Immune effector cell-associated hematotoxicity after CAR T-cell therapy: from mechanism to management

Abstract

Genetically engineered chimeric antigen receptor (CAR) T-cells have become an effective treatment option for several advanced B-cell malignancies. Hematologic side effects, recently classified as immune effector cell-associated hematotoxicity (ICAHT), are very common and can predispose for clinically relevant infections. As hematopoietic reconstitution after CAR-T differs from chemotherapy-associated myelosuppression, a novel classification system for early and late ICAHT has been introduced. Furthermore, a risk-stratification score named CAR-HEMATOTOX has been developed to identify high-risk candidates, thereby enabling risk-based interventional strategies. Therapeutically, growth factor support with G-CSF represents the mainstay of treatment, with hematopoietic stem cell boosts available in G-CSF refractory cases. While the underlying pathophysiology remains enigmatic, recent translational studies suggest that CAR-T induced inflammation and baseline hematopoietic function are key contributors to prolonged cytopenia. In this review, we provide an overview of the spectrum of hematologic toxicities after CAR T-cell therapy and offer perspectives on future translational and clinical developments.

Introduction

Cellular immunotherapies using genetically engineered CAR T-cells that target B cell antigens such as CD19 or BCMA have rapidly altered the treatment landscape of several lymphoid malignancies. This has lead to the FDA approval of six CAR-T products across a spectrum of hematologic disease indications, with many more in the developmental pipeline. Furthermore, CAR T-cell platforms are being actively explored for several solid tumors and autoimmune diseases.^1,2 The profound systemic immune response elicited by CAR T-cells upon target antigen recognition and subsequent expansion can result in a unique toxicity profile. While much attention has been paid to cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) as prototypical side effects with distinct management protocols,³ immune effector cell-associated hematotoxicity (ICAHT) represents the most common CAR-T related adverse event across clinical trials and the real-world setting.^4–6 Importantly, hematotoxicity is observed irrespective of the applied CAR-T product, target antigen, and disease entity.⁷

While it is tempting to attribute hematotoxicity as a mere consequence of the myelotoxic lymphodepleting chemotherapy applied prior to CAR-T infusion (mainly fludarabine and cyclophosphamide), cytopenias are often long-lasting and delayed in nature. They are typically characterized by an archetypal biphasic temporal course with “intermittent” recovery followed by a second dip.^8,9 In a smaller proportion of patients, severe cases of bone marrow aplasia have been described (“aplastic” phenotype).^9–12 Notably, prolonged cytopenias have been described to last months to several years after CAR T-cell infusion.¹³ Taken together, these clinical observations strongly suggest a CAR T-cell induced mechanism of myelosuppression, though the underlying pathophysiology remains enigmatic and incompletely understood.

The clinical relevance of hematotoxicity lies in the hemorrhagic diathesis and increased risk of infectious complications. Importantly, both neutropenia and lymphopenia predispose for bacterial, fungal and viral infections.¹⁴ Infection risk is further compounded by B-cell aplasia and hypogammaglobulinemia as expected on-target/off-tumor toxicities of B-cell targeting CAR T-cell therapies. Indeed, life-threatening infectious complications drive non-relapse mortality (NRM) after CAR T-cell therapy across diverse treatment settings.¹⁵ Moreover, transfusion dependency significantly contributes to therapy-related morbidity, prolonging hospital stays and increasing the utilization of health care resources.¹⁶ Overall, there remains marked heterogeneity in regard to the reporting of cytopenias and concerning the standard diagnostic work-up and management.¹⁷ For this reason, recent efforts by the European Hematology Association (EHA) and European Society of Bone Marrow Transplantation (EBMT) classified ICAHT as a distinct toxicity category of cell therapy and issued a consensus grading framework, thereby setting the stage for severity-based treatment strategies.

In this review, we share perspectives on the range of encountered hematologic side effects of CAR T-cell therapy. Specifically, we provide an overview on the expected incidence rates of early and late ICAHT across a range of lymphoid malignancies and plasma cell dyscrasias. We shed light on current classification systems and outline their potential advantages and pitfalls. Next, we describe what is currently known about clinical risk fators and potential pathomechanisms, with a particular focus on mechanistic differences based on the different patterns of post CAR-T neutrophil recovery. Finally, we discuss the current evidence base for therapeutic options such as G-CSF, thrombopoietin agonists and hematopoietic stem cell boosts (HSCB). Our overarching goal is to inform CAR-T practitioners on this clinically relevant side effect of cell therapy, to put forth a framework for future translational efforts, and to provide suggestions to improve management of cytopenias in this challenging cohort of patients.

