A Cohort Study on Lung Cavitations in Severely Ill COVID-19 Patients – Evidence for a Thromboembolic Pathogenesis

Jan M. Kruse (  jan-matthias.kruse@charite.de ) Charite Universitatsmedizin Berlin https://orcid.org/0000-0003-4250-8315 Daniel Zickler Charite University Hospital Berlin: Charite Universitatsmedizin Berlin Willie M Lüdemann Charite University Hospital Berlin: Charite Universitatsmedizin Berlin Sophie K Piper Charite University Hospital Berlin: Charite Universitatsmedizin Berlin Inka Gotthardt Charite University Hospital Berlin: Charite Universitatsmedizin Berlin David Horst Charite University Hospital Berlin: Charite Universitatsmedizin Berlin Andreas Kahl Charite University Hospital Berlin: Charite Universitatsmedizin Berlin Kai Uwe Eckardt Charite University Hospital Berlin: Charite Universitatsmedizin Berlin Sefer Elezkurtaj Charite University Hospital Berlin: Charite Universitatsmedizin Berlin


Background
The novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a beta coronavirus is causing coronavirus disease 2019 . In the past months, its spread to most countries of the world has led to a global pandemic 1 , causing more than 1 600.000 deaths to date. 2 Severe hypoxemic respiratory failure due to acute lung injury is the most common complication leading to admission to intensive care units (ICU) and one of the main causes of death. 3,4 The precise etiology of the impaired gas exchange and optimal treatment strategies remain a matter of debate. 5 Of note, some clinical aspects seem to differentiate COVID-19 from other forms of acute respiratory distress syndrome (ARDS). In particular, in many COVID-19 patients with ARDS the pulmonary compliance is not signi cantly altered, in contrast to classic ARDS. 5 Besides lung injury a prothrombotic state has emerged as an important characteristic of COVID-19. Data from both clinical studies and postmortem case series demonstrate a high incidence of thromboembolic events.6-9 These events include pulmonary artery occlusions, which may have a thrombotic or thromboembolic origin. 10 Massive pulmonary embolism has been suggested to cause out-of-hospital mortality. 11 Whether pulmonary artery occlusion contributes to hypoxemic respiratory failure due to impaired lung perfusion and dead space ventilation is controversial. 12 We noted extensive lung cavitations in single cases of COVID-19, which prompted us to perform a systematic analysis of a cohort of 39 critically ill COVID-19 patients consecutively admitted to two ICUs in a tertiary care referral center. This analysis revealed a high incidence of cavitary lung lesions. Computer tomography (CT) morphology and the results of postmortem macro-and micropathological examination point to an ischemic etiology of these lesions due to thrombotic obstruction of pulmonary artery branches.

Patients
We retrospectively analyzed clinical patient data of 39 COVID-19 patients admitted to ICU between March and May 2020 who received at least one chest CT shortly before or during their ICU stay. All patients were tested positive for SARS-CoV-2 by polymerase chain reaction (PCR). Five patients were admitted through the emergency department, one patient was directly admitted to our ICU from an outpatient setting, two patients had worsened during their stay on regular wards and 31 patients were secondary referrals from other ICUs.

Anticoagulation
All patients received unfractionated heparin (UFH) with a targeted partial thromboplastin time (PTT) of 50-55. Patients who suffered from venous thromboembolic (VTE) complications or who had other indications for therapeutic anticoagulation were dosed for a target PTT of 60-80 seconds.
Patients who did not reach the target PTT with usual doses of UFH were switched to Argatroban. Patients were also switched to Argatroban when they required extracorporeal membrane oxygenation therapy (ECMO).

Mechanical ventilation
All mechanically ventilated patients received pressure controlled ventilation. Positive endexspiratory pressure (PEEP) was titrated to reach best possible oxygenation index. Patients received low tidal volume ventilation with a target tidal volume (VT) of 6 ml/kg/PBW and a diving pressure below 15 mmHg was targeted.

