Predictors of Response to CD19 CAR T-Cell Therapy in Large B-Cell Lymphoma: A Consolidated Review

Abstract

Purpose of review:

CD19-directed chimeric antigen receptor T-cell (CAR-T) therapy has transformed outcomes for relapsed/refractory large B-cell lymphoma (LBCL), yet nearly half of treated patients relapse, and toxicities remain frequent. A deeper understanding of response predictors is urgently needed to guide patient selection, treatment optimization, and development of rational combination strategies.

Recent findings:

Emerging data reveal that response to CAR-T therapy is shaped by patient-specific, tumor-intrinsic, and treatment-related factors. Clinical variables such as age, performance status, inflammation, and microbiome composition influence efficacy. Tumor burden, disease distribution, histologic subtype, and genomic alterations correlate with resistance. Treatment factors, including bridging strategies, lymphodepletion regimen, and CAR-T product design, affect expansion, persistence, and clinical outcomes. Novel insights from immune profiling, radiomics, and single-cell transcriptomics offer further granularity and predictive potential.

Summary:

Predictors of CAR-T response span diverse biological and clinical domains and are increasingly actionable. Integrating multimodal biomarkers into routine workflows can personalize care and improve outcomes. Prospective validation, real-time monitoring, and adaptive trial designs are essential next steps toward precision CAR-T therapy.

Introduction

Chimeric antigen receptor T-cell (CAR-T) therapy targeting CD19 has transformed the therapeutic landscape for patients with r/r LBCL. Pivotal clinical trials, such as ZUMA-1, have demonstrated high initial response rates, leading to regulatory approval of CD19-directed CAR-T products for multiple LBCL subtypes.(1)Nonetheless, long-term outcomes remain suboptimal. Nearly half of patients relapse within two years of treatment, and the risk of severe, potentially fatal toxicities, particularly cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS), poses additional clinical challenges.(2,3) These limitations underscore the pressing need to identify robust predictors of response and resistance, to refine patient selection and inform treatment strategies.

Scope of This Review

In this review, we synthesize current knowledge on factors influencing response to CD19 CAR-T therapy in LBCL. We highlight key determinants across three major domains: patient-, tumor-, and treatment- factors (Table 1). By integrating clinical, genomic, and immunologic insights, we aim to provide a comprehensive overview of emerging predictors of response that may guide precision treatment approaches and improve long-term outcomes in CAR-T recipients.

Table 1.

Predictors of Response to CAR-T Therapy: Consolidated Summary

Parameter	Favorable	Unfavorable	Non-Influential / Context-Dependent
Age	—	≥75 years: shorter EFS (12)	No significant outcome differences between younger and older patients (9–11)
Sex	—	Male sex: higher risk of shorter PFS and OS in one study (14)	No difference in CAR-T outcomes observed between sexes in trials (8,17)
Performance Status	—	ECOG ≥2: predicts worse PFS and OS (14,18,19)	—
Comorbidities	—	Severe multi-organ comorbidity (Severe4 index): shorter OS and PFS (18); Pulmonary impairment (low FEV1): shorter PFS (21)	No significant differences with moderate or stable organ dysfunction (8)
Socioeconomic / Racial Factors	—	Low SES: elevated inflammation and tumor burden (22–25); Black race: lower response and shorter PFS (24)	—
Inflammation	—	Elevated CRP, ferritin, LDH, IL-6: associated with reduced response and survival (26,27); Inflammatory profiles (e.g. InflaMix): associated with poor outcomes (28–31); Immunosuppressive effects (e.g. MDSCs, T-cell exhaustion) may impair response (32,33)	Independent negative effect despite tumor burden correlation (28)
Microbiome	—	Broad-spectrum antibiotic use pre-CAR-T: associated with inferior OS and impaired response; linked to loss of beneficial gut taxa (Ruminococcus, Faecalibacterium, Bacteroides) (34–36)	Causality remains unproven and may be confounded by inflammation or disease severity (narrative)
Tumor Burden	—	High MTV and bulky disease: associated with inferior CR and shorter PFS (38–40); Irregular lesion morphology, diffuse metabolic activity, high SUVmax and TLG predict poor outcomes (41–43)	—
Disease Distribution (Extranodal)	—	CNS and multi-organ extranodal involvement linked to early relapse and poor outcomes (44,45)	—
Histologic Subtype	PMBCL and transformed DLBCL: higher response and survival (8,53)	T cell/histiocyte-rich LBCL: poor response and early relapse (55–57)	—
Genomic Alterations	—	TP53 mutation and other high-risk lesions associated with resistance to CAR-T (62–66)	TME features (e.g. stromal exclusion, PD-L1) may contribute to resistance (67,68)
Bridging Therapy	Cytoreductive bridging (e.g. radiotherapy) associated with improved outcomes in patients with high tumor burden (69–71)	—	Effectiveness may depend on pre-treatment tumor burden; bridging alone not independently harmful (8)
Prior Lines of Therapy	—	Third-line or later CAR-T associated with lower response and survival (72)	—
Lymphodepletion	Adequate fludarabine exposure associated with better outcomes (78)	Infusion delay >3 days associated with worse PFS and OS (76,77)	Bendamustine-based lymphodepletion showed comparable efficacy and lower toxicity in frail patients (76,77)
CAR-T Product	Axi-cel: higher CR and PFS vs. tisa-cel (79,80)	—	Tisa-cel vs. liso-cel: comparable efficacy (81,82)
CAR-T Cell Expansion & Phenotype	Enriched TCM/TSCM CAR-T cells associated with durable remission (83)	Exhausted or Treg-like phenotypes (e.g. PD-1+TIM3+, FOXP3+) linked to poor response (83,85)	—

