|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2001 American Society for Clinical Oncology
Risk Factors in Children With Fever and Neutropenia
Janssen Research Foundation, Titusville, NJ To the Editor:We read with interest the article by Santolaya et al,1 in which the authors compare their model of the risk for invasive bacterial infection (IBI) to the model that we identified for bacteremia in children with cancer, fever, and neutropenia.2 We disagree with the authors’ conclusion that high fever and low absolute monocyte count (AMoC) would not have been significant risk factors in the study by Rackoff et al2 "if more children had been included."1 One of us recently re-evaluated our original bacteremia risk model2 using data from the same institution, but in a new time frame during which there was a significantly different rate of bacteremia.3 Although the model specifications change somewhat in the re-evaluation, the new study corroborates the original finding that admission AMoC and temperature are significant, independent predictors of bacteremia. Including the re-evaluation, our original study,2 and the study at the Hospital for Sick Children in Toronto, Canada,4 there are five cohorts (total no. of episodes, 692) in which AMoC predicts the risk of bacteremia. The risk factors for IBI1 are different from those reported by Rackoff et al2 because the models differ, not the sample sizes. First, the IBI study population includes patients with probable as well as demonstrated infections of the blood and a number of other organ systems. Our models include only patients with positive blood cultures. Second, 19% of the patients were receiving granulocyte colony-stimulating factor in the Santolaya study, and 55% had an indwelling venous catheter. In our study, granulocyte colony-stimulating factor was in use before admission in 67% of the patients, and virtually all episodes were associated with indwelling venous catheter use. Thus, the definition of the primary outcome and the underlying characteristics of the population differ significantly. In all likelihood, these differences, and not the difference in sample sizes, explain the differences in the models. REFERENCES
1.
Santolaya ME, Alvarez AM, Becker A, et al: Prospective, multicenter evaluation of risk factors associated with invasive bacterial infection in children with cancer, neutropenia, and fever. J Clin Oncol 19: 3415-3421, 2001
2.
Rackoff WR, Gonin R, Robinson C, et al: Predicting the risk of bacteremia in children with fever and neutropenia. J Clin Oncol 14: 919-924, 1996 3. Madsen K, Rosenman M, Hui S, et al: Value of electronic data for model validation and refinement: Bacteremia risk in children with fever and neutropenia. J Pediatr Hematol Oncol (in press)
4.
Klaassen RJ, Goodman TR, Pham B, et al: "Low-risk‘ prediction rule for pediatric oncology patients presenting with fever and neutropenia. J Clin Oncol 18: 1012-1019, 2000
ResponseUniversity of Chile, Santiago, Chile In Reply:We thank Drs Rackoff and Breitfeld for adding new information that disputes our possible explanation as to why high fever and absolute monocyte count (AMC) were significant predictors of bacteremia in his study but not of invasive bacterial infection (IBI) in ours.1,2
The article by Rackoff et al1 concluded that an AMC less than 100/mm3 and a temperature of With the new information added by Rackoff’s group, although not yet available, and from the article by Klaassen et al,3 we tend to agree with Rackoff and Breitfeld that admission AMC less than 100/mm3 seems to be a good predictor for risk of bacteremia and significant bacterial infection as defined by both studies, although not for IBI as defined by us for our population. On the other hand, fever greater than 39°C seems to be less strongly associated with the risk for significant and invasive bacterial infection as reported by both Klaassen et al3 and our study.2 The variable may be associated with a risk factor for bacteremia, Rackoff’s end point. In this context, we also agree with Rackoff and Breitfeld that our models may be different because our primary outcome and some characteristics of the population differ significantly. Our model is more comparable to that used by Klaassen et al, considering that our outcome is risk for IBI and theirs is risk for "significant bacterial infection," both of which included bacteriologically demonstrated as well as nondemonstrated, albeit probable bacterial infections. We must point out that when we considered only demonstrated IBI, defined as a positive bacterial culture obtained from a usually sterile site (eg, blood, indwelling catheter, urine, and CSF), neither monocyte count nor temperature was an independent risk factor for IBI. Finally, we stress the concept that predictive models have to be prospectively evaluated in populations that share characteristics with the study population. In our common aim to consider a selective approach for febrile neutropenic children, only this type of evaluation will confirm if our models are useful for clinical practice. REFERENCES 1. Rackoff WR, Gonin R, Robinson C, et al: Predicting the risk of bacteremia in children with fever and neutropenia. J Clin Oncol 14: 919-924, 1996 2. Santolaya ME, Alvarez AM, Becker A, et al: Prospective, multicenter evaluation of risk factors associated with invasive bacterial infection in children with cancer, neutropenia, and fever. J Clin Oncol 19: 3415-3421, 2001 3. Klaassen RJ, Goodman TR, Pham B, et al: ‘Low-risk‘ prediction rule for pediatric oncology patients presenting with fever and neutropenia. J Clin Oncol 18: 1012-1019, 2000
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2001 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
39°C were significantly associated with a higher risk for bacteremia. Our comments to this article were that the 95% confidence interval for the odds ratio comparing high-risk versus intermediate-risk groups in that study was wide (1.6 to 12.9) and that both variables in our study reached significance in the univariate but not multivariate analysis. We thus postulated that a similar solution could have occurred in the study by Rackoff et al if more children had been included. 
