Chimeric antigen receptor T-cell therapy and cardiovascular outcomes in US Medicare beneficiaries

Abstract

Background and Aims

Chimeric antigen receptor T-cell (CAR-T) therapies are cellular immunotherapies that improve survival in patients with relapsed haematologic malignancies. However, their association with major adverse cardiovascular events (MACE) has received limited study, particularly in older adults. This study investigated the incidence of MACE, associated risk factors, and their impact on survival among older patients undergoing CAR-T in the USA.

Methods

Medicare fee-for-service beneficiaries over 65 who received inpatient CAR-T therapy between 2018 and 2023 were included. Baseline characteristics were assessed during the 12 months preceding CAR-T. MACE were defined as a composite of acute heart failure (HF), cardiogenic shock, myocardial infarction, cardiac tamponade, ventricular arrhythmia, complete heart block, or stroke. Multivariable models were adjusted for demographics, malignancy type, and baseline cardiovascular comorbidities.

Results

Among 3292 patients receiving CAR-T, 191 (5.8%) had MACE. Most common events were acute HF (3.1%), followed by ischaemic (1.3%) and haemorrhagic stroke (1%). Pre-treatment atrial fibrillation/flutter [adjusted odds ratio (aOR) 1.52 (1.08–2.16)], cardiomyopathy [aOR 2.49 (1.75–3.54)], and cerebrovascular disease [aOR 2.40 (1.30–4.43)] were independently associated with MACE. In 2021–23, MACE were also associated with immune effector cell-associated neurotoxicity syndrome and higher-grade cytokine release syndrome. MACE were associated with higher in-hospital mortality [aOR 16.9 (11.0–26.1)] and 1-year mortality after discharge [adjusted hazard ratio 1.91 (1.46–2.49)].

Conclusions

In the largest national sample of older adults receiving CAR-T, MACE occurred in 5.8% of patients and were associated with increased in-hospital and 1-year mortality. Further investigation into preventive and mitigating measures is needed.

Structured Graphical Abstract

Key Question

What are the incidence of major adverse cardiovascular events (MACE), the predictors, and the impact on survival among older adults (age >65 years) undergoing inpatient chimeric antigen receptor T-cell (CAR-T) therapy in the United States?

Key Finding

Medicare fee-for-service beneficiaries over 65 years of age who received inpatient CAR-T between 2018 and 2023 were included. MACE occurred in 5.8% of patients and were associated with increased in-hospital and 1-year mortality. Pretreatment atrial fibrillation/flutter, cardiomyopathy, cerebrovascular disease, as well as the development of high-grade cytokine release syndrome or immune effector cell-associated neurotoxicity syndrome were independent predictors of MACE.

Take Home Message

CAR-T can be utilized in older patients with an acceptable cardiovascular (CV) risk profile. Pretreatment CV evaluation could further aid in CV risk optimization prior to CAR-T, though additional investigation into preventive and mitigation measures is needed.

Older adults undergoing inpatient CAR-T therapy are at a modest risk of major adverse cardiovascular events (MACE), which are strongly associated with increased in-hospital and 1-year mortality. In a nationwide cohort of Medicare beneficiaries aged ≥65 years, pretreatment atrial fibrillation/flutter, cardiomyopathy, cerebrovascular disease, and severe CRS/ICANS emerged as independent predictors of cardiovascular complications, highlighting the importance of comprehensive cardiovascular risk assessment and optimization prior to CAR-T therapy.

Introduction

Chimeric antigen receptor T-cell (CAR-T) therapy is a class of cellular immunotherapy that can improve survival in patients with haematologic malignancies relapsing after conventional therapies.^1,2 CAR-T are autologous T-cells genetically engineered ex vivo to express a tumour-specific recombinant fusion protein. After the engineered cells are reinfused into the patient, they target tumour-specific antigens [such as CD19 or B-cell maturation antigen (BCMA)] and mount an exaggerated T-cell response independent of major histocompatibility complex (MHC) recognition.³ Currently there are seven commercially available CAR-T approved for the treatment of B-cell acute lymphoblastic leukaemia (ALL), B-cell non-Hodgkin lymphoma (NHL) including diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), and mantle cell lymphoma (MCL), as well as multiple myeloma (MM).⁴

Despite their success in revolutionizing the oncologic treatment landscape and improving oncologic outcomes in relapsed malignancies, CAR-T therapies are associated with life-threatening adverse events often related to immune system hyperstimulation including cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS).^1,2,5–7

Less is known about this therapy’s cardiovascular (CV) complications. CV events are often underestimated in oncology clinical trials relative to real-world experience due to under-representation of patients with pre-existing CV disease and risk factors, in addition to clinical trial designs that specifically focus on malignancy-related rather than CV outcomes.^8,9 However, over the last 5 years, a number of small studies and single-centre experiences reported an association between CAR-T and adverse CV events including arrhythmias, left ventricular dysfunction, heart failure (HF), and myocardial infarction (MI).^10–14 While the definition of major adverse cardiovascular events (MACE) varied, several studies linked the occurrence of MACE to concurrent CRS. The studies also suggested that early administration of tocilizumab during episodes of CRS could mitigate the risk of MACE,¹¹ but several gaps in knowledge still exist. For example, the incidence and predictors of CV events and their effects on short- and long-term outcomes in a large, multicentred real-world cohort remain unknown. Additionally, there is a paucity of data about the CV risk profile of CAR-T in older patients, a group with a typically higher burden of pre-existing CV risk factors and prior events compared with younger patients who were included in clinical trials and single-centre cohorts. Finally, contemporary CAR-T therapies such as those targeting BCMA for MM have received limited study,^7,15 since most prior reports investigated the CD19 CAR-T products that were first approved.^5,11,13,16

Importantly, as CAR-T is increasingly administered to older and more comorbid patients in routine practice, contemporary population-level estimates of CV risk in this age group are needed to inform pre-treatment CV evaluation, shared decision-making, and risk mitigation strategies. Our study aims to leverage large-scale Medicare beneficiary data to investigate the incidence and predictors of MACE following CAR-T, the association of MACE with CRS and ICANS, and the effect of MACE on in-hospital and post-discharge outcomes in a contemporary national sample of older adults receiving inpatient CAR-T for relapsed haematologic malignancies.

