Leukemia & Lymphoma ISSN: 1042-8194 (Print) 1029-2403 (Online) Journal homepage: https://www.tandfonline.com/loi/ilal20 Carfilzomib and dexamethasone versus eight cycles of bortezomib and dexamethasone in patients with relapsed or refractory multiple myeloma: an indirect comparison using data from the phase 3 ENDEAVOR and CASTOR trials Katja Weisel, Istvan Majer, Lucy DeCosta, Albert Oriol, Hartmut Goldschmidt, Heinz Ludwig, Marco Campioni, Zsolt Szabo & Meletios Dimopoulos To cite this article: Katja Weisel, Istvan Majer, Lucy DeCosta, Albert Oriol, Hartmut Goldschmidt, Heinz Ludwig, Marco Campioni, Zsolt Szabo & Meletios Dimopoulos (2020) Carfilzomib and dexamethasone versus eight cycles of bortezomib and dexamethasone in patients with relapsed or refractory multiple myeloma: an indirect comparison using data from the phase 3 ENDEAVOR and CASTOR trials, Leukemia & Lymphoma, 61:1, 37-46, DOI: 10.1080/10428194.2019.1648806 To link to this article: https://doi.org/10.1080/10428194.2019.1648806 © 2019 The Author(s). Published by Informa Published online: 22 Oct 2019. UK Limited, trading as Taylor & Francis Group. Submit your article to this journal Article views: 2290 View related articles View Crossmark data Citing articles: 1 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ilal20 LEUKEMIA & LYMPHOMA 2020, VOL. 61, NO. 1, 37–46 https://doi.org/10.1080/10428194.2019.1648806 ORIGINAL ARTICLE Carfilzomib and dexamethasone versus eight cycles of bortezomib and dexamethasone in patients with relapsed or refractory multiple myeloma: an indirect comparison using data from the phase 3 ENDEAVOR and CASTOR trials Katja Weisela,b, Istvan Majerc, Lucy DeCostad, Albert Oriole, Hartmut Goldschmidtf, Heinz Ludwigg, Marco Campionic, Zsolt Szaboc and Meletios Dimopoulosh a Department of Oncology and Hematology, University Medical Center of Hamburg-Eppendorf, Hamburg, Germany; bDepartment of Hematology, Oncology, Immunology, Rheumatology, and Pulmonology, Medical Clinic II, T€ubingen, Germany; cAmgen Europe GmbH, Rotkreuz, Switzerland; dAmgen Ltd, Uxbridge, UK; eHospital Germans Trias i Pujol, Institut Josep Carreras, Barcelona, Spain; fInternal Medicine V and National Center of Tumor Diseases, University Clinic Heidelberg, Heidelberg, Germany; gWilhelminen Cancer Research Institute, Wilhelminenspital, Vienna, Austria; hSchool of Medicine, National and Kapodistrian University of Athens, Alexandra Hospital, Athens, Greece ABSTRACT ARTICLE HISTORY In ENDEAVOR, carfilzomib and dexamethasone (Kd56) demonstrated significant improvement in Received 6 March 2019 progression-free survival (PFS) compared with bortezomib and dexamethasone (Vd). Both agents Accepted 12 July 2019 were administered until disease progression; the EU label for Vd, however, stipulates a max- KEYWORDS imum of eight treatment cycles. Here, matching-adjusted treatment comparison was used to Multiple myeloma; compare efficacy of Kd56 with Vd, if Vd was administered for 8 cycles (Vd-8). Data from proteasome inhibitor ENDEAVOR and CASTOR trials (which compared daratumumab, bortezomib, and dexamethasone treatment duration; with Vd-8) were used. Hazard ratios of PFS were estimated for Vd vs. Vd-8 and Kd vs. Vd-8. For matching-adjusted indirect cycles 1–8, risk reduction in PFS for Kd56 vs. Vd-8 was equal to that estimated in ENDEAVOR treatment comparison (HR: 0.53; 95% CI 0.44–0.65). Beyond eight cycles, risk reduction in PFS for Kd56 and Vd-8 was estimated to be 60% (HR: 0.40; 95% CI 0.26–0.63). The analysis suggested that PFS benefit of Kd56 over Vd increases when Vd is given for eight cycles only. Introduction approved in Europe for use in combination with dexa- methasone (Kd56; K administered at an ultimate dose The management of patients with multiple myeloma of 56 mg/m2), or with lenalidomide and dexamethasone (MM) has evolved dramatically over the past decade owing to the introduction of more effective agents, (KRd; K administered at a dose of 27 mg/m2) for the such as proteasome inhibitors and immunomodulatory treatment of adults patients with MM who have drugs [1,2]. More recently, monoclonal antibodies, received at least one prior therapy [11,12]. The approval such as elotuzumab and daratumumab, and the his- of Kd56 was based on findings from the pivotal phase 3 tone deacetylase inhibitor, panobinostat, have also randomized, controlled trial ENDEAVOR (NCT01568866), been approved [3–5]. Among proteasome inhibitors, a head-to-head comparison of Kd56 with Vd in patients bortezomib (V) was the first agent to gain approval with RRMM. In order to have a direct head-to-head for the treatment of MM [6]. The combination of dexa- comparison that was not biased by a protocol-defined methasone and bortezomib (Vd) has been established duration of treatment, Kd56 was compared with Vd as a key regimen in the treatment of relapsed and/or under the same treatment schedule, that is: until dis- refractory MM (RRMM) [7–9]. ease progression, withdrawal of consent, or unaccept- Carfilzomib (K) is a next-generation proteasome able toxicity [13,14]. Thus, in ENDEAVOR, the median inhibitor that irreversibly binds to the proteasome, thus (range) of Vd treatment duration was eight (5.0–15.0) resulting in more sustained inhibition than with the cycles, with 25% of patients receiving 15 cycles. This is reversible proteasome inhibitor V [9,10]. Carfilzomib is different from the treatment schedule in the CONTACT Katja Weisel katja.weisel@uke.de Department of Oncology and Hematology, University Medical Center of Hamburg-Eppendorf, Hamburg, Germany ß 2019 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by- nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way. 38 K. WEISEL ET AL. bortezomib summary of product information and in event (TEAE) data observed with Kd56 and Vd in other