Association of ABO and Rhesus blood groups with severe outcomes from non‐SARS‐CoV‐2 respiratory infection: A prospective observational cohort study in Bristol, UK 2020–2022

Despite significant global morbidity associated with respiratory infection, there is a paucity of data examining the association between severity of non‐SARS‐CoV‐2 respiratory infection and blood group. We analysed a prospective cohort of adults hospitalised in Bristol, UK, from 1 August 2020 to 31 July 2022, including patients with acute respiratory infection (pneumonia [n = 1934] and non‐pneumonic lower respiratory tract infection [NP‐LRTI] [n = 1184]), a negative SARS‐CoV‐2 test and known blood group status. The likelihood of cardiovascular complication, survival and hospital admission length was assessed using regression models with group O and RhD‐negative status as reference groups. Group A and RhD‐positive were over‐represented in both pneumonia and NP‐LRTI compared to a first‐time donor population (p < 0.05 in all); contrastingly, group O was under‐represented. ABO group did not influence cardiovascular complication risk; however, RhD‐positive patients with pneumonia had a reduced odds ratio (OR) for cardiovascular complications (OR = 0.77 [95% CI = 0.59–0.98]). Compared to group O, group A individuals with NP‐LRTI were more likely to be discharged within 60 days (hazard ratio [HR] = 1.17 [95% CI = 1.03–1.33]), while group B with pneumonia was less likely (HR = 0.8 [95% CI = 0.66–0.96]). This analysis provides some evidence that blood group status may influence clinical outcome following respiratory infection, with group A having increased risk of hospitalisation and RhD‐positive patients having reduced cardiovascular complications.

One of the genetic markers often associated with disease is the ABO blood grouping system.Those with non-O blood groups tend to be more susceptible to disease, including venous thromboembolism (VTE) 5,6 while asthma and chronic obstructive pulmonary disease (COPD) are less common in blood group O individuals. 7While blood groups have been linked to non-communicable disease risk, 8 infectious diseases, such as cholera and malaria, have been suggested as selection pressures driving ABO blood group distribution variations across geographical locations. 5ecently, the influence of blood groups on disease has extended to the COVID-19 pandemic.5][16][17][18][19] However, there are few studies that examine the influence of blood group on other respiratory infections.In keeping with SARS-CoV-2, earlier small studies (n < 2800) suggested blood group A associations with acute respiratory distress syndrome (n = 732), 20 tuberculosis (n = 2760) 21 and death in bronchopneumonia (n = 400). 22A recent retrospective cohort study in Chinese blood donors (n = 26 597) found blood groups A and B had lower influenza and pneumonia risks respectively. 23An agonistic Scandinavian study (n = 5.1 million) found blood group A also was associated with lower influenza risk, but higher pneumonia risk, 24 whereas a Danish study (n = 3387) concluded that no single blood group indicated risk of lower respiratory tract infection (LRTI). 25o investigate whether the severity of respiratory infection not caused by SARS-CoV-2 is associated with blood group in Bristol, UK, we analysed data collected during the first 24 months of an ongoing prospective cohort study.We sought to determine whether hospitalisation of blood group O patients with pneumonic and non-pneumonic LRTI (NP-LRTI) had better clinical outcomes than patients with other ABO blood groups, as well as any potential effect of Rhesus status.

Study design
AvonCAP is a prospective observational cohort study undertaking comprehensive surveillance of adults admitted to two acute care hospitals encompassing all admissions in Bristol, UK (ISRCTN:17354061), with full details previously published. 26We analysed submission data from adults (≥18 years) hospitalised from 1 August 2020 to 31 July 2022 and screened patients for signs and symptoms of community-acquired respiratory disease.Hospital-acquired infection was excluded as an aetiology if infection was diagnosed after 48 h.In this analysis, we included only patients with SARS-CoV-2-negative respiratory infection and with a known blood group.
Demographic and clinical data were collected from patient records and recorded using REDCap. 27Data on pre-existing medical conditions at admission included the Charlson comorbidity index (CCI). 28The admission CURB-65 29 was derived for each hospitalisation.Patient outcomes were assessed on day 30 following admission, including cardiovascular complications, such as stroke or venous thromboembolism (VTE), intensive care admission (ICU) and all-cause mortality.

