Second primary malignancies after CAR T-cell therapy: A systematic review and meta-analysis of 5,517 lymphoma and myeloma patients

Abstract

Purpose:

CAR T-cell therapy is a potent immunotherapy for hematologic malignancies, but patients can develop long-term adverse events including second primary malignancies (SPMs) that impact morbidity and mortality. To delineate the frequency and subtypes of SPMs following CAR-T in lymphoma and myeloma, we performed a systematic review and meta-analysis.

Design:

A literature search was conducted in the MEDLINE, Embase, and Cochrane CENTRAL databases. Following extraction of SPM cases and assignment of malignant origin, we analyzed SPM point estimates using random effect models.

Results:

We identified 326 SPMs across 5,517 patients from 18 clinical trials (CT) and 7 real-world studies (RWS). With a median follow-up of 21.7 months, the overall SPM point estimate was 5.8% (95%CI 4.7–7.2). SPM estimates were associated with treatment setting (CT>RWS), duration of follow-up, and number of prior treatment lines, which were each confirmed as independent study-level risk factors of SPM in a meta-regression model. A subgroup meta-analysis of the four trials that randomized CAR-T versus standard-of-care revealed a similar risk of SPM with either treatment strategy (p=0.92). In a distribution analysis of SPM subtypes, hematologic malignancies were the most common (37%), followed by solid tumors (27%) and non-melanoma skin cancers (16%). T-cell malignancies represented a small minority of events (1.5%). We noted disease- and product-specific variations in SPM distribution.

Conclusions:

These data raise awareness of SPM as a clinically relevant long-term adverse event in patients receiving CAR T-cell therapy. However, our findings do not indicate that SPM frequency is higher with CAR-T versus previous standard-of-care strategies.

Introduction

Chimeric antigen receptor (CAR) T-cell therapy has revolutionized the treatment landscape for numerous advanced B-cell malignancies and is increasingly being explored for the treatment of autoimmune diseases and solid tumors.^1–7 Despite its proven clinical efficacy, CAR T-cell therapy is also accompanied by a distinct array of immune-related adverse events, including cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS)^8–10 and the more recently recognized immune effector cell-associated hematotoxicity (ICAHT).^11–15 In severe cases, toxicities can lead to non-relapse mortality (NRM), with a recent meta-analysis demonstrating NRM estimates ranging between 5.7% and 10.6% depending on the applied CAR-T product and primary malignancy.¹⁶ While infections accounted for most non-relapse deaths in this study (>50%), a noteworthy 7.8% were attributed to second malignancies, distinct from the underlying lymphoma or myeloma. This raises important considerations regarding the incidence and distribution pattern of second primary malignancies (SPMs) following CAR T-cell therapy, including both fatal and non-fatal occurrences.

Concern about SPMs following CAR T-cell therapy has prompted recent regulatory attention, such as the United States Food and Drug Administration (FDA) issuing a class-wide black box warning on the risk of SPMs post-CAR T-cell therapy, focused especially on secondary myeloid malignancies and secondary T-cell malignancies, and rapidly spurring research in this domain.^17–22 A recent study revealed that 4.3% of all unique CAR-T adverse events reported to the FDA Adverse Events Reporting System (FAERS) database were instances of SPMs, with myeloid neoplasms (leukemias and myelodysplastic syndrome [MDS]) being the most prevalent secondary entity.²³ However, these results reflect the likelihood of reporting to the FAERS database rather than a true incidence and were subject to significant reporting bias, necessitating caution in their interpretation (i.e.., lack of a true denominator). While such pharmacovigilance studies may suggest a correlation between CAR T-cell therapy and SPMs^19,23, the mechanistic link between CAR T-cells and SPMs remains uncertain, and may be related to prior treatment (e.g. alkylators, lenalidomide, autologous hematopoietic cell transplantation [HCT], and/or other therapies), the pre-infusion lymphodepletion required to allow CAR T-cells to expand, and/or CAR T-cell induced immune suppression.^17,24,25 In addition, the emergence of T-cell malignant neoplasms after CAR T-cell therapy has garnered particular interest by the FDA due to potential concerns of insertional mutagenesis in patients receiving genetically modified cellular therapies (e.g., lentiviral or γ-retroviral transduction).^22,26 However, the absence of CAR transgenes in most tumor biopsies has also highlighted difficulties in attributing direct causation to the CAR T-cell product itself.^27,28

Understanding the landscape of SPMs post-CAR T-cell therapy is imperative for optimizing patient care and informing treatment decisions. In this systematic review and meta-analysis, we aim to delineate the incidence and distribution pattern of SPMs in lymphoma and myeloma patients treated with the FDA-approved CD19- and BCMA-directed CAR T-cell products. By synthesizing data from clinical trials and real-world studies, we outline the estimated SPM across a range of disease entities and CAR T-cell products. Finally, we describe a breakdown of SPM entities and provide context for the risk factors driving SPM development following CAR T-cell therapy.