clinical trials where Vd is generally administered ENDEAVOR were used (data cutoff 19 July 2017) [19]. according to the European label, that is, for up to eight cycles only [15]. ENDEAVOR and CASTOR study designs, patients, In the ENDEAVOR trial, Kd56 was superior to Vd for and treatments the primary outcome of progression-free survival (PFS), with 18.7 months vs. 9.4 months median PFS for The study designs and participant information of the Kd56 and Vd, respectively (hazard ratio (HR) 0.53; 95% ENDEAVOR [13] and CASTOR [17,20] trials have been confidence interval (CI) 0.44, 0.65, p < .0001) [13]. described previously. In brief, both studies included ENDEAVOR also demonstrated a significant improve- adult patients with RRMM who had received one to ment in overall survival (OS) for Kd56 over Vd (median three (ENDEAVOR) or at least one (CASTOR) previous OS 47.6 months vs. 40.0 months, HR 0.79; 95% CI 0.65, line of therapy, who had achieved at least a partial 0.96; p¼.010) [14]. Adverse drug reactions were more response to a previous line of therapy, and who had frequently reported in patients who were receiving adequate renal, hematological, and hepatic function. Kd56 than in those who were receiving Vd, including In ENDEAVOR, patients were randomized (1:1) to hypertension and cardiac failure. In contrast, the num- receive Kd56 (n ¼ 464) or Vd (n ¼ 465), administered ber of patients with peripheral neuropathy was signifi- during 28-day (Kd56) or 21-day (Vd) cycles [13]. In cantly lower in the Kd56 group than in the Vd group CASTOR, patients were randomized (1:1) to receive [13,14]. It has been demonstrated that Kd56 is gener- DVd (n ¼ 251) or Vd-8 (n ¼ 247), administered during ally well tolerated and has a favorable benefit–risk 21-day cycles as described previously [17]. Treatment profile [16]. cycles were repeated eight times; in the DVd group, The present analysis aims to characterize the clin- daratumumab was continued after cycle 8, until pro- ical benefits of Kd56 vs. Vd by estimating the differ- gressive disease [17]. Patients in the ENDEAVOR and ence in efficacy and safety between Kd56 when given CASTOR trials provided written informed consent, and until disease progression and Vd administered for up the trials were conducted in accordance with the prin- to eight cycles as per the EU label. In the absence of ciples of the Declaration of Helsinki and the direct head-to-head comparative data between Kd56 International Conference on Harmonization Good and Vd-8, this analysis was based on matching- Clinical Practice guidelines. adjusted indirect treatment comparisons (MAIC) using data from the ENDEAVOR trial and the phase 3 CASTOR trial, which compared daratumumab, bortezo- Efficacy analyses mib, and dexamethasone (DVd) with Vd-8 [17,18]. PFS outcomes with Kd56 and Vd-8 were compared using a three-step modeling approach (Figure 1). In Materials and methods the first step, to minimize any confounding bias owing to observed differences across trial populations, Data patients receiving Vd in ENDEAVOR and patients For the efficacy analysis, PFS data for Kd56 and Vd receiving Vd-8 in CASTOR were matched for observed (until progression) from ENDEAVOR, and for Vd-8 from baseline characteristics using MAIC [21]. In the base CASTOR (NCT02136134) were used. PFS was the pri- case analysis, among all available variables, those con- mary outcome in both trials, and data from the lon- sidered to be predictive and prognostic factors for PFS gest published follow-up times were selected for and OS were selected for the matching. This approach inclusion in the analysis. In ENDEAVOR, the assessment is in agreement with the methods guide issued by the of progression was based on the findings of an inde- National Institute for Health and Care Excellence for pendent review committee at the first prespecified population-adjusted indirect treatment comparisons interim analysis, data cutoff 10 November 2014 [22]. The selected variables reflected disease character- (median follow-up 11.1 months (Vd) and 11.9 months istics and treatment history, and were generally con- (Kd56)) [13]. In CASTOR, the assessment of progression sistent with validated prognostic markers currently was based on a computerized algorithm and updated used in stratifying patients into different risk groups safety and efficacy data, cutoff 30 August 2017 [23,24]. Details of the variable selection process have (median follow-up 26.9 months) [18]. For the safety been reported previously [25,26]. Table 1 presents an analysis of Kd56 vs. Vd-8, treatment-emergent adverse overview of parameters used for the MAICs. KD56 VS. 8 CYCLES OF VD IN ENDEAVOR 39 ENDEAVOR Vd patients were matched to average baseline characteristics Matching of CASTOR Vd-8 patients using matching-adjusted indirect comparison methodology. 1. Virtual patient-level data were constructed for Vd-8 in CASTOR. Estimating the 2. A piecewise Cox regression model was fitted to the matched ENDEAVOR excess risk of Vd data and the reconstructed CASTOR Vd-8 data using 8 cycles as progression or cutoff date. death associated 3. The increased PFS risk due to stopping Vd treatment after 8 cycles vs. with stopping Vd continuing Vd treatment beyond 8 cycles until progression was assessed after 8 cycles based on the results of the piecewise model. Estimating the After adjusting Vd-8 for patient selection criteria, observed, and unobserved PFS benefit of differences between CASTOR and ENDEAVOR, an overall HR was estimated Kd56 over Vd-8 for Kd56 vs. Vd-8. Figure 1. Summary of the stepwise modeling approach. HR: hazard ratio; Kd56: carfilzomib and dexamethasone; Vd: bortezomib and dexamethasone; Vd-8: bortezomib and dexamethasone given until eight cycles. Table 1. Baseline characteristics of patients receiving Vd in ENDEAVOR and