Increased frequency of pneumothorax and pneumomediastinum in COVID-19 patients admitted in the ICU: A multicentre study from Mumbai, India

Discussion

The first description of pneumothorax complicating COVID-19 was in the early days of the pandemic⁹ and since then there have been almost 200 publications on the occurrence of air leaks in these patients. The majority are scattered case reports and small series of two to three patients. Our study looks at all ICU admissions across three large hospitals in Mumbai over 7 months. Forty-two of these 1,324 ICU admissions with COVID-19 (3.2%) developed an air leak, manifest as pneumothorax and/or a pneumomediastinum. This is significantly higher than the six patients of 1,062 (0.6%) who developed air leaks in the non-COVID ICUs of our hospitals over the same period. Consistent with the prior literature, our cohort had an incidence of barotrauma that was higher than what is reported for ARDS from non-viral aetiologies. The incidence did not appear to be associated with ventilatory strategies which were reasonably standardised to ARDS lung-protective strategies across all three hospitals.

The strength of our study is its size, including as it does 1,324 consecutive ICU admissions (of 4,906 hospitalised patients) over 7 months from three large hospitals across Mumbai. There are only two other comparable multicentre series. The first by Martinelli et al.,¹⁰ included 71 patients from 16 hospitals across the UK. The second by Miro et al.¹¹ reviewed the incidence of pneumothorax in 61 emergency departments (EDs) across multiple hospitals in Spain. The former series is limited by including patients based on author recall and a targeted review of databases, unlike our study which systematically looked at admission details of all ICU patients. The Spanish series looked only at ED patients. Table 5 summarises the similarities and differences between our studies and these two.

Miro's study found only 40 patients of 71,904 (0.56%) ED admissions showing a pneumothorax on their admission X-ray. Thus air leaks are uncommon at initial presentation to hospitals with COVID-19. In contrast, our study which looked at more critically ill ICU patients with serial X-rays and frequent CT scans showed a much higher incidence of air leak with 42 cases in 1,324 ICU admissions giving an incidence of 3.17% in this patient population. All three series demonstrated striking male preponderance. About a third of patients in the Spanish and UK series had a history of pre-existing lung disease, unlike our patients, only 12% of whom had a prior history of lung disease. None of our 42 patients were smokers or reported a history of pneumothorax and only five had a history of pre-existing lung disease (three with well-controlled asthma and two with interstitial lung disease) which could have possibly contributed to their air leak.

Despite prompt diagnosis (due to daily serial X-rays), the mortality in our patients was high at 74% with 31 of these 42 patients succumbing to their disease. This contrasts with the overall mortality of 17% (226 deaths out of 1,282) in those COVID-19 patients in the ICU without an air leak. In the series of COVID-19 patients with a pneumothorax from Martinelli, similar high mortality of 63% was reported as well. Thus the occurrence of an air leak in a patient with COVID-19 pneumonia is an adverse prognostic marker signifying bad outcomes and associated with significantly higher mortality.

Interestingly our study, like that of Martinelli,¹⁰ showed that about one-third (33.3%) of air leaks occurred in patients who never received invasive ventilation. This begs the question of why COVID-19 patients are more prone to develop air leaks and secondary pneumothoraces than other ICU patients. The typical pathology of COVID-19 pneumonia and ARDS may help explain the increased frequency of occurrence. CT scans demonstrate the peripheral distribution of the initial ground-glass opacity (GGO) changes encountered in COVID-19.^12,13 The subsequent fibrosis in some of these patients have been documented to progress to is also peripheral in its distribution.¹⁴ Cyst formation and bullae in areas of airspace disease have also been noted in several radiological^15–17 and autopsy studies of COVID-19.^18,19 It is possible that these peripherally located bullae could rupture, either spontaneously or as a consequence of the positive pressure ventilation which the more severely ill patients with COVID-19 pneumonia are subject to. Thus while mechanical ventilation may certainly be contributing, it is not the sole risk factor, with significant numbers of air leaks occurring in patients who had never received ventilatory support (either invasive or non-invasive) at the time of their pneumothorax. This is possibly facilitated by the use of systemic corticosteroids, which is the standard of care in hypoxemic patients with COVID-19,²⁰ and may have contributed to a lowered immune response, vulnerability to injury, and impaired healing.^21,22 In keeping with this observation, there are other reports of air leaks in patients of COVID-19 pneumonia and ARDS who were not receiving mechanical ventilation at the time.^23,24

The early diagnosis of air leaks is essential and the strength of our study was that since these were all ICU patients, a chest radiograph was performed on a near-daily basis. This helped facilitate the prompt diagnosis of air leaks. Our patients developed their air leak a mean of 9 days (1–60 days) after admission to the ICU and these air leaks persisted for a mean of 6 days (1–60 days). A high index of suspicion is required to diagnose this condition. Obvious pneumothorax is generally easy to diagnose on a chest radiograph but a CT scan is more accurate in early cases of suspected pneumothorax and pneumomediastinum. Significant numbers of our patients (16 of 42, 38%) had pneumomediastinum in the absence of any obvious pneumothorax which may be more subtle and difficult to diagnose. This was twice the frequency noted in the UK series and maybe because unlike the patients in this series where the diagnosis of air leak was often based on physician recall and record review, diagnosis in our patients was based on daily radiographs and therefore picked up more frequently.

Prompt treatment is also crucial. Sixty-six percent of our patients required insertion of an intercostal drain at the bedside while the remainder needed careful observation without need for a drain. One patient who had a large pneumothorax and persistent air leak over 60 days underwent successful surgical repair (VATS) and was eventually discharged home. He remains well and free of recurrence highlighting the importance of surgical intervention in selected patients with persisting air leaks.

Limitation of our study was its retrospective nature with review of records at the end of 7 months period. Air leak was not actively screened for but detected as a part of regular (near-daily) chest radiograms performed as part of routine ICU protocol. Also, treatment varied across the three ICUs though most patients received drugs that were standard of care at the time of the study.

In conclusion, our study demonstrates that air leaks occur with a higher frequency in patients with COVID-19 than in other ICU patients. The incidence does not appear to be related to ventilatory strategies, as lung-protective ventilation was standard of care in all three ICUs. When present, such air leaks contributed to poor outcomes with almost 74% mortality rates in these patients which are significantly higher than the 17% mortality in patients in our other COVID-19 ICU patients without air leaks. While pneumothoraces are the commonest form of air leak, pneumomediastinum in the absence of pneumothorax occurs in 38% of patients. Air leaks are more likely to occur in patients being mechanically ventilated but a third of all air leaks occur in patients breathing spontaneously. A high index of suspicion is needed to make the diagnosis, with an air leak being considered in the differential diagnosis of any COVID-19 patient who develops sudden respiratory distress and hypoxia.