Second Primary Malignancy Risk in Patients with Multiple Myeloma Receiving CAR T-cell Therapy or Other Systemic Anticancer Treatments: A Real-World Comparative Study

Abstract

Purpose:

This study aims to compare the risk of second primary malignancy (SPM) between patients with multiple myeloma who received chimeric antigen receptor T-cell (CAR T) therapy versus other systemic anticancer therapies (SACT).

Experimental Design:

Adult patients with multiple myeloma who initiated CAR T therapy or other SACT were identified from Komodo Health claims data and weighted to balance baseline characteristics. Cumulative incidence of SPM was estimated over 24 months, and P values were calculated for the difference between 0 and 24 months.

Results:

The study included 435 patients who received CAR T therapy and 12,268 patients who received other SACT (median follow-up, 11.8 months). Compared with other SACT, CAR T therapy was associated with similar risks of any SPM (at 24 months, 24.1% vs. 22.3%; P_{0–24 months} = 0.31) and solid SPM (9.1% vs. 11.5%, P_{0–24 months} = 0.32) but significantly higher risk of hematologic SPM (17.9% vs. 13.1%, P_{0–24 months} = 0.04). In a sensitivity analysis requiring ≥ 2 claims to identify an SPM, difference in hematologic SPM risk was attenuated (5.5% vs. 4.9%, P_{0–24 months} = 0.08). Notably, bone marrow examinations were more common after CAR T therapy (e.g., 47% vs. 13% at 0–3 months).

Conclusions:

In this real-world dataset with relatively short follow-up, patients with multiple myeloma seemed to have a higher risk of hematologic SPM after CAR T therapy compared with other SACT. However, misclassification and detection bias cannot be ruled out. The association warrants further evaluation. Physicians should be vigilant for myeloid malignancies after CAR T therapy.

Translational Relevance.

Cases of second primary malignancies (SPM) have been reported following chimeric antigen receptor T-cell (CAR T) therapy. In 2024, the U.S. Food and Drug Administration issued a class-wide boxed warning for all CAR T therapies on “serious risk of T-cell malignancy.” However, evidence has been lacking regarding whether CAR T therapy was associated with the risk of SPM. In a large U.S. claims database, we found a substantial risk of SPM in heavily pretreated patients with multiple myeloma. Compared with other systemic therapies, CAR T therapy was associated with an increased risk of hematologic SPM, particularly myeloid malignancies, within 24 months of follow-up. However, potential detection bias cannot be ruled out, as suggested by more common bone marrow examinations following CAR T therapy compared with other systemic therapies. The association warrants further evaluation. The findings suggest that treating physicians should be vigilant for myeloid malignancies after CAR T therapy.

Introduction

Multiple myeloma is a malignancy of plasma cells. Currently, approximately 35,000 new cases of multiple myeloma occur each year in the United States (1), and 188,000 occur worldwide (2). Given improvement in myeloma treatment resulting in substantially longer survival of patients with multiple myeloma (3), it has become increasingly important to understand longer-term treatment outcomes, including the potential of increased risk of developing second primary malignancies (SPM).

In population-based cohorts (4–7), patients with multiple myeloma are more likely to develop hematologic malignancies compared with the general population, and the risk of myeloid malignancies such as acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) is particularly elevated (seven- to 12-fold). In addition to factors intrinsic to the patient and the disease, treatments for multiple myeloma are also associated with increased SPM risk. A high cumulative dose of alkylators, particularly melphalan, is known to contribute to the development of acute leukemia (8). Furthermore, prolonged lenalidomide maintenance therapy has been independently associated with modest but statistically significant increases in SPM risk (8).

Chimeric antigen receptor T-cell (CAR T) therapies are a newer class of therapy. So far, 6 CAR T products, including 2 for multiple myeloma treatment [idecabtagene vicleucel (ide-cel) and ciltacabtagene autoleucel (cilta-cel)], have been approved by the U.S. Food and Drug Administration (FDA). Cases of T-cell malignancies have been reported following CAR T therapy, including several CAR transgene-positive cases, which suggests a possible role of the CAR T product in the development of some malignancies (9). As a result, the FDA announced an investigation (10) and later issued a class-wide boxed warning for “serious risk of T-cell malignancy” after CAR T therapy (11). Overall, non–multiple myeloma malignancies (prior or secondary) account for 8% of nonrelapse mortality following CAR T therapies in patients with multiple myeloma (12).

To date, only 2 clinical trials have directly compared the risk of SPM following CAR T therapy versus other systemic anticancer therapy (SACT) in patients with multiple myeloma (13, 14), but the sample sizes were relatively small. To the best of our knowledge, this is the first real-world (RW) study that performed this comparison in a large U.S. claims database.

Materials and Methods

Study population

The study was conducted in the Komodo Health database, which contains deidentified medical and pharmacy claims on > 83 million enrollees since 2006. The database includes open and closed claims, which are preadjudicated claims sourced from clearinghouses and providers and adjudicated claims sourced from payers, respectively. The study data cutoff was April 30, 2024. Approval from the Institutional Review Board was not required as the study used existing deidentified data.

Patients ages 18 years or older with newly diagnosed multiple myeloma were identified using ≥ 1 inpatient or ≥ 2 outpatient International Classification of Disease (ICD)-10 code (C90.0X) between January 1, 2017, and January 31, 2024. They were required to have a look-back period of at least 1 year of continuous medical or pharmacy insurance coverage benefits before multiple myeloma diagnosis, permitting gaps ≤ 30 days.

