Fludarabine pharmacokinetics and dynamic immune monitoring as predictive biomarkers for CD19 and BCMA CAR-T

Abstract Introduction: The optimal use of fludarabine-based lymphodepletion (LD) and the requisite degree of lymphopenia for chimeric antigen receptor (CAR) T-cell efficacy and toxicity are not well-defined across different targets. We analyzed the relationship between immune cell kinetics, fludarabine exposure, and clinical outcomes in patients receiving anti-BCMA or anti-CD19 CAR-T to provide insight into optimizing LD strategies. Methods: We retrospectively studied adults ​​​​​​treated with CD19- or BCMA-directed CAR-T between 2018–2024 at our institution. Immune cell counts (absolute lymphocyte count (ALC), CD3, CD4, CD8, NK, B) and LDH levels were collected from pre-collection through 12-months post-infusion. Fludarabine area under the curve (AUC) was ​calculated​ using an established population PK model (Langenhorst et al, Clin Pharmacokinetics 2019). Overall response rate (ORR) was defined as achieving CR, PR, or VGPR. The primary endpoint was progression-free survival (PFS). Analyses included Spearman correlation, Wilcoxon tests, optimal cut-point analysis, logistic regression and Cox regression. Multivariable models adjusted for LDH and product type. Results: The cohort included 118 CAR-T recipients. The median age was 61 years (range 24–91), 67% received CD19-directed (axi-cel 45%, brexu-cel 4%, liso-cel 15%, tisa-cel 3%) and 33% BCMA-directed products (cilta-cel 25%, ide-cel 8%). The majority had large B-cell lymphoma (59%) or myeloma (31%). Ninety-five percent received fludarabine/cyclophosphamide-based LD. The median follow-up was 26 months. The median PFS for large B-cell lymphoma and multiple myeloma were 11 months and 24.6 months, respectively. At cell collection, ALC had a median of 0.85 × 10^3/μL. ALC and subset counts were similar between CD19 and BCMA except B cell counts were significantly lower in CD19 recipients (median 0, range [0 - 9596] vs 87 [2 - 670] × 10^3/μL; p<0.0001). CD4 count > 200/μL was associated with significantly better PFS among anti-BCMA recipients (HR 0.17, p=0.01) but not in anti-CD19 recipients (HR 0.78, p=0.57). Other immune subsets did not correlate with PFS or ORR. Following CAR T infusion, we identified distinct recovery patterns between CAR-T products. As expected, anti-CD19 patients generally maintained profound B-cell aplasia from collection through six months, while anti-BCMA patients had B-cell aplasia on Day 30, but significantly higher B-cell counts at later time points (all p < 0.02). Anti-BCMA recipients had improved T-cell recovery with higher median CD3 counts at Day 30 compared to anti-CD19 recipients (590 vs. 390/μL, p = 0.046). Importantly, Day 30 B-cell recovery (> 3/uL) correlated with significantly worse PFS for either target (CD19: HR=6.1, p=0.016; BCMA: HR=18.3, p<0.01). Fludarabine AUC was heterogeneous (median 17.7, range [9.6-25.9] mg·h/L) but did not differ between anti-BCMA and anti-CD19 recipients (17.3 vs 17.9 mg·h/L, p=0.43). However, fludarabine exposure correlated with profound lymphodepletion according to Day 0 ALC (Spearman rho=-0.234, p=0.02) in all patients. Higher fludarabine exposure was associated with longer PFS for both targets. We identified optimal cutoffs for fludarabine AUC that predicted improved PFS for each target: CD19: 15.2 mg·h/L (HR 0.42, p=0.013) and BCMA: 16.5 mg·h/L (HR 0.24, p=0.073). Given that baseline CD4 lymphopenia and low fludarabine exposure were both predictive of poor outcomes, we combined them in a 2-factor predictive model. Anti-BCMA recipients meeting both baseline CD4 and fludarabine AUC cutoffs had superior PFS (HR=0.08, p=0.029). For anti-CD19 recipients, fludarabine exposure independently predicted longer PFS, but baseline CD4>200 did not improve the prediction. Conclusions: We demonstrated that higher fludarabine exposure impacts the depth of lymphodepletion and improves CAR-T efficacy independent of the target. Additionally, severe CD4 lymphopenia serves as an independent predictor of PFS in anti-BCMA CAR-T recipients. Early (Day-30) B-cell recovery is a potent biomarker of treatment failure in both BCMA and CD19 CAR-T recipients. These data suggest that PK-guided fludarabine dosing and monitoring of T- and B-cell kinetics are promising strategies to personalize and optimize CAR-T therapy.