• Roland A Ammann, Andreas Hirt, Annette Ridolfi Lüthy, and Christoph Aebi.
• Division of Pediatric Hematology and Oncology, Department of Pediatrics, University of Bern, Bern, Switzerland. roland.ammann@insel.ch
• Med. Pediatr. Oncol. 2003 Nov 1; 41 (5): 436-43.

BackgroundFebrile neutropenia (FN) remains a frequent complication in pediatric oncology, requiring emergency hospitalization and empirical broad spectrum antibiotics. The distinction of patients at high versus low risk for severe bacterial infection (SBI) is not fully established. The purpose of this study was to define a rule predicting the risk to develop SBI in children and adolescents with FN, based on information accessible at presentation.ProcedureInformation accessible within 2 hr from presentation was collected retrospectively on all pediatric cancer patients presenting with FN from 1993 to 2001 in a single institution. Patients with established SBI at presentation were excluded. After univariate analyses, two multivariate models predicting the risk of SBI were constructed and their performance evaluated using crossvalidation.ResultsAn SBI developed in 106 (37%) of 285 episodes of 111 children. The logistic regression model outperformed the decision tree model in predicting SBI. It was based on seven variables: bone marrow involvement, no clinical signs of viral infection, high level of C-reactive protein, high hemoglobin, low leukocyte count, presence of central venous catheter, and diagnosis of pre-B-cell leukemia. At 96% sensitivity, the crossvalidated specificity was 26%, and the negative predictive value 91%.ConclusionsCombining selected information accessible at presentation, SBI in FN can be predicted with clinically useful specificity maintaining very high sensitivity. Induction therapy and absence of clinical signs of viral infection were identified as new predictors of SBI. The results of this study need confirmation and refinement in prospective studies, aiming at more selective management of FN in pediatric oncology.Copyright 2003 Wiley-Liss, Inc.

31 keywords...

hide…