Index line of therapy of CAR T and other SACT

Lines of therapy (LoT) were identified based on administration or dispensing data of multiple myeloma systemic therapies (Supplementary Methods S1) using an algorithm (Supplementary Methods S2) that accounts for conventional SACT and has specific considerations for stem cell transplant (SCT) and CAR T therapies. Among eligible patients, index LoT in second or later lines (2L+) that started between March 26, 2021, the date of the first FDA approval of a CAR T therapy for multiple myeloma treatment, and January 31, 2024, were eligible for inclusion. To ensure that the index treatment is for multiple myeloma, patients must have a diagnosis of multiple myeloma on or within 30 days before the index date (i.e., the LoT start date) and no diagnosis of other hematologic malignancies on or within 90 days before the index date. In addition, they must have medical insurance coverage on the index date. The first CAR T LoT in each person was included, and any LoT after the CAR T LoT were excluded. One eligible other SACT LoT per person was selected using stratified random sampling to mimic the line number distribution of the CAR T LoT. As a result, each patient could contribute up to one CAR T LoT and one other SACT LoT.

Outcome and covariate ascertainment

The outcome of interest was incident SPM other than nonmelanoma skin cancer (NMSC), ascertained by ≥ 1 claim with a diagnostic code for a new malignancy (Supplementary Methods S3). Follow-up began on the index date and continued until the earliest of the following: SPM, death, disenrollment of medical coverage, start of a subsequent CAR T treatment, 24 months of follow-up, or data cutoff date; 24 months was chosen as a practical time point after which the at-risk sample size was too small for valid estimation. To ensure that the SPM diagnoses were incident events, we excluded preexisting malignancies, defined as those with a related 3-digit ICD-10 code present during the baseline period, that is, between 12 months before multiple myeloma diagnosis and the index date. The primary outcomes were any SPM, hematologic SPM, and solid SPM. Specific types of SPM were also evaluated, which included B-cell malignancies, T-cell malignancies, MDS, and other/unspecified hematologic malignancies among hematologic SPM, as well as breast cancer, lung cancer, colorectal cancer, melanoma, and other solid malignancies among solid SPM.

Covariates, including patient demographics, lifestyle, and health histories, were ascertained in various windows during the baseline period (Supplementary Methods S4). Prior treatments for multiple myeloma were ascertained between multiple myeloma diagnosis and the index date.

Statistical analysis

Baseline characteristics for each index LoT were summarized using frequencies and means [standard deviations (SD)]. Standardized mean differences (SMD) were used to evaluate balance in baseline covariates between the treatment groups, whereby good balance was defined as SMD between −0.1 and 0.1. The other SACT group was weighted with inverse probability of treatment weighting (IPTW) constructed to achieve a similar distribution of potential risk factors of SPM (Supplementary Methods S4) as the CAR T group, so that all subsequent analyses produce estimates of the average treatment effect on the treated (15).

The main analysis estimated the cumulative incidence and incidence rate of SPM through 24 months of follow-up. The cumulative incidence of SPM was estimated with death as a competing risk using an IPTW Aalen–Johansen estimator. Its variance was estimated with nonparametric bootstrap to account for correlation between LoT from the same individual (16). An IPTW, bootstrapped Pepe–Flemming P value was used to test for differences in the areas under the cumulative incidence curves during 0 to 24 months between treatment groups (P_{0–24 months}; refs. 17, 18). The incidence rate of SPM was calculated per 100 person-years (PY) at risk, overall, and stratified by time from the index date (0–3, > 3 to 6, > 6 to 12, and > 12 to 24 months). An additional analysis was performed in which the cumulative incidence of SPM was estimated using an IPTW Kaplan–Meier estimator with death as a censoring criterion, addressing potential bias due to differential survival.

Two prespecified sensitivity analyses were conducted: (i) requiring ≥ 2 claims to identify an SPM (same three-digit ICD-10 code) for improved specificity (19) and (ii) excluding SPM diagnosis in the first 3 months of follow-up to minimize the misclassification of potential preexisting cancers. In addition, several post hoc analyses were performed to help interpret the findings. To explore potential differences in patient monitoring and diagnosis, the proportions of patients receiving blood smear and bone marrow examinations (BME, including both biopsy and aspiration) were described by time periods since the index date. In addition, the cumulative incidence of SPM after a 2-month landmark (when approximately half of the CAR T group received BME) was estimated by exposure group, stratified by receipt of postbaseline BME by the landmark; the analysis was conducted among patients who remained SPM-free and at-risk at the landmark and applied the same inverse probability of treatment weights as the main analysis (i.e., the weights were not reestimated). Lastly, the cumulative incidence of SPM and receipt of blood smear and BME were evaluated by individual CAR T therapies (cilta-cel, ide-cel, unspecific CAR T); the analysis was unweighted, given similar baseline characteristics across the groups. No statistical testing was performed in the post hoc analysis.

Statistical analyses used R version 4.2.1 (RRID: SCR_001905).

