key: cord-0732477-2qoxuvrj
authors: Harris, Ross J.; Hall, Jennifer A.; Zaidi, Asad; Andrews, Nick J.; Dunbar, J. Kevin; Dabrera, Gavin
title: Effect of Vaccination on Household Transmission of SARS-CoV-2 in England
date: 2021-06-23
journal: N Engl J Med
DOI: 10.1056/nejmc2107717
sha: 07680a80023a814f14bc44dfcbb19058d0d8f68f
doc_id: 732477
cord_uid: 2qoxuvrj

nan

. Numbers of households, contacts of index cases, and secondary cases in the data, numbers excluded according to different criteria, total exclusions, and final analysis cohort.. 6 

The creation of the routine HOSTED dataset has been described in detail elsewhere (1) . In brief, laboratory confirmed cases of COVID-19 in England which are reported to national laboratory surveillance systems(2) are linked to individuals who share the same address, using National Health Service (NHS) number and the Unique Property Reference Number (UPRN). Known institutional settings (using UPRN information) such as care homes, prisons, and households with more than 10 residents are excluded.

For this analysis, the HOSTED dataset was linked to data from the National Immunisation Management System (NIMS)(3) to obtain information such as the dates and types of COVID-19 vaccinations for all individuals vaccinated in England.

HOSTED includes individual-level socio-demographic data for cases and household contacts, including age, sex, and Index of Multiple Deprivation (IMD); information on property type, confirmed cases using PCR-based SARS-CoV-2 through national reporting systems, and linked information on hospitalisation and mortality. The vaccination data include date and type (ChAdOx1 nCoV-19 or BNT162b2) of first and second doses for all vaccinated individuals in the HOSTED dataset.

We defined Index Cases as the earliest case of laboratory-confirmed COVID-19, by diagnosis date, for a household. Household Contacts were defined as all individuals with the same address as the index cases of COVID-19; and Secondary Cases as a known household contact of an index case with a positive SARS-CoV-2 test that has a specimen date between two and 14 days after the specimen date of the index case.

The analysis cohort included households with an index case occurring between 4th January 2021 to 28th February 2021, with 14 days observable follow up for all contacts. Households in which any individual was vaccinated prior to the 4 th January were excluded, so that our analysis would be as broadly generalizable as possible to the overall vaccination campaign. Households in which the index case was vaccinated 1-14 days after testing positive for COVID-19 were also excluded, as were all contacts who had been vaccinated prior to the index case testing positive. We excluded index cases tested under 'pillar 1' of the national testing strategy, which is a proxy for a case being either hospitalised or a health worker. This was because the household contacts of hospitalised cases are likely to have differential exposure profiles compared to contacts of non-hospitalised cases. Finally, we restricted analyses to households with a single index case age 16+, and no co-primary cases (any other cases on the same or next day as the index case).

We defined vaccinated index cases as having been vaccinated 21 days or more prior to testing positive for COVID-19 based on evidence of the time needed for the vaccine to provide a sufficient level of immunity (4) . Non-vaccinated index cases were defined as not having received a vaccine prior to testing positive. Households where the index case received the vaccine less than 21 days before testing positive were excluded from this analysis.

We compared household contacts of index cases receiving either the ChAdOx1 nCoV-19 or BNT162b2 vaccines, with contacts of unvaccinated index cases, and the proportion of contacts who tested positive within 2-14 days of the index case (secondary cases) in each group: unvaccinated (base group), BNT162b2, and ChAdOx1 nCoV-19.

The odds of being a secondary case was modelled using logistic regression, adjusted for the following covariates: age of index case and contact (ages grouped as 0-15, 16-29, 30-39, 40-49, 50-59, 60-69, 70-79, and 80+), sex of index case and contact, government office region (nine groups), calendar week of index case, index of multiple deprivation quintile (IMD) and household type. Household types were defined as in previous work(1) and further subdivided by size to give the following groups: pairs/couples of adults both age under 65; pairs/couples of adults with one or both 65+; households with children (<16) of size 2-4, 5-7 and 8-10; multi-generational households (at least one adult age 65+ and age gaps of at least 16 years between this adult and two younger household members) of size 3-4, 5-7 and 8-10; and 3+ adult-only households of size 3-4, 5-7 and 8-10.

