Abstract
Background There is an emerging understanding that coronavirus disease 2019 (COVID-19) is associated with increased incidence of pneumomediastinum. We aimed to determine its incidence among patients hospitalised with COVID-19 in the United Kingdom and describe factors associated with outcome.
Methods A structured survey of pneumomediastinum and its incidence was conducted from September 2020 to February 2021. United Kingdom-wide participation was solicited via respiratory research networks. Identified patients had SARS-CoV-2 infection and radiologically proven pneumomediastinum. The primary outcomes were to determine incidence of pneumomediastinum in COVID-19 and to investigate risk factors associated with patient mortality.
Results 377 cases of pneumomediastinum in COVID-19 were identified from 58 484 inpatients with COVID-19 at 53 hospitals during the study period, giving an incidence of 0.64%. Overall 120-day mortality in COVID-19 pneumomediastinum was 195/377 (51.7%). Pneumomediastinum in COVID-19 was associated with high rates of mechanical ventilation. 172/377 patients (45.6%) were mechanically ventilated at the point of diagnosis. Mechanical ventilation was the most important predictor of mortality in COVID-19 pneumomediastinum at the time of diagnosis and thereafter (p<0.001) along with increasing age (p<0.01) and diabetes mellitus (p=0.08). Switching patients from continuous positive airways pressure support to oxygen or high flow nasal oxygen after the diagnosis of pneumomediastinum was not associated with difference in mortality.
Conclusions Pneumomediastinum appears to be a marker of severe COVID-19 pneumonitis. The majority of patients in whom pneumomediastinum was identified had not been mechanically ventilated at the point of diagnosis.
Plain language summary
Pneumomediastinum is air around the heart and structures in the middle of the chest. This survey of hospitals from around the UK found that roughly 1 in every 160 patients (0.6%) admitted to hospital with COVID-19 had pneumomediastinum identified. Most of these patients were not mechanically ventilated when pneumomediastinum was diagnosed. However, by the end of their admissions more than three quarters (76.5%) of all patients with COVID-19 and pneumomediastinum who were eligible for mechanical ventilation had been mechanically ventilated. Half of all patients with COVID-19 and pneumomediastinum died. Pneumomediastinum occurs in patients with severe COVID-19 pneumonitis but it is not clear if pneumomediastinum is a contributory factor to the high death rate. There was no difference in outcome associated with removing patients from CPAP treatment after pneumomediastinum was identified.
Glossary of terms
ARDS=acute respiratory distress syndrome; BiPaP=conscious non-invasive bi-level positive airways pressure ventilation; COVID-19=coronavirus 2019 infection; CPAP=continuous positive airways pressure; CT=computed tomography; ECMO=extracorporeal membrane oxygenation; FiO2=fraction of inspired oxygen; HFNO=high flow nasal oxygen; Mechanical ventilation=invasive mechanical ventilation; PEEP=positive end expiratory pressure; PTM=pneumomediastinum; PPV=positive pressure ventilation; UK=United Kingdom
Introduction
Pneumomediastinum (PTM) is the abnormal presence of air or gas in the mediastinum. Spontaneous PTM is rare, appearing in approximately 1 in 33 000 hospital admissions [1]. PTM has a higher reported incidence among patients receiving positive pressure ventilation (PPV), particularly those with acute respiratory distress syndrome (ARDS) [2, 3].
The COVID-19 pandemic has seen a remarkable increase in the number of patients receiving PPV within a given period with many patients with COVID-19 pneumonitis meeting ARDS criteria [4, 5]. The publication of several case reports and small series of PTM in patients with COVID-19 could be viewed in this context [5–8]. There have however, been a number of reports of PTM occurring in COVID-19 pneumonitis without positive pressure ventilation [9–11]. The true incidence of PTM in COVID-19 and its relationship to PPV remains unclear. In addition, whether management should be altered after the identification of PTM is not known.
