Figure 1. The sensitivities (A) and specificities (B) of the different imaging characteristics evaluated are reported for each observer. Error bars represent 95% confidence intervals. CT indicates computed tomography; MRI, magnetic resonance imaging.
Figure 2. Findings from T1-weighted axial postgadolinium magnetic resonance imaging (MRI) with fat saturation. A, T1-weighted axial postgadolinium MRI with fat saturation in a patient with acute lymphoblastic leukemia and subsequently histopathologically diagnosed acute fulminant invasive fungal sinusitis (AFIFS). The left middle turbinate (arrow) and left maxillary sinus mucosa (curved arrow) show extensive loss of contrast enhancement (LoCE). Sinonasal mucosa should always show enhancement on gadolinium-enhanced MRI. Notice also subtle stranding of premaxillary fat on the left side. B, T1-weighted axial postgadolinium MRI with fat saturation in a patient with acute myelogenous leukemia and subsequently histopathologically diagnosed AFIFS. A focal area of LoCE at the anterior aspect of the left middle turbinate (small arrow) is readily apparent. Fungal material within the left maxillary sinus (large arrow) and invasion into the left retromaxillary space and pterygopalatine fossa (curved arrow) are also apparent. C, T1-weighted axial postgadolinium MRI with fat saturation in a patient with Burkitt lymphoma and subsequently histopathologically diagnosed AFIFS. A focal area of LoCE of the left ethmoid sinus (small arrow) is apparent. Proptosis of the left eye and thickening and enhancement of the left medial rectus (large arrow) and preseptal, periorbital soft tissue (curved arrows) are also seen.
Figure 3. Overall survival of all patients with suspected acute fulminant invasive fungal sinusitis calculated by Kaplan-Meier product limit estimate. Steps indicate deaths and hash marks indicate censor events.
Groppo ER, El-Sayed IH, Aiken AH, Glastonbury CM. Computed Tomography and Magnetic Resonance Imaging Characteristics of Acute Invasive Fungal Sinusitis. Arch Otolaryngol Head Neck Surg. 2011;137(10):1005-1010. doi:10.1001/archoto.2011.170
Author Affiliations: Departments of Otolaryngology, Head and Neck Surgery (Drs Groppo, El-Sayed, and Glastonbury), Radiology & Biomedical Imaging (Dr Glastonbury), and Radiation Oncology (Dr Glastonbury), University of California, San Francisco; and Department of Radiology, Emory University School of Medicine, Atlanta, Georgia (Dr Aiken).
Objective To determine radiographic findings on computed tomography (CT) and magnetic resonance imaging (MRI) predictive of acute fulminant invasive fungal sinusitis (AFIFS) in an immunocompromised patient population.
Design Retrospective case-control study.
Setting Tertiary referral hospital.
Patients Cases were 17 immunocompromised patients with confirmed AFIFS after surgical debridement or biopsy. Controls were 6 immunocompromised patients histopathologically negative for AFIFS after surgical debridement or biopsy.
Main Outcome Measures Computed tomographic and MRI scans were independently reviewed by 2 neuroradiologists to identify imaging characteristics predictive of AFIFS. Operative reports and histopathologic, microbiologic, and survival data were reviewed.
Results No significant differences with regard to baseline characteristics between the 2 groups were identified. There was moderate or substantial agreement (κ = 0.40-0.77) between the 2 radiologists for all imaging parameters except MRI loss of contrast enhancement (κ = 0.16). Magnetic resonance imaging was more sensitive than CT for the diagnosis of AFIFS (sensitivity 85% and 86% for both reviewers compared with 57% and 69%). Extrasinus invasion with MRI was the most sensitive individual parameter (87% and 100%). Magnetic resonance imaging and CT had similar specificities, and perisinus invasion was the most specific individual parameter (83% and 83% for MRI compared with 81% and 83% for CT). The positive predictive values were high for both imaging modalities (93% and 94% for MRI compared with 89% and 93% for CT). The negative predictive values were lower for both modalities and varied more between reviewers (71% and 100% for MRI compared with 45% and 67% for CT).