Ethics and permissions
The Health Research Authority Research Ethics Committee East of England, Essex, approved this study, reference 20/ EE/0157.Informed consent was obtained from cognisant patients, and declarations from consultees for individuals lacking capacity.Patients who declined consent were not included in this analysis.If it was not practical to approach individuals for consent, data were included under approval from the Clinical Advisory Group under Section 251 of the 2006 NHS Act.

Case definitions
SARS-CoV-2 infection was defined as PCR-positive test, using the established assay (Hologic Panther TMA) conducted by UKHSA diagnostic laboratories (RCP Path 2021).Patients with no molecular SARS-CoV-2 test (4.0%eligible cases) were excluded from analysis.Pneumonia was defined as acute respiratory illness with confirmed radiological changes compatible with infection or when diagnosed by the treating clinician.In keeping with NICE and BTS guidelines, patients assigned a diagnosis of pneumonia were counted as a pneumonia case even if a CXR was not taken or no infiltrate was seen, due to false-negative radiology occurring, for example, when consolidation is behind thoracic structures or in severe dehydration.NP-LRTI was classified as the presence of signs and symptoms of acute lower respiratory tract infection in the absence of infective radiological change and a clinical diagnosis of pneumonia.Full case definitions can be found in Data S1.
Cardiac complications were defined as the presence of one of the following within 30 days of admission: ST elevation myocardial infarction (STEMI), non-ST elevation myocardial infarction (NSTEMI), new episode of atrial fibrillation or other arrhythmias, stroke or brain haemorrhage, deep vein thrombus (DVT), pulmonary embolism (PE), or new or worsening congestive heart failure (CHF).

Study objectives
The primary objective of this analysis was to determine whether there is an association between blood group polymorphisms (ABO and Rhesus status) and severity of SARS-CoV-2-negative respiratory infection in hospitalised patients in Bristol, UK.

Statistical analysis
Statistical analyses were performed with R (v2022.12.0+353). 30here appropriate, ABO frequencies were compared using a multinomial exact goodness of fit against frequencies reported in the official statistics provided by NHS Blood and Transplant, comprising blood group distribution in England. 31This donor pool is restricted to first-time donors only, to ameliorate some biases and minimise the overrepresentation of group O and Rhesus-negative groups.Continuous variables were assessed with the Kruskal-Wallis test.Pneumonia and NP-LRTI were compared with multinomial exact goodness of fit, using NP-LRTI cohort as the expected probability.Cox proportional regression adjusted hazard ratios (aHR) were used to analyse ABO and Rhesus blood group influence on 30-day survival and survival to hospital discharge (up to 60 days.i.e. hospital length of stay).Those who died were censored from survival analysis for hospital discharge.Blood group O and RhD-negative status, respectively, were used as the reference groups.Where not mentioned, the reference condition was the absence of the assessed covariate.Age reference was the 18-34 years age group.Kaplan-Meier survival curves for survival and hospital stay can be found in supplementary data (Data S2  and S3).Associations between blood and Rhesus groups and likelihood of cardiac complications were assessed with logistic regression, adjusting for the covariates: age at admission, sex, pneumonia (in respiratory infection), total CCI score, CURB-65 score, smoking status and Caucasian ethnicity as appropriate.The age component was removed from the total CCI and CURB-65 scores.Logistic regression models including only blood groups and pneumonia (where appropriate) were compared to this full model with all covariates (Data S4).The strength of the association was assessed via the odds ratios (OR) and 95% confidence intervals.Multicollinearity of covariates was checked against a variance inflation factor <5 (Data S5), and model fits were compared with Akaike information criterion (AIC).Three pneumonia patients lacked information on hospital length of stay and were excluded from this analysis.The 45-54 years age group was removed from NP-LRTI mortality analysis due to insufficient sample size.

