key: cord-0785845-idlmau69
authors: Pritchard, E.; Matthews, P.; Stoesser, N.; Eyre, D.; Gethings, O.; Vitha, K.-D.; Jones, J.; House, T.; VanSteenhouse, H.; Bell, I.; Bell, J.; Newton, J.; Farrar, J.; Diamond, I.; Rourke, E.; Studley, R.; Crook, D. W.; peto, t. E.; Walker, A. S.; Pouwels, K. B.; team, Coronavirus Infection Survey
title: Impact of vaccination on SARS-CoV-2 cases in the community: a population-based study using the UK COVID-19 Infection Survey
date: 2021-04-23
journal: nan
DOI: 10.1101/2021.04.22.21255913
sha: 68d8703e525b6bea4ff8773bbd89e055973f0d65
doc_id: 785845
cord_uid: idlmau69

Objectives: To assess the effectiveness of COVID-19 vaccination in preventing SARS-CoV-2 infection in the community. Design: Prospective cohort study. Setting: The UK population-representative longitudinal COVID-19 Infection Survey. Participants: 373,402 participants aged [≥]16 years contributing 1,610,562 RT-PCR results from nose and throat swabs between 1 December 2020 and 3 April 2021. Main outcome measures: New RT-PCR-positive episodes for SARS-CoV-2 overall, by self-reported symptoms, by cycle threshold (Ct) value (<30 versus [≥]30), and by gene positivity (compatible with the B.1.1.7 variant versus not). Results: Odds of new SARS-CoV-2 infection were reduced 65% (95% CI 60 to 70%; P<0.001) in those [≥]21 days since first vaccination with no second dose versus unvaccinated individuals without evidence of prior infection (RT-PCR or antibody). In those vaccinated, the largest reduction in odds was seen post second dose (70%, 95% CI 62 to 77%; P<0.001).There was no evidence that these benefits varied between Oxford-AstraZeneca and Pfizer-BioNTech vaccines (P>0.9).There was no evidence of a difference in odds of new SARS-CoV-2 infection for individuals having received two vaccine doses and with evidence of prior infection but not vaccinated (P=0.89). Vaccination had a greater impact on reducing SARS-CoV-2 infections with evidence of high viral shedding Ct<30 (88% reduction after two doses; 95% CI 80 to 93%; P<0.001) and with self-reported symptoms (90% reduction after two doses; 95% CI 82 to 94%; P<0.001); effects were similar for different gene positivity patterns. Conclusion: Vaccination with a single dose of Oxford-AstraZeneca or Pfizer-BioNTech vaccines, or two doses of Pfizer-BioNTech, significantly reduced new SARS-CoV-2 infections in this large community surveillance study. Greater reductions in symptomatic infections and/or infections with a higher viral burden are reflected in reduced rates of hospitalisations/deaths, but highlight the potential for limited ongoing transmission from asymptomatic infections in vaccinated individuals. Registration: The study is registered with the ISRCTN Registry, ISRCTN21086382.

On 8 December 2020, the UK was the first country to start a COVID-19 vaccination programme following the emergency use authorisation of the PBNT162b2 messenger RNA (mRNA) vaccine (Pfizer-BioNtech) by UK's Medicines & Healthcare Products Regulatory Agency (MHRA) 1 . Additional COVID-19 vaccines have since been approved, including the Oxford-AstraZeneca adenovirus-vector vaccine, ChAdOx1 nCOV-19 2 , and more recently an mRNA-based COVID-19 vaccine developed by Moderna, mRNA-1273 3 . To date, most vaccinated individuals in the UK received one or two doses of the Pfizer-BioNTech or Oxford-AstraZeneca vaccines.

Initially, those in care homes, over 80 years old, and frontline health and social care workers were prioritised for vaccination 4 . Clinically vulnerable people and those ≥70 years were the next priority groups, followed by remaining adults in decreasing age order. As of 14 April, over 32 million (62%) UK adults had received at least one COVID-19 vaccine dose 5 , and mostly one dose only following the extension of the dosing interval to 12 weeks to maximise initial coverage 6 .

Large randomised trials estimated efficacy against symptomatic laboratory-confirmed COVID-19 infection of 70% (95% CI 55% to 81%) after two Oxford-AstraZeneca doses 7 , and 95% (95% CI 90% to 98%) after two Pfizer-BioNTech doses 8 . Whilst trials provide unbiased effect estimates, trial participants may differ from the general population in many ways, and so it is essential to assess effectiveness in the community, particularly given differences between real-world vaccine deployment and the licenced dosing schedule. Comparing vaccine effectiveness in the community is also important as the trials used different outcome definitions (e.g. start of at-risk period 14 vs 7 days after the second dose) and populations (e.g. smaller proportion >55 years in the Oxford-AstraZeneca vaccine trial (12% 7 vs 42% for Pfizer-BioNtech 8 )).

