key: cord-0272405-berv929f
authors: Canas, L. S.; Osterdahl, M. F.; Deng, J.; Hu, C.; Selvachandran, S.; Polidori, L.; May, A.; Molteni, E.; Murray, B.; Chen, L.; Kerfoot, E.; Klaser, K.; Antonelli, M.; Hammers, A.; Spector, T.; Ourselin, S.; Steves, C. J.; Sudre, C. H.; Modat, M.; Duncan, E. L.
title: Disentangling post-vaccination symptoms from early COVID-19
date: 2021-07-22
journal: nan
DOI: 10.1101/2021.07.21.21260906
sha: eddaa73dc1e3106db3490f9fe6ede2589f164c41
doc_id: 272405
cord_uid: berv929f

Background: Identifying and testing individuals likely to have SARS-CoV-2 is critical for infection control, including post-vaccination. Vaccination is a major public health strategy to reduce SARS-CoV-2 infection globally. Some individuals experience systemic symptoms post-vaccination, which overlap with COVID-19 symptoms. This study compared early post-vaccination symptoms in individuals who subsequently tested positive or negative for SARS-CoV-2, using data from the COVID Symptom Study (CSS) app. Design: We conducted a prospective observational study in UK CSS participants who were asymptomatic when vaccinated with Pfizer-BioNTech mRNA vaccine (BNT162b2) or Oxford-AstraZeneca adenovirus-vectored vaccine (ChAdOx1 nCoV-19) between 8 December 2020 and 17 May 2021, who subsequently reported symptoms within seven days (other than local symptoms at injection site) and were tested for SARS-CoV-2, aiming to differentiate vaccination side-effects per se from superimposed SARS-CoV-2 infection. The post-vaccination symptoms and SARS-CoV-2 test results were contemporaneously logged by participants. Demographic and clinical information (including comorbidities) were also recorded. Symptom profiles in individuals testing positive were compared with a 1:1 matched population testing negative, including using machine learning and multiple models including UK testing criteria. Findings: Differentiating post-vaccination side-effects alone from early COVID-19 was challenging, with a sensitivity in identification of individuals testing positive of 0.6 at best. A majority of these individuals did not have fever, persistent cough, or anosmia/dysosmia, requisite symptoms for accessing UK testing; and many only had systemic symptoms commonly seen post-vaccination in individuals negative for SARS-CoV-2 (headache, myalgia, and fatigue). Interpretation: Post-vaccination side-effects per se cannot be differentiated from COVID-19 with clinical robustness, either using symptom profiles or machine-derived models. Individuals presenting with systemic symptoms post-vaccination should be tested for SARS-CoV-2, to prevent community spread. Funding: Zoe Limited, UK Government Department of Health and Social Care, Wellcome Trust, UK Engineering and Physical Sciences Research Council, UK National Institute for Health Research, UK Medical Research Council and British Heart Foundation, Alzheimer's Society, Chronic Disease Research Foundation, Massachusetts Consortium on Pathogen Readiness (MassCPR).

symptoms and SARS-CoV-2 test results were contemporaneously logged by participants. Demographic and clinical information (including comorbidities) were also recorded. Symptom profiles in individuals testing positive were compared with a 1:1 matched population testing negative, including using machine learning and multiple models including UK testing criteria.

Findings: Differentiating post-vaccination side-effects alone from early COVID-19 was challenging, with a sensitivity in identification of individuals testing positive of 0.6 at best. A majority of these individuals did not have fever, persistent cough, or anosmia/dysosmia, requisite symptoms for accessing UK testing; and many only had systemic symptoms commonly seen post-vaccination in individuals negative for SARS-CoV-2 (headache, myalgia, and fatigue).

Interpretation: Post-vaccination side-effects per se cannot be differentiated from COVID-19 with clinical robustness, either using symptom profiles or machine-derived models. Individuals presenting with systemic symptoms post-vaccination should be tested for SARS-CoV-2, to prevent community spread. . 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 22, 2021 