Methods

Study selection

All Medicare fee-for-service beneficiaries over the age of 65 who received inpatient CAR-T for haematologic malignancies between January 2018 and September 2023 were identified from the Medicare Virtual Research Data Center (VRDC) using the International Classification of Diseases 10th Revision (ICD-10) inpatient procedural codes for commercially available CAR-T therapies including: axicabtagene ciloleucel (Yescarta^®), tisagenlecleucel (Kymriah^®), brexucabtagene autoleucel (Tecartus^®), lisocabtagene maraleucel (Breyanzi^®), ciltacabta-gene autoleucel (Carvykty^®), or idecabtagene vicleucel (Abecma^®). Obecabtagene autoleucel (Aucatzyl^®) was not included since it was not available during the study period. Identifying the specific CAR-T product was feasible with the introduction of type-specific procedural ICD-10 codes starting in the year 2021 (see Supplementary data online, Table S1). Patients with a non-specific CAR-T procedural code without a concomitant type-specific CAR-T code of claim were labelled as ‘type unspecified’. Medicare Advantage patients were also excluded due to incomplete data (Figure 1).

Figure 1.

Flow chart of included and excluded patients. Abbreviation: CAR-T, chimeric antigen receptor T-cell therapy

For inclusion, patients were required to have 12 months of uninterrupted enrolment in Medicare Part A and Part B prior to the CAR-T index admission. Baseline characteristics including sociodemographic parameters, medical comorbidities, haematologic malignancy type, and prior oncologic therapies were identified from patients’ enrolment files and ICD-10 diagnostic and procedural codes of claims during the 12 months preceding CAR-T administration (see Supplementary data online, Table S2). Following Medicare chronic condition code protocols, to improve the specificity and accuracy of identifying baseline characteristics and minimize ‘rule-out’ coding, we required the presence of either one inpatient or two out-patient diagnostic or procedural codes of claims to identify a baseline characteristic.

Outcomes

The primary outcome was the occurrence of MACE during the CAR-T hospitalization identified using ICD-10 codes as a composite incidence of any of the following: acute HF, cardiogenic shock, acute MI, cardiac tamponade, ventricular arrhythmia leading to cardiac arrest, complete heart block, and ischaemic or haemorrhagic strokes (see Supplementary data online, Table S3). We did not include atrial arrhythmias, specifically atrial fibrillation and atrial flutter (AF), in the definition of MACE because baseline AF was modelled as a predictor variable, and administrative diagnostic codes do not reliably distinguish chronic from new-onset atrial fibrillation, creating a high risk of misclassification. In addition, our definition of MACE was intentionally restricted to more severe and life-threatening CV events. In a sensitivity analysis addressing potential diagnostic code misclassification (i.e. MACE preceding CAR-T erroneously listed as a diagnostic code during CAR-T admission), patients with MACE occurring during the 3 months preceding CAR-T were excluded from the analysis.

We studied the association between baseline characteristics and the incidence of inpatient MACE. We also investigated the association of MACE with the incidence and grade of CRS and ICANS during the index admission for the years where CRS and ICANS codes were introduced and clinically utilized (2021–23 and 2022–23, respectively). CRS incidence and grading were identified using ICD-10 diagnostic codes (D89.83x, introduced in October 2020) in addition to ICD-10 procedural codes for the administration of tocilizumab (XW033H5, XW043H5, introduced in August 2020). ICANS was identified using the ICD-10 code G92.0, introduced in October 2021 (see Supplementary data online, Table S3). We also explored the association of MACE with in-hospital all-cause mortality and length of hospitalization. We further assessed trends of CAR-T utilization before, during, and after the coronavirus disease-19 (COVID-19) pandemic and in different rurality areas of the USA.

In patients who survived hospital discharge, we studied the association between the occurrence of MACE during CAR-T hospitalization and all-cause 12-month mortality. Additionally, we explored the incidence of new-onset 12-month MACE in patients who did not have MACE during the CAR-T hospitalization.

Statistical analysis

Continuous variables were presented as means (standard deviation) or medians (interquartile range) and compared using a t-test or Wilcoxon test, depending on distribution. Categorical variables were presented as frequency (%) and compared using the χ² test. To comply with the Centers for Medicare and Medicaid Services (CMS) Cell Size Suppression Policy, cells with values of ≤10 were not reported. Cells that allow a value of ≤10 to be derived from other reported cells or information were also omitted. Multivariable logistic models with Firth penalty were fit to evaluate the association between baseline MACE and clinical and demographic baseline characteristics. The following variables were included in all models: age, sex, race, haematologic malignancy type, and medical history [AF, coronary artery disease (CAD), chronic kidney disease (CKD), HF/cardiomyopathy, cerebrovascular disease, diabetes mellitus, hypertension, ventricular arrhythmias, peripheral artery disease, and tobacco use]. Pre-CAR-T antineoplastic therapies including chemotherapy, immunotherapy, and radiation in addition to the type of CAR-T therapy were not included in the multivariable prediction model due to collinearity with haematologic malignancy type. The index hospitalization length of stay was compared for those with and without MACE using the Wilcoxon test. Post-discharge 1-year mortality was evaluated using the Cox proportional hazards model. All analyses were performed using SAS Enterprise Guide version 7.15.