Conclusions Magnetic resonance imaging is more sensitive for detecting early changes of AFIFS than CT. Both imaging modalities have similar specificities. Perisinus invasion with MRI was the most sensitive and specific single parameter evaluated.
Acute fulminant invasive fungal sinusitis (AFIFS) is a relatively uncommon disease primarily affecting immunocompromised individuals.1- 4 The most common predisposing factor for AFIFS is neutropenia, commonly associated with clinical conditions or treatments such as leukemia, bone marrow transplant, chronic immunosuppressive therapy, and AIDS.3,4 Functional neutropenia, such as that encountered in poorly controlled diabetes mellitus and diabetic ketoacidosis, is another well-known risk factor.5 The rapidity of progression, significant morbidity, and high mortality rate associated with AFIFS make it one of the most aggressive infections of the head and neck and one of the most difficult to treat. Treatment usually consists of aggressive surgical debridement, intravenous antifungal therapy, and correction of the underlying comorbidity that has resulted in compromised immunity when possible.6,7 Historically, the mortality of AFIFS has been cited as 50% to 80%.4 Recent series have reported a decreased mortality from 7% to 18%.5,8 This decrease is largely attributed to earlier detection and intervention.8,9
Accurate early diagnosis of AFIFS is difficult because the presenting symptoms of AFIFS are similar to routine viral or bacterial rhinosinusitis. In part because of the overlap in symptoms with indolent rhinosinusitis, as well as the spectrum of physical findings, a diagnosis of AFIFS hinges on histopathologic evidence of fungi invading nasal tissue (hyphal forms invading sinus mucosa, submucosa, blood vessel, or bone).10,11 Unless the disease process is widespread or confined to anterior nasal tissue, obtaining suitable tissue often requires a surgical biopsy.12 At-risk patients frequently have significant comorbidities including coagulopathy from thrombocytopenia. These comorbidities place them at elevated risk for intraoperative and postoperative complications.
Identifying noninvasive techniques to accurately evaluate high-risk patients displaying symptoms compatible with AFIFS is of utmost importance, both for early detection and to avoid unnecessary surgery in comorbid patients. The current imaging modalities advocated for AFIFS are computed tomography (CT) and magnetic resonance imaging (MRI). Previously described early CT findings of AFIFS include unilateral sinus opacification and soft-tissue thickening of the mucosa of the nasal floor and lateral nasal wall.13 Described signs of AFIFS on MRI include obliteration or infiltration of periantral fat, inflammatory changes in the orbital fat and extraocular muscles, and leptomeningeal enhancement.14- 17 The latter 2 imaging changes reflect advanced AFIFS.
Owing to the lack of appropriate control groups, to our knowledge, there are currently no studies reporting sensitivity and specificity data for either CT or MRI, and controversy over which imaging modality is best for diagnosing AFIFS persists.8,13,15,18 The purpose of this study was to review CT and MRI scans of patients with suspected AFIFS and correlate these radiographic findings with those found on endoscopy and histopathologic examination. By including a control group, we compared the sensitivity and specificity of CT and MRI.
A retrospective case-control medical chart review was performed for all patients who underwent operative endoscopy to rule out or diagnose AFIFS at the University of California, San Francisco, between 1999 and 2009. Institutional review board approval was obtained. Inclusion criteria included immunocompromised state, suspected AFIFS based on clinical history, preoperative CT and MRI findings, and operative endoscopy with tissue biopsy. Only patients with complete data sets were reviewed. Twenty-three patients (14 men and 9 women) met these criteria. Demographic data, underlying diagnosis, identified fungal organism(s), comorbid condition(s), absolute neutrophil count, CT and MRI scans, endoscopy findings, and patient outcomes were reviewed. Cases were defined as patients with histopathologic evidence of AFIFS (hyphal forms within sinus mucosa, submucosa, blood vessel, or bone). Controls were defined as patients with clinically suspected AFIFS but with a negative histopathologic biopsy result.