R E SU LTS
Of the 22 410 acute lower respiratory tract disease (aLRTD) admissions, 3118 aLRTD-related admissions occurred in patients with known blood group and a negative SARS-CoV-2 test (Figure 1): 1934 pneumonia and 1884 with NP-LRTI.The A, B, AB and O blood groups were found in 43.9%, 9.3%, 2.2% and 44.6% of the participants with respiratory infection, respectively; 82.8% were Rh-positive (Table 1).
Both pneumonia and NP-LRTI hospitalised cohorts had ABO blood group distributions that were significantly different from the blood donor population, driven by predominance of blood group A and under-representation of group O (Figure 2A) (multinomial goodness of fit, pneumonia p < 0.01; NP-LRTI p = 0.026).There was a greater ratio of RhD-positive patients hospitalised with both pneumonia and NP-LRTI compared with the ratio of the blood donor population.
Univariate analysis found that pneumonia patients had worse clinical outcomes compared to those with NP-LRTI, including higher rates of cardiovascular complications, ICU admission, 30-day mortality and longer hospital admissions (Data S6 and S7).There were fewer pneumonia patients with groups AB or B, and more with group A as compared with those in the NP-LRTI cohort (Data S7).More of the hospitalised cohort within both pneumonia and NP-LRTI were RhD-positive than expected based on the blood group representation in the blood donor population (Figure 2C; Data S7).The rate of stroke and venous thromboembolic disease (VTE) varied with patient blood group (Table 1; Data S6).Blood group A was more common in patients with cardiovascular complications than in the reference blood donor population (Figure 2D; Data S6 and S7).RhD positivity was also more among patients with respiratory infection who had cardiovascular complications (multinomial goodness of fit, p < 0.01) (Figure 2D; Data S6).Blood group distribution among patients admitted to ICU did not appear to differ significantly from the distribution of blood groups observed in the reference blood donor group, though this was a relatively smaller group of patients (n = 212).Among patients with pneumonia who died after 30 days, however, we did see evidence (p = 0.034) of a difference between the reference group, in which there were more group O and fewer group A patients (Figure 2F; Data S6).Data highlighting the proportion of patients in each ABO and Rhesus blood group suffering these clinical outcomes are shown in Data S8.
Multivariate analysis found few associations between ABO and Rhesus groups and respiratory infection (Figure 3A-C).There was suggestion of a protective effect of Rhesus-positive status against cardiovascular complications in patients with pneumonia (OR = 0.76 [95% CI = 0.59-0.98])(Figure 3B).Cox regression analyses suggested that blood groups did not appear to influence 30-day mortality in patients hospitalised with pneumonia and NP-LRTI (Figure 4A-C).The influence of blood group on survival until hospital discharge (i.e.hospital length of stay) varied across respiratory infections (Figure 5A-C): Patients with blood group A hospitalised with any respiratory infection, and when specified as NP-LRTI, were associated with a higher probability of discharge before 60 days (RI HR = 1.109 [95% CI 1.01-1.19),p = 0.031, NP-LRTI HR = 1.17 [95% CI 1.03-1.33],p = 0.018), whereas those with blood group B who had been hospitalised with pneumonia were less likely to have been discharged in this period (HR = 0.79 [95% CI 0.65-0.95],p = 0.012).Those with group B, however, had shorter time until hospital discharge in NP-LRTI (HR = 1.25 [95% CI 1.01-1.54],p = 0.04).
Sensitivity analyses were conducted with varied coefficients of the Cox model over time; the only difference of note was that group A no longer showed evidence of an influence on time to respiratory infection hospital discharge, and group B in NP-LRTI (Data S9).Repeating the survival until discharge analysis excluding ABO and Rhesus groups as covariates generated the same outcomes, apart from the 35-44 years age group reaching significance in a longer time to discharge in NP-LRTI (Data S10).