We report a multi-centre observational study of 377 cases of COVID-19 PTM from 53 hospitals in the United Kingdom between September 2020 and February 2021. We describe the incidence and risk factors associated with PTM in COVID-19 and associations with mortality.
Materials and methods
Study population
The study recruited across the United Kingdom (UK). It was advertised via national and regional trainee research networks including Pulmonary Research Inter-Site Matrix (PRISM) and North West Collaborative Respiratory Research (NCORR).
Participating institutions contributed cases of PTM in inpatients with COVID-19 identified between 01/09/2020 to 31/01/2021. The diagnosis of PTM was based on a computed tomography (CT) or plain radiograph of chest and the diagnosis of COVID-19 was based on a positive SARS-CoV-2 PCR result or evidence of COVID-19 pneumonitis on CT imaging and a clear clinical history. All participating institutions searched radiology reports using the keywords “pneumomediastinum”, “pneumothorax” or “subcutaneous emphysema”, and patient lists from medical and respiratory wards and intensive care units to ensure all cases were identified. Anonymized data were collected for each case. These included; demographics; past medical history; radiological findings; clinical outcomes and respiratory settings from all respiratory support prior to and after diagnosis of PTM. Follow up at 120 days or more was recorded for all patients and all cases of interhospital transfer were cross-checked to ensure no duplication. In order to accurately estimate incidence data were collected on the total numbers of patients admitted during this period who were coronavirus positive on SARS-CoV-2 PCR testing and the proportion who underwent CT imaging of chest at each institution.
All data pertaining to fraction of inspired oxygen (FiO2) were normalised to a uniform scale prior to analyses. This is described in online supplementary table S2 (e.g., all patients receiving 15 L of oxygen via a non-rebreather mask with reservoir were assigned an inspired FiO2 of 90%). A variety of devices were used to deliver high flow nasal oxygen (HFNO) and continuous positive airways pressure (CPAP). Positive end expiratory pressure (PEEP) and FiO2 for patients receiving CPAP were normalised to values based on data from the Association of Respiratory Technology and Physiology (2020). PEEP for patients on HFNO was estimated and normalised based on published physiological data from Groves & Tobin (2007) [12]. Details of this can be found in online supplementary tables S3a and S3b and supplementary figure S3c.
Ethical approval was obtained from the Oxford University Hospitals NHS Foundation Trust (UK) audit committee with additional ethical approval from local NHS Trusts where relevant. All data were collected retrospectively and entered by local physicians in anonymized fashion without linkage to patient identifiers.
Statistical methods
Patient data are presented as frequencies and percentages for categorical data and mean (±sd) for continuous data. Where data were not normally distributed the median and interquartile range (IQR) are presented. Differences in categorical data are presented using the chi-square test. For comparisons of continuous normally distributed data 2-sided independent t-tests were used. All outcomes quoted are outcomes at 120 days. Variable entry into regression models was performed backward stepwise. Analyses were performed using SPSS 28 (SPSS, Chicago, IL, USA). Figures were created in R version 4.0.3.
We present the following article in accordance with the STROBE reporting checklist.
Results
A total of 377 cases of pneumomediastinum were detected of whom 98.4% had a positive SARS-CoV-2 PCR and the remainder were diagnosed clinically. The diagnosis of PTM was made or confirmed on CT-scan of chest in 318 cases (84.4%). For 147/318 (46.2%) of the cases diagnosed by CT scan, PTM had not been visible on a preceding chest radiograph. Outcome data were obtained for all patients and incidence data from all included hospitals. All other data was≥95% complete for all parameters.
Incidence
There were 58 484 PCR positive inpatient admissions for the period September 1, 2020 to January 31, 2021 within the 53 participating hospitals (mean 1103±611 COVID-19 inpatients per hospital). The incidence of PTM was 0.64% (95% CI 0.58%–0.71%) per COVID-19 inpatient admission with a mean number of PTM cases of 7.1±4.8 per hospital. 12 703 of the 58 484 PCR positive inpatients (21.7%) underwent CT imaging of chest during admission with a mean number of CT scans performed per hospital of 240±200. The relationship between the number of total inpatient admissions, use of CT imaging of chest and the number of cases of PTM across hospitals is presented in online supplementary figure S4.