Results

Patient characteristics

Between January 2018 and September 2023, a total of 3292 eligible patients received CAR-T (mean 73.3 years, 40% female) (Table 1). The annual number of CAR-T performed increased over time with DLBCL being the most common haematologic malignancy treated with CAR-T initially before being exceeded by MM in 2023 (see Supplementary data online, Figure S1). Similarly, the product type of CAR-T varied over time from 2021 to 2023 to reflect the increased use of MM CAR-T products (see Supplementary data online, Figure S2). The number of CAR-T performed in different geographic regions across the country (urban vs micropolitan vs rural) before (2018–19), during (2020–21), and after (2022–23) the COVID-19 pandemic is shown in Supplementary data online, Table S4.

Table 1.

Baseline demographics and medical comorbidities in patients with and without major adverse cardiovascular events

	All patients (n = 3292)	Patients without MACE (n = 3101)	Patients with MACE (n = 191)	P-value
Demographics
Age, years, mean (SD)	73.3 (4.6)	73.3 (4.6)	73.9 (4.6)	.054
Female sex, n (%)	1307 (39.7)	1240 (40.0)	67 (35.1)	.18
Race, n (%)
White	2840 (86.3)	2677 (86.3)	163 (85.3)	.70
Non-White	452 (13.7)	426 (13.7)	26 (15.1)
ZIP income, $, median (IQR)	82 260 (33–864)	82 112 (33–840)	84 633 (34–243)	.32
Medicaid enrolment, n (%)	168 (5.1)	NA	NA	.80
Geographic region, n (%)
Midwest	721 (21.9)	687(22.2)	34(17.8)	.09
Northeast	832 (25.3)	791(25.5)	41 (21.5)
South	1072 (32.6)	995(32.1)	77(40.3)
West	667 (20.3)	62 820.3)	39(20.4)
Hospital type, n (%)
Teaching	2738 (83.2)	2583 (83.3)	155 (81.2)	.44
Rural	NA	NA	NA	1
Medical characteristics
Cancer type, n (%)				.04
Diffuse large B-cell lymphoma	1789 (54.3)	1675 (54.0)	114 (59.7)
B-cell acute lymphoblastic leukaemia	27 (0.8)	NA	NA
Follicular lymphoma	173 (5.3)	NA	NA
Mantle cell lymphoma	281 (8.5)	257 (83)	24 (12.6)
Multiple myeloma	1022 (31)	977 (31.5)	45 (23.6)
Prior cancer therapies, n (%)
Anthracyclines	779 (23.7)	728 (23.5)	51 (26.7)	.31
Proteasome inhibitors	768 (23.3)	729 (23.5)	39 (20.4)	.33
Radiation	907 (27.6)	857 (27.6)	50(26.2)	.66
Stem cell transplantation	124 (3.8)	NA	NA	.21
CV risk factors, n (%)
Hypertension	1980 (60.1)	1845 (59.5)	135 (70.7)	.002
Diabetes	663 (20.1)	610 (19.7)	53 (27.8)	.007
Hyperlipidaemia	1566 (47.6)	1470 (47.4)	96 (50.3)	.44
Chronic kidney disease	629 (19.1)	588 (19.0)	41(21.5)	.39
Obesity	389 (11.8)	359 (11.6)	30 (15.7)	.09
Smoking	158 (4.8)	146 (4.7)	12 (6.3)	.32
Prior CV events/conditions, n (%)
Atrial fibrillation/flutter	554 (16.8)	501 (16.2)	53 (27.8)	<.001
Coronary artery disease	685 (20.8)	628 (20.3)	57 (20.8)	.002
Myocardial infarction	195 (5.9)	176 (5.7)	19 (10.0)	.02
Coronary artery bypass grafting	23 (0.7)	NA	NA	.55
Percutaneous coronary intervention	26 (0.8)	NA	NA	.21
Ventricular arrhythmia	62 (1.9)	NA	NA	.06
Heart failure/cardiomyopathy	450 (13.7)	393 (12.7)	57 (29.8)	<.001
Cerebrovascular disease, haemorrhagic stroke, ischaemic stroke, or transient ischaemic attack	96 (2.9)	83 (2.7)	13 (6.8)	<.001
Peripheral arterial disease	440 (13.4)	412 (13.3)	28 (14.7)	.59
Any CV risk factor or pre-existing cardiovascular disease	2683 (81.5)	2514 (81.1)	169 (88.5)	.014

CV, cardiovascular; MACE, major adverse cardiovascular event; SD, standard deviation; IQR, interquartile range; ZIP, zone improvement plan.

Baseline demographics and comorbidities were extracted from the 12 months preceding the CAR-T admission. In univariable analysis (chi-square test), age, sex, and race were not associated with MACE, whereas patients experiencing MACE had a higher burden of baseline CV risk factors/disease (such as hypertension, diabetes, atrial fibrillation/flutter, coronary artery disease, heart failure/cardiomyopathy, and cerebrovascular disease) and were more likely to have diffuse large B-cell lymphoma or mantle cell lymphoma (vs follicular lymphoma or multiple myeloma).

To comply with the Centers for Medicare and Medicaid Services (CMS) Cell Size Suppression Policy, cells with values of ≤10 were not reported. Cells that allow a value of ≤10 to be derived from other reported cells or information were also omitted.

