FormalPara Key Summary Points

Why carry out this study?

Invasive pulmonary aspergillosis is a common infection in intensive care units. There are no consensus criteria for defining IPA in the ICU.

The study evaluated the diagnostic agreement and performance of latest three criteria (the 2020 EORTC/MSG, 2021 EORTC/MSG ICU, and M-AspICU) for diagnosing IPA in critically ill patients.

What was learned from the study?

The three criteria (the 2020 EORTC/MSG, 2021 EORTC/MSG ICU, and M-AspICU) for diagnosing IPA showed very poor diagnostic agreement.

IPA diagnosed by either the 2020 EORTC/MSG or the 2021 EORTC/MSG ICU criteria was independently associated with 28-day mortality. Although the M-AspICU criteria have the highest “sensitivity”, IPA diagnosed by M-AspICU was not an independent risk factor of 28-day mortality.

Caution is required when using the M-AspICU criteria in ICU, especially in patients with non-specific infiltration and non-classical host factors.

Introduction

Invasive pulmonary aspergillosis (IPA), a common opportunistic infection in the intensive care unit (ICU), is associated with increased mortality [1,2,3,4]. The reported incidence of IPA among critically ill patients varies widely from 9.4% to 69% [2,3,4,5]. The significant differences in these values can be attributed to the lack of lung histopathology, heterogeneity among different types of ICU patients, and most importantly, the lack of consensus criteria on how to define IPA in the ICU population. Rapid and accurate diagnosis of IPA is critical for early administration of appropriate antifungal therapy and improved prognosis.

The European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group (EORTC/MSG) first proposed diagnostic criteria for IPA in 2002 and updated these criteria in 2008 and 2020 for better use in research and clinical care [6,7,8]. However, these criteria are aimed primarily at immunocompromised populations. In 2021, EORTC/MSG proposed IPA criteria for the ICU population [9]. Blot et al. proposed a clinical algorithm for IPA (AspICU criteria) that relies on a positive Aspergillus culture in lower respiratory tract specimens (LRTs) and is used to differentiate infection from colonization [10]. Schauwvlieghe et al. modified the AspICU criteria (M-AspICU) with the purpose of improving low sensitivity to culture and ICU population inapplicability [3]. Both the EORTC/MSG and M-AspICU criteria classify patients with IPA as “proven” and “probable” (EORTC/MSG)/“putative” (M-AspICU) cases. “Proven” cases are only defined with positive lung histopathology results, but this invasive test is not suitable for ICU patients with severe respiratory failure, coagulation disorders, or other complications. The diagnosis of probable IPA, which relies on a combination of host factors, clinical symptoms, mycological findings, and radiological findings, is relatively non-invasive and thus more suitable for ICU patients. The updated EORTC/MSG and M-AspICU criteria focus on probable cases with enlargement of host factors and imaging results and the revision of mycological criteria to identify more IPA cases.

To date, only a few studies have compared different criteria for IPA diagnosis in critically ill patients. In this study, we compared the rates of IPA according to the 2020 EORTC/MSG, 2021 EORTC/MSG ICU, and M-AspICU criteria in an ICU population and evaluated the effects of the three different IPA criteria on patient outcomes.

Methods

Study Design and Population

We conducted a retrospective single-center study in two ICUs (medical and surgical) of Peking Union Medical College Hospital between November 10, 2016 and November 10, 2021. Patients with the following characteristics were included: (1) age over 18 years; (2) length of ICU stay longer than 48 h; (3) had suspected pneumonia; (4) had any of the following tests done: Aspergillus galactomannan (GM) enzyme immunoassay test in serum or bronchoalveolar lavage fluid (BALF), LRT fungal culture (sputum, BALF, bronchial brush, or aspirate), or lung histopathology; (5) had a chest computed tomography (CT) examination, and the interval between CT and mycological test was less than 7 days.

