key: cord-0428695-c1urxa4r
authors: Huppert, A.; Mor, O.; Zuckerman, N. S.; Hazan, I.; Fluss, R.; Ash, N.; Ginish, N.; Mendelson, E.; Alroy-Price, S.; Freedman, L. S.
title: BNT162b2 Vaccination efficacy is marginally affected by the SARS-CoV-2 B.1.351 variant in fully vaccinated individuals.
date: 2021-07-23
journal: nan
DOI: 10.1101/2021.07.20.21260833
sha: 89058e5fd3b4e8440828877aa2ea96234a9bd60a
doc_id: 428695
cord_uid: c1urxa4r

Background Israeli has vaccinated over 80% of its adult population, with two doses of the Pfizer BNT162b2 vaccine. This intervention has been highly successful in curtailing the coronavirus 2 outbreak. One major concern is the ability of the virus to mutate which potentially can cause SARS-CoV-2 to partially escape from the immune system. Here we evaluate the efficacy of the Pfizer vaccine against the B.1.351 variant. Methods The Ministry of Health, initiated sequencing of selected positive swab samples identified as being of interest. We used logistic regression, with variant type as the dependent variable, vaccination status as the main explanatory variable, controlling for age, sex, subpopulation, place of residence and time of sample, to estimate the odds ratio for a vaccinated case to have the B. 1.351 versus the B.1.1.7 variant, within vaccinated and unvaccinated persons who tested positive. Findings There were 19 cases of B.1.351 variant (3.2%) among those vaccinated more than 14 days before the positive sample and 88 (3.5%) among the unvaccinated. The estimated odds ratio was 1.29 [95% CI: 0.66-2.50]. From this result, assuming the efficacy against the B.1.1.7 variant to be 95%, the estimated efficacy against the B.1.351 variant was 94% [95% CI: 87-97%]. Interpretation Despite the concerns caused by the B.1.351 variant, the BNT162b2 vaccine seems to provide substantial immunity against both that variant and the B.1.1.7. Our results suggest that from 14 days following the second vaccine dose the efficacy of BNT162b2 vaccine is at most marginally affected by the B.1.351 variant. Funding No funding

The impressive success of mass vaccination on curtailing the coronavirus 2 (SARS-CoV-2) by halting transmission has paved the road for returning to "pre-pandemic life". However, several open questions challenge the triumph of controlling/ending the pandemic by vaccination. One major concern is the ability of the virus to mutate and evolve; this potentially can cause SARS-CoV-2 to partially escape from the immune system, which will reduce the effectiveness of the vaccine in preventing disease and viral transmission. Determining vaccine efficacy against variants of concern (VOC) 1 is vital for planning and modifying vaccination strategies.

On December 19 th 2020, Israel launched a massive COVID vaccination campaign based on the Pfizer BNT162b2 vaccine, and by end of May 2021, had administered over 10,500,000 doses, to approximately 5,400,000 individuals, more than 80% of the population over 16y, receiving two doses. Both in clinical trials and in real world studies, the BNT162b2 vaccine has proven to be highly effective in both averting infections and preventing severe disease and death. [2] [3] [4] [5] The Israeli vaccination campaign took place during the third and largest wave of the pandemic (see Figure 1 ). During this third wave, the B.1.1.7 variant became the dominant strain in Israel, reaching over 95% dominance. 6 Since the detection of the B.1.1.7 variant in November 2020 in the United Kingdom, it has spread rapidly and become the dominant strain in many countries.

There is also evidence that it causes higher rates of morbidity and mortality. 7 Nevertheless, the BNT162b2 vaccine, which was developed based on the original Wuhan strain sequence, has been found very effective against the B.1.1.7 variant, both in blocking transmission and reducing morbidity and mortality following infection. [2] [3] [4] [5] The B.1.351 strain, which was first documented in South Africa, is also considered a VOC mainly because in vitro experiments have demonstrated its ability to overcome previous immunity to SARS-CoV2. Specifically, experimental work demonstrated significant decrease in neutralization capacity of B.1.351. 8 However other research found that neutralizing antibodies remained sufficiently high against the B.1.351 variant. 9 Humoral protection measured by antibody responses and neutralization studies do not assess the role of cellular immunity mediated by T-cell responses. A recent study has shown that the cellular protection established following previous infection or vaccination remains high against both the B.1.1.7 and the B.1.351 variants. 10 On the other hand, two real world studies have raised concern that the BNT162b2 vaccine has reduced efficacy against the B.1.351 variant. A study from Qatar has shown that the efficacy of BNT162b2 against the B.1.351 variant was ~75% compared to ~90% against the B.1.1.7 variant. 11 A second study from Israel has estimated that the odds ratio (OR), in a matched study of SARS-Cov-2 cases occurring in unvaccinated persons versus persons who had received their second dose of vaccine at least one week previous to sample collection, was 1/8, implying considerably lower efficacy of the vaccine against the B.1.351 variant. 6 Thus, our goal was to further quantify the risk of the B.1.135 variant causing a significant breakthrough in a real world environment.