Hematologic complications of CAR-T therapy

When approaching hematologic side effects of CAR-T therapies, one can broadly separate three distinct phases: prior to CAR-T infusion, early ICAHT (day 0–30) and late ICAHT (after day +30, Figure 1). The pre-therapeutic phase is characterized by the unique patient history, the number of prior cytotoxic treatment lines and particularly the commonly applied holding or bridging therapies immediately preceding CAR T-cell infusion.^18,19 For example, chemotherapy-based bridging can result in baseline cytopenias, which can reflect impaired hematopoietic reserve.²⁰ Other relevant baseline risk factors of hematotoxicity relate to the degree of systemic inflammation (e.g. elevations of serum C-reactive protein [CRP] or ferritin) and the presence of underlying bone marrow (BM) infiltration. To risk-stratify patients for developing cytopenias and associated infections prior to lymphodepletion, the CAR-HEMATOTOX score was established in a multicenter cohort of large B-cell lymphoma (LBCL) patients and subsequently validated for use in patients with mantle cell lymphoma (MCL) and multiple myeloma (MM).^9–11,21,22 Of interest, the score also appears to be useful to identify patients at high risk for disease progression and prolonged hospitalization.

Figure 1. Timeline of hematologic toxicities of CAR T-cell therapy.

Top: Overview of the most important risk factors, clinical considerations and complications of hematotoxicity across the three critical phases in relation to CAR-T infusion. Bottom, left: Individualized risk assessment with the CAR-HEMATOTOX score, which is assessed prior to lymphodepletion and separates patients into a low (score 0–1) versus high (score ≥2) risk for severe hematoxicicty. Bottom, right: Summary of management strategies to treat cytopenias after CAR-T therapy.

The applied lymphodepleting chemotherapy (typically fludarabine [range 25–30 mg/m²] and cyclophosphamide [range 250–500 mg/m²]) facilitates an expected early nadir phase that can extend until day +10. During this early phase, a delay in count recovery can be aggravated by high-grade CRS and the associated cytokine patterns, especially IL-6 and IFN-γ.^23,24 Three typical trajectories of early neutrophil recovery have emerged. First, “quick” recovery, refers to transient and self-resolving cytopenia due to the applied lymphodepletion. Second, “intermittent” recovery, describes the commonly observed biphasic pattern with count recovery followed by a second or multiple dips. Third, the clinically challenging “aplastic” phenotype is characterized by marked BM aplasia that is often treatment-refractory and translates into a high risk for infections and NRM.

The final phase of hematotoxicity is observed after day +30 and describes prolonged and/or late cytopenias.⁵ To understand the clinical relevance of these later cytopenias, the presence or absence of antecedent count recovery must be considered (e.g. “intermittent” vs. “aplastic” patients).²⁴ Patients with sustained neutropenia lasting into the second month after CAR-T infusion (“aplastic phenotype”) present with a very high risk for severe and even fatal infections.^12,14 On the other hand, recurrent cytopenias (“intermittent phenotype”) can result in repeated hospitalizations due to infectious events or transfusion needs, though it should also be noted that these patients exhibit excellent survival outcomes and a watch-and-wait approach to management can thus be reasonable.²⁴ Nonetheless, persistently low counts can prevent potentially efficacious post-relapse therapies as cytopenias represent a common study exclusion criterion. Finally, any new-onset or unexplained cytopenias must raise concern for secondary malignancies, particularly treatment-emergent myeloid neoplasms, and should thus prompt a BM examination.²⁵

Expected incidence rate of ICAHT in clinical trials and the real-world setting

Several studies have reported on the incidence rates and the quality of hematologic side effects of CAR T-cell therapy (summarized in Table 1). In general, the expected rate of grade 3 or 4 cytopenias is very high, ranging between 28–65% (severe thrombocytopenia), 16–77% (severe anemia) and 59–95% (severe neutropenia), respectively. Notably, significant heterogeneity was noted for the definition of prolonged cytopenias (e.g. day 21 vs. day 30 vs. day 90), highlighting differences in the reporting of late cytopenias.¹⁷ In terms of disease entity, the ‘aplastic’ phenotype was more commonly encountered in lymphoma patients (e.g., LBCL, MCL) compared to myeloma patients.^12–14 Furthermore, the majority of studies outlined a preponderance towards increased hematotoxicity in patients receiving CAR-T products harboring the CD28z endodomain compared to 4–1BB.⁷ This observation was corroborated by a matched-paired comparison comparing axi-cel vs. tisa-cel from the DESCAR-T registry.²⁶ Still, any direct comparison of the incidence of hematotoxicity across clinical trials evaluating CAR-T products is complicated by substantial heterogeneity in disease entity (BCP-ALL, B-NHL, MM), CAR-T product (CD28z vs. 4–1BB) and cytopenia reporting (particularly for late cytopenias). A recent meta-analysis indicated an increased rate of high-grade cytopenias in patients with BCP-ALL, likely as a result of BM infiltration by leukemic blasts and the extensive prior treatment lines.²⁷ When applying the new EHA/EBMT consensus grading (Table 2), the incidence of severe or life-threatening early ICAHT was highest in MCL, followed by LBCL and MM (28% vs. 23% vs. 15%), though the multivariable model suggested that the disease-specific differences more fundamentally reflect variability in underlying patient features (e.g., disease burden, inflammation).²⁸

Table 1:

Overview of studies examining hematologic toxicities of CAR T-cell therapies

	Fried et al (2019)8	Jain et al (2020)66	Logue et al. (2021)6	Rejeski et al (2021)9	Juluri et al (2022)23	Penack et al (2023)75	Rejeski et al (2023a)11	Rejeski et al (2023b)10
Number of study patients	35	83	85	258	173	398	113	103
Median age (range)	27 (3.5–55)	58 (19–85)	64 (28–79)	63 (19–83)	55 (20–76)	61 (18–81)	65 (39–81)	66 (49–89)
Hematological Disease	ALL (19), B-NHL (16)	LBCL (40), BCP-ALL (37), MM (6)	LBCL	LBCL	BCP-ALL (62), CLL (48), B-NHL (63)	LBCL	MM	MCL
Median previous lines of chemotherapy	-	4 (range 1–9)	3 (range 1–8)	3 (range 2–11)	4 (range 1–11)	3 (range 1–8)	6 (95% CI 5–6)	3 (IQR 2–4)
Previous SCT:
Autologous	37%*	18%	25%	27%	13%	24%	88%	32%
Allogeneic		19%	2.4%	-	20%	3.6%	3.5%	3%
Underlying BM Involvement	-	-	-	14%	-	-	46%	37%
Serum LDH prior to CAR-T (U/L)	-	-	-	Median: 276 (95% CI 260–302)	-	-	Median: 217 (95% CI 202–242)	Median: 208 (95% CI 194–218)
Platelet Count prior to CAR-T (G/L)	-	-	-	Median: 164 (95% CI 152–178)	Median: 123 (IQR 64–196)	-	Median: 144 (95% CI 125–158)	Median: 164 (95% CI 152–178)
Serum Ferritin prior to CAR-T (ng/mL)	-	-	-	Median: 501 (95% CI 378–647)	-	-	Median: 211 (95% CI 120–354)	Median: 243 (95% CI 184–282)
Applied CAR Product	Local product	Axi-cel (30), Tisa-cel (10), Local product (43)	Axi-cel	Axi-cel (170), Tisa-cel (88)	Local product (CD28z + 4–1BB)	Axi-cel (245), Tisa-cel (153)	Ide-cel (106), Cilta-cel (7)	Brexu-cel
Target Antigen	CD19	CD19 / BCMA	CD19	CD19	CD19	CD19	BCMA	CD19
Grade 3 or higher CRS	14%	23%	9%	11%	15%	N/R	5%	6%
Any grade ≥3 Thrombocytopenia	28%	65%	N/R	62%	44%	Any grade 3 or 4 cytopenia: 100%	50%	57%
Any grade ≥3 Anemia	55%	77%	N/R	69%	16%		50%	51%
Any grade ≥3 Neutropenia	72%	95%	N/R	91%	59%		73%	88%
Prolonged Neutropenia	d+42: 62%	d+30/90 67%/20%	d+30/90: 30%/12.5%	d+21: 64%	d+28: 46%	d+30/90 cytopenia: 9%/12%	d+21: 50%	d+21: 47%
Prolonged Thrombocytopenia	d+42: 44%	d+30/90 49%/10%	d+30/90: 26%/5.4%	N/R	d+28: 34%		d+31–100: 28%	d+31–100: 35%
Phenotypes of Neutrophil Recovery	-	-	-	Q: 25% I: 52% A: 23%	-	-	Q: 40% I: 44% A: 16%	Q: 41% I: 38% A: 21%
Median follow-up time after CAR-T	N/R	6 mo	12.8 mo	N/R	40.8 mo	N/R	7.9 mo	15.4 mo
Association with infectious events	N/R	N/R	Infections linked to high-grade CRS/ICANS and steroid use; association with neutropenia duration was not studied	N/R	N/R	Infections reported both in the patients with and without severe cytopenias	Increased infection rate in patients with severe hematotoxicity, particularly bacterial infections	Increased infection rate in patients with severe hematotoxicity, particularly bacterial infections
Additional Comments	Biphasic pattern of cytopenia; late cytopenia linked to high-grade CRS, prior HSCT and SDF-1 alterations	Association between high-grade CRS/ICANS and markers of acute inflammation with delayed count recovery	Study also explores long-term immune reconstitution patterns following axi-cel	Development of CAR-HEMATOTOX (HT) score with independent validation; high scores associated with adverse treatment outcomes	Association of hematotoxicity with high-grade CRS and CRS-related inflammatory patterns (especially IL-6 elevations)	Severe cytopenia linked to the number of prior treatment lines; reduced PFS in patients with severe hematotoxicity	Validation of CAR-HEMATOTOX score, including for infections, NRM and survival	Validation of CAR-HEMATOTOX score, including for infections, NRM and survival

Data are depicted as n (%) unless otherwise specificed.

