Skip to content
Frontiers
Search

Original Research

Frontiers in Pharmacology ·

Volume 17 - 2026 | https://doi.org/10.3389/fphar.2026.1878984

Population cellular kinetics and exposure-CSR analysis of CAR-T cells in Chinese patients with relapsed/refractory B-ALL

  • Yicong Bian1,2,3
  • Jinjie Yuan1,2,3
  • Chenrong Huang1,2,3
  • Kaijie Zhang1,3
  • Ling Xue1,3
  • Fan Xia1,3,4
  • Liqing Kang5
  • Yan Wang1,3
  • Hua Zhang1,2,3
  • Lei Yu5*corresponding author
  • Depei Wu4*corresponding author
  • Liyan Miao1,2,3*corresponding author
  1. 1Department of Pharmacy, The First Affiliated Hospital of Soochow University, Suzhou, China
  2. 2National Institution of Drug Clinical Trial, The First Affiliated Hospital of Soochow University, Suzhou, China
  3. 3Institute for Interdisciplinary Drug Research and Translational Sciences, Soochow University, Suzhou, China
  4. 4National Clinical Research Center for Hematologic Diseases, Jiangsu Institute of Hematology, The First Affiliated Hospital of Soochow University, Suzhou, China
  5. 5Shanghai Unicar-Therapy Bio-Medicine Technology Co., Ltd, Shanghai, China
Article metrics0 viewsView details
PDFCiteMetrics

Abstract

CD19-targeting chimeric antigen receptor (CAR)-T cells have demonstrated great efficacy in treating relapsed and refractory B-cell acute lymphoblastic leukemia (B-ALL). Despite the high objective response rate of CAR-T cell therapy, high-risk adverse event of cytokine-release syndrome occur unexpectedly. As living gene-modified cells, CAR-T cells exhibit unique kinetics. This study aimed to explore the relationship between CAR-T therapy-related CRS and the kinetics of CAR-T cells. In this study, data for 2184 CAR-transgene copies from 131 B-ALL patients treated with IL-6 knockdown CD19-targeting CAR-T cells were included to characterize the cellular kinetics using a nonlinear mixed-effects (NLME) model. The extrinsic and intrinsic factors that may impact the kinetics of CAR-T cells were also evaluated. The CD4 + /CD8 + ratio of CAR-T cell products was shown to be significantly positively correlated with T max , suggesting that a higher CD4 + /CD8 + ratio may lead to a delay in reaching the maximum expansion, while other candidate covariates including tumor burden, CAR-T doses and comedication were not significantly associated with the kinetics of CAR-T cells. The estimated values of the cellular kinetic parameters C max , T max and Fold x were 4.696 × 10 4 copies/μg, 13.0 days, and 127.5, respectively. The estimated AUC 0-28d and AUC 0-90d accounted for only 13.0% and 18.5% of AUC 0-inf , respectively, indicating that there was long-term persistence of memory CAR-T cells in vivo . In addition, the incidence of CRS was investigated to be correlated with the early kinetics of CAR-T cells, and the optimal cutoff values of C max and AUC 0-28d for severe CRS were 6.951 × 10 4 copies/μg and 4.962 × 10 5 copies/μg × days, respectively.

Keywords

1 Introduction

Chimeric antigen receptor (CAR)-modified T Cells targeting antigen CD19 are a highly effective therapeutic approach for relapsed and refractory (r/r) B Cell acute lymphoblastic leukemia (B-ALL) (Baguet et al., 2024; Brudno et al., 2024; Shah et al., 2021; Tan et al., 2025; Zhang et al., 2023). Although the short-term objective response rate (ORR) of CAR-T cell therapy is high, the risk of post-treatment cytokine-release syndrome (CRS) remains high and unexpected. Previous studies have speculated that CRS associated with CAR-T cell therapy may be associated with CAR-T cell kinetics in patients (Baguet et al., 2024; Mueller et al., 2017; Dasyam et al., 2020), but supporting evidence remains limited. Moreover, CAR-T cell proliferation exhibits greatly interindividual variability, and how to assess CRS following CAR-T cell therapy using individual cellular kinetic parameters remains unclear.

CAR-T cells are gene-modified cells that exhibit unique cellular pharmacokinetic (PK) profiles in vivo (Mueller et al., 2017; Dasyam et al., 2020; Hu et al., 2025; Pan et al., 2025; Tegenge et al., 2025). Briefly, CAR-T cells are activated by specific antigens on tumor cells and undergo expansion in response to antigenic stimulation in patients (Pfeifer et al., 2022; Qi et al., 2022). After a period of expansion, CAR-T cells undergo depletion and numerical reduction, while a small subset of CAR-T cells may persist long-term in vivo and retain antitumor potential (Qi et al., 2022; Stein et al., 2019). As a personalized medicine, CAR-T cell proliferation is highly variable among individuals and may be influenced by multiple potential factors, including patient status, clinical treatment regimens and CAR-T cell product characteristics (Dasyam et al., 2020; Qi et al., 2022; Stein et al., 2019; Singh et al., 2021). However, how these factors affect CAR-T cellular kinetics as well as the quantitative relationship between the kinetic parameters and CRS development remain unclear.

Population pharmacokinetics (PopPK) models facilitate the characterization of drug behavior in patients and the investigation of potential factors influencing kinetics using large datasets from sparsely collected patient blood samples (Sheiner and Ludden, 1992; Dai et al., 2022; Han et al., 2024; Li et al., 2024). Given that CAR-T cell kinetics differ from those of traditional drugs, we employed a cellular kinetic model rather than a traditional compartmental model to quantify the behavior of CAR-T cells in vivo (Stein et al., 2019). A total of 2184 CAR-transgene copy measurements from 131 Chinese B-ALL patients treated with an IL-6 knockdown CD19-targeting CAR-T cell product were analyzed to characterize cellular kinetics using a nonlinear mixed-effects (NLME) model. The factors associated with CAR-T cell kinetics, as well as the relationship between CAR-T cell exposure and the occurrence of CRS were investigated in this study.

