Skip to Main content Skip to Navigation

Prédire l’infection sévère lors des épisodes de neutropénie fébrile post-chimiothérapie de l’enfant : développement d’une règle de décision clinique

Abstract : Purpose: Chemotherapy-induced febrile neutropenia (FN) is known to be a risk for severe infection and death in the absence of prompt and appropriate antibiotic therapy. Immediate hospitalization for rapid institution of empirical broad-spectrum intravenous antibiotic therapy has led to reduce the mortality. However, documented or severe infections occur in only 15-25% of cases. In 2012 paediatric guidelines suggested to adapt the management of FN episodes to the infectious risk. In a previous work, none of the published clinical decision rules (CDRs) to rule out severe infections have been validated and have only rarely been tested in an external set of children. The methodological standards used to derive and validate these CDRs were a real concern. A new CDR was previously derived as a scoring system in Lille to classify the patients at high or low risk of severe infection, with a dataset collected in 2 centers in Lille, in following methodological standards. Differences between solid tumours and blood cancers were observed in children with FN for numbers and types of infections. As a result, we considered relevant to build a decision tree model to predict the low risk for severe infection with a first division that could be the type of cancer. This new decision rule was already validated in an internal set of data, but required an external validation.The aim of this project was to calibrate the CDR as a decision tree and validate its performance a posteriori in an external set of patients, using prospectively collected data from multiple centers.Methods: the methodological standards of available CDRs were first analysed. The new CDR derived on a bicentric dataset was reused to calibrate the CDR as a decision tree, using Sipina software. A prospective multicentric observational protocol funded by 72000€ provided by “la Ligue Contre le Cancer” was developed for an external validation of the CDR to expect near 100% sensitivity (Se) and a negative likelihood ratio (LR) below 0.1. The ethical regulation was followed. Thirty-one centers were recruited in France (27/30 referent centers for management of children with cancer, and 4 proximity centers fit to manage children with FN). The CDR was not applied to the included patients, and remained confidential. The data were collected on an e-CRF “capture system”. The data were captured by an assistant of clinical research and controlled by a physician researcher after the monitoring of the data in all centers. The CDR was a posteriori applied on the dataset. The performance of the CDR between validation and derivation sets of patients was analysed in terms of Se, specificity (Sp) and negative LR.Results: the methodological standards of development of a CDR were not always followed for the development of the published CDR predicting infection for FN in children. Only one CDR followed all criteria and reached the highest level of evidence, but this CDR was built in a very different population from our and was not reproducible. A decision tree model of the CDR was built to distinguish children with FN at low risk of severe infection. For children with solid tumours, the CDR had 96% Se, 59% Sp, and a negative LR at 0.07. For children with blood cancers, the CDR had 99% Se, 52% Sp, and a negative LR at 0.03.For external validation, inclusions started in 2012 until May 2016. Of the 31 centers, 23 included 1806 cases (333 severe infections [18.4%]). The retrospective application of the CDR on all included case in ongoing. A national survey was also conducted as the same time to analyse the real management of children with FN in France in order to determine the type of management that could be proposed for low risk patients when the CDR will be tested in an impact study.Conclusion: the different steps for the construction and validation of the new CDR were conducted following standards. This CDR is in progress to reach the highest level of evidence.
Document type :
Complete list of metadatas

Cited literature [83 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Tuesday, October 2, 2018 - 5:09:06 PM
Last modification on : Sunday, October 25, 2020 - 7:06:52 AM
Long-term archiving on: : Thursday, January 3, 2019 - 3:58:09 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01886250, version 1


Mathilde Delebarre. Prédire l’infection sévère lors des épisodes de neutropénie fébrile post-chimiothérapie de l’enfant : développement d’une règle de décision clinique. Médecine humaine et pathologie. Université du Droit et de la Santé - Lille II, 2016. Français. ⟨NNT : 2016LIL2S018⟩. ⟨tel-01886250⟩



Record views


Files downloads