A cost-effectiveness analysis was conducted through a partitioned survival model (PSM) with three health states: progression-free, post-progression and death.

The perspective of the Spanish NHS was taken, and a weekly cycle length was considered in the model. To capture clinical and economic effects derived from long-term patient evolution, a lifetime horizon was used (25 years) applying a discount rate of 3% for both future costs and outcomes [21].

The demographic characteristics of a hypothetical patient cohort were sourced from the MagnetisMM-3 Cohort A patient data since they were consistent with the Spanish population. Median age was 68.0 (range: 36–89) years, with 19.5% of the cohort older than 75 years. Of all patients, 63.4% had an Eastern Cooperative Oncology Group (ECOG) performance status of 1 or 2. Patients had received a median of five (range 2–22) prior lines of therapy, with 96.7% having triple-class refractory disease and 42.3% having penta-drug refractory disease. Additionally, general mortality was adapted as per the Spanish mortality data obtained from the National Institute of Statistics (INE).

Elranatamab was compared to a treatment basket representing the PCT based on the LocoMMotion study [13] and adapted to reflect the clinical setting in Spain at the time of the analysis (Electronic Supplementary Material [ESM] Table S1), and to teclistamab.

The proportion of patients in each health state at each time point was based on the efficacy of each treatment as measured by the PFS and OS curves, with the area between both curves being the post-progression survival (PPS). Since at the time of this analysis long-term survival data from MagnetisMM-3 were not available, data from the median follow-up of 14.7 months were used, as well as data from published matching-adjusted indirect comparisons (MAICs) of elranatamab versus PCT and teclistamab, respectively [20, 22].

As clinical data did not cover the entire time horizon and extrapolations were needed, several parametric distributions (exponential, Weibull, Gompertz, lognormal, log-logistic, gamma and generalized gamma) were fitted to the clinical data obtained from MagnetisMM-3, assessing goodness-of-fit through the Akaike and Bayesian Information Criteria (AIC and BIC) and visual inspection of the resulting OS and PFS parametric fits for elranatamab (Fig. 1). Weibull distributions were used as the base case for both OS and PFS curves, and alternative distributions were explored in scenario analyses.

Overall survival (a) and progression-free survival (b) standard parametric fits for elranatamab. KM Kaplan–Meier curve, OS overall survival, PFS progression-free survival

For each comparator (teclistamab and PCT), efficacy curves were obtained through the MAIC. If the proportional hazard (PH) hypothesis was acceptable, a hazard ratio (HR) was used to calculate relative efficacy values. Otherwise, efficacy curves were adjusted directly from the MAIC parameters. HRs were used for PCT and teclistamab OS (HR values 0.62 and 0.66, respectively) and for teclistamab PFS curves (HR value 0.59). The PH hypothesis was not acceptable for the PCT PFS curve and, therefore, it was adjusted directly from the MAIC parameters.

In addition to PFS and OS data, the duration of treatment (DoT) was also considered to take into consideration that a group of patients discontinued treatment before their progression due to other causes. This parameter affected solely cost results. Since only median DoT values were available for the different treatment options, DoT curves were built through exponential distribution adjusted to the median values reported in clinical trials (elranatamab 5.6 months; teclistamab 8.5 months; PCT 4.0 months) [13, 18, 23] (ESM Fig. S1). DoT curves were adjusted to PFS when necessary.

Treatment-related grade 3–4 adverse events (AEs) with a reported incidence of ≥ 5% in any of the comparators sourced from their respective clinical trials were included in the model [13, 18, 23,24,25,26]. Special interest AEs, namely ICANS (immune effector cell-associated neurotoxicity syndrome) and CRS (cytokine release syndrome), were also included regardless of grade and incidence for elranatamab and teclistamab (ESM Table S2). Treatment with tocilizumab was used for the management of both special interest AEs (ICANS and CRS) [18, 19].

Additionally, given the high incidence of hypogammaglobulinaemia and infections during treatment with anti-BCMA bispecific antibodies, and following recent clinical recommendations, infection prophylaxis was considered for both options, elranatamab and teclistamab. Acyclovir administration was assumed for all alternatives, and intravenous immunoglobulin G was considered for elranatamab and teclistamab.