Methods

Study design and literature search

We screened all studies for the FDA-approved CAR T-cell products in multiple myeloma and lymphoma (axi-cel, tisa-cel, ide-cel, cilta-cel, liso-cel, brexu-cel). A systematic search was conducted using the MEDLINE (via PubMed (RRID:SCR_004846)), Embase (RRID:SCR_008498) and Cochrane CENTRAL (RRID:SCR_006576) databases for articles published prior to May 17th, 2024 with combined keywords for each of the CAR T-cell products together with “lymphoma” or “myeloma” and an additional search for conference abstracts investigating SPMs in the context of CAR-T cell therapy (see study protocol, supplementary material). Case studies, reviews, meta-analyses and pharmacovigilance studies were excluded. After screening titles and abstracts, publications were evaluated by two independent investigators (TT, MA) based on the following inclusion criteria, which all needed to be met:

1. adult cancer patients with either indolent lymphoma (IL), large B-cell lymphoma (LBCL), multiple myeloma (MM), or mantle cell lymphoma (MCL),

2. use of CAR T-cell products approved by the Food and Drug Administration (FDA),

3. comprehensive reporting of all SPM cases throughout the entire follow-up period.

All included articles were checked for double reporting. If two studies reporting on the same patient population were identified, the study with longer follow-up was chosen. This study followed the PRISMA(-P) guidelines (see study protocol and PRISMA Checklist, supplemental material) and was prospectively registered to the PROSPERO database (study number: CRD42024542037).²⁹

Data extraction

For the studies meeting inclusion criteria, two independent investigators (TT, MA) extracted details such as publication date, study size, time period of patient inclusion, specific disease types, CAR T-cell product, follow-up duration, median age of the cohort, treatment history (median number of prior treatment lines and previous HCT), and treatment setting. The primary endpoint was the frequency and distribution pattern of SPMs after CAR T-cell therapy. SPM proportions were calculated by dividing the number of SPM occurrences by the total patient number within each study cohort. In cases where multiple CAR T-cell products were utilized within a single study, the data was segregated into distinct cohorts and analyzed separately. In the randomized studies, SPMs occurring in the standard arm were analyzed in an intent-to-treat manner.

Quality assessment

The Joanna Brigg’s Institute appraisal tool was applied to assess study bias of included articles (Table S1). Visual inspection of funnel plot asymmetry and Egger regression tests were used to assess reporting bias (Fig. S1).^30,31

Statistical analysis

Data were analyzed in R (v4.3.1), using the metafor (v.4.4–0) and meta (v7.0–0) packages. SPM point estimates were calculated by conducting random effect meta-analyses of single proportions utilizing a generalized linear mixed model.³² The Clopper-Pearson interval was used to compute 95% confidence intervals (95% CIs) of proportions.³³ Outcome data for SPMs were visualized using forest plots. Heterogeneity was assessed with the Q statistic and quantified using I² (values of 25%, 50% and 75% reflect low, moderate, and high between-study heterogeneity, respectively).³⁴ Continuous variables were compared using Mann-Whitney test, while Kruskal-Wallis test was used to compare multiple variables assuming non-normal distribution.

Meta-analyses

Separate meta-analyses were conducted for prespecified subgroups comparing SPM point estimates by age, disease entity, CAR T-cell product, treatment setting, follow-up, and treatment lines including prior hematopoietic cell transplantation (HCT). Subgroups were compared using a test for subgroup differences (random effects model). To test the robustness of main study conclusions, additional sensitivity analyses were performed examining fixed effects models.³⁵

Meta-regression

Multivariate meta-regressions were performed to examine the association between SPM point estimates and the following variables: follow-up time (treated as a continuous variable), median age of treated cohort, median number of prior treatment lines, proportion of patients with a previous HCT, and treatment setting (i.e., clinical trial vs. RWS). Exploratory univariate meta-regressions were conducted to estimate the contribution of individual variables to between-study heterogeneity using the R² statistic. Meta-regressions were calculated based on random effects models using the maximum likelihood estimator.³⁶ Individual model coefficients and respective confidence intervals were tested using the Knapp-Hartung method.³⁷ The stability of model estimates was validated by performing permutation testing.³⁸