Furthermore, both trials were largely conducted before the SARS-CoV-2 variant, B.1.1.7, became dominant 9 . This variant is more transmissible and potentially also more severe [10] [11] [12] . Concerns have been raised that some of its defining mutations may affect the efficacy of vaccines and natural infection-derived immunity to (re)infection. A subset of 8,534 participants from the initial Oxford-AstraZeneca trial were followed for a longer period to assess protection against different viral variants, but large uncertainty meant it was difficult to conclude whether efficacy was lower against B.1.1.7 (70%, 95%CI 44% to 85%) than other lineages (82%, 95%CI 70% to 89%) 13 .

Ongoing assessment of the effectiveness of different vaccines across different subgroups is criticalespecially amongst older adults, where more limited evidence from Oxford-AstraZeneca trials has resulted in several countries deciding not to use this vaccine in the elderly despite vaccination shortages and increasing infections. Real-world studies are starting to appear, with an analysis from Israel estimating 92% (95%CI 88 to 95%) effectiveness against symptomatic PCR-confirmed infection ≥7 days after the second Pfizer-BioNTech dose 14 . Another study assessing the early effectiveness of the Pfizer-BioNTech and Oxford-AstraZeneca vaccine in older adults (≥70 years) in England showed a single dose of either vaccine was ~60% and ~80% effective against symptomatic laboratoryconfirmed infection and hospitalisation respectively 15 . The evidence on effectiveness against asymptomatic infection is limited, with one study among healthcare workers from Oxfordshire, UK, showing a 64% (95% CI 50 to 74%) reduction in any SARS-CoV-2 PCR-positive result following a single Pfizer-BioNTech or Oxford-AstraZeneca dose 9 . Another study among 3,950 healthcare workers, first responders, and other essential and frontline workers from the US estimated 80% (95%CI 59 to 90%) and 90% (95%CI 68 to 97%) vaccine effectiveness 14 or more days after 1 or 2 doses of the Pfizer-BioNTech or Moderna vaccines respectively 16 . Most recently, a study in 10,412 residents of long-term care facilities showed 65% and 68% protection against SARS-CoV-2 PCR-positive results 28-42 days after vaccination with Oxford-AstraZeneca and Pfizer-BioNtech vaccines respectively 17 .

However, existing studies have either investigated defined sub-populations 9 16 17 or have relied on results from symptomatic testing programmes 14 15 , potentially leading to bias from vaccination status influencing test-seeking behaviour of cases not requiring healthcare. Large community-based studies where testing is done in a systematic manner (independent of both vaccination status and symptoms) are lacking. We therefore used the Office for National Statistics (ONS) COVID-19 Infection Survey (CIS) -a large community-based survey of individuals aged 2 years and older living in randomly selected private households across the UK -to assess the effectiveness of Pfizer-BioNTech and Oxford-AstraZeneca vaccines against any SARS-CoV-2 PCR positive test performed in the survey 18 , where RT-PCR tests were done on a fixed schedule, irrespective of symptoms, vaccine status and prior infection. We assessed vaccine effectiveness based on overall RT-PCR positivity, and split according to self-reported symptoms, cycle threshold (Ct) value (<30 versus ≥30) as a surrogate for viral load, and gene positivity pattern (compatible with B.1.1.7 or not).

. CC-BY 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 Office for National Statistics (ONS) COVID-19 Infection Survey (CIS) is a large household survey with longitudinal follow-up (ISRCTN21086382, https://www.ndm.ox.ac.uk/covid-19/covid-19infection-survey/protocol-and-information-sheets) (details in 18 ). Private households are randomly selected on a continuous basis from address lists and previous surveys to provide a representative sample across the UK. Following verbal agreement to participate, a study worker visited each selected household to take written informed consent for individuals aged 2 years and over. Parents or carers provided consent for those aged 2-15 years; those aged 10-15 years also provided written assent.

Individuals were asked about demographics, behaviours, work, and vaccination uptake (https://www.ndm.ox.ac.uk/covid-19/covid-19-infection-survey/case-record-forms). At the first visit, participants were asked for (optional) consent for follow-up visits every week for the next month, then monthly for 12 months from enrolment. At each visit, enrolled household members provided a nose and throat self-swab following instructions from the study worker. From a random 10-20% of households, those 16 years or older were invited to provide blood monthly for antibody testing.

