Clinical and Radiological Predictors of Invasive Fungal Sinusitis

Abstract

Background

Invasive fungal sinusitis (IFS) is associated with high mortality rates, chiefly because it is often diagnosed late in the course of disease when infection has spread to the orbit, skull base, or intracranially. We aimed to define early clinical and radiologic indicators of IFS, which can be integrated into a prediction model and clinical management algorithms.

Methods

Patients with proven or probable IFS were identified in 2 tertiary-level hospitals as part of prospective monitoring of invasive mycoses. Patients with similar underlying medical conditions who underwent sinus computed tomography (CT) and nasal endoscopy because of new-onset sinus symptoms served as controls. Computed tomography images were reviewed by 2 blinded radiologists, and features were scored on structured report forms. Computed tomography and clinical data were compared between cases and controls; variables associated with IFS were further assessed in multivariate regression models.

Results

Twenty-seven patients with IFS were identified and matched 1:1 with control patients. Ocular pain, preauricular or premaxillary fat infiltration on sinus CT, and mucosal necrosis identified on nasal endoscopy were independently associated with IFS and were retained in the regression model. Model sensitivity and specificity were 0.77 and 0.96, respectively, with an area under the receiver-operating curve of 0.88.

Conclusions

Integration of radiologic and clinical features allowed early prediction of IFS, suggesting that these criteria could be used in the context of a clinical diagnostic algorithm.

Diagnosis of invasive fungal sinusitis is often delayed due to nonspecific early findings. In this case−control study, ocular pain, premaxillary fat infiltration on CT, and mucosal necrosis on nasal endoscopy were significant predictors, potentially aiding early detection.

Invasive fungal sinusitis (IFS) is a rare opportunistic mold infection characterized by hyphal extension across the mucosal surfaces of the nasal cavity, turbinates, septum, and paranasal sinuses, with high rates of associated morbidity and mortality [1, 2]. Aspergillus species and members of the order Mucorales are the most frequent causative agents [3]. Complications of IFS may include orbital and intracranial extension with cavernous sinus thrombosis, parenchymal cerebritis or abscess, mycotic aneurysm, stroke, and hematogenous dissemination [4]. Mortality rates of up to 50% have been reported [1, 5]. Patients with IFS are usually immunocompromised, often as a result of hematologic malignancy, cytotoxic chemotherapy, stem-cell transplantation, neutropenia, or poorly controlled diabetes mellitus with diabetic ketoacidosis [6, 7]. Treatment includes correction of the underlying immune or metabolic abnormalities, surgical debridement of necrotic tissue, and mold-active antifungal therapy [2, 7, 8].

Early diagnosis of IFS has been shown to decrease morbidity by facilitating the initiation of appropriate therapy during the early stages of disease [7, 9]. However, early signs and symptoms of IFS may be nonspecific, resembling those of rhinosinusitis caused by viruses or bacteria as well as chemotherapy-induced mucositis [7, 8]. Often cited features of IFS, such as bone erosion and extension of the infection beyond the paranasal sinuses, generally only manifest late in the course of the disease [8].

In the present study, we aimed to determine early clinical and radiologic predictors of IFS. We performed a case−control study of high-risk patients who underwent evaluation for sinus symptoms and compared clinical and radiological findings between patients with proven or probable IFS and control patients in whom IFS was excluded.

METHODS

This was a retrospective case−control study performed in 2 hospitals in Israel, the Tel Aviv Sourasky Medical Center and Rambam Health Care Campus. Patients with hematologic malignancies or a history of stem-cell transplantation meeting diagnostic criteria for IFS were identified during prospective monitoring of invasive fungal diseases. Invasive fungal sinusitis cases were defined as probable or proven according to the European Organization for Research and Treatment of Cancer and the Mycoses Study Group consortium (EORTC/MSG) criteria [10]. Chronic IFS cases with symptoms lasting >30 days and cases of COVID-19-associated fungal sinusitis were excluded. Control patients were identified by screening the medical record database for patients with hematologic malignancy or stem-cell transplantation who underwent evaluation for sinus-related symptoms, including computed tomography (CT) and examination by an otolaryngologist, resulting in exclusion of IFS. Patients were assigned to the control group if at least one of the following conditions was met: (1) negative histopathology and culture results from endoscopic biopsy; (2) negative endoscopic assessment without biopsy and no evidence of IFS on follow-up of at least 30 days. Computed tomography rather than magnetic resonance imaging (MRI) was assessed because the latter was not universally done and tended to be performed later in the clinical course. All patients were evaluated by an otolaryngologist, including nasal endoscopic evaluation. Controls were selected via frequency matching (1:1) to align the distributions of age, followed by neutropenic status, with those of the case group. All patients were hospitalized between 1 January 2014 and 1 March 2021. All procedures followed were in accordance with the ethical standards of the Declaration of Helsinki. This study was authorized by the Tel Aviv Sourasky Medical Center ethics committee (approval no. 009-21-TLV). The need for informed consent was waived given the retrospective noninterventional nature of this study.