With the start of the vaccination campaign in Israel, the Central Virology Laboratory (CVL) of the Ministry of Health, initiated collection and sequencing of selected swab samples that had tested positive on polymerase chain reaction (PCR). Samples were selected for sequencing i) to monitor the circulating and imported variants in Israel ii) to characterize viral variants among cases . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 23, 2021. 

The information on all cases sent for sequencing was entered in a database, containing: sociodemographical information (e.g. age, city/town/village of residence, subpopulation -Arab, Ultra-Orthodox Jewish, other), date of collection of first positive sample, date of recovery, vaccination . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 23, 2021. ;

dates, symptoms and hospitalizations, and the infecting variant as determined by whole genome sequencing. All information was retrieved from the Israeli Ministry of Health's databases.

We restricted our analyses to vaccinated and unvaccinated cases that were positive either for the variant B.1.1.7 or the variant B.1.351 using whole genome sequencing. Vaccinated cases were defined as those where the first positive sample was taken at least 14 days after the second dose. Those with unknown dates of vaccination, or who received only one vaccine dose, or for whom the sample was taken between the first dose and second dose, were excluded.

Those for whom the sample was taken less than 14 days after their second dose were excluded from the main analysis, but were included in a secondary analysis. Individuals who had acquired the infection outside Israel, those aged less than 16y (not eligible for vaccination) and individuals without information regarding their place of residence, were all excluded. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 23, 2021. ;

From this equation, when OR equals 1, then VB.1.351 = VB.1.1.7. Thus, a test of the hypothesis that OR = 1, also tests whether the vaccine is equally effective against each variant.

We estimated the odds ratio by logistic regression 12 13 This approach also provided a p-value for the test that OR=1 and a 95% confidence interval for the OR.

A secondary analysis was conducted to examine the influence of time of infection following full vaccination, using the same methods as described above and comparing unvaccinated cases versus cases where the sample from the vaccinated case was taken within the first 14 days after the second dose of vaccination.

Alternative analyses were conducted using matching to control for confounding variables, and are reported in the Supplementary Information. Unlike the main analyses presented here, such matching entails exclusion of approximately 60% of the cases. We regard their results as providing information supportive to the main analyses.

The database contained the sequencing results of 11624 samples obtained from distinct individuals. After the exclusions described in the Methods section, 596 vaccinated and 2515 unvaccinated cases were left eligible for analysis ( Figure 2 ). Characteristics of these vaccinated and unvaccinated individuals are shown in Table 1 . The vaccinated groups were on average . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; https://doi.org/10.1101/2021.07.20.21260833 doi: medRxiv preprint older and had a smaller proportion of ultra-orthodox Jews, emphasizing the need to control for these potential confounders.

The distribution of variants (B.1.1.7 and B.1.351) by vaccination status is shown in Table 2 . There were 19 cases of B.1.351 variant (3.2%) among those vaccinated more than 14d before the positive sample and 88 (3.5%) among the unvaccinated. The estimated OR (Table 3 ) was 1.29, with a P-value of 0.46 and 95% confidence limits ranging from 0.66 to 2.50. Note that the estimated OR was larger than one even though the crude proportion of B.1.351 was slightly lower (3.2% v 3.5%) among the vaccinated cases. This was due to the regression adjustment for the confounders age and subpopulation -younger age and ultra-orthodox Jews both had a negative association with the B.1.351 variant. Using the estimated OR, assuming that the vaccine efficacy against the B.1.1.7 variant is 95% 2-5 , the estimated efficacy against the B.1.351 variant is estimated to be 93% with 95% confidence limits between 87% and 97%. Results from the supportive analysis when matching was employed gave an estimated OR of 0.90 (95%CI: 0.40-2.01) (see Supplementary Tables A1-A4 for details).

The same methods as above were applied to persons who received their second dose between 1 and 13 days before collection of the sample. In the 121 cases, there were 14 (11.1%) B.1.351 variants; the estimated OR was 2.62 (P=0.015, 95%CI = 1.20-5.70). A similar estimate was obtained from the supportive analysis using matching, but with a wider confidence interval (estimated OR = 2.12, 95%CI = 0.49-9.05).

The great success of the Israeli vaccination campaign against SARS-CoV-2 can be appreciated by the fact that, as of June 1 st , Israel lifted all COVID emergency restrictions except the requirement to wear masks indoors and regulations governing international travel.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; https://doi.org/10.1101/2021.07.20.21260833 doi: medRxiv preprint Variants can affect a vaccine's impact on the transmission of an infection by two factors. First, VOC may have a higher rate of transmission that will require a higher vaccination coverage or a greater vaccine efficacy to curtail the spread of infection. 14 Limitations of our study include the relatively low number of B.1.351 cases, owing to its low prevalence in Israel, and also the fact that the sequencing was not done on a sample selected randomly from the total population of SARS-Cov2 positive cases in Israel. In particular, samples . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; https://doi.org/10.1101/2021.07.20.21260833 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; https://doi.org/10.1101/2021.07.20.21260833 doi: medRxiv preprint 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. For supplementary statistical analysis:

Each matched set comprised a single vaccinated case matched to one or more (up to 10) unvaccinated cases. Matching on all of the following variables was required: city/town/village of residence, date of taking the swab sample (± 7d), and subpopulation (Arab, Jewish ultraorthodox, other). We estimated the odds ratio by conditional logistic regression , 17 with variant type as the dependent variable and vaccination status as the explanatory variable. The analysis . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021.

was implemented using the clogit procedure in the survival package of the R language. 18 Age group (16-44, 45-64, 65-79, ≥80y) was included instead as a covariate in the regression model. Table A1 , and characteristics of the vaccinated and unvaccinated individuals in these matched sets are shown in Table A2 .