DISCUS SION
This is a unique study of a comprehensively assessed prospective cohort looking at the influence of ABO and Rhesus blood groups on hospitalisation and clinical outcomes in adults with respiratory infections other than SARS CoV-2.We found over-representation of blood group A (44% vs. 38%) and Rhesus-positive patients (83% vs. 75%) hospitalised with respiratory infection compared to a donor population.Once in hospital, Rhesus-positive status appeared to have a slight protective effect against cardiovascular complications in patients hospitalised with pneumonia.The data used in this study included those hospitalised with non-SARS-CoV-2 respiratory infection.Those entering hospitals are likely more symptomatic and may under-represent the likelihood of infection due to lack of surveillance of those with respiratory infection who do not encounter a healthcare worker.Moreover, the period of data collection was at the height of the COVID-19 pandemic (August 2020 to July 2022); this may confound under-representation of those with less severe respiratory infection and should be noted.We also observed that ABO groups were associated with the length of hospital stay but varied between pneumonia and NP-LRTI.Compared with group O, patients with blood type A hospitalised with any respiratory infection were more likely to have a shorter stay.Furthermore, group A and B NP-LRTI patients were both associated with a shorter stay.This was not evident in pneumonia patients, however.Indeed, group B individuals hospitalised with pneumonia appeared to have a higher risk of a longer hospital stay than those with group O.This discrepancy may be due to pathogen-specific effects.By contrast, Su et al. 23 found that blood patients with group B were less likely to be infected with pneumonia compared with patients with blood group O, but no significant associations between blood groups and NP-LRTI.Moreover,  no association between blood groups and length of hospitalisation in patients with any lower respiratory tract infection was found in their study.These contrasting observations between the two studies may be attributed to differences in population and location, with 87% of patients in this study being White Caucasian living in Bristol, and 99% in the study conducted by Su et al. being Han Chinese living in Shaanxi, China. 23reviously, group A has been associated with increased risk of ARDS, 20 Pseudomonas aeruginosa and smallpox, and non-O groups with respiratory disease hospitalisation and SARS infection. 32Parallel recent studies in SARS-CoV-2  T A B L E 1 (Continued) infection 18 suggest a universal mechanism by which people with group A are disadvantaged in infection and backing the consensus that non-O groups are more susceptible to disease. 5The increased binding of the SARS-CoV-2 receptor-binding domain to the group A antigen has been proposed as a possible mechanism which could explain the increased hospitalisation and poorer outcomes in group A patients with SARS-CoV-2 infection. 33Over the 2 years studied here, blood group A was over-represented, and O under-represented, among patients hospitalised with SARS-CoV-2-negative respiratory infection.However, limited evidence is available for other respiratory viruses binding the A antigen, and the significance of such binding is unclear.A more universal proposed mechanism centres on the clotting protein, von Willebrand Factor (VWF).Approximately 80% of individuals express ABO sugars on glycoproteins secreted in bodily fluids, and VWF is one such glycoprotein. 34leavage and clearance of plasma VWF secreted from activated endothelial cells during infection 35 may be impeded by the presence of A/B sugars, thereby resulting in higher levels of this clotting protein and increased VTE events. 36Thus, increased VWF observed in group A plasma may contribute to the increased risk of hospitalisation and poorer outcomes seen compared to group O, 13,37 and indeed, increased VWF levels have been reported in SARS-CoV-2 38,39 and non-SARS-CoV-2 40 respiratory infections.VWF levels have been reported to be high in SARS-CoV-2 pneumonia compared to non-SARS-CoV-2 pneumonia 41 ; though pneumonia still has a fivefold higher VTE risk within 1 year. 42Moreover, Streptococcus pneumoniae induces VWF release from pulmonary endothelial cells 43 and interacts with VWF itself. 44s 30% of the genetic variation in VWF is due to ABO, 34 it is feasible that changes in this clotting factor may influence the requirement for hospitalisation with respiratory infection due to altered physiological effects.
Rhesus-positive patients were over-represented in respiratory infection hospitalisation compared to the blood group abundance in the reference donor population, but we observed a negative association of Rhesus-positive status with cardiovascular complications in pneumonia.This analysis is, to our knowledge, the first to report an association of RhD outside of SARS-CoV-2 infection.Previous studies have reported that Rhesus-negative patients are protected against clinical outcomes in SARS-CoV-2 respiratory infection. 16However, the association of Rhesus factor with cardiovascular complications in SARS-CoV-2 infection has not been reported, due to limited power. 45verall, Rhesus blood group is less robustly associated with disease than ABO, 5 although Rhesus-negative individuals have poorer prognosis in ischaemic cardiomyopathy. 46It is therefore possible that Rhesus factor has a protective cardiovascular effect, which may be important in the preventing cardiac sequelae following respiratory infection.In contrast to A/B sugars, Rhesus blood group is solely present on red blood cells, and the defining RhD antigen likely shares functional redundancy with RhCE proteins in negative individuals. 5The role of RhD antigen is incompletely understood, and further research is required to elucidate its potential functions in the pathophysiology of infections.
