key: cord-0937893-idq9n6xb authors: Stirrup, Oliver; Boshier, Florencia; Venturini, Cristina; Guerra-Assunção, José Afonso; Alcolea-Medina, Adela; Beckett, Angela; Charalampous, Themoula; da Silva Filipe, Ana; Glaysher, Sharon; Khan, Tabassum; Kulasegaran Shylini, Raghavendran; Kele, Beatrix; Monahan, Irene; Mollett, Guy; Parker, Matthew; Pelosi, Emanuela; Randell, Paul; Roy, Sunando; Taylor, Joshua; Weller, Sophie; Wilson-Davies, Eleri; Wade, Phillip; Williams, Rachel; Copas, Andrew; Cutino-Moguel, Maria-Teresa; Freemantle, Nick; Hayward, Andrew C; Holmes, Alison; Hughes, Joseph; Mahungu, Tabitha; Nebbia, Gaia; Partridge, David; Pope, Cassie; Price, James; Robson, Samuel; Saeed, Kordo; de Silva, Thushan; Snell, Luke; Thomson, Emma; Witney, Adam A; Breuer, Judith title: SARS-CoV-2 lineage B.1.1.7 is associated with greater disease severity among hospitalised women but not men: multicentre cohort study date: 2021-09-20 journal: BMJ Open Respir Res DOI: 10.1136/bmjresp-2021-001029 sha: 23d3409c9d5fcfd05b517a672611aa68157a6bfb doc_id: 937893 cord_uid: idq9n6xb BACKGROUND: SARS-CoV-2 lineage B.1.1.7 has been associated with an increased rate of transmission and disease severity among subjects testing positive in the community. Its impact on hospitalised patients is less well documented. METHODS: We collected viral sequences and clinical data of patients admitted with SARS-CoV-2 and hospital-onset COVID-19 infections (HOCIs), sampled 16 November 2020 to 10 January 2021, from eight hospitals participating in the COG-UK-HOCI study. Associations between the variant and the outcomes of all-cause mortality and intensive therapy unit (ITU) admission were evaluated using mixed effects Cox models adjusted by age, sex, comorbidities, care home residence, pregnancy and ethnicity. FINDINGS: Sequences were obtained from 2341 inpatients (HOCI cases=786) and analysis of clinical outcomes was carried out in 2147 inpatients with all data available. The HR for mortality of B.1.1.7 compared with other lineages was 1.01 (95% CI 0.79 to 1.28, p=0.94) and for ITU admission was 1.01 (95% CI 0.75 to 1.37, p=0.96). Analysis of sex-specific effects of B.1.1.7 identified increased risk of mortality (HR 1.30, 95% CI 0.95 to 1.78, p=0.096) and ITU admission (HR 1.82, 95% CI 1.15 to 2.90, p=0.011) in females infected with the variant but not males (mortality HR 0.82, 95% CI 0.61 to 1.10, p=0.177; ITU HR 0.74, 95% CI 0.52 to 1.04, p=0.086). INTERPRETATION: In common with smaller studies of patients hospitalised with SARS-CoV-2, we did not find an overall increase in mortality or ITU admission associated with B.1.1.7 compared with other lineages. However, women with B.1.1.7 may be at an increased risk of admission to intensive care and at modestly increased risk of mortality. The emergence of SARS-CoV-2 lineage B.1.1.7 in South East England has been found to be associated with an estimated 70% increased rate of community transmission compared with previously circulating variants. 1-3 Lineage B.1.1.7 became the dominant lineage in the UK in winter 2020/2021. It has also been detected in over 120 countries outside the UK, 4 and was assigned the label of variant of concern Alpha by WHO. 5 Lineage B.1.1.7 has acquired an unusually large number of mutations and deletions in a short period of time 1-3 ; specifically 14 Key messages ► In a multicentre cohort we did not find an overall increase in mortality or intensive therapy unit admission associated with B.1.1.7 among hospitalised patients, but women with B.1.1.7 may be at an increased risk of admission to intensive care and at modestly increased risk of mortality. ► The impact of SARS-CoV-2 lineage B.1.1.7 on disease severity appears to be dependent on the patient's sex, meaning that its health impacts and healthcare burden will differ from earlier lineages of the virus and that monitoring of disease course by sex should be considered for other viral lineages with increased transmissibility. Open access non-synonymous single nucleotide polymorphisms and 3 amino acid deletions, with 8 of these 17 amino acid changes occurring in the spike protein, responsible for receptor binding and a major immunogenic target. At least three of the spike protein changes are associated with in vitro biological changes. A tyrosine substitution at position 501 in the spike protein receptor binding domain has been shown to increase binding to the ACE2 receptor, while deletion of spike protein amino acids 69/70 reduces antibody neutralisation by convalescent sera. 6 7 The potential that so many mutations might change B.1.1.7 virulence has been