Demographics
The median age was 60 years (IQR 52–78). Male patients were over-represented with 277 (73.5%) male to 100 (26.5%) female patients. The most prevalent medical comorbidities were: hypertension (32.4%), diabetes mellitus (21.5%), asthma (19.4%), obesity (10.6%), ischaemic heart disease/left ventricular systolic dysfunction (8.8%) and chronic kidney disease (4.5%). Three patients were pregnant. The median duration of symptoms prior to admission to hospital was 7 days (IQR 5–10) while the median duration from admission to the identification of PTM on imaging was 7 days (IQR 5–12.3). Chest pain was a feature of the presenting complaint in 11.9% of patients.
315/377 (83.6%) patients were considered eligible for mechanical ventilation should it be required. Eligibility for mechanical ventilation was a clinical decision recorded in the notes. Treatment of the remaining 62 patients was limited to CPAP support should it be required. Given the differences in management between these two groups they are considered separately in our regression analyses examining factors linked to patient outcome.
Management
241/315 (76.5%) of patients considered eligible for escalation to mechanical ventilation were mechanically ventilated at some point during their admission. 172 of these 241 (71.4%) patients were mechanically ventilated prior to the diagnosis of PTM. PTM was detected within 24 h following intubation in 38/241 (15.8%) of these patients. 24/241 (10.0%) of these patients went on to receive extracorporeal membrane oxygenation (ECMO). Of the 377 patients eligible to receive CPAP 256/377 (67.9%) had received CPAP prior to the diagnosis of PTM. Conscious non- invasive bi-level positive airways pressure ventilation (BiPaP) was used at some point in the admissions of 9/377 (2.4%) patients, other than its use in weaning patients from mechanical ventilation. Four of these nine patients were in type two respiratory failure. The indication for use of BiPaP for the other five patients was not clear. Given the few patients who received BiPaP and the variation in its use we have excluded it from our analyses.
The maximum respiratory support provided to all patients before and after diagnosis of PTM is described in figure 1. Four patients whose treatment was limited to non invasive respiratory support were switched from CPAP to Oxygen at the point of diagnosis of PTM as part of a decision to initiate palliative treatment. Two patients were managed on room air throughout.
Alteration of respiratory support at the time of diagnosis is illustrated in figure 1. Most patients whose respiratory support was changed after diagnosis of PTM were on CPAP. At the point of diagnosis of PTM 93 patients eligible for mechanical ventilation were on CPAP. Fifty (53.8%) of these patients were switched immediately on diagnosis of PTM to either Oxygen or HFNO therapy creating two subgroups amenable to analysis; the 50 switched to Oxygen or HFNO and the 43 continuing on CPAP. These two subgroups were retrospectively well matched at the point of diagnosis by age (CPAP mean age 57.0 years versus Oxygen or HFNO 55.6 years, p=0.51), by the maximum FiO2 they had received (CPAP mean FIO2 66% versus Oxygen or HFNO 68%, p=0.15) or by the maximum PEEP they had received (CPAP mean PEEP 10.4cmH20 versus Oxygen or HFNO 9.8cmH20, p=0.19). The subsequent trajectory of these two subgroups is illustrated in figure 2. Associations of change in mode of respiratory support and mortality for these patients was examined by ANOVA. There was no significant main effect of switching support from CPAP to Oxygen or HFNO on outcome. There was however, a main effect of mechanical ventilation as a factor associated with mortality for both subgroups (p<0.001).