Vd-8 in CASTOR. Vd Vd Vd-8 ENDEAVOR ENDEAVOR CASTOR (before matching) (after matching)a (n¼ 465) (n¼ 378)b (n¼ 247) Age, n (%) <65 years 210 (45.2%) 191 (50.6%) 125 (50.6%) 65–74 years 189 (40.6%) 133 (35.2%) 87 (35.2%) 75þ years 66 (14.2%) 54 (14.2%) 35 (14.2%) Sex, n (%) Male 229 (49.2%) 184 (48.6%) 147 (59.5%) International Staging System, n (%)c Stage 1 204 (44.1%) 147 (38.9%) 96 (38.9%) Stage 2 151 (32.5%) 153 (40.5%) 100 (40.5%) Stage 3 109 (23.4%) 78 (20.6%) 51 (20.6%) Baseline creatinine clearance >60 mL/min, n (%) 297(63.9%) 264 (70.0%) 163 (70.0%)d Type of multiple myeloma IgG 284 (61.1%) 230 (60.9%) 138 (55.9%) Other 181 (38.9%) 148 (39.1%) 109 (44.1%) Time since diagnosis in years, median 3.61 2.92 3.72 Patients with time since diagnosis 3.72 years (%) 47.6% 50.0% 50% Number of prior therapies, n (%) 1 232 (49.9%) 173 (45.7%) 113 (45.7%) 2 145 (31.2%) 113 (30.0%) 74 (30.0%) 3þ 87 (18.9%) 92 (24.3%) 60 (24.3%) Prior bortezomib, n (%) 252 (54.2%) 251 (66.4%) 164 (66.4%) Prior IMID, n (%) 348(74.8%) 303 (80.2%) 198 (80.2%) Prior lenalidomide 177 (38.1%) 155 (41.1%) 120 (48.6%) Prior PI þ IMID, n (%) 167 (35.9%) 189 (50.1%) 129 (52.2%) Prior stem cell transplantation, n (%) 272 (58.5%) 228 (60.3%) 149 (60.3%) Refractory to last prior therapy, n (%) 188 (40.4%) 130 (34.4%) 85 (34.4%) Cytogenetic profile, n (%) Standard risk 291 (62.6%) 233 (61.7%) 135 (54.7%) High risk 113 (24.3%) 93 (24.5%) 51 (20.6%) Missing 61 (13.1%) 52 (13.8%) 61 (24.7%) IMID: immunomodulatory imide drugs; PI: proteasome inhibitor; Vd: bortezomib and dexamethasone. a Patients receiving Vd in ENDEAVOR were matched to patients receiving Vd-8 in CASTOR based on observed baseline characteristics using matching-adjusted indirect treatment comparison methodology. Base case matching analysis; the following variables were used for the matching: age (65, 75, other), time from diagnosis, International Staging System (2, 3, other), creatinine clearance (>60 mL/min, other), number of prior therapies (1, 2, other), prior stem cell transplantation (yes, no), prior bortezomib (yes, no), prior IMID (yes, no), and refractory to last prior therapy (yes, no). b Sample size was calculated as the sum of weights. c ISS measured at the start of treatment was used for matching. d Based on 233 evaluable patients. 40 K. WEISEL ET AL. In the second step, in the absence of access to cycles, TEAEs with an onset date from the first dose patient-level PFS data from the CASTOR trial, the pub- until 30 days after the last dose of any study drug lished Kaplan–Meier curves were digitized to generate were included in the analysis. For patients who contin- virtual patient-level CASTOR data that replicated the ued Vd beyond cycle 8, TEAEs with an onset date Kaplan–Meier curves [27]. Then, a piecewise Cox from cycle 9 to the last dose of any study drug were regression model was fitted to the matched not included. Analysis of TEAEs included crude inci- ENDEAVOR Vd data and CASTOR Vd-8 data. In the dence proportions and exposure-adjusted incidences piecewise Cox model, the cutoff date was determined per treatment group; Kd56/Vd-8 exposure-adjusted at 24 weeks (corresponding to eight cycles of Vd). This incidence ratios and associated 95% CI were also cal- model specification allowed separate HRs to be esti- culated using the Poisson log-linear models. mated for the first eight cycles, and for the treatment time beyond cycle 8. The HR for the first period cap- Results tured all unobserved heterogeneity in baseline charac- teristics and between-trial differences between the Efficacy analyses matched PFS of ENDEAVOR Vd and the PFS of To estimate the impact of stopping Vd after eight CASTOR Vd-8 (i.e. visually, the difference between the cycles on PFS, as a first step, patients who were matched Vd and the Vd-8 PFS curves during the first receiving Vd in ENDEAVOR were matched to the base- eight cycles). The HR estimated for the period beyond line characteristics of those receiving Vd-8 in CASTOR. eight cycles captured both unobserved heterogeneity The observed baseline characteristics of patients and the impact of stopping Vd after eight cycles. receiving Vd in ENDEAVOR (n ¼ 465) [13] and those Comparing these two HRs allowed the impact of receiving Vd-8 in CASTOR (n ¼ 247) [18,20] have been unobserved heterogeneity to be quantified and subse- published previously and are summarized in Table 1. quently the net impact of stopping Vd after eight In both trials, the median age was 64–65 years, cycles in ENDEAVOR to be quantified. 46–50% had received one line of prior therapy, and In the third step, adjusting for the between-trial dif- 50–54% had received two or more lines of prior ther- ferences in terms of patient selection criteria and apy. Patients in the CASTOR trial had slightly more observed baseline patient characteristics, as well as for advanced disease (61% with International Staging unobserved differences between Vd-8 in CASTOR and System (ISS) scores  2) than those in ENDEAVOR Vd in ENDEAVOR, an overall HR was estimated for (56% with ISS scores 2). Matching for observed base- Kd56 vs. Vd-8 capturing the difference in the risk of line characteristics between the two trial populations progression or death during the first eight cycles and meant that differences were effectively eliminated beyond eight cycles. (Table 1). To assess the robustness of the results, three scen- After matching for observed baseline characteristics, ario analyses were explored: PFS for Vd in ENDEAVOR a difference in PFS (up to and beyond eight cycles) truncated at 16 months (i.e. where the number of was still evident between patients who were receiving patients at risk decreased below 5%), all available vari- Vd in ENDEAVOR and Vd-8 in CASTOR (Figure 2). This ables used for the matching, and no