Multinational Observational Cohort Study of COVID-19–Associated Pulmonary Aspergillosis

We performed an observational study to investigate intensive care unit incidence, risk factors, and outcomes of coronavirus disease–associated pulmonary aspergillosis (CAPA). We found 10%–15% CAPA incidence among 823 patients in 2 cohorts. Several factors were independently associated with CAPA in 1 cohort and mortality rates were 43%–52%.


Data Collection
For the discovery cohort, local investigators collected pseudonymized patient data from medical files into Castor Electronic Data Capture (Castor EDC, https://www.castoredc.com) electronic case report form (eCRF). Pseudonymized data for the validation cohort were entered into a SharePoint eCRF (Department of Clinical Research, Amiens University Hospital, Amiens, France).

Definitions
Clinically presumed coronavirus disease 2019  was based on clinical signs and symptoms and exposure to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), as assessed by treating physicians or multidisciplinary COVID-19 team. COVID-19-associated pulmonary aspergillosis (CAPA) diagnosis was based on the 2020 European Confederation for Medical Mycology/International Society for Human and Animal Mycology (ECMM/ISHAM) consensus classification (1). We applied several necessary modifications needed because of the eCRF's design and differences in interpretation between participating centers regarding performance of bronchial lavage (BL) or bronchoalveolar lavage (BAL). Although a BL was intended to signify a nonbronchoscopic, nondirected bronchial lavage (NBL) and a BAL a directed, bronchoscopic lavage, these terms were used differently by several centers. To avoid missing any directed BALs performed, we regarded BLs as equivalent to BALs during data analysis. Due to missing data on clinical factors and radiological results in many patients, all patients classified with CAPA were considered to demonstrate these clinical characteristics and (infiltrative) abnormalities on thoracic imaging during ICU stay. Patients who underwent none of the mycological tests required for classification as proven or probable CAPA, including autopsy, were designated CAPA not classifiable. Patients who underwent mycological tests were further evaluated for the presence of CAPA. In the validation cohort, data were not available to classify patients as possible CAPA.
According to the 2020 ECMM/ISHAM classification (1), BL or BALF galactomannan (GM) results that were qualitatively positive without a known quantitative result were not regarded as a positive mycological result. Patients were classified into 3 defined groups for further analyses. The CAPA group comprised patients with proven, probable, or possible Aspergillus tracheobronchitis, pulmonary CAPA, or both. The CAPA excluded group comprised patients who underwent diagnostic workup, but had no evidence for proven, probable, or possible CAPA, including no CAPA at autopsy, or patients without probable CAPA, but who were not classifiable as possible CAPA. The CAPA not classifiable group included patients who did not undergo any required mycological testing for proven or probable CAPA or those who were not classifiable for probable CAPA but had possible CAPA excluded.
We defined acute kidney injury (AKI) according to the Kidney Disease: Improving Global Outcomes (KDIGO; https://kdigo.org) criteria. AKI criteria include increase in serum creatinine of >26.5 µmol/L (>0.3 mg/dL) within 48 hours or increase in serum creatinine >1.5 times baseline values, which is known or presumed to have occurred within the prior 7 days or urine volume <0.5 mL/kg/h for 6 hours (2). Prior to classification, correction of volume status and obstructive causes of AKI are allowed.
We also defined stages of AKI according to the KDIGO criteria. Patients were classified according to the criteria that resulted in the highest, that is the most severe, stage of injury. Stage 1 is increase in serum creatinine to 1.5-1.9 times baseline values or increase in serum creatinine by >26.5 µmol/L (>0.3 mg/dL) or reduction in urine output to <0.5 mL/kg/hour for 6-12 hours.
Stage 2 is increase in serum creatinine to 2.0-2.9 times baseline values, or reduction in urine output to <0.5 mL/kg/hour for >12 hours. Stage 3 is increase in serum creatinine to 3.0 times baseline values, or increase in serum creatinine to >353.6 µmol/L (>4.0 mg/dL), or reduction in urine output to <0.3 mL/kg/h for >24 hours, or anuria for >12 hours, or the initiation of renal replacement therapy.

Comparisons between CAPA Classification Criteria
CAPA diagnosis according to the 2020 ECMM/ISHAM classification was compared to the IAPA expert opinion case definition (3), the modified AspICU (mAspICU) classification (4,5), and the revised European Organization for Research and Treatment of Cancer and the Mycoses Study Group Education and Research Consortium (EORTC/MSGERC) classification (6). As we did for the ECMM/ISHAM classification, we presumed all patients classified with IPA had clinical factors, signs, symptoms, and pulmonary infiltrates on thoracic imaging present for all other classifications, if applicable. As such, only subset A of the IAPA expert opinion case definition criteria for probable IPA could be evaluated in our cohorts. For the mAspICU classification, BAL fluid (BALF) GM optical density (OD) >1.0 and serum GM OD >0.5 were added as mycological criteria for the diagnosis of putative IPA and positive Aspergillus BALF culture was regarded as an entry criterion. For EORTC/MSGERC definitions, the eCRF was not designed to assess the presence of acute graft-versus-host-disease; therefore, we could not take this host factor into account. Due to frequent missing data regarding details on use of corticosteroids before admission to the ICU, all systemic corticosteroid use was considered a risk factor for invasive aspergillosis according to mAspICU and EORTC/MSGERC criteria. We compared the 2020 ECMM/ISHAM CAPA classification with the clinically reported occurrence of CAPA in patients' medical files, that is physician reported CAPA, regardless of fulfilment of any formal classification criteria.

