CARTITUDE-1 is a multi-center, phase Ib/II, open-label, single-arm study that recruited patients into its US cohort between July 2018 and October 2019. T cells from enrolled patients were obtained through leukapheresis and used to produce cilta-cel, which was administered via an infusion. Full details for CARTITUDE-1 have been previously published . The present analysis was based on an updated data cut-off of February 2021, corresponding to a median follow-up of 18 months .
Daratumumab Clinical Trials
This study used long-term follow-up data from three daratumumab RRMM clinical trials: POLLUX, CASTOR, and EQUULEUS, which provided information for a large cohort of predominantly ex-US and triple-class exposed patients treated with physician’s choice of therapy after discontinuing the trial treatments . POLLUX and CASTOR were open-label, randomized, phase III trials evaluating the efficacy of daratumumab in combination with lenalidomide-dexamethasone and bortezomib-dexamethasone, respectively [14, 15]. EQUULEUS was an open-label, non-randomized, phase Ib trial evaluating the safety, tolerability, and dose regimen of daratumumab and various combination therapies . Full study details for POLLUX, CASTOR, and EQUULEUS have been previously published [14,15,16]. Data cut-off for patients included in the present analysis was September 2019 for POLLUX and CASTOR and ranged from 2017 to 2019 for EQUULEUS (depending on study arm), corresponding to an overall median follow-up of 16.9 months across all eligible LOTs received since patients became triple-class exposed.
Study Population and Design
Individual patient-level data were available from all data sources. The main analysis included the treated population of CARTITUDE-1, which consisted of patients who were infused with cilta-cel (median of 47 days after apheresis). Patients from the daratumumab clinical trials were included in the external control arm (referred to as the physician’s choice cohort) if they met the key inclusion criteria of CARTITUDE-1 (i.e., triple-class exposed, at least three prior LOTs,Footnote 1 Eastern Cooperative Oncology Group score less than two, creatinine less than or equal to 2 mg/dL,Footnote 2 and disease progression within 12 months of the most recent LOT). Patients in the physician’s choice cohort also must have received the physician’s choice treatment after triple class exposure. Patients in the physician’s choice cohort were reweighted to make their baseline characteristics similar to those of patients in CARTITUDE-1, using average treatment effect in the treated weights derived from propensity scores. This emulated a hypothetical comparative trial in which patients were randomized to receive cilta-cel (represented by CARTITUDE-1) or physician’s choice of treatment (represented by the physician’s choice cohort).
Because patients from the long-term follow-up data from the daratumumab clinical trials were retrospectively enrolled into this comparison, the first LOT initiated by a patient after they met the key eligibility criteria for CARTITUDE-1 could be identified. This differed from CARTITUDE-1, in which cilta-cel is the treatment of interest and patients may have received additional LOTs between first meeting all eligibility criteria and entering the clinical trial. To account for this difference, if a patient in the physician’s choice cohort received multiple LOTs after becoming eligible, that patient contributed multiple observations (corresponding to all eligible LOTs) to the present analysis, provided that the patient was eligible at the start of each LOT. Prior studies have shown this approach to be associated with improved statistical efficiency compared with restricting inclusion to only a single eligible LOT (e.g., first or last) [17, 18]. An exploratory analysis using only the first eligible LOT received by patients in the physician’s choice cohort was conducted to assess potential differences in the two approaches.
To avoid survivorship bias in favor of cilta-cel, treatment lines were excluded from the physician’s choice cohort if a patient experienced progressive disease or death within 47 days of initiating the corresponding LOT, thereby mimicking the median period of 47 days between apheresis and infusion in CARTITUDE-1 . Index dates were defined for treated patients in CARTITUDE-1 as the time of the cilta-cel infusion, and as 47 days after starting the relevant LOT for observations in the physician’s choice cohort.
Baseline Characteristics for Population Alignment
Comparative efficacy estimates can be biased when baseline characteristics that are prognostic of the outcomes differ across non-randomized study cohorts and are left unadjusted . In this study, a clinician-driven process was used to select the most important baseline characteristics for adjustment. First, a list of potential factors for adjustment was identified a priori based on studies from a literature review of the clinical evidence associated with triple-class exposed RRMM. This list was presented to a panel of five independent clinical hematology experts who lead clinical and research initiatives related to MM at major centers in the UK, Germany, and the USA, and modified according to their input. Next, clinicians were asked to rank the variables in order of importance for adjustment. To aid in the process, clinicians were provided with results from univariate regression analyses evaluating the prognostic strength of each factor in terms of PFS and OS in CARTITUDE-1. Clinicians were also provided with the standardized mean differences (SMDs) between each factor in CARTITUDE-1 and the physician’s choice cohort (small, moderate, and substantial differences were represented by SMDs ≤ 0.1, > 0.1 and ≤ 0.2, and > 0.2, respectively ). Rankings were revised iteratively until a consensus was reached. The panel determined that refractory status, cytogenetic profile, International Staging System (ISS) stage, extramedullary plasmacytomas, time to progression on last regimen, number of prior LOTs, years since MM diagnosis, and age represented the minimum set of variables that should be adjusted for in all analyses to ensure clinical validity. Hence, these variables comprised the “base case”. Remaining variables were ranked in order of importance (Table S1 in the Electronic Supplementary Material [ESM]).