All radiographic and clinical material was retrospectively evaluated. Absolute neutrophil count was examined only in patients with a history of hematopoietic malignancy, solid organ transplant, or bone marrow transplantation. CD4 cell number was examined only in patients with human immunodeficiency virus (HIV). Each CT and MRI scan was independently reviewed by 2 attending neuroradiologists (A.H.A. and C.M.G.). The film reviewers were blinded to initial radiology reports, operative findings, histopathologic findings, and survival data. Each scan was evaluated for extrasinus involvement including subtle infiltration of periantral fat, soft-tissue enhancement, orbital enhancement beyond normal extraocular muscle enhancement, and leptomeningeal enhancement. Magnetic resonance images were also specifically evaluated for focal areas of loss of contrast enhancement (LoCE) of the sinonasal mucosa, which is normally seen to enhance with gadolinium or iodinated contrast. Computed tomographic scans were all postcontrast images obtained on multidetector scanners following the injection of iodinated contrast. All MRI scans were performed on 1.5T magnets. Routine imaging sequences were obtained with dedicated sinus views: axial T1-weighted images, axial and coronal T2-weighted images with fat saturation, and postcontrast axial and coronal T1-weighted images with fat saturation.
Endoscopic results were obtained directly from operative reports. Histopathologic and microbiologic results were obtained directly from transcribed results as well. Survival data were collected from clinical records.
Statistical comparisons of demographic data between the case and control groups were made using χ2 (dichotomous variables) and unpaired t tests (continuous variables), with a significance level of .05. Agreement between the 2 reviewers was compared using the Cohen κ coefficient. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated by comparing radiographic parameters with histopathologic results (gold standard). Exact binomial 95% confidence intervals were calculated. Actuarial survival data were analyzed using Kaplan-Meier product limit estimates.
Magnetic resonance imaging findings of focal areas of LoCE were correlated with endoscopic findings during surgery. Agreement between LoCE results and endoscopic findings was defined as (1) any description in the operative report of abnormal mucosa (excluding inflammation) in the anatomic region identified to have focal LoCE on MRI or (2) a description in the operative report of normal mucosa throughout the nose and paranasal sinuses and no findings of LoCE on MRI. A percentage agreement score between LoCE and abnormal mucosa on endoscopy was calculated.
A total of 23 patients met inclusion criteria (17 with histologically proven AFIFS and 6 controls). There were no significant differences in demographic data (Table 1). Seventeen of the patients had a history of hematopoietic malignancy, solid organ transplant, or bone marrow transplant. Of these, 12 had severe neutropenia (absolute neutrophil count <250/μL). Of the 6 control patients, histopathologic analysis confirmed a final diagnosis of rhinosinusitis in 4 patients and metastases in 2 patients (sinonasal B-cell lymphoma and acute lymphoblastic leukemia). Rhizopus and Aspergillus were the most common organisms isolated from patients with AFIFS (Table 2).
There was moderate or substantial agreement (κ = 0.40-0.77) between the 2 radiologists for all imaging parameters except MRI LoCE, for which there was slight agreement (κ = 0.16) (Table 3). The sensitivity, specificity, PPV, NPV, and accuracy of CT and MRI are reported in Table 4. Magnetic resonance imaging was more sensitive than CT for the diagnosis of AFIFS (sensitivity 85% and 86% for both reviewers compared with 57% and 69%) (Figure 1A). Extrasinus invasion with MRI was the most sensitive individual parameter (87% and 100%). Magnetic resonance imaging and CT had similar specificities, and extrasinus invasion was the most specific individual parameter (83% and 83% for MRI compared with 81% and 83% for CT) (Figure 1B). The PPVs were high for both imaging modalities (93% and 94% for MRI compared with 89% and 93% for CT). The NPVs were lower for both modalities and varied more between reviewers (71% and 100% for MRI compared with 45% and 67% for CT). Loss of contrast enhancement agreed with endoscopic mucosal findings 76.5% of the time. Representative MRI images showing extrasinus invasion and focal areas of LoCE are shown in Figure 2.
The in-hospital mortality rate was 29.4% for patients with AFIFS and 33.3% in the control group. All 5 deaths prior to discharge in the AFIFS group were attributable to fungal infection. Two patients in the control group died prior to discharge from complications of relapsed hematopoietic malignancy. The overall survival at 5 years (60 months) for all patients was 36.6% (Figure 3). The overall survival at 5 years for AFIFS and control patients was 39.2% and 29.2%, respectively. This difference was not significant.