Statistical Analysis
All data are expressed as no. (%) or median (interquartile range [IQR]). We compared variables by using Fisher exact test, Mann-Whitney U test, or Kruskal-Wallis test, as appropriate.
We analyzed survival differences by using the Kaplan-Meier method and log rank test. We performed binary logistic regression analysis to detect independent predictors of CAPA occurrence and ICU death. In all cases, we considered p<0.05 statistically significant. We did not apply corrections for multiple statistical testing during these analyses; readers should keep this in mind when interpreting results.
For binary logistic regression analysis, we used different independent variables as potential predictors for CAPA occurrence and ICU death. For CAPA occurrence, we used underlying conditions significantly more prevalent in the CAPA group in univariate analysis and corticosteroid use before and during ICU admission as independent variables. For ICU death, age, sex, AKI, renal replacement therapy (RRT), mechanical ventilation, use of vasopressors and/or inotropes and corticosteroids during ICU admission and presence of CAPA were used as independent variables or covariates. We performed statistical analyses by using SPSS Statistics for Windows version 25.0 (IBM Corp., https://www.ibm.com) or GraphPad Prism 5.03 for Windows (GraphPad Software Inc., https://www.graphpad.com).

Discovery Cohort
Of the 521 patients admitted to the participating ICUs in the Netherlands and Belgium during the study period, 1 patient was excluded from further analysis because <60% of data were collected due to an ICU stay of only several hours, at which point a non-ICU policy was instated and patient was discharged from the ICU. Another patient was excluded because of objection to participation ( Figure 1, panel A).
Of the 519 included patients, 4 (0.8%) did not have a positive SARS-CoV-2 PCR result, 2 of whom later were shown to be SARS-CoV-2 IgG positive. Three of the SARS-CoV-2 PCRnegative patients had radiographic findings on thoracic computed tomography (CT) and 1 on thoracic x-ray suggestive of COVID-19. One patient (0.2%) had no known SARS-CoV-2 PCR results but had radiographic findings on thoracic CT suggestive of COVID-19.

ECMM/ISHAM CAPA Classification
In the discovery cohort, tracheobronchitis could be evaluated in 187 patients who underwent bronchoscopy with or without BAL or BL, underwent autopsy, or both. Diagnostic tests to classify patients as proven tracheobronchitis or pulmonary CAPA were performed in 41/519 (8%) patients. Tests for classification as probable (pulmonary) CAPA were performed in 273/519 (53%) patients, and tests for classification as possible CAPA in 43/519 (8%). Because positive culture, GM and PCR results were reported more readily than negative ones, we could not give exact total numbers of fungal culture, GM testing, and Aspergillus PCR performed in BAL, BL, and NBL samples obtained (Tables 1, 2; Appendix Tables 7, 12).
In the validation cohort, 127/304 (42%) of patients could be evaluated for invasive Aspergillus tracheobronchitis, and 209/304 (69%) could be evaluated for probable CAPA. Data to classify patients as possible CAPA were not collected in this cohort.

CAPA Patients with COPD or Bronchiectasis
The discovery cohort contained 8 patients with CAPA and underlying COPD, whereas the validation cohort contained 2 CAPA patients with COPD. Furthermore, the validation cohort included 2 patients with CAPA and bronchiectasis. Data on bronchiectasis were not collected in the discovery cohort. The 2020 ECMM/ISHAM classification requires a positive GM test result as confirmation of a positive Aspergillus culture or PCR result in patients with COPD or another chronic respiratory disease to rule out colonization or chronic aspergillosis (1). Of the 12 patients with COPD or bronchiectasis in both cohorts, 11 (92%) had a BAL or BL GM OD >1.0, and 1 had a GM OD >6.0 in an NBL sample, indicating high fungal load with hyphal formation in the respiratory tract and probably not mere colonization.

Discovery Cohort Microbiological Results
Among the 17 patients with positive BL or BAL cultures, A. fumigatus was found in 15/17 (88%), A. nidulans in 1 (6%) patient, and both A. fumigatus and A. flavus in 1 (6%) patient. None of the 5 patients for whom susceptibility data were available demonstrated voriconazole resistance. Among 9 patients in whom positive BAL or BL PCR results were reported, A. fumigatus was reported in 3 (33%); speciation was not possible or provided in the other 6 (67%). Azole resistance PCR test results were reported in 1 patient in whom wild-type A.
fumigatus was found. A. fumigatus was found in all 6 patients with positive NBL cultures; 1 patient had a positive BALF culture and a positive NBL culture on different dates and both demonstrated A. fumigatus. In 5 of these patients, an Aspergillus PCR also was performed on the same NBL sample: in 3 patients A. fumigatus was found and no species were identified in the other 2.

Logistic Regression Analysis
In the discovery cohort, RRT was more prevalent in the CAPA group than in the CAPA excluded group. To account for any possible effects of this difference on ICU death, we explored an interaction term between RRT and CAPA in the logistic regression model, which demonstrated no interaction between the 2 variables (adjusted odds ratio [aOR] for the interaction term for ICU death 1.42, 95% CI 0.30-6.87, p = 0.66). In the validation cohort, we explored an interaction term for AKI during ICU admission and CAPA for the occurrence of ICU death, which also demonstrated no interaction between the 2 variables for ICU death (aOR 0.16, 95% CI 0.01-2.71, p = 0.20).  CAPA not classifiable, n = 240 p value † p value ‡ Days between first signs/symptoms and first positive SARS-CoV-2 PCR 7 (4-10); n = 459 9 (3-10); n = 37 7 (5-11); n = 213 7 (4-10); n = 209 0.48 0.13