A predictive model for cytarabine syndrome-associated febrile episodes in children with acute lymphoblastic leukemia.

Abstract

Objective: This study aims to develop a predictive model for cytarabine syndrome (CS)-related fever in pediatric patients with acute lymphoblastic leukemia (ALL). By integrating clinical features and laboratory indicators, we seek to establish an evidence-based diagnostic tool to assist clinicians in decision-making and optimize the use of antimicrobial agents.

Methods: Study Design: A retrospective case-control study was conducted involving 72 ALL children (32 in the CS group vs. 40 in the infection group) who developed fever (≥37.5 °C) following cytarabine treatment. Model Development: Multivariate logistic regression was used to identify independent predictors. A nomogram was constructed. The model's discriminative ability was evaluated using the receiver operating characteristic (ROC) curve, calibration was assessed via the Hosmer-Lemeshow test, and clinical utility was validated using decision curve analysis (DCA).

Results: Key Predictors: Multivariate analysis revealed that BSA (OR = 0.01, 95% CI: 0.00-0.17) and hs-CRP (OR = 0.84, 95% CI: 0.72-0.97) were the most significant predictors of CS. Model Performance: The ROC analysis demonstrated an AUC of 0.785, with internal validation via 1000 bootstrap resamples yielding a 95% confidence interval of 0.675-0.882. The calibration curve demonstrated excellent agreement between predicted and observed outcomes (MAE = 0.039). Decision curve analysis indicated that the model provided superior net benefits across most threshold probabilities (0.2-0.8) compared to strategies of universal antibiotic use or no intervention.

Conclusion: The developed predictive model demonstrates high discriminative performance and robust clinical applicability, functioning as an effective quantitative instrument for early identification of cytarabine syndrome (CS) and facilitating evidence-based antibiotic stewardship.