KarMMa-RW: comparison of idecabtagene vicleucel with real-world outcomes in relapsed and refractory multiple myeloma

Patients with relapsed and refractory multiple myeloma (RRMM) who are triple-class exposed (to an immunomodulatory agent, proteasome inhibitor, and anti-CD38 antibody) have limited treatment options and there is no standard of care. Idecabtagene vicleucel (ide-cel, bb2121), a BCMA-directed CAR T-cell therapy, demonstrated efficacy in triple-class exposed RRMM patients in the KarMMa trial (NCT03361748). In this retrospective study (KarMMa-RW), patient-level data from triple-class exposed RRMM patients were merged into a single data model and compared with KarMMa using trimmed stabilized inverse probability of treatment weighting. Endpoints included overall response rate (ORR; primary), rate of very good partial response or better (≥VGPR), progression-free survival (PFS), and overall survival (OS). Of 1949 real-world triple-class exposed RRMM patients, 190 received subsequent (index) line of therapy and met KarMMa eligibility criteria (Eligible RRMM cohort). With a median follow-up of 13.3 months in KarMMa and 10.2 months in Eligible RRMM, ORR, and ≥VGPR were significantly improved in KarMMa versus Eligible RRMM (ORR, 76.4% vs 32.2%; ≥VGPR, 57.9% vs 13.7%; both P < 0.0001) as were PFS (11.6 vs 3.5 months; P = 0.0004) and OS (20.2 vs 14.7 months; P = 0.0006). This study demonstrated that ide-cel significantly improved responses and survival compared with currently available therapies in triple-class exposed RRMM.


INTRODUCTION
Multiple myeloma is the second most common hematologic malignancy, with~160,000 newly diagnosed cases and 106,000 deaths worldwide in 2018 [1]. Over the past decade, advances in treatment of multiple myeloma have resulted in a significant improvement in overall survival (OS) [2][3][4]. This improvement has been primarily driven by more effective combination therapies of immunomodulatory agents, proteasome inhibitors (PIs), and dexamethasone coupled with consolidation using autologous stem cell transplant [3,5]. New treatment options for multiple myeloma are rapidly evolving, with the approval of anti-CD38 antibodies such as daratumumab and isatuximab, further improving outcomes [6][7][8][9]. Additional novel agents, such as the histone deacetylase inhibitor panobinostat, the anti-SLAMF-7 antibody elotuzumab, and the nuclear export inhibitor selinexor, have been approved by the Food and Drug Administration in recent years for use in combination regimens [10][11][12][13]. Despite dramatic progress in treatment, multiple myeloma remains largely incurable, and almost all patients eventually relapse, with worsening prognosis and survival at each relapse regardless of subsequent treatment [3,14,15]. Previous retrospective data for patients with multiple myeloma refractory to immunomodulatory agents, PIs, and anti-CD38 monoclonal antibodies from 14 different US academic institutions reported a median progression-free survival (PFS) of only 3−4 months and a median OS of 8−9 months [16]. Such triple-class exposed patients with relapsed and refractory multiple myeloma (RRMM) have few treatment options, even with the newly approved therapy belantamab mafodotin [17,18], and there is no clear consensus on the optimal therapy or standard of care [16,[18][19][20][21][22].
Idecabtagene vicleucel (ide-cel, bb2121) is a B-cell maturation antigen (BCMA)-directed chimeric antigen receptor (CAR) T-cell therapy that has demonstrated promising safety and efficacy in RRMM. Ide-cel is generated from autologous T cells transduced with a third-generation lentiviral vector encoding a CAR specific for human BCMA, which consists of the targeting domain of anti-BCMA, costimulatory domain of 4-1BB, and T-cell activation domain of CD3ζ [23]. The phase 1 CRB-401 study (NCT02658929) reported a confirmed ORR of 76% with ide-cel, including complete responses in 39% of patients with ≥3 prior lines of therapy (including immunomodulatory agents and PIs), or double-class refractory disease, and a median PFS of 8.8 months [24]. Frequent, deep and durable responses with ide-cel were recently reported in patients who were triple-class exposed and refractory to their last regimen in the ongoing pivotal phase 2 KarMMa study (NCT03361748) [25]. The safety profile of ide-cel was consistent across both the phase 1 and 2 studies, with mostly grade 1 or 2 adverse events of cytopenia, cytokine-release syndrome, and neurotoxicity.
Large-scale, patient-level retrospective studies can provide a better understanding of outcomes with currently available therapies and help to establish benchmarks for future clinical trials [21,22,26,27]. However, patient-level data in triple-class exposed RRMM patients are not well characterized, and the limited data that are available vary across geographies. Here we describe the demographics, disease characteristics, treatment patterns, and clinical outcomes in real-world RRMM patients with characteristics similar to the KarMMa study population. We also compare clinical outcomes from the real-world RRMM patients treated with currently available therapies and the patients treated with ide-cel in the KarMMa study.