2 Methods

2.1 Study design and patients

Data were obtained from a single-center, one-arm, open-label study at our hospital (NCT03919240). The inclusion criteria included patients diagnosed with CD19-positive r/r B-ALL; adequate cardiac, hepatic and renal function; Karnofsky performance status greater than sixty or Eastern Cooperative Oncology Group (ECOG) performance status from zero to two; life expectancy greater than 3 months; negative pregnancy test and nonlactating status (women only). All enrolled patients received treatment with IL-6 knockdown CD19-targeting CAR-T cells, and were followed up for adverse events (AEs). Patients with sparsely collected peripheral blood (PB) samples available within 90 days post-CAR-T infusion for CAR transgene copy number quantification were included in the present retrospective analysis. This retrospective analysis was approved by the ethics committee of our hospital. All procedures were performed in accordance with the Helsinki Declaration of 1975, as revised in 2024. All participants provided written informed consent.

2.2 CAR-T cell products and treatment

CAR-T cells were provided by Shanghai Unicar-Therapy Bio-Medicine Technology Co., Ltd. The product were CAR-expressing T Cells consisting of an anti-CD19 single-chain antibody fragment and an IL-6-specific shRNA, with the manufacturing process conducted as previously described (Li et al., 2022; Kang et al., 2020). The quality of CAR-T cells was tested throughout the manufacturing process.

Patients enrolled in the study received an infusion of autologous IL-6 knockdown CD19-targeting CAR-T cells at a target dose of 5∼10 × 106 cells/kg body weight after a preparative regimen consisting of fludarabine and cyclophosphamide (FC). The total CAR-T cell dose was administered in three sequential fractions over three consecutive days: 10% on the first day, 30% on the second day, and 60% on the third day.

2.3 Quantification of CAR-T cells in PB

CAR-transgene copy numbers in PB were quantified at different time points using a validated real-time quantitative polymerase chain reaction (qPCR) assay. qPCR was performed using the ABI 7500 Real-Time PCR System (Life Technologies, United States), and transgene copy numbers were calculated from a standard curve established by plasmids encoding the transgenes. The qPCR protocol was performed as described previously (Davis et al., 2021). The lower limit of quantitation (LLOQ) of the assay was 10 copies/μg DNA.

2.4 Kinetics modeling of CAR-T cells

2.4.1 Base model

The base structural model was adapted from a previously published cellular kinetics model (Stein et al., 2019). All kinetics model analyses were performed using Phoenix NLME software version 8.2 (Pharsight, Certara Inc.). The typical CAR-T cellular kinetics were simplified into three processes: exponential expansion, rapid initial decline, and slow long-term decline. A schematic representation of the cellular kinetics of CAR-T cells in vivo is shown in Figure 1. Six key cellular kinetic parameters, namely, Cmax, Tmax, α, β, Foldx, and FB , were estimated with this model. Specifically, the structural model defined ρ as the exponential expansion rate constant, which applies until the time point (Tma*x*) when the transgene copy number reaches its maximum value (Cma*x*). After Tma*x*, a biexponential decline was modeled, characterized by two rate constants: α (contraction) and β (persistence). To quantify the total transgene copy number over time following CAR-T cell infusion, three parameters (R *0* , A, B) were incorporated into Equation 1. R*0* represents the baseline transgene copy number level; A and B represent the portion of Cma*x* that decline rapidly (at slope α) and gradually (at slope β), respectively, at time Tma*x*. Mathematical expressions for these parameters are provided in Equations 2–4. The fold expansion of CAR-T cells relative to baseline (Fold*x*) is described in Equation 5. The fraction of transgene copies that declined at the rate of β at Tma*x* (F*B*) and the fraction of transgene copies declining at the rate of α at Tma*x* (1 − F*B*) are described in Equation 6.

F t = R 0 · e ρ · t , t < T max A · e − α t − T max + B · e − β · t − T max , t ≥ T max (1)

R 0 = C max / e ρ · T max (2)

A = C max · 1 − F B (3)

B = C max · F B (4)

Fold x = e ρ · T max (5)

F B = B C max (6)

E o b s = E · e E E p s (7)

A schematic representation of the cellular kinetics of CAR-T cells *in vivo*.
Figure 1. A schematic representation of the cellular kinetics of CAR-T cells in vivo.

Typical cellular kinetic parameters and their random interindividual variability (IIV) were estimated using a first-order conditional estimation method with the extended least squares method (FOCE ELS). Random effects were estimated via an exponential model. For the residual error model (Equation 7), natural logarithmic-transformed data were modeled with an additive error model in consideration of the magnitude difference in transgene copies.

2.4.2 Covariate screening

To evaluate the distribution of covariates in the population, exploratory covariate analysis was conducted based on clinical plausibility and scatter plots of covariates versus random-effects estimates (Ette and Ludden; Wu et al., 2019). Covariates were investigated using a stepwise procedure in the model, including: demographic characteristics (sex, age, height, weight, BMI); pre-infusion tumor burden (primitive naive cells); CAR-T doses and comedication (corticosteroids or tocilizumab); in vitro CAR-T cell proliferative properties (culture time, proliferation fold change, cell cycle), and transduction efficiency and CD4+/CD8+ T Cell ratio of CAR-T cell products. Categorical covariates were coded as numbers, and continuous covariates were centered at their median value (Ma et al., 2020). During forward selection, a decrease in the objective function value (OFV) exceeding 3.84 (p < 0.05) served as the covariate inclusion criterion. For backward elimination, an OFV increase greater than 6.635 (p < 0.01) was defined as covariate retention criterion.

2.4.3 Model evaluation

Model evaluation was based on the Goodness-of-fit (GOF) plots of the final model using the Phoenix NLME, including conditional weighted residuals (CWRES) over population predicted concentration (PRED) or time after dosing (TAD) and the relationship between observed concentration (OBS) and PRED or individual predicted concentration (IPRED) (Dai et al., 2022; Wu et al., 2019; Hooker et al., 2007; Karlsson and Savic, 2007). Visual predictive check (VPC) and bootstrap analysis (1000 runs) were also performed. Normalized prediction distribution error (NPDE) plots were generated using R software (version 4.6.1).

2.5 Safety assessment

Vital signs and clinical laboratory tests were included in the safety assessment. Special attention was paid to the incidence of CRS in this study. The AEs were graded according to the National Cancer Institute CTCAE version 4.03, with grade 3 or higher considered serious.