examined epidemiologically using data largely from community-collected samples. [8] [9] [10] [11] However, there are few data on the impact of B.1.1.7 infection as compared with other variants on disease outcomes in hospitalised patients. We investigated the potential associations between the B.1.1.7 variant and the outcomes of mortality and intensive therapy unit (ITU) admission both in patients admitted with COVID-19 and hospital onset COVID-19 infections (HOCIs) in the COG-UK-HOCI study. The main objective was to estimate the overall effect of the variant on each of these outcomes, and we also evaluated whether the impact of the variant differed according to patient characteristics. Sequence and patient meta-data Data were collected from five NHS hospitals within London and three outside. The first SARS-CoV-2-positive sample from all inpatients tested through hospital laboratories between 16 November 2020 and 10 January 2021 was sequenced. In addition, metadata were collected from clinical records on patient age, sex (as binary M/F), comorbidities as identified by the COVID-19 Greenbook 12 (including obesity with body mass index ≥35 kg/ m 2 ), care home residence, pregnancy, ethnicity, date of hospital admission, ward location and first SARS-CoV-2positive test for all samples plus dates of admission to the ITU and all-cause death where these events occurred. Inpatients were classified as those admitted with SARS-CoV-2 plus cases which were identified after admission, with the latter termed HOCI cases and subdivided into indeterminate healthcare-associated infections (HCAIs) diagnosed 3-7 days after admission and probable/definite HCAIs diagnosed ≥8 days postadmission. 13 The primary outcomes for analysis were the events of death and of ITU admission. Events were included in the analysis within 28 days of hospital admission for those admitted with COVID-19 and within 28 days of diagnosis for HOCI cases. Samples were sequenced by Nanopore or Illumina methods as part of the COVID-19 Genomics UK Consortium (COG-UK). To maximise success, 4/8 labs sequenced only those samples with quantitative PCR cycle thresholds (ct) values of ≤32 or equivalent. Sequences were assigned to lineages using COG-UK Pangolin. 14 Only patients with admission to hospital and HOCIs were included in the statistical analysis of the clinical outcomes of mortality and ITU admission. Mortality and ITU admission were modelled as time-to-event outcomes, from time of hospital admission for those admitted with COVID-19 and from time of diagnosis for HOCI cases, censored at 28 days. Analyses of ITU admission were also censored at patient death. Both outcomes were censored at date of data collection for these variables for each site (between 15 January and 22 February 2021) . Mixed effects Cox models were used with adjustment for sex, patient age (using 5-knot restricted cubic spline), number of comorbidities (none, one, two, ≥three), care home residence, pregnancy, ethnicity (white, black, Asian, mixed or other) and sample week with separate parameters for London sites and for other sites grouped using the R package coxme V.2.2-16. 15 A 5-knot restricted cubic spline 16 was used for patient age in all analyses to allow flexibility in modelling the relationship with each outcome while maintaining a consistent model structure. Random intercept terms were included to reflect clustering of outcomes within hospital and weekly periods nested within hospitals. Cox models were stratified by HOCI status (allowing for different baseline hazard functions in patients admitted with COVID-19 vs HOCI groups). Outcomes were analysed on a complete case basis with regard to patient characteristics. This decision was based on the availability of complete data for >90% of patients and the fact that Cox regression gives asymptotically unbiased estimates of an association of interest as long as the missingness is not dependent on both outcome (ie, death or ITU admission) and exposure (B.1.1.7 status). 