MATERIALS AND METHODS Study design and patients
In this global, noninterventional, retrospective study (KarMMa-RW), and real-world patient-level data were collected from multiple sources and merged into a single data model. Real-world patients with RRMM were initially selected based on broad inclusion and exclusion criteria (Fig. 1). Patients aged ≥18 years with a documented diagnosis of multiple myeloma, who had received ≥3 prior regimens, including an immunomodulatory agent, a PI, and an anti-CD38 antibody (received on or after 16 November 2015, the earliest approval of daratumumab by the Food and Drug Administration) were included in the broad RRMM cohort. Patients in this cohort had to have received ≥3 prior regimens by 30 September 2018, to ensure sufficient follow-up; induction with or without hematopoietic stem cell transplant and with or without maintenance therapy was considered a single regimen. Additional eligibility criteria for this cohort included ≥2 consecutive treatment cycles for each regimen, unless progressive disease was the best response for that regimen. Patients exposed to any BCMA-directed therapy or gene-modified therapy were excluded.
From the broad RRMM cohort, patients who were refractory to their last antimyeloma regimen (defined as documented progression during or within 60 days of last dose before study entry) were selected for the refractory RRMM cohort (Fig. 1). Further selection yielded a subset of patients who received ≥1 documented treatment after they became refractory to their last regimen.
The Eligible RRMM cohort was selected based on additional inclusion and exclusion criteria used for the KarMMa study and International Myeloma Working Group (IMWG) response and progression criteria [28]. Patients in the Eligible RRMM cohort were refractory to their last regimen, had measurable disease based on monoclonal-protein and/or serum free light chain levels, and ≥1 documented disease assessment after the refractory condition or documented death. Additional eligibility criteria are included in Supplementary Methods.
A subset of patients from the Eligible RRMM cohort with baseline characteristics more closely matching those of patients in the KarMMa study (Matched RRMM cohort) were selected from the real-world patientlevel data model and compared with the matched KarMMa cohort. Clinical outcomes in the real-world cohorts were compared with results in the KarMMa population using a composite across the target ide-cel dose levels of 150-450 × 10 6 CAR + T cells as well as the highest target dose level of 450 × 10 6 CAR + T cells.

Data sources
Data were collected on RRMM patients using a set of parameters (Fig. 1). Real-world data from patients in North America and Europe were obtained from three types of data sources: clinical sites, the Connect® MM Registry [29], and external research databases. Data collection was retrospective and did not change clinical practice or patient visit schedules; each data source was in compliance with applicable national and local ethical, legal, and privacy regulations.
Clinical sites in North America and Europe were selected based on predefined criteria including number of patients treated for RRMM, ability to share patient-level data, availability of key clinical information, including Selection process for real-world Eligible Cohort. Collection of patient-level data from clinical sites, the Connect MM Registry, and external research databases, and a description of the selection process for the Eligible RRMM cohort in the KarMMa-RW study. *Baseline was defined as when patients became refractory to their last regimen. † Across all ide-cel target doses in the KarMMa study (ClinicalTrials.gov: NCT03361748); at the data cutoff of 30 October 2019, 58 patients (45.3%) had discontinued from the study, 31 (24.2%) due to death, 26 (20.3%) due to study withdrawal, and 1 (0.8%) lost to follow-up. ‡ Overall, 108 patients (56.8%) discontinued from the study, all due to death. § Numbers (ranges) of matched patients from 30 imputed datasets. COTA denotes the COTA real-world evidence database, ECOG Eastern Cooperative Oncology Group, GRN Guardian Research Network, RRMM relapsed and refractory multiple myeloma. Real-world data were also collected from the Connect® MM Registry (NCT01081028), a multicenter, prospective observational cohort designed to explore the natural history and real-world management of patients with newly diagnosed multiple myeloma. The Registry currently follows 3011 patients from~250 sites in the United States.
External research databases were an enhanced collection of longitudinal data and de-identified patient-level electronic health records in the United States. Additional details can be found in Supplementary Methods.