Major adverse cardiovascular events

MACE occurred during the CAR-T hospitalization in 191 (5.8%) patients. Table 1 illustrates the univariate comparison of demographics and baseline characteristics between patients with and without MACE. Age, sex, and race were not associated with MACE, whereas patients experiencing MACE had a higher burden of baseline CV risk factors/disease (such as hypertension, diabetes, AF, CAD, HF, and cerebrovascular disease) and were more likely to have DLBCL or MCL (vs FL or MM) (Table 1). There was also a difference in the incidence of MACE between different CAR-T product types [P = .049 (years 2021–23)] likely correlating with differences in MACE among different haematologic malignancies (see Supplementary data online, Table S5). Patients who received Brexucabtagene Autoleucel, indicated for relapsed or refractory mantle cell lymphoma or B-cell ALL, experienced the highest incidence of MACE. In a post hoc analysis stratifying by disease type and line of therapy, there was no significant difference in the incidence of MACE between CAR-T products in patients with either DLBCL (axi-cel vs liso-cel vs tisa-cel) or MM (ide-cel vs cilta-cel) (see Supplementary data online, Table S6). Analyses for other malignancies were limited due to small sample sizes, a limited number of events, and/or the predominant use of a single CAR-T product within those disease groups.

The most common MACE were acute HF [103 (3.13%) cases] followed by ischaemic and haemorrhagic strokes [42 (1.3%) and 33 (1%) cases, respectively] (Figure 2). The incidence of MACE did not differ before (2018–19), during (2020–21), and after (2022–23) the COVID-19 pandemic (MACE incidence of 5.4% vs 6.2% vs 5.7% respectively, P = .83). While not included in our MACE definition, new-onset AF occurred in 5.8% of patients without prior AF diagnosis in the year preceding CAR-T. The select incidence of individual inpatient CV adverse events during the CAR-T hospitalization is shown in Supplementary data online, Figure S3.

Figure 2.

Breakdown of major adverse cardiovascular events (MACE). Most common MACE were acute HF, followed by ischaemic and haemorrhagic strokes. To comply with the Centers for Medicare and Medicaid Services (CMS) Cell Size Suppression Policy, events with values of ≤10 were not reported

In multivariable logistic regression modelling, pre-treatment (baseline) HF/cardiomyopathy [adjusted odds ratio (aOR) 2.49 (1.75–3.54), P < .0001], AF [aOR 1.52 (1.08–2.16), P = .018], and cerebrovascular disease [aOR 2.40 (1.30–4.43), P = .01] were independently associated with MACE (Table 2). Patients with MM had a non-significant trend for lower MACE incidence compared with patients with DLBCL [aOR .71 (0.49–1.02), P = .06]. The results were similar when patients who experienced MACE within 3 months prior to index hospitalization were excluded in a sensitivity analysis (see Supplementary data online, Table S7). Age as a continuous (linear) covariate was not associated with an increased risk of MACE. However, patients aged ≥75 years had a significantly higher adjusted risk of MACE compared with those aged 65–75 years [aOR 1.45 (1.07–1.95), P = .016].

Table 2.

Baseline comorbidities and treatment-related factors independently associated with major adverse cardiovascular events during CAR-T hospitalization

Variable	Main model (all years) n = 3292			CRS patients with available grading (2021–23) n = 1537			ICANS model (2022–23) n = 1905
	aOR	95% CI	P-value	aOR	95% CI	P-value	aOR	95% CI	P-value
Age	1.02	(0.98–1.05)	.35	1.02	0.97–1.07	.40	1.00	0.96–1.04	.85
Sex (female vs male)	0.95	(0.70–1.30)	.76	0.76	0.46–1.24	.27	0.86	0.57–1.31	.48
Race (non-White vs White)	1.20	(0.79–1.82)	.40	1.35	0.73–2.50	.34	1.44	0.87–2.41	.16
Cancer type (reference: DLBCL)			.06			.30			.30
B-cell ALL	0.62	0.11–3.43		0.29	0.02–5.43		0.53	0.09–3.14
Follicular lymphoma	0.82	0.38–1.74		1.50	0.54–4.18		0.99	0.40–2.45
Mantle cell lymphoma	1.52	0.96–2.42		1.82	0.92–3.59		1.59	0.87–2.89
Multiple myeloma	0.71	0.49–1.02		0.93	0.54–1.62		0.80	0.50–1.26
Atrial fibrillation/flutter	1.52	1.08–2.16	.02	1.05	0.58–1.91	.87	1.06	0.64–1.74	.83
Coronary artery disease	1.18	0.83–1.68	.35	0.89	0.50–1.59	.69	1.14	0.71–1.81	.59
Chronic kidney disease	0.89	0.61–1.29	.52	0.60	0.32–1.14	.12	0.61	0.36–1.02	.06
Heart failure/cardiomyopathy	2.49	1.75–3.54	<.0001	3.14	1.81–5.44	<.001	2.51	1.58–3.98	<.0001
Cerebrovascular disease	2.40	1.30–4.43	.01	2.46	0.88–6.86	.08	2.57	1.20–5.49	.01
Diabetes mellitus	1.36	0.96–1.92	.08	1.64	0.94–2.86	.08	1.70	1.09–2.65	.02
Hypertension	1.24	0.88–1.75	.22	1.22	0.72–2.06	.45	1.25	0.79–1.96	.34
Ventricular arrhythmia	1.27	0.57–2.84	.57	1.41	0.43–4.65	.57	2.05	0.77–5.44	.15
Peripheral arterial disease	0.83	0.54–1.27	.38	0.84	0.42–1.69	.63	0.87	0.50–1.49	.60
Tobacco use	1.19	0.65–2.20	.57	0.71	0.19–2.66	.61	1.49	0.67–3.31	.33
CRS (reference: Grade 1)						.01
Grade 2				1.18	0.71–1.97
Grade 3				1.83	0.76–4.40
Grade 4				6.00	2.11–17.02
ICANS							1.63	1.09–2.43	.02

aOR, adjusted odds ratio; CI, confidence interval; DLBCL, diffuse large B-cell lymphoma; ALL, acute lymphoblastic leukaemia; CRS, cytokine release syndrome; ICANS, immune effector cell-associated neurotoxicity syndrome; MACE, major adverse cardiovascular events.