matching. difference in PFS could be attributed to the unob- In addition, to assess whether continuing treatment, served heterogeneity in baseline characteristics, to in particular Vd treatment, until progression leads to between-trial heterogeneity that cannot be adjusted deeper responses over time than stopping treatment for by matching, and to stopping Vd after eight cycles. after eight cycles, cumulative proportions of patients To assess the net impact of stopping Vd after eight in ENDEAVOR with a complete response (CR) or better, cycles, the follow-up period was divided into two parts or with a very good partial response (VGPR) or better comprising cycles 1–8 and cycle 9 or more, respect- (for definitions of CR and VGPR, see [13]) were calcu- ively. For the period of cycles 1–8, during which any lated for Kd56 and Vd from the start of the treatment. difference in PFS between the matched ENDEAVOR Vd and CASTOR Vd patients could be attributed to unob- served heterogeneity, the HR of PFS for ENDEAVOR Vd Safety analyses vs. CASTOR Vd was 1.36 (95% CI 1.01, 1.81). For the TEAEs (coded using MedDRA version 20.0) were ana- period of cycle 9 or more, during which a difference lyzed based on the Kd56 and Vd safety data observed in PFS between the matched ENDEAVOR Vd and in ENDEAVOR [19]. For patients who were receiving CASTOR Vd patients could be due to unobserved het- Kd56, and for those receiving Vd for up to eight erogeneity and the difference in Vd treatment KD56 VS. 8 CYCLES OF VD IN ENDEAVOR 41 100 ENDEAVOR Vd ENDEAVOR Vd matched to CASTOR CASTOR Vd reconstructed 80 9+ cycles Progression-free survival (%) ENDEAVOR Vd vs CASTOR-matched ENDEAVOR Vd: 60 HR: 1.79 (1.36–2.34) 1–8 cycles 40 ENDEAVOR Vd vs CASTOR- matched ENDEAVOR Vd: HR: 1.36 20 (1.01–1.81) 0 0 6 12 18 24 30 36 Months Figure 2. Impact on PFS of stopping Vd after eight cycles vs. continuing until progression. The HRs of PFS for the periods of cycles 1–8 cycles and cycles 9þ were estimated using a piecewise Cox regression model fitted to CASTOR-matched ENDEAVOR Vd data and reconstructed CASTOR Vd data. Vd cycle length was 21 days. CI: confidence interval; HR: hazard ratio; PFS: progression- free survival; Vd: bortezomib and dexamethasone. ENDEAVOR Kd56 ENDEAVOR Vd 100 ENDEAVOR Vd 8 cycles (estimated) 9+ cycles 80 Kd56 vs Vd 8 cycles: Progression-free survival (%) HR: 0.40 (0.26–0.63) 60 1–8 cycles 40 Kd56 vs Vd 8 cycles: HR: 0.53 (0.44–0.65) 20 0 0 6 12 18 24 30 36 Months Figure 3. Difference in PFS with Kd56 vs. Vd eight cycles. The difference in PFS for Kd56 vs. Vd eight cycles was estimated using ENDEAVOR Vd data adjusted for the increased PFS risk associated with stopping Vd after eight cycles. Vd cycle length was 21 days. CI: confidence interval; HR: hazard ratio; Kd56: carfilzomib and dexamethasone; PFS: progression-free survival; Vd: bortezo- mib and dexamethasone. duration, the HR of PFS for Vd ENDEAVOR vs. Vd between Kd56 and Vd-8 was estimated, assuming Vd CASTOR was 1.79 (95% CI 1.36, 2.34). The ratio would have been stopped after eight cycles. For cycles between the two HRs provides an estimate of the net 1–8, the risk reduction in PFS for Kd56 vs. Vd-8 cycles impact on PFS of stopping Vd treatment after eight was equal to that estimated for Kd56 vs. Vd in cycles; thus, stopping Vd after eight cycles was esti- ENDEAVOR (HR: 0.53; 95% CI 0.44, 0.65) [13]. Beyond mated to increase the risk of progression or death by eight cycles, the risk reduction in PFS for Kd56 and 32% compared with Vd treatment until progression Vd-8 was estimated to be 60% (HR: 0.40; 95% CI 0.26, (HR: 1.32; 95% CI 0.89, 1.96). The difference in PFS 0.63; Figure 3). The corresponding overall risk 42 K. WEISEL ET AL. Table 2. Excess risk of progression or death associated with stopping Vd after eight cycles vs. continuous Vd until progression, and efficacy of Kd56 vs. Vd-8. Excess risk of progression or death associated with Analysis stopping Vd after 8 cycles vs. Vd until progression PFS HR for Kd56 vs. Vd-8 Base case analysisa 32% (–11%, 96%) 0.44 (0.30; 0.66) Truncated PFS at 16 months 30% (–13%, 94%) 0.45 (0.30; 0.68) All variables used for matchingb 38% (–8%, 108%) 0.43 (0.29; 0.65) No matching 41% (–3%, 106%) 0.43 (0.30; 0.61) HR: hazard ratio; IMID: immunomodulatory imide drugs; Kd56: carfilzomib and dexamethasone; PFS: progression-free survival; PI: proteasome inhibitor; Vd: bortezomib and dexamethasone; Vd-8: bortezomib and dexamethasone given until eight cycles a Base case covariates: age (65, 75, other), time from diagnosis, International Staging System (2, 3, other), creatinine clearance (>60 mL/min, other), number of prior therapies (1, 2, other), prior stem cell transplantation (yes, no), prior bortezomib (yes, no), prior IMID (yes, no), refractory to last prior therapy (yes, no). b All available covariates: besides the covariates in the base case analysis, sex (male, female), type of multiple myeloma (IgG, other), prior lenalidomide (yes, no), prior PI þ IMID (yes, no), and cytogenetic profile (standard risk, high risk, other). 