Co-occurrence of pneumothorax, subcutaneous emphysema and complications associated with pneumomediastinum
Pneumothorax was seen concurrently in 154/377 patients (40.8%) and subcutaneous emphysema was seen in 280/377 (74.3%) of patients. The co-occurrence of pneumomediastinum with pneumothorax, subcutaneous emphysema and tension phenomena and the use of intercostal drains are displayed in figure 3. The number and frequencies of intercostal chest drains inserted are presented in online supplementary figure S5. In cases associated with subcutaneous emphysema, subcutaneous drains were employed in 6 (1.6%) cases. In 5 of these 6 cases subcutaneous drains were inserted for threatened or actual tension subcutaneous emphysema. There were 4 (1.1%) instances of mediastinal drains being used. In 1 of these 4 cases the mediastinal drain was inserted as an emergency bedside procedure for suspected tension PTM and tension subcutaneous emphysema. In the other 3/4 cases the mediastinal drain was inserted to obviate possible tension PTM. These four mediastinal drains were inserted in patients at four different hospitals, each without on-site cardiothoracic services. There were 14 cases of tension pneumothorax. During 10 cases of suspected tension phenomena bilateral intercostal drains were inserted as an emergency procedure. Seven of these 10 cases were performed without prior radiographic evidence of pneumothorax.
The development of pneumothorax was not associated with increased risk of death for our cohort (table 1) including the subset of 16 patients who were mechanically ventilated before pneumothorax developed [11/116 (9.5%) were among those patients who subsequently died while 5/56 (8.9%) were among those discharged, p=0.9] There were two cases of pneumoperitoneum. Both of these cases were in mechanically ventilated patients.
In 8 cases PTM appeared following an interventional procedure that could potentially represent a separate mechanism for occurrence e.g., tracheostomy, and these cases are included in the final analysis. Analyses were performed excluding these cases without any statistically significant deviation from the results presented.
Mortality
At 120 days from admission 175/377 (46.4%) patients had been discharged and 195/377 (51.7%) patients had died. Of the seven patients still in hospital at 120 days, at time of writing one patient had died on day 162 of their admission. Three patients remained mechanically ventilated on days 131, 146 and 150 of admission. One patient had been extubated but remained within intensive care on day 132 of admission. These 5 patients were categorised with those who had died at 120 days in all outcome analyses. The remaining two patients were medically fit for discharge to rehab facilities at days 137 and 149 of admission. They were categorised with patients discharged at 120 days in all outcome analyses. A breakdown of mortality is provided in online supplementary tables S6a and S6b according to whether patients were eligible for mechanical ventilation or limited to CPAP support.
Factors of the presentation and association with outcome are presented for all patients in table 1. All factors significantly associated with mortality in univariate analyses were entered into binary regression prediction models with the exception of the use of ECMO which, was excluded as the direction of association for this variable was in favour of discharge rather than death. The variable “radiographic progression of pneumomediastinum” was excluded where the model was conducted from the point of diagnosis.
A regression model comparing the predictive utility of variables for mortality at 120 days from the point of diagnosis for patients eligible for all treatment is presented in table 2. Further models looking at the predictive utility of the same variables for mortality across the duration of hospital admission are presented for patients eligible for all treatment in online supplementary table S7 and for those limited to CPAP in online supplementary table S8.
Discussion
These data comprise the largest series of PTM in COVID-19 to date. In comparison with other series we sought to accurately represent the incidence of PTM in COVID-19 during the period of the survey – the United Kingdom's “second wave” of the pandemic. Hospital records and radiology reports were systematically reviewed in each centre. Hospitals that did not observe cases of PTM but provided accurate incidence data were included. However, hospital participation was sought via trainee research networks and this may have resulted in inclusion bias.