Data collection
CT scans were analyzed independently by two of us (JMK, WML) for parenchymal cavities, de ned as a lucency within a zone of pulmonary consolidation, a mass, or a nodule; hence, a lucent area within the lung that may or may not contain a uid level and that was surrounded by a wall of varied thickness. 13 Screening for venous thrombosis was performed using ultrasound (GE Vivid S70 ultrasound machine with a 9L-D probe) in all patients after ICU admission and repeated at least weekly.
Laboratory parameters included viscoelastic coagulation testing after ICU admission using the ROTEM Sigma System (Tem International, Munich, Germany). 14 Maximum values of C-reactive protein (CRP), ddimers, brinogen, leukocytes, interleukin-6 and procalcitonin were compared between patient groups.
Microbiology reports of all of the collected specimens from the respiratory tract and blood cultures were analyzed for pneumopathogenic species.
The highest levels of PEEP, peak inspiratory pressure and driving pressure during the ICU stay were recorded and analyzed in ventilated patients.
The highest values for SOFA and APACHE II during the course of therapy were recorded and analyzed. Mann-Whitney U tests were used to compare differences between patient groups in continuous variables while Chi-squared tests were used for categorical data. A two-sided signi cance level of 0.05 was applied without adjustment for multiple comparison. All p-values constitute exploratory data analysis and do not allow for con rmatory generalization of results.

Results
Chest CT ndings 64 CT scans were analyzed, of which 39 were performed shortly before or after ICU admission and 25 during the course of the ICU stay. In 37/39 (95%) of patients we found ground glass opacities of the lung parenchyma, characterized as "mosaic pattern". Cavitary lung lesions were found in 22 patients (56%), of which 15 patients presented with cavitations in the initial CT while seven patients exhibited cavitary lesions in a subsequent CT scan.
Among patients with cavitations the number of cavities between left and right lung were similar.
Cavitations were evenly distributed between central and peripheral parts of the lungs. Eleven patients presented with peripherally and 11 patients with centrally located cavitations. Thirteen patients showed involvement of the lower parts, while in 9 patients cavitations were found in the upper parts of the lung. Clinical characteristics Table 1 shows demographics, preexisting comorbidities and treatment parameters of the patient cohort and the two subgroups with and without lung cavitations. Most parameters were similar between groups, except that patients in the group with cavitations were older and had a higher body mass index (BMI) than those without cavitations. Laboratory and microbiology ndings Table 2 shows markers of in ammation and coagulation. These parameters did not differ signi cantly between the patients with and without cavitations.

Autopsy ndings
In all three patients in whom an autopsy was performed, lung cavities had previously been identi ed on CT scans. External examination of the lungs showed pronounced pleural brin deposits and sunken lung areas, which corresponded to bullous transformations of lung parenchyma. Findings in two patients are presented in Fig. 2. Upon opening the cavities appeared as areas of liqui ed necrosis (Fig. 2B). Careful dissection revealed connections between cavities and bronchial system (Fig. 2C). Moreover, preparation of the pulmonary vessels on frontally oriented cross-sections yielded unequivocal associations of cavitary lesions with thrombotic occlusion of the supplying pulmonary artery branches ( Fig. 2D and 2E). Microscopy of adjacent lung tissue revealed numerous thrombotic vascular occlusions and extended, partially hemorrhagic and partially anemic infarct zones in spatial association with vascular occlusions (Fig. 2F). The border zone of the infarct areas showed pronounced neutrophil in ltrations, but there was morphologic evidence for bacterial colonization. In summary, macro-and microscopic ndings in combination suggested extensive vascular occlusions of different duration with multiple pulmonary infarctions of different size, some of which had transformed into liquefying necrosis, corresponding to large cavities.