100 100 Cumulative VGPR or better rate (%) Cumulative CR or better rate (%) 80 80 ENDEAVOR Kd56 ENDEAVOR Vd 60 60 8 × 21-day cycles 40 40 20 20 0 0 0 3 6 9 12 15 18 0 3 6 9 12 15 18 Time since start of treatment (months) Time since start of treatment (months) Figure 4. Cumulative rates of patients with a complete response or better or with a very good partial response or better with Kd56 and Vd in ENDEAVOR. Cumulative rate of CR or better or VGPR or better was assessed from the start of treatment in the safety population by an independent review committee (Kd: n ¼ 463; Vd: n ¼ 456). The time at which treatment for Vd eight cycles would be discontinued (i.e. at 8  21-day cycles) is indicated by a dashed line. CR: complete response; Kd56: carfilzomib and dexamethasone; Vd: bortezomib and dexamethasone arm; VGPR: very good partial response. reduction, that is, capturing both the first eight cycles CR or better and VGPR or better were higher in and the period beyond that, in a single measure, was patients receiving Kd56 than in those receiving Vd. 56% for Kd56 vs. Vd-8 (HR: 0.44; 95% CI 0.30, 0.66). During the first eight cycles, rates of CR or better Similar results were obtained in scenario analyses. and VGPR or better increased steadily for patients The excess risk of progression or death associated receiving Vd and started to become stable during sub- with stopping Vd after eight cycles vs. continuing Vd sequent cycles. until progression was estimated to be between 30% and 41%. The overall risk reduction associated with Safety analyses Kd56 vs. Vd-8 was estimated to be between 0.43 and 0.45. Table 2 presents an overview of the results of Crude incidences and exposure-adjusted incidence the scenario analyses. rates of TEAEs in patients receiving Kd56 and Vd-8 in To assess whether the duration of treatment, in par- ENDEAVOR are shown in Table 3. Crude incidences ticular Vd treatment, had an impact on the depth of tended to be higher in the Kd56 group than in the response, cumulative rates of CR or better and VGPR Vd-8 group. However, exposure-adjusted incidences or better observed with Kd56 and Vd (given until pro- per 100 person-years were lower with Kd56 than with gression) in ENDEAVOR were assessed (Figure 4). Over Vd-8 for all TEAE categories assessed. Thus, after the entire treatment period, cumulative rates of both adjusting for exposure time, the risks of a grade 3 and KD56 VS. 8 CYCLES OF VD IN ENDEAVOR 43 above or of a serious TEAE were estimated to be Total person time is the sum of the time to first TEAE for all patients in each treatment group. If a patient had no event, the entire exposure time to study treatment was considered in the sum. For patients TEAEs were defined as any adverse event with an onset date from the first dose through 30 days after the last dose of any study drug and were coded using MedDRA version 20.0. Patients were counted only 0.7912) 0.7381) 0.7827) 0.7618) 0.4215 (0.3267, 0.5439) 0.4122 (0.3139, 0.5413) 0.4893 (0.2789, 0.8585) receiving Vd who continued study treatment beyond cycle 8, adverse events with an onset date from cycle 9 to the last dose of any study drug are not included. For patients with no events, the exposure time Kd56/Vd (95% CI) approximately 35% lower with Kd56 than with Vd-8. Kd56 vs. Vd-8 Rate ratio: (0.6093, (0.5442, (0.5310, (0.5049, The risk of a TEAE leading to discontinuation of any, or both, study drugs was estimated to be approxi- 0.6943 0.6338 0.6447 0.6202 mately 60% lower with Kd56 than with Vd-8 (Table 3). An additional safety analysis assessed specific TEAEs that had been previously found to be associated with Kd56 (cardiovascular events) or Vd (peripheral neur- (1233.62, 1481.89) (146.77, 179.50) opathy, gastrointestinal toxicities) [13] (Table 4). The Exposure-adjusted (60.02, 75.89) (50.10, 64.33) 25.14 (21.26, 29.72) 21.12 (17.66, 25.26) patient-years 5.40 (3.79, 7.67) incidence of grade 2 or above, or grade 3 or above, estimate (95% CI) per 100 peripheral neuropathies was lower in the Kd56 group than in the Vd-8 group. Likewise, the exposure- 1352.07 162.31 67.49 56.77 adjusted estimates indicated that rates were higher Table 3. Crude incidence proportions of TEAEs and exposure-adjusted incidence rates of TEAEs in Kd56 and Vd-8 patients in ENDEAVOR. Kd56 (N¼ 463) with Vd-8 than with Kd56 for diarrhea, nausea, and constipation. For the cardiovascular TEAEs of cardiac failure, hypertension, and embolic or thrombotic events (%) (98.7) (81.9) (60.3) (53.1) 137 (29.6) 120 (25.9) patients 31 (6.7) no. of Total with events, crude incidences and exposure-adjusted inci- 457 379 279 246 dence rates per 100 person-years were higher in the CI: confidence interval; Kd56: carfilzomib and dexamethasone; TEAE: treatment-emergent adverse event; Vd: bortezomib and dexamethasone Kd56 group than in the Vd-8 group for all three out- comes, but these findings were not statistically signifi- (years)a person 33.8 233.5 413.4 433.3 545.0 568.2 574.6 Total time cant. The exposure-adjusted rate ratios were 1.18 (95% CI 0.66, 21.4), 1.81 (95% CI 1.26, 2.60), and 1.85 (95% (1774.40, 2137.20) CI 0.97, 3.52) for cardiac failure, hypertension, and (228.43, 287.11) (89.70, 122.17) (77.75, 107.79) patient-years (95% CI) 59.63 (49.21, 72.27) 51.24 (41.72, 62.93) 11.03 (7.11, 17.09) embolic or thrombotic events, respectively. Exposure-adjusted estimate per 100 Discussion 1947.37 256.10 104.68 91.54 This exploratory analysis of ENDEAVOR that used MAIC Vd-8 (N¼ 456) methodology aimed to estimate the difference in effi- cacy with Kd56 vs. Vd-8. The results suggest that the events (%) (97.4) (64.5) (35.3) (31.6) 104 (22.8) 91 (20.0) patients 20 (4.4) benefit for PFS with Kd56 treatment over Vd treat- no. of Total with ment is larger if Vd is given for eight cycles only 444 294 161 144 instead of continued until progression. Furthermore, the data suggest that a prolonged treatment duration (years)a person 22.8 114.8 153.8 157.3 174.4 177.6 181.4 to study treatment was defined as the duration of the first eight cycles. of Vd improves PFS, thus adding to the growing body Total time of evidence that there may be a benefit for patients receiving proteasome inhibitor treatment until pro- gression rather than for a fixed number of cycles. TEAE leading to discontinuation of at least one study drug Grade 3 TEAEs leading to discontinuation of at least one If Vd had been given for eight cycles, the risk of progression or death was estimated to be 56% lower TEAE leading to discontinuation of bortezomib or with Kd56 than with Vd-8. This