This retrospective study was conducted at the General Intensive care unit (ICU) of the First Affiliated Hospital of Xiamen University. We screened a total of 2,316 patients with suspected pneumonia admitted between January 2022 and June 2025. To be eligible for inclusion, patients had to meet the following criteria: (1)new onset or worsening cough or dyspnea with or without fever and expectoration; the presence of new or progressive radiographic pulmonary infiltrates; (2) age ≥ 18 years; (3) an ICU length of stay ≥ 3 days; (4) mycological evidence of Aspergillus, defined by at least one positive result from fungal culture, galactomannan (GM) assay, polymerase chain reaction (PCR); and (5) availability of chest computed tomography (CT) imaging data. Patients who did not meet these criteria were excluded. Ultimately, 208 patients were enrolled in the final analysis (Figure 1).

We extracted the following variables from electronic medical records: patient demographics, comorbidities, host factors, medication use, mycological results, and the initial Sequential Organ Failure Assessment (SOFA) score. Additionally, we documented information on bronchoscopy, tracheostomy, mechanical ventilation, extracorporeal membrane oxygenation (ECMO), antifungal treatment, and the cumulative prednisone equivalent during hospital stay. Clinical outcomes included ICU and hospital length of stay as well as survival status.

Chest CT images were assessed by a senior ICU physician and a radiologist, with any disagreements resolved by consensus or by a third senior radiologist.

ICU patients are substantially heterogeneous. Therefore, we employed both the EORTC/MSG definition and the FUNDICU consensus (Donnelly et al., 2020; Bassetti et al., 2021, Bassetti et al., 2024) to diagnose probable IPA in critically ill patients with a combination of host factors, clinical criteria, and evidence of Aspergillus.

1. Host factors included both classic immunocompromised conditions defined by EORTC/MSG consensus and several risk factors expanded for ICU patients in recent years, as follows:

2. Clinical criteria were based on the presence of at least one of the following:

3. Aspergillus evidence was established by at least one of the following tests:

Patients were classified as either IPA or non-IPA based on whether they met the above diagnostic criteria. Only three patients were pathologically confirmed as proven IPA.

Due to the non-normal distribution, continuous variables are presented as median with interquartile range and were compared by Mann–Whitney U test. Categorical data are expressed as numbers, percentages and compared using the chi-square test. To identify independent risk factors for survival, variables with a P value < 0.1 in univariable COX regression were included in a multivariable Cox regression model. A two-sided P value < 0.05 was deemed statistically significant in the final model. Regarding missing data, complete case analysis was applied for the statistical analysis of individual variables (i.e., cases with missing values were excluded), and all variables included in the multivariable Cox regression analysis had no missing values. All statistical analyses were performed using the Statistical Package for the Social Sciences statistical software (version 26.0; SPSS Inc, Chicago, IL, USA). Figures were designed using Prism version 10.6.

After excluding patients without mycological workup or those with an ICU stay of less than 3 days, Aspergillus spp. was identified in 208 of the 1,835 screened patients. Among these 208 cases, the etiologies of pneumonia were bacterial (56.3%), mixed (21.6%), viral (4.3%) and unknown (17.8%). All patients with pneumonia in both groups received standard antibiotic therapy, and no statistically significant difference was observed in the antibiotic use profile between the two groups. Based on the updated EORTC/MSG and FUNDICU consensus, 131 cases were diagnosed as IPA, representing an incidence of 7.1% among the enrolled cohort.

As shown in Table 1, the IPA group demonstrated significantly higher rates of solid tumors (34.4% vs. 16.9%, P = 0.007), immunosuppressive therapy (75.6% vs. 35.1%, P < 0.001), and neutropenia (12.2% vs. 1.3%, P = 0.004) than the non-IPA group. Furthermore, IPA patients received higher cumulative corticosteroid doses (prednisone equivalents) during their hospital stay. Conversely, demographics, comorbidities, and admission SOFA scores were similar between groups. Mycologically, significantly more IPA patients tested positive in ≥ 2 different assays (67.9% vs. 16.9%, P < 0.001).

In terms of ICU management, IPA patients more likely to receive CRRT (57.3% vs. 40.3%, P = 0.022), vv-ECMO (9.2% vs 1.3%, P = 0.034), and anti-aspergillus treatment (90.1% vs. 61.0%, P < 0.001).

Figure 2 provides a summary of the performance of different mycological assays, either used separately or in combination. Culture from LRTs was the most frequently employed method (207/208), followed by serum GM (191/208) and BALF GM (158/208). BALF GM showed the highest positive rate (107/158, 67.7%), followed by molecular tests (68/102, 66.7%). On the other hand, serum GM demonstrated the lowest positive rate (60/191, 31.4%).