Distribution of SPM entities

SPMs were classified into one of the following groups: hematological malignancies, solid tumors, non-melanoma skin cancers and unknown. Hematological malignancies were further classified into MDS, MDS/AML, AML, B-NHL, T-NHL, plasma cell dyscrasias (occurring in lymphoma patients), and others. Solid tumors were further grouped in genitourinary, gastrointestinal/hepatopancreaticobiliary (GI/HPB), lung, gynecological, melanoma, sarcoma/stromal, and other solid tumors.

Data Availability Statement

All data needed to evaluate the conclusions in the paper are present in the manuscript and/or the Supplementary Materials. Data from primary studies are publicly available within the databases described in Supplementary Information. Original output data are available from the corresponding author upon reasonable request. In case of further questions, please contact the corresponding author.

Code Availability Statement

All codes were adapted using R software, v.4.3.1 (meta package 7.0–0, metafor package 4.4–0). For further questions, please contact the corresponding author.

Results

Study Cohort

We reviewed 1031 studies to identify SPM cases occurring in patients undergoing CAR T-cell therapy. In total, 811 reports were screened, 235 full-text articles were examined for eligibility, and 25 articles met criteria for further analysis (Fig. S2). The study population included 18 reports on clinical trials (CT) spanning a total of 1,991 patients and 7 real-world studies (RWS) comprising 3,526 patients (Table 1).^18,39–62 Three articles reported on multiple disease entities,^43,56,62 and one study reported on multiple CAR-T products.¹⁸ These studies were subdivided into sub-cohorts, resulting in 34 distinct patient cohorts for the final evaluation.

Table 1.

Characteristics of included records

Author	Year	Cohort	Entity	Setting	Product	Patient number	number of SPM	Follow-up [months]	Median age [years]	Prior therapy lines [median]	Previous HSCT [%]
Abramson et al.	2023		LBCL	CT	Liso-cel	89	3	17.5	60	1	0
Abramson et al.	2024		LBCL	CT	Liso-cel	270	22	19.9	63	3	35
Berning et al.	2024		LBCL	RWS	Tisa-cel + Axi-cel	172	1	8.3	65	3	32.5
Cappell et al.	2020	A	LBCL	CT	Axi-cel	28	3	42	51	4	35.7
		B	FL+MZL			7	2		57	4	42.9
		C	CLL			7	2		61	4	0
Chong et al.	2021		LBCL+FL	CT	Tisa-cel	38	6	60.7	NR	5	34.2
Dreyling et al.	2024		IL	CT	Tisa-cel	97	2	28.9	57	4	36.1
Ghilardi et al.	2024	A	LBCL	RWS	Tisa-cel + Axi-cel + Liso-cel	259	11	10.3	NR	NR	35.2
		B	MCL		Brexu-cel	28	2
		C	IL		Tisa-cel + Axi-cel + Liso-cel	30	0
		D	MM		Ide-cel	67	3
		E	MM		Cilta-cel	58	0
Jacobson et al.	2022		LBCL	RWS	Axi-cel	1297	50	12.9	62.1	3	28
Jaeger et al.	2022		LBCL	CT	Tisa-cel	115	9	32.6	56	3	49
Kato et al.	2023		LBCL	CT	Axi-cel	16	0	13.4	58	NR	37.5
Lin et al.	2023		MM	CT	Ide-cel	62	8	18.1	61	6	91.9
Martin et al.a	2023		MM	CT	Cilta-cel	97	15	27.7	61	6	90
Melody et al.	2024	A	LBCL	RWS	Axi-cel	360	26	35.3	NR	3	38.5
		B			Tisa-cel	152	9
		C			Liso-cel	70	10
Neelapu et al.b	2023		LBCL	CT	Axi-cel	101	5	63.1	58	3	25
Neelapu et al.	2024	A	FL	CT	Axi-cel	124	13	41.7	60	3	24
		B	MZL			28	5	31.8	64	3	13
Pasquini et al.	2020		LBCL	RWS	Tisa-cel	155	6	11.9	65.4	4	28.4
Rodriguez-Otero et al.	2023		MM	CT	Ide-cel	225	13	18.6	63	3	84
San Miguel et al.	2023		MM	CT	Cilta-cel	176	9	15.9	61.5	2	NR
Sehgal et al.c	2023		LBCL	CT	Liso-cel	61	2	18.2	74	1	0
Shah et al.d	2024		CLL	CT	Liso-cel	118	11	23.5	65	5	6
Sidana et al.	2023		MM	RWS	Ide-cel	603	27	6.6	65	7	NR
Spiegel et al.	2023		LBCL	RWS	Axi-cel	275	23	58	60	3	35.3
Wang et al.	2024		MCL	CT	Liso-cel	88	3	16.1	68.5	3	33
Westin et al.e	2023		LBCL	CT	Axi-cel	170	8	47.2	58	1	0
Xu et al.	2024		MM	CT	Cilta-cel	74	4	65.4	54.5	3	24.3