difference in PFS risk is greater than that observed in the original ENDEAVOR trial where Vd was administered until progression, and where the risk of progression or death was 47% lower once for each event of interest. with Kd56 than with Vd [13]. Scenario analyses sug- gested that the results were robust. A corresponding analysis for OS could not be performed, because sur- Death owing to TEAE vival data from the CASTOR trial were not as mature Grade 3 TEAE as PFS data at the time of this analysis. study drug Serious TEAEs carfilzomib Apart from the estimated difference between Kd56 Any TEAE and Vd-8, the findings of the present study may indi- cate that an extended duration of treatment with Vd, a 44 K. WEISEL ET AL. and possibly other proteasome inhibitors, beyond Total person time is the sum of the time to first TEAE for all patients in each treatment group. If a patient had no event, the entire exposure time to study treatment was considered in the sum. For patients TEAEs were defined as any adverse event with an onset date from the first dose through 30 days after the last dose of any study drug and were coded using MedDRA version 20.0. Patients were counted only 0.0657 (0.0448, 0.0965) 0.0886 (0.0454, 0.1729) 0.5149) 0.8077) 0.2547) 1.2174) 1.1865 (0.6568, 2.1434) 1.8070 (1.2579, 2.5956) 1.8504 (0.9721, 3.5222) receiving Vd who continued study treatment beyond cycle 8, adverse events with an onset date from cycle 9 to the last dose of any study drug are not included. For patients with no events, the exposure time eight cycles could result in prolonged PFS. With V, it Kd56 vs. Vd-8 Rate ratio: (95% CI) Kd56/Vd (0.3327, (0.4387, (0.1422, (0.5402, has been shown that proteasome function recovers within 72 hours after dosing, with no observed long- term effects on the proteasome [28,29] indicating that 0.4139 0.5953 0.1903 0.8109 the efficacy of V could be related to both the duration CI: confidence interval; Kd56: carfilzomib and dexamethasone; SMQN: standardized MedDRA query, narrow scope; TEAE: treatment-emergent adverse event; Vd: bortezomib and dexamethasone of therapy and cumulative dose received. Several 52.05) 28.90) 18.78) 19.51) 38.54 (32.96, 45.06) other recently published studies have also demon- Exposure-adjusted 9.33 (7.09, 12.28) 11.46 (8.88, 14.79) estimate per 100 6.00 (4.24, 8.49) 1.96 (1.09, 3.55) patient-years strated that prolonged treatment with proteasome (95% CI) (38.53, (19.88, (11.94, (12.55, inhibitors or immunomodulatory drugs provides a benefit for patients [30–34]. Subcutaneous V, which 44.78 23.97 14.97 15.65 Table 4. Crude incidence proportions and exposure-adjusted incidence rates of selected TEAE events in Kd56 and Vd-8 patients in ENDEAVOR. has been shown to be as efficacious as intravenous V Kd56 (N¼ 463) but to have a more favorable safety profile [35], Total no. of should enable patients to receive longer treatment events (%) (36.7) (23.8) (16.2) (17.1) 51 (11.0) 157 (33.9) 59 (12.7) patients 32 (6.9) 11 (2.4) with with less toxicity. Continuous treatment with prote- 170 110 75 79 asome inhibitors might lead to a benefit for improved outcomes, especially in high-risk patients, such as those with del[17/17p] cytogenic abnormalities [36]. time (years)a person Thus, the identification of patients who can tolerate 533.1 560.2 379.6 458.9 500.9 504.8 546.5 407.4 514.8 Total continuous treatment with proteasome inhibitors might be of clinical value in the management of MM. To identify possible drivers for the improved PFS 91.32 (77.38, 107.78) 126.78) 22.17 (16.20, 30.35) 51.26) 94.53) 27.15) 21.33 (15.38, 29.57) Exposure-adjusted benefit with a longer Vd treatment duration, cumulative 7.87 (4.66, 13.28) 6.19 (3.43, 11.18) estimate per 100 patient-years response rates were assessed for patients receiving (95% CI) (92.35, (31.64, (65.48, (13.72, Kd56 and Vd in ENDEAVOR (both given until progres- sion). Cumulative rates of CR or better and VGPR or 108.20 40.27 78.67 19.30 better were consistently higher with Kd56 than with Vd-8 (N¼ 456) Vd, indicative of a greater depth of response with Kd56 than with Vd, an effect that is likely to contribute to patients with Total no. of events (%) 140 (30.7) (33.6) (14.5) (25.0) 39 (8.6) (7.2) 14 (3.1) 36 (7.9) 11 (2.4) the longer duration PFS with Kd56 than with Vd. In the Vd arm, rates of CR or better and VGPR or better con- 153 66 114 33 tinued to gradually increase after cycle 8, which may imply that the improved PFS with continued Vd treat- time (years)a ment beyond eight cycles likely contributes to a greater person 153.3 175.9 141.4 163.9 144.9 171.0 178.0 168.8 177.6 Total to study treatment was defined as the duration of the first eight cycles. depth of response. However, a carryover effect of the eight cycles administered previously cannot be excluded. Continued Vd treatment duration might also provide a benefit for PFS by maintaining the response achieved during the early cycles, whereas stopping Vd after eight cycles could increase the risk of relapse. In line with this hypothesis, in ENDEAVOR the median (range) duration of response to continued Vd was 10.4 SMQN peripheral neuropathy (grade 2) SMQN peripheral neuropathy (grade 3) SMQN embolic and thrombotic events (95% CI 9.3, 13.8), whereas it was 7.9 months (95% CI 6.7, 11.3) with eight cycles of Vd in CASTOR [13,17]. once for each event of interest. Using truncated data for patients who received Peripheral neuropathy (SMQN) more than eight cycles of Vd in ENDEAVOR the differ- Gastrointestinal toxicities Cardiovascular toxicities SMQN cardiac failure ences in the safety profiles of Kd56 and Vd given for SMQN hypertension eight cycles were estimated. For any TEAE, grade 3 or Event of interest Constipation higher TEAEs, serious TEAEs, TEAEs leading to discon- Vomiting Diarrhea