Multivariable logistic regression models with Firth penalty analysed independent associations between MACE and baseline characteristics across three groups (all patients and years in the main model, only in patients with cytokine release syndrome and available grading, and in 2022 and 2023 when the diagnostic code for ICANS was available). Variables with significant association with MACE are bolded. Among all patients in the main model, pre-treatment cardiomyopathy, atrial fibrillation/flutter, and cerebrovascular disease were independently associated with MACE. In the subgroup of patients where the incidence and grade of CRS were identified, higher-grade (grade ≥2) CRS was associated with increased risk of MACE along with pre-treatment cardiomyopathy. In the third model, restricted to the years that the diagnostic code for ICANS was available, ICANS was independently associated with increased risk of MACE along with pre-treatment cardiomyopathy, cerebrovascular disease, and diabetes.

Between 2021 and 2023, any-grade CRS occurred in 1617 (67%) of patients during CAR-T admission, with MACE occurring in 75 patients with CRS (4.6%) and 45 patients without CRS (5.6%). Among the 1537 patients with CRS where the grade of CRS was specified, higher-grade (grade ≥2) CRS was associated with an increased risk of MACE in multivariable analysis {MACE incidence by CRS grade: Grade 1, 4.0% [reference group]; Grade 2, 4.9% [aOR 1.18 (0.71–1.97)]; Grade 3, 8.1% [aOR 1.83 (0.76–4.40)]; Grade 4, 20.0% [aOR 6.00 (2.11–17.02)], P = .007} (Table 2).

Given our study’s retrospective design and limitations of the Medicare database, we could not identify the timing of MACE incidence in relation to tocilizumab administration and thus could not study the effect of tocilizumab on reducing MACE (we could not assess whether tocilizumab was given before or after MACE). Nevertheless, the timing of tocilizumab administration was not associated with in-hospital mortality [median time to tocilizumab administration after CAR-T in patients who died vs survived: 8 (IQR 6–14) vs 7 (IQR 4–9) days, P = .20] (years 2021–23).

In the years 2022 and 2023, ICANS occurred in 558 (29%) patients. During these years, ICANS was independently associated with MACE in multivariable analysis [aOR 1.63 (1.09–2.43), P = .02] (Table 2).

In-hospital mortality and length of stay

A total of 118 (3.6%) patients died during the hospitalization or were discharged to hospice care, with higher mortality in patients who had MACE compared with those who did not [50 (26.2%) vs 68 (2.2%) deaths, P < .0001]. Patients who experienced MACE also had longer hospitalizations [median 21 (IQR 15–30) vs 16 (IQR 12–21) days, P < .001]. After adjusting for differences in baseline comorbidities and haematologic malignancy type, MACE were independently associated with increased all-cause mortality during the CAR-T admission [aOR 16.9 (11.0–26.1), P < .0001]. Older age was also independently associated with higher mortality [aOR 1.06 (1.02–1.10), P = .01, per year]. Also, B-cell ALL patients were more likely to die than DLBCL patients [aOR 6.41 (2.07–19.86)] (notably, only 27 patients had ALL). MM patients were less likely to die compared with DLBCL patients [aOR 0.40 (0.23–0.70), P < .0001 for haematologic malignancy type] (see Supplementary data online, Table S8). There was a non-significant trend of increased in-hospital mortality before (2018–19) compared with during (2020–21) and after (2022–23) the COVID-19 pandemic (in-hospital mortality incidence of 5.2% vs 3.6% vs 3.2%, respectively, P = .093).

One-year mortality

Among the 3174 patients alive on discharge, 809 (25.5%) died within the first year with higher unadjusted 12-month all-cause mortality among patients who suffered MACE during the index admission compared with those who did not [62 (44.0%) vs 747 (24.6%) deaths, P < 0.0001] (Figure 3). The association of MACE with increased 12-month mortality remained significant after adjusting for differences in baseline comorbidities and haematologic malignancy type [adjusted hazard ratio (aHR) 1.91 (1.46–2.49), P < .0001]. Other baseline comorbidities that were independently associated with 12-month mortality included: AF, CKD, diabetes, and hypertension. Patients undergoing CAR-T for FL, MCL, and MM had lower 12-month mortality compared with patients with DLBCL (see Supplementary data online, Table S9). In the years 2021–23, the use of tocilizumab during the CAR-T hospitalization was not associated with 12-month mortality [aHR 1.09 (0.83–1.44), P = .53].

Figure 3.

One-year survival comparison between patients with vs without MACE during the CAR-T hospitalization. Among the 3174 patients alive on discharge, 809 (25.5%) died within the first year with higher 12-month all-cause mortality among patients who suffered MACE during the CAR-T admission compared with those who did not [62 (44.0%) vs 747 (24.6%) deaths, P < .001]. To comply with the Centers for Medicare and Medicaid Services (CMS) Cell Size Suppression Policy, intervals with ≤10 deaths were not reported. Abbreviation: MACE, major adverse cardiovascular event.

A total of 3033 patients survived and remained MACE-free during the index admission. Among those, 268 (8.8%) had at least one new-onset MACE during the following 12 months, and 127 (47.4%) of them died. The median time to MACE was 52 days. Most common MACE were HF (127 events), strokes (55 events), and acute MI (63 events). In competing risk with death analysis, predictors of MACE after discharge were baseline AF, CAD, HF/cardiomyopathy, cerebrovascular disease, and diabetes (see Supplementary data online, Table S10). Results were unchanged with cause-specific hazard analysis. Comparison of in-hospital and 12-month mortality between patients receiving different CAR-T products is shown in Supplementary data online, Table S11.

