Subsequent CAR-T and engineered antibody for relapsed/refractory multiple myeloma following BCMA-targeted treatment: a systematic review and meta-analysis

Abstract

Relapse following B-cell maturation antigen (BCMA)-directed therapies represents a critical challenge in relapsed/refractory multiple myeloma (R/R MM). We performed a systematic review and meta-analysis of 34 studies (n = 1280) published between 2020 and 2025 to evaluate salvage CAR-T versus engineered antibody therapies post-BCMA exposure. The pooled overall response rate (ORR) was 60%. CAR-T therapy achieved significantly superior ORR compared to antibody-based approaches (77% vs 52%; p < 0.0001), with comparable complete response rates. While incidences of cytokine release syndrome and immune effector cell–associated neurotoxicity syndrome were similar, grade ≥3 cytopenias were more frequent with CAR-T. Notably, GPRC5D-targeted CAR-T yielded higher ORRs than BCMA-targeted constructs (86% vs 66%; p = 0.0006). Furthermore, patients with prior exposure to BCMA bispecific antibodies exhibited lower response rates compared to those with prior CAR-T exposure (71% vs 86%; p = 0.0404). In addition, studies with a median interval of ≥10 months demonstrated significantly higher ORRs than those with <10 months (70% vs 50%; p = 0.0146). In conclusion, GPRC5D CAR-T demonstrates superior efficacy to engineered antibodies in the post-BCMA setting. These findings support prioritising earlier CAR-T use and underscore the importance of antigen selection in optimising treatment sequencing.

Introduction

The robust development of immunotherapies designed to redirect the immune system to recognise and eliminate target cells has given rise to a paradigm shift in treating patients with multiple myeloma (MM). Among the most promising and extensively studied targets is B-cell maturation antigen (BCMA), a transmembrane glycoprotein highly expressed on myeloma cells but limited to normal tissues, making it an ideal therapeutic target. BCMA-directed strategies, involving antibody-drug conjugates (ADCs), bispecific antibodies (BsAbs), and chimeric antigen receptor (CAR)-T therapies, have reshaped the therapeutic landscape of MM and are being increasingly integrated into clinical practice. Belantamab mafodotin (belamaf) is a first-in-class ADC that delivers cytotoxic payloads directly to myeloma cells. Although belamaf monotherapy failed to demonstrate a progression-free survival (PFS) advantage in the DREAMM-3 trial and was withdrawn from the market [1], recent evidence from combination regimens has revived interest and led to the approval in the UK and European Union for patients with early relapse after first-line therapy [2, 3]. Two commercial CAR-T products, idecabtagene vicleucel (ide-cel) and ciltacabtagene autoleucel (cilta-cel), representing the most potent forms of BCMA-targeted treatment, have achieved deep and durable remissions in patients who have exhausted conventional therapies [4, 5]. In addition, BsAbs, particularly bispecific T-cell engagers, can facilitate the formation of immunological synapses and CD4+ and CD8 + T cells to proliferate, generate cytokines, and kill malignant cells in a polyclonal manner. Teclistamab, elranatamab, and linvoseltamab are BCMA-directed BsAbs approved for treating adults with relapsed or refractory (R/R) MM [6–8]. Although BCMA-directed therapy has shown impressive clinical activity, long-term follow-up reveals that a considerable proportion of patients ultimately relapse or progress. Notably, in the 5-year update of the CARTITUDE-1 trial, only 33% of patients remained progression-free [4, 5, 9–12].

Mechanisms underlying resistance and relapse to BCMA-targeted immunotherapies have been identified, including immune escape by tumour cells, intrinsic T-cell factors, and extrinsic factors from the immunosuppressive tumour microenvironment that impede sustained immune activity [13]. Initial investigations into the mechanisms have facilitated the advancement of personalised interventions and innovative adaptive therapeutic strategies. A second BCMA-targeted treatment is thought to be a viable option for subsequent anti-myeloma therapy (sAMT); however, the sequential use of various BCMA-targeted modalities raises concerns regarding potential cross-resistance. In early-phase clinical trials [14–18], novel targets such as GPRC5D and FcRH5 have shown promising efficacy and manageable side-effect profiles, making an immunotherapy that is feasible for patients who are relapsing on BCMA-directed treatment. Meta-analyses suggest that BCMA CAR-T therapy as a third-line or later intervention achieves higher complete response (CR) rates compared with BsAb, though it is also associated with increased rates of adverse events (AEs) [19, 20]. Conversely, a single-centre comparative study reported no significant differences in PFS between the CAR-T and BsAb cohorts, underscoring the ongoing debate regarding the optimal sequencing of immunotherapies [21]. Given the rapidly expanding population of patients previously exposed to BCMA-targeted therapies and the current lack of evidence-based guidance regarding the choice of immunotherapy modality or antigen target, we conducted a systematic review and meta-analysis of clinical trials and real-world retrospective studies evaluating CAR-T cells and engineered antibody therapies in patients with prior BCMA-targeted treatment. The review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

