Effectiveness of CAR-T Cell Therapies for Relapsed/Refractory Follicular Lymphoma: An External Control Arm Study
Article information
Abstract
Purpose
Axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), and lisocabtagene maraleucel (liso-cel) have received regulatory approval for relapsed or refractory follicular lymphoma (FL). However, the data are scarce on their comparative effectiveness against the salvage therapies available in real-world settings. This study aimed to indirectly compare treatment outcomes of axi-cel, tisa-cel, and liso-cel versus usual care in South Korean patients with FL.
Materials and Methods
To assess effectiveness in real-world data, aggregate data from the ZUMA-5, ELARA, and TRANSCEND FL studies were compared with individual patient data from the Samsung Medical Center–Lymphoma Cohort Study (SMC-LCS). Patients meeting ZUMA-5, ELARA, and TRANSCEND FL eligibility criteria were selected as the external control arm. All eligible treatment lines per patient were analyzed as independent episodes and weighted using the matching-adjusted indirect comparison method. Time-to-event outcomes were assessed with weighted Kaplan-Meier analysis, and adjusted hazard ratios (HRs) were estimated using Cox proportional hazards models.
Results
Axi-cel included 127 patients, tisa-cel included 94, liso-cel included 101, and 121 episodes from 49 patients were analyzed in the external control arm. The weighted HRs for overall survival and progression-free survival for axi-cel versus the external control were 0.37 (95% confidence interval [CI], 0.21 to 0.64) and 0.35 (95% CI, 0.20 to 0.59), respectively. For tisa-cel, the HRs were 0.24 (95% CI, 0.11 to 0.53), and 0.35 (95% CI, 0.20 to 0.60), respectively. For liso-cel, the HRs were 0.38 (95% CI, 0.13 to 1.04) and 0.36 (95% CI, 0.15 to 0.88), respectively.
Conclusion
All three chimeric antigen receptor T-cell therapies showed outstanding effectiveness compared to conventional treatments in usual care in South Korea.
Introduction
Follicular lymphoma (FL) accounts for 35% of all non-Hodgkin lymphomas (NHLs) and represents 20%-30% of all cases of indolent NHL [1,2]. FL typically follows a slow progression, with a median overall survival (OS) exceeding 15 years. However, it remains challenging to cure. While patients with FL generally have longer survival compared to those with aggressive B-cell lymphomas due to its indolent nature, they are still at risk of disease progression and subsequent mortality. Moreover, some patients experience an aggressive disease course, such as progression of disease within 24 months (POD24) or large-cell transformation, both of which are associated with poor outcomes [3-5].
Several salvage treatment options have been proposed for relapsed or refractory (r/r) FL, the prognosis for these individuals remains suboptimal [6]. According to the 2023 update on diagnosis and management of FL, patients with POD24 or heavily pre-treated disease should be referred to a specialized center for evaluation for chimeric antigen receptor T-cell (CAR-T) therapy or autologous stem cell transplantation [7]. Notably, CAR-T therapy has shown remarkable success in hematological malignancies, particularly in r/r B-cell malignancies [8]. Single-arm trials, including ZUMA-5, ELARA, and TRANSCEND FL, have confirmed the feasibility for r/r FL [9-11]. Furthermore, studies using real-world data suggest that axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), and lisocabtagene maraleucel (liso-cel) are effective treatment options for r/r FL. However, these findings were primarily derived from Western population [12-14].
FL exhibits geographic variation in incidence and characteristics. It is the second most common subtype of NHL in the United States and Europe [1], with approximately 15,000 new cases diagnosed annually [15]. In contrast, FL is less common in Asia, though its incidence has been increasing over recent decades in several Asian countries [16]. Moreover, the presentation of FL tends to vary by region [17]. For instance, high-grade, low-stage, and low B-cell lymphoma 2 (BCL-2) expression are the typical presentations in the South Korean population, whereas low-grade, high-stage, and BCL-2 translocation are more prevalent in the United States [18]. Existing clinical data on the comparative effectiveness of CAR-T therapies for FL were derived from indirect comparison with the real-world treatment outcomes from the Western population [14,19]. Given these differences, previous indirect comparisons of CAR-T therapies versus real-world treatment outcomes, primarily based on Western data, may not fully reflect treatment effectiveness in Asian populations. In this regard, this study aimed to indirectly compare treatment outcomes of axi-cel, tisa-cel, and liso-cel with usual care in South Korean patients.
