Therapeutic Drug Monitoring–Driven Prediction of Cyclosporine-Associated Liver Injury in Pediatric Thalassemia After Hematopoietic Stem Cell Transplantation

KEY POINTS: What is already knownCyclosporine A (CsA) is widely used for immunosuppression after hematopoietic stem cell transplantation (HSCT), but its narrow therapeutic window and marked pharmacokinetic variability make pediatric patients particularly vulnerable to drug-induced liver injury (DILI).What this study addsThis study establishes a therapeutic drug monitoring (TDM)-based prediction model integrating CsA trough concentrations with key concomitant medications (azole antifungals, amphotericin B (AmB), and ursodeoxycholic acid) to estimate the individualized risk of CsA-associated liver injury in children with thalassemia after HSCT.What is the clinical implicationThis model provides a practical tool to support early risk stratification, optimize CsA dose adjustment, and guide individualized immunosuppressive and antifungal therapy, thereby improving medication safety in pediatric HSCT recipients. BACKGROUND: Cyclosporine A (CsA) is a cornerstone immunosuppressant after hematopoietic stem cell transplantation (HSCT) in children, but its narrow therapeutic window and large interindividual pharmacokinetic variability confer a substantial risk of drug-induced liver injury (DILI). This study aimed to develop a therapeutic drug monitoring-driven predictive model for CsA-associated liver injury in pediatric thalassemia patients following HSCT. METHODS: A retrospective cohort of 116 pediatric patients with thalassemia who underwent HSCT between January 2020 and June 2023 was analyzed. Clinical characteristics, laboratory indices, CsA trough concentrations, and concomitant medications were collected within the first 100 days posttransplantation. Least absolute shrinkage and selection operator regression was applied for feature selection, followed by multivariate logistic regression to identify independent predictors. A nomogram was constructed and internally validated using bootstrap resampling. Model performance was evaluated using the concordance index (C-index), receiver operating characteristic curve, calibration analysis, and decision curve analysis. The events-per-variable ratio was calculated to assess model stability. With 46 DILI events and 4 predictors in the final model, the events-per-variable exceeded 10, supporting adequate model robustness. Multicollinearity was assessed using variance inflation factors, with all values <5. RESULTS: DILI occurred in 46 of 116 patients (39.7%). Four independent predictors were identified: CsA trough concentration 150-250 ng/mL (OR 5.84, P = 0.001), CsA trough concentration >250 ng/mL (OR 30.05, P < 0.001), azole antifungal use (OR 24.27, P = 0.003), and amphotericin B exposure (OR 3.66, P = 0.042). Ursodeoxycholic acid showed a protective effect. The nomogram demonstrated excellent discrimination with a C-index of 0.858 and a bootstrap-corrected C-index of 0.823. The area under the receiver operating characteristic curve was 0.858, with good calibration and favorable net clinical benefit across wide probability thresholds in decision curve analysis. CONCLUSIONS: A therapeutic drug monitoring-driven nomogram integrating CsA exposure levels and key concomitant medications accurately predicts the risk of CsA-associated liver injury in pediatric thalassemia patients after HSCT. This model may support individualized immunosuppressive management and early prevention of hepatotoxicity in this high-risk population.