The combination of any two methods produced an overall positive rate of 48.6%, with BALF GM plus molecular testing yielded the highest positive rates (55.4%) as shown in Figure 2A.

Figure 2B shows the sensitivity and specificity of different mycological methods, used individually or in pairwise combinations, for the diagnosis of IPA. BALF GM exhibited the highest sensitivity (85.5%) with the specificity of 82.9%. Low sensitivity and specificity (54.2% and 26.3% respectively) were obtained from culture. The combination of BALF GM and molecular test showed a moderate level of sensitivity (64.4%) and specificity (80.0%). Although serum GM showed limited sensitivity (41.7%), its specificity was considerably high (89.2%). Furthermore, specificity rose to more than 95% when serum GM was used in combination with either culture or molecular methods.

A. fumigatus accounted for 80% of the distribution of Aspergillus species, followed by A. flavus (11.7%), as shown in Figure 2C.

Radiological characteristics of the study population are summarized in Table 1. Patchy infiltrates were the most frequent manifestation; however, no significant difference was observed between the IPA and non-IPA groups (79.4% vs. 70.1%, P = 0.135). In contrast, the prevalence of consolidation (63.4% vs. 40.3%, P < 0.001), and nodules (22.9% vs. 9.1%, P = 0.014) was significantly higher in the IPA group. Although less common, cavities (17.6% vs. 3.9%) and tree-in-bud signs (9.9% vs. 1.3%) also showed statistically significant intergroup differences (all P < 0.05). No significant disparities were identified between the two groups regarding bronchial wall thickening or pleural effusions. Figure 3 illustrates the chest CT manifestations of four representative patients with IPA at different disease stages.

Table 2 summarizes that clinical outcomes were significantly worse in the IPA group. These patients had a longer median ICU stay (20 vs. 16 days, P = 0.041) and a significantly longer duration of invasive mechanical ventilation (14.7 vs. 6.9 days, P < 0.001). Furthermore, the mortality rate was significantly elevated in the IPA group (55.0% vs. 26.0%, P < 0.001). Kaplan-Meier analysis further validated the difference in survival (Figure 4). Cox regression analysis revealed that IPA (HR: 2.064, 95% CI = 1.211-3.518, P = 0.008) and a longer interval from admission to the first positive mycological test (HR: 1.017, 95% CI = 1.004-1.031, P = 0.011) were independent risk factors for survival. However, multivariable analysis showed that tracheostomy (HR: 0.351, 95% CI = 0.222-0.557, P < 0.001) and anti-aspergillus therapy (HR: 0.489, 95% CI = 0.275-0.871, P = 0.015) were associated with significantly reduced mortality (Table 3).

The observed IPA incidence of 7.1% aligns with the reported 5%–10% in recent literature (Loughlin et al., 2020; Bassetti et al., 2024), though slightly exceeding the 6.5% rate found using 2021 EORTC/MSG ICU criteria alone (Liu et al., 2023). Notably, only 12.2% of our cohort presented with traditional neutropenia, while 75.6% received immunosuppressive therapy, underscoring the shift toward iatrogenic host factors and chronic organ failure recognized by the FUNDICU consensus (Bassetti et al., 2024). Aspergillus fumigatus remained the dominant pathogen (80%), consistent with regional findings in China (Feng et al., 2025). The 55% crude mortality rate reinforces IPA as a devastating complication in the ICU (Denning, 2024).

The diagnostic landscape of IPA in the ICU has shifted toward non-culture-based assays due to the limitations of traditional methods. Our findings underscore the superior sensitivity of BALF GM (85.5%), which outperformed both LRT culture (54.2%) and serum GM (41.7%), consistent with previous reports (Scharmann et al., 2023; Hu et al., 2024). As a polysaccharide released during active growth, BALF GM is sampled directly from the infection site, making it the preferred diagnostic tool when bronchoscopy is feasible (Bassetti et al., 2024; Heylen et al., 2024). However, the observed specificity (82.9%) was slightly lower than literature values, likely reflecting the high prevalence of false-positive triggers and population heterogeneity inherent in critically ill cohorts (Garnacho-Montero et al., 2024).

Molecular Testing (PCR) demonstrated the second-highest sensitivity (69.7%), providing particularly valuable for culture-negative cases or patients on empirical antifungal therapy (Hu et al., 2024), despite its limited accessibility in resource-constrained settings. In contrast, serum GM exhibited low sensitivity (41.7) but high specificity (89.2%). This low sensitivity is characteristic of non-neutropenic patients, where fungal antigens often remain localized within the pulmonary parenchyma rather than circulating systemically (Leeflang et al., 2015). Nevertheless, consecutive positive serum GM results remain a robust indicator of IPA, especially when BALF collection is contraindicated (Huang et al., 2021).