Materials and Methods
1. Data source
In this study, data for the external control arm were derived from individual patient data (IPD) in Samsung Medical Center–Lymphoma Cohort Study (SMC-LCS; NCT#00822731, NCT#01877109, NCT#03117036), which has been ongoing since 2008. The diagnosis of FL was pathologically confirmed by expert lymphoma pathologists using immunohistochemistry findings [20]. The SMC-LCS comprised patients diagnosed with Hodgkin’s and non-Hodgkin’s lymphomas, with demographic information, pre-treatment details, and treatment regimens recorded at diagnosis. Variables not initially structured were extracted from electronic medical records by trained personnel and physicians, following legal procedures. Data for the intervention group, consisting of axi-cel, tisa-cel, and liso-cel, were extracted as aggregate data from the ZUMA-5, ELARA, and TRANSCEND FL studies, respectively.
2. Study design and study population
In this retrospective study, published summary data of ZUMA-5 (NCT#03105336), ELARA (NCT#03568461), and TRANSCEND FL (NCT#02631044) single-arm trials were indirectly compared with an external control arm constructed from SMC-LCS. Eligibility criteria for the external control arm were adopted from the intervention trials (ZUMA-5 for axi-cel, ELARA for tisa-cel, and TRANSCEND FL for liso-cel): (1) ≥ 18 years of age; (2) FL (grades 1, 2, 3a) confirmed before index treatment; (3) refractory to second-line or later systemic therapy including an anti-CD20 antibody and an alkylating agent, or relapsed within 6 months after completing therapy or during/within 6 months after anti-CD20 maintenance therapy, or relapsed after autologous hematopoietic stem cell transplantation (HSCT); (4) radiographically measurable disease; and (5) Eastern Cooperative Oncology Group performance status that is either 0 or 1 at index treatment [9-11].
The following patients were excluded from the external control arm: (1) evidence of histologic transformation; (2) FL grade 3b; (3) prior treatment with anti-CD19 therapy, gene therapy or any adoptive T-cell therapy; (4) prior allogeneic HSCT; (5) active central nervous system involvement by malignancy; (6) active neurological autoimmune or inflammatory disorders (e.g. Guillain-Barre syndrome, amyotrophic lateral sclerosis); (7) presence of active or prior hepatitis B or C as indicated by serology; (8) presence of human immunodeficiency virus antibody; and (9) previous or concurrent malignancy (Fig. 1).
Study flowchart of external control arm. CNS, central nervous system; HIV, human immunodeficiency virus; HSCT, hematopoietic stem cell transplantation. a)Episode refers to including a patient in the study multiple times, with each eligible line considered using its respective index date (the start date of the eligible treatment).
The treatment strategies for FL at SMC-LCS included six cycles of induction immunochemotherapy, such as R-CVP (rituximab, cyclophosphamide, vincristine, and prednisone), R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone), or BR (bendamustine and rituximab). Patients who responded to induction chemotherapy received rituximab, administered either subcutaneously or intravenously, every two months as maintenance therapy for two years. For patients who relapsed or progressed during or after induction and maintenance therapy, a wide range of subsequent treatments was employed, reflecting real-world clinical practice in South Korea. These included salvage chemotherapy regimens (e.g., GDP [gemcitabine, dexamethasone, cisplatin], DHAP [dexamethasone, high-dose cytarabine (Ara-C), cisplatin], ICE/DEXA [ifosfamide, carboplatin, etoposide, dexamethasone], R-GDP [rituximab, gemcitabine, dexamethasone, cisplatin]), immunomodulatory combinations (e.g., rituximab plus lenalidomide, thalidomide), targeted therapies such as phosphoinositide 3-kinase inhibitors (e.g., copanlisib, idelalisib), Bruton tyrosine kinase inhibitors (e.g., zanubrutinib, pirtobrutinib), and investigational agents administered as part of clinical trials. High-dose chemotherapy followed by autologous stem cell transplantation was also performed in selected patients [6].