Methods

Systematic literature search process

This systematic review was prospectively registered in PROSPERO (CRD420251083913) and conducted in accordance with the PRISMA guidelines. A comprehensive literature search was performed across PubMed, Embase, Web of Science, and the Cochrane Library, covering the period from June 12, 2020, to June 12, 2025. The search strategy was developed in collaboration with a medical information specialist and independently peer-reviewed following the PRISMA-S recommendations. The search framework incorporated controlled vocabulary terms (MeSH/Emtree: “Multiple Myeloma,” “BCMA,” “Receptors,” “Chimeric Antigen”) combined with predefined free-text keywords (e.g., “GPRC5D,” “FcRH5,” “bispecific antibody,” “CAR-T”). A population filter was applied to identify studies enrolling patients with relapsed/refractory multiple myeloma previously exposed to BCMA-directed therapies (e.g., “salvage therapy,” “sequential therapy,” “BCMA-exposed”). Intervention terms encompassed CAR-T cell therapies and engineered antibody platforms, including BsAbs and ADCs. To minimize publication bias, grey literature was systematically reviewed through manual screening of ASH, EHA (2020–2025), and ClinicalTrials.gov was queried for completed or ongoing studies lacking published results. All retrieved records were deduplicated in EndNote X9 and screened in blinded, dual-reviewer mode using Rayyan.

Study selection

Studies were eligible for inclusion if they met all of the following criteria: (1) Prospective clinical trials or retrospective observational studies; (2) enrolled patients with R/R MM, defined according to the International Myeloma Working Group criteria; (3) included patients previously treated with at least one BCMA-directed agent (CAR-T cells, BsAbs, or ADCs), while allowing prior exposure to non-BCMA-targeted therapies; (4) reported at least one prespecified clinical outcome, including CR rate, overall response rate (ORR), very good partial response rate, partial response rate, cytokine release syndrome (CRS), immune effector cell–associated neurotoxicity syndrome (ICANS), or haematologic toxicities. Only short-term AEs were included in this study, and AEs were graded according to the Common Terminology Criteria for Adverse Events version 5.0 criteria. Treatment response was evaluated according to the International Myeloma Working Group criteria. Studies were excluded if they met any of the following criteria: (1) No documented prior exposure to BCMA-targeted therapy; (2) studies employing combination regimens involving two or more immune-based therapeutics were excluded—for example, dual-targeting CAR-T constructs, BsAbs combined with ADCs, or either CAR-T or BsAb therapy combined with other active anti-myeloma agents (e.g., immunomodulatory drugs, proteasome inhibitors, anti-CD38 or anti-SLAMF7 monoclonal antibodies, gamma-secretase inhibitors, or small molecule inhibitors such as XPO1 or BCL-2 inhibitors); (3) studies did not provide sufficient data to assess efficacy or safety outcomes.

Data extraction and quality assessment

Two investigators (Qiaolin Liu and Yun Kang) independently extracted data from all eligible studies using a standardized template. Duplicate reports originating from the same study centre were manually identified and removed to avoid redundancy. Key parameters included trial identifiers (name, registration number), publication details (journal, year, first author), and clinical profiles (median age, median prior treatment lines, high-risk cytogenetic profile, penta-refractory MM). Additional parameters were extracted: the Revised International Staging System (R-ISS) stage III proportion, and extramedullary disease. Efficacy endpoints included CR rate, ORR, and mPFS (median progression-free survival). Safety outcomes included CRS, ICANS, infection, anaemia, thrombocytopenia, and neutropenia incidence.

Methodological quality and risk of bias were assessed using the Methodological Index for Non-Randomised Studies [22], with independent scoring by QLL and YK. Discrepancies were resolved through consensus or third-party adjudication. Publication bias was evaluated using Begg’s test (two-tailed P < 0.05 indicating significance). Data extraction and bias assessment workflows followed PRISMA guidelines to ensure reproducibility.

Statistical analysis

The primary outcome measure for this meta-analysis was the ORR. Secondary outcomes included CR rate, CRS, severe CRS (grade 3 or higher), ICANS, severe ICANS (grade 3 or higher), and the incidence of anaemia, thrombocytopenia, and neutropenia. Given the predominance of single-arm studies, a random-effects model was implemented to synthesize pooled estimates for all outcomes, with corresponding 95% confidence intervals (CI) and P values, using the restricted maximum likelihood method. Heterogeneity was quantified using the I² statistic.

Subgroup analyses were conducted using the Q statistic to evaluate outcomes stratified by treatment type (e.g., CAR-T therapy, BsAbs). To investigate heterogeneity within the CAR-T group, studies were further grouped by product type. Meta-regression analysis was performed to identify potential confounders influencing ORR, including the types of product, extramedullary disease, high-risk cytogenetic abnormalities, median age, median prior lines of therapy, and penta-refractory MM. Continuous variables (e.g., median age) were analysed via weighted regression, while categorical variables were assessed using the Mann-Whitney U test to compare proportions between subgroups. Sensitivity analysis was conducted using a leave-one-out approach, iteratively excluding individual studies to assess their impact on pooled effect sizes (Fig. S1). All statistical tests were two-sided, with P < 0.05 considered significant. Analyses were performed in R Studio (version 2023.12.1) using the meta and metafor packages. Data visualization was generated using forest plots and funnel plots to evaluate publication bias.