Results

The study included 435 CAR T LoT and 12,268 other SACT LoT (See Supplementary Fig. S1 for attrition). Patients were followed for a median of 11.8 months (IQR, 5.7–20.6) after the index date. The index CAR T therapies included cilta-cel (32%), ide-cel (36%), and unspecified CAR T (32%); the index other SACT included anti-CD38 monoclonal antibodies (53%), proteasome inhibitors (52%), immunomodulators (48%), and alkylators (16%), among other treatments.

Before weighting, patients who received CAR T therapy tended to be younger (61.8 vs. 66.3 years) than those who received other SACT. They were also more likely to be treated in 4L+ (83.5% vs. 26.2%), to have received certain therapies potentially relevant to SPM risks [autologous SCT (43.9% vs. 22.8%), radiotherapy (36.1% vs. 20.7%), alkylators (52.6% vs. 22.4%), and immunomodulators other than lenalidomide (77.7% vs. 21.4%)], and to have a prior hematologic (27.8% vs. 15.5%) or solid (43% vs. 33.5%) malignancy during the baseline period. After applying IPTW, all baseline characteristics were well balanced between the 2 groups, and the weighted sample size for the other SACT group was 437 LoT (Table 1).

Table 1.

Baseline characteristics among patients with multiple myeloma receiving CAR T therapy or other SACT, weighted and unweighted.

Baseline characteristic	Unweighted			Weighteda
	CAR T (N = 435)	Other SACT (N = 12,268)	SMD	CAR T (N = 435)	Other SACT (N = 437)a	SMD
Age at index date (years), mean (SD)	61.8 (8.6)	66.3 (11.1)	−0.46	61.8 (8.6)	62 (8.9)	−0.02
Sex	​	​	​	​	​	​
Female	45.7%	44.2%	0.02	45.7%	46.1%	0
Male	52.2%	53.5%	−0.01	52.2%	51.9%	0
Unknown	2.1%	2.3%	0	2.1%	2%	0
Race	​	​	​	​	​	​
White	57%	56.2%	0.01	57%	56.6%	0
African American	15.2%	18.8%	−0.04	15.2%	15%	0
Asian/Pacific Islander	2.5%	2.9%	0	2.5%	2.5%	0
Other	2.5%	3.4%	−0.01	2.5%	2.5%	0
Unknown	22.8%	18.7%	0.04	22.8%	23.4%	−0.01
Ethnicity	​	​	​	​	​	​
Hispanic	6.4%	8.4%	−0.02	6.4%	6.5%	0
Non-Hispanic	72%	72.4%	0	72%	71.5%	0
Unknown	21.6%	19.2%	0.02	21.6%	21.9%	0
Region	​	​	​	​	​	​
Northeast	30.1%	28.3%	0.02	30.1%	29.9%	0
Midwest	23.9%	24.1%	0	23.9%	23.9%	0
South	28.7%	24.8%	0.04	28.7%	28.9%	0
West	10.6%	16.1%	−0.06	10.6%	10.6%	0
Other	6.7%	6.8%	0	6.7%	6.7%	0
LoT	​	​	​	​	​	​
2L	6%	49.2%	−0.43	6%	6%	0
3L	10.6%	24.7%	−0.14	10.6%	10.5%	0
4L+	83.4%	26.2%	0.57	83.4%	83.6%	0
Time from diagnosis to index (days), mean (SD)	1,484 (643)	881 (694)	0.90	1,484 (643)	1,483 (629)	0
Prior multiple myeloma treatmentb,c	​	​	​	​	​	​
Autologous SCT	43.9%	22.8%	0.21	43.9%	43.8%	0
Allogenic SCT	2.8%	0.7%	0.02	2.8%	3%	0
Radiotherapy	36.1%	20.7%	0.15	36.1%	36.3%	0
Melphalan	14.3%	7.8%	0.06	14.3%	14.3%	0
Alkylator other than melphalan	52.6%	22.4%	0.30	52.6%	52.9%	0
Chemotherapy other than alkylator	7.6%	1.8%	0.06	7.6%	7.7%	0
Lenalidomide	74.9%	72.5%	0.02	74.9%	74.4%	0.01
Immunomodulators other than lenalidomide	77.7%	21.4%	0.56	77.7%	77.8%	0
Charlson Comorbidity Index	​	​	​	​	​	​
0	18.4%	21.5%	−0.03	18.4%	18.4%	0
1–2	37.5%	36.8%	0.01	37.5%	37.5%	0
3–4	29.4%	26.4%	0.03	29.4%	29.6%	0
5+	14.7%	15.2%	0	14.7%	14.5%	0
Prior malignancyc	​	​	​	​	​	​
Hematologic malignancy	27.8%	15.5%	0.12	27.8%	31.2%	−0.03
Solid malignancy	43%	33.5%	0.09	43%	42.8%	0

Abbreviations: 2L/3L/4L, second/third/fourth line; SD, standard deviation.

^aIPTW was applied to the other SACT group to be similar to the CAR T therapy group (see Supplementary Methods S4 for covariates). The weighted sample size of the other SACT group was 437 patients.

^bFrom multiple myeloma diagnosis to the index date.

^cThe categories do not add up to 100%.

A total of 56 and 1,170 patients had SPM through a maximum of 24 months in the unweighted CAR T and other SACT groups, respectively. The weighted incidence of SPM was 11.3% and 9.5% in the CAR T therapy and other SACT groups, respectively.