The radiologic characteristics of invasive fungal sinusitis have been described in both the radiology and otolaryngology literature. Early reports focused on boney destruction, best evaluated with CT.19- 21 Bone destruction, however, occurs relatively late in the disease process, and late diagnosis may partially explain the historically high mortality of 50% to 80%. More recent reports have focused on more subtle earlier manifestations of the infiltrative nature of this fulminant infection and specifically the development of soft-tissue abnormalities outside the confines of the sinuses. This is most notably found anterior or posterior to the maxillary sinus walls.13,15 These findings likely occur earlier in the natural history of the disease process, and the detection of extrasinus involvement may be contributing to earlier diagnosis and improved survival.8 Some authors have argued for early MRI scanning in patients with suspected AFIFS based on the possibility of underestimation of disease with CT.18 Indeed, one of the inherent advantages of MRI over CT is that it has better soft-tissue contrast resolution, which is the ability to detect subtle differences in tissues of similar appearance. However, the best imaging modality for detecting AFIFS in immunocompromised patients has yet to be established.8,13,15,18
In the present study, CT and MRI findings of extrasinus invasion and LoCE were compared against histopathologic diagnosis, which remains the gold standard. To our knowledge, this study is the first to compare an AFIFS case group with a similarly comorbid control group in which a histopathologically negative biopsy had been established. Previous studies have used control groups comprising immunocompromised patients undergoing sinus imaging, but no histopathologic data were available for these groups.13 As such, sensitivity and specificity data could not be accurately assessed against the gold standard of histopathologic analysis for establishing the diagnosis. This is a crucial distinction of this study, which aimed to replicate the true clinical imaging scenario.
While both CT and MRI were found to have similar specificities, MRI was more sensitive than CT in the diagnosis of AFIFS. Extrasinus invasion seen on MRI was the most sensitive and specific single imaging parameter evaluated. Extrasinus invasion with MRI failed to detect 2 of 17 cases (reviewer 1) and 0 of 17 cases of AFIFS (reviewer 2). Importantly, the only false-positive outcome identified with MRI extrasinus invasion (observed by both reviewers) was a case of sinonasal lymphoma, another invasive disease.
A high sensitivity is particularly important when choosing a modality for screening immunocompromised patients with sinus and facial signs and symptoms, since missing a case of AFIFS could lead to delayed initiation of treatment and poor outcome. Our results indicate that MRI appears to be better than CT at screening for AFIFS. Magnetic resonance imaging has the added benefit of conferring no ionizing radiation exposure, which is important for all patients who might be undergoing many studies over their lifetime given the multiple comorbidities and particularly important for pediatric patients. Computed tomography is often useful for operative planning and intraoperative image-guided navigation, especially in extended endoscopic surgery of the paranasal sinuses and anterior skull base in both adults and children.22 Reserving CT for treatment planning purposes in patients identified as having AFIFS by screening MRI would serve to limit both cost and radiation exposure. Furthermore, MRI alone can also be used for intraoperative image guidance, allowing radiation to be altogether avoided.
Focal areas of LoCE on MRI have previously been described in other anatomic locations and pathologic conditions including myocardial and pancreas ischemia and necrosis, but has not been described for the paranasal sinus mucosa.23,24 In the present study, LoCE was hypothesized to correlate with sinus mucosal ischemia. Early endoscopic signs of AFIFS include ischemic, dusky, or pale mucosa, with eschar and gross tissue necrosis developing later in the disease process.13 One potential theory for these endoscopic findings is the propensity of pathogenic fungi to invade blood vessels. This may result in local mucosal ischemia, surrounded by vascular congestion. As the disease process progresses, vascular invasion eventually leads to extrasinus involvement, a supposed later finding on MRI.