Discussion
Our ndings indicate that lung cavitations of variable size, at least in part a consequence of liquefying ischemic lung infarcts, contribute to lung pathology and loss of functional lung parenchyma in COVID-19 patients. These observations extend the reported spectrum of lung CT ndings that is considered as typical for COVID-19, including ground glass opacities, consolidations and a "crazy paving pattern" of the lung parenchyma. 15,16 In general, infectious causes and ventilator induced lung injury are recognized as the main etiologies of lung cavities in critically ill patients. 17 Ischemia is less commonly considered as a cause, although cavitations are described in up to 32% of patients with pulmonary embolism and are a common nding in patients suffering from chronic thromboembolic pulmonary hypertension. 18,19 While the precise pathogenesis is di cult to ascertain in individual cases in our study, we believe that several lines of evidence indicate that an ischemic pathogenesis rather than alternative causes play a major role.
First, several of our ndings are not consistent with primarily ventilator induced lung injury. Ventilator settings were chosen to minimize lung trauma and did not differ signi cantly between the groups with and without lung cavities. The distribution of the cavities with a signi cant proportion of central lesions and involvement of the lower parts of the lung argues against ventilator induced lung injury, since from our experience one would expect mainly peripheral lesions in the upper lobes, if mechanical overdistension played the main role. The observation that a high percentage of lesions occurred in preexisting opacities also seems rather untypical for classical ventilator induced lung alterations. Due to higher compliance of the less affected regions, overdistension tends to occur preferentially in non-opaque regions of the lung. Most striking is the fact, that three of ve patients, who did not need mechanical ventilation, also developed cavitations. Still mechanical ventilation might play an important role in the development of the cavitary lesions in that it aggravates the damage done by microvascular and macrovascular thrombosis. We therefore think that sticking to the principles of lung-protective ventilation is of great importance in this group of patients to minimize pulmonal destructions even if the lung compliance does not seem to be altered in every case.
Second, patients who developed cavitations did not exhibit positive cultures more often than the patients without, nor did we nd typical abscess inducing species more frequently in the patients with lung cavities. However, a secondary bacterial infection of infarcted areas, transforming into cavitations cannot be ruled out.
Third, we found evidence for a prothrombotic state both in terms of coagulation parameters and thromboembolic complications. Although thromboembolic complications were evenly distributed between patients with and without lung cavities, notably pulmonary embolism was more frequent in patients with cavities (4 vs. 1). Patients who developed cavitations during the course of their disease were signi cantly older and had a higher BMI. Obesity has been associated with a prothrombotic state and hypo brinolysis. [20][21][22][23][24] The "mosaic pattern" of the lung-parenchyma that we observed in accordance with previous publications and the lung cavities strikingly resemble ndings in patients with chronic thromboembolic pulmonary hypertension. 15,18,19 During the time of this observational study critical care resources for COVID-19 patients were in no way compromised in our regional setting, enabling long-term ICU therapy. Together with a very large proportion of secondary referrals this might explain why others have to the best of our knowledge not yet reported similar observations. However, several other observations support the concept of impaired lung perfusion in COVID-19 patients. Ackermann et al. found a high incidence of microvascular thrombi and signs of endothelitis during autopsy of seven COVID-19 patients. 6 Lang et al., using dual source computer tomography, discovered severe perfusion abnormalities in the lungs of three COVID-19 patients and postulated a signi cant contribution of altered perfusion to the etiology of respiratory failure in COVID- 19. 12 In terms of the mechanisms potentially causing pulmonary hypoperfusion, SARS-CoV-2 binds to the angiotensin converting enzyme 2 (ACE-2) receptor of alveolar epithelial cells. 25 There is evidence for consecutive downregulation of ACE-2 leading to increased levels of angiotensin II. 25 High levels of angiotensin II in the pulmonary circulation may lead to endothelial activation and vasoconstriction and promote a prothrombotic state. 26 Hypoperfusion of pulmonary artery branches either due to vasoconstriction or due to thromboembolic occlusion will lead to increased dead space ventilation and impaired gas exchange, consistent with the ventilation pattern observed in COVID-19. 5,27 Selective perfusion of the pulmonary circulation through anastomoses between the bronchial and the pulmonary circulation might further contribute to respiratory failure due to an increased right-to left shunt fraction. 28 By implying a link between a prothrombotic stage, pulmonary hypoperfusion and structural lung changes, our data add to the considerations for therapeutic anticoagulation in COVID-19 patients. Interestingly in this regard Wang et al. recently reported a positive impact of tissue plasminogen activator treatment on oxygenation in a case series. 29 An association of higher therapeutic targets of systemic anticoagulation and improved survival has also been reported. 11,30 Our study has several limitations. Being retrospective and non-interventional it can only be hypothesisgenerating. It is monocentric and focusses on severely ill patients, many of whom received ICU treatment for several weeks and our ndings may not be generalizable to less severely ill patients. The etiology of lung cavitations is possibly heterogeneous and multifactorial. Finally, the number of autopsies supporting our interpretation is small and although all point to the exact same pathogenesis of the cavitations we do not know whether the same results would have been found in the lungs of the other patients of the cohort.
In conclusion, we found that cavitating lung lesions occur frequently in severely ill COVID-19 patients and provide evidence that pulmonary hypoperfusion and occlusion of pulmonary arteries plays an important role in the pathogenesis of these lesions.

Conclusion
Our ndings underline the importance of su cient anticoagulation in the management of patients with severe COVID-19 pneumonia. Furthermore our data show, how vulnerable the malperfused lungs of these patients are, even though the compliance might not be altered in the beginning. This points to the importance of lung protective ventilation in order to avoid further damage in the poorly perfused areas of the lungs of these patients. Availability of data and material: The datasets analyzed during the current study are available from the corresponding author upon reasonable request.
Competing interests: There are no competing interests to be declared by the authors.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.