key: cord-0947339-qhe34dx5
authors: Layan, Maylis; Gilboa, Mayan; Gonen, Tal; Goldenfeld, Miki; Meltzer, Lilac; Andronico, Alessio; Hozé, Nathanaël; Cauchemez, Simon; Regev-Yochay, Gili
title: Impact of BNT162b2 vaccination and isolation on SARS-CoV-2 transmission in Israeli households: an observational study
date: 2022-03-03
journal: Am J Epidemiol
DOI: 10.1093/aje/kwac042
sha: 592e3fa95b4718e70bb1c699cdc3f492b7b65618
doc_id: 947339
cord_uid: qhe34dx5

Several studies have characterized the effectiveness of vaccines against SARS-CoV-2 infections. However, estimates of their impact on transmissibility remain limited. Here, we evaluated the impact of isolation and vaccination (7 days after the 2(nd) dose) on SARS-CoV-2 transmission within Israeli households. From December 2020 to April 2021, confirmed cases were identified among healthcare workers of the Sheba Medical Centre and their family members. Recruited households were followed up with repeated PCR for at least ten days after case confirmation. Data were analyzed using a data augmentation Bayesian framework. 210 households with 215 index cases were enrolled. 269 out of 667 (40%) susceptible household contacts developed a SARS-CoV-2 infection. Of those, 170 (63%) developed symptoms. Compared with unvaccinated and unisolated adult/teenager contacts, vaccination reduced the risk of infection among unisolated adult/teenager contacts above 12 (RR=0.21, 95% CrI credible interval 0.08-0.44) and isolation reduced the risk of infection among unvaccinated adult/teenager (RR=0.12, 95% CrI 0.06-0.21) and child contacts (RR=0.17, 95% CrI 0.08-0.32). Infectivity was reduced in vaccinated cases (RR=0.25, 95% CrI 0.06-0.77). Within households, vaccination reduces both the risk of infection and of transmission if infected. When contacts were unvaccinated, isolation also led to important reductions in the risk of transmission.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a highly transmissible virus that was first detected in Wuhan China in December 2019 (1, 2) . It is the cause of coronavirus disease 2019 (COVID- 19) , that has spread through the world, leading to a pandemic that has infected at least 250M people and caused more than 5M deaths worldwide by November 10, 2021 (3) . The advent of novel coronavirus disease 2019 (COVID- 19) vaccines has been an important breakthrough in the management of the pandemic. To determine how vaccination may modify epidemic dynamics, it is essential to estimate its effectiveness with respect to infection, transmission, and disease severity.

Multiple studies have shown that COVID-19 vaccines are effective at reducing both the risk of infection (4) (5) (6) (7) (8) , and the risk of developing severe symptoms (4, (8) (9) (10) in the general population.

Documenting vaccine impact on transmission is more challenging. This stems from the difficulty to thoroughly document chains of transmission and account for how different types of contacts may lead to different risks of transmission (11) . Households represent the perfect environment to evaluate factors impacting transmission such as vaccination because the probability of SARS-CoV-2 transmission among household members is high, ranging between 14 and 32% (12) (13) (14) . 

All HCW, regardless of their vaccination status, were required to use an electronic questionnaire to report daily any COVID-19 related symptom they, or a member of their household, had. SARS-CoV-2 PCR was readily available and HCW were encouraged to be tested for any mild symptom or suspected exposures. All HCW were instructed to notify the infection prevention and control unit if one of their household members was SARS-CoV-2 positive. All SARS-CoV-2 detected HCW as well as those with a positive SARS-CoV-2 household member were immediately contacted as part of the epidemiologic investigation for contact tracing and were provided with instructions regarding isolation precautions. All unvaccinated household members, i.e. those that did not receive the two vaccine doses at least seven days before the detection of the COVID-19 patient, were required to perform two PCR tests in the ten days after the diagnosis of the positive COVID-19 patient. Vaccinated household members were encouraged to perform two PCR tests during the ten days after detection. Household members were not required to test a second time if they had a positive test (Web Table 1 The household member who had the first positive PCR test was defined as the index case. When multiple household members had a positive PCR test on the same day, they were defined as co-index cases. We defined complete isolation as complete separation in sleeping and eating between household contacts and index case(s), i.e. they did not spend any time in the same room, and if a separate bathroom was provided for the index case(s). Partial isolation was defined if one of the above was violated but masks were continuously used, and eating was consistently separate.