We used robust standard errors in the final model to allow for dependence between individuals in the same household. Results are presented in terms of odds ratios (ORs) and 95% confidence intervals (CI). We also considered whether the effect of vaccination varied according to both the age of index case and contact, which was assessed via the inclusion of interaction terms. Models were compared via likelihood ratio tests.

In addition to a binary exposure (vaccinated 21+ days before becoming an index case vs. unvaccinated) we considered households in which the index case was vaccinated any time up to the date of testing positive, in whom there may be a partial effect on transmission. Timing of vaccination was grouped in two-to-three day intervals in the 0-24 days prior to testing positive, then at four, five, or more than six weeks prior to testing positive. The odds of being a secondary case according to vaccination timing (vs. unvaccinated index case) was modelled using the same logistic regression model described above, and we examined the timing according to calendar time of index case (weeks 1-2, 3-4, 5-6 and 7-8 in 2021).

Finally, we compared households in which the index case was vaccinated 21-35 days before testing positive with households in which the index case was vaccinated 1-10 days before testing positive. We used a similar multivariable logistic regression model to that used in the main analysis.

Data were extracted on 23 rd March 2021 and included individuals with at least 14 days of observable follow-up from the date of specimen collection from the index case (i.e. index cases between 4th January 2021 to 28th February 2021, secondary cases up to 14 th March 2021). There were 552,984 residential households of two to 10 people where there was at least one case. Table S1 shows the exclusions applied to arrive at the final analysis cohort.

The exclusion criteria did not overlap substantially, although contacts in households where the index case was vaccinated prior to 4 th January 2021 were more likely to have been vaccinated (28.6% vs. 3.7%). The final cohort consisted of 365,447 households with a single index case and 1,018,842 contacts. There were 4,107 households where the index case was vaccinated 21 days or more before testing positive (1.12%), and 20,110 where the index case was vaccinated less than 21 days before testing positive (5.51%). Table S2 shows characteristics of contacts according to vaccination status of the index case.

In all households, the majority of index cases and contacts were age <60, with a high proportion of individuals aged <40 in unvaccinated households. Over half of contacts were also aged <40 for all household groups. Male contacts were more common in households in which the index case had been vaccinated. As households where the index case had been vaccinated for 21+ days before being infected required a 3-week window, there are no households in this group with an index case between 4 th -24 th January; otherwise, index cases approximately follow the overall case rate, as described above. Distributions of IMD, region and household size were generally s imilar between the vaccinated and unvaccinated groups.

In households where the index case received the vaccine more than 21 days before testing positive, nearly two-thirds received BNT162b2. Table S3 shows the proportions of each vaccine received by index cases over time: in January, less than 10% of those vaccinated received the ChAdOx1 nCoV-19 vaccine, but this increased to around 40% by mid-February.

In households where the index case was not vaccinated before testing positive, there were 96,898 secondary cases out of 960,765 household contacts (10.1%). There were 196 secondary cases in 3,424 contacts (5.72%) where the index case received the ChAdOx1 nCoV-19 vaccine 21 days or more before testing positive, and 371 secondary cases in 5,939 contacts (6.25%) where the index case received the BNT162b2 vaccine 21 days or more before testing positive.