Microbiology

Fungal isolates from sinus aspirates, tissue biopsies, and surgical specimens were grown on Sabouraud's dextrose agar plates and slants at 30 and 35 °C. All isolates were sent to the National Reference Laboratory for Medical Mycology at the Tel Aviv Sourasky Medical Center, where identification based on morphologic features was repeated. In addition, molecular identification to the species level was done by polymerase chain reaction (PCR) and sequencing of the ribosomal RNA (rRNA) 18S–5.8S–28S internal transcribed spacer. Aspergillus species and Fusarium species were further analyzed using PCR and sequencing of the β-tubulin and translation elongation factor genes, respectively.

Data Collection

Clinical and demographic data were retrieved from electronic medical records and entered into structured case report forms. The collected data included patient age, comorbidities, underlying hematologic malignancy, neutropenia, and immunosuppressive treatment. Neutropenia was defined as an absolute neutrophil count of <500/µL for 10 or more days at the time of symptom onset. Corticosteroid use was defined as systemic treatment with a prednisone-equivalent daily dose of ≥20 mg for ≥3 weeks within the 3 months preceding symptom onset. Systemic and sinus-related signs, symptoms, and physical examination findings were extracted from clinical notes of hematologists and infectious diseases consultants. Documentation of endoscopic evaluation by otolaryngologists, including macroscopic and histopathologic findings, was reviewed.

Imaging Techniques

Sinus imaging was done using multiple-slice CT (slice width 0.8−3 mm) with a bone window. Three-dimensional reconstruction was performed in the evaluation of each patient.

Computed tomography images were interpreted independently by 2 experienced radiologists who were blinded to the clinical diagnosis. The radiologists used a predefined score sheet to standardize interpretation (see Supplementary Data). The reviewed structures included paranasal sinuses, orbit, nose, and soft tissues. Each paranasal sinus was assessed for specific findings, including the measured mucosal thickening, changes in sinus content density, presence of a fluid level, alterations in bone structure, and fat infiltration adjacent to the sinus. Associated features, including facial soft tissue and subcutaneous fat swelling, masticator space and pterygopalatine fossa fat infiltration, muscle thickening, and abscess formation, were also documented. In addition, each radiologist provided a presumptive diagnosis of IFS or non-IFS, considering their overall impression of the radiological findings.

Statistical Analysis

To check baseline balance between cases and controls, we assessed intergroup differences using the Wilcoxon rank-sum test for age and Fisher's exact test for neutropenia. Given the small sample size, a conservative significance threshold of P < .1 was used to screen for potential residual imbalances. Each clinical and radiological feature was compared between case and control groups using Fisher's exact test for categorical data and the Wilcoxon rank-sum test for continuous variables. Features that differed between groups (P < .1) were further assessed as predictive variables in a multivariable binary logistic regression model with case/control group as the dependent variable. Variables were selected using a manual backward elimination process. At each step, the variable with the highest P-value was removed, provided its exclusion did not result in a change of >20% in the coefficients of the remaining variables, indicating the presence of confounding. Variables meeting a significance level of <.05 were retained in the model. In line with our aim of early IFS detection, only symptoms present within 5 days of clinical presentation were entered into the model. Goodness-of-fit and diagnostic accuracy of competing models were compared using the Akaike information criterion (AIC) and area under the receiver-operating characteristic curve. Because of the small sample size, the logistic regression model was fitted using Firth's penalized likelihood method. Next, predictors were validated by fitting a Bayesian ridge regression model with full posterior inference. Credible intervals excluding zero for each predictor confirmed a meaningful effect on the model. Interrater reliability between radiologists was assessed by calculating Cohen's kappa coefficient for each radiological feature. A type I error of <.05 was used to define statistical significance. Calculations were performed in R (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