While we did not observe any associations between ABO and Rhesus status with 30-day mortality in patients hospitalised with pneumonia and NP-LRTI, the influence of blood group on hospital length of stay varied across respiratory infection.Blood group A was associated with a reduced hospital length of stay, supportive of findings by Su et al. 23 in patients with influenza, despite its association with increased cardiovascular complications within 30 days of hospitalisation.These findings appear incongruous but may be explained if the cardiovascular complications occur after hospital discharge, due to the persistence in thrombosis risk following infection. 42nfection-specific mechanisms may be involved in the interaction with ABO groups, and A/B carbohydrates have been proposed as putative receptors for certain infections, including group A for SARS-CoV-2. 33Different pathogens may have different direct interactions with different blood group antigens or exert effects that vary with ABO groups, such as VWF; therefore, the causative pathogen may not only affect disease severity, and thus hospital length of stay, but also risk of cardiovascular complication associated with ABO group.These variations may contribute to the altered association of ABO group with clinical outcomes, and potentially explain these findings.
This study has many strengths, including the ability to include patient data obtained under Section 251 of the 2006 NHS Act to comprehensively ascertained respiratory infection in adults hospitalised within a defined geographical area.Hospital medical records were linked with community records to obtain detailed and accurate data for each study participant.Prospective, comprehensive case ascertainment within a defined geographical area remains the gold standard epidemiological methodology for ascertaining disease burden.This study does not rely on ICD-10 coding nor solely on national data linkage and assesses each case individually by gathering complete data.However, this study has some limitations, including the use of a donor cohort as the reference group for the blood group distribution.It is likely to be over-populated with more universal group O and Rhesus-negative individuals, given the nature of blood donation preferences.However, we restricted our donors to those who were first-time donors, which may reduce some biases.Information on blood group was only available in 29.5% of the entire hospitalised cohort, and this may be biased towards those with prior hospitalisations with transfusion risk; indeed, the mean total CCI score was higher in the known blood group cohort (4.32 ± 2.7 and 3.70 ± 2.6, 2 Sample t-test, p < 0.001).We were unable to obtain data on some potential confounders, including anticoagulant use; variations in treatment after hospitalisation may affect blood group influence, and the dose of anticoagulant needed may be greater in non-O groups. 47Furthermore, we found correlations between age and other variables (Data S11), although the absolute value of all correlation coefficients was <0.7, the threshold at which collinearity is thought to severely influence model predictions. 48Moreover, collinearity between the covariates in our study were lower than the threshold variance inflation factor of 5 (Data S5).Finally, we were unable to assess the associations of blood group and Rhesus status for individual pathogens as standard-of-care microbiological testing for non-SARS-CoV-2 infection was disrupted during the pandemic.Future data may be able to provide insight into pathogen-specific ABO group and Rhesus factor interactions.
In conclusion, we found evidence that ABO and Rhesus factor status may affect the risk of severe disease in patients hospitalised with respiratory infection.Understanding the roles of ABO groups in infection is important to improve patient care and may help inform treatment regimens based on varied risks with different groups.This may have applications for public health planning, including drives for certain blood group donations based on true hospitalisation needs.Further knowledge of the mechanisms by which ABO groups influence the body's milieu during infection will be essential to this end.teams for the support provided during this study.We would also like to acknowledge the research teams at North Bristol and University Hospitals of Bristol and Weston NHS Trusts for making this study possible, including Helen Lewis-White, Rebecca Smith, Rajeka Lazarus, Jane Blazeby, Diana Benton and David Wynick.Finally, we would like to thank Aman Kaur-Singh and Kevin Sweetland for their support with this study.