The havoc wrought by SARS-CoV-2 is unprecedented in living memory, with >184 million cases of COVID-19 world-wide and >4.0 million deaths by 8 July 2021. 1,2 Extraordinary efforts directed towards rapid vaccine development meant that by late 2020 the UK Medicines and Healthcare products Regulatory Agency had authorized three vaccines: Pfizer-BioNTech mRNA (BNT162b2) 3, 4 Oxford-AstraZeneca adenovirus-vectored [5] [6] [7] and Moderna mRNA (mRNA-1273) 8, 9 A fourth vaccine (Janssen adenovirus-vectored Ad26.COV2.S) was authorised on 28 May 2021. 10 Vaccination with BNT162b2 (herein, PB) and ChAdOx1 nCoV-19 (herein, O-AZ) started in the UK on 8 December 2020 11 and 4 January 2021 12 respectively, during which time the UK was experiencing its third pandemic wave with widespread community transmission (peak UK positive specimens reported on 29 December 2020 13 ). Since then, and in the context of social distancing and stay-at-home directives, new infections, hospitalisations, and deaths from SARS-CoV-2 have fallen rapidly. 1, 2, 14 Local and systemic reactions have been observed after all vaccines for SARS-CoV-2. Considering the two vaccines used predominantly in the UK to date (O-AZ and PB), local reactions were common during their pivotal trials (76% of younger (<55 years) O-AZ recipients reported tenderness; 5,6 83% of younger PB recipients reported pain). 4 Systemic reactions were also common and included fatigue (O-AZ 76%; PB 59%), headache (O-AZ 65%; PB 52%), and fever (O-AZ 24%; PB 16%). [4] [5] [6] Observational data from the COVID Symptom Study (CSS) 15 also showed high incidence of local (62%) and systemic (26%) effects. 16 The most serious side-effect to date, observed after both O-AZ and Janssen vaccines, is vaccine-induced immune thrombotic thrombocytopenia (VITT) associated with anti-PF4 antibody production. 17, 18 As of 9 June 2021, with 69,743,980 vaccinations administered in the UK (>45 million O-AZ, including first and second doses), 395 cases of VITT have been reported, with 70 deaths. 19 Saliently, most vaccine-related side-effects (including VITT) are more common in younger individuals, whereas COVID-19 clinical severity increases with age. [4] [5] [6] 16 Prevention of SARS-CoV-2 dissemination requires rapid recognition followed by quarantining of infected individuals (along with appropriate health care). However, there is overlap between symptoms from COVID-19 20,21 and early post-vaccination systemic symptoms. [4] [5] [6] 16 Moreover, immunity to SARS-CoV-2 does not occur immediately post-vaccination, 22 with functional protection from approximately day 12. 23 Quarantining and testing every individual with systemic symptoms early post-vaccination would be onerous, expensive, and labourintensive -but given the impact of viral outbreaks might be unavoidable if SARS-CoV-2 infection cannot be excluded robustly. 20, 21 Here we aim to determine whether symptom profiles can be used to differentiate individuals with systemic sideeffects of vaccination alone from individuals with superimposed SARS-COV-2 infection.

. 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 22, 2021 

Data were acquired prospectively from the CSS, using a mobile health application launched by ZOE Limited and King's College London in March 2020 (app details and development given in Supplementary Methods). 15, 21 Briefly, individuals are asked daily to log their health status, health care access, SARS-CoV-2 testing and results, and vaccination details, with direct questions about symptoms associated with COVID-19 (Supplementary Table   S1 ). 14, 15, 21 Symptomatic individuals are prompted to undergo testing, either through standard care or through test request from ZOE/CSS. 24 Data were acquired from UK participants aged 16-90 years, between 8 December 2020 (UK vaccination start date) and 17 May 2021, who were asymptomatic when vaccinated with PB or O-AZ (first or second dose), and subsequently reported: a) at least one predefined symptom (Supplementary Table S1 ) within seven days postvaccination, and b) a SARS-CoV-2 test result (rtPCR or lateral flow antigen test [LFAT]) within ten days postvaccination. The seven-day cut-off for symptom presentation was informed by: a) serial interval for COVID-19

(mean, 5.2 days 25) ; b) incubation period for SARS-CoV-2 (mean, 5.8 days 25, 26 ); c) the timeline for acute postvaccination side-effects in both pivotal trials (one-week 4-6,8,10 ) and d) reported real-world experience of postvaccination symptoms (peak prevalence day 1 post-vaccination; mean duration one day 16 ). The ten-day cut-off for testing allowed three days' delay in accessing testing. 27 Early results indicated a large imbalance in numbers of individuals testing positive vs. negative post-vaccination, sufficient to bias analysis. 28 A 1:1 population from the negative cohort (matching age, BMI, gender, occupation,

week of testing, and comorbidities) was selected based on minimisation of Euclidean distance between positive and negative subjects considering these features, enabling a fair comparison between groups of equal size. 29, 30 However, to ensure robustness, analyses were repeated using: a) a one-hundred bootstrapping scheme selecting from the negative population; and b) the entire negative population.