Data Collection

The following variables were collected from patient records: age, gender, comorbidities, Acute Physiology and Chronic Health (APACHE II) score on admission, and Sequential Organ Failure Assessment (SOFA) score on the day of first retained GM; if the patient did not have a GM test, we selected the day of the first retained fungal culture of the LRTs. We also collected mycological tests, including fungal cultures of LRTs, serum GM, and BALF GM and chest CT data, and lung histopathology. Chest CT data were obtained by two senior ICU doctors who reviewed the chest images. If there were any disagreements, the two doctors had a discussion, and another senior ICU doctor was consulted if a consensus could not be reached. Finally, we collected invasive mechanical ventilation and survival status at 28 days.

Diagnostic Criteria

We selected three recent diagnostic criteria to diagnose IPA in critically ill patients: (1) the EORTC/MSG criteria published by Donnelly et al. in 2020 (2020 EORTC/MSG) [7]; (2) the EORTC/MSG ICU Working Group criteria published by Bassetti et al. in 2021 (2021 EORTC/MSG ICU) [9]; and (3) the algorithms published by Schauwvlieghe et al. in 2018 (M-AspICU) [3]. The detailed diagnostic criteria are listed in Supplementary Material 1. We defined “proven” and “probable” IPA as IPA cases, while “possible” IPA and cases that could not be classified by any of the diagnostic criteria were defined as patients without IPA (“No-IPA”).

Statistical Analysis

Continuous variables are presented as mean with standard deviation or median with interquartile range and were compared by t test or Mann–Whitney U test, respectively. Categorical data are reported as proportions and compared using the chi-square test. Diagnostic agreement was calculated using Cohen’s kappa. We performed a multivariate binary logistic regression analysis to detect independent risk factors of 28-day mortality and to assess the association of each criterion with 28-day mortality. P < 0.05 was considered significant. Data were analyzed using SPSS version 21.0.1. Figures were designed using Prism version 9.0.

Ethics Statement

The study was approved by the institutional review boards and Ethics Committee of Peking Union Medical College Hospital. As a result of the retrospective nature of the study, informed written consent was waived. The study was conducted in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Results

Study Population and Characteristics

A total of 2403 patients met the inclusion criteria between November 10, 2016 and November 10, 2021. (Supplementary Material 2). The patient characteristics are shown in Table 1. Of the 2403 patients, 1462 (60.8%) were male, and the mean age was 59 years. Almost one-third of the patients were admitted to the ICU for medical conditions. The patients in our study had a median APACHE II score of 18, of whom 1376 (57.3%) required invasive mechanical ventilation. None of the patients had lung histopathology, but all had fungal cultures in LRTs. Serum GM tests were performed in 2286 (95.1%) patients, and 315 also underwent BALF GM testing.

Table 1 Patient characteristics
Full size table

IPA Diagnostic Rates

The frequency of IPA varied substantially among the three criteria. According to the 2020 EORTC/MSG, 2021 EORTC/MSG ICU, and M-AspICU criteria, IPA was present in 81 (3.37%), 157 (6.53%), and 555 (23.10%) cases, respectively. Patients with IPA diagnosed according to the 2020 EORTC/MSG also met 2021 EORTC/MSG ICU criteria, while patients with IPA diagnosed according to the 2021 EORTC/MSG ICU also met the M-AspICU criteria (Fig. 1). The characteristics of patients with IPA diagnosed using the three criteria are shown in Table 2. Compared to the M-AspICU, the 2020 EORTC/MSG and 2021 EORTC/MSG ICU criteria diagnosed more severely ill patients. There were no differences in serum GM and BALF GM among patients with IPA diagnosed by the three criteria. Patients with IPA diagnosed by the 2020 EORTC/MSG had the highest rate of positive Aspergillus culture (71.6%), while patients diagnosed using the M-AspICU had the lowest rate (37.5%).