The study cohort comprised 27 immunosuppressed patients with IFS, selected based on predefined inclusion criteria, and 27 control patients in whom IFS had been excluded. The median age was 51.5 years (interquartile range [IQR], 26.7−66 years). Owing to a limited control patient pool, case patients were nonsignificantly younger than controls (median age, 45 vs 57 years, P = .109), and prolonged neutropenia was more common among controls (85% vs 57%; P = .055). Additionally, a significantly lower proportion of cases, compared with controls, were hospitalized in the hematology department and received chemotherapy (Table 1). Forty-five patients (83%) had hematologic malignancies, primarily acute myelocytic leukemia (30 patients), and 8 (14.8%) had solid malignancies. Twenty patients (37%) had undergone stem-cell transplantation, 38 (70.3%) were neutropenic for 10 days or more at the time of diagnosis, and 21 (38.8%) had received corticosteroids for 3 weeks or more. Eight patients (14.8%) had diabetes mellitus.

Table 1.

Demographic and Clinical Characteristics of Study Cohort

Variable	Cases (N = 27)	Controls (N = 27)	P-Value
Age, median (IQR)	45.00 (17.50, 62.50)	57.00 (42.50, 68.50)	.109
Sex, male	16 (59.3)	18 (66.7)	.778
Department
Hematology	18 (66.7)	25 (92.6)	.043
Internal medicine	6 (22.2)	1 (3.7)	.105
ICU	5 (18.5)	1 (3.7)	.194
Comorbidities
Charlson index, median (IQR)	3.00 (2.00, 5.50)	3.50 (2.00, 5.00)	.926
Neutropenia >10 d	15 (57.7)	23 (85.2)	.055
Chemotherapy	18 (66.7)	27 (100.0)	.003
Corticosteroid >3 wk	12 (44.4)	9 (33.3)	.577
Hematologic malignancy
Acute myelocytic leukemia	13 (48.1)	17 (63.0)	.411
Multiple myeloma	2 (7.4)	3 (11.1)	1
Chronic lymphocytic leukemia	2 (7.4)	0 (0.0)	.471
MPAL	1 (3.8)	0 (0.0)	.985
Acute lymphocytic leukemia	3 (11.1)	0 (0.0)	.235
Non-Hodgkin's lymphoma	2 (7.4)	6 (22.2)	.25
Myelodysplastic syndrome	1 (3.7)	4 (14.8)	.348
Solid organ malignancy	3 (11.1)	5 (18.5)	.702
HCT	10 (37.0)	10 (37.0)	1
Allogeneic	7 (25.9)	5 (18.5)	.743
Autologous	3 (11.1)	5 (18.5)	.702
GVHD	3 (11.1)	3 (11.1)	1
Solid organ transplant	2 (7.4)	0 (0.0)	.471
Diabetes mellitus	6 (22.2)	2 (7.4)	.25
PKD	1 (3.7)	0 (0.0)
CNI/Mtori	5 (18.5)	4 (14.8)	1
HIV infection	1 (3.7)	0 (0.0)	1

All categorical variables represented as number of patients (percent within group). Continuous variables displayed as median (interquartile range [IQR]).

Abbreviations: CNI, Calcineurin inhibitor; GVHD, graft vs host disease; HCT, hematopoietic stem-cell transplantation; HIV, human immunodeficiency virus; ICU, intensive care unit; MPAL, mixed phenotype acute leukemia; Mtori, mammalian target of rapamycin inhibitors; PKD, polycystic kidney disease.

According to the EORTC/MSG criteria, 13 case patients (48.1%) were defined as having proven and 14 (51.8%) were defined as having probable invasive fungal disease. Mycologic culture was positive for filamentous fungi in 26 case patients (96%) and none of the controls. Aspergillus species were most frequently recovered (n = 12, 44%), followed by Mucorales (n = 9, 33%), phaeohyphomycetes (n = 3, 11%), and Fusarium species (n = 1, 4%). Functional endoscopic sinus surgery was performed on 19 case patients and 9 controls (P = .15). The median time to endoscopy was 2 days from symptom onset (IQR, 1−4 days). Histopathologic analyses of tissue obtained during endoscopic biopsy of nasal mucosal lesions in 19 case patients showed hyphae in tissue (n = 18, 94.7%), necrosis (n = 10, 47.6%), angioinvasion (n = 5, 23.8%), and inflammatory infiltrates (n = 6, 28.6%). Eight case patients and 2 control patients died within 30 days of initial evaluation (odds ratio 5.88, 95% confidence interval 1.11–31.16; Fisher's exact P = .035). The most common diagnosis among control patients was “rhinosinus disease excluded,” typically following evaluation for unexplained fever and sinus-related symptoms (n = 16, 59.2%). These cases were generally attributed to chemotherapy-induced mucositis. Additional diagnoses included acute bacterial rhinosinusitis (n = 4), previously unrecognized chronic sinusitis (n = 2), sinus mass (n = 2), sinus infiltration by lymphoma (n = 1), rhinitis medicamentosa (n = 1), and sinus cyst (n = 1).