Individual symptoms (here, a symptom reported at any time within seven days post-vaccination, irrespective of duration) were compared between vaccinated individuals testing positive or negative for SARS-CoV-2, using Chi-squared tests per symptom. Duration of individual symptoms was calculated as days from first report of that symptom, until asymptomatic and/or seven days post-vaccination, noting that duration beyond seven days was not considered; however, as number of individuals experiencing each symptom were low in both groups, no statistical comparison was made. Symptom burden, defined as total symptom count per person [irrespective of symptom duration] were compared between groups using Mann-Whitney-U tests. We also considered symptom manifestation across the week post-vaccination, by dynamic profiling for each symptom (symptom frequency).

Distribution of symptom duration was assessed using Mann-Whitney-U. Correlation of individual symptoms within both positive and negative individuals was assessed by computing a Spearman-rank correlation test. Local symptoms due to vaccination per se (Supplementary Table S2 ) were excluded from analysis as unlikely to be indicative of, or influenced by, SARS-CoV-2 infection.

. 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 22, 2021. ; https://doi.org/10.1101/2021.07.21.21260906 doi: medRxiv preprint Machine learning was used to determine if post-vaccination symptoms per se could be separated from superimposed SARS-CoV-2 infection (including symptom combination, and cumulative symptom burden). 16, 21 We trained a set of binary classifiers to identify SARS-CoV-2-positive individuals. Models included random forest, logistic regression, and Bayesian mixed-effect models, exploiting their varying properties (Supplementary   Table S3 ) to improve reliability of results. We also considered whether symptoms were after first or second vaccination and the impact of vaccination order, by including vaccination ordering as a covariate. Models were trained on data without stratifying by vaccine type, due to small sample sizes. We also did not discriminate between type of SARS-CoV-2 testing (PCR vs. LFAT), or mode of testing access (NHS vs. ZOE-request), either in model development or other analyses.

We sought to reduce bias from assessing high numbers of individual symptoms by performing symptomclustering using K-means. 31 However, a relevant/accurate number of clusters was not evident from the silhouette plot and entropy (data not shown); thus further analyses using machine clustering were not pursued. Symptoms were clustered manually into clinical groupings (reviewed by ELD, MO, TS, AH, CJS) (Supplementary Table   S4 ), and analysed using the above models similarly. Lastly, clustering based on having at least one of the four symptoms required for accessing NHS testing during the timing of this study (viz., presence or absence of fever, persistent cough, anosmia and/or dysosmia) 24 were assessed.

Data were split into training and validation sets for random forest, logistic regression, and Bayesian mixed-effect models. Five folds were used on the training set, composed of 80% of the initial dataset randomly selected, to train the models in different subsamples of the population. The remaining 20% were then used to assess the performance of models, evaluating sensitivity, specificity, and balanced accuracy. The class ratio was maintained in both training and testing sets. For fair evaluation, both clinical clustering and categorisation using the NHS criteria were assessed on 20% of the data of each fold.

ZOE Limited developed the app for data collection as a not-for-profit endeavour. The funder had no role in study design, data analysis, data interpretation, or influence on report content.

The app and CSS were approved in the UK by KCL's ethics committee (REMAS no. 18210, review reference LRS-19/20-18210). All app users provided informed consent for use of their data for COVID-19 research. . 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 11 ,317 tested individuals did not report any requisite symptom, of whom 88 (0.78%) tested positive. Individuals with requisite symptoms were more likely to test positive than those without (p-value<0.0001); none-the-less, the majority (88 of 150, 59%) who tested positive did not meet current UK testing criteria. . 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 For further analyses, one positive individual (vaccinated with PB) was excluded due to invalid data entry (invalid BMI), leaving 149 symptomatic positively-tested individuals. Table 1 describes the positive and matched negative cohorts. Table S5 ).