Fig. 1
figure 1

The distribution of invasive pulmonary aspergillosis by single criteria and the overlap of three diagnostic criteria in all patients

Full size image
Table 2 Characteristics and clinical outcomes of patients classified as IPA according to the three criteria
Full size table

Diagnostic Agreement Among the Three Criteria

The diagnostic agreement between the three diagnostic criteria was poor, with a Cohen’s kappa of only 0.328 (P < 0.001). The diagnostic agreements between any two of the three criteria are shown in Fig. 2.

Fig. 2
figure 2

Cohen’s kappa correlation between any of the three criteria

Full size image

2021 EORTC/MSG ICU Criteria vs. 2020 EORTC/MSG Criteria

Compared with the 2020 EORTC/MSG criteria, the 2021 EORTC/MSG ICU criteria diagnosed 76 additional patients with IPA. Among these patients, 37 (48.7%) were diagnosed because of the broader host criteria alone, which included chronic obstructive pulmonary disease (COPD; n = 20), severe influenza (n = 12), and decompensated liver cirrhosis (n = 10), 29 (38.2%) were diagnosed because of the lower serum GM positive threshold alone, with a median of 0.59 (0.41–0.71), and 10 (13.1%) were diagnosed owing to a combination of both.

2021 EORTC/MSG ICU Criteria vs. M-AspICU Criteria

Compared with the 2021 EORTC/MSG ICU criteria, the M-AspICU criteria diagnosed 398 additional patients with IPA. Among them, 154 (38.7%) were diagnosed because of broader host factors alone, which included solid tumors (n = 28), diabetes mellitus (n = 57), acute kidney injury (n = 66), 119 (29.9%) patients were diagnosed because of the broader radiological factors alone, and the remaining 125 (31.4%) patients were diagnosed owing to both of the above. Although the BALF GM threshold is higher for the M-AspICU criteria (1.0) than for the 2021 EORTC/MSG ICU criteria (0.8), this small difference had no effect on IPA diagnosis in our study.

Outcomes of IPA Diagnosed Using the Three Criteria

Compared to those without IPA, 28-day mortality was significantly higher in patients with IPA, regardless of the criteria used: 2020 EORTC/MSG criteria (40.7% vs. 11.4%, P < 0.001); 2021 EORTC/MSG ICU criteria (34.4% vs. 10.6%, P < 0.001); and M-AspICU criteria (19.8% vs. 9.9%, P < 0.001).

2021 EORTC/MSG ICU Criteria vs. 2020 EORTC/MSG Criteria

There was no difference in 28-day mortality between patients with IPA diagnosed using the 2021 EORTC/MSG ICU and 2020 EORTC/MSG criteria (34.4% vs. 40.7%, P = 0.335) (Fig. 3). Patients diagnosed by the 2021 EORTC/MSG ICU criteria but missed by the 2020 EORTC/MSG criteria had higher 28-day mortality than patients without IPA based on the 2021 EORTC/MSG ICU criteria (27.6% vs. 10.6%, P < 0.001). The 28-day mortality rates of the additional patients diagnosed by the 2021 EORTC/MSG ICU criteria (as a result of broader host factors, lower GM threshold, or both) were higher than the patients without IPA based on the EORTC-ICU criteria (Supplementary Material 3).

Fig. 3
figure 3

Kaplan–Meier for 28-day survival of patients with IPA

Full size image

2021 EORTC/MSG ICU Criteria vs. M-AspICU Criteria

The 28-day mortality rate of patients with IPA diagnosed by the M-AspICU criteria was lower than that of patients diagnosed by the 2021 EORTC/MSG ICU criteria (19.8% vs. 34.4%, P < 0.001) (Fig. 3). Patients diagnosed with IPA using the M-AspICU criteria but missed by the 2021 EORTC/MSG ICU criteria had higher 28-day mortality than patients without IPA based on the M-AspICU criteria (14.1% vs. 9.9%, P = 0.014). The 28-day mortality rate of the additional patients diagnosed by the M-AspICU criteria as a result of either broader host factors or broader radiological factors alone was higher than patients without IPA based on the M-AspICU criteria. However, the 28-day mortality rate of additional patients diagnosed by the M-AspICU criteria owing to both broader host factors and broader radiological factors was similar to patients without IPA based on M-AspICU criteria (8.8% vs. 9.9%, P = 0.689) (Supplementary Material 3).