Health utility values were derived from patient reported outcome data gathered from the MagnetisMM-3 trial for progression-free and post-progression health states and were used to calculate quality-adjusted life-years (QALYs) [27]. As utility values were collected from the EuroQoL five dimensions questionnaire (EQ-5D-5L), they had been mapped onto EQ-5D-3L through the Hernández Alava [28] method to follow the latest clinical recommendations [29], thereby providing utility values of 0.71 and 0.63 for progression-free and the post-progression health states, respectively. These utilities were applied to all treatments included in the analysis.

The model included the healthcare resource use derived from the disease management for patients on each treatment, differentiating frequencies for progression-free and post-progression health states. All patients were assumed to make use of included resources. Treatment-specific resource use was sourced from clinical trial protocols [18, 23], as shown in Table 1.

The model assumed that 63.9% of patients progressing to the post-progression state receive a subsequent treatment, which was considered only for cost purposes, following the MagnetisMM-3 trial data. Subsequent treatment composition is based on the reported subsequent treatments observed in the MagnetisMM-3 trial and is a treatment basket which consists primarily of carfilzomib- and bortezomib-containing regimens (data on file). The same subsequent treatment was applied to all patients, regardless of their primary treatment.

As the analysis adopted the perspective of the Spanish NHS, only direct costs (€ 2024) were considered. These include drug acquisition and administration costs, AEs and disease management costs (routine tests, physician’s office visits, imaging and biopsy tests, hospitalization).

Drug acquisition costs were expressed as ex-factory list prices [31], considering, when applicable, mandatory deductions established in Royal Decree-Law 8/2010 (Boletín Oficial del Estado [32]). Given that, in clinical trials, doses were sometimes reduced in the presence of AEs, relative dose intensity (RDI) was considered for each treatment based on the values reported during their respective clinical trials (78.35%, 93.35% and 93.70% for elranatamab, PCT and teclistamab, respectively) [18, 23, 33, 34]. Vial wastage was assumed for all treatments. Drug costs for primary treatments were calculated following the DoT curves based on reported median values. Following established trial protocols and reported results, dosing regimen switches were considered when applicable for elranatamab and teclistamab (at 6 and 11.3 months after start of treatment, respectively) [18, 24].

A one-off cost of €21,084.50 was included in the model to represent terminal care prior to death. This value was obtained from relevant sources for the Spanish setting and adjusted to 2024 € [35].

Drug acquisition and administration, AE management and healthcare resource use costs were obtained from relevant publications and official sources (ESM Table S3).

Finally, all inputs and assumptions used were validated by a panel of two haematologists, both of whom are MM experts from University Hospitals in Spain, through a multi-step process consisting of a written questionnaire, an in-person advisory board and a follow-up meeting for results validation.

Our research is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.

Deterministic and probabilistic sensitivity analyses were conducted to assess the uncertainty of the variables and determine the robustness of the results.

One-way sensitivity analysis (OWSA) was performed, modifying relevant model variables individually by ± 20% compared to the base-case values.

Scenario analyses were conducted by examining changes in some assumptions or data sources as opposed to those considered in the base case and assessing their impact on the results. Scenarios with different discount rates for future costs and outcomes (0% and 5%), dosing assumptions (vial sharing instead of wastage), efficacy modelling (different extrapolations for elranatamab OS and PFS curves), administration regimens (RDI set at 100% for all treatments), healthcare resource use (duration of hospitalization for initial elranatamab and teclistamab doses set to 3 days in both cases, instead of 6 days for teclistamab) and drug costs (net prices with 40–70% discounts off list prices, and applying spending caps [36]) were explored.

In the probabilistic sensitivity analysis (PSA), 1000 simulations were run, simultaneously varying all parameters within the uncertainties obtained from reported standard errors (SE) or 95% CI assuming normal distribution, and assuming a 10% SE when these were not available. The Cholesky decomposition was used to estimate the variability of survival curve parameters.