2.6 Statistical analysis

Patient characteristics are presented using descriptive statistics, and nonparametric tests and Fisher’s exact tests were used for comparisons among groups. The PK parameters Cmax, Tmax, α, β, Foldx and FB were estimated in the cellular kinetic model described above, while the area under the concentration-time curve (AUC) from time zero to day 28 or day 90 post-dose (AUC0-28day, AUC0-90day) and the AUC form time zero to infinity (AUC *0-* ∞) were calculated from the above parameters (Stein et al., 2019).

The correlation between the CRS of CAR-T cell therapy and cellular kinetics was assessed using the nonparametric Mann-Whitney test. P < 0.05 were considered statistically significant. The optimal cutoff points of kinetics parameters for predicting the toxicity of CAR-T cell therapy were determined using receiver operating characteristic (ROC) curve analyses, and 95% confidence intervals (CIs) were calculated for all ROC AUCs. Data were analyzed using GraphPad Prism 9.3 software.

3 Results

3.1 Patients and CAR-T cell products

A total of 131 B-ALL patients who received IL-6 knockdown CD19-targeting CAR-T cell therapy and had quantifiable CAR-transgene copies in PB within 90 days after CAR-T cell treatment were included in the study. All patients received a dose of 5 × 106∼10 × 106 cells/kg CAR-T cells. The characteristics of the patients and CAR-T cell products are shown in Table 1.

TABLE 1. Characteristics of 131 Chinese r/r B-ALL patients and CAR-T cell products.

CharacteristicsN (%)/Median (IQR, min ∼ max)
Patients
Male (N)72 (54.96)
Age (years)36 (24.44, 11∼68)
Height (cm)167 (10.5, 150∼190)
Weight (kg)61 (17.85, 22∼110)
BMI (kg/m2)21.72 (5.22, 8.59∼34.44)
Tumor burden (%)1.00 (3.50, 0.00∼90.50)
Lymphodepleting chemotherapy with FC (%)131 (100%)
Weight-adjusted dose of CAR-T cells (cells/kg)5.00 × 106 (0, 5.00 × 106∼10.00 × 106)
Total dose of CAR-T cells (cells)3.25 × 108 (1.34 × 108, 1.10 × 108∼9.60 × 108)
Received steroids (N)10 (7.63)
Received tocilizumab (N)2 (1.53)
CAR-T cell products
Transduction efficiency (%)44.60 (22.99, 5.80∼77.12)
CD4+/CD8+ T Cell ratio1.95 (3.91, 0.05∼25.17)
Culture time in vitro (days)11 (4, 7∼20)
Proliferation fold change in vitro40.19 (47.87, 6.28∼357.50)
Cell cycle in vitro (days)2.10 (0.76, 1.27∼5.05)

3.2 CAR-T cellular kinetics

A total of 2184 observed CAR-transgene copies from 131 patients were sparsely collected within 90 days after CAR-T cells infusion. Each patient had an average of 16–17 observations, with 88.8% of the total observations obtained from the first day to the 28th day after CAR-T cells infusion.

In the preliminary estimates of the base model, the value of β was fixed at 0.001 due to high variable coefficient (43.96%). The shrinkage values of Cmax, Tmax and Foldx were 0.147, 0.240 and 0.175, respectively, indicating that the dataset used for model development was enough to compute theses parameters. In addition, the shrinkage values of α and FB were larger than 0.5, and that of β was close to 1. Therefore, random effects of α, β and FB were not incorporated into the base model. Through rigorous screening, the CD4+/CD8+ T Cell ratio in the CAR-T cell products was shown to be positively correlated to the parameter of Tmax, while other candidate covariates, including tumor burden, CAR-T doses and comedication were not significantly associated with the kinetic parameters (Figure 2; Supplementary Figure S1). Therefore, the CD4+/CD8+ T Cell ratio was included in the final model (Δ-2LL = 9, p < 0.01). The predictive performance of the model in terms of the GOF and VPC is shown in Figure 3, which demonstrated the exactitude of the parameter estimates and the reliability of the final model. The mean (−0.014), variance (1.04), skewness (0.088), and kurtosis (0.21) of the NPDEs collectively suggested their distribution closely matched the standard normal distribution. Accordingly, the final model exhibited no systematic bias, with no overall overprediction or underprediction of observed cellular concentrations. The final model also characterized the data well in individuals, and the typical fitting plots are shown in Figure 4.

Exploratory analysis of continuous covariates and categorical covariates plotted against *post hoc* random-effects estimates.
Figure 2. Exploratory analysis of continuous covariates and categorical covariates plotted against post hoc random-effects estimates.
Goodness-of-fit plots (GOF) and visual prediction check (VPC) of the final model. **(a)** OBS vs. PRED and **(b)** OBS vs. IPRED. OBS, observed values. PRED, population predicted values. IPRED, individual predicted values. The points were uniformly distributed around the line y = x. **(c)** CWRES vs. TAD and **(d)** CWRES vs. PRED. CWRES, conditional weighted residuals. TAD, time after dose. The weighted residuals were mainly concentrated on both sides of y = 0, within the range of ±4. **(e)** Histogram of NPDE and **(f)** NPDE vs. PRED. NPDE, normalized prediction distribution error. The distribution of the NPDEs approximated a normal distribution, and the NPDEs were randomly and symmetrically scattered around the line y = 0 across the entire predicted concentration range, showing no discernible trend or systematic bias **(g,h)** visual prediction check within 90 days and 28 days of the final model. The blue dots denoted the observed values. The red line represented the 5th, 50th and 95th percentiles of the observed concentrations. The black lines represented 5th, 50th and 95th percentiles of simulated concentrations. The shaded areas represented 95% confidence interval of the 5th, 50th and 95th prediction intervals. The 5th, 50th and 95th percentiles and these 95% CI of the observations remain within the corresponding percentile range and predicted intervals of the predicted data from the final model.
Figure 3. Goodness-of-fit plots (GOF) and visual prediction check (VPC) of the final model. (a) OBS vs. PRED and (b) OBS vs. IPRED. OBS, observed values. PRED, population predicted values. IPRED, individual predicted values. The points were uniformly distributed around the line y = x. (c) CWRES vs. TAD and (d) CWRES vs. PRED. CWRES, conditional weighted residuals. TAD, time after dose. The weighted residuals were mainly concentrated on both sides of y = 0, within the range of ±4. (e) Histogram of NPDE and (f) NPDE vs. PRED. NPDE, normalized prediction distribution error. The distribution of the NPDEs approximated a normal distribution, and the NPDEs were randomly and symmetrically scattered around the line y = 0 across the entire predicted concentration range, showing no discernible trend or systematic bias (g,h) visual prediction check within 90 days and 28 days of the final model. The blue dots denoted the observed values. The red line represented the 5th, 50th and 95th percentiles of the observed concentrations. The black lines represented 5th, 50th and 95th percentiles of simulated concentrations. The shaded areas represented 95% confidence interval of the 5th, 50th and 95th prediction intervals. The 5th, 50th and 95th percentiles and these 95% CI of the observations remain within the corresponding percentile range and predicted intervals of the predicted data from the final model.
Individual fitting plots for OBS/IPRED vs. TAD. **(a–t)** represent individual subjects. OBS, observed values, IPRED, individual predicted values, TAD, time after dose. The red dots denoted the observed values, and the green line denotes the individual predicted values.
Figure 4. Individual fitting plots for OBS/IPRED vs. TAD. (a–t) represent individual subjects. OBS, observed values, IPRED, individual predicted values, TAD, time after dose. The red dots denoted the observed values, and the green line denotes the individual predicted values.