tinuation, and fatal TEAEs, exposure-adjusted incidence Nausea rates were lower with Kd56 than with Vd-8. Consistent with previous findings, the estimated risk of peripheral a KD56 VS. 8 CYCLES OF VD IN ENDEAVOR 45 neuropathy was significantly higher with Vd given for employee of Amgen (Europe) GmbH and holds Amgen eight cycles than with Kd56. Hypertension, a known stocks. L. DeCosta is an employee of Amgen Ltd and holds Amgen Stocks. A. Oriol has been a member of advisory and manageable side-effect with carfilzomib, was more boards for Amgen, Janssen, and Takeda. H. Goldschmidt has frequent with Kd56 than with Vd-8. Cardiac failure and received research support from Amgen, BMS, Celgene, embolic and thrombotic events were also more fre- Chugai, Janssen, Mundipharma, Novartis, Sanofi, and Takeda; quent with Kd56 than with Vd-8. has participated in advisory boards for Adaptive Several factors should be considered when interpret- Biotechnology, Amgen, BMS, Celgene, Janssen, Sanofi, and ing the present data. First, the analysis used MAIC meth- Takeda; and has received honoraria for speakers’ bureaux from ArtTempi, BMS, Celgene, Chugai, Janssen, and Novartis. odology, which is considered less robust than a H. Ludwig has received honoraria for speaker’s bureau from randomized controlled trial. Compared with other indir- Amgen, BMS, Celgene, Janssen, and Takeda, as well as ect treatment comparisons such as network meta-analy- research grants from Amgen and Takeda. M. Campioni is an ses that do not adjust for differences in trial design or employee of Amgen (Europe) GmbH and holds Amgen patient populations and are thus susceptible to greater stocks. Z. Szabo is an employee of Amgen (Europe) GmbH and holds Amgen stocks. M. Dimopoulos received consulting bias, MAIC is a more robust methodology and is consid- fees from Abbvie, Amgen, Celgene, Janssen, and Takeda. ered more reliable by payers and health technology assessment authorities to fill any gaps in the available comparative evidence. Another limitation of the present Funding study is that the follow-up durations were different This analysis and the development of this manuscript were between the ENDEAVOR data (median follow-up: funded by Amgen (Europe) GmbH. Medical writing support, 11.1 months (Vd group) and 11.9 months (Kd56 group)) funded by Amgen (Europe) GmbH, was provided by Oxford PharmaGenesis, Oxford, UK. Editorial support was provided and the CASTOR data (median follow-up: 19.4 months); by Carine Thual of Amgen (Europe) GmbH. however, this is due to PFS data from the longest pub- lished follow-up available at the time of the analyses being included, in order to achieve the most robust analysis. Moreover, the ENDEAVOR PFS Kaplan–Meier References curves are mature up to about 16–18 months, allowing [1] Yong K, Delforge M, Driessen C, et al. Multiple mye- a reliable and robust comparison even beyond the first loma: patient outcomes in real-world practice. Br J eight Vd cycles. Finally, patient-level data for Vd-8 had Haematol. 2016;175:252–264. to be reconstructed, which inherently adds uncertainty [2] Abramson HN. The multiple myeloma drug pipeline- 2018: a review of small molecules and their therapeutic to the reported results. targets. Clin Lymphoma Myeloma Leuk. 2018;18:611–627. The methodological approach used in this study [3] Bhatnagar V, Gormley NJ, Luo L, et al. FDA approval attempted to adjust for the differences in trial design, summary: daratumumab for treatment of multiple notably in this case Vd treatment duration, allowing myeloma after one prior therapy. Oncologist. 2017;22: cross-trial comparisons to be undertaken. Using this 1347–1353. [4] Moore D. Panobinostat (Farydak): a novel option for approach, the analyses suggested that, if Vd had been the treatment of relapsed or relapsed and refractory given for only eight cycles instead of continued until multiple myeloma. P T. 2016;41:296–300. progression, the treatment benefit with Kd56 would [5] Magen H, Muchtar E. Elotuzumab: the first approved have been larger, with lower HR than that observed in monoclonal antibody for multiple myeloma treat- ENDEAVOR. Exposure-adjusted incidence rates of grade ment. Ther Adv Hematol. 2016;7:187–195. [6] Kane RC, Bross PF, Farrell AT, et al. Velcade: U.S. FDA 3 or worse TEAEs were lower with Kd56 than with Vd. approval for the treatment of multiple myeloma pro- Overall, it was concluded that the duration of Vd treat- gressing on prior therapy. Oncologist. 2003;8:508–513. ment has a significant impact on both the response [7] Jagannath S, Richardson PG, Barlogie B, et al. rates and duration of response, and that the relative Bortezomib in combination with dexamethasone for efficacy and safety of Kd56 vs. Vd are increased when the treatment of patients with relapsed and/or refrac- tory multiple myeloma with less than optimal response Vd treatment is provided as per label guidance. to bortezomib alone. Haematologica. 2006;91:929–934. [8] Moreau P, San Miguel J, Sonneveld P, et al. Multiple Disclosure statement myeloma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2017;28:iv52–iv61. K. Weisel has participated in advisory boards for Amgen, [9] Khan ML, Stewart AK. Carfilzomib: a novel second- BMS, Celgene, Janssen, Juno, Sanofi, and Takeda; has generation proteasome inhibitor. Future Oncol. 2011; received honoraria from Amgen, BMS, Celgene, Janssen, 7:607–612. Novartis, and Takeda; and has received research funding [10] McBride A, Klaus JO, Stockerl-Goldstein K. Carfilzomib: from Amgen, Celgene, Janssen, and Sanofi. I. M. Majer is an a second-generation proteasome inhibitor for the 46 K. WEISEL ET AL. treatment of multiple myeloma. Am J Health Syst 2016; [cited 2019 May 31]. Available from: http:// Pharm. 