Swabs were couriered directly to the UK's national Lighthouse laboratories (Glasgow and the National Biocentre in Milton Keynes (to 8 February 2021)) where samples were tested within the national testing programme using identical methodology. The presence of three SARS-CoV-2 genes (ORF1ab, nucleocapsid protein (N), and spike protein (S)) was identified using real-time polymerase chain reaction (RT-PCR) with the TaqPath RT-PCR COVID-19 kit (Thermo Fisher Scientific, Waltham, MA, USA), analysed using UgenTec Fast Finder 3.300.5 (TagMan 2019-nCoV assay kit V2 UK NHS ABI 7500 v2.1; UgenTec, Hasselt, Belgium). The assay plugin contains an assay-specific algorithm and decision mechanism that allows conversion of the qualitative amplification assay raw data into test results with little manual intervention. Samples are called positive if either N or ORF1ab, or both, are detected. The S gene alone is not considered a reliable positive 18 , but could accompany other genes (ie, one, two, or three gene positives).

Blood samples were couriered directly to the University of Oxford, where they were tested for the SARS-CoV-2 antibody using an ELISA detecting anti-trimeric spike IgG 19 . Before 26 February 2021, the assay used fluorescence detection as previously described (positivity threshold 8 million units) 19 . After this, it used a commercialised CE-marked version of the assay, the Thermo Fisher OmniPATH 384 Combi SARS-CoV-2 IgG ELISA (Thermo Fisher Scientific, Waltham, MA, USA), with the same antigen and a colorimetric detection system (positivity threshold 42 ng/ml monoclonal antibody unit equivalents, determined from 3840 samples run in parallel).

This analysis included participants aged 16 years or over (i.e. those who theoretically could have received vaccination), and all visits with positive or negative swab results from 1 December 2020 to 3 April 2021.

Participants were asked about their vaccination status at visits, including type, number of doses and date(s). Participants from England were also linked to administrative records from the National Immunisation Management Service (NIMS). We used records from NIMS where available, otherwise records from the survey, since linkage was periodic and NIMS does not contain information about vaccinations received abroad or in Northern Ireland, Scotland, and Wales. Where records were available in both, agreement on type was >98% and on dates >95% within ±7 days.

PCR-positive results may be obtained at multiple visits after infection, so we grouped positive tests into 'episodes'. Whole genome sequencing is available on only a subset of positives, and only a subsample provide monthly blood samples for antibody status, so positive episodes were defined using study PCR results. Based on the World Health Organisation (WHO) definition of re-infection as positive tests occurring at least 90 days after the onset of primary infection 20 , but also incorporating multiple consecutive negative tests, we defined the start of a new 'infection episode' as the date of either: i) the first PCR-positive test in the study (not preceded by any PCR-positive test); ii) a PCRpositive test after 4 or more consecutive negative tests; or iii) a PCR-positive test at least 90 days after the start of a previous infection episode with one or more negative tests immediately preceding this. Positive episodes were used to classify exposure groups and outcomes (see below).

At each study visit, a participant was classified into one of seven different exposure groups based on current vaccination status, and study antibody and PCR tests, as follows:

Visits from participants ≥21 days before first vaccination, including those currently with no vaccination date, with no prior PCR/antibody-positive (as defined below) ("Not vaccinated, not previously positive, ≥21 days before vaccination"); ii)

Visits from participants 1 to 21 days before first vaccination with no prior PCR/antibodypositive ("Not vaccinated, not previously positive, 1-21 days before vaccination") iii)

Visits 0 to 7 days following a first vaccination ("Vaccinated 0-7 days ago"); iv) Visits 8 to 20 days following a first vaccination ("Vaccinated 8-20 days ago"); v) Visits 21 days or more following a first vaccination ("≥21 days after 1 st dose, no second dose"); vi)

Visits after second vaccination, ≥21 days following first vaccination ("Post second dose"); vii)

Visits from participants previously PCR/antibody-positive and not (yet) vaccinated ("Not vaccinated, previously positive").

As antibody status before vaccination is not available for all participants, we defined prior positivity by having either a positive antibody measurement or PCR-positive episode >45 days before the visit date. The choice of 45 days was arbitrary, but designed to exclude ongoing infections acquired previously being misattributed to current visits. Information about self-reported or linked positive SARS-CoV-2 PCR or lateral flow tests outside the study was not considered. Visits from vaccinated individuals (groups (iii)-(vi)) were defined irrespective of previous positivity. Visits from the same participant were classified in different groups depending on their status at each visit. As very few visits occurred after a second Oxford-AstraZeneca dose (3,613, 3.5% of all visits ≥21 days after first Oxford-AstraZeneca dose), this group was pooled with Oxford-AstraZeneca one dose only in analyses of vaccine type. We chose these vaccination status categories empirically based on the odds of infection episodes when modelling days since first vaccination as a continuous effect, allowing for non-linearity by using restricted cubic splines (Supplementary Figure 1) .