Our estimate of incidence is also subject to diagnostic biases. We identified cases through radiology reports which, may not always reference a relevant finding. The main mode of diagnosis of PTM was CT imaging and there was considerable variation in the use of CT by participating hospitals (online supplementary figure S4). Many CT scans of chest were pulmonary angiogram studies assaying for pulmonary emboli, not for PTM. For 46.2% of the patients diagnosed with PTM on thoracic CT the PTM was not visible on their preceding chest radiograph. As only 21.7% of our total denominator population of 58 484 COVID-19 positive inpatients had thoracic CT imaging performed during their admissions, there is likely to be a number of undetected cases of PTM in our denominator population. These unknown cases may have had a more benign disease trajectory than the cases identified.
With these caveats these data demonstrate an incidence of PTM in COVID-19 of 0.64% per inpatient admission and 3.0% per COVID-19 inpatients undergoing thoracic CT. This incidence is similar to rates reported by two other studies of PTM in hospitalised COVID-19 populations from Brazil and Romania, of 0.51% and 0.67% respectively [13, 14]. The incidence of “spontaneous” PTM in COVID-19 in this cohort i.e., without any PPV via mechanical ventilation or CPAP, was 77/58 484 (0.13%). This is much higher than estimated background rates of non-COVID-19 “spontaneous” PTM. The largest study of non-COVID-19 “spontaneous” PTM in the literature with a defined denominator population, identified 41 cases of PTM from 1 824 967 emergency department admissions over 16 years (0.00002%) [15].
The mean age of the cohort (59.1 years) is consistent with inpatient international COVID-19 PTM cohorts from Brazil, Romania, Turkey, Pakistan and the USA [13, 16–19]. It is somewhat younger than the mean age of general COVID-19 inpatients in the UK, according to the largest epidemiological study (70.4 years) [20]. There could be pathophysiological reasons why COVID-19 inpatients who develop PTM are younger than the hospital population average (we note that background rates of non-COVID-19 PTM typically occur in younger adults) [1, 14–15]. It could reflect bias towards more frequent imaging in younger patients who are usually eligible for all treatments, with an artificial reduction in the identification of PTM in older patient groups. A younger mean age is also representative of trends in patients hospitalised with COVID-19 during the “second wave” in the UK [21].
Pneumothorax was found to co-exist with PTM in 40.3% of cases. This compares to reported rates of between 20.0% and 72.7% in other series with more than 10 patients [6, 13, 16–18, 22]. There was no finding of an effect on mortality of pneumothorax within this cohort, nor specifically for those patients who were mechanically ventilated when pneumothorax occurred. This contrasts with the findings of Marciniak et al. [23] who report an increased risk of mortality with COVID-19 pneumothorax in a large dataset of UK inpatients, and Chopra et al. [24] who found increased mortality in mechanically ventilated patients across four intensive care units in the USA. As concurrent PTM was not reported in the Marciniak et al. study and was relevant to 30% of the patients in the Chopra et al. study it is not clear how comparable these patient groups are to our cohort. The extent to which pneumothorax and PTM are manifestations of barotrauma in COVID-19 underwritten by a pathophysiological process and the extent to which they are distinct entities remains to be determined.
Subcutaneous emphysema was seen in 77.9% of COVID-19 PTM patients. Subcutaneous emphysema has been documented at rates of between 63.6% and 90.5% in other COVID-19 PTM series with more than 10 patients [13, 16–17]. This result is in keeping with high reported rates of subcutaneous emphysema in spontaneous non-COVID PTM of up to 100% [1] and in excess of lower rates of co-occurrence between subcutaneous emphysema and non-COVID-19 pneumothorax of up to 20% [25]. It would suggest that subcutaneous emphysema is a feature strongly associated with PTM and not specifically to COVID-19 PTM. It is acknowledged however, that co-occurrence of subcutaneous emphysema and PTM may be subject to diagnostic bias with patients presenting with subcutaneous emphysema more likely to have CT imaging and subsequent revealing of a diagnosis of PTM.