Validation of CART-7 score

In a recent study, Farmakis et al.¹⁷ performed a meta-analysis of previously published cohorts of CAR-T recipients to identify baseline predictors of CV adverse events and used the pooled relative risks of significant predictors to develop a risk stratification score, termed CART-7. This score assigns weighted contributions to several baseline CV risk factors, including HF, AF, CAD, hyperlipidaemia, diabetes, hypertension, and smoking status. Patients were stratified based on their score as low risk (0–8), moderate risk (9–17), high risk (18–25), or very high risk (26–33). We validated the CART-7 score in our cohort (see Supplementary data online, Table S12) and observed similar consistent findings. Specifically, higher CART-7 scores were associated with a stepwise increase in MACE risk [4.2% in the low-risk group, 7.1% in the moderate-risk group, 10.4% in the high-risk group, and 18.4% in the very high-risk group (P < .001)] with comparable incidence of MACE within each risk category to their meta-analysis.

Discussion

In the largest nationwide analysis of older patients receiving contemporary commercially available CAR-T, we observed several notable findings. MACE occurred in 5.8% of patients during CAR-T admission and was associated with a 17-fold higher in-hospital mortality, longer lengths of stay, and increased 12-month mortality following CAR-T. Patients with pre-treatment cardiomyopathy, AF, and cerebrovascular disease were at the highest risk of developing in-hospital MACE. We also observed a non-statistically significant trend of lower incidence of MACE in patients receiving MM CAR-T compared with those receiving DLBCL CAR-T. Further, both high-grade CRS and ICANS appear to be significantly associated with higher MACE occurrence (Structured Graphical Abstract).

In this cohort we provide real-world estimates of the incidence, predictors, and prognostic implications of MACE in an older population underrepresented in prior trials and registries. While the association between severe CV events and mortality may be expected, the magnitude of this risk in a contemporary, nationwide older cohort has not previously been quantified and provides important benchmarks for clinical counselling, risk stratification, and evaluation of future CAR-T constructs and toxicity mitigation strategies. Although cardiomyopathy, AF, and cerebrovascular disease are established CV risk predictors in general populations, treatment-specific CV risk profiling is a central principle of cardio-oncology. Different oncologic therapies are associated with distinct patterns of CV toxicity and therefore require therapy-specific risk models, as reflected in the European Society of Cardiology (ESC) cardio-oncology guidelines.¹⁸ Large population-based cohorts such as ours provide foundational data necessary to develop and validate CAR-T-specific risk stratification tools, as illustrated by our validation of the CART-7 score in this older real-world population.¹⁷ To aid in counselling and risk stratification, it should be noted that the reported 5.8% incidence of MACE represents a restricted composite of severe CV events rather than the overall spectrum of CV toxicity associated with CAR-T therapy.

These findings build upon previous literature and offer a distinct contribution due to our study size, older population, contemporary CAR-T studied, and strict definition of MACE to only include serious life-threatening events. As CAR-T continues to become more prevalent and with more approved indications, a more precise understanding of the CV risk involved is instrumental in preprocedural CV assessments, patient benefit–risk discussions, and risk mitigation efforts. Our study underscores the importance of rigorous CV risk assessment and follow-up of older patients receiving CAR-T and emphasizes the need for future research focusing on prevention strategies and optimal management practices to improve outcomes.

Our in-hospital MACE incidence of 5.8%, while still significant, appears lower than some prior studies largely due to a stricter definition of MACE that focused on major events and excluded AF and subclinical cardiomyopathy. In prior studies of older CAR-T products (between the years 2016 and 2019), the incidence of short-term (~30-day) MACE ranged between 12% and 21%, with varied MACE definitions and duration of follow-up.^11,12,14,19 Cardiac arrhythmias (mostly supraventricular) made up 30%–70% of prior reported MACE. A recently published meta-analysis of 13 studies comprising 1528 CAR-T recipients also found that supraventricular arrhythmias and cardiomyopathies were the most commonly reported MACE (pooled prevalence of around 8% and 9%, respectively).¹⁰ The pooled incidence of more serious clinical events of acute HF, MI, and ventricular arrhythmias were relatively similar to our study (3.8%, 0.6%, and 0.7%, respectively).¹⁰ In an analysis of 202 patients from a multicentre registry where MACE was defined by more serious adverse events excluding supraventricular arrhythmias (a composite incidence of clinical HF, cardiogenic shock, or MI), Mahmood et al.²⁰ reported a MACE incidence of 16% during a median follow-up of 297 days. This incidence is similar to our observed 1-year MACE incidence of around 15%.

As our study included Medicare beneficiaries over the age of 65, prior cohorts were significantly younger at the time of receiving CAR-T (median age of 61 in the recently published meta-analysis of all 13 studies¹⁰ vs 73 in our cohort). Nevertheless, we observed a comparable incidence of MACE among our older cohort and did not observe an association between age and MACE in our multivariable analysis that adjusted for differences in baseline medical comorbidities and haematologic malignancy type. A study by Steiner et al.¹⁹ previously reported age >60 years to be significantly associated with increased risk of 30-day MACE; however, 19 out of their observed 27 cases of MACE were non-life-threatening arrhythmic events. Similarly Lee et al.²¹ associated advanced age with adverse cardiac events following CAR-T, but 10/11 of those events were new-onset AF. These differences are also likely explained by our definition of MACE that did not include supraventricular arrhythmias (particularly AF) that are well-known to increase with age. Because our study was limited to Medicare beneficiaries aged 65 years and older, a younger comparator group was not available, precluding assessment of age-specific vulnerability. However, our results suggest that older patients may still have an acceptable CV risk profile and should be considered for CAR-T when eligible.