Results

Literature search

From an initial database query of 5273 records, 2844 studies were selected for full-text review following deduplication and title/abstract screening. Consistent with predefined exclusion criteria, 34 studies comprising 1280 patients (CAR-T cells: n = 13; engineered antibody therapies: n = 21) were ultimately included in the quantitative pooled analysis (Fig. 1).

Fig. 1. Trial selection flow diagram.

The diagram summarizes record identification, screening, eligibility assessment, and study inclusion, with boxes indicating study numbers and reasons for exclusion.

Baseline study characteristics

Baseline characteristics of the 34 included studies are summarized in Table 1. All studies were published between 2021 and 2025. Engineered antibody therapies trials evaluated teclistamab (n = 8) [23–30], talquetamab (n = 4) [31–34], cevostamab (n = 2) [18, 35], RO7297089 (n = 1) [36], modakafusp alfa (n = 1) [37], belamaf (n = 1) [38], elranatamab (n = 2) [39, 40], linvoseltamab (n = 1) [8], and RG6234 (n = 1) [41], while CAR-T cell therapy trials (n = 13) investigated commercial products (n = 4) [10, 42–44], and those still in clinical trials (n = 9) [14, 45–52]. Most patients were heavily pretreated, with a median number of prior lines of treatment (LOT) exceeding four.

Table 1.

Additional characteristics of the included trials.