. CC-BY 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) Analysis was based on visits, since these occur independently of symptoms and are therefore unbiased. Only the first test-positive visit in the first new positive infection episode starting after 1 December was used, dropping all subsequent visits in the same infection episode, to avoid misattributing ongoing PCR-positivity to visit characteristics. Primary analysis included all first new positive infection episodes. Secondary analyses considered the impact of vaccination on infection severity, by classifying positives by cycle threshold (Ct) value (<30 or ≥30) and self-reported symptoms. For each positive test, a single Ct was calculated as the arithmetic mean across detected genes (Spearman correlation>0.98), then the minimum value was taken across positives in the infection episode to reflect the greatest measured viral burden within an episode. To allow for presymptomatic positives being identified in the survey, any self-reported symptoms at any visit within 0 to 35 days after the index positive in each infection episode were included (questions elicit symptoms in the last 7 days at each visit). Finally, positive infection episodes were classified as compatible with the B.1.1.7/VOC202012/01 SARS-CoV-2 variant (those positive at least once for ORF1ab+N across the episode and never S-positive) and those that were incompatible (ORF1ab+N+S or ORF1ab+S or N+S at least once). B.1.1.7/VOC202012/01 has deletions in the S gene leading to S gene target failure, and ORF1ab+N positivity only remains a good proxy for B.1.1.7/VOC202012/01 from whole-genome sequencing from mid November 2020 21 . Positives where only a single N or single ORF1ab gene were detected were excluded from this secondary analysis.

The following potential confounders were adjusted for in all models as potential risk factors for acquiring SARS-CoV-2 infection: geographic area and age in years (see below), sex, ethnicity, index of multiple deprivation (percentile, calculated separately for each country in the UK) [22] [23] [24] [25] , working in a care-home, having a patient-facing role in health or social care, presence of long-term health conditions, household size, multigenerational household, rural-urban classification [26] [27] [28] , direct or indirect contact with a hospital or care-home, smoking status, mode of travel to work, work location, and visit frequency. Details are shown in Supplementary Table 1 . Analysis was based on complete cases (>99% observations) (Supplementary Table 2 ).

Associations between the different exposure groups and outcome (first positive test in an infection episode vs test-negative) were evaluated with generalised linear models with a logit link. Robust standard errors were used to account for multiple visits per-participant. To adjust for substantial confounding by calendar time and age, with non-linear effects of age which are also different by region, we included both as restricted cubic splines with knots at the 20%, 40%, 60%, and 80% percentiles of unique values and interactions between these splines and region/country (regions for England and country for Northern Ireland, Scotland and Wales). Furthermore, given previous observations of different positivity rates by age over time 18 , we added a tensor spline to model the interaction between age and calendar time with the restriction that the interaction is not doubly non-linear 29 . We considered effect modification by age of vaccination by fitting this same model, but also including an interaction between vaccine exposure group and age <75 vs ≥75 years, or longterm health conditions. Pairwise comparisons of the five exposure groups were performed using Tukey adjustments for the pairwise comparisons.

Members of the general public contributed to participant materials. Question wording was tested with members of the general public and amended based on their feedback. No members of the . CC-BY 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 April 23, 2021. ; https://doi.org/10.1101/2021.04.22.21255913 doi: medRxiv preprint public were asked to advise on interpretation or writing up of results. Results will be disseminated to relevant communities through news media.

. CC-BY 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 April 23, 2021. ; Table 4 ). The highest Ct values were in those who had received two vaccine doses, with a similar distribution to those not vaccinated but previously PCR/antibody-positive. Ct values were lowest in those not vaccinated and not previously PCR/antibody-positive. The percentage of PCR-positive cases self-reporting symptoms was highest in those not vaccinated and not previously PCR/antibody-positive, and lowest in those with two vaccine doses and those not vaccinated but previously PCR-/antibody-positive (Figure 2) . Wellrecognised COVID-19 symptoms (cough, fever, loss of taste/smell) were most commonly reported in unvaccinated individuals and not previously PCR/antibody-positive, while other self-reported symptoms occurred similarly across all vaccine exposure groups.

In unadjusted analyses, the percentage of positive PCR tests remained stable over the first 20 days following vaccination, but decreased from 21 days onwards regardless of having received one or two doses (Supplementary Figure 2) . Adjusting for multiple potential confounders, the odds of a new PCR-positive, with or without symptoms, were reduced by 55% (95% CI 49 to 60%) in those 8 to 20 days after vaccination versus those not vaccinated or previously PCR/antibody-positive and ≥21 days before vaccination, with no evidence of a difference versus those vaccinated 0 to 7 days ago (P=0.204). Odds were reduced 65% (95% CI 60 to 70%; P<0.001) in those ≥21 days since first vaccination with no second dose, significantly more than those vaccinated 8 to 20 days ago (P=0.004) (Figure 3A, Supplementary Table 5 ; coefficients for all factors in Supplementary Table 6 ). Odds of testing positive were reduced 72% (95% CI 69 to 74%) 1 to 21 days before first vaccination and 62% (57 to 67%) 0 to 8 days post vaccination versus those not vaccinated or previously PCR/antibody-positive and ≥21 days before vaccination.