The unadjusted odds ratio for being a secondary case if the index case was vaccinated with ChAdOx1 nCoV-19 21 days or more before testing positive (vs. index case not vaccinated) was 0.55 (95% CI 0.46, 0.67), and for BNT162b2, 0.57 (95% CI 0.49, 0.65). Full results from the multivariable logistic regression model are given in table S4. There was little evidence for effect modification by age of index case (likelihood ratio p-value=0.085) or age of contact (p=0.177). Figure S1 shows odds ratios for contacts in which the index case had been vaccinated according to the timing of vaccination, compared to households where the index case was not vaccinated. This analysis also includes households where the index case was vaccinated between 0 and 20 days before testing positive, which due to the timing of the rollout and data under consideration, are a larger group than those vaccinated 21+days before testing positive.

The results show that contacts of vaccinated cases have lower odds of being secondary cases if the index case was vaccinated 14 days or more before testing positive after controlling for calendar week, but this protective effect diminishes sharply if vaccination occurs closer to the positive test date. Of note however is that estimates diverge for the two vaccines: where index cases are recently vaccinated (less than 10 days before testing positive), the odds for contacts being a secondary case are lower for ChAdOx1 nCoV-19, but higher for BNT162b2 (vs. contacts of an unvaccinated index case). The latter may be due to priority administration of BNT162b2 early in the vaccination campaign in those with high-risk social care occupations during a peak incidence period, whose contacts may also have higher risks.

Supplementary Figure S2 show odds ratios of becoming a secondary case according to vaccination timing for different time periods of index case. This suggests that odds of transmission were lower in later calendar weeks for index cases receiving either vaccine compared to unvaccinated cases, suggesting an effect of timing as case rates declined. The difference between vaccines in ORs at less than 10 days is more pronounced in weeks 3-4 (and to some degree in weeks 1-2, although data are sparse). Again, this may suggest that those receiving BNT162b2 were targeted for early vaccine due to high exposure risk, whereas the difference is somewhat less pronounced in weeks 5-6 and 7-8.

We compared household contacts of index cases vaccinated 21-35 days before testing positive with household contacts of index cases vaccinated 1-10 days before testing positive (with the same vaccine type). The adjusted ORs from multivariable logistic regression were 0.53 (95% CI: 0.44, 0.63) for ChAdOx1 nCoV-19 and 0.49 (95% CI: 0.44, 0.56) for BNT162b2, indicating a halving in the odds of contacts becoming secondary cases if the index case was vaccinated with either ChAdOx1 nCoV-19 or BNT162b2 21-35 days before testing positive. Figure S1 . Odds ratios for contacts becoming a secondary case according to vaccination timing of the index case (days before testing positive)

By type of vaccination, vs. contacts where the index case was not vaccinated. Results from multivariable logistic regression. Odds ratios by type of vaccination, vs. contacts where the index case was not vaccinated, as in Figure  S1 . Results from the multivariable logistic regression. 

HOSTED-England's Household Transmission Evaluation Dataset: preliminary findings from a novel passive surveillance system of COVID-19

Timeliness and completeness of laboratorybased surveillance of COVID-19 cases in England. Public Health

The National Immunisation Management Service

Effectiveness of BNT162b2 mRNA Vaccine Against Infection and COVID-19 Vaccine Coverage in Healthcare Workers in England, Multicentre Prospective Cohort Study (the SIREN Study)

The authors wish to thank Heather Pinches, Phillip Bowker and Paul Ellingham as well as the rest of the NHS Digital team whose hard work and dedication enabled the creation of these data during the UK Covid19 pandemic response. The authors also wish to thank Sarah Woodhall for her integral role in developing the HOSTED project; Jamie Lopez-Bernal, Mary Ramsay, Liz Miller and the Vaccine Efficacy Working Group for their input and discussion on this project; and Jemma Walker for comments on an early version of the pre-print relating to this manuscript.

The HOSTED surveillance system was reviewed and approved by the PHE Research Ethics Governance Group. The data was collected and linked by NHS Digital. The data was processed lawfully under GDPR Article 6(1)e and 9(2)i and shared under Regulation 3(4) of the Health Service (Control of Patient Information) Regulations 2002.

This work was undertaken as part of the core functions of Public Health England in relation to the surveillance of communicable diseases and outbreak response. Funding: None