Product	Target	Types	Register No.	Study design	Median LOT	MINOR	Total pts	ORR, %	PFS/OS/DOR	With BCMA exposure
										No. of BCMA-exposed pts	First BCMA exposure	Median intervals	ORR, n (%)	PFS/OS
Modakafusp Alfa	CD38	immunocytokine	NCT03215030	prospective	BDT exposure: 7	16	146	120 mg: 32	120 mg: 4.1 mos/NA/NR	66	CAR-T (n = 21) ADC (n = 32) BsAb (n = 33)	NA	19(29)	NA
					BCMA-naive: 4			240 mg: 41	240 mg: 5.3 mos/NA/9.2 mos
Teclistamab	BCMA	BsAb	/	retrospective	6(3–14)	14	123	59	8.7mos/NR/NR	45	Ide-cel (n = 21) ADC (n = 23) both (n = 1)	6.0 mos	22(54.8)	ide-cel exposed: 1.8 mos/NA
Teclistamab	BCMA	BsAb	NCT03145181 NCT04557098	prospective	6(3–14)	16	40	/	/	40	CAR-T (n = 15) ADC (n = 29)	ADC (n = 10), 1.4(0.7–4.8) mos CAR-T (n = 8), 4.6(3.0–10.6) mos not last–line (n = 22), 1.1(0.2–2.9) mos	21(52.5)	4.5 mos/15.5 mos
Teclistamab	BCMA	BsAb	/	retrospective	6(3–13)	14	110	62	6-month PFS rate 52%, 6-month OS rate 80%	38	CAR-T (n = 18) ADC (n = 12) CAR-T and ADC (n = 2) BsAb (n = 5) ≥ 2 agents (n = 17)	NA	20(53)	NA
Teclistamab	BCMA	BsAb	/	retrospective	8(5–13)	14	22	/	/	22	CAR-T (n = 15) ADC (n = 2) both (n = 5)	12.3(1.3–33.5) mos	14(63)	NA
Teclistamab	BCMA	BsAb	/	retrospective	BDT exposure: 7 (6–9) BCMA–naive: 4 (4–6)	14	164	BCMA–naive: 68	BCMA–naive: 9.7 mos/NR/NA	60	CAR-T (n = 41) ADC (n = 22) BsAb (n = 5)	NA	32(53)	2.5 mos/9.1 mos
Teclistamab	BCMA	BsAb	/	retrospective	6(1–20)	14	210	all: 67	all: 6-month PFS 53%	84	CAR-T (n = 42) ADC (n = 27) BsAb (n = 6) ≥ 2 agents (n = 17)	249.5(30–1290) days	49(58.3)	6-month PFS 43%/6-month OS 75%
								BCMA–naive: 74	BCMA–naive: 6-month PFS 63%
Teclistamab	BCMA	BsAb	/	retrospective	6(4–17)	14	106	66	5.4 mos/10–month OS 67%/NA	56	CAR-T (n = 42) ADC (n = 24) other agents (n = 4)	250(19–1317) days within 3 mos (n = 14, ORR 42.9%) >3 mos (n = 42, ORR 63.3%)	33(59)	NA
Teclistamab	BCMA	BsAb	/	retrospective	5(2–21)	14	33	46	NA/17.7 weeks/NA	20	BDT	NA	7(35)	NA
Elranatamab	BCMA	BsAb	NCT03269136 NCT04649359 NCT05014412	retrospective	7(3–19)	14	86	/	/	86	CAR-T (n = 36) ADC (n = 58) Both (n = 8)	NA	39(45)	4.8 mos/NR
Elranatamab	BCMA	BsAb	/	retrospective	7(4–14)	14	34	53	NA	26	BDT	immediate line (n = 20) 1 LOT (n = 4) 2 LOT (n = 2)	23(88)	NA
Linvoseltamab	BCMA	BsAb	NCT03761108	prospective	200 mg: 5(2–16)	16	117	200 mg: 71	200 mg: NR/31.4mos/NA	10	ADC	NA	7(70)	NA
					50 mg: 6(3–14)			50 mg: 48
RO7297089	BCMA	BsAb	NCT04434469	prospective	7(2–12)	16	27	7	/	5	BDT	NA	1(20)	NA
Talquetamab	GPRC5D	BsAb	NCT03399799	prospective	0.4 mg/kg:5(4–6) 0.8 mg/kg:4.5(4–6) exposure:6(5–8)	16	537	0.4 mg/kg: 74	0.4 mg/kg:7.5 mos/NR/NA	78	CAR-T (n = 57) ADC (n = 27) BsAb (n = 11)	<6 mos (n = 28) 6–9 mos (n = 10) >9 mos (n = 45)	52(67)	7.7 mos/12–month OS rates 74%
			NCT04634552					0.8 mg/kg:69	0.8 mg/kg:11.2 mos/NR/NA
Talquetamab	GPRC5D	BsAb	/	retrospective	7(4–17)	14	68	71	4.9 mos/NR/8.5 mos	63	CAR-T (n = 43) ADC (n = 6) BsAb (n = 49)	<6 mos (n = 35) ≥6 mos (n = 26) immediate line (n = 25) not immediate (n = 36)	43(68.2)	4.8 mos/NR/DOR 7.6 mos
Talquetamab	GPRC5D	BsAb	/	retrospective	NA	14	55	BCMA–naive: 68	BCMA–naive: 0.67 mos/NR/NA	24	CAR-T (n = 18) BsAb (n = 6) both (n = 12)	NA	17(71)	CAR-T 3.9 mos/NR BsAb 0.77 mos/NR Both 1.63 mos/NR
Talquetamab	GPRC5D	BsAb	/	retrospective	6(2–15)	14	138	65	6.4 mos/NR	71	CAR-T (n = 12) ADC (n = 10) BsAb (n = 5)	NA	41(58)	NA
RG6234	GPRC5D	BsAb	NCT04557150	prospective	IV: 5(2–15)	16	105	IV: 71	NA	18	BDT	NA	10(56)	NA
					SC: 4(2–14)			SC: 60
Cevostamab	FcRH5	BsAb	NCT03275103	prospective	6(2–18)	16	167	BCMA–naive: 61	NA/NA/15.6 mos	96	CAR-T (n = 60) ADC (n = 34) BsAb (n = 40)	NA	29(30.2)	NA
Cevostamab	FcRH5	BsAb	NCT05535244	prospective	8(5–14)	16	42	CAR-T: 73, ADC: 60, BsAb: 10	NA	42	CAR-T (n = 11) ADC (n = 10) BsAb (n = 21)	122(68–558) days	16(38)	NA
Belamaf	BCMA	ADC	/	retrospective	6(2–14) exposure: 9	14	82	43	3.8 mos/17.2 mos/10.2 mos	18	CAR-T (n = 12) ADC (n = 1) BsAb (n = 5)	10.7 mos	5(28)	2 mos/20.4mos
HRC0202	BCMA	CAR-T	NCT04003168	prospective	7(4–11)	16	7	/	/	7	CAR-T	NA	5(71.4)	9.0 mos/NR/NA
Cilta-cel	BCMA	CAR-T	NCT04133636	prospective	8(4–13)	16	20	/	/	20	BsAb (n = 7) ADC (n = 13)	235(62–944) days	12(60)	9.1mos/NR/11.5mos
Ide-cel and Cilta-cel	BCMA	CAR-T	/	retrospective	7(3–13)	14	36	/	/	36	BsAb	NA	27(75)	9.6 mos/NA
HBI0101	BCMA	CAR-T	NCT04720313	prospective	4(3–13)	16	50	90	11.1 mos/NR/10.3 mos	12	ADC	NA	9(75)	NA
Ide-cel	BCMA	CAR-T	NCT03361748	prospective	7(4–21)	16	821	73	8.8 mos	123	BDT	20(7–31) mos	71(58)	Last BDT < 6 mos, mPFS 4.9 mos Last BDT ≥ 6 mos, mPFS 5.9 mos
Equecabtagene Autoleucel	BCMA	CAR-T	ChiCTR2000033946	prospective	4(3–23)	16	101	96	12 mos PFS rate 78.8%; 12 mos OS rate 92.2%	12	CAR-T	NA	9(75)	NR
P-BCMA-ALLO1	BCMA	CAR-T	NCT04960579	prospective	10	16	5	/	/	5	CAR-T (n = 2) BsAb (n = 2) both (n = 1)	NA	3(60)	NA
anti-GPRC5D CAR T	GPRC5D	CAR-T	ChiCTR2100048888	prospective	6(5–7)	16	37	/	/	37	CAR-T	16.2(7.3–28.9) mos	31(84)	4.5 mos/NA/NA
CC-95266	GPRC5D	CAR-T	NCT06297226	prospective	5(3–15)	16	84	87	14.5 mos/NA/NA	38	BDT	NA	30(79)	NA
OriCAR-017	GPRC5D	CAR-T	NCT05016778	prospective	5.5(3–17)	16	10	100	11.4 mos/NR/10.4 mos	5	CAR-T	NA	5(100)	19.1 mos/NR/17.2 mos
DM91	GPRC5D	CAR-T	NCT05739188	prospective	/	16	7	86	NA	3	CAR-T	NA	2(67)	NA
MCARH109	GPRC5D	CAR-T	NCT04674813	prospective	all:6(4–14);with:6	16	17	71	150×106group 7.8 mos/NR/7.8 mos	10	BDT	16.4(4.4–36.6) mos	9(90)	NA
CT071	GPRC5D	CAR-T	NCT05838131	prospective	5(1–12)	16	17	94	NA	4	CAR-T	NA	4(100)	NA