3. Time-zero
In randomized controlled trials, time-zero is typically defined as the point when a patient receives the interventional or comparator drug after enrollment, making its determination relatively straightforward. However, in observational studies using external control groups, identifying time-zero can be more complex, as patients in the external cohort may have multiple potential entry points. This is particularly relevant for patients with FL, who often have longer survival times and therefore more possible entry points. Researchers must carefully select the appropriate treatment line to define the entry point for the external control group, while addressing the potential for bias introduced by this choice.
In this study, when defining time-zero for the external control arm, we initially considered methods suitable for studies incorporating aggregate data, with options such as selecting all eligible treatment lines or a single eligible treatment line (e.g., the first or last line). We selected all eligible treatment lines in the external control arm and analyzed them as independent episodes rather than on a per-patient basis. This approach increased statistical power, enabled robust variance estimation through group-based analysis, and aimed to minimize potential bias by applying weights to balance baseline characteristics (Fig. 2) [21,22].
Diagram of time zero selection for external control arm. LoT, line of treatment.
4. Outcomes
To evaluate the effectiveness of CAR-T therapies (i.e., axi-cel, tisa-cel, and liso-cel) compared with conventional therapies, we analyzed OS and progression-free survival (PFS) in each treatment group. For the CAR-T cohort, follow-up began at the date of CAR-T infusion after experiencing relapse or refractory to two or more previous treatment lines, including anti-CD20 monoclonal antibodies and alkylating agents [12,13]. For treatment outcomes of CAR-T cohort, Kaplan-Meier curves for OS and PFS were reconstructed using the Engauge Digitizer software ver. 12.1. The primary data were extracted from publications by Neelapu et al. [9], Dreyling et al. [10], and Morschhauser et al. [11]. Additionally, the Kaplan-Meier curves of ReCORD-FL and SCHOLAR-5, which served as external control arms in previously published studies, were extracted and utilized as supplementary graphs in this study [12,13].
In the external control arm, all eligible treatment lines were selected and analyzed as independent episodes. OS was defined as the time from the start of the index treatment to death from any cause. PFS was the time from the start of the index treatment to disease progression or death from any cause. The date of the last follow-up was used for censoring subjects in the analyses of OS and PFS.
5. Statistical analysis
There were 127, 94, and 101 patients in the axi-cel, tisa-cel, and liso-cel arms, respectively. In the external control arm, 121 episodes cloned from 49 patients were analyzed. Using the matching adjusted indirect comparison (MAIC) approach, the baseline characteristics of the conventional therapy group were matched to those of the CAR-T therapy groups. In the analysis comparing tisa-cel to the external control arm, adjustments were made for the following variables measured at the time of index treatment: age, sex, disease stage at study entry, number of previous lines of chemotherapy, refractory status to the last prior therapy, POD24, Follicular Lymphoma International Prognostic Index (FLIPI), and previous autologous HSCT. Balance was assessed using the absolute standardized mean difference (aSD) for each covariate.
After assigning weights, OS and PFS were compared using statistical tests. Specifically, the log-rank test was employed to compare the survival curve of axi-cel, tisa-cel, and liso-cel arms with the weighted survival curve of the external control group. Additionally, a weighted Cox proportional hazards model was used to estimate adjusted hazard ratio with 95% confidence interval (CI), allowing for a balanced comparison between the groups. Crude values were also presented.
We conducted a sensitivity analysis to evaluate the robustness of our findings. To address potential bias in index treatment selection, we performed an analysis selecting only the first eligible treatment line for each patient.
All statistical analyses were performed using SAS ver. 9.4 (SAS Institute Inc.) and R ver. 4.5.1 (R Foundation for Statistical Computing). A two-sided p-value of < 0.05 was considered statistically significant.
Results
1. Patient characteristics
Among 366 patients with newly diagnosed FL in SMC-LCS from 2008 to February 2020, 49 patients met the eligibility criteria of ZUMA-5, ELARA, and TRANSCEND FL (Fig. 1).
