate, are above the 70% VE threshold when we disregard uncertainties in the point estimates. As a result of large uncertainties and the question whether VE estimates apply to Omicron, we refrain from inferring correlates of protection in the remaining groups. When comparing pseudovirus (PV) and live-virus (LV) neutralization assays, we see that laboratories that used pseudovirus assays to determine Omicron neutralization reported almost Omicron titers estimated against other variants than WT can be found in Extended Data Figure 7 . In the supplement, we further show the titer estimates grouped by alternative strategies, including vaccine type, assay type and cell type ordered by Omicron and WT titer (Extended Data Figures 5-9 ). Two definitive statements can be made from the aggregation of the early data on Omicron virus neutralization data. Sera from individuals who have been vaccinated twice or infected once show generally more than a 19x fold drop of titers, whereas people who have been vaccinated three times or have been vaccinated + infected show an average of approximately 7x fold drop of titers. This reduced titer drop in three times vaccinated individuals appears to be real and not an artifact of an upper limit of detection of the assay. Almost all of the data in the three times vaccinated group are, however, from sera taken within 1 month of the last vaccination, whether this reduced titer drop will persist over time is yet to be determined. Nevertheless, the substantially lower fold drop and somewhat higher titers after a third vaccination are strong early evidence for the utility of booster vaccination at increasing virus neutralization titers against Omicron, and thus potentially at increasing vaccine efficacy. Some of the data we use are estimates as numerical data or individual repeat data is not available at this early stage. In such cases we extracted individual data points from figures using Webplotgitizer 3 and this will lead to some inaccuracies. Censored titers below an assay's detection threshold further increase these uncertainties. The fold drops for the twice vaccinated group, for example, is likely substantially greater than the 19x numeric estimate. However, a deflation of fold drops due to wild type titers at or above an assay's limit of detection is unlikely, as visible in Figure 2 . The differences between Omicron and WT titers are roughly consistent across studies in the triple vaccinated and infected + vaccinated cohort and are not influenced by the absolute magnitude of WT titers. In terms of titer and fold drop estimates, the triple vaccinated group is more similar to the infected + vaccinated than to the double vaccinated group. It needs to be noted that in these studies the time point of most recent vaccination was less than 3 months prior to serum collection, in the majority of cases 1 month prior to collection. How titers against Omicron develop over time is yet to be seen.. While we did not see a difference in fold drops from WT to Omicron neutralization depending on assay type, the titers reported in studies that used pseudovirus neutralization assays exceeded titers measured by live-virus neutralization. Omicron titers in double vaccinees were on average 2 times higher in pseudovirus than in live-virus assays, and almost 3-fold higher in triple vaccinees. This could, however, be a consequence of the sera types investigated using the different assays. The majority of PV sera came from mRNA vaccinated individuals, whereas livevirus assay tested sera contained sera from J&J, SputnikV and CoronaVac vaccinated individuals which tend to induce lower titers than mRNA vaccines (Extended Data Figure 9 ). Omicron neutralization data from publicly available preprints, reports or tweets were collected and categorized according to assay type, vaccine and convalescent sera tested, and the presence of the R346K substitution in the spike in addition to the common set of Omicron spike substitutions. In most cases, datasets are named after the corresponding author. A full list of all studies considered is shown in Table 1, detailed metadata in Supplementary Tables 1-4 . We used numerical data on geometric mean titers (GMT) and Omicron titer fold drops to a reference antigen as stated in each preprint. In case of missing GMT data, GMTs were either directly extracted from the manuscripts' figures using Webplotdigitizer 2 or individual data points were extracted by the same method and GMTs subsequently calculated with the meantiter R package 3 (method = "truncated_normal", dilution_stepsize = 0), which performs a Bayesian statistics analysis to correctly handle thresholded values. Thresholded titers for individual data were set to "