Ide-cel idecabtagene vicleucel, Cilta-cel ciltacabtagene autoleucel, Belamaf belantamab mafodotin, No. number, Pts patients, ORR overall response rate, LOT lines of therapy, PFS progression-free survival, OS overall survival, DOR duration of response, CAR chimeric antigen receptor, BsAb bispecific antibodies, ADC antibody-drug conjugate, Mos months, SC subcutaneous, IV intravenous, BDT BCMA-targeted therapy, NA not available, NR not reached.

Overall, the included studies demonstrated robust quality as evaluated by the Methodological Index for Non-Randomised Studies scale (Table 1) [22]. Publication bias was formally assessed using Begg’s and Egger’s tests. Significant publication bias was detected across evaluated endpoints: response rate (Begg’s test P = 0.51; Egger’s test P < 0.05) with graphical confirmation in Figure S2A.

Comparative efficacy and safety outcomes

Efficacy of CAR-T vs. engineered antibody therapies

The aggregate pooled CR rate was 27% (95% CI 0.19–0.36). No significant difference in CR was observed between the CAR-T cell group (30%, 95% CI 0.16–0.43) and the engineered antibody therapies group (26%, 95% CI 0.16–0.37, Fig. 2A). The overall pooled proportion of ORR was 60% (95% CI 0.53–0.67). There was a significant difference in ORR between the CAR-T group and the engineered antibody therapies group (P < 0.01); 77% (95% CI 0.69–0.85) in the CAR-T group and 52% (95% CI 0.44–0.60) in the engineered antibody therapies group (Fig. 2B). Subgroup analysis stratified by CAR-T regimen and products revealed that heterogeneity primarily originated from diversity in product types. In Table 1, the median PFS period for the CAR-T group is longer than that of the antibody group. Moreover, we performed a meta-analysis of the mPFS between the CAR-T group (n = 1) and the antibody group (n = 5), revealing a mPFS of 8.9 months and 3.62 months, respectively (Figure S2B, P = 0.1174).

Fig. 2. Forest plot of the efficacy of CAR-T and engineered antibody therapies.

A Complete response (CR) or stringent complete response (sCR) rate; B overall response rate (ORR). CAR-T, chimeric antigen receptor T cell.

Safety profile: CAR-T vs. engineered antibody therapies

In the analysis related to safety, CAR-T therapy was associated with a higher incidence of AEs. The incidence of CRS was comparable between the CAR-T group (67%) and the engineered antibody therapies group (69%) (Figure S3A). Grade ≥3 CRS occurred in 3% of patients in the CAR-T group and 1% in the engineered antibody therapies group (P > 0.05) (Figure S3B). The incidence of ICANS was in the CAR-T group (10%) compared with the engineered antibody therapies group (8%) (Figure S3C). The rate of grade ≥3 ICANS was 1% in the CAR-T group and 2% in the engineered antibody therapies group (Figure S3D). About haematologic adverse effects, the rate of neutropenia was 87% in the CAR-T group and 57% in the engineered antibody therapies group (Fig. 3A, P = 0.0674). These haematologic toxicities largely reflect the expected cytopenias occurring within 30 days after CAR-T therapy. The incidence of anaemia was 67% in the CAR-T group and 56% in the engineered antibody therapies group (Fig. 3C, P = 0.0552). The incidenceof thrombocytopenia was 85% in the CAR-T group and 42% in the engineered antibody therapies group (Fig. 3E, P < 0.01). Intergroup disparities were observed in the occurrence of haematologic AEs of grade 3 or higher, notably with neutropenia, at an incidence of 79% (95% CI 0.70–0.89) in the CAR-T group and 47% (95% CI 0.34–0.59) for the engineered antibody therapies group (Fig. 3B, P < 0.01). Similarly, the rate of anaemia of grade 3 or higher was markedly higher in the CAR-T group with 60% (95% CI 0.47–0.72) while 38% (95% CI 0.28–0.47) within the engineered antibody therapies group (Fig. 3D, P < 0.01). The rate of thrombocytopenia of grade 3 or higher was markedly higher in the CAR-T group with 65% (95% CI 0.53–0.77) and 29% (95% CI 0.22–0.36) within the engineered antibody therapies group (Fig. 3F, P < 0.01). The rate of grade ≥3 infections was 22% (95% CI 0.10–0.38) in CAR-T therapy and 37% (95% CI 0.23–0.50) in engineered antibodies. However, the difference was not statistically significant (Figure S4, P = 0.114). Given the limited number of studies, these results require further validation.