Abbreviations: IL = indolent lymphoma (including follicular lymphoma [FL], marginal zone lymphoma [MZL] and chronic lymphocytic leukemia [CLL]), LBCL = large B-cell lymphoma, MCL = mantle cell lymphoma, MM = multiple myeloma, NR = not reported, SPM = second primary malignancies, RWS = real world study, CT = clinical trials.

^aPatient characteristics reported in Berdeja et al. 2021,

^bPatient characteristics reported in Locke et al. 2019,

^cPatient characteristics reported in Sehgal et al. 2022,

^dPatient characteristics reported in Siddiqi et al. 2023,

^ePatient characteristics reported in Locke et al. 2022

Overall, the most common primary malignancy was LBCL (3,614 patients), followed by MM (1,362 patients), IL (425 patients), and MCL (116 patients). The number of studies reporting on each CAR-T product was as follows: 14 for axi-cel, 8 for tisa-cel, 8 for liso-cel, 1 for brexu-cel, 4 for ide-cel, and 4 for cilta-cel. One study did not distinguish between axi-cel and tisa-cel when assessing SPM rates and was therefore excluded from product-specific comparisons.⁴⁰ Follow-up times varied from 6.6 to 65.4 months.

SPM estimates do not vary significantly across disease entities and CAR-T products

Across all patients, the overall SPM point estimate was 5.8% (95% CI 4.7–7.2%) with a median follow-up of 21.7 months. Study heterogeneity was moderate-to-high (I² = 61%, Fig. 1). We neither identified a significant risk of publication bias by funnel plot analysis (Egger’s test p=0.79) nor a significant risk of study bias among the included studies (Table S1). SPM point estimates did not significantly differ between MCL (4.3%, 95% CI 1.8–9.9%), LBCL (5.3%, 95% CI 4.3–6.9%), MM (6.0%, 95% CI 3.8–9.1%), and IL (8.7%, 95% CI 4.3–16.8% p(χ2)=0.56; Fig. 1). Additionally, the different CAR-T products did not significantly impact SPM point estimates (Cilta-cel 5.2%, Ide-cel 5.5%, Tisa-cel 5.8%, Liso-cel 6.7%, Brexu-cel 7.1%, Axi-cel 7.2%; p=0.95, Fig. S3).

Figure 1. Forest plot of SPM point estimates across all study cohorts and stratified by disease entity.

Forest plot illustrating SPM point estimates and 95% confidence intervals (95% CI) stratified by disease entity. The results of the random effects model and measures of heterogeneity are shown for each disease entity. Data is presented as point estimate proportion with error bars showing 95% CI. The p-value for the comparison of subgroups was derived from a two-sided test for subgroup differences (random effects model). Heterogeneity measures including I² are depicted (I² between 50% to 75% indicates moderate-to-high study heterogeneity). Abbreviations: IL = indolent lymphoma, LBCL = large B-cell lymphoma, MCL = mantle cell lymphoma, MM = multiple myeloma, SPM = second primary malignancy.