Compared with other SACT, CAR T therapy was associated with similar risk of any SPM [cumulative incidence (95% confidence interval) at 24 months: 24.1% (18.2%–31.9%) vs. 22.3% (19.2%–25.8%); P_{0–24 months} = 0.31] and solid SPM [9.1% (5.7%–13.1%) vs. 11.5% (9.2%–14.5%); P_{0–24 months} = 0.32] but significantly higher risk of hematologic SPM [17.9% (12.8%–24.5%) vs. 13.1% (10.6%–16.4%); P_{0–24 months} = 0.04] during 0 to 24 months of follow-up (Table 2; Fig. 1). Among hematologic SPM, the biggest difference was observed for MDS (4.3% vs. 2.8%) and other hematologic malignancies (8.8% vs. 6.7%; Table 2; Supplementary Table S1; in post hoc analyses, most of other hematologic malignancies were AML – 8.3% vs. 4.6%). When death was treated as a censoring criterion, the difference in hematologic SPM was similarly observed (P_{0–24 months} = 0.06; Supplementary Table S2). In the sensitivity analysis requiring ≥ 2 claims to identify an SPM, the cumulative incidence of SPM was reduced by approximately half in both exposure groups; the difference in the risk of hematologic SPM somewhat attenuated (5.5% vs. 4.9%; P_{0–24 months} = 0.08; Table 2; Fig. 1). In the sensitivity analysis excluding SPM diagnosis in the first 3 months, there was no difference in hematologic SPM risk (11.9% vs. 11.9%; P_{0–24 months} = 0.66; Supplementary Table S1). Among solid SPM, CAR T therapy was associated with a significantly lower risk of lung cancer (0.1% vs. 0.9%; P_{0–24 months} = 0.01; Table 2). However, the number of cases was small (0 vs. 2.3 cases), and there was no significant difference in solid SPM overall.

Table 2.

Weighted^a cumulative incidence of SPM at 24 months among patients with multiple myeloma receiving CAR T therapy or other SACT, main analysis and sensitivity analysis requiring 2+ diagnoses.

Outcome	CAR T (N = 435)		Other SACT (N = 437)a		P 0–24 months b
	# of events	Cumulative incidence at 24 months, % (95% CI)	# of events	Cumulative incidence at 24 months, % (95% CI)
Main analysis					​
Any SPM	56	24.1 (18.2–31.9)	49.6	22.3 (19.2–25.8)	0.31
Solid SPM	19	9.1 (5.7–13.1)	25.1	11.5 (9.2–14.5)	0.32
Breast	1	0.5 (0.1–1.1)	1.6	0.8 (0.4–1.5)	0.62
Lung	0	0.1 (0–0.3)	2.3	0.9 (0.4–1.3)	0.01
Prostate	1	0.7 (0.1–1.8)	1.9	0.9 (0.3–1.6)	0.64
Colorectal	3	1.9 (0.2–4.6)	2.3	1.6 (0.6–3.1)	0.59
Melanoma	1	0.5 (0.1–1.4)	2.6	1.5 (0.6–2.9)	0.22
Other solid	13	5.6 (3.3–8.6)	15.5	6.9 (5.2–9)	0.68
Hematologic SPM	42	17.9 (12.8–24.5)	29	13.1 (10.6–16.4)	0.04
B cell	11	5.9 (2.6–9.6)	11	5.9 (3.9–8)	0.81
T cell	2	0.6 (0–1.5)	0.5	0.3 (0.1–0.5)	0.25
MDS	12	4.3 (2.1–6.7)	5.8	2.8 (1.6–4.4)	0.06
Other hematologic	20	8.8 (4.3–14.4)	14.3	6.7 (4.7–9.2)	0.20
Sensitivity analysis requiring ≥ 2 diagnoses					​
Any SPM	28	11.5 (7.9–16)	19.7	9.7 (7.5–12.6)	0.05
Solid SPM	13	6.4 (3.4–9.9)	10.4	5.1 (3.6–6.9)	0.29
Breast	1	0.4 (0–1.1)	0.2	0.1 (0.1–0.2)	0.56
Lung	0	0.1 (0–0.1)	0.5	0.3 (0.1–0.5)	0.04
Prostate	1	0.7 (0.1–1.8)	1.3	0.6 (0.2–1.1)	0.99
Colorectal	3	2.1 (0.2–5.6)	2	1.7 (0.6–3.5)	0.52
Melanoma	0	0 (0–0.1)	0.5	0.2 (0.1–0.4)	0.07
Other solid	8	3.2 (1.5–5.1)	6.2	2.7 (1.8–3.7)	0.37
Hematologic SPM	16	5.5 (3.1–8)	9.7	4.9 (3.4–7)	0.08
B cell	2	0.7 (0.1–1.7)	3.6	2.4 (1.2–4.2)	0.35
T cell	0	0 (0–0)	0	0 (0–0)	0.38
MDS	7	2.8 (0.9–5)	2.6	1.5 (0.6–2.6)	0.09
Other hematologic	8	2.5 (0.9–4.3)	4	1.4 (0.8–2.3)	0.10

^aIPTW was performed on the other SACT group to be similar to the CAR T therapy group (see Supplementary Methods S4 for covariates). The weighted sample size of the other SACT group was 437 patients.

^b P value for difference in the area under the cumulative incidence curve during 0–24 months between the 2 groups.