Fig. 3. Forest plot of the adverse events rate on CAR-T and engineered antibody therapies.

A Neutropenia; B neutropenia (grade ≥3); C anaemia; D anaemia (grade ≥3); E thrombocytopenia; F thrombocytopenia (grade ≥3). CAR-T, chimeric antigen receptor T cell.

Meta-regression insights and clinical implications

To explore potential sources of heterogeneity, we conducted mixed-effects meta-regression analyses incorporating key moderators, including patient demographics, disease characteristics, and methodological variables (Table 2). This approach enabled simultaneous evaluation of heterogeneity determinants while adjusting for relevant covariates. In univariate meta-regression, treatment therapeutics emerged as a significant moderator of ORR, with GPRC5D-directed CAR-T therapy demonstrating superior efficacy compared with teclistamab (P < 0.001). For FcRH5-targeted BsAbs, cevostamab demonstrated a significantly lower ORR compared with teclistamab (P = 0.004). Subgroup analyses based on therapeutic targets further supported these findings. ORR was higher in GPRC5D CAR-T (86%; 95% CI 0.79–0.93) than in BCMA CAR-T (66%; 95% CI 0.58–0.75) (Fig. 4A, P < 0.001). Among antibody-based therapies, ORR varied by target: 52% (95% CI 0.43–0.61) for BCMA-directed agents, 63% (95% CI 0.56–0.69) for GPRC5D-directed agents, and 44% (95% CI 0.16–0.71) for FcRH5-directed agents (Fig. 4B, P < 0.001). Talquetamab also demonstrated a higher ORR than teclistamab (65% [95% CI 0.59–0.71] vs. 53% [95% CI 0.48–0.58], P = 0.0037) (Fig. 4C). In the CAR-T subgroup, ORR was 65% (95% CI 0.55–0.76) for commercial products and 84% (95% CI 0.77–0.90) for investigational CAR-T therapies (Figure S5A, P < 0.01). In the BsAbs group, ORR was 57% (95% CI 0.50–0.64) for commercial products and 36% (95% CI 0.18–0.54) for BsAb therapies in trials (Figure S5B, P = 0.0031).

Table 2.

Univariate meta-regression analysis to explore the factors that affect the overall response rate.

Variables	Coefficient	SE	95% CI	P
Commercial CAR-T	0.444	0.253	(0.949–2.562)	0.08
GPRC5D CAR-T	1.517	0.332	(2.378–8.746)	<0.001
BCMA CAR-T	0.747	0.489	(0.809–5.509)	0.127
Cevostamab	−0.802	0.278	(0.260–0.773)	0.004
Talquetamab	0.486	0.224	(1.048–2.522)	0.03
Extramedullary disease	−0.021	0.013	(0.954–1.005)	0.114
High-risk cytogenetic profile	0.002	0.046	(0.916–1.096)	0.965
R-ISS stage III	−0.044	0.073	(0.830–1.103)	0.543
Median age (years)	−0.088	0.073	(0.794–1.057)	0.228
Median prior lines of therapy	−0.258	0.223	(0.499–1.195)	0.246
Penta refractory MM	−0.021	0.023	(0.936–1.023)	0.343

CAR chimeric antigen receptor, R-ISS The Revised International Staging System, MM multiple myeloma, SE standard error, CI confidence interval.

Fig. 4. Forest plot of the overall response rate in different subgroups.

A Different targets in the CAR-T group; B different targets in the engineered antibody group; C teclistamab and talquetamab; D different types of BCMA exposure in the CAR-T group; E different types of BCMA exposure in the engineered antibody group; F subgroup analysis of pooled objective response rate stratified by treatment interval ( < 10 months vs >10 months). ORR, overall response rate. CAR-T, chimeric antigen receptor T cell. Mos, months.

Prior BCMA-directed exposure influenced CAR-T outcomes: ORR was higher in patients previously treated with BCMA CAR-T (86%; 95% CI 0.77–0.94) than in those exposed to BCMA BsAbs (71%; 95% CI 0.61–0.82) (Fig. 4D, P = 0.04). In contrast, for the BsAb subgroup, prior exposure type did not significantly affect outcomes among patients receiving BsAbs (Fig. 4E; P = 0.664). A subgroup analysis was performed to assess the effect of the interval between prior BCMA-targeted treatment and sAMT on ORR using a cutoff of 10 months. Studies with a median interval of <10 months showed a pooled ORR of 50% (95% CI 0.40–0.59), whereas those with an interval of >10 months showed a higher pooled ORR of 70% (95% CI 0.57–0.84). The difference between the subgroups was statistically significant (P = 0.0146) (Fig. 4F). However, the median prior LOT was not significantly associated with ORR (P = 0.246). Subgroup analysis stratified by the median prior LOT using a cutoff of seven revealed distinct patterns across modalities. For CAR-T therapy, studies enrolling patients with 5–7 prior LOT showed a numerically higher pooled ORR than those with >7 prior LOT (80% vs. 60%; P = 0.07) (Figure S6A). For BsAbs, no significant difference in pooled ORR was observed between the two subgroups (44% for 5–7 LOT vs. 51% for >7 LOT, P = 0.52) (Figure S6B).