In those vaccinated, the largest reduction in odds was seen in those post second vaccine dose (70%, 95% CI 62 to 77%; P<0.001); however, there was no evidence this differed compared with having received only one dose ≥21 days previously (P=0.889). There was no evidence that reductions in . CC-BY 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 April 23, 2021. ; https://doi.org/10.1101/2021.04.22.21255913 doi: medRxiv preprint odds of testing positive differed between having received two vaccine doses and not being vaccinated but previously PCR/antibody-positive (P=1.00) (Supplementary Table 5 ).

The benefits associated with vaccination were much greater for infection episodes with Ct<30 as evidence of high levels of viral shedding compared with Ct≥30 ( Figure 3B) , with a 88% reduction (95% CI 80 to 93%; P<0.001) in odds of testing positive with Ct<30 post-second dose, a marginally greater reduction compared with one dose ≥21 days ago (P=0.050) and with no evidence of difference versus those not vaccinated but previously PCR/antibody-positive (P=1.00). Similarly, benefits associated with vaccination were much greater for self-reported symptomatic infection episodes (Figure 3C) , with an 90% reduction (95% CI 82 to 94%; P<0.001) in odds of testing positive post-second dose with self-reported symptoms, significantly greater than with one dose ≥21 days ago (P=0.012) (Supplementary Table 5 ), but again without evidence of difference versus those not vaccinated but previously PCR/antibody-positive (P=0.992). In comparison, the reduction in odds of new infection episodes with no self-reported symptoms was 49% (95% CI 31 to 62%; P<0.001) postsecond dose. Whilst overlapping, positives with Ct<30 also differed to positives reporting symptoms e.g. 4377 (35%) of all positives had Ct <30 and symptoms reported, and 2,332 (19%) had Ct<30 and no symptoms reported (Supplementary Table 4 ). Effects of vaccination on infections compatible and incompatible with the B.1.1.7 variant appeared similar, but small numbers of positives in the latter group led to large uncertainty in estimates ( Figure 3D ; Supplementary Table 5) .

There was no evidence that reductions in odds of new infections differed between the Pfizer-BioNTech and Oxford-AstraZeneca vaccine (Figure 4A; Supplementary Table 7 ) whether the vaccine was received 0 to 7 days ago (P=0.965), 8 to 20 days ago (P=1.00), or ≥21 days ago (P=0.998 for Pfizer-BioNTech ≥21 days ago, one dose only, vs Oxford-AstraZeneca ≥21 days ago, one or two doses). There was also no evidence that reductions in odds of new infections differed between those post second Pfizer dose and those not vaccinated but previously PCR/antibody-positive (P=1.00). Effects were similar considering infections with Ct<30 and ≥30 (Figure 4B ), and with and without self-reported symptoms (Figure 4C) , with the impact of both vaccines attenuated for infections with Ct≥30 and without self-reported symptoms.

There was evidence of differences in the effect of vaccination on new infection between those aged under or over 75 years (global heterogeneity for all vaccination terms P=0.014), with the reduction in odds of new infections post-vaccination being slightly greater in those aged ≥75 (Figure 6A) . The greatest numeric difference was in those ≥21 days since first vaccination with no second dose, where reductions in odds were 76% in those aged ≥75 (95% CI 68% to 82% reduction) and 62% in those <75 (95% CI 56 to 67%) (interaction P=0.002). There was no evidence of differences in the effect of vaccination on new infection between those reporting or not reporting long-term health conditions (global heterogeneity for all vaccination terms P=0.840).

. CC-BY 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 April 23, 2021.

The results from this large community surveillance study show that vaccination against COVID-19 significantly reduced the odds of individuals testing PCR-positive with a new SARS-CoV-2 infection, with greatest reductions in new infections with Ct<30 and self-reported symptoms, and in those who had received 2 vaccine doses. Reductions afforded by vaccination were similar to those provided by natural immunity. The protective effect of vaccination was attenuated in infections with Ct≥30 and without self-reported symptoms. There was no evidence of any difference in effectiveness between Pfizer-BioNTech and Oxford-AstraZeneca vaccines, or in those with long-term health conditions. We observed greater reductions in new infections in those aged ≥75 years versus those under 75.