Influence of study features on SPM point estimates

To assess the influence of possible study features on SPM development, we next examined SPM point estimates by follow-up, number of prior treatment lines, prior hematopoietic cell transplantation (HCT), age, and treatment setting (Fig. S4–8). Unsurprisingly, we found a significant positive correlation between follow-up time and SPM estimates (r=0.44, p=0.009, Fig. S9). Accordingly, follow-up greater than the median of 21.7 months was associated with higher SPM point estimates than less than median follow-up (8.5% vs. 4.2%, p<0.001, Fig. 2A). Studies with an increased proportion of patients with previous HCT (e.g., above the median of 35.2%) also had significantly higher SPM point estimates (8.1% vs. 5.0%, p=0.02, Fig. 2B). Similarly, exposure to more than 3 previous therapy lines was associated with numerically higher SPM point estimates compared to the studies with a median of 3 or fewer (8.7% vs. 5.7%, p=0.13, Fig. 2C). Paradoxically, studies with a median age below 61 years were associated with higher SPM point estimates (7.5% vs. 4.9%, p=0.06, Fig. 2D). However, these studies also had longer follow-up (Fig. S10). Finally, patients treated within a clinical trial reported significantly higher SPM point estimate than the patients in RWS (7.3% vs. 4.6%, p=0.04, Fig. 2E).

Figure 2. Subgroup analyses of SPM point estimates.

Comparison of aggregated SPM point estimates with 95% CIs for A follow-up time (above median n = 17, below median n = 17), B inclusion of patients with prior HCT (> 35% n = 12, < 35% n = 20), C number of prior lines (> 3 n = 10, < 3 n = 19), D age (< 61 years n = 14, > 61 years n = 11), and E treatment setting (CT n = 21, RW n = 13). To compare follow-up times, prior HCT and therapy line exposure as well as for age, we classified studies using the respective median values. P-values for the comparison of SPM point estimates were derived from the test for subgroup differences (random effects model). Size of included patients per study is illustrated by point size. Abbreviations: CT = clinical trial, HCT = hematopoietic cell therapy, SPM = secondary primary malignancy, RW = real-world.

To account for potential collinearity across study features, we next performed a multivariate meta-regression analysis incorporating the above covariates (p<0.2 on univariate analysis, Fig. 2). This analysis revealed CAR-T treatment in clinical trials (p=0.049), extended follow-up (p=0.035), and a higher number of previous therapy lines (p=0.016) to be independent factors associated with increased SPM (Table 2). Conversely, age and prior HCT were not retained as independent SPM predictors.

Table 2.

Meta-regression analysis

Covariate*	Estimate	Confidence interval		p-value
		Lower limit	Upper limit
Follow-up (months)	0.016	0.001	0.031	0.035
Prior HCT (% of patients)	0.001	−0.009	0.011	0.878
Prior Lines (number)	0.268	0.056	0.479	0.016
Age (years)	0.016	−0.063	0.095	0.669
Treatment setting  - Real-world [ref.]  - Clinical Trials	0.415	0.003	0.828	0.049

Abbreviations: HCT = hematopoietic cell transplantation.

Follow-up is measured in months, Prior HCT as percentage of the respective study cohort, prior lines is the median number of previous therapies received by the study cohort, age demarks the median age, reference category for treatment setting is RWS.

^*23 studies with complete data for meta-regression

Subgroup analysis of randomized clinical trials suggests similar SPM rates between CAR-T and standard-of-care therapy.

While most of the published clinical trials evaluating CAR T-cells have reported on single arm studies, there have so far been five published randomized clinical trials investigating CAR-T therapy against the current standard-of-care. Due to the direct comparison, these trials are of particular interest as they enable an assessment of the relative risk of developing SPMs with CAR-T compared to established therapies, in patients with similar underlying diseases and exposure to similar prior therapies. We were able to extract SPM information for four of these trials. KarMMa-3 and CARTITUDE-4 randomized the BCMA-directed CAR T-cell products Ide-cel and Cilta-cel against common treatment strategies in patients with relapsed/refractory MM. In the ZUMA-7 and TRANSFORM trials, LBCL patients were randomized to receive CD19 CAR-T or standard-of-care salvage therapy followed by autologous HCT in chemosensitive patients. Information from the negative BELINDA trial examining tisagenlecleucel in second-line LBCL patients was unavailable. Using a meta-analysis of binomial outcomes including a total of 1,253 patients, we detected a pooled odds ratio of 1.04 (95% CI 0.32–3.41, p=0.92, Fig. 3), indicating that SPMs occurred with the same frequency in both arms. The SPM frequency in the CAR-T arms was 5.0% (95% CI 3.6–6.9%) compared to 4.9% (95% CI 3.4–6.9%) in the SOC arms. The study heterogeneity was low-to-moderate (I² = 35%, p=0.2) despite inclusion of different entities, CAR-T products, and comparator strategies.

Figure 3. Forest plot of SPM point estimates in CAR-T versus standard-of-care arms in a subgroup of randomized clinical trials.

