Population Pharmacokinetic Model Identifies an Optimal Fludarabine Exposure for Improved Outcomes after CD19-Directed CAR T Cell Therapy for Aggressive B-NHL: Analysis from the Cell Therapy Consortium

Background: Fludarabine (Flu) is one of the most common lymphodepleting chemotherapy (LDC) agents given before CD19 CAR T cells. Because the main pharmacokinetic (PK) predictors of Flu exposure are weight and glomerular filtration rate (GFR) and not body-surface area (BSA), standard BSA-dosed Flu leads to highly variable drug exposure among patients (pts) (Langenhorst et al., 2019). We hypothesized that variable Flu exposure influences outcomes for pts with relapsed/refractory (rel/ref) aggressive B-NHL receiving CD19 CAR T cells (Fabrizio et al.; Dekker et al., Blood Advances 2022). To test this, we estimated Flu exposure and evaluated its association with key outcomes, aiming to identify a target exposure that optimized efficacy and tolerability.

Methods: This is an IRB-approved retrospective study using data from the Cell Therapy Consortium (CTC), a multicenter working group/registry of 8 U.S. academic institutions performing real-world, observational studies on pts receiving commercial CAR T cells. This analysis included adults (≥18 y/o) with rel/ref aggressive B-NHL treated with axicabtagene ciloleucel (axi-cel) between 4/2018-6/2021. Eligible pts received LDC per the FDA label: Flu 30 mg/m² & cyclophosphamide 500 mg/m² daily x 3 days starting day -5. Exclusions included: weight unsupported by the PK model (>125 kg), >3-days between end of LDC and cell infusion, and/or LDC given on an interrupted schedule. We estimated cumulative Flu exposure (AUC; mg*h/L) for each pt using a validated population PK model and the following variables: GFR, actual body weight, height, and the daily doses (mg) of Flu given (Langenhorst et al., 2019). Main outcomes were PFS, OS, and cumulative incidences (CIs) of rel/POD, CRS, and ICANS. To define an optimal Flu exposure, univariable (UV) hazard ratio (HR) plots were drawn with P-splines, and exposure thresholds were set by assessing where the HRs (and confidence intervals) crossed to <1. The chosen Flu exposure thresholds were used as categorical variables in UV and multivariable (MV) Cox proportional models.

Results: Characteristics of the 199 included pts are shown in Table 1. Among all pts, median estimated Flu exposure was 21.1 (range 12-40, IQR 19-24). Three Flu exposure thresholds of <23 (N=134), 23-25 (N=29), and >25 (N=35) were selected for analyses. Median follow-up was 18.4 months (mos). CIs of rel/POD at 6 mos by exposure group were 53% (44-61%), 27% (12-45%), and 32% (17-48%), respectively. PFS rates at 6 mos were 43% (35-52%), 72% (58-91%), and 56% (42-76%), respectively (Figure 1). OS rates at 6 mos were 70% (62-78%), 86% (75-100%), and 68% (54-85%), respectively. In each exposure group, 64, 8, and 18 pts died, respectively. In the <23 group, causes of death (COD) were rel/POD (N=57; 89%), infection (N=3; 5%), and toxicity (N=4; 6%); in the 23-25 group COD were rel/POD (N=7; 87%) and infection (N=1; 12%); and in the >25 group COD were rel/POD (N=8; 44%), infection (N=4; 22%), and toxicity (N=5; 28%) with 3 of these deaths due to ICANS.

In UV analyses (not shown), Flu exposure of 23-25 was significantly associated with the lowest risk of rel/POD, and highest PFS and OS compared to the other groups. Flu exposure of <23 was associated with the highest risk of rel/POD and lowest PFS, while an exposure of >25 was associated with comparable PFS and OS to <23. Flu exposure was not associated with risk of CRS, but exposure >25 was associated with a higher risk of any-grade ICANS. In MV analyses, Flu exposure of 23-25 was associated with the highest PFS (HR 0.48; 0.26-0.91, p=0.02) and lowest risk of rel/POD (HR 0.5; 0.28-0.96, p=0.04) without an increased risk of any-grade ICANS (HR 1.5; 0.86-2.5, p=0.16). Higher LDH and bulky disease were associated with poorer PFS and higher rates of rel/POD. Flu exposure of >25 was associated with the highest risk of any-grade ICANS (HR 1.6; 1.1-2.5, p=0.03).

Conclusions: These CTC registry findings suggest that there is a Flu exposure target window associated with improved survival in pts with rel/ref aggressive B-NHL receiving axi-cel. Flu underexposure is associated with a higher risk of disease-related treatment failure while overexposure is associated with a higher risk of toxicities. This optimal Flu exposure is achievable through personalized, PK-directed dosing and is a novel, easily modifiable strategy to improve outcomes after CAR T cell therapy. These data require further external validation which is currently on-going.