Background: 

Invasive fungal diseases (IFDs) are opportunistic infections that result in significant morbidity and mortality in pediatric oncology patients. Predictive risk tools for IFD in pediatric cancer are not available.

Methods: 

We conducted a 7-year retrospective study of pediatric oncology patients with a diagnosis of febrile neutropenia at UCM Comer Children’s Hospitals. Fourteen clinical, laboratory, and treatment-related risk factors for IFD were analyzed. Stepwise variable selection for multiple logistic regression was used to develop a risk prediction model for IFD. Two comparative analyses have been conducted: (i) all suspected IFD cases and (ii) all proven and probable IFD cases.

Results: 

A total of 667 febrile neutropenia episodes were identified in 265 patients. IFD was diagnosed in 62 episodes: 13 proven, 27 probable, and 22 possible. In the final multiple logistic regression models, 5 variables were independently significant for both analyses: fever days, neutropenia days, hypotension, and absolute lymphocyte count <250 at the time of diagnosis. The odds ratio and a relative weight for each factor were then calculated and summed to calculate a predictive score. A risk score of ≤4 and ≤5 (10/11 maximum) for each model signifies low risk, respectively (<1.2% incidence). Model discrimination was evaluated by the area under the receiver operator characteristics curve with an area under the curve of 0.95/0.94 for each model.

Conclusion: 

Our prediction IFD risk models perform well, are easy–to-use, and are based on readily available clinical data. Profound lymphopenia absolute lymphocyte count <250 mm³ could serve as a new important prognostic marker for the development of IFD in pediatric cancer and hematopoietic stem cell transplant patients.