Data integration
Data from all sources were reformatted into the standardized study data model. Program and mapping specifications for variable transformations and derivations were created for each data element within each data source. Variable transformations included creation of common data types, formats, taxonomy, ontology, as well as data structure. Variable derivations were implemented to define a consistent definition for regimen, baseline, index date (study day 1), study entry, outcome response, comorbidities, and imputation of missing date values across all data sources. All transformations and analytic derivation decisions and their lineages to the source data were documented.

Endpoints and assessments
The primary endpoint was ORR, defined as partial response (PR) or better per IMWG criteria [28]. Secondary endpoints included complete response (CR) rate, rate of very good partial response or better (≥VGPR), PFS, OS, duration of response (DOR), and time to response (TTR). Subgroup assessments of ORR and PFS stratified by sex, age, double-refractory status (immunomodulatory agents and PIs), and number of prior antimyeloma regimens per year were performed.

Statistical analysis
Propensity score balancing was used to summarize the impact of covariates on treatment selection into scalar values. These values were then used for weighting individual patients in both the Eligible RRMM cohort and the KarMMa cohort, using inverse probability treatment weighting (IPTW). Furthermore, a subset of the two cohorts were matched in order to improve the balance of covariates between groups [30][31][32].
Steps utilized in the propensity score balancing process are reported in Supplementary Methods. Using this methodology, results in the KarMMa cohort were adjusted to account for differences in patient characteristics between cohorts and are therefore slightly different from results reported with the KarMMa study.
For primary analyses, trimmed stabilized IPTW was used to compare clinical outcomes/endpoints of interest. Baseline prognostic variables considered for IPTW were selected and ranked by a scientific steering committee. Multiple imputation was used to address missing values with propensity scores generated in each of 30 separate datasets.
The initial study protocol plan was to have variable ratio matching using the propensity score modeling to achieve an overall 2:1 ratio of the KarMMa cohort to Eligible RRMM cohort; however, due to the resulting small sample size, a matched-pair comparison between the KarMMa cohort and the Matched RRMM cohort was included as a supporting sensitivity analysis instead. An untrimmed stabilized IPTW was used to compare the KarMMa and RRMM cohorts as an additional sensitivity analysis.
Poisson regression was used to analyze ORR and ≥VGPR rates due to nonconvergence of the binomial regression, and Cox proportional hazards models were used to analyze PFS, OS, and DOR. Because analyzing CR requires a bone marrow biopsy per IMWG criteria, which is generally not available for real-world data, analyses were summarized for ≥VGPR rate to avoid underestimating depth of response in the real-world setting. All models were adjusted for unbalanced covariates and overall summary estimates generated from the 30 separate datasets.
Analyses were performed using SAS version 9.4. Descriptive analyses were performed to gain an understanding of the quality of the data and statistical distributions of characteristics of the real-world patients. Continuous variables were described by mean with standard deviation and 95% confidence interval (CI), median, upper and lower quartiles and range values. Categorical variables were reported as number and percentage with 95% CIs. Two-sided P < 0.05 was considered statistically significant without multiplicity adjustment.

Patients
Patient-level data were collected from 1949 real-world patients with RRMM who received ≥3 prior regimens, including an immunomodulatory agent, a PI, and an anti-CD38 antibody. Of these 1949 realworld triple-class exposed RRMM patients, 1171 were refractory to their last regimen at baseline (RRMM cohort). The median age of patients in the RRMM cohort was 68 years, the median number of prior regimens was 5, and 19.0% of patients had received ≥7 prior regimens; 40.9% were triple-class refractory (Table 1). Among the 1171 patients in the RRMM cohort, 528 had received a subsequent line of therapy (hereafter referred to as the subsequent [index] line therapy); 643 patients were excluded for having no new treatment or due to death (n = 441) and lack of follow-up (n = 202). Additional eligibility criteria from the KarMMa study [33] were applied to the 528 patients, resulting in the selection of 190 patients for the Eligible RRMM cohort. A total of 128 ide-cel treated patients from the KarMMa study were compared with the Eligible RRMM cohort (Fig.  1). At the data cutoff (30 October 2019), 108 patients (57%) in the Eligible RRMM cohort had died.