Individual symptom prevalence after first vaccination is shown in Figure 2 and Supplementary Table S5 . Although some symptoms were more common in individuals testing positive vs. negative (sore throat (p-value = 0.0187), . 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 22, 2021. ; https://doi.org/10.1101/2021.07.21.21260906 doi: medRxiv preprint 10 sneezing (p value = 0.0474) and persistent cough (p-value = 0.0396)), others were more common in the negative group (palpitations (p-value = 0.0284). The numbers of individuals reporting each symptom were small (e.g., sore throat, n=17; persistent cough, n =12) (Supplementary Table S5 ). Table 2 presents prevalence of each symptom over the week post-vaccination, divided into three windows. Some symptoms increased over time in both positive and negative individuals (e.g., headache, myalgia) whereas others increased in positive individuals only (e.g., sneezing, hoarse voice). Although fever and sore throat increased across the week in the negative individuals, there was a suggestion of a biphasic response in the positive individuals, also observed with persistent cough. The numbers of individuals were too small for formal testing; moreover, the exact date of infection in positive individuals was unknown. 

Negative SARS-CoV-2 . 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 Individual symptom duration is shown in Supplementary Table S8 . There were no significant differences in symptom duration in positive vs. negative individuals after first vaccination. Importantly, symptom assessment was truncated at seven days, noting as above that some symptoms were increasing in prevalence with time.

Interestingly, amongst individuals testing negative, dysosmia and delirium had the longest duration (median, 2 days for each).

. 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 22, 2021 Figure S1) ; and with the negative population as a whole (Supplementary Figure S2) . Some symptoms were significantly more common in negative individuals when using the entire negative population (e.g., brain fog), driven by the extremely large negative sample size, which supports our use of a selected matched population to avoid bias from unbalanced sample size.

There was no correlation between symptoms in either the positive or negative populations, assessed using Spearman-rank test (Supplementary Figure S3) . As a sensitivity analysis we assessed the impact of a (self-logged) previous COVID-19 diagnosis; this made no difference to our results.

Model performance including receiver operator curves, using all reported symptoms, are shown in Table 3 and Table S4) and categorisation of individuals using NHS screening criteria, were no better than chance. . 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 22, 2021. ; https://doi.org/10.1101/2021.07.21.21260906 doi: medRxiv preprint 

Here we aimed to develop a clinically useful algorithm predictive of SARS-CoV-2 infection early postvaccination, by parsing symptoms according to proven infection status in symptomatic individuals. Such an algorithm would be extremely useful, particularly in countries with limited health resources, as testing could be targeted towards those predicted positive, with quarantining of these individuals until an available result. To our knowledge, this is the first study with this aim. However, we were unable to differentiate post-vaccination symptoms per se from superimposed SARS-CoV-2 infection with robustness. Although two models showed ROC AUC significantly greater than 0.5, neither came close to approaching clinical utility -for most clinical tests, conventionally given as 0.8; but for a highly infectious agent with devastating consequences from community spread the necessary AUC is much higher.

Although one third of the one million vaccinated app users reported symptoms previously associated with COVID-19 early post-vaccination, only 4% of symptomatic individuals reported testing for SARS-CoV-2 even with allowance for delayed testing access. Considering those individuals who reported at least one of the symptoms fulfilling NHS criteria for testing (266,502 overall), 40% (107,929) were tested. During the study period, testing was widely available in the UK and it is unclear why more symptomatic people (including those with the widely advertised core symptoms) were not tested. 33 Conversely, of 149 individuals who tested positive, only 62 (41%) had symptoms that met current UK testing criteria. We do not know why the other 88 positive individuals were tested (e.g., contact tracing, routine workplace testing, direct personal request through the app).

Our data also suggest sensitivity of using core symptoms for COVID-19 may be lower post-vaccination than in pre-vaccination times (here 48%, previously 73%.). 34 Although individuals with core symptoms were more likely to test positive than those without, the overall sensitivity and AUC suggests current UK testing policy is suboptimal for pandemic management particularly now that rapid testing capacity is much greater than when these criteria were established. 34 Notably, current UK testing criteria are more limited than WHO guidelines 2 and that of many other jurisdictions of similar GDP (including France, Germany, USA, and Australia).

. 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 22, 2021 Although there were some differences in symptom prevalence and distribution between positive and negative individuals, these could not be used robustly to discriminate between groups, including using machine-learning.