It is not possible to determine the effect of different ventilatory strategies on outcome within an observational study such as this. However, we examined this for those patients eligible for mechanical ventilation who were on CPAP when PTM was diagnosed. The role of CPAP in patients with PTM is a clinically important question: Analysis of changes in respiratory support after diagnosis of PTM permits an exploration of physician preferences regarding respiratory support, and by inference use of PEEP, in PTM. Those patients who remained on CPAP immediately after diagnosis of PTM were retrospectively well matched with those patients who were switched immediately to Oxygen or HFNO by age, maximum FiO2 and maximum PEEP. There was no difference in survival at 120 days between these subgroups. Thus, the current data do not support a policy of taking patients off CPAP when PTM is diagnosed, although we acknowledge potential confounders.
The 120-day mortality rate for patients with COVID-19 PTM of 51.7% is in keeping with reported mortality rates of 47.7% - 72.2% in other COVID-19 PTM cohorts [13, 16, 17]. The severity of COVID-19 illness is demonstrated by the high mean levels of FiO2 and PEEP before and after the diagnosis of PTM was made (fig. 1). Only two patients (0.5%) were managed on room air throughout admission. The number of patients who were mechanically ventilated at some point during their admission was remarkable at 76.5% of those eligible, in comparison to the UK average for mechanical ventilation of COVID-19 inpatients of 8.8% [20]. Mechanical ventilation was unsurprisingly an important prognostic factor and dominant variable in outcome prediction models (table 2). It is a ubiquitous event in the trajectory of a deteriorating patient eligible for this support. Only one eligible patient in our cohort died without having been mechanical ventilated.
High rates of mechanical ventilation in COVID-19 PTM have been reported in other general hospital inpatient COVID-19 studies [13, 16]. This may reflect a confounding relationship between more severe illness and higher rates of CT scanning and detection in high-care environments. It may also indicate an important role for mechanical ventilation in the development of PTM in COVID-19. However, the majority of this cohort, 205/377 patients (54.4%), were not mechanically ventilated at the point the diagnosis of PTM was made. Mechanical ventilation was therefore not a sufficient or necessary mechanism of PTM for the majority of patients.
Different mechanisms of PTM are described in the literature, including posterior membrane tracheal lesion or rupture due to coughing [26]. The “Macklin effect” [27] describes PTM secondary to the rupture of marginal alveoli due to a steeply increased pressure gradient between the alveolus and the interstitial space. After rupture of the alveolus air dissects centripetally along the sheaths of the broncho-vascular bundles into the mediastinum. Depending on volume and pressure, air can be decompressed along cervical fascial planes into the subcutaneous tissues of the chest wall, neck or face. Air may rupture the relatively thin mediastinal pleura to enter the pleural space causing unilateral or bilateral pneumothorax and /or pneumopericardium/pneumoperitoneum. Macklin and Macklin believed the effect could be benign or result in circulatory collapse if air directly compressed the pericardium or venous return – a tension PTM or pneumothorax . Air in the broncho-vascular bundles could also have a pernicious splinting effect leading to hyperinflation and low compliance with vascular compression and poor gas exchange, “malignant interstitial emphysema”. The “Macklin effect” was inspired by physician descriptions of “pulmonary interstitial emphysema” in patients suffering severe respiratory illness during the 1918–20 influenza pandemic [28, 29], the pathophysiology of which may bear comparison with the COVID-19 pandemic.
The “Macklin effect” offers a plausible mechanism for PTM in COVID-19 whereby the pneumonitis creates an altered diathesis for the rupture of alveoli and the emergence of PTM. The proposition that COVID-19 PTM patients have severe pneumonitis is supported by cohort studies that describe high radiological scores of pneumonitis in COVID-19 PTM [10, 16–18]. The complimentary findings in our cohort of; high levels of respiratory support (taken to represent severe pneumonitis); high rates of subcutaneous emphysema; episodes of tension phenomena, and low rates of chest pain (compared to spontaneous PTM) support the Macklin effect as the likely mechanism of PTM in COVID-19. The previous 2002–4 SARS epidemic also saw an increase in case reports of PTM [30] and this may reflect similar pathophysiology.
