Aspergillus, a ubiquitous opportunistic filamentous fungus, primarily invades the human host through the respiratory tract, potentially leading to a spectrum of pulmonary pathologies collectively termed pulmonary aspergillosis (PA). PA predominantly affects immunocompromised patients (Kanj et al., 2018), with particularly high morbidity in pediatric populations due to immature immune system and frequent comorbidities (notably hematologic malignancies and primary immunodeficiencies) (Wattier et al., 2015). Epidemiological data from the European Registry of Invasive Fungal Infections in Children (ECIFIG) reveal a pediatric incidence rate of invasive PA at 1.2 cases per 100,000 population. Notably, pediatric mortality rates (40%–60%) significantly exceed those observed in adult populations (Groll et al., 2014). The risk escalates dramatically in high-risk cohorts, particularly hematopoietic stem cell transplant recipients and chemotherapy patients, where the incidence of PA reaches 15–20% (Zaoutis et al., 2006).

The diagnosis of PA in pediatric populations remains clinically challenging due to three limitations: (1) non-specific clinical presentation (e.g., fever, cough, dyspnea) that overlap with common respiratory infections; (2) technical difficulties in obtaining adequate lower respiratory tract specimens, particularly bronchoalveolar lavage fluid (BALF), result in suboptimal sensitivity (<50%) when using conventional microbiological diagnostic methods (Dinand et al., 2016); (3) lack of standardized diagnostic thresholds for serological biomarkers, particularly galactomannan (GM) assays in pediatric population (Warris et al., 2019). Consequently, these diagnostic constraints highlight the urgent need for the identification of pediatric-specific risk factors and the establishment of predictive models for pediatric PA to improve early detection and clinical outcomes.

In recent years, research on risk factors for PA in pediatric populations has gained increasing attention. While classical predisposing factors—neutropenia and glucocorticoid therapy—remain clinically significant, emerging evidence underscores unique pediatric risk factors, including a history of preterm birth, congenital heart disease requiring surgical intervention, and Epstein-Barr virus coinfection (Wattier et al., 2015). Notably, developmental immunometabolism features in children may exacerbate disease progression, for example, heightened iron metabolism could facilitate Aspergillus fumigatus iron acquisition pathways (Schrettl and Haas, 2011). Despite these insights, current research remains disproportionately focused on adult populations, resulting in critical knowledge gaps regarding pediatric-specific risk stratification, validated biomarker thresholds, and multifactorial pathogenic interactions. These limitations significantly hinder the development of precision diagnostic and therapeutic frameworks for PA in children. Thus, we performed a comprehensive analysis of PA risk factors in pediatric patients through a retrospective case-control study encompassing 97 pulmonary infections (38 PA cases vs 59 controls) in this study.

PA in pediatric populations remains a life-threatening infection with high morbidity and mortality, necessitating early and accurate diagnosis (Cadena et al., 2021; Olivier-Gougenheim et al., 2021; Jia et al., 2023). This study identified 6 independent risk factors for pediatric PA—surgical history, elevated serum BDG level, viral coinfection, absence of fever, shorter symptom-to-admission interval, and hematologic diseases—and established a predictive model with an AUC of 0.93. These discoveries redefined pediatric PA diagnosis by identifying child-specific biomarkers.

This study identified prior surgery as a key risk factor for pediatric PA (OR=9.52, 95% CI:1.96–46.23). This aligns with existing literature linking postoperative immunosuppression to increased susceptibility to invasive fungal infections in adults (Friol et al., 2025). However, this association between recent surgery and PA is stronger in children, likely due to immature immune regulation. Pediatric patients show impaired fungal clearance for the neutrophil dysfunction, with reduced reactive oxygen species (ROS) production and delayed apoptosis (Urban and Backman, 2020). Surgical stress exacerbates this by suppressing Th17-mediated immunity, essential for Aspergillus defense (Alexander et al., 2021). In our cohort, 60.53% of PA patients received corticosteroids, further weakening Th17 responses by inhibiting IL-23/IL-17 signaling (Gaffen et al., 2014). Surgical trauma combined with corticosteroid use creates an environment favorable to Aspergillus invasion (Lionakis et al., 2023).

In our predictive model, an elevated BDG level (cutoff: 61.275 pg/mL) demonstrated superior diagnostic performance over GM. This divergence from guideline recommendations (Warris et al., 2019); (Subspecialty Group of Respiratory, t.S.o.P.C.M.A et al., 2024) may be attributed to three pediatric-specific factors: 1) Antifungal Prophylaxis Impact: Triazole prophylaxis (e.g., posaconazole) in pediatric cohorts effectively suppresses Aspergillus galactomannan (GM) release by inhibiting fungal membrane synthesis, while BDG—a stable cell wall component—remains detectable (de Heer et al., 2019). This explains the predominantly negative serum GM results (<1.0 μg/L) in our study. 2) Pathogen Species Variation: Non-fumigatus Aspergillus species (e.g., A. flavus, A. niger) are more prevalent in children and produce lower GM levels than A. fumigatus (Hon et al., 2024). BDG, as a pan-fungal marker, is less affected by species-specific variations. 3) Age-Specific Diagnostic Thresholds: Our optimized BDG cutoff (61.275 pg/mL) exceeds the adult negative range (20–60 pg/mL), underscoring the necessity for pediatric-specific criteria—a gap highlighted in recent consensus guidelines (Lass-Florl et al., 2021). When used alone, BDG (>61.28 pg/mL) showed limited sensitivity (34.21%, 95% CI: 21.21–50.11%) but high specificity (94.92%, 95% CI: 86.08–98.61%). However, our composite model significantly improved sensitivity to 78.95% (95% CI: 63.65–88.93%) while maintaining specificity at 88.14% (95% CI: 77.48–94.13%) ( Supplementary Table 1 ). This demonstrates the model’s capacity to integrate BDG with other predictors for enhanced pediatric PA detection. This aligns with findings that BDG sensitivity may surpass GM in specific clinical contexts (Dichtl et al., 2020).