In the axi-cel group, significant differences were observed in several baseline characteristics compared to external controls: age ≥ 65 years (31.5% vs. 5.0%; aSD, 0.73), disease stage 1-2 (14.2% vs. 3.3%; aSD, 0.39), disease stage 3-4 (85.8% vs. 96.7%; aSD, 0.39), POD24 (55.1% vs. 70.3%; aSD, 0.32), and refractory to prior line of treatment (68.5% vs. 47.9%; aSD, 0.43). After weighting, axi-cel and external control groups had an aSD < 0.1 for all categorical variables, confirming good balance between these groups.
In the tisa-cel group, significant differences were observed in several baseline characteristics compared to external controls: age ≥ 65 years (25.5% vs. 5.0%; aSD, 0.6), disease stage 1-2 (14.4% vs. 3.3%; aSD, 0.4), disease stage 3-4 (85.6% vs. 96.7%; aSD, 0.4), refractory to prior line of treatment (78.7% vs. 47.9%; aSD, 0.67), FLIPI high (60.6% vs. 38.0%; aSD, 0.46), FLIPI low/intermediate (39.4% vs. 62.0%; aSD, 0.46). After weighting, tisa-cel and external control groups had an aSD < 0.1 for all categorical variables.
In the liso-cel group, significant differences were observed in several baseline characteristics compared to external controls: age ≥ 65 years (40.6% vs. 5.0%; aSD, 0.94), disease stage 1-2 (11.2% vs. 3.3%; aSD, 0.31), disease stage 3-4 (88.8% vs. 96.7%; aSD, 0.31), POD24 (43.6% vs. 70.3%; aSD, 0.56), and FLIPI high (57.4% vs. 38.0%; aSD, 0.40). After weighting, liso-cel and external control groups achieved good balance with aSD < 0.1 for most categorical variables (Table 1). Detailed baseline characteristics before weighting are provided in S1-S3 Tables.
2. Comparison between axi-cel and external control arm
Fig. 3A shows weighted Kaplan-Meier curves for OS and PFS for axi-cel (ZUMA-5) versus conventional therapy (SMC-LCS). The median OS was not reached (NR) in axi-cel versus 36.5 months (95% CI, 5.13 to not estimable [NE]) in SMC-LCS. The median PFS was 40.2 (95% CI, 28.9 to NE) in axi-cel versus 5.7 months (95% CI, NE to NE) in SMC-LCS, corresponding to weighted hazard ratios (HRs) of 0.37 (95% CI, 0.21 to 0.64) for OS and 0.35 (95% CI, 0.20 to 0.59) for PFS (Table 2, S4 Table). A comparison of the point estimates of median OS and PFS in SCHOLAR-5 and SMC-LCS, weighted against axi-cel (ZUMA-5), the median OS in SCHOLAR-5 was 59.8 months (95% CI, 21.9 to NE), while 36.5 months (95% CI, 5.13 to NE) in SMC-LCS. The median PFS was 12.7 months (95% CI, 6.2 to 14.7) in SCHOLAR-5 and 5.7 months (95% CI, NE to NE) in SMC-LCS. These results suggest a trend toward higher OS and PFS in SCHOLAR-5 (Fig. 3A).
Weighted Kaplan-Meier plots comparing chimeric antigen receptor T-cell (CAR-T) therapies (axicabtagene ciloleucel [axi-cel] [A], tisagenlecleucel [tisa-cel] [B], and lisocabtagene maraleucel [liso-cel] [C]) versus conventional therapies for overall survival and progression-free survival. CI, confidence interval; SMC-LCS, Samsung Medical Center–Lymphoma Cohort Study. a)Matching-adjusted indirect comparison weights were based on each CAR-T therapy dataset, including parameters such as age, sex, disease stage at study entry, number of previous lines of antineoplastic therapy, last prior therapy refractory status, progression of disease within 24 months, and previous autologous hematopoietic stem cell transplantation. Conventional therapies (SCHOLAR-5, ReCORD-FL) were converted into summary.