To exclude the potential influence of study design, a subgroup analysis was performed according to prospective and retrospective studies separately for the teclistamab and talquetamab cohorts. In the teclistamab cohort, the pooled ORR was 55% in retrospective studies and 53% in prospective studies, with no significant difference observed between the two subgroups (P = 0.739). Similarly, in the talquetamab cohort, the ORR was 64% in retrospective studies and 67% in prospective studies, with no significant subgroup difference (P = 0.745) (Figure S6C-6D). These findings suggest that study design (prospective vs retrospective) did not significantly influence the estimated treatment response rates.

Discussion

This meta-analysis indicates that CAR-T therapy induced more favourable response rates than engineered antibody treatment. These benefits, however, were accompanied by a higher incidence of severe haematologic toxicities within 30 days of treatment, including neutropenia, anaemia, and thrombocytopenia. The rates of ICANS and CRS were comparable between CAR-T and engineered antibody therapies, suggesting that neurotoxicity and cytokine-driven toxicities are not disproportionately increased with either treatment approach. Our analysis revealed that GPRC5D represents a more advantageous subsequent target than BCMA, as GPRC5D CAR-T products demonstrate higher response rates, even in the early stages of development. Besides, studies with a median interval of at least 10 months between prior BCMA-targeted therapy and subsequent CAR-T or antibody-based therapy had higher pooled ORRs than those with a median interval of <10 months. In the BsAb cohort, the target antigen correlates with efficacy; across 12 studies evaluating talquetamab and teclistamab, talquetamab consistently yielded superior response rates. Meta-regression analyses further confirmed that GPRC5D CAR-T therapy (P < 0.001) and talquetamab (P = 0.03) maintained superior response rates even after adjusting for confounders such as prior lines of therapy and target selection. However, cevostamab was found to be an adverse factor (P = 0.004). Furthermore, patients with prior CAR-T therapy exhibited a higher ORR upon subsequent immunotherapy compared to those previously treated with BsAb, implying that BsAb exposure was a risk factor for regaining remission.

Previous BsAb exposure may compromise the effectiveness of subsequent T-cell-redirected therapies. In the CAMMA 2 trial of cevostamab, the ORR for patients with previous exposure to BsAb was 10%, significantly lower than in those with prior ADC (60%) or CAR-T therapy (73%) [35]. Similarly, in the MonumenTAL-1 study of talquetamab, patients with exposure to BCMA CAR-T achieved better ORR compared to those with prior BsAb exposure (74% vs 58%) [31]. The interval between prior targeted therapy and subsequent treatment represents an important clinical consideration. Patients who had BsAb as the immediate previous line of therapy had an ORR of 29%, whereas those treated with BsAb prior to the final line achieved an ORR of 68%, suggesting that the impact of BsAb on the immune system may diminish over time [53]. A similar pattern has been observed with CAR-T therapy. In real-world data of standard-of-care ide-cel, the mPFS for patients with prior BCMA BsAbs was 2.8 months, compared with 9.0 months in those without prior BCMA therapy, while mPFS was not reached in patients with prior CAR-T exposure [43]. In addition, a multi-centre study in Germany reported that patients with dual exposure to teclistamab and talquetamab had the shortest mPFS (4.4 months) following commercial BCMA CAR-T across all exposure groups [54]. Our analysis revealed no significant difference in mPFS was observed between the CAR-T group and the engineered antibody group; however, the limited number of eligible studies warrants cautious interpretation, and further survival data are needed. Overall, the sequence of T-cell–engaging therapy may affect the therapeutic benefit, as repeated BsAb exposure may contribute to T-cell exhaustion and reduced efficacy. These observations underline the need to optimise treatment sequencing, given their potentially distinct effects on subsequent immunotherapy outcomes.

The divergent outcomes of CAR-T therapies and engineered antibody therapies as sAMT could be partially attributed to multiple factors. Among engineered antibody therapies, BsAb stood out. BsAbs require repeated administration and function by redirecting preexisting T cells to eliminate malignant cells, the response to which is highly dependent on the endogenous T-cell repertoire [55]. Impaired cytotoxicity of BsAbs has been associated with an increased proportion of T cells expressing exhaustion markers, as well as expansion of regulatory T cells (Tregs) [56]. Likewise, exhausted-like T-cell clones in the bone marrow have been implicated in primary resistance or early relapse in patients receiving BCMA-CD3 BsAbs [56].

