ORIGINAL STUDIES The Role of Bronchoalveolar Lavage Galactomannan in the Diagnosis of Pediatric Invasive Aspergillosis Rishi Desai, MD,* Lawrence A. Ross, MD,*† and Jill A. Hoffman, MD*† Background: Molecular biomarkers such as the galactomannan assay are of increasing interest in clinical settings for diagnosis of invasive aspergil- losis (IA). The use of bronchoalveolar lavage galactomannan (BAL GM) is being validated in adult populations, but has not been systematically evaluated in pediatric patients. Methods: A retrospective analysis was performed of patients for whom GM assays from BAL were submitted between November 1, 2006 and November14, 2007 at Childrens Hospital Los Angeles. Medical charts were reviewed and patients were categorized as having no, possible, probable, and proven IA, per established definitions. Results: Of 85 pediatric patients who had a BAL GM submitted during the study, 59 were immunocompromised. Three patients were identified as having proven IA, 6 had probable IA, 37 had possible IA, and 39 had no evidence of IA; 38 had a concurrent serum GM performed. A positive, linear correlation was established between BAL and serum GM, using OD index values (  0.48, P  0.002). Among immunocompromised pa- tients, receiver operating characteristic curves demonstrated an optimal BAL GM OD cut-off value of 0.87, that yielded a sensitivity for probable/ proven IA of 78% and a specificity of 100%. At 0.87, the positive and negative predictive values among immunocompromised patients were 58% and 96%, respectively. Conclusions: We found a correlation between BAL GM values and a diagnosis of IA. We also noted a linear relationship between serum and BAL GM values. Receiver operating characteristic curves obtained from our pediatric data validate the current cut off for serum and suggest a possible cut off for BAL specimens. Key Words: galactomannan, bronchoalveolar lavage, pediatrics, aspergillosis (Pediatr Infect Dis J 2009;28: 283–286) I nvasive aspergillosis (IA) is an important cause of morbidity and mortality in immunocompromised pediatric patients. Timely di- agnosis is challenging because comorbid conditions and immuno- modulation often result in nonspecific signs and symptoms of disease. In addition, in pediatric patients, classic radiologic find- ings are identified at a lower incidence than in adults. 1 Although the standard for diagnosis is lung biopsy, this approach engenders risk in medically fragile, neutropenic patients and has unclear sensitivity. 2 Focus on early diagnosis has resulted in identifying surrogate markers such as galactomannan (GM), a cell wall poly- saccharide released by Aspergillus hyphae. The platelia Aspergil- lus enzyme immunoassay (PA-EIA) has been used internationally to detect serum GM, and is FDA approved for this use. Serum GM was detected by this technique in carefully controlled studies in high-risk adults at a mean of up to 8 days before diagnosis of IA by other methods. 3,4 The role of this assay has also been evaluated in bronchoalveolar lavage (BAL) performed on immunocompro- mised adults; cut-off values for BAL GM predictive of IA have ranged from 0.5 to 1. 5–7 This retrospective review was de- signed to evaluate the role of BAL GM in pediatric patients with suspected IA. METHODS A retrospective analysis was performed on all patients for whom GM assays from BAL were submitted between November 1, 2006 and November 14, 2007 at Childrens Hospital Los Ange- les. Medical records were reviewed and patients were categorized as having no, possible, probable, and proven IA, by established definitions. 8 Because the goal of the study was to evaluate BAL GM cut-off values, this was not used as microbiologic criterion for IA, despite inclusion in the definition for probable IA. Serum GM was assessed using the PA-EIA and resultant optical density (OD) measurements were recorded according to the manufacturer’s instruc- tions. Clinical samples were frozen at 70°C, and batch run twice weekly. Samples with a serum GM value 0.5 in patients with no prior positive values were repeated on the same clinical sample and on a new clinical sample. This confirmatory analysis was not done for BAL GM values 0.5. Mean BAL GM values were compared among patients with no, possible, and probable/proven IA to generate box plots and confidence intervals. Receiver operating characteristic (ROC) curves were generated using a STATA statistical package for BAL GM and serum GM. A second set of curves was generated for the subset of the patient population that was immunocompromised. Based on these curves, optimal cut-off values were selected and corresponding sensitivity and specificity for BAL GM and serum GM results were calculated. A Spearman analysis was conducted among patients with BAL GM and serum GM assays done con- currently. The effect of piperacillin/tazobactam on serum and BAL GM values was evaluated using 2-sample t test analysis. The study was approved by the Childrens Hospital Los Angeles Institutional Review Board in its entirety and informed consent was not required. RESULTS Eighty-five BAL GM assays were submitted at the discre- tion of the attending clinicians. Fifty-nine samples were performed on immunocompromised patients, most of whom had hematologic malignancies. An additional 26 immunocompetent patients also had BAL GM submitted (Tables 1, 2). Three patients were iden- tified as having proven IA, 6 had probable IA, and 37 had possible IA. Thirty-nine patients had no evidence of IA. Mean BAL GM values for the probable/proven IA, possible IA, and no IA groups were 3.4, 0.72, and 0.34, respectively (P  0.002). ROC data demonstrated an optimal BAL GM OD cut-off value of 0.98 that yielded sensitivity for probable/proven IA of 78% and a specificity of 92% (Fig. 1). At this cut off, the positive predictive and negative Accepted for publication September 24, 2008. From the *Division of Infectious Disease, Childrens Hospital Los Angeles, Los Angeles, CA; and †Department of Pediatrics, Keck School of Medicine, University of Southern California, Los Angeles, CA. Address for correspondence: Jill Hoffman, MD, Division of Infectious Diseases, Childrens Hospital Los Angeles, 4650 West Sunset Blvd, MS #51, Los Angeles, CA 90027. E-mail: jhoffman@chla.usc.edu. Copyright © 2009 by Lippincott Williams & Wilkins ISSN: 0891-3668/09/2804-0283 DOI: 10.1097/INF.0b013e31818f0934 The Pediatric Infectious Disease Journal • Volume 28, Number 4, April 2009 283