Future studies among mechanically ventilated patients with COVID-19 may elucidate whether strategies which modify trans-alveolar pressure have any association with development or progression of PTM. We notice that the 40/377 (10.6%) patients in our cohort with obesity were not at increased risk of death compared to other patients with PTM and speculate whether this could relate to mass loading around the chest wall and/or abdomen with reduction of alveolar compliance and/or trans-alveolar gradients. Propensity matched cohort analyses may address whether the development of PTM confers increased mortality risk, beyond severe pneumonitis, or whether development of PTM is affected by disease modifying drugs such as dexamethasone (standard of care for our cohort) or different variants of coronavirus 2019.
In summary this study is the largest reported series of PTM in COVID-19 disease. PTM appears to be a marker of severe pneumonitis, and not necessarily as a result of the use of PPV. There was no evidence of increased harm by continuing CPAP in COVID-19 patients who developed PTM.
Acknowledgements
We would like to thank Pallav Shah, Lupei Cai, Muhammad Tariq, Benjamin Jones, and Emma Helm for their help with data collection, Maria Tsakok and Nick Tessier for assistance with imaging and Simon Couillard and Sanjay Ramakrishnan for review of the manuscript. The research was funded by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.
Footnotes
Data Sharing: De-identified participant data from the study will be made available with publication to medical researchers on a not for profit basis by email request to the corresponding author for the purposes of propensity matching or meta-analysis.
Author Contributions: J.M conceived and designed the study, collated and analysed data and drafted the manuscript. A.A and R.J.H contributed data, contributed to study design and data analysis and drafted the manuscript. F.C, E.T, E.S, N. P, J.D, P.N, B.V, M.T, C.D, G.L, V.A, T.J, A.A, B.I, A.H, V.K, E.R, A.Y.K.C.N, S.U, M.S, N.M, P.E, M.N, F.A, G.D, M.B, C.M, M.I, M.B, N.H, Y.M, R.H, H.B, A.A, K.J, G.T, I.Z, D.S, N.M, S.K, H.I, J.N, H.W, C.C, H.L, R.S, N.J, A.G, T.S, K.B, C.R, B.P, E.B, G.W, R.W, R,T, K.M, N.B, V.K, S.N, K.S, M.A.K, A.I.M, Y.M.M.M, H.G, B.L & A.R, contributed data and revised the manuscript for important intellectual content, W.L designed the figures and carried out statistical analyses and N.M.R helped design the study and revised the manuscript for important intellectual content
Support statement: The research was funded by the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC).
Conflict of interest: James Melhorn has nothing to disclose.
Conflict of interest: Andrew Achaiah has nothing to disclose.
Conflict of interest: Francesca M. Conway has nothing to disclose.
Conflict of interest: Elizabeth M. F. Thompson has nothing to disclose.
Conflict of interest: Erik W. Skyllberg has nothing to disclose.
Conflict of interest: Joseph Durrant has nothing to disclose.
Conflict of interest: Neda A. Hasan has nothing to disclose.
Conflict of interest: Yasser Madani has nothing to disclose.
Conflict of interest: Prasheena Naran has nothing to disclose.
Conflict of interest: Bavithra Vijayakumar has nothing to disclose.
Conflict of interest: Matthew J. Tate has nothing to disclose.
Conflict of interest: Gareth E. Trevelyan has nothing to disclose.
Conflict of interest: Irfan Zaki has nothing to disclose.
Conflict of interest: Catherine A. Doig has nothing to disclose.
Conflict of interest: Geraldine Lynch has nothing to disclose.
Conflict of interest: Gill Warwick has nothing to disclose.
Conflict of interest: Avinash Aujayeb has nothing to disclose.
Conflict of interest: Karl A. Jackson has nothing to disclose.
Conflict of interest: Hina Iftikhar has nothing to disclose.