Patient-Related Predictors

Age:

Median age at diagnosis of DLBCL is approximately 67 years, (4) aligning with the typical age of many CAR-T recipients. Despite older patients often presenting with higher-risk disease features, both clinical trial analyses and real-world data generally support comparable CAR-T therapy efficacy across age groups.(5–8) In ZUMA-1, patients aged ≥65 had similar response rates and overall survival (OS) as younger patients.(9) Data from the CIBMTR registry also showed no meaningful differences in outcomes with tisa-cel or axi-cel between older and younger adults.(10) Similarly, Berning et al. reported comparable response rates and progression-free survival (PFS) in patients <70 versus ≥70 years.(11) In contrast, a Medicare-based analysis by Chihara et al. noted shorter event-free survival (EFS) in patients ≥75 compared to those aged 65–74.(12) While CAR-T therapy remains effective in older adults regardless of product, the potential for increased neurotoxicity in this population warrants attention and careful evaluation.

Sex:

Sex-based differences in immune activation and function have prompted investigation into whether CAR-T therapy outcomes vary by sex.(13) Retrospective data suggest a potential advantage for female patients.(14–16) In a multicenter real-world study, Nastoupil et al. reported nearly a twofold increased risk of shortened PFS and OS in male patients compared to females treated with axi-cel.(14) However, this association was not observed in pivotal trials or other large observational cohorts.(8,17) Overall, while some real-world data suggest a sex-related survival difference, the evidence remains inconsistent and further studies are needed to clarify the role of sex in CAR-T therapy efficacy.

Performance Status:

Although pivotal CAR-T therapy trials typically exclude patients with poor performance status,(7,8,17) such individuals are commonly encountered in real-world practice. Impaired performance status, assessed by ECOG or Karnofsky scores, is often associated with aggressive disease biology and extensive prior treatment. Multiple retrospective analyses have consistently linked poor performance status with reduced response rates and inferior progression-free and overall survival.(14,18,19) Given its prognostic significance, performance status should be carefully assessed prior to CAR-T therapy, and efforts to optimize functional status may improve clinical outcomes.

Comorbidities:

While comorbidities are established predictors of non-relapse mortality in hematopoietic cell transplantation,(20) their role in CAR-T therapy is less clearly defined. Individual comorbidities may have context-dependent effects: for instance, Sdayoor et al. found that lower FEV1 was associated with shorter PFS,(21) while the TRANSCEND trial showed no difference in response rates between patients with or without impaired renal function or reduced ejection fraction.(8)Composite scores, such as the Severe4 index,(18) which captures severe dysfunction in key organ systems, have demonstrated prognostic value for both OS and PFS. Overall, while comorbidities may inform risk assessment, their ability to predict CAR-T therapy efficacy remains limited and requires further refinement.