We also considered time of symptom onset and symptom duration post-vaccination, (previous trials and postmarketing observational data have examined these parameters but not with respect to SARS-CoV-2 status). [4] [5] [6] 8, 16 Whether positive or negative, median symptom peak burden was day 3 in both groups, concordant with vaccination side-effect profiles reported previously. [4] [5] [6] 8, 16 As time progressed, some symptoms started to become more common in the positive group only (e.g., persistent cough, hoarse voice), which timing coincides with the serial interval and incubation period of SARS-CoV-2; 26 however, the critical public health importance of identifying and isolating cases early does not allow the luxury of a watch-and-wait approach.

We do not know the circumstances contributing to infection of the positive group (whether prior to, peri-, or Overall, CSS app users are not fully representative of the UK population (younger, more likely to be female, of higher educational status, and over-representative of healthcare workers) 21 . Although the population in the current study shares some of these biases, the median age of vaccinated individuals at the time of our analysis (64 years) was older than for app users overall (47 years), which is not surprising as the UK vaccination schedule began with the oldest individuals in the community. We considered the implications of this with respect to the likelihood of an infected person presenting for testing: although asymptomatic SARS-CoV-2 infection is well-recognised, it is less common in older people. 25, 26 We also acknowledge that different economic and cultural experiences may influence presentation of SARS-CoV-2 infection and reporting of post-vaccination side-effects. 45

Our approach in comparing symptom profiles for individuals testing positive or negative for SARS-CoV-2 required a 1:1 matched population, so that comparison of symptom prevalence was fair and unbiased by the greatly different sample sizes of the two populations. However, this methodological choice is less reliable when used for the outcome of SARS-CoV-2 test prediction; and the forced balance of the classes can lead to an overestimation . 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 22, 2021. ; https://doi.org/10.1101/2021.07.21.21260906 doi: medRxiv preprint of the likelihood of being positive in the modelling. Thus we have also presented extended analyses, using both boot-strapping, and entire-cohort approaches in the Supplementary Results. We acknowledge that the predictive power of our optimised models may be hampered if there are brand-specific post-vaccination side-effects, which were not considered during model optimisation. 16 However, although there were some differences in frequencies, most symptoms were reported in both PB and O-AZ pivotal trials. 4, 5, 8, 10 We also did not consider type of SARS-CoV-testing (PCR vs. LFAT), or mode of testing access (NHS vs. ZOE-request), which may also contribute variability to these models. A strength of our study was our very large cohort of vaccinated participants, in a country that was an early adopter of vaccination; and our timeframe included the UK pandemic "third wave". Prospective real-time symptom logging through the app minimised recall bias; and our symptom assessment included direct ascertainment of core symptoms for accessing UK testing. However, the sharp decline in cases in the first six months of 2021 resulted in only 150 positive cases to inform our modelling which number we acknowledge is small -though we were also able to draw upon large numbers of tested negative individuals for comparisons, reinforcing the consistency and generalisation of our results. We also acknowledge that the demographic features of the app population especially those parameters considered for model estimation (e.g., age, gender, BMI) may be different in other populations within the UK and elsewhere.

The implication of our results will vary depending on the population prevalence of SARS-CoV-2 and pace of vaccination roll-out. At the time of writing Australia has negligible community spread of SARS-CoV-2 but is early in vaccination roll-out. It would be very interesting to repeat this study in these different circumstances.

Testing of all symptomatic individuals comes at a cost (e.g., testing kits, infrastructure). Here, the UK is a resource-rich country; the impact of our results in countries with fewer health resources needs careful consideration.

In conclusion, post-vaccination symptoms cannot be distinguished with clinical confidence from early SARS-CoV-2 infection. Our study highlights the critical importance of testing symptomatic individuals -even if recently vaccinated -to ensure early detection of SARS-CoV-2 infection and help prevent future waves of COVID-19.

. 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 22, 2021. ; https://doi.org/10.1101/2021.07.21.21260906 doi: medRxiv preprint

COVID-19 Coronavirus Pandemic

Public Assessment Report Authorisation for Temporary Supply COVID-19 mRNA Vaccine BNT162b2 (BNT162b2 RNA) concentrate for solution for injection

Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine

Safety and immunogenicity of the ChAdOx1 nCoV-19 vaccine against SARS-CoV-2: a preliminary report of a phase 1/2, single-blind, randomised controlled trial

Safety and immunogenicity of ChAdOx1 nCoV-19 vaccine administered in a prime-boost regimen in young and old adults (COV002): a single-blind, randomised, controlled, phase 2/3 trial