Figure 1.

Weighted^a cumulative incidence of SPM among patients with multiple myeloma receiving CAR T therapy (N = 435) or other SACT (N = 437)^a, main analysis and sensitivity analysis requiring 2+ diagnoses. ^aIPTW was performed on the other SACT group to be similar to the CAR T therapy group (see Supplementary Methods S4 for covariates). The weighted sample size of the other SACT group was 437 patients.

The incidence rates per 100 PY (95% CI) of any SPM, solid SPM, and hematologic SPM were 22 (16.6–28.6) versus 18.9 (14–25), 7 (4.2–10.9) versus 9.1 (5.9–13.4), and 15.9 (11.4–21.5) versus 10.6 (7.1–15.2) following CAR T therapy and other SACT, respectively. Requiring ≥ 2 claims to identify an SPM reduced the rates by approximately half in both groups (Table 3). When stratified by time from the index date, the rate of SPM tended to decrease over time. The rate for solid SPM was highest at 0 to 3 and > 3 to 6 months for both treatment groups; the rate of hematologic SPM was highest at 0 to 3 months following CAR T therapy and > 3 to 6 months following other SACT (Fig. 2). Rates of specific SPM followed similar temporal patterns as described above (Supplementary Fig. S2).

Table 3.

Weighted^a incidence rate of SPM through 24 months among patients with multiple myeloma receiving CAR T therapy or other SACT, main analysis and sensitivity analysis requiring ≥ 2 claims.

Outcome	CAR T (N = 435)			Other SACT (N = 437)a
	# of events	PY at risk	Incidence rate, per 100 PY (95% CI)	# of events	PY at risk	Incidence rate, per 100 PY (95% CI)
Main analysis
Any SPM	56	254.7	22 (16.6–28.6)	49.6	262.3	18.9 (14–25)
Solid SPM	19	271.9	7 (4.2–10.9)	25.1	276.7	9.1 (5.9–13.4)
Breast	1	280.5	0.4 (0–2)	1.6	288.9	0.6 (0–2.3)
Lung	0	282.6	0 (0–1.3)	2.3	288.7	0.8 (0.1–2.7)
Prostate	1	282	0.4 (0–2)	1.9	288.9	0.7 (0.1–2.5)
Colorectal	3	281	1.1 (0.2–3.1)	2.3	289.1	0.8 (0.1–2.6)
Melanoma	1	282.4	0.4 (0–2)	2.6	288.1	0.9 (0.2–2.9)
Other solid	13	276.4	4.7 (2.5–8)	15.5	282.3	5.5 (3.1–9)
Hematologic SPM	42	264.6	15.9 (11.4–21.5)	29	273.2	10.6 (7.1–15.2)
B cell	11	277	4 (2–7.1)	11	283.7	3.9 (1.9–6.9)
T cell	2	280.6	0.7 (0.1–2.6)	0.5	289.5	0.2 (0–1.6)
MDS	12	278.2	4.3 (2.2–7.5)	5.8	286.9	2 (0.7–4.5)
Other hematologic	20	275.3	7.3 (4.4–11.2)	14.3	281.5	5.1 (2.8–8.5)
Sensitivity analysis requiring ≥ 2 claims
Any SPM	28	268.5	10.4 (6.9–15.1)	19.7	281.2	7 (4.3–10.8)
Solid SPM	13	273.7	4.8 (2.5–8.1)	10.4	285.4	3.6 (1.8–6.6)
Breast	1	280.6	0.4 (0–2)	0.2	289.8	0.1 (0–1.4)
Lung	0	282.6	0 (0–1.3)	0.5	289.5	0.2 (0–1.6)
Prostate	1	282	0.4 (0–2)	1.3	289.2	0.4 (0–2.1)
Colorectal	3	281.2	1.1 (0.2–3.1)	2	289.4	0.7 (0.1–2.5)
Melanoma	0	282.6	0 (0–1.3)	0.5	289.7	0.2 (0–1.6)
Other solid	8	277.8	2.9 (1.2–5.7)	6.2	287.3	2.2 (0.8–4.6)
Hematologic SPM	16	277.1	5.8 (3.3–9.4)	9.7	285.6	3.4 (1.6–6.3)
B cell	2	282.2	0.7 (0.1–2.6)	3.6	287.8	1.3 (0.3–3.4)
T cell	0	282.6	0 (0–1.3)	0	289.9	0 (0–1.3)
MDS	7	279.9	2.5 (1–5.2)	2.6	289	0.9 (0.2–2.8)
Other hematologic	8	279.3	2.9 (1.2–5.6)	4	288.4	1.4 (0.4–3.5)

^aIPTW was performed on the other SACT group to be similar to the CAR T therapy group (see Supplementary Methods S4 for covariates). The weighted sample size of the other SACT group was 437 patients.

Figure 2.

Weighted^a incidence rate per 100 PY (95% CI) of SPM in patients with multiple myeloma receiving CAR T therapy (N = 435) or other SACT (N = 437)^a, stratified by time since the index date, main analysis and sensitivity analysis requiring ≥2 claims. ^aIPTW was performed on the other SACT group to be similar to the CAR T therapy group (see Supplementary Method S4 for covariates). The weighted sample size of the other SACT group was 437 patients.