2015;72:353–360. www.nicedsu.org.uk [11] Chim CS, Kumar SK, Orlowski RZ, et al. Management [23] Rajkumar SV. Multiple myeloma: 2016 update on of relapsed and refractory multiple myeloma: novel diagnosis, risk-stratification, and management. Am J agents, antibodies, immunotherapies and beyond. Hematol. 2016;91:719–734. Leukemia. 2018;32:252–262. [24] Chng WJ, Dispenzieri A, Chim CS, et al. IMWG consen- [12] European Medicines Agency. Kyprolis (carfilzomib) sus on risk stratification in multiple myeloma. summary of product characteristics. 2018; [cited 2018 Leukemia. 2014;28:269–277. Jul]. Available from: http://www.ema.europa.eu/docs/ [25] Majer IM, Castaigne JG, Palmer S, et al. Modeling en_GB/document_library/EPAR_-_Product_Information/ covariate-adjusted survival for economic evaluations human/003790/WC500197692.pdf in oncology. Pharmacoeconomics. 2019;37:727–737. [13] Dimopoulos MA, Moreau P, Palumbo A, et al. [26] National Institute for Health and Care Excellence. Carfilzomib and dexamethasone versus bortezomib Single technology appraisal. Carfilzomib for previously and dexamethasone for patients with relapsed or treated multiple myeloma; 2016; [cited 2019 May 31]. refractory multiple myeloma (ENDEAVOR): a rando- Available from: https://www.nice.org.uk/guidance/ mised, phase 3, open-label, multicentre study. Lancet ta457/documents/committee-papers Oncol. 2016;17:27–38. [27] Guyot P, Ades AE, Ouwens MJ, et al. Enhanced sec- [14] Dimopoulos MA, Goldschmidt H, Niesvizky R, et al. ondary analysis of survival data: reconstructing the Carfilzomib or bortezomib in relapsed or refractory data from published Kaplan–Meier survival curves. multiple myeloma (ENDEAVOR): an interim overall sur- BMC Med Res Methodol. 2012;12:9. vival analysis of an open-label, randomised, phase 3 [28] Orlowski RZ, Stinchcombe TE, Mitchell BS, et al. Phase trial. Lancet Oncol. 2017;18:1327–1337. I trial of the proteasome inhibitor PS-341 in patients [15] European Medicines Agency. Velcade (bortezomib) sum- with refractory hematologic malignancies. J Clin mary of product characteristics. 2017; [cited 2018 Jul 18]. Oncol. 2002;20:4420–4427. Available from: http://www.ema.europa.eu/ema/index. [29] Schwartz R, Davidson T. Pharmacology, pharmacokin- jsp?curl=pages/medicines/human/medicines/004207/ etics, and practical applications of bortezomib. human_med_001995.jsp&mid=WC0b01ac058001d124 Oncology (Williston Park, NY). 2004;18:14–21. [16] Jakubowiak AJ, DeCara JM, Mezzi K. Cardiovascular [30] Palumbo A, Gay F, Cavallo F, et al. Continuous ther- events during carfilzomib therapy for relapsed mye- apy versus fixed duration of therapy in patients with loma: practical management aspects from two case newly diagnosed multiple myeloma. J Clin Oncol. studies. Hematology. 2017;22:585–591. 2015;33:3459–3466. [17] Palumbo A, Chanan-Khan A, Weisel K, et al. [31] Mateos MV, Oriol A, Martınez-Lo pez J, et al. Matched- Daratumumab, bortezomib, and dexamethasone for pairs analysis of outcomes with bortezomib, melphalan, multiple myeloma. N Engl J Med. 2016;375:754–766. and prednisone (VMP) treatment for previously untreated [18] Spencer A, Hungria V, Mateos M, et al. Daratumumab, multiple myeloma (MM) using long-term follow-up data bortezomib, and dexamethasone (DVd) versus borte- from the phase 3 VISTA and PETHEMA/GEM05 trials. Clin zomib and dexamethasone (Vd) in relapsed or refrac- Lymphoma Myeloma Leuk. 2015;15:e73–e74. tory multiple myeloma (RRMM): updated efficacy and [32] Harousseau JL, Dimopoulos MA, Wang M, et al. Better safety analysis of CASTOR. 59th American Society of quality of response to lenalidomide plus dexametha- Hematology (ASH) Annual Meeting & Exposition; Dec sone is associated with improved clinical outcomes in 9–12; Atlanta, Georgia; 2017. patients with relapsed or refractory multiple mye- [19] Orlowski R, Moreau P, Ludwig H, et al. Carfilzomib loma. Haematologica. 2010;95:1738–1744. and dexamethasone (Kd56) vs bortezomib and dexa- [33] San-Miguel JF, Dimopoulos MA, Stadtmauer EA, et al. methasone (Vd) in relapsed or refractory multiple Effects of lenalidomide and dexamethasone treatment myeloma (RRMM): updated overall survival (OS), duration on survival in patients with relapsed or safety, and subgroup analysis of ENDEAVOR. J Clin refractory multiple myeloma treated with lenalido- Oncol. 2018;36:8032–8032. mide and dexamethasone. Clin Lymphoma Myeloma [20] Mateos M-V, Estell J, Barreto W, et al. Efficacy of dara- Leuk. 2011;11:38–43. tumumab, bortezomib, and dexamethasone versus [34] Hari P, Romanus D, Palumbo A, et al. Prolonged dur- bortezomib and dexamethasone in relapsed or refrac- ation of therapy is associated with improved survival tory myeloma based on prior lines of therapy: in patients treated for relapsed/refractory multiple updated analysis of castor. Blood. 2016;128:1150. myeloma in routine clinical care in the United States. [21] Signorovitch J, Erder MH, Xie J, et al. Comparative Clin Lymphoma Myeloma Leuk. 2018;18:152–160. effectiveness research using matching-adjusted indirect [35] Moreau P, Pylypenko H, Grosicki S, et al. comparison: an application to treatment with guanfa- Subcutaneous versus intravenous administration of cine extended release or atomoxetine in children with bortezomib in patients with relapsed multiple mye- attention-deficit/hyperactivity disorder and comorbid loma: a randomised, phase 3, non-inferiority study. oppositional defiant disorder. Pharmacoepidemiol Drug Lancet Oncol. 2011;12:431–440. Saf. 2012;21:130–137. [36] Sonneveld P, Avet-Loiseau H, Lonial S, et al. [22] Phillippo DM, Ades AE, Dias S, et al. NICE DSU tech- Treatment of multiple myeloma with high-risk cyto- nical support document 18: methods for population- genetics: a consensus of the International Myeloma adjusted indirect comparisons in submission to NICE; Working Group. Blood. 2016;127:2955–2962.