Overall response rate
The efficacy parameters were significantly improved in the KarMMa cohort of all ide-cel treated patients across all target doses, compared with the Eligible RRMM cohort (  Table 3).
Median OS was significantly improved with ide-cel in KarMMa across all target doses, versus the Eligible RRMM cohort (20.2 months vs 14.7 months; HR, 0.45 [95% CI, 0.28-0.71]; P = 0.0006); median follow-up among surviving patients was 14.4 months in the KarMMa cohort and 15.0 months in the Eligible RRMM cohort (Fig. 3A). Median OS was not reached at the highest target dose in the KarMMa cohort, versus 14.2 months in the Eligible RRMM cohort (HR, 0.32 [95% CI, 0.15-0.72]; P = 0.0055) (Fig. 3B). The estimated 12-month probability of surviving was 80% in the KarMMa cohort across all target doses and 56% in the Eligible RRMM cohort. The estimated 12-month probability of surviving was 82% at the target dose of 450 × 10 6 CAR + T cells in the KarMMa cohort and 53% in the Eligible RRMM cohort. Furthermore, consistent improvements in ORR and PFS in the KarMMa cohort versus the Eligible RRMM cohort were demonstrated across all subgroups, including patients aged ≥65 years, patients with double-class refractory disease, and those with >1 prior antimyeloma regimen per year (Fig. 4).   Table 4). The median DOR was 11.1 months (95% CI, 10.8-11.5) in the KarMMa cohort and 9.0 months (95% CI, 7.