ALL, acute lymphoblastic leukemia; B-NHL, B Non Hodgkin-Lymphoma; LBCL, large B-cell lympoma; CLL, chronic lymphocytic leukemia; MM, multiple myeloma; MCL, mantle cell lymphoma; SCT, stem cell transplantation; Axi-cel, axicabtagene ciloleucel; Tisa-cel, tisagenlecleucel; Liso-cel; lisocabtagene maraleucel; ide-cel, idecabtagene vicleucel; Cilta-cel, ciltacabtagene autoleucel; Brexu-cel, brexucabtagene autoleucel; CD, cluster of differentiation; mo, months

Table 2:

Overview of hematotoxicity grading systems.

Classification Systems		Grade 1	Grade 2	Grade 3	Grade 4
CTCAE	Neutropenia	ANC <LLN – 1500/μL	ANC <1500–1000/μL	ANC <1000–500/μL	ANC <500/μL
	Anemia	Hgb <LLN – 10 g/dL	Hgb <10.0 – 8.0 g/dL	Hgb <8.0 g/dL; transfusion	life-threatening intervention
	Thrombo-cytopenia	Platelet Count: <LLN – 75 G/L	<75 – 50 G/L	<50 – 25 G/L	<25 G/L
ICAHT Grading	Early (day 0–30)	ANC <500/μL* for 1–6 days	ANC <500/μL for 7–13 days	ANC <500/μL for ≥14 days ANC <100/μL** for ≥7 days***	Never above ANC 500/μL ANC <100/μL for ≥14 days
	Late (after day +30)	ANC <1500/μL	ANC <1000/μL	ANC <500/μL	ANC <100/μL
Phenotypes of Neutrophil Recovery		Quick Recovery: sustained neutrophil recovery without a second dip below an ANC <1000/μL. Intermittent Recovery: neutrophil recovery (ANC >1500/μL) followed by a second dip below an ANC <1000/μL. Aplastic Recovery: continuous severe neutropenia (ANC <500/μL) for ≥14 days.

Based on ASCO/IDSA consensus grading of cancer-related infection risk for *severe neutropenia (ANC <500/μL), **profound neutropenia (ANC <100/μL), ***protracted neutropenia (≥7 days).

Abbreviations: ANC = Absolute Neutrophil Count; Hgb = Hemoglobin; LLN = lower limit of normal.

Defining Immune Effector Cell-Associated Hematotoxicity

The initial clinical trials exploring CAR T-cell therapies primarily attributed cytopenias according to the common terminology criteria for adverse events (CTCAE) (Table 2, top). However, such a purely quantitative grading system fails to capture the unique quality of post CAR-T hematotoxicity. More importantly, the CTCAE criteria do not reflect the risk of infections due to neutropenia, which is based not only on the depth but also the duration of severe neutropenia (e.g. protracted neutropenia lasting longer than 7 days).²⁹ To account for this, an expert panel from EHA and EBMT recently developed a new grading system for ICAHT, which separates early (day 0–30) versus late (after day +30) ICAHT. Early ICAHT assesses the duration of continuous severe (ANC <500/μL) or profound (ANC <100/μL) neutropenia and thereby closely mirrors the ASCO/IDSA guidelines for cancer-related infection risk. The grading of early ICAHT follows the severity categories of mild, moderate, severe, and life-threatening – similar to the broadly implemented ASTCT grading sytems for CRS and ICANS. Importantly, a recent study showed that the early ICAHT grading closely reciprocates the clinically relevant phenotypes of neutrophil recovery.²⁸ Concomitantly, patients with severe or life-threatening ICAHT very frequently displayed ‘aplastic’ neutrophil recovery, consistent with profound BM aplasia in this small subset of patients. While ICAHT severity was linked to clinically meaningful endpoints such as infections, NRM, transfusion use, duration of hospitalization, and adverse treatment outcomes, the utility of the grading system still needs to be prospectively evaluated. Nonetheless, a standardized grading system has specific advantages such as enabling comparability across disease entities, CAR-T products, and treatment settings.

Pathophysiology of hematotoxicity after CAR T-cell therapy

A range of clinical risk factors contribute to the development of cytopenias after CAR-T, which can be broadly separated into host-, disease, and treatment-related factors (appendix page 1). These factors provide crucial context for understanding the underlying (patho-)physiology of hematotoxicity. The heterogeneity of clinical variables associated with prolonged cytopenias also sheds light on the fact that hematotoxicity is unlikely to be mediated by any one factor alone. Instead, a variety of features relating to hematopoietic stem cell reserve, the BM microenvironment, systemic inflammatory mediators and (CAR) T-cell expansion characteristics likely act together either in concert or independently (e.g. multifactorial origin; Figure 2).

Figure 2.