In the early stage of presentation (within 5 days of symptom onset), ocular pain was reported among 19% of cases and none of the controls (P = .054). At later time points, facial numbness and proptosis were more frequent among cases versus controls (Table 2). There were no significant differences in the occurrence of fever between groups. Necrotic mucosal lesions were identified more frequently on endoscopic examination among cases than controls (33% vs 4%, P = .02). Pulmonary infiltrates were observed in 37% of cases versus 11.1% of controls (P = .056). Orbital or ocular findings were reported in 6 (22.2%) case patients versus none of the control patients (P = .022; Table 2). Findings included orbital cellulitis (n = 3), proptosis (n = 1), optic neuropathy (n = 1), and chemosis (n = 1).

Table 2.

Clinical Symptoms and Findings Among Case and Control Patients

Variable	Cases (N = 27)	Controls (N = 27)	P-Value
Signs and symptoms
Early (≤5 d)
Fever	14 (53.8)	18 (66.7)	.501
Visual disturbance	3 (11.1%)	0 (0.0)	.235
Ocular pain	5 (19.2%)	0 (0.0)	.054
Numbness	0 (0)	0 (0)	NA
Facial pain	7 (26.9%)	5 (18.5%)	.687
Facial swelling	7 (26.9)	4 (14.8)	.455
Cranial nerve palsy	1 (3.8)	0 (0.0)	.985
Nasal congestion	3 (11.5)	7 (25.9)	.324
Nasal discharge	3 (11.1)	3 (11.1)	1.0
Epistaxis	0 (0.0)	4 (14.8)	.128
Anosmia	0 (0.0)	1 (3.7)	1.0
Headache	7 (26.9)	3 (11.1)	.263
Mental status change	0 (0)	0 (0)	NA
Neurologic deficit	2 (7.7)	0 (0.0)	.454
Proptosis	1 (3.8%)	0 (0.0)	.985
Ptosis	2 (8.0)	0 (0.0)	.453
Ophthalmoplegia	0	0	NA
Absent pupillary reflexes	0	0	NA
Late (>5 d)
Fever	12 (50.0)	10 (40.0)	.677
Visual disturbance	5 (20.0)	1 (4.0)	.192
Ocular pain	6 (24.0)	1 (4.0)	.103
Numbness	5 (20.0)	0 (0.0)	.059
Facial pain	11 (44.0)	5 (20.0)	.13
Facial swelling	8 (32.0)	3 (12.0)	.172
Cranial nerve palsy	2 (8.0)	0 (0.0)	.47
Nasal congestion	5 (20.0)	6 (24.0)	1.0
Epistaxis	2 (8.0)	2 (8.0)	1.0
Nasal discharge	2 (8.0)	2 (8.0)	1.0
Anosmia	1 (4.0)	1 (4.0)	1.0
Headache	5 (19.2)	3 (12.0)	.745
Mental status change	0 (0)	0 (0)	NA
Neurologic deficit	0 (0)	0 (0)	NA
Proptosis	6 (24.0)	0 (0.0)	.03
Ptosis	3 (12.0)	0 (0.0)	.234
Ophthalmoplegia	2 (8.0)	0 (0.0)	.47
Absent pupillary reflexes	2 (8.0)	0 (0.0)	.47
Endoscopic findings
Necrotic mucosa	9 (33.3)	1 (4.0)	.02
Ischemic (pale) mucosa	6 (22.2)	2 (8.0)	.3
Mucosal ulcer	2 (7.4)	1 (4.0)	1
Purulent discharge	8 (29.6)	5 (20.0)	.631
Nasal septum perforation	1 (3.7)	0 (0.0)	1
Extrapulmonary infection
Pulmonary infiltrates	10 (37.0)	3 (11.1)	.056
Orbital infection	6 (22.2)	0 (0.0)	.022
Brain abscess	2 (7.4)	0 (0.0)	.471

Abbreviation: NA, not applicable.