For HCW, nasopharyngeal swabs were collected by trained personnel and RT-qPCR was performed using the Allplex™ 2019-nCov RT-qPCR assay (Seegene Inc., South Korea) and expressed by cycle threshold (Ct). Other household members reported the results of their COVID-19 test performed by their healthcare providers.

Confirmed SARS-CoV-2 infections were defined by a positive PCR test, i.e., with a Ct value lower than 40. Symptomatic cases were defined as confirmed cases with the presence of at least one symptom among the following ones: fever, cough, myalgia, headache, congestion, diarrhea, vomiting, anosmia or ageusia. Contacts who reported at least one of the above-mentioned symptoms but were

not confirmed because they performed no PCR test (n=6) or a single one at inclusion (n=2) were also considered as symptomatic cases. Asymptomatic cases were defined as confirmed cases who did not report any symptom over the follow-up period of the household.

We evaluated transmission in households using two metrics: the Secondary Attack Rate (SAR) defined as the proportion of susceptible household contacts that are infected after the index case is detected (22) , and the person-to-person probability of transmission defined as the per-capita probability that an infected individual transmits to a susceptible household contact. The first metric includes tertiary (i.e. household contacts infected by a household member that is not the index case) and community cases (i.e. household contacts infected in the community) contrary to the second metric. In both cases, we assumed that individuals who reported past infection of SARS-CoV-2 confirmed by PCR over the year preceding the detection of the household index case (n=20) were protected from infection and therefore, did not count as susceptible household contacts.

Baseline characteristics of the index cases and household contacts were described according to their vaccination status. All individuals above 12 years old are considered as adults/teenagers. We calculated the SAR for different categories of household contacts: unisolated and unvaccinated adults/teenagers, unisolated and vaccinated adults/teenagers, isolated and unvaccinated adults/teenagers, vaccinated and isolated adults/teenagers, unisolated children, and isolated children.

Here, isolation corresponds to complete or partial isolation between household contacts and the index case. We also defined the SAR of vaccinated and unvaccinated index cases as the proportion of infected household contacts in households with vaccinated or unvaccinated index cases, respectively.

In a sensitivity analysis, the SAR calculation was restricted to households in which a single index case was identified (Web Table 2 in Web Appendix 3). We also report the 95% Confidence Interval (CI) of the SAR.

We developed a statistical model to evaluate the effect of age, isolation precautions, BNT162b2

vaccination and household size on SARS-CoV-2 transmission dynamics in households (Web Appendix 4). The model uses the sequence of symptom onset dates and positive molecular test dates

to estimate the person-to-person risk of transmission within the household while accounting for the community hazard of infection (i.e., household contacts infected outside the household) and the possibility of tertiary transmissions (i.e., household contacts infected by a member of the household that is not the index case) (23) . The person-to-person risk of transmission is decomposed into the baseline person-to-person risk of infection that depends on household size, the relative infectivity of the infector that depends on their vaccination status (reference group: unvaccinated cases), and the relative susceptibility of the infectee that depends on their age, isolation behavior, and vaccination status. The relative susceptibility is estimated separately for unisolated children, isolated children, isolated and unvaccinated adults/teenagers, unisolated and vaccinated adults/teenagers, and adults/teenagers that are both isolated and vaccinated, considering the group of adults/teenagers that are unisolated and unvaccinated as the reference group. None of the children were vaccinated at the time of the study. This formulation accommodates for the potential confounding effects between the three variables characterizing household contacts (i.e. being vaccinated, isolated or being a child). We assumed that individuals whose isolation behavior was missing (n=6) did not comply with isolation precautions.

Model parameters were estimated using a Bayesian Markov Chain Monte Carlo sampling with data augmentation (23) (Web Appendix 5). Data were augmented with the probable date of infection of confirmed cases. For symptomatic cases, the date of infection was reconstructed from the date of symptom onset, using the probabilistic distribution of the incubation period (24) . For asymptomatic cases, we assumed that the date of infection could occur up to ten days prior to their molecular detection based on a meta-analysis (25) .