Matching on the same variables as the main analysis, but also including age, yielded a total of 290 matched sets, with 290 vaccinated and 563 matched unvaccinated cases. Table A1 shows the distribution of the number of unvaccinated case matches per vaccinated case. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; https://doi.org/10.1101/2021.07.20.21260833 doi: medRxiv preprint Computer simulations to check on effect of selection on the estimate of the odds ratio

The simulation was written in the R language, and the code is given below. The computer generates a very large study based on sampling preferentially more vaccinated cases and more B.1.351 cases than are representatively found in the population of cases. However, the preferential selection of vaccinated cases is without reference to whether they have the B.1.351 or the B.1.1.7 variant, and the preferential selection of B.1.351 cases is without reference to whether they are vaccinated or unvaccinated. The selected study data are then used to estimate the odds ratio and this estimate is then compared to the "true odds ratio" that was used to generate the data.

In the following R code notation is as follows: p0 = probability of infection by B. for(i in 1:nsim) { vac<-rbinom(n,1,pvac) pdis0<-p0*(1-vac*v0) pdis1<-p1*(1-vac*v1) dis0<-rbinom(n,1,pdis0) dis1<-rbinom(n,1,pdis1) pselvac <-vac*svac1 + (1-vac)*svac0 pselvar <-dis0*svar0*(1-dis1) + dis1*svar1*(1-dis0) psel <-pselvar*pselvac sel <-rbinom(n,1,psel)

} c(sum(n00),sum(n01),sum(n10),sum(n11),sum(n11)*sum(n00)/(sum(n10)*sum(n 01)),sum(nsel00),sum(nsel01),sum(nsel10),sum(nsel11),sum(nsel11)*sum(ns el00)/(sum(nsel10)*sum(nsel01))) . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 23, 2021. ; https://doi.org/10.1101/2021.07.20.21260833 doi: medRxiv preprint } In the above program, the term sum(n11)*sum(n00)/(sum(n10)*sum(n01)) is the odds ratio in the population and estimates the ratio (1-v1)/(1-v0).

The term sum(nsel11)*sum(nsel00)/(sum(nsel10)*sum(nsel01)) is the odds ratio in the cases selected for screening.

The following table gives the results of running this program for different vaccine efficacies, and selection proportions. Comparing the final column of the table with the true odds ratio (3 rd column) or the estimated OR in the population, one can see that the estimated odds ratio in the selected sample appears to have little or no bias.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 23, 2021. ; https://doi.org/10.1101/2021.07.20.21260833 doi: medRxiv preprint Demonstration of the relationship between vaccine efficacy and the odds ratio in a matched case-case design

The following theory shows that estimating the odds ratio of having the B.1.351 variant among the vaccinated to unvaccinated cases will allow one to estimate the efficacy of the vaccine against the B. 

Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine

BNT162b2 mRNA Covid-19 Vaccine in a Nationwide Mass Vaccination Setting

Impact and effectiveness of mRNA BNT162b2 vaccine against SARS-CoV-2 infections and COVID-19 cases, hospitalizations, and deaths following a nationwide vaccination campaign in Israel: an observational study using national surveillance data

Protection of previous SARS-CoV-2 infection is similar to that of BNT162b2 vaccine protection: A three-month nationwide experience from Israel

Evidence for increased breakthrough rates of SARS-CoV-2 variants of concern in BNT162b2 mRNA vaccinated individuals

Increased mortality in community-tested cases of SARS-CoV-2 lineage B.1.1.7

Sensitivity of SARS-CoV-2 B.1.1.7 to mRNA. vaccine-elicited antibodies

Neutralizing Activity of BNT162b2-Elicited Serum -Preliminary Report

Negligible impact of SARS-CoV-2 variants on CD4 + and CD8 + T cell reactivity in COVID-19 exposed donors and vaccines bioRxiv

Effectiveness of the BNT162b2 Covid-19 Vaccine against the B. 1.1. 7 and B. 1.351 Variants

Online ahead of print

Applied Logistic Regression

Linear Mixed Effects Model using 'Eigen' and S4

Modeling Infectious Diseases in Humans and Animals

Escape of SARS-CoV-2 501Y.V2 from neutralization by convalescent plasma

The importance of time post-vaccination in determining the decrease in vaccine efficacy against SARS-CoV-2 variants of concern

The Analysis of Case-Control Studies

Modeling Survival Data: Extending the Cox Model