Forest plot illustrating Odds ratios and 95% confidence intervals (95% CI) for the occurrence of SPM in comparison of CAR-T therapy versus standard-of-care (SOC) derived from randomized clinical trials. Pooled Odds ratio was calculated based on a meta-analysis of binary outcomes using a random-effects model. Data is presented as Odds ratio with error bars showing 95% CI. Heterogeneity measures including I² are depicted (I² between 25% to 50% indicates low-to-moderate study heterogeneity). Abbreviations: OR = Odds ratio, SOC = standard-of-care, SPM = second primary malignancy.

Hematological malignancies are the most prevalent SPM entity after CAR-T therapy

To examine the patterns of SPMs arising after CAR-T cell therapy, we next performed a detailed analysis of the distribution of SPM subtypes. The most prevalent disease entities were hematological malignancies (121 events, 37%), followed by solid tumors (79 events, 27%) and non-melanoma skin cancers (61 events, 16%) (Fig. 4A); 65 SPM cases (20%) were not further specified in the respective studies. Myelodysplastic syndromes (MDS) represented the predominant hematological malignancy (61.2%), followed by acute myeloid leukemia (AML, 14.0%), MDS/AML (12.4%), and MPN (4.1%) (Fig. 4B, Table S2). Only 5 cases of T-cell malignancy were described, accounting for just 4.1% of all hematological SPMs and 1.5% of all SPMs. Of note, the point estimate for T-cell malignancies across the entire study population was exceedingly low at 0.09% (95% CI 0.04%−0.2%). Of the five reported cases of T-cell lymphomas, three were tested for the presence of CAR transgene, with only the case from the CARTITUDE-4 trial being classified as positive in a preliminary report.²⁰

Figure 4. Characterization and distribution of SPM subtypes across disease entities and CAR-T products.

A Donut plot displaying major SPM subtypes among the entire study cohort: hematologic (red), solid tumors (blue), non-melanoma skin cancers (dark grey) and not further specified SPM (light grey). B-C Pie charts subdivide hematologic SPMs (red tones, B) and solid tumors (blue tones, C). D-E Comparison of distribution of hematologic and solid SPMs (excluding non-melanoma skin cancers) across disease entities (D) and between individual CAR-T products (E). Studies reporting on multiple entities or CAR-T products without further separation were excluded from the analysis. Chi-square distribution test was used for statistical testing. Abbreviations: MDS = myelodysplastic syndrome, AML = acute myeloid leukemia, MPN = myeloproliferative neoplasms, T/B-NHL = T/B cell non-Hodgkin lymphoma, HPB = hepatopancreaticobiliary, GIST = gastrointestinal stromal tumor, NOS = not otherwise specified, SPM = second primary malignancy, MCL = mantle cell lymphoma, IL = indolent lymphoma, MM = multiple myeloma, LBCL = large B cell lymphoma.

Solid tumors following CAR-T cell therapy encompassed a diverse range of subtypes (Fig. 4C, Table S3). Genitourinary tumors were the most prevalent, accounting for 16 cases (20.3%), predominantly prostate cancer (12 cases), and 4 cases of bladder cancer. The gastrointestinal and hepatopancreatobiliary (GI/HPB) category followed, representing 14 cases (17.7%). Within this group, colorectal cancer (CRC) was the most common (4 cases), with other cancers including anal, esophageal, hepatocellular carcinoma (HCC), anal/rectal cancer not otherwise specified (NOS), appendix cancer, pancreatic cancer, and small intestine cancer each contributing between one to two cases (Table S3). Lung cancers were noted in 10 cases (12.7%), predominantly non-small cell lung cancer (NSCLC) (7 cases). Melanoma was also identified in 10 cases (12.7%). Eight patients developed gynecologic tumors (10.1%), with breast cancer being the most frequent (5 cases), followed by endometrial cancer (2 cases) and cervical cancer (1 case). Sarcomas and gastrointestinal stromal tumors (GIST) comprised 5 cases (6.33%), distributed evenly among angiosarcoma, GIST, histiocytic sarcoma, myxofibrosarcoma, and spindle cell sarcoma. Other solid tumors also numbered 5 cases (6.33%), including adenocarcinoma NOS, laryngeal cancer, mesothelioma, neuroendocrine carcinoma, and neuroendocrine tumor.

