The predictive value of clinical and radiographic features in fungal pathogen identification in immunocompromised patients with pulmonary invasive fungal infection

Abstract

Objective To investigate the predictive value of clinical and radiographic features in fungal pathogen identification in immunocompromised patients with pulmonary invasive fungal infection (IFI). Methods All consecutive immunocompromised adult patients with pulmonary IFI in respiratory intensive care unit (ICU) in the First Affiliated Hospital of Xinjiang Medical University were recruited during a 2 year period. All patients met the 2008 European Organization for Research and Treatment of Cancer and Mycoses Study Group (EORTC/MSG) criteria were studied for proved or probable IFI responding to antifungal agents. The data of demographic, clinical and radiographic features, as well as serological test results of the patients were collected. Differences in the clinical and radiographic features of pulmonary IFIs caused by yeasts and molds were compared by χ2 test. A logistic regression model was used to perform discriminant analysis, and the effect of discrimination was assessed for accuracy. Results The study included 143 patients with a probable diagnosis of IFI who had the following risk factors: diabetes mellitus (43.4%), chronic lung disease (32.2%), broad-spectrum antibiotics administration (≥14 days; 35.7%), malignancy (23.1%), corticosteroid therapy (≥14 days; 23.1%), chronic renal failure and renal replacement therapy (16.1%), and immunological disease (10.5%). Frequent broad-spectrum antibiotics administration was associated with yeast infection (P<0.05), while mold infection was associated with chronic lung disease (P<0.05). Yeast was more often isolated from patients with concurrent bacterial infection and on mechanical ventilation (P<0.05). Thoracic high-resolution computed tomography (HRCT) showed the following images: bronchial pneumonia/pulmonary consolidation (53.1%), massive shadowing (29.4%), small nodules (24.5%), large nodules (18.9%), pleural effusion (18.9%), halo sign (14%), and cavity (9.8%). Imaging showed that mold was more common than yeast in patients with pleural and pericardial effusions (P<0.05). Logistic regression modeling showed that broad-spectrum antibiotics administration, prolonged mechanical ventilation, and pleural and pericardial effusions were statistically significant in fungal identification (P<0.05), with a predictive accuracy of 77.6%. Conclusions For immunocompromised patients with pulmonary IFI, most of the risk factors , the main clinical and chest HRCT features did not help to predict the type of fungal pathogen, and yeast but not cryptococcus may be accompanied or colonized. Key words: Invasive fungal infection; Pathogens; Pneumonia; Immunocompromised; Clinical features; Radiographic features