Population cellular kinetic parameters and estimation precision of both the base model and final model are presented in Table 2. The estimated cellular kinetic parameters of Cmax, Tmax and Foldx in the final model were 4.696 × 104 copies/μg, 13.0 days and 127.5, respectively. The expansion rate constant ρ and the half-life of the rapid contraction rate (ln2/α) were calculated as 0.374/day and 3.077 days, while the half-life of the gradual decrease rate (ln2/β) could not be estimated due to the fixed value of β. As shown in Table 2. The parameters of AUC *0-* ∞ of CAR-T cells was estimated as 4.610 × 106 copies/μg × days on average, with a range from 2.805 ×105 to 3.924 ×107 copies/μg × days, and the estimated AUC0-28day and AUC0-90day accounted for only 13.0% and 18.5% of the AUC *0-* ∞, respectively.

TABLE 2. Population cellular kinetic parameter estimates and bootstrap results.

​Base modelFinal model1000 bootstrap replicates
​EstimateCV%2.5% CI97.5% CIEstimateCV%2.5% CI97.5% CIMedianCV%2.5% CI97.5% CI
Kinetic parameters
Cmax (copies/μg)4.519 × 10413.0083.366 × 1045.672 × 1044.696 × 10414.4413.366 × 1046.025 × 1044.571 × 10413.9933.474 × 1046.148 × 104
Tmax (days)12.4002.29311.84212.95712.9502.76712.24713.65212.5153.22711.70513.337
Foldx133.11717.77486.715179.519127.48019.87877.784177.176123.51724.48971.596197.024
Alpha (1/day)0.2075.1650.1860.2280.2258.0870.1900.2610.21313.4290.1590.272
Beta (1/day)0.001///0.001/0.0010.0010.001/0.0010.001
FB0.0517.7100.0430.0580.0508.1440.0420.0580.04927.6060.0270.080
Covariate effect on Tmax
CD4+/CD8+ T Cell ratio////0.09223.5460.0490.1340.09023.1610.0370.110
Inter-individual variability
ωC *max*1.17220.8660.6931.6511.15220.5640.6881.6171.32518.0320.7441.558
ωT *max*0.2613.8730.2410.2810.2927.0060.2520.3320.0714.9120.2420.293
ωFoldx1.84336.6750.5183.1671.80836.4840.5153.1013.32537.2650.4923.155
Random residual variability
Epsilon (ε)1.2981.0981.2701.3261.2951.1231.2661.3231.3002.9251.2231.376

3.3 Correlation between CRS and CAR-T cell exposure

Among the 131 patients, 72 patients (55.0%) developed CRS. Mild to moderate CRS (grade 1-2) occurred in 63 patients (48.1%), and severe CRS (grade 3-4) occurred in nine patients (6.9%). Baseline characteristics were comparable between patients with and without severe CRS. The severe CRS group consisted of four males and five females (mean age, 41 years; mean body weight, 60 kg), while patients without severe CRS included 68 males and 54 females (mean age, 37 years; mean body weight, 62 kg). The cellular kinetic parameters were analyzed between patients with and without severe CRS, and the results showed that patients who developed severe CRS had significantly higher Cmax, AUC0-28day and AUC *0-* ∞ after CAR-T cell therapy (P = 0.0004, 0.0007 and 0.0004, respectively), as shown in Figure 5. These results indicated that high exposure to CAR-T cells in vivo was associated with a higher risk of CRS incidence.

Correlation of CAR-T cellular kinetics with the incidence of CRS **(a)** C<sub>max</sub>, AUC<sub>0-28day</sub> and *AUC* <sub>*0-*</sub> ∞ of CAR-T cells in patients with moderate (n = 63) and severe (n = 9) CRS; **(b)** ROC curve analysis of C<sub>max</sub> vs. CRS, AUC<sub>0-28day</sub> vs. CRS and *AUC* <sub>*0-*</sub> ∞ vs. CRS.
Figure 5. Correlation of CAR-T cellular kinetics with the incidence of CRS (a) Cmax, AUC0-28day and AUC *0-* ∞ of CAR-T cells in patients with moderate (n = 63) and severe (n = 9) CRS; (b) ROC curve analysis of Cmax vs. CRS, AUC0-28day vs. CRS and AUC *0-* ∞ vs. CRS.

ROC curves were constructed to determine the optimal cutoff points of Cmax, AUC0-28day and AUC *0-* ∞ for severe CRS in CAR-T cell therapy. As presented in Figure 5, the areas under the ROC curves for Cmax, AUC0-28day and AUC *0-* ∞ of CAR-T cells were 0.8361 (95% CI 0.6947–0.9775, p = 0.0008), 0.8233 (95% CI 0.6730–0.9736, p = 0.0012) and 0.8342 (95% CI 0.6897–0.9788, p = 0.0008), respectively, and the cutoff values were 6.951 × 104 copies/μg, 4.962 × 105 copies/μg × days and 3.905 × 106 copies/μg × days. The specificities of the cutoff values for Cmax, AUC0-28day and AUC *0-* ∞ were all 0.8889, and the corresponding sensitivities for Cmax, AUC0-28day and AUC *0-* ∞ were 0.7049, 0.6639 and 0.6967, respectively. These results suggested that exposure to CAR-T cells in vivo could be used to effectively evaluate the occurrence of CRS.