Finally, we compared the relative distribution of SPMs by disease entity (Fig. 4D) and CAR-T product (Fig. 4E). We observed that hematologic malignancies predominated relative to solid tumors for LBCL patients (68.2%) and with the CAR T-cell products axicabtagene ciloleucel (69.3%) and lisocabtagene maraleucel (66.6%).

Discussion

This systematic review and meta-analysis provides a comprehensive overview of the incidence and distribution patterns of SPMs following CAR T-cell therapy in patients with hematologic malignancies. Our findings suggest that the risk factors driving SPM development are multi-faceted, highlighting the need for a nuanced discussion with patients about the myriad of potential contributing factors before attributing causation to the CAR T-cells themselves.

Our analysis revealed an overall SPM point estimate of 5.8% after a median follow-up of approximately 2 years. This estimate aligns with SPM rates reported in previous studies. A recently published analysis of the FAERS database found 536 SPM reports among 12,394 adverse event reports from CAR T-cell patients, equating to 4.3%.²³ In a report by Levine and colleagues incorporating topline results from more than 11,000 patients within the CIBMTR registry, an identical proportion of patients (4.3%) developed SPMs.¹⁷ However, neither publication reported risk factors for SPM. Interestingly, our analysis did not identify a significant difference in SPM rates between different CAR T-cell products or disease entities, indicating that the risk of developing SPMs may be more closely related to patient-specific factors and treatment history rather than the specific CAR T-cell product. Correspondingly, our meta-regression analysis identified several independent factors associated with increased SPM risk: higher number of prior therapy lines, extended follow-up time and treatment within clinical trials.

The risk of SPM after CAR-T therapy is unlikely to stem from a single risk factor; instead, their etiologies are complex and multifactorial. Patients’ exposure to lymphodepleting chemotherapy prior to CAR T-cell infusion and the associated risk of genomic instability therein is likely contributory. For example, the rate of secondary malignancy after conventional first-line ‘FCR’ chemoimmunotherapy in chronic lymphocytic leukemia patients (a similar fludarabine and cyclophosphamide dosing regimen to that used in lymphodepletion prior to CAR-T) has been reported to be as high as 33.7% at 4 years.⁶³ Indeed, we noted numerically higher SPM point estimates with the CAR-T products that have higher lymphodepletion doses (Figure S2). Furthermore, in the current indications, CAR T-cell patients have often been exposed to multiple lines of pro-tumorigenic therapies, including alkylating agents, lenalidomide, and the high-dose chemotherapy included prior to autologous HCT.^64–66 Other potential risk factors include pre-existing precursor conditions such as Clonal Hematopoiesis of Indeterminate Potential (CHiP), which may transform to overt myeloid neoplasia in the setting of CAR T-cells and the associated hyperinflammation and/or immunosuppression (e.g., decreased immune surveillance).^67–70 Notably, extended follow-up was associated with significantly higher SPM. One consequence of CAR-T therapy may thus be that precisely its efficacy provides patients the life-years to develop other diseases, including secondary malignancies. Given that patients in clinical trials are typically younger and healthier than the average patient with a given disease, the observation that studies conducted within clinical trials reported higher SPM point estimates compared to real-world studies may be a result of closer follow-up and reporting mechanisms in clinical trials than in RWS.⁷¹ However, this also raises concerns about potential underreporting in real-world settings, underscoring the need for standardized and comprehensive reporting practices.⁷²

A comparison of the derived SPM estimate with the incidence of SPM following other therapies found no increased signal following CAR T-cell therapy. For example, a study of DLBCL patients reported a 5-year SPM incidence of approximately 5% in the first line setting.⁷³ In the Childhood Cancer Survivor Study cohort, the 30-year cumulative incidence of SPM in NHL survivors was around 12%.⁷⁴ For patients undergoing autologous HCT, the 10-year SPM incidence has been estimated at about 21%.⁷⁵ Notably, the risk of secondary hematological malignancies is higher with high-dose chemotherapy followed by autologous HCT compared to other therapies.^65,76 Interestingly, our subgroup analysis of the limited number of randomized phase III studies found no significant increase in the risk of SPM for CAR-T cell treated patients compared to those who received standard-of-care therapies (including auto-HCT for LBCL patients). Furthermore, the meta-regression analysis revealed that prior HCT was not an independent risk factor for SPM when controlling for the number of previous therapies. Instead, the extent of pre-treatment was a more relevant factor. Taken together, these data do not suggest that CAR-T therapy is intrinsically associated with an increased risk for SPM development, rather that the population of patients receiving this therapy is often already at high risk for SPMs (e.g., advanced age, multiple prior lines of potentially oncogenic therapies).