Conflict of interest: Jonathan H. Noble has nothing to disclose.
Conflict of interest: Anthony Y. K. C. Ng has nothing to disclose.
Conflict of interest: Mark Nugent has nothing to disclose.
Conflict of interest: Philip J. Evans has nothing to disclose.
Conflict of interest: A. Hastings has nothing to disclose.
Conflict of interest: Harry R. Bellenberg has nothing to disclose.
Conflict of interest: Hannah Lawrence has nothing to disclose.
Conflict of interest: Rachel L. Saville has nothing to disclose.
Conflict of interest: Nikolas T. Johl has nothing to disclose.
Conflict of interest: Adam N. Grey has nothing to disclose.
Conflict of interest: Huw C. Ellis has nothing to disclose.
Conflict of interest: Cheng Chen has nothing to disclose.
Conflict of interest: Thomas L. Jones has nothing to disclose.
Conflict of interest: Nadeem Maddekar has nothing to disclose.
Conflict of interest: Shahul Leyakathali Khan has nothing to disclose.
Conflict of interest: Ambreen Iqbal Muhammad has nothing to disclose.
Conflict of interest: Hakim Ghani has nothing to disclose.
Conflict of interest: Yadee Maung Maung Myint has nothing to disclose.
Conflict of interest: Cecillia Rafique has nothing to disclose.
Conflict of interest: Benjamin J. Pippard has nothing to disclose.
Conflict of interest: Benjamin R. H. Irving has nothing to disclose.
Conflict of interest: Fawad Ali has nothing to disclose.
Conflict of interest: Viola H. Asimba has nothing to disclose.
Conflict of interest: Aqeem Azam has nothing to disclose.
Conflict of interest: Eleanor C. Barton has nothing to disclose.
Conflict of interest: Malvika Bhatnagar has nothing to disclose.
Conflict of interest: Matthew P. Blackburn has nothing to disclose.
Conflict of interest: Kate J. Millington has nothing to disclose.
Conflict of interest: Nicholas J. Budhram has nothing to disclose.
Conflict of interest: Katherine L. Bunclark has nothing to disclose.
Conflict of interest: Toshit P. Sapkal has nothing to disclose.
Conflict of interest: Giles Dixon has nothing to disclose.
Conflict of interest: Andrew J. E. Harries has nothing to disclose.
Conflict of interest: Mohammad Ijaz has nothing to disclose.
Conflict of interest: Vijayalakshmi Karunanithi has nothing to disclose.
Conflict of interest: Samir Naik has nothing to disclose.
Conflict of interest: Malik Aamaz Khan has nothing to disclose.
Conflict of interest: Karishma Savlani has nothing to disclose.
Conflict of interest: Vimal Kumar has nothing to disclose.
Conflict of interest: Beatriz Lara Gallego has nothing to disclose.
Conflict of interest: Noor A. Mahdi has nothing to disclose.
Conflict of interest: Caitlin Morgan has nothing to disclose.
Conflict of interest: Neena Patel has nothing to disclose.
Conflict of interest: Elen W. Rowlands has nothing to disclose.
Conflict of interest: Matthew S. Steward has nothing to disclose.
Conflict of interest: Richard S. Thorley has nothing to disclose.
Conflict of interest: Rebecca L. Wollerton has nothing to disclose.
Conflict of interest: Sana Ullah has nothing to disclose.
Conflict of interest: David M. Smith has nothing to disclose.
Conflict of interest: Wojciech Lason has nothing to disclose.
Conflict of interest: Anthony J Rostron has nothing to disclose.
Conflict of interest: Najib M Rahman has nothing to disclose.
Conflict of interest: Rob J Hallifax has nothing to disclose.
- Received September 19, 2021.
- Accepted January 12, 2022.
- Copyright ©The authors 2022.
This version is distributed under the terms of the Creative Commons Attribution Licence 4.0.