Results of radiologic assessments are presented in Table 3 and Figure 1. The median time to performance of a sinus CT scan was 2 days (IQR, 0−3 days) from symptom onset. Mucosal thickening was the most frequent radiological finding across all sinuses in both cases and controls, with the maxillary sinuses being the most frequently affected. Other common findings in both groups included nasal concha or septum soft tissue thickening and lacrimal duct thickening. Findings significantly associated with a diagnosis of IFS were swelling of the premaxillary and preauricular subcutaneous fat, observed in 27.7% of case patients versus 0% of controls (P < .001) and intraconal fat infiltration (16.6% vs 0%, P = .002). Bone changes in the cribriform plate were significantly more frequent in controls than in patients with IFS (29.6% vs 5.7%, P = .0018). Review of these cases revealed underlying chronic sinus disease in the control patients.

Table 3.

Computed Tomography Imaging Findings in Case and Control Patients

	Cases	Controls	P-value
N sites analyzed	54	54
Mucosal thickening (>4 mm)
Frontal	12 (22.2)	20 (37.0)	.13
Maxillary	43 (79.6)	42 (77.7)	1.0
Anterior ethmoid	23 (42.5)	22 (40.7)	1.0
Posterior ethmoid	17 (31.4)	14 (25.9)	.67
Sphenoid	14 (25.9)	16 (29.6)	.83
Fluid level
Frontal	1 (1.8)	0 (0)	1.0
Maxillary	1 (1.8)	0 (0)	1.0
Increased density
Frontal	5 (9.2)	4 (7.4)	1.0
Maxillary	13 (24.0)	10 (18.5)	.63
Anterior ethmoid	3 (5.5)	2 (3.7)	1.0
Posterior ethmoid	5 (9.2)	0 (0)	.056
Sphenoid	5 (9.2)	10 (18.5)	.26
Bone changesa
Frontal	3 (5.5)	0 (0)	.24
Maxillary	16 (29.6)	16 (29.6)	1.0
Anterior ethmoid	13 (24.0)	20 (37.0)	.20
Posterior ethmoid	9 (16.6)	10 (18.5)	1.0
Sphenoid	7 (12.9)	4 (7.4)	.52
Pterygoid plate	4 (7.4)	2 (3.7)	.67
Greater sphenoid wing	1 (1.8)	4 (7.4)	.36
Cribriform plate	3 (5.7)	16 (29.6)	.0018
Fat infiltration
Frontal	6 (11.1)	2 (3.7)	.27
Anterior ethmoid	6 (11.1)	10 (18.5)	.41
Pterygopalatine fossa	14 (25.9)	10 (18.5)	.48
Sphenopalatine foramen	11 (20.3)	6 (11.1)	.29
Retromaxillary	15 (27.7)	16 (29.6)	1.0
Subcutaneous infiltration
Preauricular and premaxillary	20 (37.0)	0 (0)	<.001
Nose	17 (31.4)	14 (25.9)	.67
Orbital structures
Proptosis	6 (11.1)	6 (11.1)	1.0
Inferior orbital fissure	14 (25.9)	18 (33.3)	.52
Intraconal fat infiltration	9 (16.6)	0 (0)	.002
Extraconal fat infiltration	6 (11.1)	2 (3.7)	.27
Ocular muscle thickening	6 (11.1)	2 (3.7)	.27
Orbit bone destruction	0 (0)	2 (3.7)	.49
Lacrimal gland thickening	6 (11.1)	4 (7.4)	.74
Lacrimal duct infiltration	24 (44.4)	22 (40.7)	.69
Preseptal thickening	13 (24.0)	12 (22.2)	1.0
Orbital apex infiltration	4 (7.4)	4 (7.4)	1.0
Masticator space muscle thickening	8 (14.8)	8 (14.8)	1.0
Cavernous sinus involvement	4 (7.4)	0 (0)	.11
Nasal septum/concha thickening	25 (46.2)	32 (59.2)	.24

Values in bold represent statistically significant differences between case and control groups.

^aBone changes include erosion, thinning, irregularity, and dehiscence.

Figure 1.

