The clinical course of pneumomediastinum in patients with SARS-CoV-2 before invasive mechanical ventilation

Discussion

Compared with non-SARS-CoV-2 cohorts, patients with COVID-19 were at increased risk for developing pneumothorax with or without pneumomediastinum. Table 3 summarises the studies that show an increased association but one of the limitations in these studies has been the lack of clarity about pneumomediastinum alone, especially before invasive ventilation, and this is largely due to limited numbers.^4–8 We evaluated this further and focused on three main clinical domains: potential predisposing factors, complication burden and clinical outcomes. Our data suggest that one in five patients hospitalised with COVID-19 had moderate to severe disease needing either non-invasive respiratory support and/or IMV. PneumoCoV was more common in men with a higher BMI (>25 kg/m²) and the majority were non-smokers.⁹ We used the Charlson Comorbidity Index to categorise the comorbidities' burden; a third of patients had airways disease (asthma and COPD) and none had any previous episode of pneumomediastinum.¹⁰ Association of asthma is not specific to SARS-CoV-2 and has been reported in a quarter of patients with spontaneous pneumomediastinum in the non-SARS-CoV-2 cohort.¹¹ The ISARIC-developed risk stratification tool was used at admission, and majority of patients (60%) had a high to very high 4C score signifying poor outcomes.¹²

Despite excluding patients with SARS-CoV-2 with pneumothorax alone, our data are consistent with a recently published estimate.⁴ PneumoCoV was encountered in patients predominately with moderate to severe disease requiring non-invasive respiratory support (HFOT, CPAP or NIV) and seldom seen in mild illness. Proper clinical assessment and a low threshold for chest X-rays are necessary as this complication can be seen at three stages: at admission, during the course of illness and as an incidental finding. The majority of patients who developed pneumomediastinum had clinical deterioration (worsening gas exchange, increased respiratory rate or unexplained desaturations) on non-invasive respiratory support but one in five patients had an incidental diagnosis, most commonly seen during nasogastric tube check prior to initiating nutritional support, thus the severity of the illness matters. Following the evidence from the RECOVERY trial, dexamethasone (6 mg once per day for 10 days) was considered as the standard of care and we witnessed increased cases following this pharmacological implementation.¹³ Two patients developed pneumomediastinum prior to steroids being a standard of care as compared with the remaining 36 patients. This observation was consistent with a recently reported study and, although an observational case series cannot establish the temporal contributory relation, the role of steroids needs to be evaluated further.⁴ Procalcitonin as a biomarker for secondary infections was part of routine investigations and, as per local protocol, a level of >0.5 ng/mL was considered sufficient to suspect secondary bacterial infections and triggered a course of antibiotics. An interesting observational finding was that two-thirds of patients were initiated on antibiotics (5–7 days) and no positive yield from blood cultures was isolated. Increased procalcitonin levels predicted survival on subgroup analysis and further large-scale studies are needed to evaluate the temporal relationship of elevated procalcitonin as a contributory factor for developing PneumoCoV. We did not find any association with other parameters (C-reactive protein, D-dimer, lactate dehydrogenase, creatine kinase, troponin and ferritin).

Patients were managed as per national guidelines and we were biased towards CPAP.² This was mainly due to three reasons: inconsistent advice from different national societies and guidelines, some of which discouraged the use of HFOT; discordance between oxygen demand and supply with a risk of exhausting hospital oxygen supply; and the risk of aerosol generating procedure. Due to high patient burden leading to increased oxygen demand, HFOT was limited to those who were intolerant to CPAP and this provision was only available in ARCU and ICU. CPAP was provided with non-invasive ventilators (Trilogy100/202, V60 and Trilogy Evo) and the median duration of CPAP was 4 days. All patients were maintained on a positive end-expiratory pressure (PEEP) ranging from 8–12 cm of water depending on patient tolerance and targeted oxygen saturations (92%–96% as standard, and 88%–92% for those not suitable for IMV). The majority of patients underwent awake proning (full prone or semi-prone) that was carried out by a multidisciplinary proning team with constant monitoring to assess clinical stability or to identify the deteriorating patient.

Pneumomediastinum is not specific to SARS-CoV-2, with reported incidence of approximately 1 in 30,000 in the general population and may be either primary (idiopathic) or secondary to any underlying secondary disease or interventions.¹⁰ Likely barotrauma, viral-associated tracheal oedema, recruitment manoeuvres and cytokine storm leading to Macklin's phenomenon (increased intra-alveolar pressure leading to alveolar rupture causing the air leak to pulmonary interstitial spaces with further extension of the air leak along the perivascular sheaths that dissect into the mediastinum leading to mediastinal emphysema) may all be plausible explanations for the development of PneumoCoV, shown in Fig 1 (chest X-ray and CT).^14,15 Following an episode of pneumomediastinum, 45% (n=17) of patients were transferred to ICU for IMV but a third of our patients were deemed unsuitable for escalation beyond ARCU due to poor performance status and or extensive comorbidities. A third of patients continued on CPAP but with lesser PEEP (6–8 cm of water) and the rest had either conventional oxygen therapy or were considered for palliation due to poor clinical outcome. Close clinical monitoring, appropriate escalation plan and a patient-centred clinical decision about continuing respiratory support post-PneumoCoV may be beneficial for best outcomes.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

a) Chest X-ray showing pneumomediastinum and SARS-CoV-2 pneumonia. b) Computed tomography showing subcutaneous air leak and parenchymal changes of severe COVID-19.

Pneumothorax can be associated with pneumomediastinum and this is usually unilateral (more commonly on the right side). The preponderance is likely due to SARS-CoV-2 targeting the right more than the left lung.¹⁶ Further evaluation with CT was not routinely considered due to patient haemodynamic instability and due to strict infection control measures limiting the safe transfer of patients from either ARCU or ICU to the radiology department. This was only considered when further characterisation of the complication was required at the clinician's discretion or prior to planning any specific invasive intervention, specifically chest drain. Seldinger chest drain insertion was required only in patients with significant pneumothorax (>2 cm) and five patients underwent this intervention.⁵ As per our data, the associated mortality rate in PneumoCoV was higher when compared with our overall mortality in patients requiring either ARCU and/or ICU (55% vs 43%). Thus, patients developing pneumomediastinum alone or with associated pneumothorax have a worse outcome. Additionally, on subgroup analysis, increasing age, higher BMI, lack of resolution of pneumomediastinum, presence of concurrent pneumothorax and the need for antibiotics for secondary infections based on procalcitonin predicted survival.

Our case series had limitations: data were from a single centre and, thus, an accurate estimate of the incidence and causal relationships cannot be extrapolated. PneumoCoV may have an impact on length of stay and we did not evaluate this but, at the time of reporting, two patients were still being managed as inpatients. We did not undertake multivariate analysis given the limited numbers. Further large-scale multicentre studies are needed for a better understanding of possible predisposing factors for PneumoCoV. Finally, a national registry to collate data and clinical characteristics may help in better understanding this complication.