Potential pathomechanisms of ICAHT

1). Role of the hematopoietic stem cell reserve

Hematopoietic stem and progenitor cells (HSPCs) reside within a specialized niche in the bone marrow that is surrounded by endothelial and mesenchymal stromal cells, where they serve as precursors to a wide array of cells of the innate and adaptive immune systems.³⁰ The regenerative capacity of HSCs and their ability to respond to external stimuli in the CAR-T patient is dependent on a multitude of factors including the cumulative cytotoxic stress conferred by prior genotoxic chemotherapies (especially lenalidomide or alkylating agents like melphalan),^31,32 the process of natural aging,³³ and direct or indirect interactions between the underlying disease and HSPCs.³⁴ The acquisition of somatic mutations due to these factors can facilitate the development of age-related clonal hematopoiesis and clonal hematopoiesis of indeterminate potential (CHiP), which is defined by the manifestation of cancer-related somatic driver mutations with a variant allele frequency (VAF) of greater than 2% in peripheral blood. The prevalence of CHiP is inherently age-dependent, with an expected rate of 10–20% in individuals aged 70 or older.³⁵ However, the prevalence is higher in lymphoma patients (~30% prior to auto-HCT) and has been associated with adverse treatment outcomes.³⁶ In the CAR-T population, several recent studies indicate a prevalence between 34 to 56%.^37–41 It is critical to understand that HSCs not only react to infections and inflammatory stimuli but also serve as the foundation of the subsequent host immune response by replenishing certain immune cell populations.⁴² Accordingly, the presence of CH clones may potentiate the host inflammatory response to CAR T-cells. Furthermore, recent evidence suggests that CH clones may be gradually selected for because they are more resistant to the deleterious impact of inflammation and aging.⁴³ In line with this observation, clonal expansion of CH clones has been observed following CAR-T, with a trend towards more pronounced post-CAR-T cytopenias in these patients.^37,44 This would indicate context-dependent selection of pre-existing CH clones following CAR T-cells, which may be accelerated with certain genotypes (e.g., TP53).⁴⁵

2). Role of the BM microenvironment

The BM microenvironment is orchestrated by the complex interplay of cells and factors that regulate hematopoiesis including mesenchymal stem cells, a vascular niche formed by endothelial cells and perivascular stromal cells, as well as adipocytes and bone lineage cells which contribute to the microenvironment’s metabolic and structural dynamics.³⁰ Soluble factors like cytokines and growth factors mediate critical communication and regulatory pathways within this niche.⁴⁶ Kitamura and colleagues reported that the BM niche is severely disrupted in CAR-T patients with prolonged cytopenias, identifying an impairment of CD271⁺ stromal cells using 3-dimensional imaging analyses from BM biopsy specimens.⁴⁷ Furthermore, the authors found that CXC chemokine ligand 12 (CXCL12) and stem cell factor (SCF), both niche factors essential for hematopoietic recovery, were significantly decreased in the BM of patients with prolonged cytopenia, indicating reduced niche cell function. The presence of underlying BM infiltration (e.g. extranodal involvement of the lymphoma) likely disrupts the intricate balance within the niche. Indeed, BM involvement represents one of the strongest independent predictors of severe post-CAR-T hematotoxicity across several disease entities.^10,11,24 One potential explanation for this observation lies in the transmigration of CAR T-cells to target cells within the BM, resulting in local hyperinflammation and the release of cytokines and growth factors in close vicinity to hematopoietic progenitor cells. Even in the absence of lymphoma cells in the BM, interactions between CAR T-cells and endogenous CD19- or BCMA-positive B-cell precursor populations (e.g., on-target/off-tumor toxicity) may contribute to local inflammatory processes and microenvironmental alterations that subsequently result in prolonged cytopenias.

3). Role of systemic inflammatory mediators

While inflammation-induced activation of HSCs and cytokines like IFN-γ can cause HSCs to lose quiescence and proliferate in the short-term, chronic exposure can lead to functional impairment and depletion of these cells.⁴⁸ Specifically, IFN-γ has been demonstrated to reduce stem cell cycling and plays a key regulatory role for the proliferation and differentiation of human HSPCs.^49,50 Chronic inflammation is particularly deleterious, causing long-term changes to the BM microenvironment, promoting aging-related changes, and potentially leading to BM failure.⁴⁸

In the context of CAR-T, severe CRS and several inflammatory markers have been implicated in the development of severe hematotoxicity. Juluri et al found that higher peak IL-6 serum concentrations were associated with slower hematopoietic recovery,²³ which have also been observed locally within the BM niche.⁴⁷ The authors also described higher serum TGF-β levels in patients with improved hematopoietic recovery, a pleiotropic cytokine which can mediate the proliferation of myeloid-producing HSCs.⁵¹ Focusing on CAR-T patients with aplastic neutrophil recovery, serum proteomic studies revealed a signature displaying hallmarks of immune dysregulation and macrophage activation (e.g. elevations of IL-15, IL-18 and MCP-1), endothelial dysfunction (e.g. increasing Angiopoietin-2-to-1 ratio), and T-cell suppression (e.g. upregulation of soluble T-cell checkpoint ligands).²⁴ Together with increased IFN-γ and serum ferritin in ‘aplastic’ patients, this study indicated some mechanistic overlap with immune effector cell-associated hemophagocytic lymphohistiocytosis-like syndrome (IEC-HS), which also frequently presents with pancytopenia and represents a well-characterized side effect of CAR-T.⁵² Importantly, many of the perturbations of these systemic inflammatory mediators were already present prior to the application of CAR T-cells, underlining the importance of pre-existing inflammation for the subsequent development of severe hematotoxicity. Patients with impaired hematopoietic function at baseline may be at particular risk for inflammation-mediated myelosuppression induced by the infusion of CAR T-cells.⁵³