The impact of MACE on survival, clinical outcomes, and healthcare utilization following CAR-T has been inadequately explored, with previous studies often limited by small sample sizes, low event rates, and short follow-up. Steiner et al.¹⁹ reported a 16% 30-day MACE incidence among 165 patients, primarily arrhythmias, but found no association with progression-free or overall survival at a median follow-up of 16 months. Similarly, Lee et al.²¹ observed no difference in 1-year overall survival between patients with and without MACE in a cohort of 90 CAR-T recipients. A smaller study of 78 patients receiving idecabtagene vicleucel suggested a non-statistically significant trend towards worse survival in patients with MACE during a 6-month median follow-up.¹⁵ However, in a larger multicentre registry of 202 patients receiving anti-CD19 CAR-T, Mahmood et al.²⁰ reported higher adjusted overall mortality and non-relapse mortality among 33 (16%) patients who developed severe cardiac events (heart failure, cardiogenic shock, or myocardial infarction) following CAR-T. Leveraging our large Medicare cohort with uninterrupted follow-up data, we confirmed an independent association between MACE and higher in-hospital mortality [aOR 16.9 (11.0–26.1)], longer hospital stays, and elevated 12-month mortality post-discharge [aHR 1.91 (1.46–2.49)].

Our results also contribute further insights into the relationship between CRS, ICANS, and CV outcomes. In our cohort, 67% of patients experienced CRS, similar to prior studies.^16,22 MACE occurred with similar frequency in patients with and without CRS in our cohort, although increasing CRS severity was significantly associated with an increased risk of MACE in multivariable analysis. Earlier literature with first-generation CAR-T originally reported an association between MACE and the incidence and grade of CRS. Alvi et al.¹¹ observed a 12% MACE incidence among patients with grade ≥2 CRS and no MACE events otherwise, noting a graded relationship between CRS severity, elevated troponin, and CV events, and reported reduced CV events with earlier tocilizumab administration. Subsequent studies similarly associated higher-grade CRS with increased MACE risk.^12,14,20 Consequently, the ESC cardio-oncology guidelines recognized the possible association between MACE and CRS, suggesting early treatment with tocilizumab and/or dexamethasone to possibly reduce this risk.¹⁸ In contrast, other recent reports did not observe an association between the occurrence of CRS and MACE in patients receiving CAR-T.^23–26

Several theories may explain discrepancies in the reported association between CRS and MACE in addition to the limitations expected from small, single-centre cohorts with fewer observed events. First, it is very plausible that higher-grade CRS drives the association with MACE, as was seen in our analysis and others.^11–13,20 Second, the association between CRS and MACE was previously more often observed in studies that included AF in the definition of MACE,^11,14,19 and the known relationship between AF and inflammatory conditions like CRS and sepsis may amplify this association in smaller studies.^27–29 Indeed, arrhythmias constitute the majority of CV complications reported post CAR-T [77.6% of CV complications in the Food and Drug Administration (FDA) Adverse Events Reporting System (FAERS)].³⁰ Third, the association between MACE and CRS in older studies may have been influenced by patients receiving first-generation commercial CAR-Ts mostly for DLBCL^11,14,19,20 as opposed to newer studies with more recently approved CAR-T including those for MM.^24,25 While subgroup analysis by CAR-T and haematologic malignancy type in those cohorts was not reported, likely due to low event rates, it is very possible that different CAR-T products have different CV risk profiles and association with CRS. Pre-CAR-T oncologic therapies with potential cardiotoxicity and adverse CV risk profiles vary by malignancy type and may influence subsequent event rates. Regardless of the underlying mechanism, we observed a non-statistically significant trend towards a higher risk of MACE among patients with DLBCL and MCL compared with those with MM. This difference may reflect the more frequent exposure of lymphoma patients to cardiotoxic treatments prior to CAR-T therapy, including anthracyclines, cyclophosphamide, and mediastinal radiation, relative to patients with MM, but other factors may also be contributing. Across all patients with lymphoma (rather than DLBCL specifically) or myeloma, there was no evidence of an increased risk of MACE in lymphoma patients after controlling for baseline CV risk factors in a prior analysis of a nationwide administrative database.³¹ Further investigation will be warranted to determine whether different CAR-T constructs (e.g. CD19- vs BCMA-directed therapies) have distinct intrinsic CV risk profiles. Finally, enhanced clinicians’ awareness of MACE risks and early treatment of CRS might have mitigated this association in recent literature, although high-grade CRS continues to be an important risk factor in our contemporary cohort.

The association between ICANS and MACE has received limited study. Steiner et al.¹⁹ and Korell et al.²⁴ did not observe an increased risk of MACE among patients with ICANS. In contrast, Lee et al.¹⁵ observed over five-fold higher incidence of ICANS in patients who had an adverse cardiac event vs those who did not in a cohort of 78 patients receiving the (BCMA)-targeting idecabtagene vicleucel (ide-cel). We similarly observed a higher incidence of MACE in patients with vs without ICANS (7.4% vs 4%); however, our analysis was limited to the years 2022–23 when the diagnostic codes of ICANS became available and utilized. It is also possible that those findings may be related to an overlap in the clinical presentation between cerebrovascular accidents and ICANS which could have led to over-coding both conditions and thus potentially overestimating the association.