Social and Racial Disparities:

Social and racial disparities affect both access to and outcomes of CAR-T therapy in LBCL.(22–24) Patients from minoritized groups and lower socioeconomic backgrounds are less likely to receive CAR-T therapy and may present with more aggressive disease features; Knight et al. reported that low-SES patients had elevated inflammatory markers and tumor burden at baseline.(25) Additionally, Black patients have shown lower response rates and shorter PFS following axi-cel in some real-world and clinical trial cohorts.(24) These findings highlight the need to address systemic and biological contributors to inequity in CAR-T therapy.

Inflammation:

Baseline systemic inflammation and immune dysregulation are strongly associated with inferior CAR-T therapy outcomes in LBCL. Elevated markers such as CRP, ferritin, LDH, and IL-6 correlate with lower response rates, and shorter PFS and OS.(26,27) While systemic inflammation is often correlated with tumor burden, its detrimental impact on CAR-T therapy efficacy appears to be independent of it.(28) Models like InflaMix have identified inflammatory patient subsets with poor outcomes, though normalization of markers pre-infusion may mitigate this risk(28) Other scores incorporating inflammatory markers are also predictive of inferior disease control.(29–31) Inflammation may impair response through immunosuppressive effects on the tumor microenvironment, expansion of myeloid-derived suppressor cells, and T-cell exhaustion.(32,33) Collectively, these findings highlight the importance of controlling inflammation and restoring immune fitness prior to CAR-T therapy.

Microbiome:

The gut microbiome has emerged as a potential factor influencing CAR-T therapy efficacy in LBCL. Several studies have shown that exposure to broad-spectrum antibiotics, particularly piperacillin-tazobactam and meropenem, prior to infusion is associated with reduced survival and impaired response.(34,35) These associations are linked to loss of beneficial taxa such as Ruminococcus, Faecalibacterium, and Bacteroides, and alterations in microbial metabolites, including short-chain fatty acids.(36) Prasad et al. demonstrated that such dysbiosis correlates with impaired CAR-T therapy outcomes in both patients and mouse models.(37) While causality remains unproven and confounding by disease severity or inflammation is possible, these findings support further investigation into microbiome-based risk stratification and therapeutic modulation.

Disease-Related Predictors

Tumor Burden:

Baseline disease burden is among the most consistent predictors of CAR-T therapy outcomes. High metabolic tumor volume (MTV), as measured by PET-CT, is associated with inferior CR rates and shorter remissions.(38–40) Tumors with high glycolytic activity or bulky involvement may sequester CAR-T cells, exhaust their cytotoxic capacity, and limit complete disease eradication. Advanced imaging techniques such as radiomics provide additional insight, quantifying lesion shape, texture, and metabolic heterogeneity to predict resistance. For example, irregular lesion morphology and diffuse metabolic activity patterns have been associated with suboptimal response.(41–43) Total lesion glycolysis and maximum standardized uptake value (SUVmax) also independently predict outcomes.(41,42) These findings underscore the rationale for pre-infusion cytoreduction via bridging therapy to improve CAR-T therapy efficacy. Moreover, newer tools such as machine learning models incorporating imaging features, laboratory values, and clinical variables are being developed to improve risk stratification before infusion.

Disease Distribution and Extranodal Sites:

The anatomic spread of disease influences therapeutic response. Extranodal involvement, particularly in sanctuary sites such as the central nervous system, lung, pleura, gastrointestinal tract, or peritoneum, is linked to early relapse.(44,45) These environments may hinder CAR-T cell trafficking or foster local immunosuppression. CNS involvement poses a unique challenge due to limited CAR-T penetration and neurotoxicity risk. Bridging strategies, including localized radiotherapy or intrathecal chemotherapy, are often required to mitigate this risk.(46,47) Even within nodal compartments, multifocal or disseminated disease portends poorer outcomes.(44,45) Patients with limited disease at infusion tend to derive more durable benefit, reinforcing the importance of early referral and disease control prior to CAR-T therapy.(44) Variation in disease genetic and phenotypic features may drive site-specific tropism and alter therapeutic efficacy.(48–52)