In the post hoc analysis, both blood smear and BME were more common in patients who received CAR T therapy than other SACT across time periods, particularly at 0 to 3 months (blood smear: 51.3% vs. 24.1%; BME: 47.4% vs. 13%; Fig. 3). When stratified by receipt of postbaseline BME by month 2, weighted cumulative incidences (95% CI) of hematologic SPM after month 2 did not differ between patients who received CAR T therapies versus other SACT, both among recipients [19.4% (4.9%–33.9%) vs. 17.5% (8.6%–26.3%) at 24 months] and among nonrecipients [8.9% (1.9%–15.8%) vs. 11.4% (8.4%–14.3%); Supplementary Table S3]. When stratified by individual CAR T therapies, unweighted cumulative incidences (95% CI) of hematologic SPM seemed highest in the cilta-cel group [27.7% (12.7%–42.6%)], followed by unspecific CAR T [16% (6.3%–25.8%)] and ide-cel [12.1% (4%–20.1%)], with wide and overlapping CI (Supplementary Table S4). Postbaseline, the cilta-cel group was also more likely to receive BME [e.g., 38% (cilta-cel), 16% (ide-cel), and 22% (unspecific) at 0–3 months] and less likely to receive blood smear [e.g., 53% (cilta-cel), 84% (ide-cel), and 81% (unspecific) at 0–3 months; Supplementary Fig. S3].

Figure 3.

Weighted^a proportions of patients with multiple myeloma who received blood smear and bone marrow aspiration/biopsy before and following treatment with CAR T therapy (N = 435) or other SACT (N = 437)^a. ^aIPTW was performed on the other SACT group to be similar to the CAR T therapy group (see Supplementary Method S4 for covariates). The weighted sample size of the other SACT group was 437 patients.

Discussion

To the best of our knowledge, this is the first RW study to directly compare the risk of SPM following CAR T therapy versus other SACT in patients with multiple myeloma. CAR T therapy was associated with a higher risk of hematologic SPM through 24 months. In a sensitivity analysis, the difference attenuated when ≥ 2 claims were required to identify an SPM. Blood smear and BME were also more common among patients who received CAR T therapy compared with other SACT, particularly within the first 3 months after therapy started.

This study reports a substantial risk of SPM among heavily pretreated patients with multiple myeloma drawn from a large national claims database. Specifically, 12.9% and 11.3% of patients developed SPM at a median of 11.8 months’ follow-up in the CAR T and other SACT groups, respectively. The incidences were 22 and 18.9 per 100 PY, respectively. Two other claims-based, RW studies in patients with triple-class exposed (TCE) multiple myeloma reported a similar incidence of SPM (excluding NMSC) at 14.8 to 19.3 per 100 PY (20, 21).

In comparison, clinical trials among patients with TCE multiple myeloma reported variable incidence of SPM (including NMSC), ranging from 1.2% to 16.5% over a median follow-up of 6 to 27 months (13, 14, 22–27). Furthermore, facility-based RW studies in patients receiving CAR T therapy for multiple indications reported a lower incidence of SPM, ranging from 1.7% to 4.3% over variable follow-up periods (28). In addition, trials and facility-based RW studies also reported a relatively low incidence of SPM among patients with various hematologic malignancies after CAR T therapies (approximately 6% over 2 years; refs. 29–31).

The exact reason for the relatively higher incidence of SPM estimated from claims-based RW studies is unknown. We speculate that the claims data may include suspected diagnoses of malignancies in which a confirmatory workup was not performed. In contrast, some clinical trials seem to have shorter SPM ascertainment windows than the current study, censored shortly after treatment discontinuation (22–25), whereas facility-based RW studies could miss malignancies diagnosed in a different setting.

The study found a higher risk of hematologic SPM associated with CAR T therapy through 24 months compared with other SACT, with the biggest difference in MDS and AML. The association cannot be explained by longer survival in patients who received CAR T therapy, as shown in the additional analysis in which death was treated as a censoring event (i.e., estimating risk in a hypothetical cohort in which no death occurs, thus accounting for potential bias due to differential survival). Of note, the study did not find any evidence of association between CAR T therapy and T-cell malignancies. However, it was specifically powered to detect a difference in such a low-incidence event. Furthermore, CAR T therapy was not associated with an increased risk of solid SPM.

In the randomized controlled trials (RCT), similar proportions of patients developed any SPM following CAR T therapies versus standard of care [SOC; 4.3% vs. 6.7% in CARTITUDE-4 (13) and 5.8% vs. 4% in KarMMa-3 (14)]. Similar findings were reported in a meta-analysis of RCT in multiple myeloma and large B-cell lymphoma (29). However, in both RCT for multiple myeloma treatment, hematologic SPM, primarily consisting of MDS and AML, only occurred in patients following CAR T therapy but not SOC [3/208 (1.4%) vs. 0/208 (0%) in CARTITUDE-4 (13) and 3/225 (1.3%) vs. 0/126 (0%) in KarMMa-3 (14)]. The sample size was too small to support statistical inferences. Similar to our study, the RCT in multiple myeloma did not find any difference in solid SPM between multiple myeloma treatment [3/208 (1.4%) vs. 4/208 (1.9%) in CARTITUDE-4 (13) and 6/225 (2.7%) vs. 3/126 in KarMMa-3 (2.4%; ref. 14)]. Prominence of myeloid malignancies (62%) after CAR T therapies was also supported by an analysis of 536 SPM cases from the FDA Adverse Events Reporting System (32).