The main study strength is its design as a large-scale community survey recruiting from randomly selected private residential households, providing a representative sample of the UK general population. Participants are tested regardless of symptoms, allowing us to additionally consider vaccine effectiveness against infection without reported symptoms. The availability of Ct values allowed us to compare vaccine impact on viral loads, using Ct as a proxy 30 . Scheduled visits provide an unbiased sampling frame which we exploited for our logistic regression, rather than having to censor individuals at last tests in the study using time-to-event analyses, and assume all infections between visits were identified. Participants were asked about demographics, behaviours, and work, allowing us to control for a wide range of potential confounders that are unavailable in record linkage studies performed to date. 15 The design also has limitations, particularly with individuals tested initially at weekly and then monthly visits. Any positive episodes occurring between visits will be missed, leading to contamination of the "not vaccinated, no previous PCR/antibody-positive" groups, possibly diluting the effect of vaccination. Because participants can only test positive at scheduled visits, some of the "new" positives episodes may in fact have occurred sometime previously; we therefore stratified time from vaccination to reduce the impact of this. Older infections would be expected to have a higher Ct values, so this may also partly explain the differences between positives with Ct<30 and ≥30, at least shortly after vaccination. Imperfect sensitivity of SARS-CoV-2 PCR tests may bias absolute risk, but would result in unbiased relative risk provided that misclassification is nondifferential to vaccination status and all non-cases are correctly classified (i.e. 100% specificity). PCR test specificity is likely very high 12 18 , and therefore any bias here is expected to be small. Due to relatively small numbers of infections post-vaccination, power to detect differences between vaccine types and differential vaccine effectiveness in subgroups was relatively low.

An important potential issue with observational studies evaluating vaccine effectiveness is that individuals are not supposed to be vaccinated if they recently tested positive, and individuals may reduce their number of contacts in response to the knowledge that they will soon receive a vaccination. Interestingly, we found that 613 individuals tested positive 1 to 21 days before receiving their vaccination -due to the design and logistics of the survey they may have received their test results after the date of vaccination -suggesting that ensuring social distancing at vaccination locations remains important. The reduced risk observed in the 21 days prior and 0-7 days after vaccination is likely due to this reverse causality, specifically changes in behaviour due to either receiving the vaccination invitation letter or knowledge that individuals from their age or risk group are about to get vaccinated in their area, rather than a biological effect. Because a reduction in contacts in the week before vaccination will also reduce the likelihood of testing positive in the . CC-BY 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 April 23, 2021. ; https://doi.org/10.1101/2021.04.22.21255913 doi: medRxiv preprint following week, it will be important for future studies trying to evaluate the effectiveness of vaccination to carefully construct the appropriate comparator. Here we used study visits from those that are not vaccinated, not previously positive, ≥21 days before vaccination as comparator to overcome these issues when estimated the impact of the vaccination itself.

Our estimated effect of two vaccine doses on symptomatic infections is similar to other studies which have considered this outcome 9 14-16,17 , but is slightly lower than that reported in the key Phase III clinical trials 7 8 . The clinical trials had a more intensive testing schedule, whereas we may have missed some infections due to monthly testing in the majority of participants. Another explanation could be differences with our general population sample, in particular our vaccinated participants being, on average, older due to their prioritisation in the UK's vaccine rollout 4 , combined with decreased immunological competence (immunosenescence) in an older population 31 (although we did not identify any loss of benefit in older individuals in subgroup analyses). Higher Ct in infections identified post vaccination has also been demonstrated in older adults in care homes 17 . Our estimated effectiveness is also slightly lower than studies in healthcare workers 9 16 ; these studies had antibody tests in the majority of participants so were likely able to identify previous infection more accurately, avoiding misclassification in our control "not vaccinated, no previous PCR/antibodypositive" group. Our estimated reduction in risk of infection for those not vaccinated but previously PCR/antibody positive was slightly lower than the ~80% (95% CI 75.4 to 84.5%) estimated elsewhere 32 .

Consistent with two recent studies 9,13 , we found vaccination to be as effective against the B.1.1.7 variant as non-B.1.1.7 variants. Our study supports this in a broader population, including positives from individuals not reporting symptoms and for the Pfizer-BioNTech vaccine in addition to the Oxford-AstraZeneca vaccine. Our study had good power to estimate vaccine effectiveness against the B.1.1.7 variant as it was conducted over the period when B.1.1.7 became dominant in the UK. This is particularly relevant as the variant has now been detected in over 40 countries worldwide 33 34 , and the major Phase III vaccine trials were conducted before this strain was dominant 7 8 . We observed a slightly greater reduction in new infection episodes in those vaccinated and aged ≥75 years, compared with those <75 years, potentially due to the combination of vaccination with reduced social contact in the former group. We currently do not have evidence of the vaccine being less effective in older individuals as seen elsewhere with natural re-infections 32 , although would note that, as described above, vaccine effectiveness also includes a non-biological behavioural component and there may be compensation for lower biological activity in older individuals with lower behavioural risk.

Similar to other studies 7 9 16 , we found greater reductions in new positives after two vaccine doses compared with one dose, particularly in reducing infections with self-reported symptoms and low Ct/high viral load. In the UK, the interval between vaccine doses was extended to 12 weeks to maximise initial coverage and reduce hospitalisations/deaths; our findings highlight the importance for increased protection of individuals getting the second vaccine dose. Nonetheless, the significant reduction in positivity after only one dose supports the decision to maximise initial vaccination coverage.