F U N DI NG I N FOR M AT ION
AvonCAP is a University of Bristol sponsored study which is investigator-led and funded under a collaborative agreement by Pfizer Inc.The study funder had no role in data collection, but collaborated in study design, data interpretation and analysis and writing this manuscript.The corresponding author had full access to all study data and had final responsibility for the publication decision.AH PhD studentship is supported by funding from the British Heart Foundation (FS/4yPhD/F/21/34162).

C ON F L IC T OF I N T E R E S T S TAT E M E N T
CH is Principal Investigator of the AvonCAP study which is an investigator-led University of Bristol study funded by Pfizer and has previously received support from the NIHR in an Academic Clinical Fellowship.JO is a Co-Investigator on the AvonCAP Study.LD is further supported by UKRI through the JUNIPER consortium (grant number MR/ V038613/1), MRC (grant number MC/PC/19067), EPSRC (EP/V051555/1 and The Alan Turing Institute, grant EP/ N510129/1).AF is a member of the UK Joint Committee on Vaccination and Immunization (JCVI) and chair of the World Health Organization European Technical Advisory Group of Experts on Immunization (ETAGE) committee.In addition to receiving funding from Pfizer as Chief Investigator of this study, he leads another project investigating transmission of respiratory bacteria in families jointly funded by Pfizer and the Gates Foundation and is an investigator in trials of COVID-19 vaccines including ChAdOx1nCOV-19, Janssen and Valneva vaccines.The other authors have no relevant conflicts of interest to declare.The AvonCAP study is a University of Bristol sponsored study which is investigatorled and funded under a collaborative agreement by Pfizer Inc.

DATA AVA I L A BI L I T Y S TAT E M E N T
The data used in this study are sensitive and cannot be made publicly available without breaching patient confidentiality rules.Therefore, individual participant data and a data dictionary are not available to other researchers.

F I G U R E 1
Study flow diagram.Of the 20 614 episodes of adults hospitalised with aLRTD during the study period, there were 3118 admissions with SARS-CoV-2-negative respiratory infection in patients with a known blood group: 1934 with pneumonia and 1184 admissions with NP-LRTI.aLRTD, acute lower respiratory tract disease; NP-LRTI, non-pneumonia lower respiratory tract infection.

b
VTE included deep vein thrombosis (DVT) and pulmonary embolic disease (PE).cCardiovascular complications included: ST-elevation myocardial infarction (STEMI), non-ST-elevation myocardial infarction (NSTEMI), new episodes of atrial fibrillation or other arrhythmias, stroke or brain haemorrhage, DVT, PE and new or worsening congestive heart failure.

F I G U R E 2
Hospital admission and clinical outcome patient distributions by blood group.Respiratory infection, pneumonia and non-pneumonia lower respiratory tract infection (NP-LRTI) multinominal exact goodness of fit plots against (A) ABO blood group alone, (B) ABO blood group and Rhesus group, (C) Rhesus group alone.Respiratory infection, pneumonia and NP-LRTI multinomial exact goodness of fit plots for ABO blood group distribution in (D) cardiovascular complications, (E) ICU admissions, (F) 30-day mortality.Clinical outcomes in patients hospitalised with SARS-CoV-2negative respiratory infection are shown in Data S6.F I G U R E 3 Logistic regression modelling of cardiovascular complications and ICU admission.Logistic regression odds ratios for cardiovascular complications in patients hospitalised with (A) acute lower respiratory tract disease and (B) pneumonia, and (C) ICU admission in patients hospitalised with acute lower respiratory tract disease.F I G U R E 4 Cox proportional hazard models of 30-day mortality.Cox proportional hazard model forest plots for 30-day mortality in patients hospitalised with (A) respiratory infection, (B) pneumonia and (C) non-pneumonia lower respiratory tract infection (NP-LRTI).F I G U R E 5 Cox proportional hazard models for hospital length of stay.Cox proportional hazard model forest plots for hospital length of stay in patients hospitalised with (A) respiratory infection, (B) pneumonia and (C) non-pneumonia lower respiratory tract infection (NP-LRTI).
Characteristics and outcomes of adults hospitalised with SARS-CoV-2-negative respiratory infection by ABO secretor status.