4 Discussion

In this study, we developed a population cellular kinetic model to characterize the in vivo kinetics of an IL-6 knockdown CD19-targeted CAR-T cell product in Chinese patients with B-ALL. Notably, we report the first systematic population cellular kinetic characterization of this unique CAR-T construct administered via a three-day split-infusion regimen in a Chinese patient cohort. As reflected in the model, the CAR-T cellular kinetics process was divided into three stages, including exponential expansion, rapid contraction and long-term persistence, which fitted perfectly with the dataset in this study and was consistent with previous study (Stein et al., 2019). Although CAR-T cells have been reported to exhibit a transient decline due to tissue distribution after the first infusion (Hardiansy et al., 2019), the early distribution phase has only been observed in a limited number of studies (Liu et al., 2021). In the current work, no obvious early distribution phase of CAR-T cells was detected, which may be partly due to the continuous infusion of CAR-T cells masking the early distribution. Thus, we selected the cellular kinetic model without the distribution phase.

Samples collected over 90 days captured the phases of exponential expansion, rapid contraction and the early stage of long-term decline. Most sampling timepoints fell within 30 days post-infusion, providing sufficient data to characterize CAR-T expansion and subsequent decline kinetics. Key parameters assessed included Cmax, Tmax (with covariate analysis) and Foldx. The shrinkage values of α, β, and FB were high and similar to those reported by Stein et al. (Stein et al., 2019). High shrinkage indicates that individual empirical Bayes estimates regress strongly toward population typical values. Thus, reliable quantification of inter-individual variability for these parameters were challenging, and investigating covariate effects on these parameters was not recommended. In the base model, β was initially estimated at 0.001 (CV 43.96%), but its inter-individual variance could not be stably estimated across multiple initial value trials (shrinkage >0.5). Given that only approximately 10% of samples were obtained after day 30 with sparse long-term elimination phase data, β was fixed at 0.001 to guarantee model convergence and avoid over-parameterization. This value also matches the magnitude reported by Stein et al. (Stein et al., 2019).

Key parameters of Cmax, Tmax and AUC0-28day are commonly employed to characterize the early PK profiles of CAR-T cells. Our results demonstrate that CAR-T cells expand in patients and reach a peak at approximately 13 days, with the peak CAR-T cell count reaching 127.5-fold of the baseline. The estimated parameters of Cmax and Tmax were comparable to those in prior studies, while the value of Foldx was lower (Stein et al., 2019; Ogasawara et al., 2021). The difference in Foldx may be attributed to the implementation of a three-day continuous CAR-T cell infusion protocol in this study. Alternatively, it may also stem from differences in CAR-T cell products and patient ethnicity between the two studies. The estimated AUC0-28day and AUC0-90day accounted for only 13.0% and 18.5% of AUC *0-* ∞, respectively, indicating that there was long-term persistence of memory CAR-T cells in vivo, which was consistent with previous studies (Mueller et al., 2017).

A range of factors may affect CAR-T cellular kinetics, including but not limited to CAR design, patient disease type, tumor burden, lymphodepleting chemotherapy and concomitant medication (Dasyam et al., 2020; Qi et al., 2022; Stein et al., 2019; Singh et al., 2021). In this study, all patients enrolled were diagnosed with B-ALL and underwent FC lymphodepleting conditioning, followed by infusion of CAR-T cells manufactured via a standardized process and supplied by the same company, which means that CAR design, disease type and conditioning regimens of patients were relatively unitary in the study. Following rigorous covariate screening, CD4+/CD8+ T Cell ratio of the administered CAR-T cell products was found to exert a significant effect on Tmax, while patient characteristics, clinical treatment with steroids or tocilizumab, and other characteristics of CAR-T cell products were not significantly associated with cellular kinetics.

Unlike traditional drugs, the dose-exposure relationship of CAR-T cells was not observed in our study, as evidenced by the lack of correlation between dose and Cmax. The weak dose-exposure relationship might be explained by the proliferative capacity of CAR-T cells, and is consistent with findings from previous studies (Qi et al., 2022; Sheih et al., 2020). Furthermore, no significant correlation between tumor burden and cell expansion was observed, and cotreatment with tocilizumab or corticosteroids had no impact on CAR-T cellular kinetics, which was consistent with some published literature (Stein et al., 2019; Li et al., 2022; Liu et al., 2021; Porter et al., 2015; C et al., 2019), but not with others (Hardiansy et al., 2019; Mu et al., 2022). The impacts of CAR-T cell product characteristics on cellular kinetics were also investigated. Our results showed that a higher CD4+/CD8+ T Cell ratio of the CAR-T cell product may delay the achievement of maximum CAR-T cell expansion. Additionally, the relationship between the CD4+/CD8+ T Cell ratios of CAR-T cell products and Cmax was presented as a polyline with a zenith, suggesting that a balanced proportion of CD4+ and CD8+ CAR-T cells could promote in vivo proliferation. The mechanism underlying the observation remains unclear. We propose that a favorable CD4+/CD8+ T Cell ratio coordinates the function of helper and cytotoxic CAR-T subsets. CD4+ CAR-T cells constantly produce supportive cytokines that sustain persistent CAR-T cell proliferation and attenuate early cellular contraction, which may ultimately delay the Tmax of CAR-T cells in peripheral blood (Liu et al., 2026; Lee et al., 2023; Boulch et al., 2023; Bove et al., 2024). It should be noted that limited statistical power may hinder the identification of covariates with modest yet clinically meaningful effects. Larger cohorts are required to validate these potential associations in future investigations.