While some infections, particularly those with Ct≥30, could represent historical infections contracted prior to vaccination, given the timescales and prior negatives post vaccination, some will . CC-BY 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 April 23, 2021. ; https://doi.org/10.1101/2021.04.22.21255913 doi: medRxiv preprint undoubtedly reflect new infections after vaccination. Together with other evidence, this suggests that vaccination does not completely prevent infection following virus exposure, yet minimises progression to more severe infection 14 . The fact that vaccinated individuals can still be infected, even if predominantly with lower viral burden/asymptomatic infections, means that onwards transmission remains a possibility, albeit at lower efficiency 35 . Maintaining measures such a social distancing may therefore still be needed to control virus spread until enough of the population is vaccinated.

We have also shown two vaccine doses to be as effective as prior natural infection. This could be an important consideration during policy development over COVID-status certification or "COVID passports", and supports considering both prior PCR/serological testing and vaccination data for this 36 .

Looking forward, one key question will be whether immunisation offers long-term protection against COVID-19. A recent study showed the rate of waning and longevity of neutralising antibodies varies greatly amongst individuals with prior COVID-19 infection and suggested that, if similar rates of waning are seen after vaccination, annual vaccine administration is likely needed 37 .

Overall, we have shown COVID-19 vaccination to be effective in reducing the number of new SARS-CoV2 infections, with the greatest benefit received after two vaccinations, and against symptomatic and high viral burden infections, and no difference between the Pfizer-BioNTech and Oxford-AstraZeneca vaccine.

What is already known on this topic -Large randomised trials have shown high efficacy of Oxford-AstraZeneca and Pfizer-BioNTech vaccines against symptomatic laboratory-confirmed SARS-CoV-2 infection -The effectiveness of these vaccines in the real world against any SARS-CoV-2 infection, including those without symptoms is less clear, especially among the elderly that were underrepresented in the Oxford-AstraZeneca trial

-SARS-CoV-2 infections fall substantially after a first dose of either vaccine; two doses of the Pfizer-BioNTech vaccine provided even greater protection, to a similar degree as previous infection with SARS-CoV2 -Vaccination and previous infection were most effective at reducing symptomatic infections, and infections with high viral burden, with lower reductions in infections not causing symptoms and with lower viral burden. -Both vaccines appear to be highly effective against infections compatible with B.1.1.7

. CC-BY 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 April 23, 2021. ; https://doi.org/10.1101/2021.04.22.21255913 doi: medRxiv preprint The study was designed and planned by ASW, JF, JB, JN, IB, ID and KBP and is being  conducted by ASW, IB, RS and ER. This specific analysis was designed by ASW, KBP, PCM, NS, DWE,  TH, DC, TEAP, K-DV, and EP . EP, KBP, OG, and JJ contributed to the statistical analysis of the survey data. HVS conducted analysis of the RT-PCR data. EP, ASW and KBP drafted the manuscript. All authors contributed to interpretation of the study results, and revised and approved the manuscript for intellectual content. KBP and ASW are the guarantors and accept full responsibility for the work and conduct of the study, had access to the data, and controlled the decision to publish. The corresponding author (KBP) attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. The funder/sponsor did not have any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. All authors had full access to all data analysis outputs (reports and tables) and take responsibility for their integrity and accuracy.

Competing interests: All authors have completed the ICMJE uniform disclosure from at www.icmje.org/coi_disclore.pdf and declare: DWE declares lecture fees from Gilead, outside the submitted work; EP, PCM, NS, DWE, JIB, DC, TEAP, ASW, and KBP are employees of the University of Oxford, but not involved in the development or production of the vaccine; JIB act as an unpaid advisor to HMG on Covid but does not sit on the vaccine task force and it not involved in procurement decisions, sits on the Board of OSI who has an investment in Vaccitech who have a royalty from the Oxford-AstraZeneca vaccine when, if ever, it makes a profit; HVS reports personal fees from BioSpyder Technologies, Inc, outside the submitted work; ASW besides funding mentioned above, also received grants from Medical Research Council UK during the conduct of the study; there are no other relationships or activities that could appear to have influenced the submitted work. . CC-BY 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 April 23, 2021. ;  Transparency The lead authors affirm that the manuscript is an honest, accurate, and transparent account of the study design being reported, no important aspects of the study have been omitted, and any discrepancies from the study as originally planned (and, if relevant, registered) have been explained.

Dissemination to participants and related patient and public communities: Results of individual tests were communicated to the participants. Overall study results were disseminated through the preprint of the study. Findings were disseminated in lay language in the national and local press. Provenance and peer review: Not commissioned; externally peer reviewed.