LRT Culture performed poorly (sensitivity 54.2%, specificity 26.3%), as slow growth. and colonization issues hinder its reliability (Garnacho-Montero et al., 2024; Hu et al., 2024). Consequently, our data support the FUNDICU recommendation for multimodal strategies (Bassetti et al., 2024); specifically, combining BALF GM and PCR yield the most balanced profile (64.4% sensitivity, 80.0% specificity). In scenarios where BALF is unavailable, pairing serum GM with either culture or PCR can enhance specificity. Crucially, our findings caution against over-reliance on LRT culture alone or a single positive PCR result, advocating instead for the integration of multiple mycological markers to guide clinical decision making (Bassetti et al., 2021, Bassetti et al., 2024).

The radiographic presentation of IPA in our cohort was largely nonspecific, with patchy infiltrates being the most common finding, while characteristic CT signs such as cavities, nodules, and halo sign were infrequent. This variability underscores that CT manifestations of IPA are dynamic and highly heterogeneous, shaped by underlying comorbidities, host immune status, and potential co-infections (Alexander et al., 2021; Brandi et al., 2022).

Our findings suggest a correlation between the extent of CT abnormalities and the severity of immunosuppression. Patients with profound immune impairment (e.g., diabetes complicated by severe viral pneumonia) showed diffuse consolidation, while those with milder factors presented with more localized lesions (Fischer et al., 2022; Lu et al., 2024). Notably, classic cavities often emerged only after one week, suggesting early reliance on “typical” signs may cause diagnostic delays. Furthermore, early airway-invasive signs like the tree-in-bud pattern, often seen in COPD and influenza patients, are critical but frequently overlooked markeres of Aspergillus bronchitis (Liu et al., 2020; Alexander et al., 2021; Feys et al., 2024). The resolution of lesions following timely antifungal therapy versus irreversible injury in delayed cases underscores the need for serial CT scans performed at appropriate intervals (Slavin et al., 2021).

IPA patients faced significantly worse outcomes, including longer ICU stays and nearly double the hospital mortality rate (55.0% vs. 26.0%), consistent with meta-analyses (Loughlin et al., 2020; Denning, 2024). Multivariable Cox regression identified, IPA (HR: 2.064) and delayed diagnosis as independent risks. Conversely, anti-Aspergillus therapy (HR: 0.489) was significantly associated with reduced mortality, This supports guideline-based pre-emptive therapy for high-risk, test-positive patients (Ledoux and Herbrecht, 2023; Bassetti et al., 2024) and highlights the consequences of undertreatment, a well-documented problem in COPD and ICU patients due to low awareness (Denning, 2024).

Notably, tracheostomy (HR: 0.351) was associated with significantly lower mortality. This may be related to that patients with IPA often require prolonged mechanical ventilation and assisted sputum suction. tracheostomy may improve outcomes by facilitating airway clearance, reducing ventilator-associated pneumonia (a common complication in IPA patients), and enabling earlier weaning from mechanical ventilation (Herritt et al., 2018; Chorath et al., 2021; Merola et al., 2024). However, this result may also be influenced by selection bias, as clinicians are more likely to perform this procedure on patients expected to survive longer.

Our findings differ from a recent study on COVID-19-associated invasive pulmonary aspergillosis in critically ill patients, which identified advanced age, mechanical ventilation, and high CRP levels as independent risk factors for mortality (Xiao et al., 2025). This discrepancy is likely attributable to differences in the study populations.

Our study has several strengths that enhance its generalizability to similar tertiary care settings. As it was conducted in a large, mixed ICU cohort, the research captures real-world clinical diversity among critically ill patients beyond narrow, specific subpopulations. Our observed IPA incidence (7.1%) aligns with multicenter studies (5%–15%) in Europe and North America, and our diagnostic results match international meta-analysis. The application of standardized FUNDICU and EORTC/MSG criteria ensures our results are comparable across different geographical regions. Furthermore, our identification of IPA itself and delayed diagnosis as independent prognostic risk factors, provides actionable insights for ICU clinicians.

However, several limitations necessitate a nuanced interpretation of our findings. First, the single-center retrospective design may introduce potential selection and indication biases. In particular, the observed association between tracheostomy and improved survival likely reflects survivor bias rather than direct protective effect. Second, the scarcity of “proven” IPA cases—due to the inherent risks of invasive biopsy (Ledoux and Herbrecht, 2023)—along with the occasional lack of serial chest CT imaging, may have led to an underestimation of disease progression in our cohort.