Risk Index Score for Bacteremia in Febrile Neutropenic Episodes in Children with Malignancies

BACKGROUND: Nowadays febrile neutropenia is regarded as heterogeneous syndrome and adequate risk stratification of patients as a basis for tailoring empiric antibiotic therapy is attempted. We conducted this prospective study to determine risk factors for bacteremia in febrile neutropenic children that could be assessed in the first days from the onset of fever and based on logistic regression analysis to construct a risk index score for prediction of bacteremia.

MATERIAL AND METHODS: We studied prospectively 199 consecutive episodes of febrile neutropenia for a 4-years period (2000 – 2004), occurring in 80 children with malignancies, treated by conventional chemotherapy. A standardized computer database with a set of variables for each febrile neutropenic episode was used. C-reactive protein was measured at the first fever and depending on its duration - on the 3^-rd and 5^-th day from the beginning of antibiotic therapy by immunoturbidimetric method. At the onset of fever and before initiation of antibiotic therapy blood cultures were drawn and injected directly into a culture bottle BACTEC Peds plus.

RESULTS: By multivariate logistic regression analysis we determined 5 variables as independent risk factors for bacteremia: the underlying malignant disease (leukemia, NHL-gr.IV), chills, perianal cellulitis, presence of central venous catheter and degree of CRP rise ≥34.5 mg/L between the onset of fever and the 3^rd day of empiric antibiotic therapy. This logistic regression model has sensitivity of 93%, specificity of 25%, negative predictive value of 89% and positive predictive value of 36% for the presence of bacteremia (Chi-square: 44, −2 log likelihood:200, Cox & Snell R square: 0.2).

CONCLUSIONS: The present prospective study is characterized by prevalence of hematologic malignancies and relatively high frequency of bacteremia. We identified a low-risk group of 19.1%, comparable to the reported by other groups. Based on the statistical significance of serial measurement of C-reactive protein, which allows definite risk stratification of febrile neutropenic episodes on the 3^rd day from the onset of fever, we would recommend sequential empiric antibiotic approach with switch to oral or even outpatient treatment in the low-risk group on the 3^rd day.