We collected partially prospective and partially retrospective data for 823 patients in 2 cohorts. The discovery cohort comprised patients with PCR-confirmed or clinically presumed COVID-19 admitted to 4 ICUs in the Netherlands and 4 ICUs in Belgium during February 28–May 27, 2020. The validation cohort comprised patients with PCR-confirmed COVID-19 admitted because of respiratory insufficiency to 3 participating ICUs in France during April 7–May 31, 2020 (Appendix Methods, Table 1).

Figure 1

Figure 1. Flowchart of the study inclusion process for a multinational observational study of CAPA in 3 countries in Europe, 2020. A) Discovery cohort; B) validation cohort. For further analyses, patients with...

We applied ECMM/ISHAM classification criteria for CAPA (11). We considered bronchial lavage (BL) equivalent to bronchoalveolar lavage (BAL). We assumed all CAPA classified patients demonstrated clinical factors and radiographic abnormalities. We defined 3 patient groups: CAPA, CAPA-excluded, and CAPA not classifiable (Figure 1; Appendix).

We included 519 patients in the discovery cohort; median age was 64 years, 73% were male, and 82% required invasive mechanical ventilation during ICU admission (Table 1; Appendix Table 2, 3, 4). Among patients in the discovery cohort, 279 (54%) were classifiable: 6 (2%) as CAPA proven, 32 (12%) as probable CAPA, and 4 (1%) as possible CAPA (Figure 1, panel A; Appendix Results, Tables 5, 6). CAPA incidence among classifiable patients was 15% (42/279); 85% were CAPA-excluded. Among patients in the discovery cohort, 46% (240/519) were not classifiable, including 3 who did not fulfill the criteria for possible CAPA (Figure 1, panel A). In patients with any EORTC/MSGERC host factor, CAPA incidence was 30% (13/44), compared with 16% (26/161) in patients with no host factors (p = 0.053).

Chronic obstructive pulmonary disease (COPD; p = 0.04) and HIV/AIDS (p = 0.01) were more prevalent in CAPA patients (Table 1; Appendix Table 2). Among CAPA patients, 33% had >1 EORTC/MSGERC host factor, compared with 19% of CAPA-excluded patients (p = 0.053). Corticosteroid use was not more prevalent in the CAPA group (p = 0.14), in contrast to other immunosuppressant drugs (p = 0.01). In logistic regression analysis, corticosteroid use at any dose before or during ICU admission was not independently associated with CAPA development. However, COPD, HIV/AIDS, and use of other immunosuppressant drugs before ICU admission were associated with CAPA (Appendix Figure 1, panel A).

Among CAPA patients who underwent BAL or BL, Aspergillus culture was positive in 42%, GM was positive (optical density [OD] >1.0) in 78%, and Aspergillus PCR was positive in 17%. Among CAPA patients who underwent nonbronchoscopic lavage, 67% had positive cultures. Serum GM was positive in 11% of tested CAPA patients. Median time between ICU admission and first positive mycologic test was 6 (interquartile range [IQR] 3–9) days (Table 1; Appendix Table 7).

Figure 2

Figure 2. Kaplan-Meier survival curves comparing patients with CAPA and those classified as CAPA excluded in a multinational observational study. A) Discovery cohort; B) validation cohort. Survival analysis performed by using Mantel-Cox...

The proportion of patients receiving systematic corticosteroid treatment in ICUs was not significantly different between CAPA and CAPA-excluded groups (p = 0.40), nor was corticosteroid dose (p = 0.88) (Table 1; Appendix Table 4). Antifungal treatment was administered to 16% (83/519) of patients, 88% of CAPA patients, and 15% of CAPA-excluded patients (Appendix Table 8). ICU mortality rates were significantly higher in CAPA patients (52%) than in CAPA-excluded patients (34%) (p = 0.04; Table 1; Appendix Table 4); mortality rates were 67% for patients with positive serum GM. CAPA patients demonstrated reduced survival (p = 0.02) (Figure 2, panel A); estimated median survival was 42 days after ICU admission. When correcting for covariates, CAPA was not independently associated with ICU mortality rates, but older age and acute kidney injury (AKI) during ICU stay were (Appendix Figure 1, panel B).