The role of CAR-T expansion in driving cytopenias is not fully resolved and may be dependent on the pattern of cytopenia. Interestingly, patients with biphasic neutrophil recovery (e.g., recurrent neutrophil dips) displayed markedly higher CAR T-cell expansion and persistence compared to patients with ‘aplastic’ recovery. Intermittent cytopenia may thus reflect extravasation of immune cells including CAR T-cells into the periphery, BM, and to the lymphomatous tissue. On the other hand, the type of immune dysregulation that is both inflammatory (e.g, high IFN-γ, IL-18) and T-cell suppressive (e.g, increased soluble T-cell checkpoint ligands) may provide a potential explanation for the paradoxical finding of lower CAR T-cell expansion in the ‘aplastic’ patients that show pronounced myelosuppression. These results suggest that CAR T-cell expansion is not the sole driver of cytopenias, but rather that CAR T-cell expansion ignites pre-existing inflammation, thereby inducing an injury to the marrow.

4). Role of clonal T cell expansion phenomena and T/B cell imbalances

An early correlative study of hematologic toxicity following CD19 CAR-T by Fried et al showed perturbations of SDF-1 levels in cases of late neutropenia. This chemokine is essential for B-cell development and the trafficking of neutrophils and HSCs, and has been previously implicated in cases of late onset neutropenia after B-cell depleting treatment with Rituximab.⁵⁴ The authors postulate that early recovery of B-cells after CD19 CAR-T may lead to alterations of SDF-1 levels in the marrow microenvironment with concomitantly reduced neutrophil egress from the bone marrow. Furthermore, the association between B-cell depleting therapies and neutropenia has been linked to the clonal expansion of T-cells – most likely due to T-cell imbalances facilitated by diminished T-/B-cell interactions.⁵⁵ In line with this, detailed single-cell RNA and T-cell receptor (TCR) sequencing from a patient with acquired BM failure following CAR-T showed marked oligoclonal (CAR) T-cell expansion, particularly of a CD8⁺CD57⁺ T-cell population. This was accompanied with a shift from multiclonal to oligoclonal TCR usage, with clonality levels rivalling those of a reference T-LGLL patient population.⁵⁶ Similar clonal expansion phenomena were noted by Strati and colleagues, who observed a significant increase in the frequency of clonally expanded CXCR1^high cytotoxic effector T-cells in CAR-T patients with prolonged cytopenia. These expanded T_EFF cell subsets expressed high IFN-γ and cytokine signaling gene sets, while corresponding HSC populations in the same patients expressed IFN-γ response signatures.⁵⁷

Management

Identifying patients with a high-risk profile for severe ICAHT

The CAR-HEMATOTOX score was developed to enable early risk-stratification of CAR-T patients into a high versus low risk of developing severe hematotoxicity.⁹ The score is calculated prior to the initiation of lymphodepletion and incorporates the complete blood count (e.g., ANC, hemoglobin, platelet count) and two serum inflammatory markers (e.g., CRP, ferritin). Patients deemed high-risk (score ≥2) displayed an increased rate of severe and prolonged neutropenia, as well as severe thrombocytopenia and anemia compared to low-risk patients (score 0–1). Next to risk-stratification for severe ICAHT, high CAR-HEMATOTOX scores have also been linked to severe infections, increased NRM and poor treatment outcomes, indicating broad utility of the score.²¹ Furthermore, the score was validated for use in patients treated with BCMA-directed CAR-T in multiple myeloma and treated with brexucabtagene autoleucel for mantle cell lymphoma.^10,11 While the individual score components also appear to be relevant for adult and pediatric B-ALL patients, a high number of patients are classified as high-risk based on a score threshold of 2 and it is likely that further refinements of the score for this disease entity are required.⁵⁸ It is important to be cognizant of the limitations of the score including its low positive predictive value (i.e., better at ruling out than in) and the fact that it remains to be determined if the score is also predictive at earlier time points (i.e., prior to leukapheresis) which would enable prophylactic collection of autologous CD34+ stem cells as a potential rescue strategy in very high-risk patients without underlying BM infiltration.

Diagnostic Algorithm

In case of severe cytopenia prior to lymphodepletion, the presence of underlying BM involvement should be strongly considered, and confirmed with histopathologic studies. Knowledge of the extent of BM infiltration as a highly relevant risk factor can help with the interpretation of subsequent cytopenia trajectories and guide therapeutic strategies. Assessing the presence of pre-existing CHiP clones with next-generation sequencing (NGS) does not currently represent a standard-of-care. However, it can be prudent to cryopreserve the BM aspirate or peripheral blood mononuclear cells (PBMCs) to enable testing for such clones in case a patient develops secondary BM failure after CAR-T and, more generally, to contribute to our growing understanding of the pro-inflammatory role of CHiP in CAR-T patients.

