A network meta-analysis of randomized trials found no definitive efficacy advantage for ciltacabtagene autoleucel (cilta-cel), teclistamab plus daratumumab, or talquetamab-based combinations in early relapsed multiple myeloma (Merz et al., 2026).
Teclistamab-daratumumab had the most favorable PFS estimates, but the advantage disappeared when between-trial heterogeneity was allowed for. Toxicity, sequencing, treatment logistics, and drug-acquisition cost therefore carry more weight in selection as T-cell-redirecting therapies move into earlier lines.
Background: T-Cell-Redirecting Therapy Moves Into Earlier Relapse
BCMA-directed CAR T cells, particularly cilta-cel, have produced deep and durable responses in heavily pretreated relapsed or refractory myeloma. Bispecific antibodies targeting BCMA or GPRC5D are also highly active and need no individualized cell manufacturing. Earlier indirect comparisons and real-world studies generally favored CAR T cells in later-line disease, but most of that evidence came from single-arm or nonrandomized cohorts with extensive prior treatment. It may not apply to earlier relapse, where patients have received fewer therapies, may have better-preserved T-cell function, and can access both modalities.
Several phase III trials have evaluated these therapies after one or more prior lines, but none has compared cilta-cel directly with teclistamab- or talquetamab-based combinations. The analysis tested whether the randomized evidence can establish an efficacy hierarchy, and whether toxicity, delivery, and cost can guide selection when efficacy estimates overlap.
Methods
The systematic review included four randomized phase III trials with 2,174 patients: CARTITUDE-4 (cilta-cel), MajesTEC-3 (teclistamab-daratumumab), MonumenTAL-3 (talquetamab-daratumumab, with or without pomalidomide), and APOLLO (daratumumab, pomalidomide, dexamethasone [DPd], an additional control-arm reference). A network meta-analysis anchored to DPd-type control regimens allowed indirect comparison of treatments never tested against each other. PFS and OS were the primary outcomes.
Patient-level time-to-event data were reconstructed from published survival curves and analyzed with Bayesian fixed-effect and random-effects models, with frequentist analyses as a consistency check. A fixed-effect model assumes treatment effects are comparable across trials without meaningful between-trial variation. A random-effects model allows that variation. Each experimental regimen came from a single randomized comparison, so heterogeneity couldn’t be estimated from the network and was examined by varying the assumptions. Safety was evaluated descriptively, and a separate US cost-consequence analysis compared drug-acquisition costs with modeled progression-free time.
Results
Despite broadly similar DPd-type controls, median PFS differed about 2.5-fold. Median PFS in the DPd-type control arm:
- APOLLO, lenalidomide-refractory subgroup: 9.9 months
- CARTITUDE-4: 11.8 months
- MajesTEC-3: 18.1 months
- MonumenTAL-3: 24.4 months
The populations weren’t interchangeable. CARTITUDE-4 and the APOLLO subgroup were entirely lenalidomide-refractory, and CARTITUDE-4 alone enrolled anti-CD38-refractory patients (about 24%). Prior treatment, refractoriness, enrollment periods, and control regimens could all have contributed. Such differences can distort the relative effects compared indirectly.
PFS: Teclistamab-Daratumumab Ranked Highest, but Not Robustly
Under the fixed-effect model, teclistamab-daratumumab had the most favorable PFS estimates, with a P-score of 98%. The P-score is the average probability of outranking competing treatments, not a 98% probability of being best. One of six pairwise comparisons was significant under fixed-effect assumptions: teclistamab-daratumumab versus talquetamab-daratumumab (HR 0.55). Versus cilta-cel, the HR was 0.62.
With between-trial heterogeneity allowed, no comparison remained significant. Versus cilta-cel, the HR stayed at 0.62 but the interval widened to 0.21-1.81. Versus talquetamab-daratumumab, the HR was 0.55. These intervals are compatible with clinically meaningful differences in either direction.
Restricting the analysis to more comparable populations, to address the anti-CD38-refractory patients in CARTITUDE-4, strengthened the fixed-effect estimate favoring teclistamab-daratumumab over cilta-cel, but the difference again disappeared under random-effects assumptions. Across sensitivity analyses, the PFS advantage depended on assuming minimal between-trial differences.
OS: No Hierarchy
Teclistamab-daratumumab versus cilta-cel gave an OS HR of 0.94. No pairwise OS comparison was significant under either model. Because survival curves crossed in several trials, restricted mean survival time was also evaluated. At 24 months, OS gains over control ranged from 0.7 to 1.6 months across regimens.
Cilta-cel showed a later-emerging benefit: its restricted mean OS advantage over control reached 2.3 months at 36 months, with a wide confidence interval. The indirect comparison with teclistamab-daratumumab at 36 months was not significant. The estimates were too imprecise to exclude clinically important differences.
Safety
The analysis couldn’t reliably compare overall toxicity burden across trials, because populations, exposure, adverse-event reporting, supportive care, and follow-up differed. The regimens did show distinct profiles.