Although the composition of the T-cell compartment at leukapheresis also influences CAR-T efficacy [57], the genetic engineering and ex vivo expansion endow CAR-T cells with enhanced receptor signalling capacity, robust proliferative potential, and sustained immune surveillance independent of MHC restriction. These properties may mitigate the effects of patient heterogeneity on T-cell fitness. The efficiency of allogeneic off-the-shelf CAR-T products as salvage therapies further shows the potential to overcome host T-cell heterogeneity [47]. Beyond mechanistic differences, treatment accessibility and timing may also influence outcomes. CAR-T therapy involves personalised manufacturing and lymphodepleting treatment before infusion, whereas BsAbs are “off-the-shelf” and can be administered rapidly, which is advantageous in aggressive disease. Nevertheless, CAR-T tends to produce more profound and more durable responses, emphasizing its importance as a more potent therapeutic option [20]. However, the higher incidence of severe haematological toxicity in CAR-T therapies is likely related to the myelotoxicity of conditioning chemotherapy [58]. Besides, delayed haematopoietic recovery after CAR-T infusion was linked to reduced niche function and elevated inflammation-related cytokines in the bone marrow, which can damage haematopoietic stem cells and hinder long-term bone marrow reconstitution [59].

In addition to the availability of immunotherapy modalities, several tumour-associated factors account for the target selection after BCMA treatment in the context. Poor PFS after initial BCMA-targeted therapy correlates with specific baseline characteristics, including high-risk cytogenetics, R-ISS stage III disease, extramedullary disease, poor performance status, and elevated metabolic tumour volume [60, 61]. Indeed, real-world outcomes with ide-cel and cilta-cel pointed out that prior BCMA-targeted therapy negatively affects PFS [53, 62]. Nonetheless, BCMA-negative relapse following CAR-T therapy, driven by genomic loss or mutation, appears to be relatively uncommon [63, 64].

Biallelic loss of TNFRSF17 at relapse has been associated with reduced BCMA expression and the expansion of dominant myeloma clones under the selective pressure of BCMA-targeted therapy. Among patients who initially responded to CAR-T but later relapsed, monoallelic and biallelic BCMA losses were detected in 12% and 6% of cases, respectively, compared with 4% at baseline [65]. Emerging evidence further suggests that high metabolic tumour volume combined with low pretreatment soluble BCMA levels may indicate plasma cell populations with absent or down-regulated BCMA expression, potentially conferring a higher risk of antigen-negative relapse [61]. Even though antigen escape might not be the main cause of relapse, previous BCMA-targeted therapy could reduce the effectiveness of subsequent BCMA-directed treatments. GPRC5D is an alternative target of sAMT, with over 50% antigen expression on CD138 + MM cells detected in 65% of patient samples, and demonstrates a distribution pattern complementary to BCMA [66]. MCARH109, a GPRC5D CAR-T with a human B-cell-derived scFv, has shown clinical responses at all dose levels in patients with prior BCMA-targeted CAR-T therapies, which were slightly higher than those without BCMA treatment exposure (ORR 75% vs 71%; CR 37.5% vs 35%) [14]. Similarly, OriCAR-017, a bi-epitopic GPRC5D CAR-T, achieved an ORR of 100% and a sCR rate of 40% among five patients with prior BCMA CAR-T therapy, including two with BCMA-negative relapse [16]. These results collectively emphasize the potential of GPRC5D as a prospective therapeutic target in the future and establish the capacity to circumvent resistance mechanisms that restrict the efficacy of repeated BCMA-directed strategies.

This meta-analysis acknowledges several inherent constraints. As a single-arm trial synthesis, the investigation lacked the capacity to assess patient-level variables comprehensively. Additionally, the limited cohort size may restrict broader applicability. Notably, while significant intergroup heterogeneity was detected through multiple outcome measures, subsequent stratification analyses were not pursued to elucidate these variations. Despite these methodological drawbacks, our rigorous inclusion protocols and advanced statistical approaches validated the therapeutic advantages of T cell-mediated interventions for R/R MM. The systematic methodology employing stringent selection criteria and robust analytic frameworks maintained scientific validity despite the acknowledged limitations.

In summary, our study demonstrated the superior therapeutic potential of CAR-T therapy as a subsequent intervention for patients with R/R MM who failed prior BCMA-directed treatments, though with elevated AEs. Future studies should focus on refining patient selection, optimising sequential therapy strategies, and exploring alternative antigen targets such as GPRC5D and FcRH5 to overcome resistance and further improve clinical outcomes in this high-risk population. Treatment selection necessitates the strict integration of individualized patient profiles, incorporating molecular biomarkers, prior therapy responses, and comorbidities guided by precision oncology principles (NCCN Guidelines v3.2023).

Author contributions

MH and KY designed the study. KY and LQL screened studies, interpreted data, performed statistical analysis, and wrote the manuscript. LQL and KY performed the literature search. MH, KY, and SCY verified extracted data and contributed to manuscript revision. All authors approved the final version of the manuscript and take full responsibility for the decision to submit for publication.

Funding

This work was supported by grants from the National Natural Science Foundation of China (No. 82330005 to HM, No. 82350103 to HM), and the National Science Fund for Distinguished Young Scholars (No. 82425003 to HM).

Data availability

All data included in this manuscript are publicly available from the original publications from which data were extracted.

Competing interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies

During the preparation of this work, the authors used ChatGPT in order to refine and polish the writing and enhance readability. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41408-026-01564-1.