Since cytopenias are to be expected in the first week after CAR-T, we recommend initiating more comprehensive diagnostic studies in patients with persisting severe neutropenia beyond day +10.^7,59 A first step should include ruling out other pertinent causes causes of neutropenia like drug-induced myelosuppression (e.g., co-trimoxazole and other antibiotics), vitamin deficiencies, and coincident infections (e.g., viral infections, sepsis).⁶⁰ In patients with rapid elevations of serum ferritin levels, IEC-HS should be considered as an important differential diagnosis.⁵² A more advanced work-up should be initiated in patients with severe or life-threatening early ICAHT, refractory to G-CSF support (e.g., no count recovery despite at least 5 days of G-CSF). This should incorporate extended viral studies and BM studies to rule out persistent infiltration (e.g., progressive disease) and evaluate for signs of hemophagocytosis or myelodysplasia which can emerge rapidly after CAR-T infusion.²⁵ However, the typical finding is a hypocellular marrow without dysplastic changes.⁶ Because treatment-emerging myeloid neoplasms are a diagnostic concern after CAR-T therapy, in-depth cytogenetic studies and NGS with a myeloid panel should be performed in case of any new-onset or unexplained cytopenia, or non-resolving ICAHT beyond day 30.

Therapeutic Strategies

When CAR-T therapies first entered the clinical routine, there was a reluctance to apply growth factors for the management of cytopenias as preclinical studies had suggested that the use of GM-CSF may promote inflammatory toxicity and induce neuroinflammation.⁶¹ Because of this, the application of G-CSF was commonly deferred until acute CAR-T immunotoxicities had abated (typically week 3). However, several real-world studies have since demonstrated that G-CSF may be applied as early as the first week or even prophylactically, with no significant increase in grade 3 or higher CRS or ICANS.^62–65 For example, Lievin et al showed that early G-CSF administration (starting day +2) in neutropenic patients was associated with a reduction in the rate of febrile neutropenia, with no negative impact on CAR T-cell expansion or clinical outcomes. A further retrospective study of 197 patients by Miller and colleagues examined the impact of prophylactic G-CSF, with the majority of patients receiving pegylated G-CSF prior to CAR-T. While there was a slight increase in the rate of grade 2 (but not grade 3) CRS, prophylactic G-CSF was associated with faster neutrophil recovery and a trend towards shorter intravenous antibiotic exposure.⁶⁴ Furthermore, the authors showed that the initiation of G-CSF in patients with grade 1 CRS did not exacerbate CRS severity. Nonetheless, these studies were not prospective and more research will be needed to further confirm the safety of early G-CSF and identify the optimal treatment protocol for each disease entity (early vs. prophylactic vs. ANC-triggered / non-pegylated vs. pegylated). Scores like the CAR-HEMATOTOX may be useful to guide early G-CSF and anti-infective strategies and thereby restrict these interventions to the patients who are most likely to benefit (e.g. high-risk, score ≥2).²¹

It is important to recognize that the majority of CAR-T patients will ultimately either spontaneously recover their counts or display prompt count improvement with G-CSF.^6,66 However, a minority of patients do not respond to G-CSF (<20%) and these patients can be clinically challenging due to their high risk for life-threatening infections. If cryopreserved CD34⁺ stem cells are available from a previous autologous or allogeneic hematopoietic cell transplantation (HCT), the use of a stem cell boost should be the preferred rescue strategy based on the encouraging rates of engraftment.^67–69 At the same time, a recent EHA/EBMT survey shed light on the fact that HSCB were often not available even when they were considered as a therapeutic avenue.¹⁷ Patients with multiple myeloma may be a notable exception as some younger patients may have collected additional cells for a potential second consolidative transplant, as was recently demonstrated by Mohan and colleagues.⁷⁰ Prophylactic stem cell collection in high-risk CAR-T candidates has been performed successfully in individual cases,⁶⁸ but can be associated with additional costs and logistic burden and should not delay the application of CAR T-cells.⁷¹ Other options for G-CSF refractory cases include thrombopoietin (TPO) receptor agonists like eltrombopag or romiplostim, though evidence is limited and it remains unclear if their use is superior to a watch-and-wait approach.^72,73 Both TPO agonists and IFN-neutralizing antibodies like emapalumab would target the aberrant IFN signalling outlined above.⁵⁷ In cases of grade 3 or 4 ICAHT with a clear inflammatory stressor like IEC-HS and persistently increased inflammatory markers, a trial using anti-inflammatory agents like pulse-dose corticosteroids and/or anti-cytokine therapies (e.g., siltuximab, anakinra) can be attempted. Should all the above measures prove to be futile and grade 4 ICAHT persists (<5% of cases²⁸), allogeneic HCT can be offered as a last salvage option. However, this will invariably result in the eradication of CAR T-cells and this decision should carefully weigh several factors: donor suitability/availability, patient’s goals of care, the possibility of spontaneuous count recovery, risk of fatal infections, and likelihood of disease recurrence.⁵⁹ When pursuing observation, optimizing supportive strategies (e.g., avoiding sick contacts, prophylactic anti-infectives, IVIG) and infectious disease consultation is recommended.