17 18 The variable of obesity was analysed as 'morbid obesity' versus 'no record of morbid obesity' on examination of case notes, and was included in statistical models within the ordinal comorbidities variable. The primary aim of the analysis was to estimate the overall association between the B.1.1.7 vs non-B.1.1.7 strain and the risk of each of the outcomes considered. Exploratory secondary analyses also evaluated interactions between B.1.1.7 status and patient characteristics in estimating the effect on each outcome. Analyses were conducted in R V.4.0.2, using tidyverse collection of packages with all plots generated using ggplot2 and survminer. [19] [20] [21] [22] Patient and public involvement This was an analysis of retrospectively collected data, for which the planning, data collection and analysis were all carried out during a period of extreme pressure on the UK health service due to the COVID-19 epidemic. As such, there was no patient and public engagement in the Open access conduct of this research and it would not be possible to disseminate the findings directly to participants. Between 16 November 2020 and 10 January 2021, SARS-CoV-2 RNA-positive upper respiratory tract samples from 2341 inpatients were sequenced from the eight participating sites (table 1 and online supplemental figure S1). Analysis of clinical outcomes was carried out in 2147 inpatients with all data available. The prevalence of lineage B.1.1.7 was highest in London and Hampshire (South of England), but substantially increased at all sites over the study period (online supplemental figure S2 ). Death within 28 days was reported in 527 (22.5%) of the 2341 patients. Death was recorded as having occurred following discharge with date of death missing in five, and these patients have been excluded from analyses. Death within 28 days was recorded in 494/2147 (23.0%) of the patients with all data available, with full 28 days of follow-up in 939/1653 (56.8%) of the other patients. On mixed effects multivariable Cox regression, the overall HR for mortality of lineage B.1.1.7 was 1.01 (95% CI 0.79 to 1.28, p=0.94) (figure 1, online supplemental table S1). Male sex was found to be a substantial risk factor for mortality (HR 1.46 vs female, 95% CI 1.22 to 1.75; p<0.001) and age was also strongly associated with the risk of death (figure 2). The risk of death was higher in care home residents (HR 1.39, 95% CI 1.02 to 1.90, p=0.04) and those with one or more significant comorbidities (HR 1.78 (95% CI 1.26 to 2.52) for one comorbidity, 2.03 (95% CI 1.43 to 2.88) for two and 2.89 (95% CI 2.04 to 4.08) for ≥three versus none; p<0.001). Those with ethnicity other than white were estimated to be at higher risk of death, but ethnicity was not a statistically significant predictor when evaluated over all categories (p=0. 36) . No pregnant women died and so this variable was dropped from the model as a perfect predictor. The addition of an interaction term between B. , online supplemental table S1 ). Within this model, male sex was a substantial risk factor for ITU admission (HR 1.33, 95% CI 1.05 to 1.68; p=0.02). Age was also strongly associated with the risk of ITU admission, although the relationship estimated was non-linear with the greatest risk of this outcome at 65 years of age (figure 2). The risk of ITU admission was higher in those with one or two significant comorbidities (HR 1.25 (95% CI 0.92 to 1.71) for one comorbidity, 1.24 (95% CI 0.89 to 1.74) for two and 0.79 (95% CI 0.54 to 1.15) for ≥three versus none; p=0.03). Those with ethnicity other than white were estimated to be at higher risk of ITU admission, but ethnicity was not a statistically significant predictor evaluated over all categories (p=0.09). Pregnant women were found to be at lower risk of ITU admission (HR 0.13, 95% CI 0.02 to 0.98, p=0.048). The addition of an interaction term between B. figure 4 (also provided separately for non-HOCI and HOCI inpatients in online supplemental figures S5 and S6, with HR estimates in online supplemental table S2). Our findings provide the largest dataset on disease severity in hospitalised patients with lineage B.1.1.7 and the only one based on routine sequencing of all specimens from multiple hospitals. The overall hazard of mortality and ITU were unchanged for patients with lineage B.1.1.7 in comparison to other viral variants (HR 1.01, 95% CI 0.79 to 1.28 and HR 1.01, 95% CI 0.75 to 1.37, respectively). These findings are in line with the results of a much smaller analysis of 341 (n=198 with B.1.1.7) hospital inpatients with viral sequencing over a similar time period in London, which found an adjusted mortality risk ratio for B.1.1.7 of 1.02 (95% CI 0.76 to 1.38). 