Predictive Ability of the Three Criteria for Mortality

IPA diagnosed by the 2020 EORTC/MSG or 2021 EORTC/MSG ICU criteria was independently associated with 28-day mortality, even after adjustment for intensive mechanical ventilation, acute kidney injury, and SOFA score (Fig. 4). IPA diagnosed by the M-AspICU criteria is an independent risk factor of 28-day mortality [odds ratio (OR) = 1.431, 95% confidence interval (CI) = 1.034–1.981, P = 0.031] only when excluding patients who fulfilled neither the host factors nor radiological features of 2021 EORTC/MSG ICU.

Fig. 4
figure 4

Forest plots of risk factors for 28-day mortality. Model 1: Analysis of risk factors based on the model includes IPA diagnosed by the 2020 EORTC/MSG criteria and other univariable factors. Model 2: Analysis risk factors based on the model includes IPA diagnosed by the 2021 EORTC/MSG ICU criteria and other univariable factors. Model 3: Analysis of risk factors based on the model includes IPA diagnosed by the M-AspICU criteria and other univariable factors. Displayed are only those variables that remained in the model during variable selection. OR odds ratio, DM diabetes mellitus, IMV invasive mechanical ventilation, SOFA Sequential Organ Failure Assessment, IPA invasive pulmonary aspergillosis, EORTC/MSG Consensus Definitions of IPA from European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group and the National Institute of Allergy and Infectious Diseases Mycoses Study Group, M-AspICU modified algorithm for IPA

Full size image

Discussion

We compared the diagnostic and prognostic performance of the 2020 EORTC/MSG, 2021 EORTC/MSG ICU, and M-AspICU criteria for IPA in an ICU population. The main findings of our study can be summarized as follows: (1) the diagnostic rate of IPA in the ICU population ranged from 3.34% based on the 2020 EORTC/MSG criteria to 23.1% based on the M-Asp ICU criteria; (2) IPA diagnosed by either the 2020 EORTC/MSG or 2021 EORTC/MSG ICU criteria was independently associated with 28-day mortality; (3) IPA diagnosed by the M-AspICU criteria is an independent risk factor of 28-day mortality when excluding patients who fulfilled neither the host factors nor radiological features of 2021 EORTC/MSG ICU.

The diagnostic rate of IPA according to the M-AspICU criteria was four times higher than that based on the 2021 EORTC/MSG ICU criteria. These findings are in alignment with the results of Schroeder et al., who compared the two criteria in critically ill patients with positive LRT Aspergillus cultures (Cohen’s kappa = 0.12) [11]. The poor diagnostic agreement between the 2020 EORTC/MSG and M-AspICU criteria was similar to the results of Szabo et al.’s study of patients with corona virus disease 2019 (COVID-19) (Cohen’s kappa = 0.154) [12].

The 2020 EORTC/MSG criteria are not suitable for ICU patients because they only can identify patients with hematological malignancies and severe immunosuppression. Patients with IPA would be missed in the ICU according to the 2020 EORTC/MSG criteria. To reduce the likelihood of missing IPA diagnoses in the ICU, the diagnostic criteria were broadened. Compared with the 2020 EORTC/MSG criteria, additional host factors including influenza and COPD were added to the 2021 EORTC/MSG ICU criteria, and the positive threshold for serum GM was decreased from 1.0 to 0.5. Schauwvlieghe et al. concluded that influenza was independently associated with IPA [3]. An autopsy of 67 patients with proven IPA indicated that half had underlying diseases with COPD [13]. According to our findings, the use of the 2021 EORTC/MSG ICU criteria increased the rate of IPA diagnosis by 48.4% compared with the 2020 EORTC/MSG criteria. Among the additional patients diagnosed with IPA using the 2021 EORTC/MSG ICU criteria, those diagnosed because of expanded host factors alone had higher mortality than patients without IPA. Similarly, those diagnosed because of the lower GM threshold also had higher mortality than those without IPA. These findings support the diagnostic value of the 2021 EORTC/MSG ICU criteria in the ICU setting.