Conclusions and future perspectives

The last years have seen increasing recognition of ICAHT as a distinct and clinically relevant side effect of cell therapy. By clearly defining hematotoxicity, the EHA/EBMT consensus grading system provides a framework for severity-based best practice recommendations, similar to what already exists for CRS and ICANS.³ Moreover, the grading provides clear criteria for the reporting of ICAHT, thus enabling standardized comparisons across disease entities and CAR-T products. Yet, several unresolved clinical and translational research questions still remain (panel 1).

Panel 1. Future clinical and translational research questions regarding hematotologic complications of CAR T-cell therapy.

Clinical research questions:

• How does prior treatment before immune effector cell therapy shape the risk of developing ICAHT? Is there a reduction of incidence rates when moving CAR-T into earlier treatment lines?

• What is the contribution of lymphodepletion for the development of severe hematotoxicity? Does bendamustine-based lymphodepletion reduce cytopenia incidence?

• Does the presence of clonal hematopoiesis prior to CAR T-cell therapy impact the subsequent development of cytopenias?

• Can the predictive capacity of the CAR-HEMATOTOX score be validated in a prospective manner and is the score helpful in guiding G-CSF and anti-infective therapy?

• What strategies can be used to better identify patients that will develop treatment-refractory BM aplasia? What is the role of pro-inflammatory serum biomarkers like IL-6 or IFN-γ?

• Are there criteria that may guide the decision to prophylactically collect stem cells in certain high-risk candidates as a rescue strategy in case of severe hematoxicity?

• What is the optimal timepoint to initiate growth factor support? Is there an advantage to applying pegylated vs. non-peglyated G-CSF?

• Does early and/or prophylactic G-CSF reduce antibiotic exposure or the rate of severe infections?

• What is the incidence of ICAHT with other immune effector cell therapies, such as tumor-infiltrating lymphocytes (TILs) or bispecific antibodies, and are the underlying risk factors similar?

• What is the relationship between ICAHT and the development second primary malignancies, particularly treatment-emergent myeloid neoplasms?

Translational research questions:

• Can a syngeneic mouse model be generated that reciprocates the unique qualities of CAR-T related cytopenia?

• What is the role of CAR T-cells in driving the expansion of clonal hematopoiesis clones into overt myeloid malignancy? Are CHiP clones more susceptible to CAR-T mediated inflammation?

• How do CRS and inflammatory patterns specifically influence hematopoietic function?

• How do endogenous B-cell populations and their early recovery contribute to long-term cytopenias? Do CAR T-cells localize to the hematopoietic niche in the bone marrow?

• What precise mechanisms underlie the superior treatment outcomes in patients with ‘intermittent’ neutrophil recovery?

This review has focused on CAR T-cells, however, hematological toxicities are also among the most common side effects of bispecific antibody therapies.⁷⁴ Future studies may evaluate the qualitative features of cytopenia with bispecifics and study if the same risk factors apply. We anticipate that large multicenter studies will be needed to elucidate the impact of specific prior therapies like bendamustine, examine the influence of different lymphodepletion regimens, and establish whether transitioning CAR-T to earlier treatment lines mitigates the risk of severe ICAHT. The diagnostic accuracy of the CAR-HEMATOTOX score may be further improved by integrating dynamic risk factors like inflammatory markers (e.g., IL-6, IFN-γ) or by making disease-specific adjustments. Potential applications of the score include restricting antibiotic prophylaxis or early G-CSF to high-risk patients, which will ideally be confirmed prospectively. Studying different ICAHT mitigation strategies in clinical trials will help to crystallize the optimal timing and sequence of G-CSF, TPO agonists, and HSCBs. Critical clinical endpoints to consider include time to neutrophil recovery, the rate of febrile neutropenia and infections, but also measures such as antibiotic exposure and duration of hospitalization.

While translational efforts have provided some insights into underlying mechanisms of ICAHT, it is unlikely that there is one unifying pathophysiology. Future preclinical studies will therefore have to take a panoply of host- and disease-related features into account. Importantly, mechanistic studies will require structured sample collection that is harmonized across centers and should leverage emerging technologies such as multiomic and spatial transcriptomic approaches. Furthermore, the paucity of preclinical and animal models studying the effects of CAR-T on hematopoiesis will need to be overcome, which would enable the systematic evaluation of novel therapeutics that ameliorate severe ICAHT. Ultimately, addressing these emerging research questions will require dedicated research efforts that integrate multilateral collaborations, registry studies and well-designed clinical studies. The latter should carefully evaluate specific management strategies, which would provide a blueprint for other immune effector cell therapies like bispecifics.