23 However, in contrast with this smaller study we also found evidence that B.1.1.7 infection appears to have a different impact on the disease course according to sex among hospitalised patients with SARS-CoV-2 infection, with increased hazard of both mortality and ITU admission associated with the variant for female but not male patients. Several larger studies of disease severity in the UK have used PCR Spike (S) gene target failure (SGTF) as a surrogate for lineage B.1.1.7. [8] [9] [10] [11] These studies, based on community testing data, all found evidence of an overall increased risk of mortality associated with lineage B. 7 (B, D) . Models were all also adjusted by age using natural cubic splines (as shown in figure 2 ). Challen et al, 8 11 In the UK, SGTF is only available as a marker for a subset of those patients who were first positive for SARS-CoV-2 on testing within the community; most people who die of COVID-19 were not previously tested within the community 8 and the relevant PCR assay is not used by all laboratories, meaning that SGTF status is only available for 8.6% of deaths. 9 SGTF is an imperfect predictor of lineage B.1.1.7, and is much less accurate as a marker when prevalence of the variant is low (before mid-November 2020 in the UK). 24 The apparent overall differences in mortality risk observed in the SGTF analyses in comparison to our study do not necessarily represent inconsistent findings. Studies that are limited to patients who test positive in the community may be subject to selection biases linked to propensity to present for testing or rapidity of disease progression, while analyses that include only data from inpatients will not reflect the characteristics of the population as a whole. For example, increased disease severity may result in a higher proportion of subjects reaching the threshold for admission to hospital but not affect the mortality rate among those who are admitted to hospital. Our study also includes a subset of patients with probable nosocomial infection, whose characteristics and comorbidity profile differs greatly from the UK population as whole. 25 Individuals testing positive in the community for an SGTF-associated variant had higher risk of hospitalisation, with OR of 1.58 (95% CI 1.50 to 1.67). 26 This result was confirmed by a study of national health registerdata from Denmark including 18 499 patients with viral genomes available in the period 1 January to 9 February 2021 which found an adjusted OR of 1.64 (95% CI 1.32 to 2.04) for hospitalisation for B.1.1.7 compared with other lineages. 27 Taken together with the findings regarding mortality in the UK, [8] [9] [10] [11] these results are consistent with an increased risk of mortality and hospitalisation among patients testing positive for B.1.1.7 in the community but no overall increase in mortality among the subset of patients admitted to hospital. We found a significantly increased risk of both mortality (30%) and ITU admission (82%) in hospitalised female patients infected with B.1.1.7 but not in male patients. In contrast, studies of community-tested individuals found no interaction with sex for the effect of B.1.1.7 on mortality, 9 11 critical care admission 11 or risk of hospitalisation. 26 However, these studies were all conducted among patients who first tested positive for SARS-CoV-2 within the community, and therefore they cannot rule out an interaction with sex for the impact of B.1.1.7 on disease severity among all people infected with the virus or among all those admitted to hospital. Nationally collated data show that females accounted for 33.2% of patients admitted to ITU with COVID-19 in London, East and South East England between 1 September and 30 November 2020 rising to 36.2%, between 1 December 2020 and 21 January 2021 when lineage B.1.1.7 predominated. 28 There is evidence that the total number of pregnant women requiring intensive care in the UK was higher in late 2020/early 2021 in comparison to the first wave of the COVID-19 epidemic in Spring 2020, 29 and that the proportion of symptomatic pregnant women requiring admission to ITU increased as lineage B.1.1.7 became dominant. 30 This would be consistent with a differential impact of the B.1.1.7 variant on women in comparison to men. However, we should also note that our analyses found pregnancy itself to be negatively associated with Open access ITU admission, conditional on age, sex, viral variant, comorbidities and other patient characteristics. This could possibly be due to unobserved or residual confounding not fully captured by our recording and analysis of patient characteristics. An impact of lineage B.1.1.7 on females that is not observed in males could potentially be explained by physiological differences. For example, increased ACE2 expression in females has been posited as one explanation for the relatively lower mortality and morbidity observed for COVID-19 for women in comparison to men. 31 32 Lineage B.1.1.7 has mutations that increase binding of the viral spike protein to ACE2, thereby providing a plausible mechanism by which the new variant might have a differential effect on disease severity in males and females. 