CAR T cells – in CARTITUDE-4, cytokine release syndrome (CRS) occurred in about 75% of cilta-cel recipients, with grade 3 events in about 1%. Immune effector cell-associated neurotoxicity syndrome (ICANS) occurred in about 5%, limited to grades 1 and 2. Cilta-cel also carries risks of delayed neurotoxicity, including movement disorders, cranial nerve palsies, and rare treatment-emergent parkinsonism, as well as prolonged cytopenias, infections, and other uncommon severe delayed toxicities.
Bispecific antibodies – CRS was frequent but mostly low grade and concentrated in step-up dosing. In MajesTEC-3, CRS occurred in about 60% of teclistamab-daratumumab recipients and ICANS in about 1%. With talquetamab combinations, CRS occurred in 58%-68% and ICANS in about 2%-3%. Talquetamab regimens carried GPRC5D-related effects: taste disturbance, skin and nail changes, weight loss, and occasional neurologic complications. Continuous BCMA-directed treatment raises concern for prolonged hypogammaglobulinemia and serious infection, which immunoglobulin replacement, antimicrobial prophylaxis, and schedule modification may reduce.
Second primary malignancies – at comparable median follow-up, these occurred in 13.0% of cilta-cel recipients and 12.4% of teclistamab-daratumumab recipients, numerically above their control arms. Talquetamab combinations showed lower incidences, but follow-up was shorter. Hematologic second primary malignancies were more frequent in each experimental arm than in its control. Longer follow-up is needed to separate the contributions of prior therapy, immune effects, and treatment-specific risk.
Indirect Rankings Are Unreliable Here
Network meta-analysis assumes that differences between study populations don’t materially distort relative effects. With control-arm PFS ranging from about 10 to 24 months, that assumption is doubtful.
Fixed-effect analysis suggested better PFS with teclistamab-daratumumab, but even modest allowance for heterogeneity removed every significant difference. Prior nonrandomized analyses that favored CAR T cells in heavily pretreated disease don’t automatically extend to earlier relapse.
Sequencing
BCMA-directed CAR T cells and bispecific antibodies can both select for BCMA loss or alteration, but frequency may differ: BCMA loss has been reported in up to half of patients relapsing after BCMA-directed bispecifics and appears less frequent after CAR T cells. Continuous T-cell engagement during bispecific treatment may also impair T-cell fitness and affect later autologous CAR T manufacturing and expansion.
CAR T therapy is a single infusion followed by expansion and contraction, which allows immune recovery. These differences have driven interest in giving CAR T cells before bispecifics when both are feasible. Observational studies suggest better outcomes with CAR T followed by a bispecific than the reverse, but this is hypothesis-generating, and randomized sequencing trials in early relapse are lacking. Optimal sequence is unresolved, particularly as agents against other antigens arrive.
Cost Considerations
From a US payer perspective, CAR T therapy carries a large upfront cost, while bispecific costs accumulate with administration. Modeled cilta-cel acquisition cost was $547,940. At the observed median treatment duration, teclistamab-daratumumab cost about $1.18 million. Using a population-standardized 30-month PFS estimate, drug-acquisition cost per progression-free month was:
- Cilta-cel: $18,300
- Teclistamab-daratumumab: $43,452
- Talquetamab-daratumumab: $34,976
- Talquetamab-daratumumab-pomalidomide: $58,629
The gap narrows with shorter bispecific courses: teclistamab-daratumumab reached cost parity with cilta-cel at about 9.7 months of treatment, talquetamab-daratumumab at 8.7 months, and talquetamab-daratumumab-pomalidomide at 5.1 months. This was not a formal cost-effectiveness analysis.
It used US wholesale acquisition prices, which may differ from actual payer costs and from other health systems, and it omitted hospitalization, manufacturing infrastructure, toxicity management, immunoglobulin replacement, caregiver burden, and quality-adjusted life-years. It describes the economic consequences of treatment structure under stated assumptions.
Limitations
Comparisons were indirect through a shared control, and each experimental regimen contributed one trial, so treatment-effect heterogeneity couldn’t be estimated directly.
- Control-arm variation points to differences in populations and prior treatment.
- Some time-to-event data were reconstructed from published curves, not original patient-level data.
- About 15% of patients assigned to cilta-cel in CARTITUDE-4 didn’t receive the planned infusion, which matters when intention-to-treat outcomes are set beside immediately available therapies.
- The shared DPd-type comparator requirement excluded KarMMa-3 and MajesTEC-9. MagnetisMM-5 was excluded because detailed published survival data weren’t yet available.
Additional randomized studies and more mature follow-up may improve precision. Prospective work is needed on optimal sequencing, whether fixed-duration bispecific therapy preserves efficacy, and long-term infection, neurologic, and second-malignancy risk, questions that will grow as these therapies enter newly diagnosed myeloma and high-risk smoldering disease.
Clinical Implications
Treatment selection depends on disease characteristics, prior therapy, anticipated toxicity, urgency, access to cellular therapy, patient preference, and consequences for later sequencing. Single-infusion CAR T therapy may have a lower modeled drug-acquisition cost per progression-free month than continuous bispecific combinations, but the advantage depends on treatment duration and pricing and doesn’t establish superior cost-effectiveness.
Source: Merz M, et al. Systematic review and network meta-analysis of CAR T-cell therapy and bispecific antibodies in early relapsed multiple myeloma. Journal of Hematology & Oncology. 2026.