Weighted OS and PFS of CAR-T therapies (axi-cel, tisa-cel, and liso-cel) versus conventional therapies
3. Comparison between tisa-cel and external control arm
Fig. 3B shows weighted Kaplan-Meier curves for OS and PFS for tisa-cel (ELARA) versus conventional therapy (SMC-LCS). The median OS was NR in tisa-cel versus 33.3 months (95% CI, NE to NE) in SMC-LCS. The median PFS was NR in tisa-cel versus 5.3 months (95% CI, NE to NE) in SMC-LCS, corresponding to weighted HRs of 0.24 (95% CI, 0.11 to 0.53) for OS and 0.35 (95% CI, 0.20 to 0.60) for PFS (Table 2, S5 Table). A comparison of the point estimates of median OS and PFS in ReCORD-FL and SMC-LCS, weighted on the tisa-cel group (ELARA), the median OS in ReCORD-FL was NR, while in SMC-LCS 20.5 months (95% CI, 8.7 to 38.5). The median PFS was 13.1 months (95% CI, 8.1 to NR) in ReCORD-FL and 3 months (95% CI, 2.0 to 5.2) in SMC-LCS. These results suggest a trend toward higher OS and PFS in ReCORD-FL (Fig. 3B).
4. Comparison between liso-cel and external control arm
Fig. 3C shows weighted Kaplan-Meier curves for OS and PFS for liso-cel (TRANSCEND FL) versus conventional therapy (SMC-LCS). The median OS was NR in liso-cel versus 20.7 months (95% CI, 12.1 to 36.5) in SMC-LCS. The median PFS was NR (95% CI, 19.0 to NE) in liso-cel versus 3.0 months (95% CI, 2.1 to 4.8) in SMC-LCS. After MAIC adjustment, the median OS was NR (95% CI, 33.3 to NE) in liso-cel versus SMC-LCS, corresponding to weighted HRs of 0.38 (95% CI, 0.13 to 1.04) for OS and 0.36 (95% CI, 0.15 to 0.88) for PFS. The median PFS after MAIC was 16.8 months in liso-cel (Table 2, S6 Table).
5. Sensitivity analysis
The sensitivity analysis yielded point estimates similar to those of the main analysis for axi-cel and liso-cel, but the CIs were wide due to the reduced sample size (from 121 episodes to 49 patients), which became even more pronounced with the application of weighting. For tisa-cel, the HRs approached unity with substantially wider CIs, suggesting reduced treatment effect when limiting to first eligible treatment lines (S7-S9 Tables).
Discussion
In this study, we indirectly compared published literature on CAR-T therapies (axi-cel, tisa-cel, and liso-cel) with usual care recorded in the real-world data in South Korea to determine the effectiveness of CAR-T therapies in improving OS and PFS in patients with r/r FL. We successfully controlled for confounding factors by securing comparability between groups using the MAIC method to address limitations due to the absence of IPD. Our findings were consistent with the previous indirect treatment comparison studies, confirming the effectiveness of CAR-T therapies in comparison with the usual care for r/r FL in the Korean population.
According to previous literature, after adjusting for baseline factors using weighting, a comparison between CAR-T therapies (ZUMA-5 and ELARA) and usual care (SCHOLAR-5, ReCORD-FL) demonstrated improved outcomes with CAR-T therapies in terms of OS and PFS. The corresponding HRs are summarized in Fig. 3 [12]. The external control cohorts used as the comparators comprised patients who met the eligibility criteria for the respective single-arm clinical trials and included patient records from international cancer centers. The SMC-LCS group demonstrated shorter median outcome times for OS and PFS compared to these external control cohorts, indicating that the protective effect of CAR-T therapies observed in this study was even more pronounced than the results reported in previous literature [12]. The outcomes observed in our external control group appeared worse than those reported in SCHOLAR-5 and ReCORD-FL, particularly in terms of OS and PFS. These differences in outcomes between the present study and the Western control cohorts may be partially attributed to several factors. First, although baseline characteristics were adjusted using MAIC, unmeasured confounding due to factors such as disease history prior to the study period cannot be ruled out. Second, biological and clinical differences in FL presentation between regions may have contributed. For example, high-grade, low-stage FL with low BCL-2 expression is more frequently observed in the Korean population, whereas low-grade, high-stage disease with BCL-2 translocation is more common in Western populations. High-grade FL is often associated with more aggressive disease and worse prognosis. Third, a portion of patients in Western control cohorts were drawn from clinical trials, and their data were collected using structured case report forms, which may reflect a more selectively defined population than the current study. These differences should be considered when interpreting cross-study comparisons.