Furthermore, we exploratively identified Cmax and AUC0-28day cutoff values associated with the occurrence of severe CRS, providing quantitative evidence to guide clinical safety management of this novel CAR-T cell product. Generally, CRS and immune effector cell-associated neurotoxicity syndrome (ICANS) are recognized as the most important CAR-T cell related toxicities. While only two patients developed mild ICANS (grade 1) in our study, CRS was analyzed as the only toxicity index in our study. Cmax and AUC0-28day are the most commonly used pharmacokinetic parameters to characterize the early exposure of CAR-T cells. Our results suggested that high exposure of CAR-T cells in vivo was associated with a high risk of severe CRS and that patients with Cmax and AUC0-28day above the cutoff values of 6.951 × 104 copies/μg and 4.962 × 105 copies/μg × days tended to develop severe CRS due to CAR-T cell therapy. These two kinetic parameters may function as risk-stratification biomarkers in clinical practice. Patients with values above the thresholds may face a markedly higher risk of severe CRS, requiring intensified clinical monitoring and prompt implementation of standardized early intervention protocols. However, it should be emphasized that the present study only identifies an association between exposure and toxicity and cannot establish causality. Given that CRS is driven by robust T Cell activation and the systemic release of IL-6-dominated cytokines (Rial Saborido et al., 2022; Jiang et al., 2021; Jia et al., 2021), the specificity of these cutoff values may be validated in other CAR-T products, as the CAR-T cells in this study featured targeted IL-6 knockdown.

We further assessed correlations between CAR-T cell therapy response and early cellular expansion. The responders trended toward higher Cmax and AUC0-28day than nonresponders, yet no clear link emerged between 12-month clinical response and CAR-T cellular kinetics (data not shown). The pronounced interindividual variability in kinetic parameters precludes definitive conclusions on the kinetics-efficacy relationship, and longer-term follow-up is ongoing to elucidate this association. Given the high interindividual variability in CAR-T cellular kinetics, it remains challenging to draw definitive conclusions regarding the relationship between CAR-T cellular kinetics and clinical efficacy.

The main limitations are outlined below. First, this single-center retrospective cohort exclusively enrolled Chinese patients, and its generalizability is restricted by the homogeneous ethnic population and small sample size. Additionally, several potential confounders (e.g., molecular disease characteristics, baseline inflammatory biomarkers, cytokine profiles) were excluded from the analyses, as complete datasets for these indicators could not be fully retrieved under this retrospective design. Accordingly, prospective multicenter studies with larger cohorts and standardized clinical data collection protocols are needed to refine our kinetic model and conduct external validation in subsequent investigations. Second, given the use of an IL-6 knockdown CAR-T cell product and the limited number of severe CRS cases in this study, further validation is required to determine whether the exposure-toxicity cutoffs identified herein are applicable to other CD19-targeted CAR-T cell products.

5 Conclusion

In conclusion, we characterized the population cellular kinetics of CAR-T cells in 131 patients with r/r B-ALL using a population cellular kinetics model. The typical kinetics of rapid expansion and long-term persistence of CAR-T cells were elucidated. Additionally, CRS incidence during CAR-T cell therapy was found to be correlated with the exposure of CAR-T cells, and the optimal cutoff values for the early cellular kinetic parameters Cmax and AUC0-28day were exploratively identified to evaluate the occurrence of severe CRS.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fphar.2026.1878984/full#supplementary-material

Exploratory analysis of continuous covariates and categorical covariates plotted against post hoc random-effects estimates (continuation of Figure 1). Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Yu-Wei Lin, Monash University, Australia

Reviewed by: Mohamed A. Yassin, Qatar University, Qatar

Xiaoqi Wang, Xinqiao Hospital, China

Statements

Conflict of interest

Authors LK and LY were employed by Shanghai Unicar-Therapy Bio-Medicine Technology Co.

The remaining author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This work was funded by National Key New Drug Creation Special Programs (2017ZX09304-021), Key R&D Program of Jiangsu Province (BE2021644), Science and Technology Plan of Suzhou (SKJYD2021159), Hospital Pharmacy Committee of Chinese Pharmaceutical Association (CPA-Z05-ZC-2021-002) and Shanghai Unicar-Therapy Bio-Medicine Technology Co., Ltd.

Data availability

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics approval

The studies involving humans were approved by the ethics committee of First Affiliated Hospital of Soochow University. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Author contributions

YB: Formal Analysis, Funding acquisition, Methodology, Validation, Visualization, Writing – original draft. JY: Data curation, Formal Analysis, Methodology, Software, Validation, Visualization, Writing – original draft. CH: Formal Analysis, Supervision, Writing – original draft. KZ: Data curation, Writing – original draft. LX: Formal Analysis, Software, Writing – original draft. FX: Project administration, Writing – original draft. LK: Conceptualization, Resources, Supervision, Validation, Writing – original draft. YW: Data curation, Project administration, Writing – original draft. HZ: Data curation, Project administration, Supervision, Writing – review and editing. LY: Conceptualization, Resources, Supervision, Validation, Writing – review and editing. DW: Conceptualization, Project administration, Supervision, Writing – review and editing. LM: Conceptualization, Funding acquisition, Validation, Writing – review and editing.

Acknowledgements

The authors would like to thank the volunteers who took part in the trial, as well as the staff that assisted with the trial at the site. The authors would also like to thank American Journal Experts for medical writing support.