We included 304 patients in the validation cohort (Figure 1, panel B); median age was 63 years, 25% were male, and 76% required invasive mechanical ventilation (Table 2; Appendix Tables 9, 10). Ultimately, 209/304 (69%) patients were classifiable for CAPA: 21 (10%) probable CAPA and 188 (90%) CAPA excluded (Figure 1, panel B; Appendix Results, Tables 5, 11). Among patients with EORTC/MSGERC host factors, CAPA incidence was 13% (3/23), compared with 10% (18/186) among patients without host factors (p = 0.71).

All 21 probable CAPA patients were female; cardiovascular disease, excluding hypertension (p = 0.02), and bronchiectasis (p = 0.03) were more prevalent in this group (Table 2; Appendix Table 9). Use of corticosteroids before or during ICU admission or other immunosuppressant drugs before ICU admission were not independently associated with CAPA (Appendix Figure 1, panel C). In the validation cohort, 19% received antifungal treatment; 57% of the CAPA group received antifungal treatment (Appendix Table 8).

Corticosteroid use during ICU stay was not significantly different between the CAPA and CAPA-excluded groups (p = 0.82) in the validation cohort. ICU mortality rates were higher in the CAPA group than the CAPA-excluded group (43% vs. 25%; p = 0.12) (Table 2; Figure 2, panel B; Appendix Table 10). The ICU mortality rate was 50% in patients with positive serum GM. CAPA was not independently associated with ICU death, but older age and AKI during ICU admission were (Appendix Table 10, Figure 1, panel D).

We found CAPA incidence was 10%–15%, corresponding to the 14%–19% reported in other studies (8,9). Discovery cohort CAPA incidence was similar to influenza-associated pulmonary aspergillosis (IAPA) incidence in ICUs (12,13). CAPA seems to develop later after ICU admission than IAPA. Median time to first positive mycologic test in our study was 6 days after ICU admission, similar to other studies reporting 4–8 days (7–9) but in contrast to the median 3 days reported for IAPA (12,14).

Corticosteroids were not associated with CAPA in our study, consistent with previous reports (7–9), but contrasting associations seen with invasive pulmonary aspergillosis (IPA) and IAPA (12). This finding might be explained by possible dual effects of corticosteroids in COVID-19, impairing anti-Aspergillus immunity while simultaneously ameliorating the hyperinflammatory immune dysregulation and associated tissue damage conducive to IPA.

We found CAPA ICU mortality rates were 43%–52%, in line with previous reports (7–9) and comparable to those for IAPA (12). We could not assess antifungal treatment effects on mortality rates, but CAPA patients in the validation cohort who received antifungal treatment demonstrated a trend toward improved survival (Appendix Figure 2).

The first limitation of our study is that assuming clinical and imaging factors were available for all patients classified with CAPA possibly led to overreporting of CAPA. Excluding CAPA based on 1 negative mycologic test might have led to underreporting. Another limitation was that patients undergoing mycologic workup were likely more severely ill, which becomes apparent when comparing baseline and outcome data of the CAPA not classifiable group to the other 2 groups (Appendix Tables 5–12). Several classifications have been published or updated after we initiated this study; therefore, not all diagnostic modalities were evaluated, and we used some terms, such as BAL and BL, interchangeably (11,15).

In conclusion, we report CAPA incidence of 10%–15% in COVID-19 patients admitted to ICUs, CAPA ICU mortality rates of 43%–52%, and decreased survival over time. Clinicians should be aware of CAPA and that underlying factors, including COPD, immunosuppressant drugs other than corticosteroids, and HIV/AIDS, can increase the risk for CAPA.