Optimizing fludarabine exposure using pharmacokinetic modeling in pediatric patients receiving tisagenlecleucel: A step toward personalized lymphodepletion

Alex Hoover et al
University of Minnesota, Minneapolis, MN, United States

Background
  • Tisagenlecleucel (tisa-cel), a CD19-directed CAR T-cell therapy, has transformed therapy in relapsed/refractory pediatric B-cell acute lymphoblastic leukemia. Over 80% of patients will achieve a complete remission after tisa-cel, however 40–50% of responders relapse within a year. Lymphodepleting chemotherapy with fludarabine (FLU) and cyclophosphamide (Cy) enhances CAR-T expansion and persistence, and previous studies have shown the importance of achieving minimum levels of FLU exposure to improve these outcomes.
  • There is substantial interpatient variability in FLU pharmacokinetics (PK) with package insert body surface area dosing in pediatrics. A priori PK modeling allows for predicted FLU exposure estimates based on patient-specific variables.
  • This study evaluated the relationship between model-predicted FLU area-under-the-curve (AUC, mg·hr/L) and clinical outcomes in pediatric patients receiving tisa-cel to identify optimal exposure targets.
Methods
  • This retrospective single center cohort study included pediatric patients who received tisa-cel with standard lymphodepletion (FLU + Cy) between November 2017 and October 2024. FLU was administered for all patients at 30 mg/m²/day for 4 days. Individual AUCs for FLU were retrospectively calculated using a validated population PK model via Bayesian estimation (Brooks, 2022).
  • Outcomes evaluated included 1- year overall survival (OS), relapse-free survival (RFS), and relapse-free, loss of B-cell aplasia (LBCA)-free survival. The cumulative incidence of cytokine release syndrome (CRS) and immune effector cell associated neurotoxicity syndrome (ICANS) were evaluated at Day 100.
  • Disease burden immediately prior to lymphodepletion and infusion defined as High (>5%), Low (<5%), or Negative as measured by flow cytometry. Recursive partitioning (RPART) was utilized to identify optimal AUC cutoffs for RFS.
  • Survival outcomes and toxicity incidence were analyzed using Kaplan-Meier, cumulative incidence functions with competing risks, and Fine and Gray regression. Group comparisons used Wilcoxon rank-sum and chi-square tests.
Results
  • 34 patients were included with a median age of 14.4 years (range 1.2–24.4) and median weight of 50.0 kg
  • (range 9.8–97.3). RPART defined three FLU AUC groups: LOW (<18.5 mg·hr/L, n=8), OPTIMAL (18.5–21.7, n=11), and HIGH (≥21.7, n=15). There were no statistically significant differences in age, weight or disease burden across the FLU AUC groups.
  • RFS of the full study cohort at 1 year was 59% (95% CI:40–73%). RFS by AUC group was: LOW 25% (4–56%), OPTIMAL 81% (42–95%), and HIGH 60% (32–80%) (p=0.04). Low or negative disease burden was associated with higher RFS (69%, 46–84%) vs. high burden (36%, 13–63%) (p=0.01).
  • Full cohort 1-year OS was 78% (95% CI:60–89%). OS by AUC group was: LOW 57% (17–84%), OPTIMAL 100% (100–100%), and HIGH 73% (44–89%) (p=0.12). Patients with low disease burden had higher OS (90%, 66–97%) than those with high burden (55%, 23–78%) (p<0.01).
  • Relapse-free, LBCA-free survival at 1 year was 38% (95% CI:22–54%) overall, with a trend towards higher rates in patients in the OPTIMAL and HIGH cohorts: LOW 13% (1–42%), OPTIMAL 45% (17–71%), and HIGH 47% (21–69%) (p=0.20).
  • The cumulative incidence of relapse at 1 year was 39% (95% CI:22–55%). Relapse by AUC group was: LOW 75% (40–100%), OPTIMAL 19% (0–49%), and HIGH 33% (10–57%) (p=0.03). High disease burden was associated with higher relapse (55%, 25–84%) than low burden (31%, 12–50%) (p=0.04).
  • In multivariate analysis controlling for disease burden, hazard ratio for RFS in the OPTIMAL cohort was 0.16 (95% CI:0.03-0.83), HIGH 0.42 (0.13-1.34), relative to LOW at 1.00 (p=0.07). Hazard ratio for relapse in OPTIMAL was 0.15 (0.03-0.73), HIGH 0.29 (0.09-0.90), relative to LOW 1.00 (p=0.02).
  • The cumulative incidence of CRS was 62% (95% CI:44–80%). CRS did not differ by AUC group but was significantly higher in patients with high disease burden (91%, 61–100%) vs. low burden (48%, 27–68%) (p=0.01). The cumulative incidence of ICANS was 15% (3–26%), with no significant differences by AUC group or disease burden
Conclusion
  • Model-predicted FLU AUC appears to correlate with key clinical outcomes in pediatric patients receiving tisa-cel.
  • These findings support the utility of model-informed target FLU AUCs to guide personalized lymphodepletion. Personalized lymphodepletion may offer a feasible and impactful approach to enhance the long-term efficacy of CAR T-cell therapy in children.