In the sensitivity analysis requiring ≥ 2 claims to identify an SPM, the absolute risk of SPM was reduced by approximately half in both groups. The risk of hematologic SPM remained slightly higher following CAR T therapy. Two potential reasons may explain why a diagnosis of hematologic SPM was not repeated in a patient. First, the study had a relatively short follow-up; the patients may have been censored administratively or died before receiving a repeat diagnosis. Second, patients may have hematologic toxicities from index treatment that were initially interpreted as emerging MDS or AML but not confirmed later, or they may have had early-stage malignancies that did not require immediate attention in the context of advanced multiple myeloma. This result suggests that misclassification bias in the primary findings cannot be ruled out. In the other sensitivity analysis, excluding SPM diagnosis within the first 3 months, no difference in hematologic SPM was identified. However, the prespecified 3-month cutoff was arbitrary and could result in bias, particularly as hematologic SPM was diagnosed most frequently at 0 to 3 months following CAR T therapy versus 3 to 6 months following other SACT. Although little data exist on the timing of potential therapy-related SPM after CAR T therapy, its development within 90 days of CAR T therapy is not implausible (33).

In the post hoc analysis, we found that blood smear and BME were more prevalent among patients who received CAR T therapy than among patients who received other SACT. When stratified by receipt of BME at month 2, the risk of hematologic SPM no longer differed between CAR T therapy and other SACT. Importantly, BME is used to evaluate prolonged cytopenia, which could be a manifestation of multiple myeloma–associated inflammation and infection, CAR T treatment, lymphodepletion therapy, or secondary myeloid malignancy per se (34–36). BME is also used to ascertain treatment response in terms of minimal residual disease (37). Although our findings regarding BME could suggest a potential detection bias against the CAR T therapy group, they are also consistent with the role of BME in mediating a potential causal effect (i.e., BME could have been performed due to suspicion of SPM). In the latter scenario, conditioning on BME would obscure a true exposure–outcome association (38, 39). Unfortunately, the current analysis cannot differentiate between these 2 scenarios.

In the post hoc analysis, we also found a numerically higher cumulative incidence of hematologic SPM in the cilta-cel group compared with the ide-cel and unspecific CAR T groups, but the CI were wide and overlapping. Again, with higher BME in the cilta-cel group, it is not clear whether this finding could be explained by detection bias or reflects a true difference, with increased testing due to early signs of SPM. Similarly, CARTITUDE-1 (27) reported a numerically higher risk of hematologic SPM [9.3% (9/97)] than KarMMa-1 [3% (2/67); ref. 26]; however, cross-trial comparisons could be affected by differences in median follow-up time (27 months vs. 16 months), patient characteristics, and/or SPM ascertainment. Additionally, a previous RW study directly comparing cilta-cel with ide-cel reported a similar risk of MDS, AML, and T-cell lymphoma combined [OR, 0.94 (95% CI, 0.26–3.47)]. Lastly, the respective phase III trials also showed a numerically similar risk of hematologic SPM between cilta-cel–treated and ide-cel–treated patients, as summarized above (13, 14). Overall, the evidence is sparse and conflicting and not sufficient to support a difference in the risk of hematologic SPM across individual CAR T therapies.

The mechanisms for potential development of MDS/AML following CAR T therapy are not clear. However, clonal hematopoiesis of indeterminate potential (CHIP) might play a role. Subjects with CHIP exhibit somatic mutations in hematopoietic stem cells that enable clonal expansion but are not associated with cytopenias or dysplasia (40). In the general population, CHIP has a low risk of malignant transformation at 0.5% to 1% each year (40). However, rapid onset of MDS has been observed within 90 days after CAR T therapy in patients with lymphoma with CHIP (33), and a cumulative incidence of 19% for treatment-related myeloid malignancies after CAR T therapy was observed among patients with lymphoma with preexisting CHIP (41). Furthermore, multiple cases of T-cell lymphomas after CAR T therapy suggest a role of CHIP-derived malignant transformation (thought to be more important than CAR vector insertional mutation), and CHIP mutations can affect both lymphoid and myeloid lineages (42). It is possible that CAR T–related lymphodepletion, inflammation, and/or rapid T-cell expansion accelerate the development of MDS/AML in patients with existing CHIP (28, 34, 42).

This study has potential limitations. First, no validated claims-based algorithm to ascertain SPM was identified. We chose to use ≥ 1 claim to identify an SPM to prioritize sensitivity, complemented by a sensitivity analysis requiring ≥ 2 claims to reduce false-positive cases. Second, a detection bias due to the higher frequency of blood smear and BME following CAR T therapy cannot be ruled out. Third, although SPM could develop over a prolonged window after index treatment, this study only had available data to evaluate SPM through the first 24 months, which may bias findings toward types of SPM with early onset and miss types with delayed onset. Lastly, although many potential confounders were accounted for, there was still a risk of residual confounding due to unmeasured factors, such as disease severity, performance status, treatment setting (academic vs. community), and pretherapy evaluations. Despite these limitations, this is the first RW study, to our knowledge, that assessed the association between CAR T therapy and SPM risk. It has the following strengths: (i) using a large national claims database; (ii) having a relevant and clearly defined comparison group; (iii) requiring a sufficient baseline period to ascertain the LoT and multiple myeloma treatment history; (iv) balancing a comprehensive list of measured SPM risk factors between exposure groups; (v) offering transparent, granular findings on specific types of SPM and estimates of risk by time period; and (vi) performing additional comprehensive analysis to evaluate the impact of competing risk of death, alternative outcome definitions, and potential ascertainment bias.