References

  1. Baguet C., Larghero J., Mebarki M. (2024). Early predictive factors of failure in autologous CAR T-cell manufacturing And/or efficacy in hematologic malignancies. Blood Adv. 8 (2), 337 – 342. doi:10.1182/bloodadvances.2023011992
  2. Boulch M., Cazaux M., Cuffel A., Ruggiu M., Allain V., Corre B. (2023). A major role for CD4(+) T cells in driving cytokine release syndrome during CAR T cell therapy. Cell Rep. Med. 4 (9), 101161. doi:10.1016/j.xcrm.2023.101161
  3. Bove C., Maher J., Glover M. (2024). The role of CD4(+) CAR T cells in cancer immunotherapy. Transl. Cancer Res. 13 (5), 2580 – 2586. doi:10.21037/tcr-23-2044
  4. Brudno J. N., Maus M. V., Hinrichs C. S. (2024). CAR T cells and T-Cell therapies for cancer: a translational science review. JAMA 332 (22), 1924 – 1935. doi:10.1001/jama.2024.19462
  5. Cohen A. D., Garfall A. L., Stadtmauer E. A., Melenhorst J. J., Lacey S. F., Lancaster E. (2019). B cell maturation antigen-specific CAR T cells are clinically active in multiple myeloma. J. Clin. Invest. 129 (6), 2210 – 2221. doi:10.1172/JCI126397
  6. Dai H. R., Liu Y., Lu K. Y., He X., Guo H. L., Hu Y. H. (2022). Population pharmacokinetic modeling of caffeine in preterm infants with apnea of prematurity: new findings from concomitant erythromycin and AHR genetic polymorphisms. Pharmacol. Res. 184, 106416. doi:10.1016/j.phrs.2022.106416
  7. Dasyam N., George P., Weinkove R. (2020). Chimeric antigen receptor T-cell therapies: optimising the dose. Br. J. Clin. Pharmacol. 86 (9), 1678 – 1689. doi:10.1111/bcp.14281
  8. Davis L., Riccitelli N., Valencia N., Ch'en I. L., Tangri S., Brogdon J. L. (2021). Monitoring of tisagenlecleucel transgene DNA using a quantitative polymerase chain reaction assay. Mol. Ther. Methods Clin. Dev. 20, 535 – 541. doi:10.1016/j.omtm.2020.12.002
  9. Ette E. I., Ludden T. M. (1995). Population pharmacokinetic modeling: the importance of informative graphics. Pharm. Res. 12 (12), 1845 – 1855. doi:10.1023/a:1016215116835
  10. Han X., Sun J., Zhang Y., Jiang T., Zheng Q., Peng H. (2024). Population pharmacokinetics of ainuovirine and exposure-response analysis in human immunodeficiency virus-infected individuals. Chin. Med. J. Engl. 137 (20), 2473 – 2482. doi:10.1097/CM9.0000000000002917
  11. Hardiansyah D., Ng C. M. (2019). Quantitative systems pharmacology model of chimeric antigen receptor T-Cell therapy. Clin. Transl. Sci. 12 (4), 343 – 349. doi:10.1111/cts.12636
  12. Hooker A. C., Staatz C. E., Karlsson M. O. (2007). Conditional weighted residuals (CWRES): a model diagnostic for the FOCE method. Pharm. Res. 24 (12), 2187 – 2197. doi:10.1007/s11095-007-9361-x
  13. Hu H., Li Y., Piasecki J., Hosseyni D., Yan Z., Liu X. (2025). Population pharmacokinetics of orvacabtagene autoleucel, an autologous BCMA-directed chimeric antigen receptor T-Cell product, in patients with relapsed/refractory multiple myeloma. Clin. Cancer Res. 31, 1163 – 1171. doi:10.1158/1078-0432.CCR-24-2753
  14. Jia Q., Qin D., He F., Xie Q., Ying Z., Zhang Y. (2021). Peripheral eosinophil counts predict efficacy of anti-CD19 CAR-T cell therapy against B-lineage non-Hodgkin lymphoma. Theranostics 11 (10), 4699 – 4709. doi:10.7150/thno.54546
  15. Jiang Z., Liao R., Lv J., Li S., Zheng D., Qin L. (2021). IL-6 trans-signaling promotes the expansion and anti-tumor activity of CAR T cells. Leukemia 35 (5), 1380 – 1391. doi:10.1038/s41375-020-01085-1
  16. Kang L., Tang X., Zhang J., Li M., Xu N., Qi W. (2020). Interleukin-6-knockdown of chimeric antigen receptor-modified T cells significantly reduces IL-6 release from monocytes. Exp. Hematol. Oncol. 9, 11. doi:10.1186/s40164-020-00166-2
  17. Karlsson M. O., Savic R. M. (2007). Diagnosing model diagnostics. Clin. Pharmacol. Ther. 82 (1), 17 – 20. doi:10.1038/sj.clpt.6100241
  18. Lee S. Y., Lee D. H., Sun W., Cervantes-Contreras F., Basom R. S., Wu F. (2023). CD8(+) chimeric antigen receptor T cells manufactured in absence of CD4(+) cells exhibit hypofunctional phenotype. J. Immunother. Cancer 11 (11), e007803. doi:10.1136/jitc-2023-007803
  19. Li M., Xue S. L., Tang X., Xu J., Chen S., Han Y. (2022). The differential effects of tumor burdens on predicting the net benefits of ssCART-19 cell treatment on r/r B-ALL patients. Sci. Rep. 12 (1), 378. doi:10.1038/s41598-021-04296-3
  20. Li M., Gao L., Wang Z., Zeng L., Chen C., Wang J. (2024). Population pharmacokinetics and dose optimization of ceftazidime in critically ill children. Front. Pharmacol. 15 15 – 2024. doi:10.3389/fphar.2024.1470350
  21. Liu C., Ayyar V. S., Zheng X., Chen W., Zheng S., Mody H. (2021). Model-based cellular kinetic analysis of chimeric antigen Receptor-T cells in humans. Clin. Pharmacol. Ther. 109 (3), 716 – 727. doi:10.1002/cpt.2040
  22. Liu X. D., Wang H. N., Zeng L. J., Liang Y. Z., Li F. L., Zhu J. Y. (2026). Association of baseline immune cell composition with CAR-T cell expansion and survival in relapsed/refractory large B-Cell lymphoma. J. Transl. Med. 24, 1041. doi:10.1186/s12967-026-08394-6
  23. Ma K. F., Liu Y. X., Jiao Z., Lv J. H., Yang P., Wu J. Y. (2020). Population pharmacokinetics of vancomycin in kidney transplant recipients: model building and parameter optimization. Front. Pharmacol. 11, 563967. doi:10.3389/fphar.2020.563967
  24. Mu W., Long X., Cai H., Chen C., Hu G., Lou Y. (2022). A model perspective explanation of the long-term sustainability of a fully human BCMA-targeting CAR (CT103A) T-Cell immunotherapy. Front. Pharmacol. 13, 803693. doi:10.3389/fphar.2022.803693
  25. Mueller K. T., Maude S. L., Porter D. L., Frey N., Wood P., Han X. (2017). Cellular kinetics of CTL019 in relapsed/refractory B-cell acute lymphoblastic leukemia and chronic lymphocytic leukemia. Blood 130 (21), 2317 – 2325. doi:10.1182/blood-2017-06-786129
  26. Ogasawara K., Dodds M., Mack T., Lymp J., Dell'Aringa J., Smith J. (2021). Population cellular kinetics of lisocabtagene maraleucel, an autologous CD19-Directed chimeric antigen receptor T-Cell product, in patients with relapsed/refractory large B-Cell lymphoma. Clin. Pharmacokinet. 60 (12), 1621 – 1633. doi:10.1007/s40262-021-01039-5
  27. Pan J., Tan Y., Shan L., Seery S., Deng B., Ling Z. (2025). Allogeneic CD5-specific CAR-T therapy for relapsed/refractory T-ALL: a phase 1 trial. Nat. Med. 31 (1), 126 – 136. doi:10.1038/s41591-024-03282-2
  28. Pfeifer R., Henze J., Wittich K., Gosselink A., Kinkhabwala A., Gremse F. (2022). A multimodal imaging workflow for monitoring CAR T cell therapy against solid tumor from whole-body to single-cell level. Theranostics 12 (11), 4834 – 4850. doi:10.7150/thno.68966
  29. Porter D. L., Hwang W. T., Frey N. V., Lacey S. F., Shaw P. A., Loren A. W. (2015). Chimeric antigen receptor T cells persist and induce sustained remissions in relapsed refractory chronic lymphocytic leukemia. Sci. Transl. Med. 7 (303), 303ra139. doi:10.1126/scitranslmed.aac5415
  30. Qi T., McGrath K., Ranganathan R., Dotti G., Cao Y. (2022). Cellular kinetics: a clinical and computational review of CAR-T cell pharmacology. Adv. Drug Deliv. Rev. 188, 114421. doi:10.1016/j.addr.2022.114421
  31. Rial Saborido J., Volkl S., Aigner M., Mackensen A., Mougiakakos D. (2022). Role of CAR T cell metabolism for therapeutic efficacy. Cancers (Basel) 14 (21), 5442. doi:10.3390/cancers14215442
  32. Shah N. N., Lee D. W., Yates B., Yuan C. M., Shalabi H., Martin S. (2021). Long-term Follow-Up of CD19-CAR T-Cell therapy in children and young adults with B-ALL. J. Clin. Oncol. 39 (15), 1650 – 1659. doi:10.1200/JCO.20.02262
  33. Sheih A., Voillet V., Hanafi L. A., DeBerg H. A., Yajima M., Hawkins R. (2020). Clonal kinetics and single-cell transcriptional profiling of CAR-T cells in patients undergoing CD19 CAR-T immunotherapy. Nat. Commun. 11 (1), 219. doi:10.1038/s41467-019-13880-1
  34. Sheiner L. B., Ludden T. M. (1992). Population pharmacokinetics/dynamics. Annu. Rev. Pharmacol. Toxicol. 32, 185 – 209. doi:10.1146/annurev.pa.32.040192.001153
  35. Singh A. P., Chen W., Zheng X., Mody H., Carpenter T. J., Zong A. (2021). Bench-to-bedside translation of chimeric antigen receptor (CAR) T cells using a multiscale systems pharmacokinetic-pharmacodynamic model: a case study with anti-BCMA CAR-T. CPT Pharmacometrics Syst. Pharmacol. 10 (4), 362 – 376. doi:10.1002/psp4.12598
  36. Stein A. M., Grupp S. A., Levine J. E., Laetsch T. W., Pulsipher M. A., Boyer M. W. (2019). Tisagenlecleucel model-based cellular kinetic analysis of chimeric antigen Receptor-T cells. CPT Pharmacometrics Syst. Pharmacol. 8 (5), 285 – 295. doi:10.1002/psp4.12388
  37. Tan X., Wang J., Chen S., Liu L., Li Y., Tu S. (2025). Novel CD19 Fast-CAR-T cells vs. CD19 conventional CAR-T cells for the treatment of relapsed/refractory CD19-positive B-cell acute lymphoblastic leukemia. Chin. Med. J. Engl. 138 (19), 2491 – 2497. doi:10.1097/CM9.0000000000003479
  38. Tegenge M. A., Wang X., Liu J., Zhu H., Fashoyin-Aje L. A. (2025). FDA experience on CAR T cell pharmacokinetics/pharmacodynamics and model-based assessments. Clin. Pharmacol. Ther. 118, 324 – 330. doi:10.1002/cpt.3703
  39. Wu Y., Feng X., Li J., Wang X., Yang C., Zhao L. (2019). Model-based meta-analysis in ankylosing spondylitis: a quantitative comparison of biologics and small targeted molecules. Clin. Pharmacol. Ther. 105 (5), 1244 – 1255. doi:10.1002/cpt.1305
  40. Zhang Y., Xu Y., Dang X., Zhu Z., Qian W., Liang A. (2023). Challenges and optimal strategies of CAR T therapy for hematological malignancies. Chin. Med. J. Engl. 136 (3), 269 – 279. doi:10.1097/CM9.0000000000002476