Public Assessment Report Authorisation for Temporary Supply COVID-19 Vaccine AstraZeneca, solution for injection in multidose container COVID-19 Vaccine (ChAdOx1-S [recombinant]) Department of Health and Social Care (DHSC) AstraZeneca AB

Efficacy and Safety of the mRNA-1273 SARS-CoV-2 Vaccine

Public Assessment Report National Procedure COVID-19 Vaccine Moderna, 0.20 mg/mL dispersion for injection (COVID-19 mRNA Vaccine

/UKPAR_COVID_19_Vaccine_Moderna_07.04.2021_CMA_Reliance_PAR__-__final.pdf

Safety and Efficacy of Single-Dose Ad26.COV2.S Vaccine against Covid-19

Covid-19 vaccine: First person receives Pfizer jab in UK . BBC News

Press 19 13 Government of the United Kingdom. Coronavirus (COVID-19) in the UK -Cases by specimen date

Quantifying additional COVID-19 symptoms will save lives

Rapid implementation of mobile technology for real-time epidemiology of COVID-19

Vaccine side-effects and SARS-CoV-2 infection after vaccination in users of the COVID Symptom Study app in the UK: a prospective observational study. The Lancet Infectious Diseases 2021

Thrombotic Thrombocytopenia after ChAdOx1 nCov-19 Vaccination

Thrombosis and Thrombocytopenia after ChAdOx1 nCoV-19 Vaccination

Medicines & Healthcare products Regulatory Agency. Coronavirus vaccine -weekly summary of Yellow Card reporting

Signs and symptoms to determine if a patient presenting in primary care or hospital outpatient settings has COVID-19 disease

Real-time tracking of self-reported symptoms to predict potential COVID-19

Dynamics of Humoral Immunity Post COVID Vaccination -A Case Report

Testing for SARS-CoV-2 antibodies

Estimates of serial interval for COVID-19: A systematic review and metaanalysis

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

Anosmia and other SARS-CoV-2 positive test-associated symptoms, across three national, digital surveillance platforms as the COVID-19 pandemic and response unfolded: an observation study

Learning from Imbalanced Data Sets

Points of Significance: Sampling distributions and the bootstrap

The use of bootstrapping when using propensity-score matching without replacement: a simulation study

Pattern Recognition and Machine Learning

Post-vaccination SARS-CoV-2 infection: risk factors and illness profile in a prospective, observational community-based case-control study

Knowledge barriers in the symptomatic-COVID-19 testing programme in the UK: an observational study

Estimates of the rate of infection and asymptomatic COVID-19 disease in a population sample from SE England

Control of a Nosocomial Outbreak of COVID-19 in a University Hospital

Nosocomial COVID-19 infection: examining the risk of mortality. The COPE-Nosocomial Study (COVID in Older PEople)

Pandemic peak SARS-CoV-2 infection and seroconversion rates in London frontline health-care workers

The safety of anaesthetists and intensivists during the first COVID-19 surge supports extension of use of airborne protection PPE to ward staff

COVID-19 vaccine coverage in health-care workers in England and effectiveness of BNT162b2 mRNA vaccine against infection (SIREN): a prospective, multicentre, cohort study

Association Between Vaccination With BNT162b2 and Incidence of

ZOE Limited provided in-kind support for all aspects of building, running and supporting the app and service to all users worldwide. This work is supported by the Wellcome EPSRC Centre for Medical Engineering at King's 

LSC, MM and ELD contributed to study concept and design. CHS, JCP, BM, TS, CJS, SO contributed to acquisition of data. All the authors had access to the raw data underlying the study. LSC, JD, ELD contributed to data analysis and verified the underlying data. LSC, MO, MM and ELD contributed to drafting of the manuscript.All authors contributed to interpretation of data and critical revision of the manuscript. MM and ELD contributed to study supervision.

ELD and CJS report grants from the Chronic Disease Research Foundation (CDRF) during the conduct of the study. CH, SS, LP, AM report other from ZOE Limited, during the conduct of the study. TS is a scientific advisor to ZOE Limited. CHS reports grants from Alzheimer's Society, during the conduct of the study. SO reports grants from the Wellcome Trust, Innovate UK (UKRI), and Chronic Disease Research Foundation (CDRF), during the conduct of the study. . 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 22, 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. (which was not certified by peer review)The copyright holder for this preprint this version posted July 22, 2021. ; https://doi.org/10.1101/2021.07.21.21260906 doi: medRxiv preprint