6 31 33 Our results suggest a reduction in the risk of mortality or ITU admission associated with B.1.1.7 in comparison to other viral lineages among male inpatients, although this finding was not definitive with HR 95% CIs that included no effect for both outcomes. However, this is against a backdrop of the established increased risk for male versus female patients for pre-B.1.1.7 strains. 34 Our results indicate an approximate equalisation of risk for otherwise equivalent male and female inpatients with B.1.1.7 infection. Although ours is substantially the largest study of hospitalised patients with confirmed lineage B.1.1.7 and non-B.1.1.7 SARS-Cov-2 infection, it has a number of limitations. Primarily, while evaluation of disease severity among only hospital inpatients can give useful Open access information on disease course and progression, analysis of only these patients cannot provide information on disease severity across all SARS-CoV-2 infections in the population as a whole. In addition, ITU admission can be difficult to interpret as a measure of disease severity among inpatients. For instance, admission to ITU may reflect the presence of severe disease and local decisions around the benefit or lack thereof to frail patients, which may be influenced by bed numbers and availability of respiratory support in non-critical care settings. Our primary analysis also includes cases of hospital-acquired infection but exclusion of these HOCI cases from our analyses yielded similar findings (online supplemental table S2). A further limitation of our analysis is that we do not have any information on vaccination status for individual patients. Our dataset covers a period in which a national vaccination programme was being initiated for HCWs and the elderly population in the UK, starting with those aged 80 years and above from 8 December 2020. This is a potential explanation for the observed protective interaction effect between care home residence and B.1.1.7 on mortality, as care home residents were prioritised for vaccination around the time that this viral variant was increasing in prevalence. Vaccine breakthrough infections are well described, particularly in partially vaccinated subjects. 35 Our analysis was focused on the effect of lineage B.1.1.7 as a single exposure of interest for each outcome variable. status, plotted according to patient sex and age categories. Date of sampling is used as the 'zero' time point for hospitalonset COVID-19 infections, with date of admission used for other patients. Naïve 95% CIs are plotted for illustrative purposes (these are not derived from the multilevel Cox models described). However, we also considered 5 potential interactions with viral lineage for the outcome of mortality and 6 for ITU admission, giving a total of 13 hypothesis tests of interest. A conservative use of the Bonferroni correction gives a p value cut-off adjusted to 0.0038 from the commonly used threshold of 0.05, using which we would consider there to be robust statistical evidence for an interaction with sex on the outcome of ITU admission but not for the interactions between sex or care home status on mortality. Although lineage B.1.1.7 was not associated with an overall increase in mortality among hospitalised patients, our investigation suggests that lineage B.1.1.7 may be associated with higher ITU admission and death in females compared with non-B.1.1.7 within this group. The dominance of lineage B.1.1.7 in the UK precluded further comparison with earlier non-B.1.1.7 variants, and there is now concern regarding the spread of other lineages in the UK and elsewhere. 36 There is a need for ongoing large-scale sequencing of SARS-CoV-2 cases linked to data on patient characteristics and outcomes in order to generate timely information regarding the associations between viral lineages and disease severity. Monitoring of disease course by sex should be considered for other viral lineages with increased transmissibility. Open access funded sequencing costs for the submitted work; NF reports grants from UKRI, during the conduct of the study; personal fees from Aimmune, personal fees from ALK, personal fees from AstraZeneca, personal fees from MSD, personal fees from Sanofi Aventis, personal fees from Novartis, personal fees from Ipsen, personal fees from Gedeon Richter, personal fees from Galderma, personal fees from Vertex, outside the submitted work. The remaining authors do not have any declarations of interest. All other authors declare