Histologic Subtypes:

Histologic subtype strongly influences CAR-T therapy outcomes in LBCL. Patients with PMBCL and transformed DLBCL demonstrate superior responses and survival, as shown in TRANSCEND and real-world studies.(8) For example, Galtier et al. reported 2-year PFS and OS rates of 70% and 87% in PMBCL versus 40% and 49% in DLBCL, attributed to favorable clinical features and a less suppressive tumor microenvironment.(53) Similar findings by Chiappella et al. suggest the possible role of checkpoint blockade during bridging.(54)

In contrast, T cell/histiocyte-rich large B-cell lymphoma are associated with poor outcomes. These subtypes exhibit resistance to CAR-T therapy, potentially due to PD-L1–rich tumor-associated macrophages and exhausted T cells.(55) Pophali et al. and Bastos-Oreiro et al. reported 2-year OS rates below 45%, with frequent early relapse. Notably, long-term survival was observed only in patients treated with checkpoint inhibitors or bispecifics post-CAR-T therapy, suggesting a need for tailored salvage strategies.(56,57)

Genomic and Microenvironmental Barriers:

Next-generation sequencing has significantly advanced our understanding of LBCL biology and refined treatment strategies, although its role in CAR-T therapy is emerging. (58–61)TP53 genomic alteration has a detrimental impact in patients with LBCL treated with CD19 CAR-T therapy.(62) Subsequent studies linked chromosomal deletions, mutational signatures, and other genomic abnormalities to poor CAR-T therapy outcomes, highlighting the potential for genomic alterations to predict treatment resistance.(63–66) However, the phenotypic consequences of these alterations and their interaction with the tumor microenvironment (TME) over time remain largely unexplored in the context of CAR-T therapy. Recent spatial and single-cell transcriptomic profiling studies suggest that resistance may stem not only from tumor-intrinsic alterations but also from the evolution of an immune-excluded or suppressive TME, characterized by increased stromal remodeling, reduced cytotoxic T-cell infiltration, and upregulation of inhibitory ligands.(67,68) These data underscore the importance of integrating genomic profiling with TME context to fully capture the determinants of CAR-T therapy failure and to inform rational combinatorial strategies.

Treatment-Related Predictors

Bridging Therapy:

Bridging therapy (BT) is often used in patients with high tumor burden during CAR-T cell manufacturing and may reflect more aggressive disease biology. While BT itself does not appear to compromise efficacy, evidenced by comparable outcomes in trials allowing (e.g., TRANSCEND) and excluding BT (e.g., ZUMA-1), real-world data suggest outcomes vary by BT type and tumor burden. Achieving MTV reduction through BT is linked to outcomes similar to those with initially low MTV, suggesting cytoreduction may mitigate poor-risk features. Although not yet studied prospectively, deeper responses to BT are likely to predict improved CAR-T therapy outcomes by reducing tumor burden and shaping a more favorable immune environment.(69–71) Thus, BT should be considered not only a temporizing measure but also a potential strategy to enhance CAR-T therapy trefficacy in selected high-risk patients.

Timing and Prior Lines of Therapy:

The cumulative impact of prior therapies influences CAR T-therapy efficacy. Patients treated earlier in their disease course (e.g., second-line) demonstrate better outcomes compared to those with multiple relapses. However, relapse still remains a significant barrier.(72) Early intervention may preserve T-cell fitness and reduce tumor heterogeneity. The TRANSFORM and ZUMA-7 trials confirmed the superiority of CAR-T therapy over salvage chemotherapy in second-line settings.(73,74) Nevertheless, CAR-T therapy remains effective even in third- or later-line use, emphasizing that patient selection should be individualized. Referral to CAR-T therapy evaluation should not be delayed based on the presumption that patients are “not ready” or should exhaust other therapies first. The concept of T-cell exhaustion from cumulative chemotherapy exposure also raises questions about whether leukapheresis should be performed earlier in the treatment journey to preserve cellular quality for later use.