. CC-BY 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 April 23, 2021. ; https://doi.org/10.1101/2021.04.22.21255913 doi: medRxiv preprint . CC-BY 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 April 23, 2021 Supplementary Table 4. . CC-BY 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. 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 April 23, 2021. 23 * Not vaccinated, but with a positive antibody result in the study >45 days previously or a previous positive episode in the study Note: Odds ratios given in Supplementary Table 5 . ≥30 (B) , self-reported symptoms (C). All odds ratios are compared to the reference category of "Not vaccinated, not previously positive and ≥21 days before vaccination" * Not vaccinated, but with a positive antibody result in the study >45 days previously or a previous positive episode in the study Note: Odds ratios given in Supplementary Table 7. . CC-BY 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 April 23, 2021. 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 April 23, 2021.

.

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 April 23, 2021. 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 April 23, 2021.

. CC-BY 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 April 23, 2021. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) . CC-BY 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. 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 April 23, 2021. ;

Odds ratio 95% CI P-value Figure 4 .

. CC-BY 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 April 23, 2021. ; https://doi.org/10.1101/2021.04.22.21255913 doi: medRxiv preprint

Medicines and Healthcare Products Regulatory Agency. Vaccine BNT162b2: conditions of authorisation under Regulation

Medicines and Healthcare products Regulatory Agency. Regulatory approval of COVID-19 Vaccine AstraZeneca

Medicines and Healthcare products Regulatory Agency. Regulatory approval of COVID-19 Vaccine Moderna

Joint Committee on Vaccination and Immunisation. Joint Committee on Vaccination and Immunisation: advice on priority groups for COVID-19 vaccination

Vaccinations in United Kingdom

Epidemiological and evolutionary considerations of SARS-CoV-2 vaccine dosing regimes

Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine

Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK

An observational cohort study on the incidence of SARS-CoV-2 infection and B. 1.1. 7 variant infection in healthcare workers by antibody and vaccination status. medRxiv

Estimated transmissibility and impact of SARS-CoV-2 lineage B.1.1.7 in England

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

Increased infections, but not viral burden, with a new SARS-CoV-2 variant. medRxiv

Efficacy of ChAdOx1 nCoV-19 (AZD1222) vaccine against SARS-CoV-2 variant of concern 202012/01 (B.1.1.7): an exploratory analysis of a randomised controlled trial

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

Early effectiveness of COVID-19 vaccination with BNT162b2 mRNA vaccine and ChAdOx1 adenovirus vector vaccine on symptomatic disease, hospitalisations and mortality in older adults in England

Interim Estimates of Vaccine Effectiveness of BNT162b2 and mRNA-1273 COVID-19 Vaccines in Preventing SARS-CoV-2 Infection Among Health Care Personnel, First Responders, and Other Essential and Frontline Workers -Eight

Vaccine effectiveness of the first dose of ChAdOx1 nCoV-19 and BNT162b2 against SARS-CoV-2 infection in residents of Long-Term Care Facilities (VIVALDI study)

Community prevalence of SARS-CoV-2 in England from

Performance characteristics of five immunoassays for SARS-CoV-2: a head-to-head benchmark comparison

Pan American Health Organisation. Interim guidelines for detecting cases of reinfection by SARS-CoV-2

Investigation of novel SARS-CoV-2 variants of concern

Ministry of Housing CLG. English indices of deprivation 2019

Index of Multiple Deprivation (full Index update with ranks

Scottish Index of Multiple Deprivation 2020

Northern Ireland Multiple Deprivation Measure

Department for Environment FRA. Rural Urban Classification

Scottish Government Urban Rural Classification

Regression modeling strategies: with applications to linear models, logistic and ordinal regression, and survival analysis

Duration of infectiousness and correlation with RT-PCR cycle threshold values in cases of COVID-19

Tackling immunosenescence to improve COVID-19 outcomes and vaccine response in older adults

Assessment of protection against reinfection with SARS-CoV-2 among 4 million PCR-tested individuals in Denmark in 2020: a population-level observational study

Phylogenetic analyses of SARS-CoV-2 B.1.1.7 lineage suggest a single origin followed by multiple exportation events versus convergent evolution

An observational study of SARS-CoV-2 infectivity by viral load and demographic factors and the utility lateral flow devices to prevent transmission

*Pairwise p-value: p-value testing whether the OR for each vaccine status group is different to the vaccine status group below; so respectively "Vaccinated 0 to 7 days ago, 1 dose" vs "Not vaccinated, no prior positive, 1-21 days before vaccination", "Vaccinated 8 to 20 days ago" vs "Vaccinated 0 to 7 days ago", "Vaccinated ≥ 21 days ago, 2 doses" vs "Vaccinated ≥ 21 days ago, 1 dose" and "Not vaccinated, but swab or antibody positive >45 days ago" vs "Vaccinated ≥ 21 days ago, 2 doses".Note: all odds ratios are compared to the reference category of Not vaccinated, no prior positive (>45 days ago), >21 days before vaccination. Results shown graphically in