Radiological findings on sinus computed tomography of patients with invasive fungal sinusitis. Arrows indicate main findings. A, Ocular fat infiltration and proptosis. B, Mucosal thickening (>4 mm) in the maxillary sinus. C, Premaxillary fat infiltration. D, Premaxillary fat infiltration (arrows) and thickening of premaxillary muscles (levator labii superioris and ala nasi, arrowhead).

Overall, blinded radiologists assessed a similar proportion of case and control patients as having IFS (both assessors, 16/27 and 15/27, respectively, P = .78). When asked to grade their level of suspicion, a high suspicion of IFS was reported by radiologist 1 for 55.5% of cases versus 37.0% of controls (P = .035) and by radiologist 2 for 40.7% of cases versus 37.0% of controls (P = .14). Interobserver agreement between the radiologists varied significantly across radiological parameters. The median Cohen's kappa was 0.73, with an IQR of 0.48–0.86 (Supplementary Figure 1). Agreement was highest for maxillary sinuses (median K 0.87, IQR 0.81–0.72), followed by sphenoid (K 0.76, IQR 0.52–0.88), ethmoid (K 0.72, IQR 0.48–0.84), and frontal sinuses (K 0.59, IQR 0.37–0.80). Cohen's kappa for premaxillary/preauricular and intraconal fat infiltration was 0.87 and 0.66, respectively.

Clinical and radiologic variables that showed an association with IFS at a significance level of P < .1 were assessed for inclusion in a multivariable regression model. The final model included 3 variables: ocular pain (odds ratio, 36.5; P ampa#thinsp;.0018), premaxillary or preauricular fat infiltration (odds ratio, 61.5; P < .001), and necrotic mucosal lesion on endoscopic examination (odds ratio, 14.8; P = .0031; Figure 2). Sensitivity analyses were conducted to evaluate the impact of hospitalization in the hematology department, receipt of chemotherapy, and neutropenia on model performance, given observed imbalances between cases and controls. Inclusion of these variables resulted in substantial increases in AIC, indicating reduced model parsimony without added explanatory value (Supplementary Table 1), supporting the decision to exclude them from the final model. The model had a sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio of 0.77, 0.96, 21, and 0.231, respectively. The area under the receiver-operator characteristics (ROC) curve was 0.88, indicating good overall performance. Results of Bayesian ridge regression were consistent with the primary regression model. All 3 predictors contributed meaningfully to the outcome, with premaxillary/preauricular infiltration having the largest posterior effect size (95% credible interval [CrI], 0.41–3.16) followed by endoscopic evidence of mucosal necrosis (95% CrI, 0.27–2.85) and ocular pain (95% CrI, 0.05–3.05; Figure 2). These predictors remained independently associated with IFS after adjusting for variables identified in univariate analyses, including neutropenia, hematology department admission, and pulmonary infiltrates (Supplementary Table 2).

Figure 2.

Regression models for invasive fungal sinusitis. A, Results of logistic regression using Firth's penalized likelihood method. Odds ratios and confidence intervals are displayed for each variable found to contribute significantly to model performance. B, Bayesian ridge regression performed to validate variables identified in the logistic regression model. Histograms show the posterior distributions of each variable. Posterior probabilities indicate the likelihood that each variable positively affects the log-odds of the model outcome. Trace plots show 4 Markov chains with 1500 iterations for each parameter.

DISCUSSION

Invasive fungal sinusitis is associated with high mortality rates, which can be mitigated by prompt sinus debridement surgery and initiation of appropriate antifungal therapy [7, 9]. However, the nonspecific signs and symptoms characteristic of early IFS often lead to delayed diagnosis, facilitating disease progression, intracranial involvement, and increased mortality [5, 7]. To address this problem, we systematically compared patients with IFS to controls without IFS and evaluated early clinical and radiological features with the aim of developing a prediction model. Ocular pain, premaxillary fat infiltration viewed on a CT scan, and an endoscopic finding of a mucosal eschar emerged as independent and significant predictors of IFS.