The pathophysiology underlying the association between CAR-T and CV events remains incompletely understood but involves several proposed inflammatory and non-inflammatory mechanisms.^22,32 Pro-inflammatory cytokines during CRS may cause vascular (capillary) leak syndrome, which, combined with aggressive hydration during CAR-T, can result in hypotension, oedema, and hyponatraemia mimicking acute HF.^33,34 Also, inflammatory cytokines and catecholamine surges in CRS may be arrhythmogenic and prothrombotic, leading to secondary cardiac dysfunction, acute HF, arterial, and venous thromboembolic events.³² Additionally, sympathetic overstimulation with physical and emotional stress associated with CAR-T may lead to reversible stress-induced (takotsubo) cardiomyopathy.^35,36 At the molecular level, oxidative stress, altered calcium signalling, and elevated levels of tumour necrosis factor α during CRS may lead to endothelial injury, and increased expression of tissue factors, von Willebrand factor, and angiopoietin-2 all leading to endothelial activation and the activation of coagulation cascades and coagulopathy.^37–39 Alternatively, consumptive hypofibrinogenaemia and thrombocytopenia related to CRS and inflammation are associated with increased risk of bleeding and may be contributing to the haemorrhagic strokes seen in our study.⁴⁰ Potentially, these inflammatory mechanisms may contribute to MACE whether or not the patient is formally diagnosed with CRS or ICANS. It is worth noting that CRS and ICANS should not be interpreted as mechanistically independent predictors of MACE, but rather as clinically identifiable manifestations of a shared inflammatory and endothelial activation cascade that may contribute to cardiovascular toxicity. In this framework, CRS and ICANS likely function both as markers of heightened systemic inflammatory burden and as mediators of haemodynamic stress, arrhythmogenic substrate, and thrombo-inflammatory injury. Given the limitations of administrative data, we are unable to determine temporal sequencing or causality. Accordingly, our findings reflect clinical co-occurrence and risk stratification relevance rather than mechanistic independence.

Non-inflammatory mechanisms behind MACE events remain more poorly understood, but direct myocardial damage from CAR-T-cells’ recognition of cardiac antigens due to antigen mimicry or alloreactivity has been hypothesized.²² Fatal myocarditis was reported in patients receiving CAR-T targeting the MAGE-A3 tumour antigen in patients with MM and melanoma, likely due to cross-reactivity with the titin sarcomeric protein in cardiac myocytes.⁴¹ Similar to graft-vs-host disease, alloreactivity where reinfused T-cells mount an immune response against unrecognized major histocompatibility complex variants in host cardiac tissue has been also postulated.⁴²

Our results should be interpreted in the context of some important limitations. First, this is a retrospective analysis of an administrative dataset that uses ICD-10 codes to capture diagnoses and procedures, which carries inherent risk for confounding and diagnostic misclassification. However, we used codes utilized in previously published studies³¹ and employed various statistical models to adjust for possible confounders to reduce those risks. We included patients with uninterrupted Part A and Part B coverage for at least 12 months prior to CAR-T to ensure adequate data availability to capture baseline demographics, comorbidities, and malignancy type. Although it is theoretically possible that cardiac events occurring more than 12 months prior to CAR-T may have been missed, we mitigated this risk by incorporating additional diagnostic codes to complement procedural codes (e.g. coronary artery disease diagnostic codes to identify patients with remote coronary artery bypass surgery). Additionally, to improve the accuracy of identifying incident MACE and minimize diagnostic code misclassification (i.e. MACE preceding CAR-T erroneously listed as a diagnostic code during CAR-T admission), we conducted a sensitivity analysis excluding patients with reported MACE in the 3 months preceding CAR-T, and results were unchanged. Lastly, we required the occurrence of one inpatient or two out-patient codes of claim to capture baseline characteristics or MACE to increase our specificity and accuracy. Second, we did not include atrial arrhythmias in the definition of MACE since baseline AF was used as a potential predictor variable, and diagnostic codes do not reliably distinguish chronic from acute events. We also aimed to focus on more serious and life-threatening events in our analysis and definition of MACE. Thus, comparing our observed incidence and predictors of MACE to previous studies that included AF may be limited. Third, the Medicare database lacks cancer staging and disease burden which may impact the outcomes of CAR-T. Also, we did not include prior oncologic therapies in our multivariable models due to collinearity with haematologic malignancy type. Fourth, we identified all-cause mortality without differentiating malignancy vs CV-related deaths since the cause of death is not readily reported in the database. It is also important to note that mortality differences seen across different haematologic malignancies are only reported as a reference and are not meant to suggest comparative effects of CAR-T, since different malignancies are inherently diverse in disease aggressiveness, prognosis, and risk of death. Fifth, the database lacks clinically important information such as laboratory findings, inflammatory markers, and cardiovascular imaging, which may further refine risk stratification and mechanistic interpretation of CV events. Finally, longer follow-up is required to better understand late CV events and risk among survivors, particularly in those receiving more contemporary anti-myeloma CAR-T.

Conclusions

In the largest, nationwide analysis of older patients hospitalized for CAR-T, MACE occurred in around 6% of patients during the CAR-T admission and in 9% within the first year after discharge. From a clinical standpoint, the highest-risk phenotype emerging from our data appears to be CAR-T recipients with pre-existing cardiomyopathy, AF, or cerebrovascular disease who subsequently develop high-grade CRS and/or ICANS during hospitalization. This subgroup may warrant intensified haemodynamic monitoring, early cardio-oncology involvement, and aggressive management of inflammatory complications. However, while MACE occurred more commonly in patients with high-grade CRS, a substantial portion of our patients suffered MACE without CRS. The occurrence of MACE was associated with longer hospitalization and increased in-hospital and 12-month all-cause mortality after accounting for differences in baseline demographics, medical comorbidities, and haematologic malignancy type. Although our study population was significantly older than previous reports, we observed a relatively similar risk of MACE compared with prior younger cohorts. Our findings may suggest that carefully selected older patients may still have an acceptable CV risk profile and should be considered for CAR-T when eligible.

As CAR-T continues to develop several key unanswered questions remain. Large multicentre prospective studies are needed to build risk stratification tools, develop optimal monitoring protocols for cardiotoxicity, investigate the role of different cardiac biomarkers and multimodality imaging for early detection of cardiac injury, and evaluate different prophylactic and therapeutic interventions especially in patients who develop CRS. Clinicians’ awareness of the risk of adverse CV events following CAR-T, and their significant associated increase in mortality is important to facilitate CV risk factor optimization, close monitoring strategies, and early lifesaving interventions.

Supplementary Material

Supplementary data

Supplementary data are available at European Heart Journal online.

Data Availability

The data underlying this article are available on the Medicare Virtual Research Data Center (VRDC).