Since the study was conducted during the vaccine rollout, participants were enrolled at varying stages of their vaccination process. We assumed that vaccines reach their full effect seven days after receiving their 2 nd dose (4, 9, 10 

estimates changed when the analysis was restricted to households in which all negative contacts had performed at least one or two PCR tests in the ten days following the detection of the index case. In the baseline scenario, we assumed that asymptomatic cases are 40% less infectious than symptomatic cases based on a meta-analysis (26) , and investigated whether assuming the same level of infectivity in asymptomatic and symptomatic cases modified our estimates. Finally, in our baseline analysis, we chose a log-normal with log-mean=0 and log-standard deviation=1 prior distribution for the relative infectivity and relative susceptibility parameters and explored smaller and larger values (log-standard deviation=0.7, 2) in a sensitivity analysis.

We compared the observed and expected distributions of the number of cases per household size to assess the goodness-of-fit of the model (Web Table 3 in Web Appendix 6). We report the posterior median and the 95% Credible Interval (CrI) of estimated parameters. We also report the posterior probability that isolated and vaccinated adult/teenager contacts are less susceptible than vaccinated adult/teenager contacts that do not isolate. To measure the strength of evidence of a reduced susceptibility in isolated contacts among vaccinated ones, we report the associated Bayes Factor (BF). Here, it directly corresponds to the posterior odds of a reduced susceptibility in isolated contacts among vaccinated ones. Additional details are available in Web Appendix 1-6.

The study was approved by the Sheba Medical Center IRB committee (approval #8130-21). (Table 3 ). This proportion dropped to 28% (11 out of 40) among those who were unisolated and vaccinated, 29% (71 out of 245) among those whot were isolated but unvaccinated, and 11% (9 out of 83) among those who were isolated and vaccinated. 65% (66 out of 101) of child contacts that were unisolated got infected by SARS-CoV-2. This proportion declined to 33% (29 out of 87) for isolated child contacts. The proportion of asymptomatic cases varied from 26% (46 out of 174) among adult/teenager contact cases to 56% (53 out of 95) among child contact cases.

The SAR also varied with the vaccination status of the index case regardless of the contacts' characteristics. Among the 622 household contacts whose index case was unvaccinated, 261 (42%) developed a SARS-CoV-2 infection (Table 3 ). This proportion dropped to 19% (8 out of 42) among household contacts whose index case was vaccinated. Finally, the SAR was relatively invariant with

household size: 31%, 40%, 32%, and 32% for households of size 2 to 5, respectively (Web Figure 1 in Web Appendix 6).

Our statistical model makes it possible to perform a multivariate analysis of the drivers of SARS-CoV-2 transmission in households. We estimate that, relative to adult/teenager contacts who were unisolated and unvaccinated, the relative risk of being infected was 0.21 (95% CrI 0.08-0.44) among adult/teenager household contacts who were vaccinated but unisolated (Figure 2A , Web Figure 2B , Web Table 4 in Web Appendix 7).

Overall, we estimate that, in a household of size 4, the person-to-person probability of SARS-CoV-2 transmission is 61% (95% CrI 48-72) between an unvaccinated case and an unvaccinated and unisolated adult/teenager. This probability drops to 4% (95% CrI 1-16) between two vaccinated adults/teenagers who do not follow isolation rules (Figure 3 , Web Table 5 in Web Appendix 7). The person-to-person probability of transmission from an unvaccinated case to a child who does not isolate is 37% (95% CrI 27-48). This probability drops to 11% (95% CrI 3-31) if the case is vaccinated and to 14% (95% CrI 7-25) if the child contact is isolated.

In general, our estimates of relative susceptibility and relative infectivity were robust to model assumptions ( Figure 4 ). When the analysis was restricted to households in which all contacts performed at least one or two PCR tests in the ten days following the recruitment of the index case, the relative susceptibility of vaccinated adult/teenager contacts who did not isolate was slightly higher Table 4 in Web Appendix 7). In the alternative scenarios, the number of individuals included was substantially lower, increasing CrIs (Web Figures 2 and 3 , Web Tables 6-9 in Web Appendix 8). Similarly, the relative susceptibility of vaccinated adult/teenager contacts who did isolate increased from 0.07 (95% CrI 0.03-0.16) in the baseline scenario to 0.12 (95% CrI 0.04-0.28) in the analysis with at least one PCR, and 0.13 (95% CrI 0.04-0.32) in the one with at least two PCR tests. Consequently, the posterior probability that isolated and vaccinated adult/teenager contacts were less susceptible than vaccinated adult/teenager contacts that did not isolate dropped from 96% to 88% with one PCR and 89% with two PCR tests. Still, the statistical support was high with a BF equal to 7 and 8, respectively. Relative infectivity and relative susceptibility were slightly sensitive to their prior distribution (Web Table 10 in Web Appendix 8).