Citation

Bian, Y., et al. (2026). Population cellular kinetics and exposure-CSR analysis of CAR-T cells in Chinese patients with relapsed/refractory B-ALL. Frontiers in Pharmacology, Volume 17. https://doi.org/
Received
Accepted
Published
Volume
17 - 2026

Copyright

Copyright © 2026 Bian, Yuan, Huang, Zhang, Xue, Xia, Kang, Wang, Zhang, Yu, Wu and Miao.

License: https://creativecommons.org/licenses/by/4.0/

*Correspondence: Lei Yu; Depei Wu; Liyan Miao

Article metrics

Views
0
Downloads
0
Citations
0
Shares
0

Related articles

ReviewSilencing without clearing: pharmacological modulation of the SASP in aging and diseaseLukas Fresser, Zofia Chrienova, Jan Gebrt, Patrik Oleksak, Kamil Kuca, Eugenie NepovimovaReviewFlavonoids in cancer pharmacology: molecular mechanisms, tumor microenvironment regulation, and drug resistanceZhengyuan Yu, Zhanyi Zhou, Ni Yin, Hongcheng GeMethodsA high-throughput 96-well plate method to measure human uterine contractilityEmilie Huon, Justine Rayssac, Marie-Charlotte Lamau, Carole Abo, Gauthier Dietrich, Ylann Abrahami, Jordan Kouby, Julien Adam, Nicolas R. Chevalier