Overall response rate (ORR), complete response (CR) or better rate (≥ CR rate), PFS, time to next treatment (TTNT), and OS were assessed. Response and progression were evaluated by an independent review committee in CARTITUDE-1 and by an investigator assessment in the physician’s choice cohort. Overall response rate was calculated as the proportion of patients who achieved a partial response or better according to International Myeloma Working Group (IMWG) criteria . The ≥ CR rate was defined as the percentage of patients who achieved a CR or stringent CR according to IMWG criteria . Progression-free survival was defined as the duration from the index date to the date of progression or death, whichever occurred first. For patients who had not progressed and were alive at the data cut-off, data were censored at the last disease evaluation before the start of any subsequent antimyeloma therapy or the retreatment of cilta-cel (CARTITUDE-1) and at the last follow-up date (physician’s choice cohort). Time to next treatment was defined as the time from the index date to the initiation of the next LOT or death, whichever occurred first, with the last known date alive used in censoring. Overall survival was derived as the duration from the index date to the date of death. Patients who remained alive at the data cut-off or had an unknown vital status were censored at the last known date alive. Efficacy outcomes were evaluated monthly in CARTITUDE-1, every 12 weeks for post-trial treatments received in POLLUX and EQUULEUS (every 6 months as of May 2018), and every 16 weeks for post-trial treatments received in CASTOR.
Inverse probability of treatment weighting (IPTW) was conducted to adjust for baseline differences between patient cohorts . Propensity scores were calculated using a logistic regression model that predicted assignment in the CARTITUDE-1 cohort as a function of baseline covariates. Average treatment effect in the treated weighting was used, wherein patients from CARTITUDE-1 were kept as observed (i.e., assigned a weight of one), whereas patients in the physician’s choice cohort were assigned a weight of p/(1 − p), where p is the propensity score predicting inclusion in the CARTITUDE-1 cohort . Consequently, patients in the physician’s choice cohort who had similar characteristics to CARTITUDE-1 were given a larger weight, making the physician’s choice cohort similar to CARTITUDE-1. The effective sample size (ESS) was calculated to reflect the impact of weighting on the available information in the individual patient-level data . The ESS is equivalent to the number of non-weighted observations needed to produce an effect estimate with the same precision as the indirect treatment comparison, had they been treated in a head-to-head trial .
For all outcomes, comparative efficacy was assessed for the unadjusted comparison (i.e., cilta-cel vs physician’s choice of treatment prior to IPTW) and for the adjusted comparison (i.e., with IPTW). For response outcomes (ORR and ≥CR rate), odds ratios and 95% confidence intervals (CIs) were estimated using logistic regression (with weights applied for the adjusted comparison). The corresponding relative risk (RR) of response and 95% CIs were derived from a Cox regression model with a constant in the time variable (with weights applied for the adjusted comparison) . For time to event outcomes (PFS, TTNT, and OS), hazard ratios (HRs) and 95% CIs were derived from a Cox proportional hazards model (with weights applied for the adjusted comparison). Irrespective of outcome, in the adjusted comparisons, the models also adjusted for the selected covariates, producing doubly robust results . The cluster-robust sandwich variance estimator was used to account for within-person clustering of observations arising from the inclusion of all eligible LOTs in the physician’s choice cohort. The proportional hazards assumption for time to event outcomes was confirmed by a visual inspection of the log-cumulative hazard plot.
For the physician’s choice cohort, variables with missing values (applicable to ISS stage only) were imputed using the most commonly reported category. Imputation was not necessary for the CARTITUDE-1 data, as there were no missing values. Statistical analyses and graphical interpretation were conducted using R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria).
Sensitivity and Subgroup Analyses
Four separate sensitivity analyses and one subgroup analysis were conducted to assess the impact of modifying the patient population, statistical methods, covariates, and outcome definitions. For each analysis, one of these criteria was modified, while all other specifications remained consistent with those outlined in the preceding sections. A comprehensive overview of each analysis is provided in Table S2 in the ESM.
For the first sensitivity analysis, the patient population was expanded to all 113 enrolled patients in CARTITUDE-1, including patients who underwent apheresis but withdrew prior to the cilta-cel infusion. To make the physician’s choice cohort comparable to the enrolled population of CARTITUDE-1, all events from the initiation of physician’s choice of treatment were included. In the second analysis, a multivariable regression model that included the base-case covariates was conducted without the use of weights. The third analysis adjusted for the following variables: hemoglobin level, prior stem cell transplant, Eastern Cooperative Oncology Group status, race, sex, and type of MM, in addition to the base-case variables. The fourth analysis used a modified definition of PFS in the physician’s choice cohort, defined as the time from the index date to disease progression, switch to subsequent treatment, or death due to any cause, whichever occurred first. Last, a subgroup analysis was conducted considering only patients from the treated population of CARTITUDE-1 and the physician’s choice cohort who were penta-refractory (to at least two IMiDs, two PIs, and an anti-CD38 MoAB).
Research Ethics Statement
The CARTITUDE-1 trial protocol was reviewed and approved by an independent ethics committee/institutional review board at all participating sites. All patients participating in the trial provided written informed consent. Similarly, the protocols of the clinical trials of daratumumab (POLLUX , CASTOR , and EQUULEUS ) were approved by local or independent institutional review boards or ethics committees at participating sites and all patients provided written informed consent. The current analyses were conducted in accordance with a protocol and statistical analysis plan developed prior to the start of data analysis.