In the present study, MRI LoCE exhibited similar but more variable sensitivity and specificity when compared with MRI extrasinus invasion (64%-87% sensitive and 40%-83% specific for LoCE compared with 87%-100% and 83%-83% for extrasinus invasion). This variability produced a Cohen κ coefficient of 0.16, corresponding to slight agreement between reviewers. This may indicate varying thresholds for identifying LoCE between reviewers, and a review by consensus might help establish standards for LoCE identification. Despite this variability, LoCE was identified in at least 11 of 16 cases of AFIFS and is an MRI finding that should not be ignored.
Limitations of the present study include small sample size, although the case group (AFIFS) is similar in size to other reported series.13,15 Small sample size indicates both the rarity of the disease as well as the difficulty of capturing patients and controls retrospectively, given a lack of specific International Statistical Classification of Diseases and Current Procedural Terminology codes for the diagnosis and treatment of AFIFS. In this study, the ratio of cases to controls was 2.8:1, which is inversely proportional to the incidence of AFIFS in immunocompromised patients. Positive and negative predictive values depend not only on the intrinsic characteristics of the test, but also on the prevalence of the disease state being tested.25 For rare diseases such as AFIFS, PPVs may be artificially increased and NPVs may be artificially decreased.
In addition, inclusion criteria required preoperative MRI, which may have resulted in a selection bias toward more advanced disease. At our institution, CT is generally used as the primary imaging modality, and MRI is reserved for patients with either CT findings or clinical scenario highly concerning for AFIFS. This may partially explain the relatively high mortality rate (31%) compared with other contemporary series.5,8 A selection bias toward more advanced disease may also artificially increase sensitivity because advanced disease could be more clinically apparent on imaging.
Unfortunately, there are no standard criteria for radiographic diagnosis of AFIFS and the clinician must maintain a high index of suspicion. Immunocompromised patients displaying signs and symptoms of sinusitis should undergo radiographic evaluation expeditiously. The present study indicates that MRI should be used as a screening modality for AFIFS because of its higher sensitivity and zero radiation exposure. Computed tomography should be considered as a second-line tool, reserved for surgical planning and intraoperative image guidance.
Radiographic imaging has particular utility in certain clinical situations, such as those in which the clinical signs and symptoms are more suggestive of uncomplicated bacterial or viral rhinosinusitis. In these scenarios, imaging may reveal early signs of AFIFS with subtle infiltration of the facial fat anterior or posterior to the maxillary sinus walls. Radiographic findings of either extrasinus invasion or focal areas of LoCE should prompt an endoscopic evaluation and tissue biopsy to establish the diagnosis of AFIFS.
Correspondence: Christine M. Glastonbury, MBBS, Radiology and Biomedical Imaging, Otolaryngology–Head & Neck Surgery, and Radiation Oncology, University of California, San Francisco, UCSF Box 0628, Room L-358, 505 Parnassus Ave, San Francisco, CA 94143 (Christine.Glastonbury@ucsf.edu).
Submitted for Publication: May 18, 2011; final revision received July 26, 2011; accepted July 27, 2011.
Author Contributions: Dr Glastonbury had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Groppo, El-Sayed, and Glastonbury. Acquisition of data: Groppo, Aiken, and Glastonbury. Analysis and interpretation of data: Groppo, Aiken, and Glastonbury. Drafting of the manuscript: Groppo and El-Sayed. Critical revision of the manuscript for important intellectual content: Groppo, Aiken, and Glastonbury. Statistical analysis: Groppo. Administrative, technical, and material support: Glastonbury. Study supervision: El-Sayed, Aiken, and Glastonbury.
Financial Disclosure: None reported.
Funding/Support: This work was supported in part by National Institutes of Health/National Center for Research Resources UCSF-CTSI grant number UL1 RR024131.
Disclaimer: The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.
Previous Presentation: This work was presented in poster form at the Triological Society Poster Session during the Combined Otolaryngology Spring Meetings (COSM); April 28–May 2, 2010; Las Vegas, Nevada. It received the First Place Sinus-Rhinology Poster Award in the scientific poster competition.
Additional Contributions: Erin Madden, MPH, and John Kornak, PhD, of the University of California, San Francisco, Clinical and Translational Sciences Institute (UCSF CTSI) assisted with statistical analysis.