The distribution of SPM entities after CAR T-cell therapy presents a diverse range of categories. Similar to the FAERS database studies^19,23, hematological malignancies were the most frequent SPMs, followed by solid tumors including non-melanoma skin neoplasms. Solid tumors following CAR T-cell therapy also showed a varied distribution. However, the observed rate of prostate cancer (0.22%), colorectal cancer (0.07%) and breast cancer (0.09%) in our study cohort was comparable to the annual incidences in populations between 60 and 64 years in the US (prostate cancer: 0.38% of males, CRC: 0.076%, breast cancer: 0.31% of females), suggesting that CAR-T cell therapy does not play a substantial role in the development of these malignancies.

Our study identified only 5 cases of T-cell malignancies (TCMs), equaling to a point estimate of only 0.09% across all CAR-T recipients. This is in line with a recent report by Verdun and Marks from the FDA outlining 22 TCM cases across more than 27,000 CAR-T treated patients²², resulting in a similar point estimate of 0.08%. Despite intensive and ongoing investigations, only a handful of TCM cases have been classified as CAR positive to date.^20,22,77 In addition, several case reports performing in-depth tissue analysis have questioned to what extent insertional mutagenesis contributed to TCM development.^18,78 Furthermore, the underlying bidirectional risk for TCMs in patients with B-cell malignancies,⁷⁹ and the aforementioned exposure to prior tumorigenic therapies, need to be considered. Ultimately, the low frequency of TCMs, the small proportion of CAR positive cases, and the inconclusive evidence regarding the pathogenetic significance of CAR vector insertion question the significant media attention that has followed the announcement of the FDA warning label. Nonetheless, the reporting of long-term adverse events, including T-cell lymphomas, remains pivotal in this vulnerable patient population and further studies investigating the specific mechanisms of CAR-driven insertional oncogenesis are required.

Our study has several limitations. The heterogeneity among included studies, particularly regarding follow-up duration and the included patient populations, may have introduced variability in the reported SPM rates. Furthermore, the potential overlap of patient cohorts in some real-world studies cannot be entirely excluded, which may have affected our estimates. Additionally, SPM entities could not be attributed in ~20% of cases (“others” and “unknown” groups), even in the setting of clinical trials where patients are ideally followed closely given their experimental nature. Some key clinical trials could not be included in this meta-analysis due to insufficient reporting,^80–84 demonstrating the necessity of integrating SPM in a structured reporting system. The lack of individual patient data precluded adjustment for specific covariates, such as comorbidities and prior treatment details, which could influence SPM development. Overall, our study does not allow, by design, the establishment of causal relationships. To account for potential differences related to the statistic modelling (random effects vs. fixed effects modelling), we performed additional sensitivity analyses, demonstrating stable study results (Table S4).

Future research should focus on elucidating the true mechanistic link between CAR T-cell therapy and SPMs, considering the roles of prior therapies, immune suppression, and genomic instability. Additionally, there is a need for large-scale, long-term studies to better understand the incidence and risk factors of SPMs in diverse patient populations. Enhanced pharmacovigilance and standardized reporting practices will be crucial in improving the detection and management of SPMs in patients undergoing CAR T-cell therapy.

In conclusion, this systematic review and meta-analysis highlights the importance of recognizing and addressing the risk of SPMs following CAR T-cell therapy. By understanding the factors contributing to SPM development and implementing effective monitoring strategies, healthcare providers can optimize patient outcomes and ensure the long-term safety of CAR T-cell therapies.

Translational Relevance.

Second primary malignancies (SPMs) arising after CAR T-cell therapy have recently garnered considerable research interest following the FDA’s issuance of a class-wide blackbox warning. Here, we identify a SPM frequency of approximately 6% at two years and demonstrate that secondary myeloid neoplasms represent the predominant subtype. This information will help inform conversations with patients regarding expected long-term adverse events. Additionally, our meta-analysis indicates that the frequency of SPMs with CAR-T is not increased when compared to previous standard-of-care strategies. We found that SPM occurrence after CAR-T therapy is associated with the duration of follow-up, the number of prior therapy lines, and treatment in a clinical trial setting, but not the disease entity or CAR-T product. Ultimately, these findings question the FDA’s class-wide warning, suggesting that SPM development rather reflects exposure to previous therapies and extended follow-up than being intrinsically linked to CAR T-cell therapy itself.