no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years; and no other relationships or activities that could appear to have influenced the submitted work. Patient consent for publication Not required. Ethics approval Ethical approval for the HOCI study, with an amendment to cover the substudy reported in this research paper, is provided by REC 20/EE/0118. Provenance and peer review Not commissioned; externally peer reviewed. Data availability statement The sequence data analysed are included within publicly available datasets (https://www. cogconsortium. uk/ data/). However, due to data governance restrictions it is not possible to share the associated patient characteristics and clinical outcome data for the analysis described, as these are considered sensitive and full anonymisation is not possible. The corresponding author (OTS) affirms that the manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted and that any discrepancies from the study as planned have been explained. Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise. Open access This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https:// creativecommons. org/ licenses/ by/ 4. 0/. Estimated transmissibility and impact of SARS-CoV-2 lineage B.1.1.7 in England Lineage-specific growth of SARS-CoV-2 B.1.1.7 during the English national lockdown -SARS-CoV-2 coronavirus / nCoV-2019 Genomic Epidemiology Assessing transmissibility of SARS-CoV-2 lineage B.1.1.7 in England World Health Organization. Tracking SARS-CoV-2 variants Deep mutational scanning of SARS-CoV-2 receptor binding domain reveals constraints on folding and ACE2 binding SARS-CoV-2 evolution during treatment of chronic infection Risk of mortality in patients infected with SARS-CoV-2 variant of concern 202012/1: matched cohort study Increased mortality in community-tested cases of SARS-CoV-2 lineage B.1.1.7 Case fatality risk of the SARS-CoV-2 variant of concern B.1.1.7 in England Analysis of severe outcomes associated with the SARS-CoV-2 variant of concern 202012/01 in England using ICNARC case mix programme and QResearch databases COVID-19: the green book COVID-19: epidemiological definitions of outbreaks and clusters cov-lineages/pangolin: CoV-lineages 15 coxme: mixed effects cox models A comparison of methods to adjust for continuous covariates in the analysis of randomised trials Asymptotically unbiased estimation of exposure odds ratios in complete records logistic regression Accounting for missing data in statistical analyses: multiple imputation is not always the answer R Foundation for Statistical Computing. R: A language and environment for statistical computing ggplot2: elegant graphics for data analysis Welcome to the Tidyverse survminer: drawing survival curves using 'ggplot2 Genomic characteristics and clinical effect of the emergent SARS-CoV-2 B.1.1.7 lineage in London, UK: a whole-genome sequencing and hospital-based cohort study Reliability of spike gene target failure for ascertaining SARS-CoV-2 lineage B.1.1.7 prevalence in a hospital setting Healthcare-associated COVID-19 in England: a national data linkage study Risk of hospital admission for patients with SARS-CoV-2 variant B.1.1.7: cohort analysis Risk of hospitalisation associated with infection with SARS-CoV-2 lineage B.1.1.7 in Denmark: an observational cohort study ICNARC report on COVID-19 in critical care: England, Wales and Northern Ireland Were pregnant women more affected by COVID-19 in the second wave of the pandemic? Impact of SARS-CoV-2 variant on the severity of maternal infection and perinatal outcomes: data from the UK obstetric surveillance system national cohort ACE2 expression and sex disparity in COVID-19 Gender susceptibility to COVID-19: a review of the putative role of sex hormones and X chromosome The high infectivity of SARS-CoV-2 B.1.1.7 is associated with increased interaction force between Spike-ACE2 caused by the viral N501Y mutation Risk stratification of patients admitted to hospital with covid-19 using the ISARIC WHO clinical characterisation protocol: development and validation of the 4C mortality score Effectiveness of the BNT162b2 Covid-19 vaccine against the B.1.1.7 and B.1.351 variants Why is India having a covid-19 surge? Competing interests All authors have completed the ICMJE uniform disclosure form at www. icmje. org/ coi_ disclosure. pdf and declare: OTS has received funding for the submitted work through the COG-UK-HOCI study, funded by COG-UK consortium, supported by funding from UK Research & Innovation, National Institute of Health Research and Wellcome Sanger Institute; the COG-UK consortium