Several previous studies have attempted to define clinical and radiological indicators of IFS. Mahomva et al. observed that prolonged fever, facial pain, low platelet count, low neutrophil count, and abnormal CT findings were associated with invasive fungal infection among patients undergoing biopsy for suspected IFS [6]. DelGaudio et al [8] identified severe unilateral thickening of the nasal cavity mucosa in most patients with IFS, while noting that this finding was also frequent in other forms of rhinosinusitis. Findings of bone erosion and extrasinus extension were rare. The authors concluded that CT evaluation is of limited utility early in the course of IFS, underscoring the importance of prompt endoscopy with biopsy and empirical antifungal treatment in patients with suspected IFS [8]. Silverman and Mancuso [11] described 2 patients in whom periantral fat infiltration was the only early radiological sign for IFS, whereas this finding could not be detected in 112 control CT images. Expanding on these findings, the same group later found a significant correlation between IFS and involvement of the pterygopalatine fossa, periantral fat, nasolacrimal duct, and lacrimal sac [12]. A higher degree of mucosal opacity in multiple sites correlated positively with IFS. They developed a 7-point prediction model that included, in addition to the aforementioned variables, orbital involvement, septal ulceration, and bone dehiscence.

In contrast to earlier studies, we developed a comprehensive prediction model by integrating detailed radiological assessments with clinical variables. Consistent with prior reports [8, 11, 12], early CT scans did not reveal a higher frequency of bone erosion or extension across sinus walls in patients with IFS versus controls. Mucosal thickening was the most frequent radiological finding in patients with IFS, but as reported previously [8], it was similarly common in patients with non-IFS-related sinus symptoms. Unexpectedly, bone changes in the cribriform plate were more frequent in controls than in IFS cases. These findings likely reflected bone remodeling secondary to unrecognized chronic sinus diseases. In contrast, fat infiltration involving intraconal, premaxillary, or preauricular tissue was significantly associated with IFS. The logistic regression model incorporating premaxillary/preauricular fat infiltration with clinical variables of ocular pain and mucosal eschar provided very good sensitivity and specificity for IFS, as reflected in the area under the ROC curve of 0.88. Interobserver agreement was high overall, with a kappa value of 0.87 for subcutaneous fat infiltration, underscoring the model's robustness.

Collectively, these results can be used to support a clinical pathway for the diagnosis and management of IFS. Invasive fungal sinusitis should be suspected in high-risk patients presenting with new-onset sinus symptoms, particularly when accompanied by ocular pain. Further evaluation of such patients with sinus CT and nasal endoscopy is warranted. Imaging findings of premaxillary or preauricular fat infiltration or identification of a mucosal eschar on endoscopy are highly suggestive of IFS. Prompt mucosal biopsy and initiation of empirical antifungal therapy are recommended while awaiting microbiological and histopathological confirmation. Other findings reported in the literature, such as proptosis, nasal discharge, and cranial nerve palsies [9], represent advanced disease with extensive orbital or intracranial extension. Therefore, these findings have limited utility in the early stages of disease.

This study is limited by a relatively small sample size, and the fact that model testing was not performed on an independent patient cohort. However, alignment of model components, specifically orbital pain [6], periantral fat infiltration [11, 12], and mucosal eschar with previous studies, as well as internal validation using Bayesian ridge regression, provide support for our findings. External validation should be performed on an independent patient cohort exhibiting comparable clinical characteristics, with data collection following standardized ENT and imaging protocols consistent with those employed in the current study, to ensure methodological alignment and reproducibility. We did not analyze MRI, which may be a more sensitive modality than CT for detecting orbital or intracranial invasion [13]. In our center, CT scans are the initial imaging modality for patients with suspected IFS, whereas MRI is performed later, often following positive findings on sinus CT. Thus, case patients were more likely to receive MRI than controls, implying that analysis of MRI images would be unbalanced. Case−control matching by neutropenic status was imperfect, reflecting limitations in the available control pool, which restricted our ability to achieve optimal 1:1 matching on both age range and neutropenia. A higher proportion of chemotherapy in the control versus the case group could represent a lower clinical threshold for investigating mild sinus symptoms among controls, potentially introducing spectrum bias and inflating model specificity. However, sensitivity analysis incorporating these imbalanced covariates did not reveal a significant impact on overall model performance. Finally, because the study cohort was restricted to patients with malignancy or stem-cell transplantation, the findings should not be generalized to other immune compromised and immune competent populations.

In summary, this study identified 2 clinical and 1 radiological feature that differentiated IFS cases from controls. These features may be integrated into a predictive model to support clinical decision-making. Future research should focus on validating these and other predictors in larger patient cohorts, with the goal of optimizing the use and interpretation of imaging studies and nasal endoscopy to facilitate early diagnosis and improve patient outcomes.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.