This study identified a notable absence of fever in pediatric PA patients (61.02 vs. 18.42%, P < 0.001) compared to non-PA patients, highlighting atypical presentations in immunocompromised children (Wakai and Hess, 2023). Patients with PA do not exhibit specific fever symptoms compared to those infected with other pathogens (bacteria or viruses). This difference likely results from corticosteroid-mediated cytokine suppression (e.g., IL-6, TNF-α) and T-cell exhaustion, which attenuate febrile responses (Bjelakovic et al., 2010). These findings emphasize the need for increased clinical suspicion for PA in afebrile immunocompromised children, particularly as classic radiographic signs, such as the halo sign, are uncommon in pediatric cases (Thomas et al., 2019).

Hematologic diseases significantly increased the risk of pediatric PA (OR: 11.68, 95% CI: 0.89–153.62, P = 0.062), particularly in acute myeloid leukemia (AML; incidence: 3.7%) and hematopoietic stem cell transplantation (HSCT; incidence: 4.5%) (Niang et al., 2023; Wakai and Hess, 2023). Viral coinfection emerged as the strong risk factor (OR: 15.99, 95% CI: 3.55–72.00, P < 0.001), consistent with reports of influenza-associated (IAPA; prevalences: 15.3%) and COVID-19- (Wauters et al., 2021)associated pulmonary aspergillosis (CAPA, prevalences: 13%) in children (Wediasari et al., 2020; Ogawa et al., 2023; Li et al., 2024). Viruses may predispose to PA through airway epithelial damage and immune dysregulation, worsened by corticosteroid use (Costantini et al., 2020; Goncalves et al., 2024). Additionally, a shorter symptom-to-admission interval (OR: 38.68, 95% CI: 5.38–277.94, P < 0.001) indicates acute, severe presentations, potentially driven by viral coinfections or profound immunosuppression, as observed in invasive pulmonary aspergillosis (IPA) cases requiring urgent intervention (McMillan et al., 2020; Niang et al., 2023).

Our novel predictive model for pediatric PA achieves superior diagnostic accuracy (AUC: 0.93, R² = 0.96) compared to existing algorithms, which showed limited sensitivity (75-80%) (Niang et al., 2023; Wakai and Hess, 2023). The model’s enhanced discriminatory capacity appears to derive from incorporating two novel predictors: (1) viral coinfection and (2) symptom-to-admission interval. This marks a significant advancement over conventional models that relied primarily on neutropenia or radiographic findings (e.g., nodules or halo signs). Notably, the model demonstrates excellent calibration (Hosmer-Lemeshow p=0.34), indicating strong agreement between predicted and observed probabilities. These characteristics highlight its potential for clinical use, particularly for early identification of high-risk pediatric patients in critical care or immunocompromised settings.

mNGS revealed a high prevalence of polymicrobial infections in PA cases (86.84% vs. 18.64%, P < 0.001), highlighting the complexity of pediatric PA, particularly in viral-associated contexts. This finding aligns with reports of bacterial and viral co-pathogens in IAPA and CAPA (Ogawa et al., 2023; Li et al., 2024). mNGS proves valuable for detecting co-infections often missed by conventional culture or GM testing (Wakai and Hess, 2023). The polymicrobial burden may complicate antifungal therapy and worsen outcomes, necessitating tailored antimicrobial strategies.

This study has several limitations that warrant consideration. First, the study’s retrospective nature introduces potential selection bias, which may affect the generalizability of our findings. Second, the optimal BDG cutoff for pediatric populations remains uncertain due to ongoing debate regarding its diagnostic utility in children. And the lack of BALF GM data represents a key constraint, as serum GM (which was uniformly low in our cohort) is known to have lower sensitivity for PA compared to BALF testing. Third, the routine use of mNGS may be limited by its high cost and restricted accessibility, particularly in resource-limited settings. Our model’s performance requires validation in cohorts with complete GM testing (both serum and BALF) to assess its generalizability across different clinical settings. Multicenter prospective studies incorporating both BALF GM and standardized BDG testing are needed to: (a) establish pediatric-specific cutoff values, and (b) clarify the complementary roles of these biomarkers in different clinical scenarios (e.g., with/without antifungal prophylaxis).

This study identifies critical risk factors and a high-performance predictive model for pediatric PA. The high polymicrobial rate revealed by mNGS emphasizes the need for comprehensive diagnostic approaches. These findings enable targeted risk stratification and early intervention, addressing a critical gap in pediatric PA management.