Notably, all three CAR-T therapies demonstrated comparable effectiveness against conventional therapy when adjusted for baseline characteristics. The similar HRs observed for OS and PFS across axi-cel, tisa-cel, and liso-cel suggest that these therapies may offer similar clinical benefits for Korean patients with r/r FL, though direct head-to-head comparisons would be needed to definitively establish relative efficacy. The ideal approach would be an IPD versus IPD comparison; however, accessing IPD data for relatively novel therapies has been challenging. To address these limitations, we utilized summary data, which are more readily accessible, and applied the MAIC method to ensure comparability between groups.
However, careful interpretation is necessary when evaluating this study, considering the following limitations. First, in selecting the index treatment line, this study used the eligible treatment line as the unit of analysis. Consequently, one patient might have been included multiple times depending on the number of eligible treatment lines, which could introduce selection bias. Nevertheless, a simulation study on index treatment selection found that the method used in this study correctly identified the treatment effect when appropriate measures were applied (e.g., group-based robust variance estimator, standardized mortality ratio weighting) [21]. Second, despite using MAIC to match baseline characteristics of eligible patients in SMC-LCS with summary characteristics from CAR-T studies, there may still be potential bias due to residual imbalance. Third, the selection/exclusion criteria applied in the CAR-T single-arm clinical trials could not be fully applied to conventional therapy. Fourth, information on whether patients met Groupe d’Etude des Lymphomes Folliculaires (GELF) criteria at the time of relapse was not collected. Although GELF was originally developed to guide treatment initiation at diagnosis, it is now increasingly used to inform treatment decisions in later lines of therapy as well. Therefore, documenting GELF status at relapse could provide valuable context for evaluating disease burden and interpreting outcomes in relapsed/refractory FL. Future studies should consider incorporating this information into patient characterization. Fifth, the CAR-T trials (ZUMA-5, ELARA, and TRANSCEND FL) reported outcomes based on the treated population rather than an intention-to-treat cohort. This may lead to overestimation of treatment efficacy, as patients who were enrolled but did not receive CAR-T therapy were excluded from the analysis. Lastly, the external control group consisted of 49 patients, which may limit the statistical power and generalizability of the findings. This limitation was addressed by applying appropriate weighting to enhance comparability between groups. Given the relatively small sample size, the findings should be interpreted with caution, and future studies with larger cohorts are warranted to validate these results. Nonetheless, this study reinforced existing evidence by demonstrating that CAR-T therapy is expected to improve OS and PFS in Korean populations with r/r FL [12,13].
Electronic Supplementary Material
Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).
Notes
Ethical Statement
The study utilized anonymized data and obtained approval from the Institutional Review Board (IRB No. SMC 2023-12-143-001) of Samsung Medical Center, waiving the requirement for informed consent.
Author Contributions
Conceived and designed the analysis: Seo HJ, Kim JH, Shin JY, Kim SJ.
Collected the data: Yoon SE, Kim WS, Kim SJ.
Contributed data or analysis tools: Seo HJ, Kim JH, Yon DK, Shin JY, Kim SJ.
Performed the analysis: Seo HJ, Kim JH.
Wrote the paper: Seo HJ, Kim JH.
Conflicts of Interest
J.Y.S. received grants from the Ministry of Food and Drug Safety, the National Research Foundation of Korea, and grants from Pfizer, Johnson & Johnson, and GSK. No other relationships or activities have influenced the submitted work. All other authors declared no competing interests for this work.
Funding
This research was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: HR20C0025).
This research was also supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: HE23C002800).
This research was additionally supported by a grant from the Ministry of Food and Drug Safety, Korea (grant number: RS-2024-00332632).