The 2021 EORTC/MSG ICU criteria add COPD and influenza in addition to the classic host factors. ICU patients with diabetes, solid tumors, and other underlying conditions, on the other hand, are susceptible to IPA [14, 15]. We may require broader IPA criteria for ICU patients. Compared to the 2021 EORTC/MSG ICU criteria, the M-AspICU criteria further broaden the diagnostic criteria with no host factor limitations and any infiltrates on lung imaging. The M-AspICU criteria diagnosed more than three times as many patients with IPA in this study. These patients with IPA had higher mortality than patients without IPA based on the M-AspICU criteria. More than one-third of the patients were diagnosed with IPA using the M-AspICU as a result of the broadened host factors alone. Among them, 37% had DM, and 19% had solid tumors. Several studies have supported diabetes mellitus (DM) and solid tumors as risk factors for IPA in critically ill patients. This finding is consistent with the results of Taccone et al., who found that either DM or solid tumors were present in nearly 20% of patients with proven IPA [16]. Cornillet et al. reported that ground-glass opacities and pleural effusions were more common in ICU patients with IPA [14]. Huang et al. demonstrated that patchiness was the most common CT sign among critically ill patients with both IPA and COPD [17]. In the present study, nearly one-third of the patients that were diagnosed with IPA using the M-AspICU criteria but missed by the 2021 EORTC/MSG ICU criteria as a result of non-specific infiltration alone had higher mortality than patients without IPA.

Although the M-AspICU criteria diagnosed the largest cohort of patients with IPA among the three criteria, IPA based on the criteria was not an independent risk factor of mortality. Schroder et al. found a high false positive rate of IPA (up to 71%) based on M-AspICU criteria in patients with proven IPA [11]. In the present study, we found that the subgroup of additional patients diagnosed by the M-AspICU criteria owing to both the expanded host factors and radiological features had a low mortality rate that did not differ from the rate among patients without IPA. Although all of these patients met the mycological criteria, Aspergillus colonization and false positives should be considered. After we excluded these additional patients mentioned above, IPA diagnosed by M-AspICU became an independent risk factor of 28-day mortality. Thus, the clinical significance of this subgroup of patients warrants further evaluation.

Our study has some limitations. First, all patients did not have lung histopathology. Second, the 2020 EORTC/MSG criteria included Aspergillus PCR; however, this technique is not available in our hospital, which may have caused us to underestimate the actual diagnostic rate of the 2020 EORTC/MSG criteria. Siopi et al. confirmed that cases of probable IPA increased by 22% when PCR was used [18]. Third, there were no patients with COVID-19 in our study. Although the inclusion dates from November 2016 to November 2021, our hospital is not a designated hospital for COVID-19 patients. However, there are other influenza data in our study. Fourth, BALF GM is more valuable than serum GM for diagnosing IPA in the ICU population [2, 19]. However, only 13.1% of patients in our study had BALF GM. Fifth, this was a single-center study, and the results may be biased; finally, radiologic features other than nodes, cavities, and consolidation were not collected.

Conclusion

The diagnostic rate of IPA in critically ill patients varied considerably based on the diagnostic criteria used, with very poor agreement between the three evaluated criteria. Compared with the 2020 EORTC/MSG criteria, the 2021 EORTC/MSG ICU criteria identified a group of patients with similarly devastating outcomes. Although the M-AspICU criteria have the highest sensitivity since they identify most patients as “probable” given the broadened criteria, this study shows that these patients are less ill. Whether broadened host factors and/or broadened radiologic features should be incorporated into the diagnostic criteria should be further evaluated in prospective cohort studies.