Conclusions

In this RW study using a large claims database, patients with multiple myeloma seemed to have a higher risk of hematologic SPM through 24 months following CAR T therapy compared with other SACT, after balancing baseline covariates. However, contributions of potential misclassification bias and detection bias cannot be ruled out, as suggested by attenuated findings when ≥ 2 claims are required to identify an SPM and by a higher observed rate of BME following CAR T therapy. Future studies should further evaluate the potential association between hematologic malignancy and CAR T therapy, consider potential sources of bias, and cover longer follow-up beyond 24 months. Any potential risk of SPM should be weighed against the treatment benefits, including improved survival (43). Our findings suggest that treating physicians should be vigilant for myeloid malignancies after CAR T therapy.

Supplementary Material

Supplementary Figure S1

Figure S1. Attrition of patients with multiple myeloma and their index treatment periods with CAR T therapy and other SACT LoT

Supplementary Figure S2

Figure S2. Weighted incidence rate (95% CI) of specific types of solid SPM (left) and hematologic SPM (right) in patients with multiple myeloma receiving CAR T therapy or other SACT in the main analysis

Supplementary Figure S3

Figure S3. Unweighted proportions of patients with multiple myeloma who received blood smear and bone marrow aspiration/biopsy before and following treatment with individual CAR T therapies

Supplementary Methods S1

supplementary methods

Supplementary Table S1

Table S1. Weighted cumulative incidence of SPM with death as competing risk in patients with multiple myeloma receiving CAR T therapy or other SACT, main analysis and sensitivity analyses

Supplementary Table S2

Table S2. Weighted cumulative incidence of SPM with death as a censoring criterion in patients with multiple myeloma receiving CAR T therapy or other SACT

Supplementary Table S3

Table S3. Weighted cumulative incidence of SPM with death as competing risk in patients with multiple myeloma receiving CAR T therapy or other SACT and remained at-risk at Month 2, stratified by receipt of post-baseline bone marrow examinations by Month 2

Supplementary Table S4

Table S4. Unweighted cumulative incidence of SPM with death as competing risk in patients with multiple myeloma receiving individual CAR T therapies

Data Availability

The data analyzed in this study are available from Komodo Health (info@komodohealth.com). Restrictions apply to the availability of these data, which were used under license for this study.

Authors’ Disclosures

A. Suvannasankha reports grants and personal fees from Regeneron Pharmaceuticals, Inc., Bristol Myers Squibb, Janssen Oncology, Sanofi, Pfizer, and GSK outside the submitted work. M. Li reports employment at Regeneron Pharmaceuticals, Inc. and stock and stock options of Regeneron Pharmaceuticals, Inc. O. Omofuma is an employee of Regeneron Pharmaceuticals, Inc. and holder of Regeneron Pharmaceuticals, Inc. stocks. C. Hampp reports personal fees from Regeneron Pharmaceuticals, Inc. during the conduct of the study and outside the submitted work and is an employee and shareholder of Regeneron Pharmaceuticals, Inc. A. Breskin reports other support from Regeneron Pharmaceuticals, Inc. during the conduct of the study and outside the submitted work. P. Shao reports personal fees and other support from Regeneron Pharmaceutical, Inc. during the conduct of the study. T. Roccia reports other support from Regeneron Pharmaceuticals, Inc. during the conduct of the study. N. Mukherjee is an employee and shareholder of Regeneron Pharmaceuticals, Inc. D. Lee reports personal fees from Cellectis outside the submitted work. J. Glass reports other support from Regeneron Pharmaceuticals, Inc. during the conduct of the study. G.S. Kroog reports personal fees and other support from Regeneron Pharmaceuticals, Inc. during the conduct of the study and outside the submitted work; in addition, G.S. Kroog has a patent for WO 2025/244973 pending. K. Rodriguez Lorenc reports other support from Regeneron Pharmaceuticals, Inc. outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

A. Suvannasankha: Writing–original draft, writing–review and editing. M. Li: Conceptualization, data curation, visualization, methodology, writing–original draft, writing–review and editing. O. Omofuma: Conceptualization, data curation, visualization, methodology, writing–original draft, writing–review and editing. C. Hampp: Conceptualization, methodology, writing–review and editing. M. Phelan: Conceptualization, methodology, writing–review and editing. A. Breskin: Conceptualization, methodology, writing–review and editing. P. Shao: Conceptualization, data curation, methodology, writing–review and editing. T. Roccia: Conceptualization, supervision, writing–review and editing. N. Mukherjee: Conceptualization, writing–review and editing. A. Shrestha: Conceptualization, data curation, methodology, writing–review and editing. D. Lee: Conceptualization, data curation, methodology, writing–review and editing. J. Glass: Conceptualization, supervision, writing–review and editing. G.S. Kroog: Conceptualization, supervision, writing–review and editing. K. Rodriguez Lorenc: Conceptualization, writing–review and editing.