Lymphodepletion:

Lymphodepleting chemotherapy, typically fludarabine and cyclophosphamide, is essential for CAR-T cell engraftment and expansion. It enhances homeostatic cytokine release (e.g., IL-7, IL-15), suppresses regulatory T cells, and creates a permissive immune environment.(75) While standard regimens are widely used, bendamustine-based lymphodepletion is increasingly considered in frail patients, with retrospective studies showing comparable efficacy and potentially lower neurotoxicity. (76,77) Importantly, delays between lymphodepletion and CAR-T infusion, particularly beyond 2–3 days, are associated with inferior outcomes, likely due to cytokine decay and immune reconstitution.(78) A recent pharmacokinetic analysis identified an optimal fludarabine exposure range associated with improved expansion and survival,(79) underscoring the importance of dosing precision. Future strategies may include tailoring lymphodepletion intensity based on patient-specific factors such as lymphocyte counts or cytokine profiles. Operational efficiency and product availability remain critical to ensure timely infusion and therapeutic success.

Product Selection:

Multiple CAR-T products are approved for LBCL, each with distinct constructs. Axi-cel (CD28 co-stimulation) offers rapid expansion and high CR rates but with increased neurotoxicity, while Tisa-cel and Liso-cel (4-1BB co-stimulation) may provide more gradual expansion and favorable safety profiles.(80). Real-world comparisons suggest Axi-cel achieves higher response rates compared to Tisa-cel.(80,81) Whether efficacy outcomes differ between Axi-cel and Liso-cel remains an open question.(82,83) Emerging products incorporating dual antigen targets, suicide switches, or engineered cytokine release profiles aim to further refine efficacy and safety profiles.

CAR T-Cell Expansion and Immunophenotype:

In vivo CAR-T cell expansion is necessary but not sufficient for durable response. Patients achieving high peak expansion (e.g., day 7–10) tend to have initial tumor control, but persistence of functional CAR-T cells is critical for long-term remission. T-cell products enriched for central memory (Tcm) and stem-cell memory (Tscm) phenotypes exhibit superior persistence and antitumor activity. Markers such as CCR7, CD27, CD28, and CD45RA help define these subsets. Conversely, terminal differentiation and exhaustion markers (e.g., PD-1, TIM3, LAG3) correlate with early relapse.(84) Manufacturing conditions, including culture duration and cytokine environment, shape T-cell phenotype. Some centers now explore strategies like PI3K inhibition during culture or monocyte depletion to preserve memory subsets.(85) Recent single-cell analyses reveal that the presence of Helios+ FOXP3+ CAR Tregs and monocyte contamination correlates with non-response. (86)Immune profiling before and after infusion may enable real-time risk stratification and timely intervention. Integration of single-cell RNA-seq, CITE-seq, and TCR clonotype analysis into correlative studies will enhance our understanding of CAR-T cell fate and mechanisms of treatment failure.

Conclusion

CAR-T therapy for LBCL offers curative potential but is limited by primary resistance and relapse in a substantial fraction of patients. Predictors of efficacy span patient-related characteristics (e.g., performance status, inflammation), disease-related factors (e.g., tumor burden, histology, genomics), and treatment elements (e.g., lymphodepletion, bridging, CAR-T product). Many of these variables are modifiable or measurable, opening avenues for personalization. For example, patients with high tumor burden may benefit from radiotherapy bridging; those with inflamed cytokine profiles could receive targeted anti-inflammatory therapies; and patients with adverse genomic features may warrant consolidation or combination regimens. Integrating predictive biomarkers into clinical workflows will improve patient selection, optimize outcomes, and reduce toxicity. Future research should prioritize prospective validation of predictive signatures, real-time immune monitoring, and trials of rational combinations. As the field evolves, a precision CAR-T approach, matching patient, tumor, and product, will be essential to maximize the therapeutic potential of this transformative modality.

Key Points.

• Response to CD19 CAR-T therapy in LBCL is influenced by patient-, disease-, and treatment-related factors.

• Patient factors such as inflammation, comorbidity burden, and microbiome disruption influence CAR-T efficacy and toxicity.

• Tumor characteristics including high metabolic tumor volume, extranodal involvement, and TP53 alterations are linked to inferior outcomes.

• Treatment-related variables, from bridging and lymphodepletion to CAR-T product, are strongly associated with disease control after CAR-T therapy.