DISCUSSION
Results from KarMMa-RW, a retrospective study of patients with triple-class exposed RRMM, confirmed that there is no standard of care therapy in this difficult-to-treat patient population. Real-world patients analyzed in this study received 94 different treatment regimens as a subsequent (index) line therapy, most commonly triplet combinations of immunotherapy, PIs or chemotherapy, and dexamethasone. Consistent with previous studies, demonstrating a poor prognosis in triple-class exposed RRMM patients [3,16], clinical outcomes were suboptimal in the real-world RRMM patients, with an ORR of 32.2% and a median PFS of 3.5 months.
With no clear standard of care, a more precise understanding of the expected outcomes with available therapies in RRMM in the real-world is needed. In real-world RRMM patients, comprehensive large-scale data are limited. A comparative analysis of results from the STORM part 2 study in patients with triple-class exposed RRMM who received selinexor plus low-dose dexamethasone and a retrospective cohort from the Flatiron Health Analytic Database was conducted; however, the analysis had a limited sample size of <70 patients per group [34,35]. The KarMMa-RW study is a large-scale, patientlevel examination of outcomes with currently available treatments in real-world RRMM patients and the first study to directly compare results in triple-class exposed RRMM patients receiving ide-cel in the KarMMa study with outcomes in a similar group of real-world patients.
In KarMMa-RW, responses and survival outcomes were significantly improved in ide-cel−treated patients in the KarMMa study versus real-world RRMM patients (ORR and ≥VGPR, P < 0.0001; PFS, P = 0.0004; OS, P = 0.0006). Of note, the OS data were considered immature at the time of analysis. In addition, patients in the KarMMa study were more heavily pretreated and more had double-or triple-class refractory disease than real-world RRMM patients. A trend toward improved response durability was observed in favor of ide-cel (P = 0.1372), which may be explained by the smaller sample sizes of the compared cohorts due to the inclusion of only patients with a ≥PR (KarMMa cohort, n = 94; Eligible RRMM cohort, n = 58). Furthermore, consistent improvements in depth of response and survival were observed with idecel across subgroups, including older patients, patients with double-class refractory disease, and those who received multiple prior regimens per year. As the ongoing KarMMa study demonstrated greatest efficacy at the highest target dose of 450 × 10 6 CAR + T cells (ORR, 82%; median DOR, 11.3 months; median PFS, 12.1 months) [25], outcomes in real-world RRMM patients were compared with those observed in KarMMa patients who received 450 × 10 6 CAR + T cells. Results were consistent with the comparison across ide-cel dose levels, with significant improvements in response rates (ORR and ≥VGPR, P < 0.0001) and survival (PFS; P < 0.0001, OS; P = 0.0055) observed with ide-cel. Given the small sample size at the highest target dose, a comparison of DOR and TTR was not feasible.
Patients with RRMM who progress on anti-CD38 antibody therapy have poor survival outcomes, as demonstrated in the retrospective MAMMOTH study, with a median PFS of 3.4 months and a median OS of 9.3 months [16]. Results are similar with recently approved therapies; selinexor plus dexamethasone treatment resulted in a median PFS of 3.7 months and median OS of 8.6 months in the STORM part 2 study and belantamab mafodotin resulted in a median PFS of 2.8 months (2.5 mg/kg) and 3.9 months (3.4 mg/kg) [36,37]. To compare outcomes with selinexor plus dexamethasone or belantamab mafodotin with those of a one-time infusion of ide-cel in triple-class exposed RRMM, a matching adjusted indirect comparison was performed which demonstrated improvements in ORR, PFS, and OS in idecel-treated patients [18,38].
In KarMMa-RW, the stringent inclusion and exclusion criteria applied to select patients with comparable baseline features as well as the propensity score methods ensured robust and reliable comparisons with the KarMMa study population. Sensitivity analyses, which included only patients with an Eastern Cooperative Oncology Group performance status of 0 or 1 within the realworld Eligible RRMM cohort, confirmed the overall results. However, rigorous selection criteria for matching resulted in only 76-80 patients in the Eligible RRMM and KarMMa cohorts. Thus, the exact number of patients matched varied for the 30 different datasets generated through multiple imputations. Throughout the study, trimmed stabilized IPTW was applied to optimize the number of adjusted and compared patients in the Eligible RRMM and KarMMa cohorts. All covariates were well balanced for the trimmed IPTW analysis except age and calcium, which were further balanced after matching. As the results were presented with trimmed stabilized IPTW to account for differences in patient characteristics between the two cohorts, outcomes from KarMMa differ slightly from previously reported outcomes in the primary KarMMa study [25]. Limitations of this study include potential bias and the fact that unmeasured confounders could not be controlled, which may have influenced the balancing of the two cohorts. Additionally, since this study was designed to develop a synthetic cohort reflective of the KarMMa population and excluded patients not healthy enough to receive ide-cel treatment as a next line of therapy, the 190 real-world Eligible RRMM patients selected from the broad RRMM cohort (N = 1949) may not fully represent the general RRMM population. For this reason, clinical implications of the improved outcomes observed with ide-cel therapy should be carefully drawn, accounting for a patient's health status and treatment history in real-world practice settings.
Response rate, depth of response, and time to disease progression decrease with each subsequent line of therapy, making late-stage multiple myeloma difficult to treat effectively [15,[39][40][41]. As confirmed by the KarMMa-RW study, outcomes are suboptimal with current treatment options in real-world RRMM patients who have received three or more prior therapies, including an immunomodulatory agent, a PI, and an anti-CD38 antibody. This study demonstrated a clear benefit with ide-cel treatment over currently available therapies, with significant increases in efficacy. Thus, ide-cel offers a promising new treatment option in triple-class exposed RRMM.

DATA SHARING STATEMENT
Bristol-Myers Squibb Company policy on data sharing may be found at https://www.bms.com/researchers-and-partners/ independent-research/data-sharing-request-process.html. Fig. 4 Subgroup analyses of overall response rate and progression-free survival. A shows overall response rate (confirmed partial response or better) by patient demographics and disease characteristics. B shows progression-free survival by patient demographics and disease characteristics. Patients with a partial response or better according to International Myeloma Working Group criteria applying Food and Drug Administration censoring rules were included. Multiple imputations were performed to create 30 datasets; estimates for the analyses were then obtained using Rubin's rule to combine the individual estimates from each dataset. *Double-class refractory was defined as refractory to an immunomodulatory agent and a proteasome inhibitor. ORR denotes overall response rate, PFS progression-free survival.