When the log-standard deviation increased, estimates were pulled towards lower values.

We evaluated the impact of BNT162b2 vaccination on case infectivity and the mitigating effect of age, isolation from the index case, and BNT162b2 vaccination on susceptibility to infection in household settings. Our approach accounts for infections in the community, potential tertiary infections within the households, the reduced infectivity of asymptomatic cases, potential misidentification of the index case(s), and varying follow-up periods between households.

In our analysis, the SAR in unvaccinated adult/teenager contacts who did not isolate was estimated at around 76%, which is substantially higher than previous estimates obtained in household settings (12) (13) (14) 18, 27, 28) . In meta-analyses (12) (13) (14) , the average SAR ranged between 14% and 32%;

however, in some studies, it could be as high as 90% (13) . Most of these studies date back to the time when historical lineages were still dominant. In contrast, our study took place when the Alpha variant represented up to 90% of infections in Israel (21) . Our higher estimate could be at least partly explained by the fact that the Alpha variant is substantially more transmissible than historical lineages (21, (29) (30) (31) .

In agreement with previous reports, we found that children are less susceptible to SARS-CoV-2 infections than adults/teenagers (12) (13) (14) 32) . We further estimated that, seven days after their second dose, vaccinated adults/teenagers benefit from a 79% reduction in the risk of infection compared to unvaccinated adults/teenagers. Consistently with previous studies (21, 33) , we show that BNT162b2 vaccination is highly effective against infection by the Alpha variant. In general population studies, vaccine effectiveness for symptomatic infections ranged from 57% 14 days after the 1 st dose (4) to 89% (4), and 97% 7 days after the 2 nd dose (9) . For asymptomatic infections, vaccine effectiveness against infection was 79% ten days after the 1 st dose (5) , and 94% 14 days after the 2 nd dose (7). Our estimate of vaccine effectiveness in household settings is lower than those obtained in the general population. This is consistent with estimates obtained in households (19, 20, 33) (18) (19) (20) .

To our knowledge, this is the first study estimating the effect of isolation on SARS-CoV-2 transmission in households that are partially vaccinated. We showed that isolation precautions markedly reduce the overall infection risk in both adult/teenager and child contacts even when considering partial physical distancing measures. We estimated a similar reduction of infection in adult/teenager contacts that were vaccinated but did not isolate. There was a signal in the data that isolation also benefited vaccinated individuals although credible intervals were larger and further investigations are required to confirm this finding.

Our study has several limitations. First, household studies such as ours may be affected by multiple sources of bias. On the one hand, we may overestimate the SAR if we are more likely to detect households with multiple cases. On the other hand, we might underestimate it if some asymptomatic, or pauci-symptomatic cases are missed during follow-up. Second, we estimated an important reduction of infectivity in vaccinated cases with 2 doses compared to unvaccinated cases as

previously shown (18) (19) (20) 34) . However, this is associated with important uncertainty due to the small number of cases (15 vaccinated index cases, and 21 vaccinated secondary cases). Thus, more data are needed to reduce the size of credible intervals. Third, we assumed an all-or-nothing effect of the vaccine that started seven days after the second dose (or 15 days after the first dose in our sensitivity analysis, Web Table 11 in Web Appendix 8). In practice, the effect of the vaccine is likely to be progressive, which might push down estimates of effectiveness since individuals with early partial protection would be considered as unvaccinated. However, excluding households with the earlyvaccinated index cases did not impact our estimates (Web Figure 4 and Web Table 12 in Web Appendix 8). The limited number of households does not make it possible to dissociate early vs full protection conferred by the vaccine nor to investigate the infectivity of children relative to adults/teenagers. Fourth, testing instructions were different for vaccinated and unvaccinated household contacts, as well as HCW and non-HCW. Most vaccinated contacts were HCW at the Sheba Medical Center who complied with testing instructions to go back to work, leading to high testing rates in vaccinated individuals with 67% having at least two PCR tests and 70% having one positive PCR or at least two PCR tests in the ten days following case detection (Web Table 1 in Web Appendix 1). Among unvaccinated contacts, 49% had at least two PCR tests and 79% had one positive PCR or at least two PCR tests in the ten days following case detection. This higher testing rate is notably due to the high proportion of single positive tests (30%). These differential testing behaviors and positivity rates between vaccinated, unvaccinated, HCW, and non-HCW contacts make it difficult to anticipate the directionality of a potential bias. When restricting our evaluation to households where all negative contacts were tested at least once or twice, estimates remained relatively similar to the baseline ones. In the analysis with at least two tests for all negative contacts, we observed a slight reduction in the point estimate for vaccine effectiveness against infection that remained difficult to interpret given the very broad credible intervals (17%-91%). Fifth, the measurement of isolation precautions may be subjected to recall bias and/or overreporting, as they represent a socially desirable behavior. The timing and evolution of isolation precautions were not measured, and thus, not integrated in our model. Nevertheless, our estimate of isolation effectiveness is consistent with a 10-day period quarantine in modelling studies (35) . In the baseline scenario, we assumed that vaccination was effective from 7 days after the 2 nd dose, the relative infectivity of asymptomatic cases compared to symptomatic cases was equal to 60% and the log-standard deviation of the relative infectivity and relative susceptibility prior distributions was equal to 1. The posterior median and its associated 95% Bayesian credible interval are reported. 

Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding

COVID-19: A review of therapeutic strategies and vaccine candidates

A global database of COVID-19 vaccinations

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

Vaccine on Asymptomatic Infection Among Patients Undergoing Preprocedural

FDA-authorized mRNA COVID-19 vaccines are effective per real-world evidence synthesized across a multi-state health system

Real-World Evidence Confirms High Effectiveness of Pfizer-BioNTech

Vaccine and Profound Public Health Impact of Vaccination One Year After Pandemic Declared

COVID-19-Vaccine-and-Profound-Public-Health-Impact-of-Vaccination-One-Year-After-Pandemic-Declared.html)

Effectiveness of COVID-19 vaccines in preventing SARS-CoV-2 infection and hospitalisation

Impact and effectiveness of mRNA BNT162b2 vaccine against SARS-CoV-2 infections and COVID-19 cases, hospitalisations, 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

Epidemiology and transmission of COVID-19 in 391 cases and 1286 of their close contacts in Shenzhen, China: a retrospective cohort study

Household Transmission of SARS-CoV-2

Household transmission of COVID-19-a systematic review and meta-analysis

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Setting-specific Transmission Rates: A Systematic Review and Meta-analysis

World Health Organization (WHO)

Effect of Vaccination on Transmission of SARS-CoV-2

Effect of Vaccination on Household Transmission of SARS-CoV-2 in England

Vaccination with BNT162b2 reduces transmission of SARS-CoV-2 to household contacts in Israel

Vaccine effectiveness against SARS-CoV-2 transmission and infections among household and other close contacts of confirmed cases, the Netherlands

BNT162b2 vaccination effectively prevents the rapid rise of SARS-CoV-2 variant B.1.1.7 in high-risk populations in Israel

Secondary attack rate and superspreading events for SARS-CoV-2

A Bayesian MCMC approach to study transmission of influenza: Application to household longitudinal data

Incubation period of COVID-19: A rapid systematic review and meta-analysis of observational research

SARS-CoV-2, SARS-CoV, and MERS-CoV viral load dynamics, duration of viral shedding, and infectiousness: a systematic review and metaanalysis. The Lancet Microbe

Estimating the extent of asymptomatic COVID-19 and its potential for community transmission: Systematic review and meta-analysis

Household secondary attack rate of COVID-19 and associated determinants in Guangzhou, China: a retrospective cohort study

Household transmission of SARS-CoV-2 and risk factors for susceptibility and infectivity in Wuhan: a retrospective observational study

Viral dynamics of SARS-CoV-2 variants in vaccinated and unvaccinated individuals

Effects of non-pharmaceutical interventions on COVID-19 cases, deaths, and demand for hospital services in the UK: a modelling study

Assessing transmissibility of SARS-CoV-2 lineage B.1.1.7 in England

Susceptibility to SARS-CoV-2 Infection Among Children and Adolescents Compared With Adults

Impact of vaccination on new SARS-CoV-2 infections in the United Kingdom

Decreased infectivity following BNT162b2 vaccination: A prospective cohort study in Israel

Quantifying the impact of quarantine duration on covid-19 transmission

The impact of SARS-CoV-2 vaccination on Alpha and Delta variant transmission

World Health Organization. Weekly epidemiological update on COVID-19 -13

Worl Health Organization